Loading...
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
17#define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
18static const char ice_driver_string[] = DRV_SUMMARY;
19static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
20
21/* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
22#define ICE_DDP_PKG_PATH "intel/ice/ddp/"
23#define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
24
25MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
26MODULE_DESCRIPTION(DRV_SUMMARY);
27MODULE_LICENSE("GPL v2");
28MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
29
30static int debug = -1;
31module_param(debug, int, 0644);
32#ifndef CONFIG_DYNAMIC_DEBUG
33MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
34#else
35MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
36#endif /* !CONFIG_DYNAMIC_DEBUG */
37
38static struct workqueue_struct *ice_wq;
39static const struct net_device_ops ice_netdev_safe_mode_ops;
40static const struct net_device_ops ice_netdev_ops;
41static int ice_vsi_open(struct ice_vsi *vsi);
42
43static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
44
45static void ice_vsi_release_all(struct ice_pf *pf);
46
47/**
48 * ice_get_tx_pending - returns number of Tx descriptors not processed
49 * @ring: the ring of descriptors
50 */
51static u16 ice_get_tx_pending(struct ice_ring *ring)
52{
53 u16 head, tail;
54
55 head = ring->next_to_clean;
56 tail = ring->next_to_use;
57
58 if (head != tail)
59 return (head < tail) ?
60 tail - head : (tail + ring->count - head);
61 return 0;
62}
63
64/**
65 * ice_check_for_hang_subtask - check for and recover hung queues
66 * @pf: pointer to PF struct
67 */
68static void ice_check_for_hang_subtask(struct ice_pf *pf)
69{
70 struct ice_vsi *vsi = NULL;
71 struct ice_hw *hw;
72 unsigned int i;
73 int packets;
74 u32 v;
75
76 ice_for_each_vsi(pf, v)
77 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
78 vsi = pf->vsi[v];
79 break;
80 }
81
82 if (!vsi || test_bit(__ICE_DOWN, vsi->state))
83 return;
84
85 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
86 return;
87
88 hw = &vsi->back->hw;
89
90 for (i = 0; i < vsi->num_txq; i++) {
91 struct ice_ring *tx_ring = vsi->tx_rings[i];
92
93 if (tx_ring && tx_ring->desc) {
94 /* If packet counter has not changed the queue is
95 * likely stalled, so force an interrupt for this
96 * queue.
97 *
98 * prev_pkt would be negative if there was no
99 * pending work.
100 */
101 packets = tx_ring->stats.pkts & INT_MAX;
102 if (tx_ring->tx_stats.prev_pkt == packets) {
103 /* Trigger sw interrupt to revive the queue */
104 ice_trigger_sw_intr(hw, tx_ring->q_vector);
105 continue;
106 }
107
108 /* Memory barrier between read of packet count and call
109 * to ice_get_tx_pending()
110 */
111 smp_rmb();
112 tx_ring->tx_stats.prev_pkt =
113 ice_get_tx_pending(tx_ring) ? packets : -1;
114 }
115 }
116}
117
118/**
119 * ice_init_mac_fltr - Set initial MAC filters
120 * @pf: board private structure
121 *
122 * Set initial set of MAC filters for PF VSI; configure filters for permanent
123 * address and broadcast address. If an error is encountered, netdevice will be
124 * unregistered.
125 */
126static int ice_init_mac_fltr(struct ice_pf *pf)
127{
128 enum ice_status status;
129 struct ice_vsi *vsi;
130 u8 *perm_addr;
131
132 vsi = ice_get_main_vsi(pf);
133 if (!vsi)
134 return -EINVAL;
135
136 perm_addr = vsi->port_info->mac.perm_addr;
137 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
138 if (!status)
139 return 0;
140
141 /* We aren't useful with no MAC filters, so unregister if we
142 * had an error
143 */
144 if (vsi->netdev->reg_state == NETREG_REGISTERED) {
145 dev_err(ice_pf_to_dev(pf), "Could not add MAC filters error %s. Unregistering device\n",
146 ice_stat_str(status));
147 unregister_netdev(vsi->netdev);
148 free_netdev(vsi->netdev);
149 vsi->netdev = NULL;
150 }
151
152 return -EIO;
153}
154
155/**
156 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
157 * @netdev: the net device on which the sync is happening
158 * @addr: MAC address to sync
159 *
160 * This is a callback function which is called by the in kernel device sync
161 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
162 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
163 * MAC filters from the hardware.
164 */
165static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
166{
167 struct ice_netdev_priv *np = netdev_priv(netdev);
168 struct ice_vsi *vsi = np->vsi;
169
170 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
171 ICE_FWD_TO_VSI))
172 return -EINVAL;
173
174 return 0;
175}
176
177/**
178 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
179 * @netdev: the net device on which the unsync is happening
180 * @addr: MAC address to unsync
181 *
182 * This is a callback function which is called by the in kernel device unsync
183 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
184 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
185 * delete the MAC filters from the hardware.
186 */
187static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
188{
189 struct ice_netdev_priv *np = netdev_priv(netdev);
190 struct ice_vsi *vsi = np->vsi;
191
192 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
193 ICE_FWD_TO_VSI))
194 return -EINVAL;
195
196 return 0;
197}
198
199/**
200 * ice_vsi_fltr_changed - check if filter state changed
201 * @vsi: VSI to be checked
202 *
203 * returns true if filter state has changed, false otherwise.
204 */
205static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
206{
207 return test_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags) ||
208 test_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags) ||
209 test_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
210}
211
212/**
213 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
214 * @vsi: the VSI being configured
215 * @promisc_m: mask of promiscuous config bits
216 * @set_promisc: enable or disable promisc flag request
217 *
218 */
219static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
220{
221 struct ice_hw *hw = &vsi->back->hw;
222 enum ice_status status = 0;
223
224 if (vsi->type != ICE_VSI_PF)
225 return 0;
226
227 if (vsi->vlan_ena) {
228 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
229 set_promisc);
230 } else {
231 if (set_promisc)
232 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
233 0);
234 else
235 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
236 0);
237 }
238
239 if (status)
240 return -EIO;
241
242 return 0;
243}
244
245/**
246 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
247 * @vsi: ptr to the VSI
248 *
249 * Push any outstanding VSI filter changes through the AdminQ.
250 */
251static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
252{
253 struct device *dev = ice_pf_to_dev(vsi->back);
254 struct net_device *netdev = vsi->netdev;
255 bool promisc_forced_on = false;
256 struct ice_pf *pf = vsi->back;
257 struct ice_hw *hw = &pf->hw;
258 enum ice_status status = 0;
259 u32 changed_flags = 0;
260 u8 promisc_m;
261 int err = 0;
262
263 if (!vsi->netdev)
264 return -EINVAL;
265
266 while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state))
267 usleep_range(1000, 2000);
268
269 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
270 vsi->current_netdev_flags = vsi->netdev->flags;
271
272 INIT_LIST_HEAD(&vsi->tmp_sync_list);
273 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
274
275 if (ice_vsi_fltr_changed(vsi)) {
276 clear_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
277 clear_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
278 clear_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
279
280 /* grab the netdev's addr_list_lock */
281 netif_addr_lock_bh(netdev);
282 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
283 ice_add_mac_to_unsync_list);
284 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
285 ice_add_mac_to_unsync_list);
286 /* our temp lists are populated. release lock */
287 netif_addr_unlock_bh(netdev);
288 }
289
290 /* Remove MAC addresses in the unsync list */
291 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
292 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
293 if (status) {
294 netdev_err(netdev, "Failed to delete MAC filters\n");
295 /* if we failed because of alloc failures, just bail */
296 if (status == ICE_ERR_NO_MEMORY) {
297 err = -ENOMEM;
298 goto out;
299 }
300 }
301
302 /* Add MAC addresses in the sync list */
303 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
304 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
305 /* If filter is added successfully or already exists, do not go into
306 * 'if' condition and report it as error. Instead continue processing
307 * rest of the function.
308 */
309 if (status && status != ICE_ERR_ALREADY_EXISTS) {
310 netdev_err(netdev, "Failed to add MAC filters\n");
311 /* If there is no more space for new umac filters, VSI
312 * should go into promiscuous mode. There should be some
313 * space reserved for promiscuous filters.
314 */
315 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
316 !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC,
317 vsi->state)) {
318 promisc_forced_on = true;
319 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
320 vsi->vsi_num);
321 } else {
322 err = -EIO;
323 goto out;
324 }
325 }
326 /* check for changes in promiscuous modes */
327 if (changed_flags & IFF_ALLMULTI) {
328 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
329 if (vsi->vlan_ena)
330 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
331 else
332 promisc_m = ICE_MCAST_PROMISC_BITS;
333
334 err = ice_cfg_promisc(vsi, promisc_m, true);
335 if (err) {
336 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
337 vsi->vsi_num);
338 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
339 goto out_promisc;
340 }
341 } else {
342 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
343 if (vsi->vlan_ena)
344 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
345 else
346 promisc_m = ICE_MCAST_PROMISC_BITS;
347
348 err = ice_cfg_promisc(vsi, promisc_m, false);
349 if (err) {
350 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
351 vsi->vsi_num);
352 vsi->current_netdev_flags |= IFF_ALLMULTI;
353 goto out_promisc;
354 }
355 }
356 }
357
358 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
359 test_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags)) {
360 clear_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
361 if (vsi->current_netdev_flags & IFF_PROMISC) {
362 /* Apply Rx filter rule to get traffic from wire */
363 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
364 err = ice_set_dflt_vsi(pf->first_sw, vsi);
365 if (err && err != -EEXIST) {
366 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
367 err, vsi->vsi_num);
368 vsi->current_netdev_flags &=
369 ~IFF_PROMISC;
370 goto out_promisc;
371 }
372 ice_cfg_vlan_pruning(vsi, false, false);
373 }
374 } else {
375 /* Clear Rx filter to remove traffic from wire */
376 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
377 err = ice_clear_dflt_vsi(pf->first_sw);
378 if (err) {
379 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
380 err, vsi->vsi_num);
381 vsi->current_netdev_flags |=
382 IFF_PROMISC;
383 goto out_promisc;
384 }
385 if (vsi->num_vlan > 1)
386 ice_cfg_vlan_pruning(vsi, true, false);
387 }
388 }
389 }
390 goto exit;
391
392out_promisc:
393 set_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
394 goto exit;
395out:
396 /* if something went wrong then set the changed flag so we try again */
397 set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
398 set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
399exit:
400 clear_bit(__ICE_CFG_BUSY, vsi->state);
401 return err;
402}
403
404/**
405 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
406 * @pf: board private structure
407 */
408static void ice_sync_fltr_subtask(struct ice_pf *pf)
409{
410 int v;
411
412 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
413 return;
414
415 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
416
417 ice_for_each_vsi(pf, v)
418 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
419 ice_vsi_sync_fltr(pf->vsi[v])) {
420 /* come back and try again later */
421 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
422 break;
423 }
424}
425
426/**
427 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
428 * @pf: the PF
429 * @locked: is the rtnl_lock already held
430 */
431static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
432{
433 int v;
434
435 ice_for_each_vsi(pf, v)
436 if (pf->vsi[v])
437 ice_dis_vsi(pf->vsi[v], locked);
438}
439
440/**
441 * ice_prepare_for_reset - prep for the core to reset
442 * @pf: board private structure
443 *
444 * Inform or close all dependent features in prep for reset.
445 */
446static void
447ice_prepare_for_reset(struct ice_pf *pf)
448{
449 struct ice_hw *hw = &pf->hw;
450 unsigned int i;
451
452 /* already prepared for reset */
453 if (test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
454 return;
455
456 /* Notify VFs of impending reset */
457 if (ice_check_sq_alive(hw, &hw->mailboxq))
458 ice_vc_notify_reset(pf);
459
460 /* Disable VFs until reset is completed */
461 ice_for_each_vf(pf, i)
462 ice_set_vf_state_qs_dis(&pf->vf[i]);
463
464 /* clear SW filtering DB */
465 ice_clear_hw_tbls(hw);
466 /* disable the VSIs and their queues that are not already DOWN */
467 ice_pf_dis_all_vsi(pf, false);
468
469 if (hw->port_info)
470 ice_sched_clear_port(hw->port_info);
471
472 ice_shutdown_all_ctrlq(hw);
473
474 set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
475}
476
477/**
478 * ice_do_reset - Initiate one of many types of resets
479 * @pf: board private structure
480 * @reset_type: reset type requested
481 * before this function was called.
482 */
483static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
484{
485 struct device *dev = ice_pf_to_dev(pf);
486 struct ice_hw *hw = &pf->hw;
487
488 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
489 WARN_ON(in_interrupt());
490
491 ice_prepare_for_reset(pf);
492
493 /* trigger the reset */
494 if (ice_reset(hw, reset_type)) {
495 dev_err(dev, "reset %d failed\n", reset_type);
496 set_bit(__ICE_RESET_FAILED, pf->state);
497 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
498 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
499 clear_bit(__ICE_PFR_REQ, pf->state);
500 clear_bit(__ICE_CORER_REQ, pf->state);
501 clear_bit(__ICE_GLOBR_REQ, pf->state);
502 return;
503 }
504
505 /* PFR is a bit of a special case because it doesn't result in an OICR
506 * interrupt. So for PFR, rebuild after the reset and clear the reset-
507 * associated state bits.
508 */
509 if (reset_type == ICE_RESET_PFR) {
510 pf->pfr_count++;
511 ice_rebuild(pf, reset_type);
512 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
513 clear_bit(__ICE_PFR_REQ, pf->state);
514 ice_reset_all_vfs(pf, true);
515 }
516}
517
518/**
519 * ice_reset_subtask - Set up for resetting the device and driver
520 * @pf: board private structure
521 */
522static void ice_reset_subtask(struct ice_pf *pf)
523{
524 enum ice_reset_req reset_type = ICE_RESET_INVAL;
525
526 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
527 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
528 * of reset is pending and sets bits in pf->state indicating the reset
529 * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set
530 * prepare for pending reset if not already (for PF software-initiated
531 * global resets the software should already be prepared for it as
532 * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
533 * by firmware or software on other PFs, that bit is not set so prepare
534 * for the reset now), poll for reset done, rebuild and return.
535 */
536 if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
537 /* Perform the largest reset requested */
538 if (test_and_clear_bit(__ICE_CORER_RECV, pf->state))
539 reset_type = ICE_RESET_CORER;
540 if (test_and_clear_bit(__ICE_GLOBR_RECV, pf->state))
541 reset_type = ICE_RESET_GLOBR;
542 if (test_and_clear_bit(__ICE_EMPR_RECV, pf->state))
543 reset_type = ICE_RESET_EMPR;
544 /* return if no valid reset type requested */
545 if (reset_type == ICE_RESET_INVAL)
546 return;
547 ice_prepare_for_reset(pf);
548
549 /* make sure we are ready to rebuild */
550 if (ice_check_reset(&pf->hw)) {
551 set_bit(__ICE_RESET_FAILED, pf->state);
552 } else {
553 /* done with reset. start rebuild */
554 pf->hw.reset_ongoing = false;
555 ice_rebuild(pf, reset_type);
556 /* clear bit to resume normal operations, but
557 * ICE_NEEDS_RESTART bit is set in case rebuild failed
558 */
559 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
560 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
561 clear_bit(__ICE_PFR_REQ, pf->state);
562 clear_bit(__ICE_CORER_REQ, pf->state);
563 clear_bit(__ICE_GLOBR_REQ, pf->state);
564 ice_reset_all_vfs(pf, true);
565 }
566
567 return;
568 }
569
570 /* No pending resets to finish processing. Check for new resets */
571 if (test_bit(__ICE_PFR_REQ, pf->state))
572 reset_type = ICE_RESET_PFR;
573 if (test_bit(__ICE_CORER_REQ, pf->state))
574 reset_type = ICE_RESET_CORER;
575 if (test_bit(__ICE_GLOBR_REQ, pf->state))
576 reset_type = ICE_RESET_GLOBR;
577 /* If no valid reset type requested just return */
578 if (reset_type == ICE_RESET_INVAL)
579 return;
580
581 /* reset if not already down or busy */
582 if (!test_bit(__ICE_DOWN, pf->state) &&
583 !test_bit(__ICE_CFG_BUSY, pf->state)) {
584 ice_do_reset(pf, reset_type);
585 }
586}
587
588/**
589 * ice_print_topo_conflict - print topology conflict message
590 * @vsi: the VSI whose topology status is being checked
591 */
592static void ice_print_topo_conflict(struct ice_vsi *vsi)
593{
594 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
595 case ICE_AQ_LINK_TOPO_CONFLICT:
596 case ICE_AQ_LINK_MEDIA_CONFLICT:
597 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
598 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
599 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
600 netdev_info(vsi->netdev, "Possible mis-configuration of the Ethernet port detected, please use the Intel(R) Ethernet Port Configuration Tool application to address the issue.\n");
601 break;
602 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
603 netdev_info(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");
604 break;
605 default:
606 break;
607 }
608}
609
610/**
611 * ice_print_link_msg - print link up or down message
612 * @vsi: the VSI whose link status is being queried
613 * @isup: boolean for if the link is now up or down
614 */
615void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
616{
617 struct ice_aqc_get_phy_caps_data *caps;
618 const char *an_advertised;
619 enum ice_status status;
620 const char *fec_req;
621 const char *speed;
622 const char *fec;
623 const char *fc;
624 const char *an;
625
626 if (!vsi)
627 return;
628
629 if (vsi->current_isup == isup)
630 return;
631
632 vsi->current_isup = isup;
633
634 if (!isup) {
635 netdev_info(vsi->netdev, "NIC Link is Down\n");
636 return;
637 }
638
639 switch (vsi->port_info->phy.link_info.link_speed) {
640 case ICE_AQ_LINK_SPEED_100GB:
641 speed = "100 G";
642 break;
643 case ICE_AQ_LINK_SPEED_50GB:
644 speed = "50 G";
645 break;
646 case ICE_AQ_LINK_SPEED_40GB:
647 speed = "40 G";
648 break;
649 case ICE_AQ_LINK_SPEED_25GB:
650 speed = "25 G";
651 break;
652 case ICE_AQ_LINK_SPEED_20GB:
653 speed = "20 G";
654 break;
655 case ICE_AQ_LINK_SPEED_10GB:
656 speed = "10 G";
657 break;
658 case ICE_AQ_LINK_SPEED_5GB:
659 speed = "5 G";
660 break;
661 case ICE_AQ_LINK_SPEED_2500MB:
662 speed = "2.5 G";
663 break;
664 case ICE_AQ_LINK_SPEED_1000MB:
665 speed = "1 G";
666 break;
667 case ICE_AQ_LINK_SPEED_100MB:
668 speed = "100 M";
669 break;
670 default:
671 speed = "Unknown";
672 break;
673 }
674
675 switch (vsi->port_info->fc.current_mode) {
676 case ICE_FC_FULL:
677 fc = "Rx/Tx";
678 break;
679 case ICE_FC_TX_PAUSE:
680 fc = "Tx";
681 break;
682 case ICE_FC_RX_PAUSE:
683 fc = "Rx";
684 break;
685 case ICE_FC_NONE:
686 fc = "None";
687 break;
688 default:
689 fc = "Unknown";
690 break;
691 }
692
693 /* Get FEC mode based on negotiated link info */
694 switch (vsi->port_info->phy.link_info.fec_info) {
695 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
696 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
697 fec = "RS-FEC";
698 break;
699 case ICE_AQ_LINK_25G_KR_FEC_EN:
700 fec = "FC-FEC/BASE-R";
701 break;
702 default:
703 fec = "NONE";
704 break;
705 }
706
707 /* check if autoneg completed, might be false due to not supported */
708 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
709 an = "True";
710 else
711 an = "False";
712
713 /* Get FEC mode requested based on PHY caps last SW configuration */
714 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
715 if (!caps) {
716 fec_req = "Unknown";
717 an_advertised = "Unknown";
718 goto done;
719 }
720
721 status = ice_aq_get_phy_caps(vsi->port_info, false,
722 ICE_AQC_REPORT_SW_CFG, caps, NULL);
723 if (status)
724 netdev_info(vsi->netdev, "Get phy capability failed.\n");
725
726 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
727
728 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
729 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
730 fec_req = "RS-FEC";
731 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
732 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
733 fec_req = "FC-FEC/BASE-R";
734 else
735 fec_req = "NONE";
736
737 kfree(caps);
738
739done:
740 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",
741 speed, fec_req, fec, an_advertised, an, fc);
742 ice_print_topo_conflict(vsi);
743}
744
745/**
746 * ice_vsi_link_event - update the VSI's netdev
747 * @vsi: the VSI on which the link event occurred
748 * @link_up: whether or not the VSI needs to be set up or down
749 */
750static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
751{
752 if (!vsi)
753 return;
754
755 if (test_bit(__ICE_DOWN, vsi->state) || !vsi->netdev)
756 return;
757
758 if (vsi->type == ICE_VSI_PF) {
759 if (link_up == netif_carrier_ok(vsi->netdev))
760 return;
761
762 if (link_up) {
763 netif_carrier_on(vsi->netdev);
764 netif_tx_wake_all_queues(vsi->netdev);
765 } else {
766 netif_carrier_off(vsi->netdev);
767 netif_tx_stop_all_queues(vsi->netdev);
768 }
769 }
770}
771
772/**
773 * ice_set_dflt_mib - send a default config MIB to the FW
774 * @pf: private PF struct
775 *
776 * This function sends a default configuration MIB to the FW.
777 *
778 * If this function errors out at any point, the driver is still able to
779 * function. The main impact is that LFC may not operate as expected.
780 * Therefore an error state in this function should be treated with a DBG
781 * message and continue on with driver rebuild/reenable.
782 */
783static void ice_set_dflt_mib(struct ice_pf *pf)
784{
785 struct device *dev = ice_pf_to_dev(pf);
786 u8 mib_type, *buf, *lldpmib = NULL;
787 u16 len, typelen, offset = 0;
788 struct ice_lldp_org_tlv *tlv;
789 struct ice_hw *hw;
790 u32 ouisubtype;
791
792 if (!pf) {
793 dev_dbg(dev, "%s NULL pf pointer\n", __func__);
794 return;
795 }
796
797 hw = &pf->hw;
798 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
799 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
800 if (!lldpmib) {
801 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
802 __func__);
803 return;
804 }
805
806 /* Add ETS CFG TLV */
807 tlv = (struct ice_lldp_org_tlv *)lldpmib;
808 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
809 ICE_IEEE_ETS_TLV_LEN);
810 tlv->typelen = htons(typelen);
811 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
812 ICE_IEEE_SUBTYPE_ETS_CFG);
813 tlv->ouisubtype = htonl(ouisubtype);
814
815 buf = tlv->tlvinfo;
816 buf[0] = 0;
817
818 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
819 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
820 * Octets 13 - 20 are TSA values - leave as zeros
821 */
822 buf[5] = 0x64;
823 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
824 offset += len + 2;
825 tlv = (struct ice_lldp_org_tlv *)
826 ((char *)tlv + sizeof(tlv->typelen) + len);
827
828 /* Add ETS REC TLV */
829 buf = tlv->tlvinfo;
830 tlv->typelen = htons(typelen);
831
832 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
833 ICE_IEEE_SUBTYPE_ETS_REC);
834 tlv->ouisubtype = htonl(ouisubtype);
835
836 /* First octet of buf is reserved
837 * Octets 1 - 4 map UP to TC - all UPs map to zero
838 * Octets 5 - 12 are BW values - set TC 0 to 100%.
839 * Octets 13 - 20 are TSA value - leave as zeros
840 */
841 buf[5] = 0x64;
842 offset += len + 2;
843 tlv = (struct ice_lldp_org_tlv *)
844 ((char *)tlv + sizeof(tlv->typelen) + len);
845
846 /* Add PFC CFG TLV */
847 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
848 ICE_IEEE_PFC_TLV_LEN);
849 tlv->typelen = htons(typelen);
850
851 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
852 ICE_IEEE_SUBTYPE_PFC_CFG);
853 tlv->ouisubtype = htonl(ouisubtype);
854
855 /* Octet 1 left as all zeros - PFC disabled */
856 buf[0] = 0x08;
857 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
858 offset += len + 2;
859
860 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
861 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
862
863 kfree(lldpmib);
864}
865
866/**
867 * ice_link_event - process the link event
868 * @pf: PF that the link event is associated with
869 * @pi: port_info for the port that the link event is associated with
870 * @link_up: true if the physical link is up and false if it is down
871 * @link_speed: current link speed received from the link event
872 *
873 * Returns 0 on success and negative on failure
874 */
875static int
876ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
877 u16 link_speed)
878{
879 struct device *dev = ice_pf_to_dev(pf);
880 struct ice_phy_info *phy_info;
881 struct ice_vsi *vsi;
882 u16 old_link_speed;
883 bool old_link;
884 int result;
885
886 phy_info = &pi->phy;
887 phy_info->link_info_old = phy_info->link_info;
888
889 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
890 old_link_speed = phy_info->link_info_old.link_speed;
891
892 /* update the link info structures and re-enable link events,
893 * don't bail on failure due to other book keeping needed
894 */
895 result = ice_update_link_info(pi);
896 if (result)
897 dev_dbg(dev, "Failed to update link status and re-enable link events for port %d\n",
898 pi->lport);
899
900 /* Check if the link state is up after updating link info, and treat
901 * this event as an UP event since the link is actually UP now.
902 */
903 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
904 link_up = true;
905
906 vsi = ice_get_main_vsi(pf);
907 if (!vsi || !vsi->port_info)
908 return -EINVAL;
909
910 /* turn off PHY if media was removed */
911 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
912 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
913 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
914
915 result = ice_aq_set_link_restart_an(pi, false, NULL);
916 if (result) {
917 dev_dbg(dev, "Failed to set link down, VSI %d error %d\n",
918 vsi->vsi_num, result);
919 return result;
920 }
921 }
922
923 /* if the old link up/down and speed is the same as the new */
924 if (link_up == old_link && link_speed == old_link_speed)
925 return result;
926
927 if (ice_is_dcb_active(pf)) {
928 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
929 ice_dcb_rebuild(pf);
930 } else {
931 if (link_up)
932 ice_set_dflt_mib(pf);
933 }
934 ice_vsi_link_event(vsi, link_up);
935 ice_print_link_msg(vsi, link_up);
936
937 ice_vc_notify_link_state(pf);
938
939 return result;
940}
941
942/**
943 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
944 * @pf: board private structure
945 */
946static void ice_watchdog_subtask(struct ice_pf *pf)
947{
948 int i;
949
950 /* if interface is down do nothing */
951 if (test_bit(__ICE_DOWN, pf->state) ||
952 test_bit(__ICE_CFG_BUSY, pf->state))
953 return;
954
955 /* make sure we don't do these things too often */
956 if (time_before(jiffies,
957 pf->serv_tmr_prev + pf->serv_tmr_period))
958 return;
959
960 pf->serv_tmr_prev = jiffies;
961
962 /* Update the stats for active netdevs so the network stack
963 * can look at updated numbers whenever it cares to
964 */
965 ice_update_pf_stats(pf);
966 ice_for_each_vsi(pf, i)
967 if (pf->vsi[i] && pf->vsi[i]->netdev)
968 ice_update_vsi_stats(pf->vsi[i]);
969}
970
971/**
972 * ice_init_link_events - enable/initialize link events
973 * @pi: pointer to the port_info instance
974 *
975 * Returns -EIO on failure, 0 on success
976 */
977static int ice_init_link_events(struct ice_port_info *pi)
978{
979 u16 mask;
980
981 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
982 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
983
984 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
985 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
986 pi->lport);
987 return -EIO;
988 }
989
990 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
991 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
992 pi->lport);
993 return -EIO;
994 }
995
996 return 0;
997}
998
999/**
1000 * ice_handle_link_event - handle link event via ARQ
1001 * @pf: PF that the link event is associated with
1002 * @event: event structure containing link status info
1003 */
1004static int
1005ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1006{
1007 struct ice_aqc_get_link_status_data *link_data;
1008 struct ice_port_info *port_info;
1009 int status;
1010
1011 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1012 port_info = pf->hw.port_info;
1013 if (!port_info)
1014 return -EINVAL;
1015
1016 status = ice_link_event(pf, port_info,
1017 !!(link_data->link_info & ICE_AQ_LINK_UP),
1018 le16_to_cpu(link_data->link_speed));
1019 if (status)
1020 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1021 status);
1022
1023 return status;
1024}
1025
1026enum ice_aq_task_state {
1027 ICE_AQ_TASK_WAITING = 0,
1028 ICE_AQ_TASK_COMPLETE,
1029 ICE_AQ_TASK_CANCELED,
1030};
1031
1032struct ice_aq_task {
1033 struct hlist_node entry;
1034
1035 u16 opcode;
1036 struct ice_rq_event_info *event;
1037 enum ice_aq_task_state state;
1038};
1039
1040/**
1041 * ice_wait_for_aq_event - Wait for an AdminQ event from firmware
1042 * @pf: pointer to the PF private structure
1043 * @opcode: the opcode to wait for
1044 * @timeout: how long to wait, in jiffies
1045 * @event: storage for the event info
1046 *
1047 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1048 * current thread will be put to sleep until the specified event occurs or
1049 * until the given timeout is reached.
1050 *
1051 * To obtain only the descriptor contents, pass an event without an allocated
1052 * msg_buf. If the complete data buffer is desired, allocate the
1053 * event->msg_buf with enough space ahead of time.
1054 *
1055 * Returns: zero on success, or a negative error code on failure.
1056 */
1057int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1058 struct ice_rq_event_info *event)
1059{
1060 struct ice_aq_task *task;
1061 long ret;
1062 int err;
1063
1064 task = kzalloc(sizeof(*task), GFP_KERNEL);
1065 if (!task)
1066 return -ENOMEM;
1067
1068 INIT_HLIST_NODE(&task->entry);
1069 task->opcode = opcode;
1070 task->event = event;
1071 task->state = ICE_AQ_TASK_WAITING;
1072
1073 spin_lock_bh(&pf->aq_wait_lock);
1074 hlist_add_head(&task->entry, &pf->aq_wait_list);
1075 spin_unlock_bh(&pf->aq_wait_lock);
1076
1077 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1078 timeout);
1079 switch (task->state) {
1080 case ICE_AQ_TASK_WAITING:
1081 err = ret < 0 ? ret : -ETIMEDOUT;
1082 break;
1083 case ICE_AQ_TASK_CANCELED:
1084 err = ret < 0 ? ret : -ECANCELED;
1085 break;
1086 case ICE_AQ_TASK_COMPLETE:
1087 err = ret < 0 ? ret : 0;
1088 break;
1089 default:
1090 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1091 err = -EINVAL;
1092 break;
1093 }
1094
1095 spin_lock_bh(&pf->aq_wait_lock);
1096 hlist_del(&task->entry);
1097 spin_unlock_bh(&pf->aq_wait_lock);
1098 kfree(task);
1099
1100 return err;
1101}
1102
1103/**
1104 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1105 * @pf: pointer to the PF private structure
1106 * @opcode: the opcode of the event
1107 * @event: the event to check
1108 *
1109 * Loops over the current list of pending threads waiting for an AdminQ event.
1110 * For each matching task, copy the contents of the event into the task
1111 * structure and wake up the thread.
1112 *
1113 * If multiple threads wait for the same opcode, they will all be woken up.
1114 *
1115 * Note that event->msg_buf will only be duplicated if the event has a buffer
1116 * with enough space already allocated. Otherwise, only the descriptor and
1117 * message length will be copied.
1118 *
1119 * Returns: true if an event was found, false otherwise
1120 */
1121static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1122 struct ice_rq_event_info *event)
1123{
1124 struct ice_aq_task *task;
1125 bool found = false;
1126
1127 spin_lock_bh(&pf->aq_wait_lock);
1128 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1129 if (task->state || task->opcode != opcode)
1130 continue;
1131
1132 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1133 task->event->msg_len = event->msg_len;
1134
1135 /* Only copy the data buffer if a destination was set */
1136 if (task->event->msg_buf &&
1137 task->event->buf_len > event->buf_len) {
1138 memcpy(task->event->msg_buf, event->msg_buf,
1139 event->buf_len);
1140 task->event->buf_len = event->buf_len;
1141 }
1142
1143 task->state = ICE_AQ_TASK_COMPLETE;
1144 found = true;
1145 }
1146 spin_unlock_bh(&pf->aq_wait_lock);
1147
1148 if (found)
1149 wake_up(&pf->aq_wait_queue);
1150}
1151
1152/**
1153 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1154 * @pf: the PF private structure
1155 *
1156 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1157 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1158 */
1159static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1160{
1161 struct ice_aq_task *task;
1162
1163 spin_lock_bh(&pf->aq_wait_lock);
1164 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1165 task->state = ICE_AQ_TASK_CANCELED;
1166 spin_unlock_bh(&pf->aq_wait_lock);
1167
1168 wake_up(&pf->aq_wait_queue);
1169}
1170
1171/**
1172 * __ice_clean_ctrlq - helper function to clean controlq rings
1173 * @pf: ptr to struct ice_pf
1174 * @q_type: specific Control queue type
1175 */
1176static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1177{
1178 struct device *dev = ice_pf_to_dev(pf);
1179 struct ice_rq_event_info event;
1180 struct ice_hw *hw = &pf->hw;
1181 struct ice_ctl_q_info *cq;
1182 u16 pending, i = 0;
1183 const char *qtype;
1184 u32 oldval, val;
1185
1186 /* Do not clean control queue if/when PF reset fails */
1187 if (test_bit(__ICE_RESET_FAILED, pf->state))
1188 return 0;
1189
1190 switch (q_type) {
1191 case ICE_CTL_Q_ADMIN:
1192 cq = &hw->adminq;
1193 qtype = "Admin";
1194 break;
1195 case ICE_CTL_Q_MAILBOX:
1196 cq = &hw->mailboxq;
1197 qtype = "Mailbox";
1198 break;
1199 default:
1200 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1201 return 0;
1202 }
1203
1204 /* check for error indications - PF_xx_AxQLEN register layout for
1205 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1206 */
1207 val = rd32(hw, cq->rq.len);
1208 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1209 PF_FW_ARQLEN_ARQCRIT_M)) {
1210 oldval = val;
1211 if (val & PF_FW_ARQLEN_ARQVFE_M)
1212 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1213 qtype);
1214 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1215 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1216 qtype);
1217 }
1218 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1219 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1220 qtype);
1221 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1222 PF_FW_ARQLEN_ARQCRIT_M);
1223 if (oldval != val)
1224 wr32(hw, cq->rq.len, val);
1225 }
1226
1227 val = rd32(hw, cq->sq.len);
1228 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1229 PF_FW_ATQLEN_ATQCRIT_M)) {
1230 oldval = val;
1231 if (val & PF_FW_ATQLEN_ATQVFE_M)
1232 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1233 qtype);
1234 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1235 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1236 qtype);
1237 }
1238 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1239 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1240 qtype);
1241 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1242 PF_FW_ATQLEN_ATQCRIT_M);
1243 if (oldval != val)
1244 wr32(hw, cq->sq.len, val);
1245 }
1246
1247 event.buf_len = cq->rq_buf_size;
1248 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1249 if (!event.msg_buf)
1250 return 0;
1251
1252 do {
1253 enum ice_status ret;
1254 u16 opcode;
1255
1256 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1257 if (ret == ICE_ERR_AQ_NO_WORK)
1258 break;
1259 if (ret) {
1260 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1261 ice_stat_str(ret));
1262 break;
1263 }
1264
1265 opcode = le16_to_cpu(event.desc.opcode);
1266
1267 /* Notify any thread that might be waiting for this event */
1268 ice_aq_check_events(pf, opcode, &event);
1269
1270 switch (opcode) {
1271 case ice_aqc_opc_get_link_status:
1272 if (ice_handle_link_event(pf, &event))
1273 dev_err(dev, "Could not handle link event\n");
1274 break;
1275 case ice_aqc_opc_event_lan_overflow:
1276 ice_vf_lan_overflow_event(pf, &event);
1277 break;
1278 case ice_mbx_opc_send_msg_to_pf:
1279 ice_vc_process_vf_msg(pf, &event);
1280 break;
1281 case ice_aqc_opc_fw_logging:
1282 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1283 break;
1284 case ice_aqc_opc_lldp_set_mib_change:
1285 ice_dcb_process_lldp_set_mib_change(pf, &event);
1286 break;
1287 default:
1288 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1289 qtype, opcode);
1290 break;
1291 }
1292 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1293
1294 kfree(event.msg_buf);
1295
1296 return pending && (i == ICE_DFLT_IRQ_WORK);
1297}
1298
1299/**
1300 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1301 * @hw: pointer to hardware info
1302 * @cq: control queue information
1303 *
1304 * returns true if there are pending messages in a queue, false if there aren't
1305 */
1306static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1307{
1308 u16 ntu;
1309
1310 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1311 return cq->rq.next_to_clean != ntu;
1312}
1313
1314/**
1315 * ice_clean_adminq_subtask - clean the AdminQ rings
1316 * @pf: board private structure
1317 */
1318static void ice_clean_adminq_subtask(struct ice_pf *pf)
1319{
1320 struct ice_hw *hw = &pf->hw;
1321
1322 if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1323 return;
1324
1325 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1326 return;
1327
1328 clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1329
1330 /* There might be a situation where new messages arrive to a control
1331 * queue between processing the last message and clearing the
1332 * EVENT_PENDING bit. So before exiting, check queue head again (using
1333 * ice_ctrlq_pending) and process new messages if any.
1334 */
1335 if (ice_ctrlq_pending(hw, &hw->adminq))
1336 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1337
1338 ice_flush(hw);
1339}
1340
1341/**
1342 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1343 * @pf: board private structure
1344 */
1345static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1346{
1347 struct ice_hw *hw = &pf->hw;
1348
1349 if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1350 return;
1351
1352 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1353 return;
1354
1355 clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1356
1357 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1358 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1359
1360 ice_flush(hw);
1361}
1362
1363/**
1364 * ice_service_task_schedule - schedule the service task to wake up
1365 * @pf: board private structure
1366 *
1367 * If not already scheduled, this puts the task into the work queue.
1368 */
1369void ice_service_task_schedule(struct ice_pf *pf)
1370{
1371 if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
1372 !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
1373 !test_bit(__ICE_NEEDS_RESTART, pf->state))
1374 queue_work(ice_wq, &pf->serv_task);
1375}
1376
1377/**
1378 * ice_service_task_complete - finish up the service task
1379 * @pf: board private structure
1380 */
1381static void ice_service_task_complete(struct ice_pf *pf)
1382{
1383 WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
1384
1385 /* force memory (pf->state) to sync before next service task */
1386 smp_mb__before_atomic();
1387 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1388}
1389
1390/**
1391 * ice_service_task_stop - stop service task and cancel works
1392 * @pf: board private structure
1393 *
1394 * Return 0 if the __ICE_SERVICE_DIS bit was not already set,
1395 * 1 otherwise.
1396 */
1397static int ice_service_task_stop(struct ice_pf *pf)
1398{
1399 int ret;
1400
1401 ret = test_and_set_bit(__ICE_SERVICE_DIS, pf->state);
1402
1403 if (pf->serv_tmr.function)
1404 del_timer_sync(&pf->serv_tmr);
1405 if (pf->serv_task.func)
1406 cancel_work_sync(&pf->serv_task);
1407
1408 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1409 return ret;
1410}
1411
1412/**
1413 * ice_service_task_restart - restart service task and schedule works
1414 * @pf: board private structure
1415 *
1416 * This function is needed for suspend and resume works (e.g WoL scenario)
1417 */
1418static void ice_service_task_restart(struct ice_pf *pf)
1419{
1420 clear_bit(__ICE_SERVICE_DIS, pf->state);
1421 ice_service_task_schedule(pf);
1422}
1423
1424/**
1425 * ice_service_timer - timer callback to schedule service task
1426 * @t: pointer to timer_list
1427 */
1428static void ice_service_timer(struct timer_list *t)
1429{
1430 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1431
1432 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1433 ice_service_task_schedule(pf);
1434}
1435
1436/**
1437 * ice_handle_mdd_event - handle malicious driver detect event
1438 * @pf: pointer to the PF structure
1439 *
1440 * Called from service task. OICR interrupt handler indicates MDD event.
1441 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1442 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1443 * disable the queue, the PF can be configured to reset the VF using ethtool
1444 * private flag mdd-auto-reset-vf.
1445 */
1446static void ice_handle_mdd_event(struct ice_pf *pf)
1447{
1448 struct device *dev = ice_pf_to_dev(pf);
1449 struct ice_hw *hw = &pf->hw;
1450 unsigned int i;
1451 u32 reg;
1452
1453 if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state)) {
1454 /* Since the VF MDD event logging is rate limited, check if
1455 * there are pending MDD events.
1456 */
1457 ice_print_vfs_mdd_events(pf);
1458 return;
1459 }
1460
1461 /* find what triggered an MDD event */
1462 reg = rd32(hw, GL_MDET_TX_PQM);
1463 if (reg & GL_MDET_TX_PQM_VALID_M) {
1464 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1465 GL_MDET_TX_PQM_PF_NUM_S;
1466 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1467 GL_MDET_TX_PQM_VF_NUM_S;
1468 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1469 GL_MDET_TX_PQM_MAL_TYPE_S;
1470 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1471 GL_MDET_TX_PQM_QNUM_S);
1472
1473 if (netif_msg_tx_err(pf))
1474 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1475 event, queue, pf_num, vf_num);
1476 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1477 }
1478
1479 reg = rd32(hw, GL_MDET_TX_TCLAN);
1480 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1481 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1482 GL_MDET_TX_TCLAN_PF_NUM_S;
1483 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1484 GL_MDET_TX_TCLAN_VF_NUM_S;
1485 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1486 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1487 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1488 GL_MDET_TX_TCLAN_QNUM_S);
1489
1490 if (netif_msg_tx_err(pf))
1491 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1492 event, queue, pf_num, vf_num);
1493 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1494 }
1495
1496 reg = rd32(hw, GL_MDET_RX);
1497 if (reg & GL_MDET_RX_VALID_M) {
1498 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1499 GL_MDET_RX_PF_NUM_S;
1500 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1501 GL_MDET_RX_VF_NUM_S;
1502 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1503 GL_MDET_RX_MAL_TYPE_S;
1504 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1505 GL_MDET_RX_QNUM_S);
1506
1507 if (netif_msg_rx_err(pf))
1508 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1509 event, queue, pf_num, vf_num);
1510 wr32(hw, GL_MDET_RX, 0xffffffff);
1511 }
1512
1513 /* check to see if this PF caused an MDD event */
1514 reg = rd32(hw, PF_MDET_TX_PQM);
1515 if (reg & PF_MDET_TX_PQM_VALID_M) {
1516 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1517 if (netif_msg_tx_err(pf))
1518 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1519 }
1520
1521 reg = rd32(hw, PF_MDET_TX_TCLAN);
1522 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1523 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1524 if (netif_msg_tx_err(pf))
1525 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1526 }
1527
1528 reg = rd32(hw, PF_MDET_RX);
1529 if (reg & PF_MDET_RX_VALID_M) {
1530 wr32(hw, PF_MDET_RX, 0xFFFF);
1531 if (netif_msg_rx_err(pf))
1532 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1533 }
1534
1535 /* Check to see if one of the VFs caused an MDD event, and then
1536 * increment counters and set print pending
1537 */
1538 ice_for_each_vf(pf, i) {
1539 struct ice_vf *vf = &pf->vf[i];
1540
1541 reg = rd32(hw, VP_MDET_TX_PQM(i));
1542 if (reg & VP_MDET_TX_PQM_VALID_M) {
1543 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1544 vf->mdd_tx_events.count++;
1545 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1546 if (netif_msg_tx_err(pf))
1547 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1548 i);
1549 }
1550
1551 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1552 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1553 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1554 vf->mdd_tx_events.count++;
1555 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1556 if (netif_msg_tx_err(pf))
1557 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1558 i);
1559 }
1560
1561 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1562 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1563 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1564 vf->mdd_tx_events.count++;
1565 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1566 if (netif_msg_tx_err(pf))
1567 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1568 i);
1569 }
1570
1571 reg = rd32(hw, VP_MDET_RX(i));
1572 if (reg & VP_MDET_RX_VALID_M) {
1573 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1574 vf->mdd_rx_events.count++;
1575 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1576 if (netif_msg_rx_err(pf))
1577 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1578 i);
1579
1580 /* Since the queue is disabled on VF Rx MDD events, the
1581 * PF can be configured to reset the VF through ethtool
1582 * private flag mdd-auto-reset-vf.
1583 */
1584 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1585 /* VF MDD event counters will be cleared by
1586 * reset, so print the event prior to reset.
1587 */
1588 ice_print_vf_rx_mdd_event(vf);
1589 ice_reset_vf(&pf->vf[i], false);
1590 }
1591 }
1592 }
1593
1594 ice_print_vfs_mdd_events(pf);
1595}
1596
1597/**
1598 * ice_force_phys_link_state - Force the physical link state
1599 * @vsi: VSI to force the physical link state to up/down
1600 * @link_up: true/false indicates to set the physical link to up/down
1601 *
1602 * Force the physical link state by getting the current PHY capabilities from
1603 * hardware and setting the PHY config based on the determined capabilities. If
1604 * link changes a link event will be triggered because both the Enable Automatic
1605 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1606 *
1607 * Returns 0 on success, negative on failure
1608 */
1609static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1610{
1611 struct ice_aqc_get_phy_caps_data *pcaps;
1612 struct ice_aqc_set_phy_cfg_data *cfg;
1613 struct ice_port_info *pi;
1614 struct device *dev;
1615 int retcode;
1616
1617 if (!vsi || !vsi->port_info || !vsi->back)
1618 return -EINVAL;
1619 if (vsi->type != ICE_VSI_PF)
1620 return 0;
1621
1622 dev = ice_pf_to_dev(vsi->back);
1623
1624 pi = vsi->port_info;
1625
1626 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1627 if (!pcaps)
1628 return -ENOMEM;
1629
1630 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
1631 NULL);
1632 if (retcode) {
1633 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1634 vsi->vsi_num, retcode);
1635 retcode = -EIO;
1636 goto out;
1637 }
1638
1639 /* No change in link */
1640 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1641 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1642 goto out;
1643
1644 /* Use the current user PHY configuration. The current user PHY
1645 * configuration is initialized during probe from PHY capabilities
1646 * software mode, and updated on set PHY configuration.
1647 */
1648 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1649 if (!cfg) {
1650 retcode = -ENOMEM;
1651 goto out;
1652 }
1653
1654 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1655 if (link_up)
1656 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1657 else
1658 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1659
1660 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1661 if (retcode) {
1662 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1663 vsi->vsi_num, retcode);
1664 retcode = -EIO;
1665 }
1666
1667 kfree(cfg);
1668out:
1669 kfree(pcaps);
1670 return retcode;
1671}
1672
1673/**
1674 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1675 * @pi: port info structure
1676 *
1677 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1678 */
1679static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1680{
1681 struct ice_aqc_get_phy_caps_data *pcaps;
1682 struct ice_pf *pf = pi->hw->back;
1683 enum ice_status status;
1684 int err = 0;
1685
1686 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1687 if (!pcaps)
1688 return -ENOMEM;
1689
1690 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_NVM_CAP, pcaps,
1691 NULL);
1692
1693 if (status) {
1694 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1695 err = -EIO;
1696 goto out;
1697 }
1698
1699 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1700 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1701
1702out:
1703 kfree(pcaps);
1704 return err;
1705}
1706
1707/**
1708 * ice_init_link_dflt_override - Initialize link default override
1709 * @pi: port info structure
1710 *
1711 * Initialize link default override and PHY total port shutdown during probe
1712 */
1713static void ice_init_link_dflt_override(struct ice_port_info *pi)
1714{
1715 struct ice_link_default_override_tlv *ldo;
1716 struct ice_pf *pf = pi->hw->back;
1717
1718 ldo = &pf->link_dflt_override;
1719 if (ice_get_link_default_override(ldo, pi))
1720 return;
1721
1722 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1723 return;
1724
1725 /* Enable Total Port Shutdown (override/replace link-down-on-close
1726 * ethtool private flag) for ports with Port Disable bit set.
1727 */
1728 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1729 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1730}
1731
1732/**
1733 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1734 * @pi: port info structure
1735 *
1736 * If default override is enabled, initialized the user PHY cfg speed and FEC
1737 * settings using the default override mask from the NVM.
1738 *
1739 * The PHY should only be configured with the default override settings the
1740 * first time media is available. The __ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1741 * is used to indicate that the user PHY cfg default override is initialized
1742 * and the PHY has not been configured with the default override settings. The
1743 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1744 * configured.
1745 */
1746static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1747{
1748 struct ice_link_default_override_tlv *ldo;
1749 struct ice_aqc_set_phy_cfg_data *cfg;
1750 struct ice_phy_info *phy = &pi->phy;
1751 struct ice_pf *pf = pi->hw->back;
1752
1753 ldo = &pf->link_dflt_override;
1754
1755 /* If link default override is enabled, use to mask NVM PHY capabilities
1756 * for speed and FEC default configuration.
1757 */
1758 cfg = &phy->curr_user_phy_cfg;
1759
1760 if (ldo->phy_type_low || ldo->phy_type_high) {
1761 cfg->phy_type_low = pf->nvm_phy_type_lo &
1762 cpu_to_le64(ldo->phy_type_low);
1763 cfg->phy_type_high = pf->nvm_phy_type_hi &
1764 cpu_to_le64(ldo->phy_type_high);
1765 }
1766 cfg->link_fec_opt = ldo->fec_options;
1767 phy->curr_user_fec_req = ICE_FEC_AUTO;
1768
1769 set_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1770}
1771
1772/**
1773 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1774 * @pi: port info structure
1775 *
1776 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1777 * mode to default. The PHY defaults are from get PHY capabilities topology
1778 * with media so call when media is first available. An error is returned if
1779 * called when media is not available. The PHY initialization completed state is
1780 * set here.
1781 *
1782 * These configurations are used when setting PHY
1783 * configuration. The user PHY configuration is updated on set PHY
1784 * configuration. Returns 0 on success, negative on failure
1785 */
1786static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1787{
1788 struct ice_aqc_get_phy_caps_data *pcaps;
1789 struct ice_phy_info *phy = &pi->phy;
1790 struct ice_pf *pf = pi->hw->back;
1791 enum ice_status status;
1792 struct ice_vsi *vsi;
1793 int err = 0;
1794
1795 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1796 return -EIO;
1797
1798 vsi = ice_get_main_vsi(pf);
1799 if (!vsi)
1800 return -EINVAL;
1801
1802 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1803 if (!pcaps)
1804 return -ENOMEM;
1805
1806 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1807 NULL);
1808 if (status) {
1809 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1810 err = -EIO;
1811 goto err_out;
1812 }
1813
1814 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1815
1816 /* check if lenient mode is supported and enabled */
1817 if (ice_fw_supports_link_override(&vsi->back->hw) &&
1818 !(pcaps->module_compliance_enforcement &
1819 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1820 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1821
1822 /* if link default override is enabled, initialize user PHY
1823 * configuration with link default override values
1824 */
1825 if (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN) {
1826 ice_init_phy_cfg_dflt_override(pi);
1827 goto out;
1828 }
1829 }
1830
1831 /* if link default override is not enabled, initialize PHY using
1832 * topology with media
1833 */
1834 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1835 pcaps->link_fec_options);
1836 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1837
1838out:
1839 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1840 set_bit(__ICE_PHY_INIT_COMPLETE, pf->state);
1841err_out:
1842 kfree(pcaps);
1843 return err;
1844}
1845
1846/**
1847 * ice_configure_phy - configure PHY
1848 * @vsi: VSI of PHY
1849 *
1850 * Set the PHY configuration. If the current PHY configuration is the same as
1851 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1852 * configure the based get PHY capabilities for topology with media.
1853 */
1854static int ice_configure_phy(struct ice_vsi *vsi)
1855{
1856 struct device *dev = ice_pf_to_dev(vsi->back);
1857 struct ice_aqc_get_phy_caps_data *pcaps;
1858 struct ice_aqc_set_phy_cfg_data *cfg;
1859 struct ice_port_info *pi;
1860 enum ice_status status;
1861 int err = 0;
1862
1863 pi = vsi->port_info;
1864 if (!pi)
1865 return -EINVAL;
1866
1867 /* Ensure we have media as we cannot configure a medialess port */
1868 if (!(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1869 return -EPERM;
1870
1871 ice_print_topo_conflict(vsi);
1872
1873 if (vsi->port_info->phy.link_info.topo_media_conflict ==
1874 ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1875 return -EPERM;
1876
1877 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
1878 return ice_force_phys_link_state(vsi, true);
1879
1880 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1881 if (!pcaps)
1882 return -ENOMEM;
1883
1884 /* Get current PHY config */
1885 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
1886 NULL);
1887 if (status) {
1888 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1889 vsi->vsi_num, ice_stat_str(status));
1890 err = -EIO;
1891 goto done;
1892 }
1893
1894 /* If PHY enable link is configured and configuration has not changed,
1895 * there's nothing to do
1896 */
1897 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1898 ice_phy_caps_equals_cfg(pcaps, &pi->phy.curr_user_phy_cfg))
1899 goto done;
1900
1901 /* Use PHY topology as baseline for configuration */
1902 memset(pcaps, 0, sizeof(*pcaps));
1903 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1904 NULL);
1905 if (status) {
1906 dev_err(dev, "Failed to get PHY topology, VSI %d error %s\n",
1907 vsi->vsi_num, ice_stat_str(status));
1908 err = -EIO;
1909 goto done;
1910 }
1911
1912 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1913 if (!cfg) {
1914 err = -ENOMEM;
1915 goto done;
1916 }
1917
1918 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
1919
1920 /* Speed - If default override pending, use curr_user_phy_cfg set in
1921 * ice_init_phy_user_cfg_ldo.
1922 */
1923 if (test_and_clear_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING,
1924 vsi->back->state)) {
1925 cfg->phy_type_low = pi->phy.curr_user_phy_cfg.phy_type_low;
1926 cfg->phy_type_high = pi->phy.curr_user_phy_cfg.phy_type_high;
1927 } else {
1928 u64 phy_low = 0, phy_high = 0;
1929
1930 ice_update_phy_type(&phy_low, &phy_high,
1931 pi->phy.curr_user_speed_req);
1932 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
1933 cfg->phy_type_high = pcaps->phy_type_high &
1934 cpu_to_le64(phy_high);
1935 }
1936
1937 /* Can't provide what was requested; use PHY capabilities */
1938 if (!cfg->phy_type_low && !cfg->phy_type_high) {
1939 cfg->phy_type_low = pcaps->phy_type_low;
1940 cfg->phy_type_high = pcaps->phy_type_high;
1941 }
1942
1943 /* FEC */
1944 ice_cfg_phy_fec(pi, cfg, pi->phy.curr_user_fec_req);
1945
1946 /* Can't provide what was requested; use PHY capabilities */
1947 if (cfg->link_fec_opt !=
1948 (cfg->link_fec_opt & pcaps->link_fec_options)) {
1949 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
1950 cfg->link_fec_opt = pcaps->link_fec_options;
1951 }
1952
1953 /* Flow Control - always supported; no need to check against
1954 * capabilities
1955 */
1956 ice_cfg_phy_fc(pi, cfg, pi->phy.curr_user_fc_req);
1957
1958 /* Enable link and link update */
1959 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
1960
1961 status = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1962 if (status) {
1963 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
1964 vsi->vsi_num, ice_stat_str(status));
1965 err = -EIO;
1966 }
1967
1968 kfree(cfg);
1969done:
1970 kfree(pcaps);
1971 return err;
1972}
1973
1974/**
1975 * ice_check_media_subtask - Check for media
1976 * @pf: pointer to PF struct
1977 *
1978 * If media is available, then initialize PHY user configuration if it is not
1979 * been, and configure the PHY if the interface is up.
1980 */
1981static void ice_check_media_subtask(struct ice_pf *pf)
1982{
1983 struct ice_port_info *pi;
1984 struct ice_vsi *vsi;
1985 int err;
1986
1987 /* No need to check for media if it's already present */
1988 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
1989 return;
1990
1991 vsi = ice_get_main_vsi(pf);
1992 if (!vsi)
1993 return;
1994
1995 /* Refresh link info and check if media is present */
1996 pi = vsi->port_info;
1997 err = ice_update_link_info(pi);
1998 if (err)
1999 return;
2000
2001 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2002 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state))
2003 ice_init_phy_user_cfg(pi);
2004
2005 /* PHY settings are reset on media insertion, reconfigure
2006 * PHY to preserve settings.
2007 */
2008 if (test_bit(__ICE_DOWN, vsi->state) &&
2009 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2010 return;
2011
2012 err = ice_configure_phy(vsi);
2013 if (!err)
2014 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2015
2016 /* A Link Status Event will be generated; the event handler
2017 * will complete bringing the interface up
2018 */
2019 }
2020}
2021
2022/**
2023 * ice_service_task - manage and run subtasks
2024 * @work: pointer to work_struct contained by the PF struct
2025 */
2026static void ice_service_task(struct work_struct *work)
2027{
2028 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2029 unsigned long start_time = jiffies;
2030
2031 /* subtasks */
2032
2033 /* process reset requests first */
2034 ice_reset_subtask(pf);
2035
2036 /* bail if a reset/recovery cycle is pending or rebuild failed */
2037 if (ice_is_reset_in_progress(pf->state) ||
2038 test_bit(__ICE_SUSPENDED, pf->state) ||
2039 test_bit(__ICE_NEEDS_RESTART, pf->state)) {
2040 ice_service_task_complete(pf);
2041 return;
2042 }
2043
2044 ice_clean_adminq_subtask(pf);
2045 ice_check_media_subtask(pf);
2046 ice_check_for_hang_subtask(pf);
2047 ice_sync_fltr_subtask(pf);
2048 ice_handle_mdd_event(pf);
2049 ice_watchdog_subtask(pf);
2050
2051 if (ice_is_safe_mode(pf)) {
2052 ice_service_task_complete(pf);
2053 return;
2054 }
2055
2056 ice_process_vflr_event(pf);
2057 ice_clean_mailboxq_subtask(pf);
2058 ice_sync_arfs_fltrs(pf);
2059 /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
2060 ice_service_task_complete(pf);
2061
2062 /* If the tasks have taken longer than one service timer period
2063 * or there is more work to be done, reset the service timer to
2064 * schedule the service task now.
2065 */
2066 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2067 test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
2068 test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
2069 test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2070 test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
2071 mod_timer(&pf->serv_tmr, jiffies);
2072}
2073
2074/**
2075 * ice_set_ctrlq_len - helper function to set controlq length
2076 * @hw: pointer to the HW instance
2077 */
2078static void ice_set_ctrlq_len(struct ice_hw *hw)
2079{
2080 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2081 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2082 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2083 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2084 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2085 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2086 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2087 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2088}
2089
2090/**
2091 * ice_schedule_reset - schedule a reset
2092 * @pf: board private structure
2093 * @reset: reset being requested
2094 */
2095int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2096{
2097 struct device *dev = ice_pf_to_dev(pf);
2098
2099 /* bail out if earlier reset has failed */
2100 if (test_bit(__ICE_RESET_FAILED, pf->state)) {
2101 dev_dbg(dev, "earlier reset has failed\n");
2102 return -EIO;
2103 }
2104 /* bail if reset/recovery already in progress */
2105 if (ice_is_reset_in_progress(pf->state)) {
2106 dev_dbg(dev, "Reset already in progress\n");
2107 return -EBUSY;
2108 }
2109
2110 switch (reset) {
2111 case ICE_RESET_PFR:
2112 set_bit(__ICE_PFR_REQ, pf->state);
2113 break;
2114 case ICE_RESET_CORER:
2115 set_bit(__ICE_CORER_REQ, pf->state);
2116 break;
2117 case ICE_RESET_GLOBR:
2118 set_bit(__ICE_GLOBR_REQ, pf->state);
2119 break;
2120 default:
2121 return -EINVAL;
2122 }
2123
2124 ice_service_task_schedule(pf);
2125 return 0;
2126}
2127
2128/**
2129 * ice_irq_affinity_notify - Callback for affinity changes
2130 * @notify: context as to what irq was changed
2131 * @mask: the new affinity mask
2132 *
2133 * This is a callback function used by the irq_set_affinity_notifier function
2134 * so that we may register to receive changes to the irq affinity masks.
2135 */
2136static void
2137ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2138 const cpumask_t *mask)
2139{
2140 struct ice_q_vector *q_vector =
2141 container_of(notify, struct ice_q_vector, affinity_notify);
2142
2143 cpumask_copy(&q_vector->affinity_mask, mask);
2144}
2145
2146/**
2147 * ice_irq_affinity_release - Callback for affinity notifier release
2148 * @ref: internal core kernel usage
2149 *
2150 * This is a callback function used by the irq_set_affinity_notifier function
2151 * to inform the current notification subscriber that they will no longer
2152 * receive notifications.
2153 */
2154static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2155
2156/**
2157 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2158 * @vsi: the VSI being configured
2159 */
2160static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2161{
2162 struct ice_hw *hw = &vsi->back->hw;
2163 int i;
2164
2165 ice_for_each_q_vector(vsi, i)
2166 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2167
2168 ice_flush(hw);
2169 return 0;
2170}
2171
2172/**
2173 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2174 * @vsi: the VSI being configured
2175 * @basename: name for the vector
2176 */
2177static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2178{
2179 int q_vectors = vsi->num_q_vectors;
2180 struct ice_pf *pf = vsi->back;
2181 int base = vsi->base_vector;
2182 struct device *dev;
2183 int rx_int_idx = 0;
2184 int tx_int_idx = 0;
2185 int vector, err;
2186 int irq_num;
2187
2188 dev = ice_pf_to_dev(pf);
2189 for (vector = 0; vector < q_vectors; vector++) {
2190 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2191
2192 irq_num = pf->msix_entries[base + vector].vector;
2193
2194 if (q_vector->tx.ring && q_vector->rx.ring) {
2195 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2196 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2197 tx_int_idx++;
2198 } else if (q_vector->rx.ring) {
2199 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2200 "%s-%s-%d", basename, "rx", rx_int_idx++);
2201 } else if (q_vector->tx.ring) {
2202 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2203 "%s-%s-%d", basename, "tx", tx_int_idx++);
2204 } else {
2205 /* skip this unused q_vector */
2206 continue;
2207 }
2208 err = devm_request_irq(dev, irq_num, vsi->irq_handler, 0,
2209 q_vector->name, q_vector);
2210 if (err) {
2211 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2212 err);
2213 goto free_q_irqs;
2214 }
2215
2216 /* register for affinity change notifications */
2217 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2218 struct irq_affinity_notify *affinity_notify;
2219
2220 affinity_notify = &q_vector->affinity_notify;
2221 affinity_notify->notify = ice_irq_affinity_notify;
2222 affinity_notify->release = ice_irq_affinity_release;
2223 irq_set_affinity_notifier(irq_num, affinity_notify);
2224 }
2225
2226 /* assign the mask for this irq */
2227 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2228 }
2229
2230 vsi->irqs_ready = true;
2231 return 0;
2232
2233free_q_irqs:
2234 while (vector) {
2235 vector--;
2236 irq_num = pf->msix_entries[base + vector].vector;
2237 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2238 irq_set_affinity_notifier(irq_num, NULL);
2239 irq_set_affinity_hint(irq_num, NULL);
2240 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2241 }
2242 return err;
2243}
2244
2245/**
2246 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2247 * @vsi: VSI to setup Tx rings used by XDP
2248 *
2249 * Return 0 on success and negative value on error
2250 */
2251static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2252{
2253 struct device *dev = ice_pf_to_dev(vsi->back);
2254 int i;
2255
2256 for (i = 0; i < vsi->num_xdp_txq; i++) {
2257 u16 xdp_q_idx = vsi->alloc_txq + i;
2258 struct ice_ring *xdp_ring;
2259
2260 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2261
2262 if (!xdp_ring)
2263 goto free_xdp_rings;
2264
2265 xdp_ring->q_index = xdp_q_idx;
2266 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2267 xdp_ring->ring_active = false;
2268 xdp_ring->vsi = vsi;
2269 xdp_ring->netdev = NULL;
2270 xdp_ring->dev = dev;
2271 xdp_ring->count = vsi->num_tx_desc;
2272 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2273 if (ice_setup_tx_ring(xdp_ring))
2274 goto free_xdp_rings;
2275 ice_set_ring_xdp(xdp_ring);
2276 xdp_ring->xsk_umem = ice_xsk_umem(xdp_ring);
2277 }
2278
2279 return 0;
2280
2281free_xdp_rings:
2282 for (; i >= 0; i--)
2283 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2284 ice_free_tx_ring(vsi->xdp_rings[i]);
2285 return -ENOMEM;
2286}
2287
2288/**
2289 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2290 * @vsi: VSI to set the bpf prog on
2291 * @prog: the bpf prog pointer
2292 */
2293static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2294{
2295 struct bpf_prog *old_prog;
2296 int i;
2297
2298 old_prog = xchg(&vsi->xdp_prog, prog);
2299 if (old_prog)
2300 bpf_prog_put(old_prog);
2301
2302 ice_for_each_rxq(vsi, i)
2303 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2304}
2305
2306/**
2307 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2308 * @vsi: VSI to bring up Tx rings used by XDP
2309 * @prog: bpf program that will be assigned to VSI
2310 *
2311 * Return 0 on success and negative value on error
2312 */
2313int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2314{
2315 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2316 int xdp_rings_rem = vsi->num_xdp_txq;
2317 struct ice_pf *pf = vsi->back;
2318 struct ice_qs_cfg xdp_qs_cfg = {
2319 .qs_mutex = &pf->avail_q_mutex,
2320 .pf_map = pf->avail_txqs,
2321 .pf_map_size = pf->max_pf_txqs,
2322 .q_count = vsi->num_xdp_txq,
2323 .scatter_count = ICE_MAX_SCATTER_TXQS,
2324 .vsi_map = vsi->txq_map,
2325 .vsi_map_offset = vsi->alloc_txq,
2326 .mapping_mode = ICE_VSI_MAP_CONTIG
2327 };
2328 enum ice_status status;
2329 struct device *dev;
2330 int i, v_idx;
2331
2332 dev = ice_pf_to_dev(pf);
2333 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2334 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2335 if (!vsi->xdp_rings)
2336 return -ENOMEM;
2337
2338 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2339 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2340 goto err_map_xdp;
2341
2342 if (ice_xdp_alloc_setup_rings(vsi))
2343 goto clear_xdp_rings;
2344
2345 /* follow the logic from ice_vsi_map_rings_to_vectors */
2346 ice_for_each_q_vector(vsi, v_idx) {
2347 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2348 int xdp_rings_per_v, q_id, q_base;
2349
2350 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2351 vsi->num_q_vectors - v_idx);
2352 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2353
2354 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2355 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2356
2357 xdp_ring->q_vector = q_vector;
2358 xdp_ring->next = q_vector->tx.ring;
2359 q_vector->tx.ring = xdp_ring;
2360 }
2361 xdp_rings_rem -= xdp_rings_per_v;
2362 }
2363
2364 /* omit the scheduler update if in reset path; XDP queues will be
2365 * taken into account at the end of ice_vsi_rebuild, where
2366 * ice_cfg_vsi_lan is being called
2367 */
2368 if (ice_is_reset_in_progress(pf->state))
2369 return 0;
2370
2371 /* tell the Tx scheduler that right now we have
2372 * additional queues
2373 */
2374 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2375 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2376
2377 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2378 max_txqs);
2379 if (status) {
2380 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2381 ice_stat_str(status));
2382 goto clear_xdp_rings;
2383 }
2384 ice_vsi_assign_bpf_prog(vsi, prog);
2385
2386 return 0;
2387clear_xdp_rings:
2388 for (i = 0; i < vsi->num_xdp_txq; i++)
2389 if (vsi->xdp_rings[i]) {
2390 kfree_rcu(vsi->xdp_rings[i], rcu);
2391 vsi->xdp_rings[i] = NULL;
2392 }
2393
2394err_map_xdp:
2395 mutex_lock(&pf->avail_q_mutex);
2396 for (i = 0; i < vsi->num_xdp_txq; i++) {
2397 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2398 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2399 }
2400 mutex_unlock(&pf->avail_q_mutex);
2401
2402 devm_kfree(dev, vsi->xdp_rings);
2403 return -ENOMEM;
2404}
2405
2406/**
2407 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2408 * @vsi: VSI to remove XDP rings
2409 *
2410 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2411 * resources
2412 */
2413int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2414{
2415 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2416 struct ice_pf *pf = vsi->back;
2417 int i, v_idx;
2418
2419 /* q_vectors are freed in reset path so there's no point in detaching
2420 * rings; in case of rebuild being triggered not from reset reset bits
2421 * in pf->state won't be set, so additionally check first q_vector
2422 * against NULL
2423 */
2424 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2425 goto free_qmap;
2426
2427 ice_for_each_q_vector(vsi, v_idx) {
2428 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2429 struct ice_ring *ring;
2430
2431 ice_for_each_ring(ring, q_vector->tx)
2432 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2433 break;
2434
2435 /* restore the value of last node prior to XDP setup */
2436 q_vector->tx.ring = ring;
2437 }
2438
2439free_qmap:
2440 mutex_lock(&pf->avail_q_mutex);
2441 for (i = 0; i < vsi->num_xdp_txq; i++) {
2442 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2443 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2444 }
2445 mutex_unlock(&pf->avail_q_mutex);
2446
2447 for (i = 0; i < vsi->num_xdp_txq; i++)
2448 if (vsi->xdp_rings[i]) {
2449 if (vsi->xdp_rings[i]->desc)
2450 ice_free_tx_ring(vsi->xdp_rings[i]);
2451 kfree_rcu(vsi->xdp_rings[i], rcu);
2452 vsi->xdp_rings[i] = NULL;
2453 }
2454
2455 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2456 vsi->xdp_rings = NULL;
2457
2458 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2459 return 0;
2460
2461 ice_vsi_assign_bpf_prog(vsi, NULL);
2462
2463 /* notify Tx scheduler that we destroyed XDP queues and bring
2464 * back the old number of child nodes
2465 */
2466 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2467 max_txqs[i] = vsi->num_txq;
2468
2469 /* change number of XDP Tx queues to 0 */
2470 vsi->num_xdp_txq = 0;
2471
2472 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2473 max_txqs);
2474}
2475
2476/**
2477 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2478 * @vsi: VSI to setup XDP for
2479 * @prog: XDP program
2480 * @extack: netlink extended ack
2481 */
2482static int
2483ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2484 struct netlink_ext_ack *extack)
2485{
2486 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2487 bool if_running = netif_running(vsi->netdev);
2488 int ret = 0, xdp_ring_err = 0;
2489
2490 if (frame_size > vsi->rx_buf_len) {
2491 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2492 return -EOPNOTSUPP;
2493 }
2494
2495 /* need to stop netdev while setting up the program for Rx rings */
2496 if (if_running && !test_and_set_bit(__ICE_DOWN, vsi->state)) {
2497 ret = ice_down(vsi);
2498 if (ret) {
2499 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2500 return ret;
2501 }
2502 }
2503
2504 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2505 vsi->num_xdp_txq = vsi->alloc_rxq;
2506 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2507 if (xdp_ring_err)
2508 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2509 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2510 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2511 if (xdp_ring_err)
2512 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2513 } else {
2514 ice_vsi_assign_bpf_prog(vsi, prog);
2515 }
2516
2517 if (if_running)
2518 ret = ice_up(vsi);
2519
2520 if (!ret && prog && vsi->xsk_umems) {
2521 int i;
2522
2523 ice_for_each_rxq(vsi, i) {
2524 struct ice_ring *rx_ring = vsi->rx_rings[i];
2525
2526 if (rx_ring->xsk_umem)
2527 napi_schedule(&rx_ring->q_vector->napi);
2528 }
2529 }
2530
2531 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2532}
2533
2534/**
2535 * ice_xdp - implements XDP handler
2536 * @dev: netdevice
2537 * @xdp: XDP command
2538 */
2539static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2540{
2541 struct ice_netdev_priv *np = netdev_priv(dev);
2542 struct ice_vsi *vsi = np->vsi;
2543
2544 if (vsi->type != ICE_VSI_PF) {
2545 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2546 return -EINVAL;
2547 }
2548
2549 switch (xdp->command) {
2550 case XDP_SETUP_PROG:
2551 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2552 case XDP_SETUP_XSK_UMEM:
2553 return ice_xsk_umem_setup(vsi, xdp->xsk.umem,
2554 xdp->xsk.queue_id);
2555 default:
2556 return -EINVAL;
2557 }
2558}
2559
2560/**
2561 * ice_ena_misc_vector - enable the non-queue interrupts
2562 * @pf: board private structure
2563 */
2564static void ice_ena_misc_vector(struct ice_pf *pf)
2565{
2566 struct ice_hw *hw = &pf->hw;
2567 u32 val;
2568
2569 /* Disable anti-spoof detection interrupt to prevent spurious event
2570 * interrupts during a function reset. Anti-spoof functionally is
2571 * still supported.
2572 */
2573 val = rd32(hw, GL_MDCK_TX_TDPU);
2574 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2575 wr32(hw, GL_MDCK_TX_TDPU, val);
2576
2577 /* clear things first */
2578 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2579 rd32(hw, PFINT_OICR); /* read to clear */
2580
2581 val = (PFINT_OICR_ECC_ERR_M |
2582 PFINT_OICR_MAL_DETECT_M |
2583 PFINT_OICR_GRST_M |
2584 PFINT_OICR_PCI_EXCEPTION_M |
2585 PFINT_OICR_VFLR_M |
2586 PFINT_OICR_HMC_ERR_M |
2587 PFINT_OICR_PE_CRITERR_M);
2588
2589 wr32(hw, PFINT_OICR_ENA, val);
2590
2591 /* SW_ITR_IDX = 0, but don't change INTENA */
2592 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2593 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2594}
2595
2596/**
2597 * ice_misc_intr - misc interrupt handler
2598 * @irq: interrupt number
2599 * @data: pointer to a q_vector
2600 */
2601static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2602{
2603 struct ice_pf *pf = (struct ice_pf *)data;
2604 struct ice_hw *hw = &pf->hw;
2605 irqreturn_t ret = IRQ_NONE;
2606 struct device *dev;
2607 u32 oicr, ena_mask;
2608
2609 dev = ice_pf_to_dev(pf);
2610 set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
2611 set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2612
2613 oicr = rd32(hw, PFINT_OICR);
2614 ena_mask = rd32(hw, PFINT_OICR_ENA);
2615
2616 if (oicr & PFINT_OICR_SWINT_M) {
2617 ena_mask &= ~PFINT_OICR_SWINT_M;
2618 pf->sw_int_count++;
2619 }
2620
2621 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2622 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2623 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
2624 }
2625 if (oicr & PFINT_OICR_VFLR_M) {
2626 /* disable any further VFLR event notifications */
2627 if (test_bit(__ICE_VF_RESETS_DISABLED, pf->state)) {
2628 u32 reg = rd32(hw, PFINT_OICR_ENA);
2629
2630 reg &= ~PFINT_OICR_VFLR_M;
2631 wr32(hw, PFINT_OICR_ENA, reg);
2632 } else {
2633 ena_mask &= ~PFINT_OICR_VFLR_M;
2634 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
2635 }
2636 }
2637
2638 if (oicr & PFINT_OICR_GRST_M) {
2639 u32 reset;
2640
2641 /* we have a reset warning */
2642 ena_mask &= ~PFINT_OICR_GRST_M;
2643 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2644 GLGEN_RSTAT_RESET_TYPE_S;
2645
2646 if (reset == ICE_RESET_CORER)
2647 pf->corer_count++;
2648 else if (reset == ICE_RESET_GLOBR)
2649 pf->globr_count++;
2650 else if (reset == ICE_RESET_EMPR)
2651 pf->empr_count++;
2652 else
2653 dev_dbg(dev, "Invalid reset type %d\n", reset);
2654
2655 /* If a reset cycle isn't already in progress, we set a bit in
2656 * pf->state so that the service task can start a reset/rebuild.
2657 * We also make note of which reset happened so that peer
2658 * devices/drivers can be informed.
2659 */
2660 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
2661 if (reset == ICE_RESET_CORER)
2662 set_bit(__ICE_CORER_RECV, pf->state);
2663 else if (reset == ICE_RESET_GLOBR)
2664 set_bit(__ICE_GLOBR_RECV, pf->state);
2665 else
2666 set_bit(__ICE_EMPR_RECV, pf->state);
2667
2668 /* There are couple of different bits at play here.
2669 * hw->reset_ongoing indicates whether the hardware is
2670 * in reset. This is set to true when a reset interrupt
2671 * is received and set back to false after the driver
2672 * has determined that the hardware is out of reset.
2673 *
2674 * __ICE_RESET_OICR_RECV in pf->state indicates
2675 * that a post reset rebuild is required before the
2676 * driver is operational again. This is set above.
2677 *
2678 * As this is the start of the reset/rebuild cycle, set
2679 * both to indicate that.
2680 */
2681 hw->reset_ongoing = true;
2682 }
2683 }
2684
2685 if (oicr & PFINT_OICR_HMC_ERR_M) {
2686 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
2687 dev_dbg(dev, "HMC Error interrupt - info 0x%x, data 0x%x\n",
2688 rd32(hw, PFHMC_ERRORINFO),
2689 rd32(hw, PFHMC_ERRORDATA));
2690 }
2691
2692 /* Report any remaining unexpected interrupts */
2693 oicr &= ena_mask;
2694 if (oicr) {
2695 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2696 /* If a critical error is pending there is no choice but to
2697 * reset the device.
2698 */
2699 if (oicr & (PFINT_OICR_PE_CRITERR_M |
2700 PFINT_OICR_PCI_EXCEPTION_M |
2701 PFINT_OICR_ECC_ERR_M)) {
2702 set_bit(__ICE_PFR_REQ, pf->state);
2703 ice_service_task_schedule(pf);
2704 }
2705 }
2706 ret = IRQ_HANDLED;
2707
2708 ice_service_task_schedule(pf);
2709 ice_irq_dynamic_ena(hw, NULL, NULL);
2710
2711 return ret;
2712}
2713
2714/**
2715 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2716 * @hw: pointer to HW structure
2717 */
2718static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2719{
2720 /* disable Admin queue Interrupt causes */
2721 wr32(hw, PFINT_FW_CTL,
2722 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2723
2724 /* disable Mailbox queue Interrupt causes */
2725 wr32(hw, PFINT_MBX_CTL,
2726 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2727
2728 /* disable Control queue Interrupt causes */
2729 wr32(hw, PFINT_OICR_CTL,
2730 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2731
2732 ice_flush(hw);
2733}
2734
2735/**
2736 * ice_free_irq_msix_misc - Unroll misc vector setup
2737 * @pf: board private structure
2738 */
2739static void ice_free_irq_msix_misc(struct ice_pf *pf)
2740{
2741 struct ice_hw *hw = &pf->hw;
2742
2743 ice_dis_ctrlq_interrupts(hw);
2744
2745 /* disable OICR interrupt */
2746 wr32(hw, PFINT_OICR_ENA, 0);
2747 ice_flush(hw);
2748
2749 if (pf->msix_entries) {
2750 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2751 devm_free_irq(ice_pf_to_dev(pf),
2752 pf->msix_entries[pf->oicr_idx].vector, pf);
2753 }
2754
2755 pf->num_avail_sw_msix += 1;
2756 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2757}
2758
2759/**
2760 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2761 * @hw: pointer to HW structure
2762 * @reg_idx: HW vector index to associate the control queue interrupts with
2763 */
2764static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2765{
2766 u32 val;
2767
2768 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2769 PFINT_OICR_CTL_CAUSE_ENA_M);
2770 wr32(hw, PFINT_OICR_CTL, val);
2771
2772 /* enable Admin queue Interrupt causes */
2773 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2774 PFINT_FW_CTL_CAUSE_ENA_M);
2775 wr32(hw, PFINT_FW_CTL, val);
2776
2777 /* enable Mailbox queue Interrupt causes */
2778 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2779 PFINT_MBX_CTL_CAUSE_ENA_M);
2780 wr32(hw, PFINT_MBX_CTL, val);
2781
2782 ice_flush(hw);
2783}
2784
2785/**
2786 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2787 * @pf: board private structure
2788 *
2789 * This sets up the handler for MSIX 0, which is used to manage the
2790 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2791 * when in MSI or Legacy interrupt mode.
2792 */
2793static int ice_req_irq_msix_misc(struct ice_pf *pf)
2794{
2795 struct device *dev = ice_pf_to_dev(pf);
2796 struct ice_hw *hw = &pf->hw;
2797 int oicr_idx, err = 0;
2798
2799 if (!pf->int_name[0])
2800 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2801 dev_driver_string(dev), dev_name(dev));
2802
2803 /* Do not request IRQ but do enable OICR interrupt since settings are
2804 * lost during reset. Note that this function is called only during
2805 * rebuild path and not while reset is in progress.
2806 */
2807 if (ice_is_reset_in_progress(pf->state))
2808 goto skip_req_irq;
2809
2810 /* reserve one vector in irq_tracker for misc interrupts */
2811 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2812 if (oicr_idx < 0)
2813 return oicr_idx;
2814
2815 pf->num_avail_sw_msix -= 1;
2816 pf->oicr_idx = (u16)oicr_idx;
2817
2818 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2819 ice_misc_intr, 0, pf->int_name, pf);
2820 if (err) {
2821 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2822 pf->int_name, err);
2823 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2824 pf->num_avail_sw_msix += 1;
2825 return err;
2826 }
2827
2828skip_req_irq:
2829 ice_ena_misc_vector(pf);
2830
2831 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2832 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2833 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2834
2835 ice_flush(hw);
2836 ice_irq_dynamic_ena(hw, NULL, NULL);
2837
2838 return 0;
2839}
2840
2841/**
2842 * ice_napi_add - register NAPI handler for the VSI
2843 * @vsi: VSI for which NAPI handler is to be registered
2844 *
2845 * This function is only called in the driver's load path. Registering the NAPI
2846 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2847 * reset/rebuild, etc.)
2848 */
2849static void ice_napi_add(struct ice_vsi *vsi)
2850{
2851 int v_idx;
2852
2853 if (!vsi->netdev)
2854 return;
2855
2856 ice_for_each_q_vector(vsi, v_idx)
2857 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2858 ice_napi_poll, NAPI_POLL_WEIGHT);
2859}
2860
2861/**
2862 * ice_set_ops - set netdev and ethtools ops for the given netdev
2863 * @netdev: netdev instance
2864 */
2865static void ice_set_ops(struct net_device *netdev)
2866{
2867 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2868
2869 if (ice_is_safe_mode(pf)) {
2870 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2871 ice_set_ethtool_safe_mode_ops(netdev);
2872 return;
2873 }
2874
2875 netdev->netdev_ops = &ice_netdev_ops;
2876 ice_set_ethtool_ops(netdev);
2877}
2878
2879/**
2880 * ice_set_netdev_features - set features for the given netdev
2881 * @netdev: netdev instance
2882 */
2883static void ice_set_netdev_features(struct net_device *netdev)
2884{
2885 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2886 netdev_features_t csumo_features;
2887 netdev_features_t vlano_features;
2888 netdev_features_t dflt_features;
2889 netdev_features_t tso_features;
2890
2891 if (ice_is_safe_mode(pf)) {
2892 /* safe mode */
2893 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2894 netdev->hw_features = netdev->features;
2895 return;
2896 }
2897
2898 dflt_features = NETIF_F_SG |
2899 NETIF_F_HIGHDMA |
2900 NETIF_F_NTUPLE |
2901 NETIF_F_RXHASH;
2902
2903 csumo_features = NETIF_F_RXCSUM |
2904 NETIF_F_IP_CSUM |
2905 NETIF_F_SCTP_CRC |
2906 NETIF_F_IPV6_CSUM;
2907
2908 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2909 NETIF_F_HW_VLAN_CTAG_TX |
2910 NETIF_F_HW_VLAN_CTAG_RX;
2911
2912 tso_features = NETIF_F_TSO |
2913 NETIF_F_TSO_ECN |
2914 NETIF_F_TSO6 |
2915 NETIF_F_GSO_GRE |
2916 NETIF_F_GSO_UDP_TUNNEL |
2917 NETIF_F_GSO_GRE_CSUM |
2918 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2919 NETIF_F_GSO_PARTIAL |
2920 NETIF_F_GSO_IPXIP4 |
2921 NETIF_F_GSO_IPXIP6 |
2922 NETIF_F_GSO_UDP_L4;
2923
2924 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2925 NETIF_F_GSO_GRE_CSUM;
2926 /* set features that user can change */
2927 netdev->hw_features = dflt_features | csumo_features |
2928 vlano_features | tso_features;
2929
2930 /* add support for HW_CSUM on packets with MPLS header */
2931 netdev->mpls_features = NETIF_F_HW_CSUM;
2932
2933 /* enable features */
2934 netdev->features |= netdev->hw_features;
2935 /* encap and VLAN devices inherit default, csumo and tso features */
2936 netdev->hw_enc_features |= dflt_features | csumo_features |
2937 tso_features;
2938 netdev->vlan_features |= dflt_features | csumo_features |
2939 tso_features;
2940}
2941
2942/**
2943 * ice_cfg_netdev - Allocate, configure and register a netdev
2944 * @vsi: the VSI associated with the new netdev
2945 *
2946 * Returns 0 on success, negative value on failure
2947 */
2948static int ice_cfg_netdev(struct ice_vsi *vsi)
2949{
2950 struct ice_pf *pf = vsi->back;
2951 struct ice_netdev_priv *np;
2952 struct net_device *netdev;
2953 u8 mac_addr[ETH_ALEN];
2954 int err;
2955
2956 err = ice_devlink_create_port(pf);
2957 if (err)
2958 return err;
2959
2960 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
2961 vsi->alloc_rxq);
2962 if (!netdev) {
2963 err = -ENOMEM;
2964 goto err_destroy_devlink_port;
2965 }
2966
2967 vsi->netdev = netdev;
2968 np = netdev_priv(netdev);
2969 np->vsi = vsi;
2970
2971 ice_set_netdev_features(netdev);
2972
2973 ice_set_ops(netdev);
2974
2975 if (vsi->type == ICE_VSI_PF) {
2976 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
2977 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
2978 ether_addr_copy(netdev->dev_addr, mac_addr);
2979 ether_addr_copy(netdev->perm_addr, mac_addr);
2980 }
2981
2982 netdev->priv_flags |= IFF_UNICAST_FLT;
2983
2984 /* Setup netdev TC information */
2985 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
2986
2987 /* setup watchdog timeout value to be 5 second */
2988 netdev->watchdog_timeo = 5 * HZ;
2989
2990 netdev->min_mtu = ETH_MIN_MTU;
2991 netdev->max_mtu = ICE_MAX_MTU;
2992
2993 err = register_netdev(vsi->netdev);
2994 if (err)
2995 goto err_free_netdev;
2996
2997 devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
2998
2999 netif_carrier_off(vsi->netdev);
3000
3001 /* make sure transmit queues start off as stopped */
3002 netif_tx_stop_all_queues(vsi->netdev);
3003
3004 return 0;
3005
3006err_free_netdev:
3007 free_netdev(vsi->netdev);
3008 vsi->netdev = NULL;
3009err_destroy_devlink_port:
3010 ice_devlink_destroy_port(pf);
3011 return err;
3012}
3013
3014/**
3015 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3016 * @lut: Lookup table
3017 * @rss_table_size: Lookup table size
3018 * @rss_size: Range of queue number for hashing
3019 */
3020void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3021{
3022 u16 i;
3023
3024 for (i = 0; i < rss_table_size; i++)
3025 lut[i] = i % rss_size;
3026}
3027
3028/**
3029 * ice_pf_vsi_setup - Set up a PF VSI
3030 * @pf: board private structure
3031 * @pi: pointer to the port_info instance
3032 *
3033 * Returns pointer to the successfully allocated VSI software struct
3034 * on success, otherwise returns NULL on failure.
3035 */
3036static struct ice_vsi *
3037ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3038{
3039 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3040}
3041
3042/**
3043 * ice_ctrl_vsi_setup - Set up a control VSI
3044 * @pf: board private structure
3045 * @pi: pointer to the port_info instance
3046 *
3047 * Returns pointer to the successfully allocated VSI software struct
3048 * on success, otherwise returns NULL on failure.
3049 */
3050static struct ice_vsi *
3051ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3052{
3053 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3054}
3055
3056/**
3057 * ice_lb_vsi_setup - Set up a loopback VSI
3058 * @pf: board private structure
3059 * @pi: pointer to the port_info instance
3060 *
3061 * Returns pointer to the successfully allocated VSI software struct
3062 * on success, otherwise returns NULL on failure.
3063 */
3064struct ice_vsi *
3065ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3066{
3067 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3068}
3069
3070/**
3071 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3072 * @netdev: network interface to be adjusted
3073 * @proto: unused protocol
3074 * @vid: VLAN ID to be added
3075 *
3076 * net_device_ops implementation for adding VLAN IDs
3077 */
3078static int
3079ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3080 u16 vid)
3081{
3082 struct ice_netdev_priv *np = netdev_priv(netdev);
3083 struct ice_vsi *vsi = np->vsi;
3084 int ret;
3085
3086 if (vid >= VLAN_N_VID) {
3087 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
3088 vid, VLAN_N_VID);
3089 return -EINVAL;
3090 }
3091
3092 if (vsi->info.pvid)
3093 return -EINVAL;
3094
3095 /* VLAN 0 is added by default during load/reset */
3096 if (!vid)
3097 return 0;
3098
3099 /* Enable VLAN pruning when a VLAN other than 0 is added */
3100 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3101 ret = ice_cfg_vlan_pruning(vsi, true, false);
3102 if (ret)
3103 return ret;
3104 }
3105
3106 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3107 * packets aren't pruned by the device's internal switch on Rx
3108 */
3109 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3110 if (!ret) {
3111 vsi->vlan_ena = true;
3112 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3113 }
3114
3115 return ret;
3116}
3117
3118/**
3119 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3120 * @netdev: network interface to be adjusted
3121 * @proto: unused protocol
3122 * @vid: VLAN ID to be removed
3123 *
3124 * net_device_ops implementation for removing VLAN IDs
3125 */
3126static int
3127ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3128 u16 vid)
3129{
3130 struct ice_netdev_priv *np = netdev_priv(netdev);
3131 struct ice_vsi *vsi = np->vsi;
3132 int ret;
3133
3134 if (vsi->info.pvid)
3135 return -EINVAL;
3136
3137 /* don't allow removal of VLAN 0 */
3138 if (!vid)
3139 return 0;
3140
3141 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3142 * information
3143 */
3144 ret = ice_vsi_kill_vlan(vsi, vid);
3145 if (ret)
3146 return ret;
3147
3148 /* Disable pruning when VLAN 0 is the only VLAN rule */
3149 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3150 ret = ice_cfg_vlan_pruning(vsi, false, false);
3151
3152 vsi->vlan_ena = false;
3153 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3154 return ret;
3155}
3156
3157/**
3158 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3159 * @pf: board private structure
3160 *
3161 * Returns 0 on success, negative value on failure
3162 */
3163static int ice_setup_pf_sw(struct ice_pf *pf)
3164{
3165 struct ice_vsi *vsi;
3166 int status = 0;
3167
3168 if (ice_is_reset_in_progress(pf->state))
3169 return -EBUSY;
3170
3171 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3172 if (!vsi)
3173 return -ENOMEM;
3174
3175 status = ice_cfg_netdev(vsi);
3176 if (status) {
3177 status = -ENODEV;
3178 goto unroll_vsi_setup;
3179 }
3180 /* netdev has to be configured before setting frame size */
3181 ice_vsi_cfg_frame_size(vsi);
3182
3183 /* Setup DCB netlink interface */
3184 ice_dcbnl_setup(vsi);
3185
3186 /* registering the NAPI handler requires both the queues and
3187 * netdev to be created, which are done in ice_pf_vsi_setup()
3188 * and ice_cfg_netdev() respectively
3189 */
3190 ice_napi_add(vsi);
3191
3192 status = ice_set_cpu_rx_rmap(vsi);
3193 if (status) {
3194 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3195 vsi->vsi_num, status);
3196 status = -EINVAL;
3197 goto unroll_napi_add;
3198 }
3199 status = ice_init_mac_fltr(pf);
3200 if (status)
3201 goto free_cpu_rx_map;
3202
3203 return status;
3204
3205free_cpu_rx_map:
3206 ice_free_cpu_rx_rmap(vsi);
3207
3208unroll_napi_add:
3209 if (vsi) {
3210 ice_napi_del(vsi);
3211 if (vsi->netdev) {
3212 if (vsi->netdev->reg_state == NETREG_REGISTERED)
3213 unregister_netdev(vsi->netdev);
3214 free_netdev(vsi->netdev);
3215 vsi->netdev = NULL;
3216 }
3217 }
3218
3219unroll_vsi_setup:
3220 ice_vsi_release(vsi);
3221 return status;
3222}
3223
3224/**
3225 * ice_get_avail_q_count - Get count of queues in use
3226 * @pf_qmap: bitmap to get queue use count from
3227 * @lock: pointer to a mutex that protects access to pf_qmap
3228 * @size: size of the bitmap
3229 */
3230static u16
3231ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3232{
3233 unsigned long bit;
3234 u16 count = 0;
3235
3236 mutex_lock(lock);
3237 for_each_clear_bit(bit, pf_qmap, size)
3238 count++;
3239 mutex_unlock(lock);
3240
3241 return count;
3242}
3243
3244/**
3245 * ice_get_avail_txq_count - Get count of Tx queues in use
3246 * @pf: pointer to an ice_pf instance
3247 */
3248u16 ice_get_avail_txq_count(struct ice_pf *pf)
3249{
3250 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3251 pf->max_pf_txqs);
3252}
3253
3254/**
3255 * ice_get_avail_rxq_count - Get count of Rx queues in use
3256 * @pf: pointer to an ice_pf instance
3257 */
3258u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3259{
3260 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3261 pf->max_pf_rxqs);
3262}
3263
3264/**
3265 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3266 * @pf: board private structure to initialize
3267 */
3268static void ice_deinit_pf(struct ice_pf *pf)
3269{
3270 ice_service_task_stop(pf);
3271 mutex_destroy(&pf->sw_mutex);
3272 mutex_destroy(&pf->tc_mutex);
3273 mutex_destroy(&pf->avail_q_mutex);
3274
3275 if (pf->avail_txqs) {
3276 bitmap_free(pf->avail_txqs);
3277 pf->avail_txqs = NULL;
3278 }
3279
3280 if (pf->avail_rxqs) {
3281 bitmap_free(pf->avail_rxqs);
3282 pf->avail_rxqs = NULL;
3283 }
3284}
3285
3286/**
3287 * ice_set_pf_caps - set PFs capability flags
3288 * @pf: pointer to the PF instance
3289 */
3290static void ice_set_pf_caps(struct ice_pf *pf)
3291{
3292 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3293
3294 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3295 if (func_caps->common_cap.dcb)
3296 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3297 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3298 if (func_caps->common_cap.sr_iov_1_1) {
3299 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3300 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3301 ICE_MAX_VF_COUNT);
3302 }
3303 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3304 if (func_caps->common_cap.rss_table_size)
3305 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3306
3307 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3308 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3309 u16 unused;
3310
3311 /* ctrl_vsi_idx will be set to a valid value when flow director
3312 * is setup by ice_init_fdir
3313 */
3314 pf->ctrl_vsi_idx = ICE_NO_VSI;
3315 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3316 /* force guaranteed filter pool for PF */
3317 ice_alloc_fd_guar_item(&pf->hw, &unused,
3318 func_caps->fd_fltr_guar);
3319 /* force shared filter pool for PF */
3320 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3321 func_caps->fd_fltr_best_effort);
3322 }
3323
3324 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3325 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3326}
3327
3328/**
3329 * ice_init_pf - Initialize general software structures (struct ice_pf)
3330 * @pf: board private structure to initialize
3331 */
3332static int ice_init_pf(struct ice_pf *pf)
3333{
3334 ice_set_pf_caps(pf);
3335
3336 mutex_init(&pf->sw_mutex);
3337 mutex_init(&pf->tc_mutex);
3338
3339 INIT_HLIST_HEAD(&pf->aq_wait_list);
3340 spin_lock_init(&pf->aq_wait_lock);
3341 init_waitqueue_head(&pf->aq_wait_queue);
3342
3343 /* setup service timer and periodic service task */
3344 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3345 pf->serv_tmr_period = HZ;
3346 INIT_WORK(&pf->serv_task, ice_service_task);
3347 clear_bit(__ICE_SERVICE_SCHED, pf->state);
3348
3349 mutex_init(&pf->avail_q_mutex);
3350 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3351 if (!pf->avail_txqs)
3352 return -ENOMEM;
3353
3354 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3355 if (!pf->avail_rxqs) {
3356 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3357 pf->avail_txqs = NULL;
3358 return -ENOMEM;
3359 }
3360
3361 return 0;
3362}
3363
3364/**
3365 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3366 * @pf: board private structure
3367 *
3368 * compute the number of MSIX vectors required (v_budget) and request from
3369 * the OS. Return the number of vectors reserved or negative on failure
3370 */
3371static int ice_ena_msix_range(struct ice_pf *pf)
3372{
3373 struct device *dev = ice_pf_to_dev(pf);
3374 int v_left, v_actual, v_budget = 0;
3375 int needed, err, i;
3376
3377 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3378
3379 /* reserve one vector for miscellaneous handler */
3380 needed = 1;
3381 if (v_left < needed)
3382 goto no_hw_vecs_left_err;
3383 v_budget += needed;
3384 v_left -= needed;
3385
3386 /* reserve vectors for LAN traffic */
3387 needed = min_t(int, num_online_cpus(), v_left);
3388 if (v_left < needed)
3389 goto no_hw_vecs_left_err;
3390 pf->num_lan_msix = needed;
3391 v_budget += needed;
3392 v_left -= needed;
3393
3394 /* reserve one vector for flow director */
3395 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3396 needed = ICE_FDIR_MSIX;
3397 if (v_left < needed)
3398 goto no_hw_vecs_left_err;
3399 v_budget += needed;
3400 v_left -= needed;
3401 }
3402
3403 pf->msix_entries = devm_kcalloc(dev, v_budget,
3404 sizeof(*pf->msix_entries), GFP_KERNEL);
3405
3406 if (!pf->msix_entries) {
3407 err = -ENOMEM;
3408 goto exit_err;
3409 }
3410
3411 for (i = 0; i < v_budget; i++)
3412 pf->msix_entries[i].entry = i;
3413
3414 /* actually reserve the vectors */
3415 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3416 ICE_MIN_MSIX, v_budget);
3417
3418 if (v_actual < 0) {
3419 dev_err(dev, "unable to reserve MSI-X vectors\n");
3420 err = v_actual;
3421 goto msix_err;
3422 }
3423
3424 if (v_actual < v_budget) {
3425 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3426 v_budget, v_actual);
3427/* 2 vectors each for LAN and RDMA (traffic + OICR), one for flow director */
3428#define ICE_MIN_LAN_VECS 2
3429#define ICE_MIN_RDMA_VECS 2
3430#define ICE_MIN_VECS (ICE_MIN_LAN_VECS + ICE_MIN_RDMA_VECS + 1)
3431
3432 if (v_actual < ICE_MIN_LAN_VECS) {
3433 /* error if we can't get minimum vectors */
3434 pci_disable_msix(pf->pdev);
3435 err = -ERANGE;
3436 goto msix_err;
3437 } else {
3438 pf->num_lan_msix = ICE_MIN_LAN_VECS;
3439 }
3440 }
3441
3442 return v_actual;
3443
3444msix_err:
3445 devm_kfree(dev, pf->msix_entries);
3446 goto exit_err;
3447
3448no_hw_vecs_left_err:
3449 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3450 needed, v_left);
3451 err = -ERANGE;
3452exit_err:
3453 pf->num_lan_msix = 0;
3454 return err;
3455}
3456
3457/**
3458 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3459 * @pf: board private structure
3460 */
3461static void ice_dis_msix(struct ice_pf *pf)
3462{
3463 pci_disable_msix(pf->pdev);
3464 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3465 pf->msix_entries = NULL;
3466}
3467
3468/**
3469 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3470 * @pf: board private structure
3471 */
3472static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3473{
3474 ice_dis_msix(pf);
3475
3476 if (pf->irq_tracker) {
3477 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3478 pf->irq_tracker = NULL;
3479 }
3480}
3481
3482/**
3483 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3484 * @pf: board private structure to initialize
3485 */
3486static int ice_init_interrupt_scheme(struct ice_pf *pf)
3487{
3488 int vectors;
3489
3490 vectors = ice_ena_msix_range(pf);
3491
3492 if (vectors < 0)
3493 return vectors;
3494
3495 /* set up vector assignment tracking */
3496 pf->irq_tracker =
3497 devm_kzalloc(ice_pf_to_dev(pf), sizeof(*pf->irq_tracker) +
3498 (sizeof(u16) * vectors), GFP_KERNEL);
3499 if (!pf->irq_tracker) {
3500 ice_dis_msix(pf);
3501 return -ENOMEM;
3502 }
3503
3504 /* populate SW interrupts pool with number of OS granted IRQs. */
3505 pf->num_avail_sw_msix = (u16)vectors;
3506 pf->irq_tracker->num_entries = (u16)vectors;
3507 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3508
3509 return 0;
3510}
3511
3512/**
3513 * ice_is_wol_supported - get NVM state of WoL
3514 * @pf: board private structure
3515 *
3516 * Check if WoL is supported based on the HW configuration.
3517 * Returns true if NVM supports and enables WoL for this port, false otherwise
3518 */
3519bool ice_is_wol_supported(struct ice_pf *pf)
3520{
3521 struct ice_hw *hw = &pf->hw;
3522 u16 wol_ctrl;
3523
3524 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3525 * word) indicates WoL is not supported on the corresponding PF ID.
3526 */
3527 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3528 return false;
3529
3530 return !(BIT(hw->pf_id) & wol_ctrl);
3531}
3532
3533/**
3534 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3535 * @vsi: VSI being changed
3536 * @new_rx: new number of Rx queues
3537 * @new_tx: new number of Tx queues
3538 *
3539 * Only change the number of queues if new_tx, or new_rx is non-0.
3540 *
3541 * Returns 0 on success.
3542 */
3543int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3544{
3545 struct ice_pf *pf = vsi->back;
3546 int err = 0, timeout = 50;
3547
3548 if (!new_rx && !new_tx)
3549 return -EINVAL;
3550
3551 while (test_and_set_bit(__ICE_CFG_BUSY, pf->state)) {
3552 timeout--;
3553 if (!timeout)
3554 return -EBUSY;
3555 usleep_range(1000, 2000);
3556 }
3557
3558 if (new_tx)
3559 vsi->req_txq = (u16)new_tx;
3560 if (new_rx)
3561 vsi->req_rxq = (u16)new_rx;
3562
3563 /* set for the next time the netdev is started */
3564 if (!netif_running(vsi->netdev)) {
3565 ice_vsi_rebuild(vsi, false);
3566 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3567 goto done;
3568 }
3569
3570 ice_vsi_close(vsi);
3571 ice_vsi_rebuild(vsi, false);
3572 ice_pf_dcb_recfg(pf);
3573 ice_vsi_open(vsi);
3574done:
3575 clear_bit(__ICE_CFG_BUSY, pf->state);
3576 return err;
3577}
3578
3579/**
3580 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3581 * @pf: PF to configure
3582 *
3583 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3584 * VSI can still Tx/Rx VLAN tagged packets.
3585 */
3586static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3587{
3588 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3589 struct ice_vsi_ctx *ctxt;
3590 enum ice_status status;
3591 struct ice_hw *hw;
3592
3593 if (!vsi)
3594 return;
3595
3596 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3597 if (!ctxt)
3598 return;
3599
3600 hw = &pf->hw;
3601 ctxt->info = vsi->info;
3602
3603 ctxt->info.valid_sections =
3604 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3605 ICE_AQ_VSI_PROP_SECURITY_VALID |
3606 ICE_AQ_VSI_PROP_SW_VALID);
3607
3608 /* disable VLAN anti-spoof */
3609 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3610 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3611
3612 /* disable VLAN pruning and keep all other settings */
3613 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3614
3615 /* allow all VLANs on Tx and don't strip on Rx */
3616 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3617 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3618
3619 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3620 if (status) {
3621 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3622 ice_stat_str(status),
3623 ice_aq_str(hw->adminq.sq_last_status));
3624 } else {
3625 vsi->info.sec_flags = ctxt->info.sec_flags;
3626 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3627 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3628 }
3629
3630 kfree(ctxt);
3631}
3632
3633/**
3634 * ice_log_pkg_init - log result of DDP package load
3635 * @hw: pointer to hardware info
3636 * @status: status of package load
3637 */
3638static void
3639ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3640{
3641 struct ice_pf *pf = (struct ice_pf *)hw->back;
3642 struct device *dev = ice_pf_to_dev(pf);
3643
3644 switch (*status) {
3645 case ICE_SUCCESS:
3646 /* The package download AdminQ command returned success because
3647 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3648 * already a package loaded on the device.
3649 */
3650 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3651 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3652 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3653 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3654 !memcmp(hw->pkg_name, hw->active_pkg_name,
3655 sizeof(hw->pkg_name))) {
3656 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3657 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3658 hw->active_pkg_name,
3659 hw->active_pkg_ver.major,
3660 hw->active_pkg_ver.minor,
3661 hw->active_pkg_ver.update,
3662 hw->active_pkg_ver.draft);
3663 else
3664 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3665 hw->active_pkg_name,
3666 hw->active_pkg_ver.major,
3667 hw->active_pkg_ver.minor,
3668 hw->active_pkg_ver.update,
3669 hw->active_pkg_ver.draft);
3670 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3671 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3672 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",
3673 hw->active_pkg_name,
3674 hw->active_pkg_ver.major,
3675 hw->active_pkg_ver.minor,
3676 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3677 *status = ICE_ERR_NOT_SUPPORTED;
3678 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3679 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3680 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",
3681 hw->active_pkg_name,
3682 hw->active_pkg_ver.major,
3683 hw->active_pkg_ver.minor,
3684 hw->active_pkg_ver.update,
3685 hw->active_pkg_ver.draft,
3686 hw->pkg_name,
3687 hw->pkg_ver.major,
3688 hw->pkg_ver.minor,
3689 hw->pkg_ver.update,
3690 hw->pkg_ver.draft);
3691 } else {
3692 dev_err(dev, "An unknown error occurred when loading the DDP package, please reboot the system. If the problem persists, update the NVM. Entering Safe Mode.\n");
3693 *status = ICE_ERR_NOT_SUPPORTED;
3694 }
3695 break;
3696 case ICE_ERR_FW_DDP_MISMATCH:
3697 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");
3698 break;
3699 case ICE_ERR_BUF_TOO_SHORT:
3700 case ICE_ERR_CFG:
3701 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3702 break;
3703 case ICE_ERR_NOT_SUPPORTED:
3704 /* Package File version not supported */
3705 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3706 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3707 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3708 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3709 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3710 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3711 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3712 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",
3713 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3714 break;
3715 case ICE_ERR_AQ_ERROR:
3716 switch (hw->pkg_dwnld_status) {
3717 case ICE_AQ_RC_ENOSEC:
3718 case ICE_AQ_RC_EBADSIG:
3719 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");
3720 return;
3721 case ICE_AQ_RC_ESVN:
3722 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");
3723 return;
3724 case ICE_AQ_RC_EBADMAN:
3725 case ICE_AQ_RC_EBADBUF:
3726 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3727 /* poll for reset to complete */
3728 if (ice_check_reset(hw))
3729 dev_err(dev, "Error resetting device. Please reload the driver\n");
3730 return;
3731 default:
3732 break;
3733 }
3734 fallthrough;
3735 default:
3736 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3737 *status);
3738 break;
3739 }
3740}
3741
3742/**
3743 * ice_load_pkg - load/reload the DDP Package file
3744 * @firmware: firmware structure when firmware requested or NULL for reload
3745 * @pf: pointer to the PF instance
3746 *
3747 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3748 * initialize HW tables.
3749 */
3750static void
3751ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3752{
3753 enum ice_status status = ICE_ERR_PARAM;
3754 struct device *dev = ice_pf_to_dev(pf);
3755 struct ice_hw *hw = &pf->hw;
3756
3757 /* Load DDP Package */
3758 if (firmware && !hw->pkg_copy) {
3759 status = ice_copy_and_init_pkg(hw, firmware->data,
3760 firmware->size);
3761 ice_log_pkg_init(hw, &status);
3762 } else if (!firmware && hw->pkg_copy) {
3763 /* Reload package during rebuild after CORER/GLOBR reset */
3764 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3765 ice_log_pkg_init(hw, &status);
3766 } else {
3767 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3768 }
3769
3770 if (status) {
3771 /* Safe Mode */
3772 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3773 return;
3774 }
3775
3776 /* Successful download package is the precondition for advanced
3777 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3778 */
3779 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3780}
3781
3782/**
3783 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3784 * @pf: pointer to the PF structure
3785 *
3786 * There is no error returned here because the driver should be able to handle
3787 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3788 * specifically with Tx.
3789 */
3790static void ice_verify_cacheline_size(struct ice_pf *pf)
3791{
3792 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3793 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3794 ICE_CACHE_LINE_BYTES);
3795}
3796
3797/**
3798 * ice_send_version - update firmware with driver version
3799 * @pf: PF struct
3800 *
3801 * Returns ICE_SUCCESS on success, else error code
3802 */
3803static enum ice_status ice_send_version(struct ice_pf *pf)
3804{
3805 struct ice_driver_ver dv;
3806
3807 dv.major_ver = 0xff;
3808 dv.minor_ver = 0xff;
3809 dv.build_ver = 0xff;
3810 dv.subbuild_ver = 0;
3811 strscpy((char *)dv.driver_string, UTS_RELEASE,
3812 sizeof(dv.driver_string));
3813 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3814}
3815
3816/**
3817 * ice_init_fdir - Initialize flow director VSI and configuration
3818 * @pf: pointer to the PF instance
3819 *
3820 * returns 0 on success, negative on error
3821 */
3822static int ice_init_fdir(struct ice_pf *pf)
3823{
3824 struct device *dev = ice_pf_to_dev(pf);
3825 struct ice_vsi *ctrl_vsi;
3826 int err;
3827
3828 /* Side Band Flow Director needs to have a control VSI.
3829 * Allocate it and store it in the PF.
3830 */
3831 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3832 if (!ctrl_vsi) {
3833 dev_dbg(dev, "could not create control VSI\n");
3834 return -ENOMEM;
3835 }
3836
3837 err = ice_vsi_open_ctrl(ctrl_vsi);
3838 if (err) {
3839 dev_dbg(dev, "could not open control VSI\n");
3840 goto err_vsi_open;
3841 }
3842
3843 mutex_init(&pf->hw.fdir_fltr_lock);
3844
3845 err = ice_fdir_create_dflt_rules(pf);
3846 if (err)
3847 goto err_fdir_rule;
3848
3849 return 0;
3850
3851err_fdir_rule:
3852 ice_fdir_release_flows(&pf->hw);
3853 ice_vsi_close(ctrl_vsi);
3854err_vsi_open:
3855 ice_vsi_release(ctrl_vsi);
3856 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3857 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3858 pf->ctrl_vsi_idx = ICE_NO_VSI;
3859 }
3860 return err;
3861}
3862
3863/**
3864 * ice_get_opt_fw_name - return optional firmware file name or NULL
3865 * @pf: pointer to the PF instance
3866 */
3867static char *ice_get_opt_fw_name(struct ice_pf *pf)
3868{
3869 /* Optional firmware name same as default with additional dash
3870 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3871 */
3872 struct pci_dev *pdev = pf->pdev;
3873 char *opt_fw_filename;
3874 u64 dsn;
3875
3876 /* Determine the name of the optional file using the DSN (two
3877 * dwords following the start of the DSN Capability).
3878 */
3879 dsn = pci_get_dsn(pdev);
3880 if (!dsn)
3881 return NULL;
3882
3883 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3884 if (!opt_fw_filename)
3885 return NULL;
3886
3887 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3888 ICE_DDP_PKG_PATH, dsn);
3889
3890 return opt_fw_filename;
3891}
3892
3893/**
3894 * ice_request_fw - Device initialization routine
3895 * @pf: pointer to the PF instance
3896 */
3897static void ice_request_fw(struct ice_pf *pf)
3898{
3899 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3900 const struct firmware *firmware = NULL;
3901 struct device *dev = ice_pf_to_dev(pf);
3902 int err = 0;
3903
3904 /* optional device-specific DDP (if present) overrides the default DDP
3905 * package file. kernel logs a debug message if the file doesn't exist,
3906 * and warning messages for other errors.
3907 */
3908 if (opt_fw_filename) {
3909 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3910 if (err) {
3911 kfree(opt_fw_filename);
3912 goto dflt_pkg_load;
3913 }
3914
3915 /* request for firmware was successful. Download to device */
3916 ice_load_pkg(firmware, pf);
3917 kfree(opt_fw_filename);
3918 release_firmware(firmware);
3919 return;
3920 }
3921
3922dflt_pkg_load:
3923 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3924 if (err) {
3925 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3926 return;
3927 }
3928
3929 /* request for firmware was successful. Download to device */
3930 ice_load_pkg(firmware, pf);
3931 release_firmware(firmware);
3932}
3933
3934/**
3935 * ice_print_wake_reason - show the wake up cause in the log
3936 * @pf: pointer to the PF struct
3937 */
3938static void ice_print_wake_reason(struct ice_pf *pf)
3939{
3940 u32 wus = pf->wakeup_reason;
3941 const char *wake_str;
3942
3943 /* if no wake event, nothing to print */
3944 if (!wus)
3945 return;
3946
3947 if (wus & PFPM_WUS_LNKC_M)
3948 wake_str = "Link\n";
3949 else if (wus & PFPM_WUS_MAG_M)
3950 wake_str = "Magic Packet\n";
3951 else if (wus & PFPM_WUS_MNG_M)
3952 wake_str = "Management\n";
3953 else if (wus & PFPM_WUS_FW_RST_WK_M)
3954 wake_str = "Firmware Reset\n";
3955 else
3956 wake_str = "Unknown\n";
3957
3958 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
3959}
3960
3961/**
3962 * ice_probe - Device initialization routine
3963 * @pdev: PCI device information struct
3964 * @ent: entry in ice_pci_tbl
3965 *
3966 * Returns 0 on success, negative on failure
3967 */
3968static int
3969ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
3970{
3971 struct device *dev = &pdev->dev;
3972 struct ice_pf *pf;
3973 struct ice_hw *hw;
3974 int err;
3975
3976 /* this driver uses devres, see
3977 * Documentation/driver-api/driver-model/devres.rst
3978 */
3979 err = pcim_enable_device(pdev);
3980 if (err)
3981 return err;
3982
3983 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
3984 if (err) {
3985 dev_err(dev, "BAR0 I/O map error %d\n", err);
3986 return err;
3987 }
3988
3989 pf = ice_allocate_pf(dev);
3990 if (!pf)
3991 return -ENOMEM;
3992
3993 /* set up for high or low DMA */
3994 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
3995 if (err)
3996 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
3997 if (err) {
3998 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
3999 return err;
4000 }
4001
4002 pci_enable_pcie_error_reporting(pdev);
4003 pci_set_master(pdev);
4004
4005 pf->pdev = pdev;
4006 pci_set_drvdata(pdev, pf);
4007 set_bit(__ICE_DOWN, pf->state);
4008 /* Disable service task until DOWN bit is cleared */
4009 set_bit(__ICE_SERVICE_DIS, pf->state);
4010
4011 hw = &pf->hw;
4012 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4013 pci_save_state(pdev);
4014
4015 hw->back = pf;
4016 hw->vendor_id = pdev->vendor;
4017 hw->device_id = pdev->device;
4018 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4019 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4020 hw->subsystem_device_id = pdev->subsystem_device;
4021 hw->bus.device = PCI_SLOT(pdev->devfn);
4022 hw->bus.func = PCI_FUNC(pdev->devfn);
4023 ice_set_ctrlq_len(hw);
4024
4025 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4026
4027 err = ice_devlink_register(pf);
4028 if (err) {
4029 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4030 goto err_exit_unroll;
4031 }
4032
4033#ifndef CONFIG_DYNAMIC_DEBUG
4034 if (debug < -1)
4035 hw->debug_mask = debug;
4036#endif
4037
4038 err = ice_init_hw(hw);
4039 if (err) {
4040 dev_err(dev, "ice_init_hw failed: %d\n", err);
4041 err = -EIO;
4042 goto err_exit_unroll;
4043 }
4044
4045 ice_request_fw(pf);
4046
4047 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4048 * set in pf->state, which will cause ice_is_safe_mode to return
4049 * true
4050 */
4051 if (ice_is_safe_mode(pf)) {
4052 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4053 /* we already got function/device capabilities but these don't
4054 * reflect what the driver needs to do in safe mode. Instead of
4055 * adding conditional logic everywhere to ignore these
4056 * device/function capabilities, override them.
4057 */
4058 ice_set_safe_mode_caps(hw);
4059 }
4060
4061 err = ice_init_pf(pf);
4062 if (err) {
4063 dev_err(dev, "ice_init_pf failed: %d\n", err);
4064 goto err_init_pf_unroll;
4065 }
4066
4067 ice_devlink_init_regions(pf);
4068
4069 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4070 if (!pf->num_alloc_vsi) {
4071 err = -EIO;
4072 goto err_init_pf_unroll;
4073 }
4074
4075 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4076 GFP_KERNEL);
4077 if (!pf->vsi) {
4078 err = -ENOMEM;
4079 goto err_init_pf_unroll;
4080 }
4081
4082 err = ice_init_interrupt_scheme(pf);
4083 if (err) {
4084 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4085 err = -EIO;
4086 goto err_init_vsi_unroll;
4087 }
4088
4089 /* In case of MSIX we are going to setup the misc vector right here
4090 * to handle admin queue events etc. In case of legacy and MSI
4091 * the misc functionality and queue processing is combined in
4092 * the same vector and that gets setup at open.
4093 */
4094 err = ice_req_irq_msix_misc(pf);
4095 if (err) {
4096 dev_err(dev, "setup of misc vector failed: %d\n", err);
4097 goto err_init_interrupt_unroll;
4098 }
4099
4100 /* create switch struct for the switch element created by FW on boot */
4101 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4102 if (!pf->first_sw) {
4103 err = -ENOMEM;
4104 goto err_msix_misc_unroll;
4105 }
4106
4107 if (hw->evb_veb)
4108 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4109 else
4110 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4111
4112 pf->first_sw->pf = pf;
4113
4114 /* record the sw_id available for later use */
4115 pf->first_sw->sw_id = hw->port_info->sw_id;
4116
4117 err = ice_setup_pf_sw(pf);
4118 if (err) {
4119 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4120 goto err_alloc_sw_unroll;
4121 }
4122
4123 clear_bit(__ICE_SERVICE_DIS, pf->state);
4124
4125 /* tell the firmware we are up */
4126 err = ice_send_version(pf);
4127 if (err) {
4128 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4129 UTS_RELEASE, err);
4130 goto err_send_version_unroll;
4131 }
4132
4133 /* since everything is good, start the service timer */
4134 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4135
4136 err = ice_init_link_events(pf->hw.port_info);
4137 if (err) {
4138 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4139 goto err_send_version_unroll;
4140 }
4141
4142 err = ice_init_nvm_phy_type(pf->hw.port_info);
4143 if (err) {
4144 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4145 goto err_send_version_unroll;
4146 }
4147
4148 err = ice_update_link_info(pf->hw.port_info);
4149 if (err) {
4150 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4151 goto err_send_version_unroll;
4152 }
4153
4154 ice_init_link_dflt_override(pf->hw.port_info);
4155
4156 /* if media available, initialize PHY settings */
4157 if (pf->hw.port_info->phy.link_info.link_info &
4158 ICE_AQ_MEDIA_AVAILABLE) {
4159 err = ice_init_phy_user_cfg(pf->hw.port_info);
4160 if (err) {
4161 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4162 goto err_send_version_unroll;
4163 }
4164
4165 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4166 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4167
4168 if (vsi)
4169 ice_configure_phy(vsi);
4170 }
4171 } else {
4172 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4173 }
4174
4175 ice_verify_cacheline_size(pf);
4176
4177 /* Save wakeup reason register for later use */
4178 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4179
4180 /* check for a power management event */
4181 ice_print_wake_reason(pf);
4182
4183 /* clear wake status, all bits */
4184 wr32(hw, PFPM_WUS, U32_MAX);
4185
4186 /* Disable WoL at init, wait for user to enable */
4187 device_set_wakeup_enable(dev, false);
4188
4189 if (ice_is_safe_mode(pf)) {
4190 ice_set_safe_mode_vlan_cfg(pf);
4191 goto probe_done;
4192 }
4193
4194 /* initialize DDP driven features */
4195
4196 /* Note: Flow director init failure is non-fatal to load */
4197 if (ice_init_fdir(pf))
4198 dev_err(dev, "could not initialize flow director\n");
4199
4200 /* Note: DCB init failure is non-fatal to load */
4201 if (ice_init_pf_dcb(pf, false)) {
4202 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4203 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4204 } else {
4205 ice_cfg_lldp_mib_change(&pf->hw, true);
4206 }
4207
4208 /* print PCI link speed and width */
4209 pcie_print_link_status(pf->pdev);
4210
4211probe_done:
4212 /* ready to go, so clear down state bit */
4213 clear_bit(__ICE_DOWN, pf->state);
4214 return 0;
4215
4216err_send_version_unroll:
4217 ice_vsi_release_all(pf);
4218err_alloc_sw_unroll:
4219 ice_devlink_destroy_port(pf);
4220 set_bit(__ICE_SERVICE_DIS, pf->state);
4221 set_bit(__ICE_DOWN, pf->state);
4222 devm_kfree(dev, pf->first_sw);
4223err_msix_misc_unroll:
4224 ice_free_irq_msix_misc(pf);
4225err_init_interrupt_unroll:
4226 ice_clear_interrupt_scheme(pf);
4227err_init_vsi_unroll:
4228 devm_kfree(dev, pf->vsi);
4229err_init_pf_unroll:
4230 ice_deinit_pf(pf);
4231 ice_devlink_destroy_regions(pf);
4232 ice_deinit_hw(hw);
4233err_exit_unroll:
4234 ice_devlink_unregister(pf);
4235 pci_disable_pcie_error_reporting(pdev);
4236 pci_disable_device(pdev);
4237 return err;
4238}
4239
4240/**
4241 * ice_set_wake - enable or disable Wake on LAN
4242 * @pf: pointer to the PF struct
4243 *
4244 * Simple helper for WoL control
4245 */
4246static void ice_set_wake(struct ice_pf *pf)
4247{
4248 struct ice_hw *hw = &pf->hw;
4249 bool wol = pf->wol_ena;
4250
4251 /* clear wake state, otherwise new wake events won't fire */
4252 wr32(hw, PFPM_WUS, U32_MAX);
4253
4254 /* enable / disable APM wake up, no RMW needed */
4255 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4256
4257 /* set magic packet filter enabled */
4258 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4259}
4260
4261/**
4262 * ice_setup_magic_mc_wake - setup device to wake on multicast magic packet
4263 * @pf: pointer to the PF struct
4264 *
4265 * Issue firmware command to enable multicast magic wake, making
4266 * sure that any locally administered address (LAA) is used for
4267 * wake, and that PF reset doesn't undo the LAA.
4268 */
4269static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4270{
4271 struct device *dev = ice_pf_to_dev(pf);
4272 struct ice_hw *hw = &pf->hw;
4273 enum ice_status status;
4274 u8 mac_addr[ETH_ALEN];
4275 struct ice_vsi *vsi;
4276 u8 flags;
4277
4278 if (!pf->wol_ena)
4279 return;
4280
4281 vsi = ice_get_main_vsi(pf);
4282 if (!vsi)
4283 return;
4284
4285 /* Get current MAC address in case it's an LAA */
4286 if (vsi->netdev)
4287 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4288 else
4289 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4290
4291 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4292 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4293 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4294
4295 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4296 if (status)
4297 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4298 ice_stat_str(status),
4299 ice_aq_str(hw->adminq.sq_last_status));
4300}
4301
4302/**
4303 * ice_remove - Device removal routine
4304 * @pdev: PCI device information struct
4305 */
4306static void ice_remove(struct pci_dev *pdev)
4307{
4308 struct ice_pf *pf = pci_get_drvdata(pdev);
4309 int i;
4310
4311 if (!pf)
4312 return;
4313
4314 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4315 if (!ice_is_reset_in_progress(pf->state))
4316 break;
4317 msleep(100);
4318 }
4319
4320 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4321 set_bit(__ICE_VF_RESETS_DISABLED, pf->state);
4322 ice_free_vfs(pf);
4323 }
4324
4325 set_bit(__ICE_DOWN, pf->state);
4326 ice_service_task_stop(pf);
4327
4328 ice_aq_cancel_waiting_tasks(pf);
4329
4330 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4331 if (!ice_is_safe_mode(pf))
4332 ice_remove_arfs(pf);
4333 ice_setup_mc_magic_wake(pf);
4334 ice_devlink_destroy_port(pf);
4335 ice_vsi_release_all(pf);
4336 ice_set_wake(pf);
4337 ice_free_irq_msix_misc(pf);
4338 ice_for_each_vsi(pf, i) {
4339 if (!pf->vsi[i])
4340 continue;
4341 ice_vsi_free_q_vectors(pf->vsi[i]);
4342 }
4343 ice_deinit_pf(pf);
4344 ice_devlink_destroy_regions(pf);
4345 ice_deinit_hw(&pf->hw);
4346 ice_devlink_unregister(pf);
4347
4348 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4349 * do it via ice_schedule_reset() since there is no need to rebuild
4350 * and the service task is already stopped.
4351 */
4352 ice_reset(&pf->hw, ICE_RESET_PFR);
4353 pci_wait_for_pending_transaction(pdev);
4354 ice_clear_interrupt_scheme(pf);
4355 pci_disable_pcie_error_reporting(pdev);
4356 pci_disable_device(pdev);
4357}
4358
4359/**
4360 * ice_shutdown - PCI callback for shutting down device
4361 * @pdev: PCI device information struct
4362 */
4363static void ice_shutdown(struct pci_dev *pdev)
4364{
4365 struct ice_pf *pf = pci_get_drvdata(pdev);
4366
4367 ice_remove(pdev);
4368
4369 if (system_state == SYSTEM_POWER_OFF) {
4370 pci_wake_from_d3(pdev, pf->wol_ena);
4371 pci_set_power_state(pdev, PCI_D3hot);
4372 }
4373}
4374
4375#ifdef CONFIG_PM
4376/**
4377 * ice_prepare_for_shutdown - prep for PCI shutdown
4378 * @pf: board private structure
4379 *
4380 * Inform or close all dependent features in prep for PCI device shutdown
4381 */
4382static void ice_prepare_for_shutdown(struct ice_pf *pf)
4383{
4384 struct ice_hw *hw = &pf->hw;
4385 u32 v;
4386
4387 /* Notify VFs of impending reset */
4388 if (ice_check_sq_alive(hw, &hw->mailboxq))
4389 ice_vc_notify_reset(pf);
4390
4391 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4392
4393 /* disable the VSIs and their queues that are not already DOWN */
4394 ice_pf_dis_all_vsi(pf, false);
4395
4396 ice_for_each_vsi(pf, v)
4397 if (pf->vsi[v])
4398 pf->vsi[v]->vsi_num = 0;
4399
4400 ice_shutdown_all_ctrlq(hw);
4401}
4402
4403/**
4404 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4405 * @pf: board private structure to reinitialize
4406 *
4407 * This routine reinitialize interrupt scheme that was cleared during
4408 * power management suspend callback.
4409 *
4410 * This should be called during resume routine to re-allocate the q_vectors
4411 * and reacquire interrupts.
4412 */
4413static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4414{
4415 struct device *dev = ice_pf_to_dev(pf);
4416 int ret, v;
4417
4418 /* Since we clear MSIX flag during suspend, we need to
4419 * set it back during resume...
4420 */
4421
4422 ret = ice_init_interrupt_scheme(pf);
4423 if (ret) {
4424 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4425 return ret;
4426 }
4427
4428 /* Remap vectors and rings, after successful re-init interrupts */
4429 ice_for_each_vsi(pf, v) {
4430 if (!pf->vsi[v])
4431 continue;
4432
4433 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4434 if (ret)
4435 goto err_reinit;
4436 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4437 }
4438
4439 ret = ice_req_irq_msix_misc(pf);
4440 if (ret) {
4441 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4442 ret);
4443 goto err_reinit;
4444 }
4445
4446 return 0;
4447
4448err_reinit:
4449 while (v--)
4450 if (pf->vsi[v])
4451 ice_vsi_free_q_vectors(pf->vsi[v]);
4452
4453 return ret;
4454}
4455
4456/**
4457 * ice_suspend
4458 * @dev: generic device information structure
4459 *
4460 * Power Management callback to quiesce the device and prepare
4461 * for D3 transition.
4462 */
4463static int __maybe_unused ice_suspend(struct device *dev)
4464{
4465 struct pci_dev *pdev = to_pci_dev(dev);
4466 struct ice_pf *pf;
4467 int disabled, v;
4468
4469 pf = pci_get_drvdata(pdev);
4470
4471 if (!ice_pf_state_is_nominal(pf)) {
4472 dev_err(dev, "Device is not ready, no need to suspend it\n");
4473 return -EBUSY;
4474 }
4475
4476 /* Stop watchdog tasks until resume completion.
4477 * Even though it is most likely that the service task is
4478 * disabled if the device is suspended or down, the service task's
4479 * state is controlled by a different state bit, and we should
4480 * store and honor whatever state that bit is in at this point.
4481 */
4482 disabled = ice_service_task_stop(pf);
4483
4484 /* Already suspended?, then there is nothing to do */
4485 if (test_and_set_bit(__ICE_SUSPENDED, pf->state)) {
4486 if (!disabled)
4487 ice_service_task_restart(pf);
4488 return 0;
4489 }
4490
4491 if (test_bit(__ICE_DOWN, pf->state) ||
4492 ice_is_reset_in_progress(pf->state)) {
4493 dev_err(dev, "can't suspend device in reset or already down\n");
4494 if (!disabled)
4495 ice_service_task_restart(pf);
4496 return 0;
4497 }
4498
4499 ice_setup_mc_magic_wake(pf);
4500
4501 ice_prepare_for_shutdown(pf);
4502
4503 ice_set_wake(pf);
4504
4505 /* Free vectors, clear the interrupt scheme and release IRQs
4506 * for proper hibernation, especially with large number of CPUs.
4507 * Otherwise hibernation might fail when mapping all the vectors back
4508 * to CPU0.
4509 */
4510 ice_free_irq_msix_misc(pf);
4511 ice_for_each_vsi(pf, v) {
4512 if (!pf->vsi[v])
4513 continue;
4514 ice_vsi_free_q_vectors(pf->vsi[v]);
4515 }
4516 ice_clear_interrupt_scheme(pf);
4517
4518 pci_save_state(pdev);
4519 pci_wake_from_d3(pdev, pf->wol_ena);
4520 pci_set_power_state(pdev, PCI_D3hot);
4521 return 0;
4522}
4523
4524/**
4525 * ice_resume - PM callback for waking up from D3
4526 * @dev: generic device information structure
4527 */
4528static int __maybe_unused ice_resume(struct device *dev)
4529{
4530 struct pci_dev *pdev = to_pci_dev(dev);
4531 enum ice_reset_req reset_type;
4532 struct ice_pf *pf;
4533 struct ice_hw *hw;
4534 int ret;
4535
4536 pci_set_power_state(pdev, PCI_D0);
4537 pci_restore_state(pdev);
4538 pci_save_state(pdev);
4539
4540 if (!pci_device_is_present(pdev))
4541 return -ENODEV;
4542
4543 ret = pci_enable_device_mem(pdev);
4544 if (ret) {
4545 dev_err(dev, "Cannot enable device after suspend\n");
4546 return ret;
4547 }
4548
4549 pf = pci_get_drvdata(pdev);
4550 hw = &pf->hw;
4551
4552 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4553 ice_print_wake_reason(pf);
4554
4555 /* We cleared the interrupt scheme when we suspended, so we need to
4556 * restore it now to resume device functionality.
4557 */
4558 ret = ice_reinit_interrupt_scheme(pf);
4559 if (ret)
4560 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4561
4562 clear_bit(__ICE_DOWN, pf->state);
4563 /* Now perform PF reset and rebuild */
4564 reset_type = ICE_RESET_PFR;
4565 /* re-enable service task for reset, but allow reset to schedule it */
4566 clear_bit(__ICE_SERVICE_DIS, pf->state);
4567
4568 if (ice_schedule_reset(pf, reset_type))
4569 dev_err(dev, "Reset during resume failed.\n");
4570
4571 clear_bit(__ICE_SUSPENDED, pf->state);
4572 ice_service_task_restart(pf);
4573
4574 /* Restart the service task */
4575 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4576
4577 return 0;
4578}
4579#endif /* CONFIG_PM */
4580
4581/**
4582 * ice_pci_err_detected - warning that PCI error has been detected
4583 * @pdev: PCI device information struct
4584 * @err: the type of PCI error
4585 *
4586 * Called to warn that something happened on the PCI bus and the error handling
4587 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4588 */
4589static pci_ers_result_t
4590ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4591{
4592 struct ice_pf *pf = pci_get_drvdata(pdev);
4593
4594 if (!pf) {
4595 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4596 __func__, err);
4597 return PCI_ERS_RESULT_DISCONNECT;
4598 }
4599
4600 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4601 ice_service_task_stop(pf);
4602
4603 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4604 set_bit(__ICE_PFR_REQ, pf->state);
4605 ice_prepare_for_reset(pf);
4606 }
4607 }
4608
4609 return PCI_ERS_RESULT_NEED_RESET;
4610}
4611
4612/**
4613 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4614 * @pdev: PCI device information struct
4615 *
4616 * Called to determine if the driver can recover from the PCI slot reset by
4617 * using a register read to determine if the device is recoverable.
4618 */
4619static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4620{
4621 struct ice_pf *pf = pci_get_drvdata(pdev);
4622 pci_ers_result_t result;
4623 int err;
4624 u32 reg;
4625
4626 err = pci_enable_device_mem(pdev);
4627 if (err) {
4628 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4629 err);
4630 result = PCI_ERS_RESULT_DISCONNECT;
4631 } else {
4632 pci_set_master(pdev);
4633 pci_restore_state(pdev);
4634 pci_save_state(pdev);
4635 pci_wake_from_d3(pdev, false);
4636
4637 /* Check for life */
4638 reg = rd32(&pf->hw, GLGEN_RTRIG);
4639 if (!reg)
4640 result = PCI_ERS_RESULT_RECOVERED;
4641 else
4642 result = PCI_ERS_RESULT_DISCONNECT;
4643 }
4644
4645 err = pci_aer_clear_nonfatal_status(pdev);
4646 if (err)
4647 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4648 err);
4649 /* non-fatal, continue */
4650
4651 return result;
4652}
4653
4654/**
4655 * ice_pci_err_resume - restart operations after PCI error recovery
4656 * @pdev: PCI device information struct
4657 *
4658 * Called to allow the driver to bring things back up after PCI error and/or
4659 * reset recovery have finished
4660 */
4661static void ice_pci_err_resume(struct pci_dev *pdev)
4662{
4663 struct ice_pf *pf = pci_get_drvdata(pdev);
4664
4665 if (!pf) {
4666 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4667 __func__);
4668 return;
4669 }
4670
4671 if (test_bit(__ICE_SUSPENDED, pf->state)) {
4672 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4673 __func__);
4674 return;
4675 }
4676
4677 ice_restore_all_vfs_msi_state(pdev);
4678
4679 ice_do_reset(pf, ICE_RESET_PFR);
4680 ice_service_task_restart(pf);
4681 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4682}
4683
4684/**
4685 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4686 * @pdev: PCI device information struct
4687 */
4688static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4689{
4690 struct ice_pf *pf = pci_get_drvdata(pdev);
4691
4692 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4693 ice_service_task_stop(pf);
4694
4695 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4696 set_bit(__ICE_PFR_REQ, pf->state);
4697 ice_prepare_for_reset(pf);
4698 }
4699 }
4700}
4701
4702/**
4703 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4704 * @pdev: PCI device information struct
4705 */
4706static void ice_pci_err_reset_done(struct pci_dev *pdev)
4707{
4708 ice_pci_err_resume(pdev);
4709}
4710
4711/* ice_pci_tbl - PCI Device ID Table
4712 *
4713 * Wildcard entries (PCI_ANY_ID) should come last
4714 * Last entry must be all 0s
4715 *
4716 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4717 * Class, Class Mask, private data (not used) }
4718 */
4719static const struct pci_device_id ice_pci_tbl[] = {
4720 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4721 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4722 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4723 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4724 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4725 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4726 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4727 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4728 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4729 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4730 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4731 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4732 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4733 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4734 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4735 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4736 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4737 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4738 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4739 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4740 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4741 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4742 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4743 /* required last entry */
4744 { 0, }
4745};
4746MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4747
4748static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4749
4750static const struct pci_error_handlers ice_pci_err_handler = {
4751 .error_detected = ice_pci_err_detected,
4752 .slot_reset = ice_pci_err_slot_reset,
4753 .reset_prepare = ice_pci_err_reset_prepare,
4754 .reset_done = ice_pci_err_reset_done,
4755 .resume = ice_pci_err_resume
4756};
4757
4758static struct pci_driver ice_driver = {
4759 .name = KBUILD_MODNAME,
4760 .id_table = ice_pci_tbl,
4761 .probe = ice_probe,
4762 .remove = ice_remove,
4763#ifdef CONFIG_PM
4764 .driver.pm = &ice_pm_ops,
4765#endif /* CONFIG_PM */
4766 .shutdown = ice_shutdown,
4767 .sriov_configure = ice_sriov_configure,
4768 .err_handler = &ice_pci_err_handler
4769};
4770
4771/**
4772 * ice_module_init - Driver registration routine
4773 *
4774 * ice_module_init is the first routine called when the driver is
4775 * loaded. All it does is register with the PCI subsystem.
4776 */
4777static int __init ice_module_init(void)
4778{
4779 int status;
4780
4781 pr_info("%s\n", ice_driver_string);
4782 pr_info("%s\n", ice_copyright);
4783
4784 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4785 if (!ice_wq) {
4786 pr_err("Failed to create workqueue\n");
4787 return -ENOMEM;
4788 }
4789
4790 status = pci_register_driver(&ice_driver);
4791 if (status) {
4792 pr_err("failed to register PCI driver, err %d\n", status);
4793 destroy_workqueue(ice_wq);
4794 }
4795
4796 return status;
4797}
4798module_init(ice_module_init);
4799
4800/**
4801 * ice_module_exit - Driver exit cleanup routine
4802 *
4803 * ice_module_exit is called just before the driver is removed
4804 * from memory.
4805 */
4806static void __exit ice_module_exit(void)
4807{
4808 pci_unregister_driver(&ice_driver);
4809 destroy_workqueue(ice_wq);
4810 pr_info("module unloaded\n");
4811}
4812module_exit(ice_module_exit);
4813
4814/**
4815 * ice_set_mac_address - NDO callback to set MAC address
4816 * @netdev: network interface device structure
4817 * @pi: pointer to an address structure
4818 *
4819 * Returns 0 on success, negative on failure
4820 */
4821static int ice_set_mac_address(struct net_device *netdev, void *pi)
4822{
4823 struct ice_netdev_priv *np = netdev_priv(netdev);
4824 struct ice_vsi *vsi = np->vsi;
4825 struct ice_pf *pf = vsi->back;
4826 struct ice_hw *hw = &pf->hw;
4827 struct sockaddr *addr = pi;
4828 enum ice_status status;
4829 u8 flags = 0;
4830 int err = 0;
4831 u8 *mac;
4832
4833 mac = (u8 *)addr->sa_data;
4834
4835 if (!is_valid_ether_addr(mac))
4836 return -EADDRNOTAVAIL;
4837
4838 if (ether_addr_equal(netdev->dev_addr, mac)) {
4839 netdev_warn(netdev, "already using mac %pM\n", mac);
4840 return 0;
4841 }
4842
4843 if (test_bit(__ICE_DOWN, pf->state) ||
4844 ice_is_reset_in_progress(pf->state)) {
4845 netdev_err(netdev, "can't set mac %pM. device not ready\n",
4846 mac);
4847 return -EBUSY;
4848 }
4849
4850 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
4851 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
4852 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
4853 err = -EADDRNOTAVAIL;
4854 goto err_update_filters;
4855 }
4856
4857 /* Add filter for new MAC. If filter exists, just return success */
4858 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
4859 if (status == ICE_ERR_ALREADY_EXISTS) {
4860 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
4861 return 0;
4862 }
4863
4864 /* error if the new filter addition failed */
4865 if (status)
4866 err = -EADDRNOTAVAIL;
4867
4868err_update_filters:
4869 if (err) {
4870 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
4871 mac);
4872 return err;
4873 }
4874
4875 /* change the netdev's MAC address */
4876 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4877 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
4878 netdev->dev_addr);
4879
4880 /* write new MAC address to the firmware */
4881 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
4882 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
4883 if (status) {
4884 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
4885 mac, ice_stat_str(status));
4886 }
4887 return 0;
4888}
4889
4890/**
4891 * ice_set_rx_mode - NDO callback to set the netdev filters
4892 * @netdev: network interface device structure
4893 */
4894static void ice_set_rx_mode(struct net_device *netdev)
4895{
4896 struct ice_netdev_priv *np = netdev_priv(netdev);
4897 struct ice_vsi *vsi = np->vsi;
4898
4899 if (!vsi)
4900 return;
4901
4902 /* Set the flags to synchronize filters
4903 * ndo_set_rx_mode may be triggered even without a change in netdev
4904 * flags
4905 */
4906 set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
4907 set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
4908 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
4909
4910 /* schedule our worker thread which will take care of
4911 * applying the new filter changes
4912 */
4913 ice_service_task_schedule(vsi->back);
4914}
4915
4916/**
4917 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
4918 * @netdev: network interface device structure
4919 * @queue_index: Queue ID
4920 * @maxrate: maximum bandwidth in Mbps
4921 */
4922static int
4923ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
4924{
4925 struct ice_netdev_priv *np = netdev_priv(netdev);
4926 struct ice_vsi *vsi = np->vsi;
4927 enum ice_status status;
4928 u16 q_handle;
4929 u8 tc;
4930
4931 /* Validate maxrate requested is within permitted range */
4932 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
4933 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
4934 maxrate, queue_index);
4935 return -EINVAL;
4936 }
4937
4938 q_handle = vsi->tx_rings[queue_index]->q_handle;
4939 tc = ice_dcb_get_tc(vsi, queue_index);
4940
4941 /* Set BW back to default, when user set maxrate to 0 */
4942 if (!maxrate)
4943 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
4944 q_handle, ICE_MAX_BW);
4945 else
4946 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
4947 q_handle, ICE_MAX_BW, maxrate * 1000);
4948 if (status) {
4949 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
4950 ice_stat_str(status));
4951 return -EIO;
4952 }
4953
4954 return 0;
4955}
4956
4957/**
4958 * ice_fdb_add - add an entry to the hardware database
4959 * @ndm: the input from the stack
4960 * @tb: pointer to array of nladdr (unused)
4961 * @dev: the net device pointer
4962 * @addr: the MAC address entry being added
4963 * @vid: VLAN ID
4964 * @flags: instructions from stack about fdb operation
4965 * @extack: netlink extended ack
4966 */
4967static int
4968ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
4969 struct net_device *dev, const unsigned char *addr, u16 vid,
4970 u16 flags, struct netlink_ext_ack __always_unused *extack)
4971{
4972 int err;
4973
4974 if (vid) {
4975 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
4976 return -EINVAL;
4977 }
4978 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
4979 netdev_err(dev, "FDB only supports static addresses\n");
4980 return -EINVAL;
4981 }
4982
4983 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
4984 err = dev_uc_add_excl(dev, addr);
4985 else if (is_multicast_ether_addr(addr))
4986 err = dev_mc_add_excl(dev, addr);
4987 else
4988 err = -EINVAL;
4989
4990 /* Only return duplicate errors if NLM_F_EXCL is set */
4991 if (err == -EEXIST && !(flags & NLM_F_EXCL))
4992 err = 0;
4993
4994 return err;
4995}
4996
4997/**
4998 * ice_fdb_del - delete an entry from the hardware database
4999 * @ndm: the input from the stack
5000 * @tb: pointer to array of nladdr (unused)
5001 * @dev: the net device pointer
5002 * @addr: the MAC address entry being added
5003 * @vid: VLAN ID
5004 */
5005static int
5006ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5007 struct net_device *dev, const unsigned char *addr,
5008 __always_unused u16 vid)
5009{
5010 int err;
5011
5012 if (ndm->ndm_state & NUD_PERMANENT) {
5013 netdev_err(dev, "FDB only supports static addresses\n");
5014 return -EINVAL;
5015 }
5016
5017 if (is_unicast_ether_addr(addr))
5018 err = dev_uc_del(dev, addr);
5019 else if (is_multicast_ether_addr(addr))
5020 err = dev_mc_del(dev, addr);
5021 else
5022 err = -EINVAL;
5023
5024 return err;
5025}
5026
5027/**
5028 * ice_set_features - set the netdev feature flags
5029 * @netdev: ptr to the netdev being adjusted
5030 * @features: the feature set that the stack is suggesting
5031 */
5032static int
5033ice_set_features(struct net_device *netdev, netdev_features_t features)
5034{
5035 struct ice_netdev_priv *np = netdev_priv(netdev);
5036 struct ice_vsi *vsi = np->vsi;
5037 struct ice_pf *pf = vsi->back;
5038 int ret = 0;
5039
5040 /* Don't set any netdev advanced features with device in Safe Mode */
5041 if (ice_is_safe_mode(vsi->back)) {
5042 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5043 return ret;
5044 }
5045
5046 /* Do not change setting during reset */
5047 if (ice_is_reset_in_progress(pf->state)) {
5048 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5049 return -EBUSY;
5050 }
5051
5052 /* Multiple features can be changed in one call so keep features in
5053 * separate if/else statements to guarantee each feature is checked
5054 */
5055 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5056 ret = ice_vsi_manage_rss_lut(vsi, true);
5057 else if (!(features & NETIF_F_RXHASH) &&
5058 netdev->features & NETIF_F_RXHASH)
5059 ret = ice_vsi_manage_rss_lut(vsi, false);
5060
5061 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5062 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5063 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5064 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5065 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5066 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5067
5068 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5069 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5070 ret = ice_vsi_manage_vlan_insertion(vsi);
5071 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5072 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5073 ret = ice_vsi_manage_vlan_insertion(vsi);
5074
5075 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5076 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5077 ret = ice_cfg_vlan_pruning(vsi, true, false);
5078 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5079 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5080 ret = ice_cfg_vlan_pruning(vsi, false, false);
5081
5082 if ((features & NETIF_F_NTUPLE) &&
5083 !(netdev->features & NETIF_F_NTUPLE)) {
5084 ice_vsi_manage_fdir(vsi, true);
5085 ice_init_arfs(vsi);
5086 } else if (!(features & NETIF_F_NTUPLE) &&
5087 (netdev->features & NETIF_F_NTUPLE)) {
5088 ice_vsi_manage_fdir(vsi, false);
5089 ice_clear_arfs(vsi);
5090 }
5091
5092 return ret;
5093}
5094
5095/**
5096 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5097 * @vsi: VSI to setup VLAN properties for
5098 */
5099static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5100{
5101 int ret = 0;
5102
5103 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5104 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5105 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5106 ret = ice_vsi_manage_vlan_insertion(vsi);
5107
5108 return ret;
5109}
5110
5111/**
5112 * ice_vsi_cfg - Setup the VSI
5113 * @vsi: the VSI being configured
5114 *
5115 * Return 0 on success and negative value on error
5116 */
5117int ice_vsi_cfg(struct ice_vsi *vsi)
5118{
5119 int err;
5120
5121 if (vsi->netdev) {
5122 ice_set_rx_mode(vsi->netdev);
5123
5124 err = ice_vsi_vlan_setup(vsi);
5125
5126 if (err)
5127 return err;
5128 }
5129 ice_vsi_cfg_dcb_rings(vsi);
5130
5131 err = ice_vsi_cfg_lan_txqs(vsi);
5132 if (!err && ice_is_xdp_ena_vsi(vsi))
5133 err = ice_vsi_cfg_xdp_txqs(vsi);
5134 if (!err)
5135 err = ice_vsi_cfg_rxqs(vsi);
5136
5137 return err;
5138}
5139
5140/**
5141 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5142 * @vsi: the VSI being configured
5143 */
5144static void ice_napi_enable_all(struct ice_vsi *vsi)
5145{
5146 int q_idx;
5147
5148 if (!vsi->netdev)
5149 return;
5150
5151 ice_for_each_q_vector(vsi, q_idx) {
5152 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5153
5154 if (q_vector->rx.ring || q_vector->tx.ring)
5155 napi_enable(&q_vector->napi);
5156 }
5157}
5158
5159/**
5160 * ice_up_complete - Finish the last steps of bringing up a connection
5161 * @vsi: The VSI being configured
5162 *
5163 * Return 0 on success and negative value on error
5164 */
5165static int ice_up_complete(struct ice_vsi *vsi)
5166{
5167 struct ice_pf *pf = vsi->back;
5168 int err;
5169
5170 ice_vsi_cfg_msix(vsi);
5171
5172 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5173 * Tx queue group list was configured and the context bits were
5174 * programmed using ice_vsi_cfg_txqs
5175 */
5176 err = ice_vsi_start_all_rx_rings(vsi);
5177 if (err)
5178 return err;
5179
5180 clear_bit(__ICE_DOWN, vsi->state);
5181 ice_napi_enable_all(vsi);
5182 ice_vsi_ena_irq(vsi);
5183
5184 if (vsi->port_info &&
5185 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5186 vsi->netdev) {
5187 ice_print_link_msg(vsi, true);
5188 netif_tx_start_all_queues(vsi->netdev);
5189 netif_carrier_on(vsi->netdev);
5190 }
5191
5192 ice_service_task_schedule(pf);
5193
5194 return 0;
5195}
5196
5197/**
5198 * ice_up - Bring the connection back up after being down
5199 * @vsi: VSI being configured
5200 */
5201int ice_up(struct ice_vsi *vsi)
5202{
5203 int err;
5204
5205 err = ice_vsi_cfg(vsi);
5206 if (!err)
5207 err = ice_up_complete(vsi);
5208
5209 return err;
5210}
5211
5212/**
5213 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5214 * @ring: Tx or Rx ring to read stats from
5215 * @pkts: packets stats counter
5216 * @bytes: bytes stats counter
5217 *
5218 * This function fetches stats from the ring considering the atomic operations
5219 * that needs to be performed to read u64 values in 32 bit machine.
5220 */
5221static void
5222ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5223{
5224 unsigned int start;
5225 *pkts = 0;
5226 *bytes = 0;
5227
5228 if (!ring)
5229 return;
5230 do {
5231 start = u64_stats_fetch_begin_irq(&ring->syncp);
5232 *pkts = ring->stats.pkts;
5233 *bytes = ring->stats.bytes;
5234 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5235}
5236
5237/**
5238 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5239 * @vsi: the VSI to be updated
5240 * @rings: rings to work on
5241 * @count: number of rings
5242 */
5243static void
5244ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5245 u16 count)
5246{
5247 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5248 u16 i;
5249
5250 for (i = 0; i < count; i++) {
5251 struct ice_ring *ring;
5252 u64 pkts, bytes;
5253
5254 ring = READ_ONCE(rings[i]);
5255 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5256 vsi_stats->tx_packets += pkts;
5257 vsi_stats->tx_bytes += bytes;
5258 vsi->tx_restart += ring->tx_stats.restart_q;
5259 vsi->tx_busy += ring->tx_stats.tx_busy;
5260 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5261 }
5262}
5263
5264/**
5265 * ice_update_vsi_ring_stats - Update VSI stats counters
5266 * @vsi: the VSI to be updated
5267 */
5268static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5269{
5270 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5271 struct ice_ring *ring;
5272 u64 pkts, bytes;
5273 int i;
5274
5275 /* reset netdev stats */
5276 vsi_stats->tx_packets = 0;
5277 vsi_stats->tx_bytes = 0;
5278 vsi_stats->rx_packets = 0;
5279 vsi_stats->rx_bytes = 0;
5280
5281 /* reset non-netdev (extended) stats */
5282 vsi->tx_restart = 0;
5283 vsi->tx_busy = 0;
5284 vsi->tx_linearize = 0;
5285 vsi->rx_buf_failed = 0;
5286 vsi->rx_page_failed = 0;
5287 vsi->rx_gro_dropped = 0;
5288
5289 rcu_read_lock();
5290
5291 /* update Tx rings counters */
5292 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5293
5294 /* update Rx rings counters */
5295 ice_for_each_rxq(vsi, i) {
5296 ring = READ_ONCE(vsi->rx_rings[i]);
5297 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5298 vsi_stats->rx_packets += pkts;
5299 vsi_stats->rx_bytes += bytes;
5300 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5301 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5302 vsi->rx_gro_dropped += ring->rx_stats.gro_dropped;
5303 }
5304
5305 /* update XDP Tx rings counters */
5306 if (ice_is_xdp_ena_vsi(vsi))
5307 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5308 vsi->num_xdp_txq);
5309
5310 rcu_read_unlock();
5311}
5312
5313/**
5314 * ice_update_vsi_stats - Update VSI stats counters
5315 * @vsi: the VSI to be updated
5316 */
5317void ice_update_vsi_stats(struct ice_vsi *vsi)
5318{
5319 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5320 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5321 struct ice_pf *pf = vsi->back;
5322
5323 if (test_bit(__ICE_DOWN, vsi->state) ||
5324 test_bit(__ICE_CFG_BUSY, pf->state))
5325 return;
5326
5327 /* get stats as recorded by Tx/Rx rings */
5328 ice_update_vsi_ring_stats(vsi);
5329
5330 /* get VSI stats as recorded by the hardware */
5331 ice_update_eth_stats(vsi);
5332
5333 cur_ns->tx_errors = cur_es->tx_errors;
5334 cur_ns->rx_dropped = cur_es->rx_discards + vsi->rx_gro_dropped;
5335 cur_ns->tx_dropped = cur_es->tx_discards;
5336 cur_ns->multicast = cur_es->rx_multicast;
5337
5338 /* update some more netdev stats if this is main VSI */
5339 if (vsi->type == ICE_VSI_PF) {
5340 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5341 cur_ns->rx_errors = pf->stats.crc_errors +
5342 pf->stats.illegal_bytes +
5343 pf->stats.rx_len_errors +
5344 pf->stats.rx_undersize +
5345 pf->hw_csum_rx_error +
5346 pf->stats.rx_jabber +
5347 pf->stats.rx_fragments +
5348 pf->stats.rx_oversize;
5349 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5350 /* record drops from the port level */
5351 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5352 }
5353}
5354
5355/**
5356 * ice_update_pf_stats - Update PF port stats counters
5357 * @pf: PF whose stats needs to be updated
5358 */
5359void ice_update_pf_stats(struct ice_pf *pf)
5360{
5361 struct ice_hw_port_stats *prev_ps, *cur_ps;
5362 struct ice_hw *hw = &pf->hw;
5363 u16 fd_ctr_base;
5364 u8 port;
5365
5366 port = hw->port_info->lport;
5367 prev_ps = &pf->stats_prev;
5368 cur_ps = &pf->stats;
5369
5370 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5371 &prev_ps->eth.rx_bytes,
5372 &cur_ps->eth.rx_bytes);
5373
5374 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5375 &prev_ps->eth.rx_unicast,
5376 &cur_ps->eth.rx_unicast);
5377
5378 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5379 &prev_ps->eth.rx_multicast,
5380 &cur_ps->eth.rx_multicast);
5381
5382 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5383 &prev_ps->eth.rx_broadcast,
5384 &cur_ps->eth.rx_broadcast);
5385
5386 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5387 &prev_ps->eth.rx_discards,
5388 &cur_ps->eth.rx_discards);
5389
5390 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5391 &prev_ps->eth.tx_bytes,
5392 &cur_ps->eth.tx_bytes);
5393
5394 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5395 &prev_ps->eth.tx_unicast,
5396 &cur_ps->eth.tx_unicast);
5397
5398 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5399 &prev_ps->eth.tx_multicast,
5400 &cur_ps->eth.tx_multicast);
5401
5402 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5403 &prev_ps->eth.tx_broadcast,
5404 &cur_ps->eth.tx_broadcast);
5405
5406 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5407 &prev_ps->tx_dropped_link_down,
5408 &cur_ps->tx_dropped_link_down);
5409
5410 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5411 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5412
5413 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5414 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5415
5416 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5417 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5418
5419 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5420 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5421
5422 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5423 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5424
5425 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5426 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5427
5428 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5429 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5430
5431 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5432 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5433
5434 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5435 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5436
5437 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5438 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5439
5440 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5441 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5442
5443 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5444 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5445
5446 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5447 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5448
5449 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5450 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5451
5452 fd_ctr_base = hw->fd_ctr_base;
5453
5454 ice_stat_update40(hw,
5455 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5456 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5457 &cur_ps->fd_sb_match);
5458 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5459 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5460
5461 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5462 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5463
5464 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5465 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5466
5467 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5468 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5469
5470 ice_update_dcb_stats(pf);
5471
5472 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5473 &prev_ps->crc_errors, &cur_ps->crc_errors);
5474
5475 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5476 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5477
5478 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5479 &prev_ps->mac_local_faults,
5480 &cur_ps->mac_local_faults);
5481
5482 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5483 &prev_ps->mac_remote_faults,
5484 &cur_ps->mac_remote_faults);
5485
5486 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5487 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5488
5489 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5490 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5491
5492 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5493 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5494
5495 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5496 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5497
5498 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5499 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5500
5501 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5502
5503 pf->stat_prev_loaded = true;
5504}
5505
5506/**
5507 * ice_get_stats64 - get statistics for network device structure
5508 * @netdev: network interface device structure
5509 * @stats: main device statistics structure
5510 */
5511static
5512void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5513{
5514 struct ice_netdev_priv *np = netdev_priv(netdev);
5515 struct rtnl_link_stats64 *vsi_stats;
5516 struct ice_vsi *vsi = np->vsi;
5517
5518 vsi_stats = &vsi->net_stats;
5519
5520 if (!vsi->num_txq || !vsi->num_rxq)
5521 return;
5522
5523 /* netdev packet/byte stats come from ring counter. These are obtained
5524 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5525 * But, only call the update routine and read the registers if VSI is
5526 * not down.
5527 */
5528 if (!test_bit(__ICE_DOWN, vsi->state))
5529 ice_update_vsi_ring_stats(vsi);
5530 stats->tx_packets = vsi_stats->tx_packets;
5531 stats->tx_bytes = vsi_stats->tx_bytes;
5532 stats->rx_packets = vsi_stats->rx_packets;
5533 stats->rx_bytes = vsi_stats->rx_bytes;
5534
5535 /* The rest of the stats can be read from the hardware but instead we
5536 * just return values that the watchdog task has already obtained from
5537 * the hardware.
5538 */
5539 stats->multicast = vsi_stats->multicast;
5540 stats->tx_errors = vsi_stats->tx_errors;
5541 stats->tx_dropped = vsi_stats->tx_dropped;
5542 stats->rx_errors = vsi_stats->rx_errors;
5543 stats->rx_dropped = vsi_stats->rx_dropped;
5544 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5545 stats->rx_length_errors = vsi_stats->rx_length_errors;
5546}
5547
5548/**
5549 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5550 * @vsi: VSI having NAPI disabled
5551 */
5552static void ice_napi_disable_all(struct ice_vsi *vsi)
5553{
5554 int q_idx;
5555
5556 if (!vsi->netdev)
5557 return;
5558
5559 ice_for_each_q_vector(vsi, q_idx) {
5560 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5561
5562 if (q_vector->rx.ring || q_vector->tx.ring)
5563 napi_disable(&q_vector->napi);
5564 }
5565}
5566
5567/**
5568 * ice_down - Shutdown the connection
5569 * @vsi: The VSI being stopped
5570 */
5571int ice_down(struct ice_vsi *vsi)
5572{
5573 int i, tx_err, rx_err, link_err = 0;
5574
5575 /* Caller of this function is expected to set the
5576 * vsi->state __ICE_DOWN bit
5577 */
5578 if (vsi->netdev) {
5579 netif_carrier_off(vsi->netdev);
5580 netif_tx_disable(vsi->netdev);
5581 }
5582
5583 ice_vsi_dis_irq(vsi);
5584
5585 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5586 if (tx_err)
5587 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5588 vsi->vsi_num, tx_err);
5589 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5590 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5591 if (tx_err)
5592 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5593 vsi->vsi_num, tx_err);
5594 }
5595
5596 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5597 if (rx_err)
5598 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5599 vsi->vsi_num, rx_err);
5600
5601 ice_napi_disable_all(vsi);
5602
5603 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5604 link_err = ice_force_phys_link_state(vsi, false);
5605 if (link_err)
5606 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5607 vsi->vsi_num, link_err);
5608 }
5609
5610 ice_for_each_txq(vsi, i)
5611 ice_clean_tx_ring(vsi->tx_rings[i]);
5612
5613 ice_for_each_rxq(vsi, i)
5614 ice_clean_rx_ring(vsi->rx_rings[i]);
5615
5616 if (tx_err || rx_err || link_err) {
5617 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5618 vsi->vsi_num, vsi->vsw->sw_id);
5619 return -EIO;
5620 }
5621
5622 return 0;
5623}
5624
5625/**
5626 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5627 * @vsi: VSI having resources allocated
5628 *
5629 * Return 0 on success, negative on failure
5630 */
5631int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5632{
5633 int i, err = 0;
5634
5635 if (!vsi->num_txq) {
5636 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5637 vsi->vsi_num);
5638 return -EINVAL;
5639 }
5640
5641 ice_for_each_txq(vsi, i) {
5642 struct ice_ring *ring = vsi->tx_rings[i];
5643
5644 if (!ring)
5645 return -EINVAL;
5646
5647 ring->netdev = vsi->netdev;
5648 err = ice_setup_tx_ring(ring);
5649 if (err)
5650 break;
5651 }
5652
5653 return err;
5654}
5655
5656/**
5657 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5658 * @vsi: VSI having resources allocated
5659 *
5660 * Return 0 on success, negative on failure
5661 */
5662int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5663{
5664 int i, err = 0;
5665
5666 if (!vsi->num_rxq) {
5667 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5668 vsi->vsi_num);
5669 return -EINVAL;
5670 }
5671
5672 ice_for_each_rxq(vsi, i) {
5673 struct ice_ring *ring = vsi->rx_rings[i];
5674
5675 if (!ring)
5676 return -EINVAL;
5677
5678 ring->netdev = vsi->netdev;
5679 err = ice_setup_rx_ring(ring);
5680 if (err)
5681 break;
5682 }
5683
5684 return err;
5685}
5686
5687/**
5688 * ice_vsi_open_ctrl - open control VSI for use
5689 * @vsi: the VSI to open
5690 *
5691 * Initialization of the Control VSI
5692 *
5693 * Returns 0 on success, negative value on error
5694 */
5695int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5696{
5697 char int_name[ICE_INT_NAME_STR_LEN];
5698 struct ice_pf *pf = vsi->back;
5699 struct device *dev;
5700 int err;
5701
5702 dev = ice_pf_to_dev(pf);
5703 /* allocate descriptors */
5704 err = ice_vsi_setup_tx_rings(vsi);
5705 if (err)
5706 goto err_setup_tx;
5707
5708 err = ice_vsi_setup_rx_rings(vsi);
5709 if (err)
5710 goto err_setup_rx;
5711
5712 err = ice_vsi_cfg(vsi);
5713 if (err)
5714 goto err_setup_rx;
5715
5716 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5717 dev_driver_string(dev), dev_name(dev));
5718 err = ice_vsi_req_irq_msix(vsi, int_name);
5719 if (err)
5720 goto err_setup_rx;
5721
5722 ice_vsi_cfg_msix(vsi);
5723
5724 err = ice_vsi_start_all_rx_rings(vsi);
5725 if (err)
5726 goto err_up_complete;
5727
5728 clear_bit(__ICE_DOWN, vsi->state);
5729 ice_vsi_ena_irq(vsi);
5730
5731 return 0;
5732
5733err_up_complete:
5734 ice_down(vsi);
5735err_setup_rx:
5736 ice_vsi_free_rx_rings(vsi);
5737err_setup_tx:
5738 ice_vsi_free_tx_rings(vsi);
5739
5740 return err;
5741}
5742
5743/**
5744 * ice_vsi_open - Called when a network interface is made active
5745 * @vsi: the VSI to open
5746 *
5747 * Initialization of the VSI
5748 *
5749 * Returns 0 on success, negative value on error
5750 */
5751static int ice_vsi_open(struct ice_vsi *vsi)
5752{
5753 char int_name[ICE_INT_NAME_STR_LEN];
5754 struct ice_pf *pf = vsi->back;
5755 int err;
5756
5757 /* allocate descriptors */
5758 err = ice_vsi_setup_tx_rings(vsi);
5759 if (err)
5760 goto err_setup_tx;
5761
5762 err = ice_vsi_setup_rx_rings(vsi);
5763 if (err)
5764 goto err_setup_rx;
5765
5766 err = ice_vsi_cfg(vsi);
5767 if (err)
5768 goto err_setup_rx;
5769
5770 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5771 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
5772 err = ice_vsi_req_irq_msix(vsi, int_name);
5773 if (err)
5774 goto err_setup_rx;
5775
5776 /* Notify the stack of the actual queue counts. */
5777 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
5778 if (err)
5779 goto err_set_qs;
5780
5781 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
5782 if (err)
5783 goto err_set_qs;
5784
5785 err = ice_up_complete(vsi);
5786 if (err)
5787 goto err_up_complete;
5788
5789 return 0;
5790
5791err_up_complete:
5792 ice_down(vsi);
5793err_set_qs:
5794 ice_vsi_free_irq(vsi);
5795err_setup_rx:
5796 ice_vsi_free_rx_rings(vsi);
5797err_setup_tx:
5798 ice_vsi_free_tx_rings(vsi);
5799
5800 return err;
5801}
5802
5803/**
5804 * ice_vsi_release_all - Delete all VSIs
5805 * @pf: PF from which all VSIs are being removed
5806 */
5807static void ice_vsi_release_all(struct ice_pf *pf)
5808{
5809 int err, i;
5810
5811 if (!pf->vsi)
5812 return;
5813
5814 ice_for_each_vsi(pf, i) {
5815 if (!pf->vsi[i])
5816 continue;
5817
5818 err = ice_vsi_release(pf->vsi[i]);
5819 if (err)
5820 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
5821 i, err, pf->vsi[i]->vsi_num);
5822 }
5823}
5824
5825/**
5826 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
5827 * @pf: pointer to the PF instance
5828 * @type: VSI type to rebuild
5829 *
5830 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
5831 */
5832static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
5833{
5834 struct device *dev = ice_pf_to_dev(pf);
5835 enum ice_status status;
5836 int i, err;
5837
5838 ice_for_each_vsi(pf, i) {
5839 struct ice_vsi *vsi = pf->vsi[i];
5840
5841 if (!vsi || vsi->type != type)
5842 continue;
5843
5844 /* rebuild the VSI */
5845 err = ice_vsi_rebuild(vsi, true);
5846 if (err) {
5847 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
5848 err, vsi->idx, ice_vsi_type_str(type));
5849 return err;
5850 }
5851
5852 /* replay filters for the VSI */
5853 status = ice_replay_vsi(&pf->hw, vsi->idx);
5854 if (status) {
5855 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
5856 ice_stat_str(status), vsi->idx,
5857 ice_vsi_type_str(type));
5858 return -EIO;
5859 }
5860
5861 /* Re-map HW VSI number, using VSI handle that has been
5862 * previously validated in ice_replay_vsi() call above
5863 */
5864 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
5865
5866 /* enable the VSI */
5867 err = ice_ena_vsi(vsi, false);
5868 if (err) {
5869 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
5870 err, vsi->idx, ice_vsi_type_str(type));
5871 return err;
5872 }
5873
5874 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
5875 ice_vsi_type_str(type));
5876 }
5877
5878 return 0;
5879}
5880
5881/**
5882 * ice_update_pf_netdev_link - Update PF netdev link status
5883 * @pf: pointer to the PF instance
5884 */
5885static void ice_update_pf_netdev_link(struct ice_pf *pf)
5886{
5887 bool link_up;
5888 int i;
5889
5890 ice_for_each_vsi(pf, i) {
5891 struct ice_vsi *vsi = pf->vsi[i];
5892
5893 if (!vsi || vsi->type != ICE_VSI_PF)
5894 return;
5895
5896 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
5897 if (link_up) {
5898 netif_carrier_on(pf->vsi[i]->netdev);
5899 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
5900 } else {
5901 netif_carrier_off(pf->vsi[i]->netdev);
5902 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
5903 }
5904 }
5905}
5906
5907/**
5908 * ice_rebuild - rebuild after reset
5909 * @pf: PF to rebuild
5910 * @reset_type: type of reset
5911 *
5912 * Do not rebuild VF VSI in this flow because that is already handled via
5913 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
5914 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
5915 * to reset/rebuild all the VF VSI twice.
5916 */
5917static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
5918{
5919 struct device *dev = ice_pf_to_dev(pf);
5920 struct ice_hw *hw = &pf->hw;
5921 enum ice_status ret;
5922 int err;
5923
5924 if (test_bit(__ICE_DOWN, pf->state))
5925 goto clear_recovery;
5926
5927 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
5928
5929 ret = ice_init_all_ctrlq(hw);
5930 if (ret) {
5931 dev_err(dev, "control queues init failed %s\n",
5932 ice_stat_str(ret));
5933 goto err_init_ctrlq;
5934 }
5935
5936 /* if DDP was previously loaded successfully */
5937 if (!ice_is_safe_mode(pf)) {
5938 /* reload the SW DB of filter tables */
5939 if (reset_type == ICE_RESET_PFR)
5940 ice_fill_blk_tbls(hw);
5941 else
5942 /* Reload DDP Package after CORER/GLOBR reset */
5943 ice_load_pkg(NULL, pf);
5944 }
5945
5946 ret = ice_clear_pf_cfg(hw);
5947 if (ret) {
5948 dev_err(dev, "clear PF configuration failed %s\n",
5949 ice_stat_str(ret));
5950 goto err_init_ctrlq;
5951 }
5952
5953 if (pf->first_sw->dflt_vsi_ena)
5954 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
5955 /* clear the default VSI configuration if it exists */
5956 pf->first_sw->dflt_vsi = NULL;
5957 pf->first_sw->dflt_vsi_ena = false;
5958
5959 ice_clear_pxe_mode(hw);
5960
5961 ret = ice_get_caps(hw);
5962 if (ret) {
5963 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
5964 goto err_init_ctrlq;
5965 }
5966
5967 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
5968 if (ret) {
5969 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
5970 goto err_init_ctrlq;
5971 }
5972
5973 err = ice_sched_init_port(hw->port_info);
5974 if (err)
5975 goto err_sched_init_port;
5976
5977 /* start misc vector */
5978 err = ice_req_irq_msix_misc(pf);
5979 if (err) {
5980 dev_err(dev, "misc vector setup failed: %d\n", err);
5981 goto err_sched_init_port;
5982 }
5983
5984 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
5985 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
5986 if (!rd32(hw, PFQF_FD_SIZE)) {
5987 u16 unused, guar, b_effort;
5988
5989 guar = hw->func_caps.fd_fltr_guar;
5990 b_effort = hw->func_caps.fd_fltr_best_effort;
5991
5992 /* force guaranteed filter pool for PF */
5993 ice_alloc_fd_guar_item(hw, &unused, guar);
5994 /* force shared filter pool for PF */
5995 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
5996 }
5997 }
5998
5999 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6000 ice_dcb_rebuild(pf);
6001
6002 /* rebuild PF VSI */
6003 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6004 if (err) {
6005 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6006 goto err_vsi_rebuild;
6007 }
6008
6009 /* If Flow Director is active */
6010 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6011 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6012 if (err) {
6013 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6014 goto err_vsi_rebuild;
6015 }
6016
6017 /* replay HW Flow Director recipes */
6018 if (hw->fdir_prof)
6019 ice_fdir_replay_flows(hw);
6020
6021 /* replay Flow Director filters */
6022 ice_fdir_replay_fltrs(pf);
6023
6024 ice_rebuild_arfs(pf);
6025 }
6026
6027 ice_update_pf_netdev_link(pf);
6028
6029 /* tell the firmware we are up */
6030 ret = ice_send_version(pf);
6031 if (ret) {
6032 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6033 ice_stat_str(ret));
6034 goto err_vsi_rebuild;
6035 }
6036
6037 ice_replay_post(hw);
6038
6039 /* if we get here, reset flow is successful */
6040 clear_bit(__ICE_RESET_FAILED, pf->state);
6041 return;
6042
6043err_vsi_rebuild:
6044err_sched_init_port:
6045 ice_sched_cleanup_all(hw);
6046err_init_ctrlq:
6047 ice_shutdown_all_ctrlq(hw);
6048 set_bit(__ICE_RESET_FAILED, pf->state);
6049clear_recovery:
6050 /* set this bit in PF state to control service task scheduling */
6051 set_bit(__ICE_NEEDS_RESTART, pf->state);
6052 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6053}
6054
6055/**
6056 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6057 * @vsi: Pointer to VSI structure
6058 */
6059static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6060{
6061 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6062 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6063 else
6064 return ICE_RXBUF_3072;
6065}
6066
6067/**
6068 * ice_change_mtu - NDO callback to change the MTU
6069 * @netdev: network interface device structure
6070 * @new_mtu: new value for maximum frame size
6071 *
6072 * Returns 0 on success, negative on failure
6073 */
6074static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6075{
6076 struct ice_netdev_priv *np = netdev_priv(netdev);
6077 struct ice_vsi *vsi = np->vsi;
6078 struct ice_pf *pf = vsi->back;
6079 u8 count = 0;
6080
6081 if (new_mtu == (int)netdev->mtu) {
6082 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6083 return 0;
6084 }
6085
6086 if (ice_is_xdp_ena_vsi(vsi)) {
6087 int frame_size = ice_max_xdp_frame_size(vsi);
6088
6089 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6090 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6091 frame_size - ICE_ETH_PKT_HDR_PAD);
6092 return -EINVAL;
6093 }
6094 }
6095
6096 if (new_mtu < (int)netdev->min_mtu) {
6097 netdev_err(netdev, "new MTU invalid. min_mtu is %d\n",
6098 netdev->min_mtu);
6099 return -EINVAL;
6100 } else if (new_mtu > (int)netdev->max_mtu) {
6101 netdev_err(netdev, "new MTU invalid. max_mtu is %d\n",
6102 netdev->min_mtu);
6103 return -EINVAL;
6104 }
6105 /* if a reset is in progress, wait for some time for it to complete */
6106 do {
6107 if (ice_is_reset_in_progress(pf->state)) {
6108 count++;
6109 usleep_range(1000, 2000);
6110 } else {
6111 break;
6112 }
6113
6114 } while (count < 100);
6115
6116 if (count == 100) {
6117 netdev_err(netdev, "can't change MTU. Device is busy\n");
6118 return -EBUSY;
6119 }
6120
6121 netdev->mtu = (unsigned int)new_mtu;
6122
6123 /* if VSI is up, bring it down and then back up */
6124 if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
6125 int err;
6126
6127 err = ice_down(vsi);
6128 if (err) {
6129 netdev_err(netdev, "change MTU if_up err %d\n", err);
6130 return err;
6131 }
6132
6133 err = ice_up(vsi);
6134 if (err) {
6135 netdev_err(netdev, "change MTU if_up err %d\n", err);
6136 return err;
6137 }
6138 }
6139
6140 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6141 return 0;
6142}
6143
6144/**
6145 * ice_aq_str - convert AQ err code to a string
6146 * @aq_err: the AQ error code to convert
6147 */
6148const char *ice_aq_str(enum ice_aq_err aq_err)
6149{
6150 switch (aq_err) {
6151 case ICE_AQ_RC_OK:
6152 return "OK";
6153 case ICE_AQ_RC_EPERM:
6154 return "ICE_AQ_RC_EPERM";
6155 case ICE_AQ_RC_ENOENT:
6156 return "ICE_AQ_RC_ENOENT";
6157 case ICE_AQ_RC_ENOMEM:
6158 return "ICE_AQ_RC_ENOMEM";
6159 case ICE_AQ_RC_EBUSY:
6160 return "ICE_AQ_RC_EBUSY";
6161 case ICE_AQ_RC_EEXIST:
6162 return "ICE_AQ_RC_EEXIST";
6163 case ICE_AQ_RC_EINVAL:
6164 return "ICE_AQ_RC_EINVAL";
6165 case ICE_AQ_RC_ENOSPC:
6166 return "ICE_AQ_RC_ENOSPC";
6167 case ICE_AQ_RC_ENOSYS:
6168 return "ICE_AQ_RC_ENOSYS";
6169 case ICE_AQ_RC_EMODE:
6170 return "ICE_AQ_RC_EMODE";
6171 case ICE_AQ_RC_ENOSEC:
6172 return "ICE_AQ_RC_ENOSEC";
6173 case ICE_AQ_RC_EBADSIG:
6174 return "ICE_AQ_RC_EBADSIG";
6175 case ICE_AQ_RC_ESVN:
6176 return "ICE_AQ_RC_ESVN";
6177 case ICE_AQ_RC_EBADMAN:
6178 return "ICE_AQ_RC_EBADMAN";
6179 case ICE_AQ_RC_EBADBUF:
6180 return "ICE_AQ_RC_EBADBUF";
6181 }
6182
6183 return "ICE_AQ_RC_UNKNOWN";
6184}
6185
6186/**
6187 * ice_stat_str - convert status err code to a string
6188 * @stat_err: the status error code to convert
6189 */
6190const char *ice_stat_str(enum ice_status stat_err)
6191{
6192 switch (stat_err) {
6193 case ICE_SUCCESS:
6194 return "OK";
6195 case ICE_ERR_PARAM:
6196 return "ICE_ERR_PARAM";
6197 case ICE_ERR_NOT_IMPL:
6198 return "ICE_ERR_NOT_IMPL";
6199 case ICE_ERR_NOT_READY:
6200 return "ICE_ERR_NOT_READY";
6201 case ICE_ERR_NOT_SUPPORTED:
6202 return "ICE_ERR_NOT_SUPPORTED";
6203 case ICE_ERR_BAD_PTR:
6204 return "ICE_ERR_BAD_PTR";
6205 case ICE_ERR_INVAL_SIZE:
6206 return "ICE_ERR_INVAL_SIZE";
6207 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6208 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6209 case ICE_ERR_RESET_FAILED:
6210 return "ICE_ERR_RESET_FAILED";
6211 case ICE_ERR_FW_API_VER:
6212 return "ICE_ERR_FW_API_VER";
6213 case ICE_ERR_NO_MEMORY:
6214 return "ICE_ERR_NO_MEMORY";
6215 case ICE_ERR_CFG:
6216 return "ICE_ERR_CFG";
6217 case ICE_ERR_OUT_OF_RANGE:
6218 return "ICE_ERR_OUT_OF_RANGE";
6219 case ICE_ERR_ALREADY_EXISTS:
6220 return "ICE_ERR_ALREADY_EXISTS";
6221 case ICE_ERR_NVM_CHECKSUM:
6222 return "ICE_ERR_NVM_CHECKSUM";
6223 case ICE_ERR_BUF_TOO_SHORT:
6224 return "ICE_ERR_BUF_TOO_SHORT";
6225 case ICE_ERR_NVM_BLANK_MODE:
6226 return "ICE_ERR_NVM_BLANK_MODE";
6227 case ICE_ERR_IN_USE:
6228 return "ICE_ERR_IN_USE";
6229 case ICE_ERR_MAX_LIMIT:
6230 return "ICE_ERR_MAX_LIMIT";
6231 case ICE_ERR_RESET_ONGOING:
6232 return "ICE_ERR_RESET_ONGOING";
6233 case ICE_ERR_HW_TABLE:
6234 return "ICE_ERR_HW_TABLE";
6235 case ICE_ERR_DOES_NOT_EXIST:
6236 return "ICE_ERR_DOES_NOT_EXIST";
6237 case ICE_ERR_FW_DDP_MISMATCH:
6238 return "ICE_ERR_FW_DDP_MISMATCH";
6239 case ICE_ERR_AQ_ERROR:
6240 return "ICE_ERR_AQ_ERROR";
6241 case ICE_ERR_AQ_TIMEOUT:
6242 return "ICE_ERR_AQ_TIMEOUT";
6243 case ICE_ERR_AQ_FULL:
6244 return "ICE_ERR_AQ_FULL";
6245 case ICE_ERR_AQ_NO_WORK:
6246 return "ICE_ERR_AQ_NO_WORK";
6247 case ICE_ERR_AQ_EMPTY:
6248 return "ICE_ERR_AQ_EMPTY";
6249 case ICE_ERR_AQ_FW_CRITICAL:
6250 return "ICE_ERR_AQ_FW_CRITICAL";
6251 }
6252
6253 return "ICE_ERR_UNKNOWN";
6254}
6255
6256/**
6257 * ice_set_rss - Set RSS keys and lut
6258 * @vsi: Pointer to VSI structure
6259 * @seed: RSS hash seed
6260 * @lut: Lookup table
6261 * @lut_size: Lookup table size
6262 *
6263 * Returns 0 on success, negative on failure
6264 */
6265int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6266{
6267 struct ice_pf *pf = vsi->back;
6268 struct ice_hw *hw = &pf->hw;
6269 enum ice_status status;
6270 struct device *dev;
6271
6272 dev = ice_pf_to_dev(pf);
6273 if (seed) {
6274 struct ice_aqc_get_set_rss_keys *buf =
6275 (struct ice_aqc_get_set_rss_keys *)seed;
6276
6277 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
6278
6279 if (status) {
6280 dev_err(dev, "Cannot set RSS key, err %s aq_err %s\n",
6281 ice_stat_str(status),
6282 ice_aq_str(hw->adminq.sq_last_status));
6283 return -EIO;
6284 }
6285 }
6286
6287 if (lut) {
6288 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6289 lut, lut_size);
6290 if (status) {
6291 dev_err(dev, "Cannot set RSS lut, err %s aq_err %s\n",
6292 ice_stat_str(status),
6293 ice_aq_str(hw->adminq.sq_last_status));
6294 return -EIO;
6295 }
6296 }
6297
6298 return 0;
6299}
6300
6301/**
6302 * ice_get_rss - Get RSS keys and lut
6303 * @vsi: Pointer to VSI structure
6304 * @seed: Buffer to store the keys
6305 * @lut: Buffer to store the lookup table entries
6306 * @lut_size: Size of buffer to store the lookup table entries
6307 *
6308 * Returns 0 on success, negative on failure
6309 */
6310int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6311{
6312 struct ice_pf *pf = vsi->back;
6313 struct ice_hw *hw = &pf->hw;
6314 enum ice_status status;
6315 struct device *dev;
6316
6317 dev = ice_pf_to_dev(pf);
6318 if (seed) {
6319 struct ice_aqc_get_set_rss_keys *buf =
6320 (struct ice_aqc_get_set_rss_keys *)seed;
6321
6322 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
6323 if (status) {
6324 dev_err(dev, "Cannot get RSS key, err %s aq_err %s\n",
6325 ice_stat_str(status),
6326 ice_aq_str(hw->adminq.sq_last_status));
6327 return -EIO;
6328 }
6329 }
6330
6331 if (lut) {
6332 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6333 lut, lut_size);
6334 if (status) {
6335 dev_err(dev, "Cannot get RSS lut, err %s aq_err %s\n",
6336 ice_stat_str(status),
6337 ice_aq_str(hw->adminq.sq_last_status));
6338 return -EIO;
6339 }
6340 }
6341
6342 return 0;
6343}
6344
6345/**
6346 * ice_bridge_getlink - Get the hardware bridge mode
6347 * @skb: skb buff
6348 * @pid: process ID
6349 * @seq: RTNL message seq
6350 * @dev: the netdev being configured
6351 * @filter_mask: filter mask passed in
6352 * @nlflags: netlink flags passed in
6353 *
6354 * Return the bridge mode (VEB/VEPA)
6355 */
6356static int
6357ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6358 struct net_device *dev, u32 filter_mask, int nlflags)
6359{
6360 struct ice_netdev_priv *np = netdev_priv(dev);
6361 struct ice_vsi *vsi = np->vsi;
6362 struct ice_pf *pf = vsi->back;
6363 u16 bmode;
6364
6365 bmode = pf->first_sw->bridge_mode;
6366
6367 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6368 filter_mask, NULL);
6369}
6370
6371/**
6372 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6373 * @vsi: Pointer to VSI structure
6374 * @bmode: Hardware bridge mode (VEB/VEPA)
6375 *
6376 * Returns 0 on success, negative on failure
6377 */
6378static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6379{
6380 struct ice_aqc_vsi_props *vsi_props;
6381 struct ice_hw *hw = &vsi->back->hw;
6382 struct ice_vsi_ctx *ctxt;
6383 enum ice_status status;
6384 int ret = 0;
6385
6386 vsi_props = &vsi->info;
6387
6388 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6389 if (!ctxt)
6390 return -ENOMEM;
6391
6392 ctxt->info = vsi->info;
6393
6394 if (bmode == BRIDGE_MODE_VEB)
6395 /* change from VEPA to VEB mode */
6396 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6397 else
6398 /* change from VEB to VEPA mode */
6399 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6400 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6401
6402 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6403 if (status) {
6404 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6405 bmode, ice_stat_str(status),
6406 ice_aq_str(hw->adminq.sq_last_status));
6407 ret = -EIO;
6408 goto out;
6409 }
6410 /* Update sw flags for book keeping */
6411 vsi_props->sw_flags = ctxt->info.sw_flags;
6412
6413out:
6414 kfree(ctxt);
6415 return ret;
6416}
6417
6418/**
6419 * ice_bridge_setlink - Set the hardware bridge mode
6420 * @dev: the netdev being configured
6421 * @nlh: RTNL message
6422 * @flags: bridge setlink flags
6423 * @extack: netlink extended ack
6424 *
6425 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6426 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6427 * not already set for all VSIs connected to this switch. And also update the
6428 * unicast switch filter rules for the corresponding switch of the netdev.
6429 */
6430static int
6431ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6432 u16 __always_unused flags,
6433 struct netlink_ext_ack __always_unused *extack)
6434{
6435 struct ice_netdev_priv *np = netdev_priv(dev);
6436 struct ice_pf *pf = np->vsi->back;
6437 struct nlattr *attr, *br_spec;
6438 struct ice_hw *hw = &pf->hw;
6439 enum ice_status status;
6440 struct ice_sw *pf_sw;
6441 int rem, v, err = 0;
6442
6443 pf_sw = pf->first_sw;
6444 /* find the attribute in the netlink message */
6445 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6446
6447 nla_for_each_nested(attr, br_spec, rem) {
6448 __u16 mode;
6449
6450 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6451 continue;
6452 mode = nla_get_u16(attr);
6453 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6454 return -EINVAL;
6455 /* Continue if bridge mode is not being flipped */
6456 if (mode == pf_sw->bridge_mode)
6457 continue;
6458 /* Iterates through the PF VSI list and update the loopback
6459 * mode of the VSI
6460 */
6461 ice_for_each_vsi(pf, v) {
6462 if (!pf->vsi[v])
6463 continue;
6464 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6465 if (err)
6466 return err;
6467 }
6468
6469 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6470 /* Update the unicast switch filter rules for the corresponding
6471 * switch of the netdev
6472 */
6473 status = ice_update_sw_rule_bridge_mode(hw);
6474 if (status) {
6475 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6476 mode, ice_stat_str(status),
6477 ice_aq_str(hw->adminq.sq_last_status));
6478 /* revert hw->evb_veb */
6479 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6480 return -EIO;
6481 }
6482
6483 pf_sw->bridge_mode = mode;
6484 }
6485
6486 return 0;
6487}
6488
6489/**
6490 * ice_tx_timeout - Respond to a Tx Hang
6491 * @netdev: network interface device structure
6492 * @txqueue: Tx queue
6493 */
6494static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6495{
6496 struct ice_netdev_priv *np = netdev_priv(netdev);
6497 struct ice_ring *tx_ring = NULL;
6498 struct ice_vsi *vsi = np->vsi;
6499 struct ice_pf *pf = vsi->back;
6500 u32 i;
6501
6502 pf->tx_timeout_count++;
6503
6504 /* Check if PFC is enabled for the TC to which the queue belongs
6505 * to. If yes then Tx timeout is not caused by a hung queue, no
6506 * need to reset and rebuild
6507 */
6508 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6509 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6510 txqueue);
6511 return;
6512 }
6513
6514 /* now that we have an index, find the tx_ring struct */
6515 for (i = 0; i < vsi->num_txq; i++)
6516 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6517 if (txqueue == vsi->tx_rings[i]->q_index) {
6518 tx_ring = vsi->tx_rings[i];
6519 break;
6520 }
6521
6522 /* Reset recovery level if enough time has elapsed after last timeout.
6523 * Also ensure no new reset action happens before next timeout period.
6524 */
6525 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6526 pf->tx_timeout_recovery_level = 1;
6527 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6528 netdev->watchdog_timeo)))
6529 return;
6530
6531 if (tx_ring) {
6532 struct ice_hw *hw = &pf->hw;
6533 u32 head, val = 0;
6534
6535 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6536 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6537 /* Read interrupt register */
6538 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6539
6540 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6541 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6542 head, tx_ring->next_to_use, val);
6543 }
6544
6545 pf->tx_timeout_last_recovery = jiffies;
6546 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6547 pf->tx_timeout_recovery_level, txqueue);
6548
6549 switch (pf->tx_timeout_recovery_level) {
6550 case 1:
6551 set_bit(__ICE_PFR_REQ, pf->state);
6552 break;
6553 case 2:
6554 set_bit(__ICE_CORER_REQ, pf->state);
6555 break;
6556 case 3:
6557 set_bit(__ICE_GLOBR_REQ, pf->state);
6558 break;
6559 default:
6560 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6561 set_bit(__ICE_DOWN, pf->state);
6562 set_bit(__ICE_NEEDS_RESTART, vsi->state);
6563 set_bit(__ICE_SERVICE_DIS, pf->state);
6564 break;
6565 }
6566
6567 ice_service_task_schedule(pf);
6568 pf->tx_timeout_recovery_level++;
6569}
6570
6571/**
6572 * ice_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
6573 * @netdev: This physical port's netdev
6574 * @ti: Tunnel endpoint information
6575 */
6576static void
6577ice_udp_tunnel_add(struct net_device *netdev, struct udp_tunnel_info *ti)
6578{
6579 struct ice_netdev_priv *np = netdev_priv(netdev);
6580 struct ice_vsi *vsi = np->vsi;
6581 struct ice_pf *pf = vsi->back;
6582 enum ice_tunnel_type tnl_type;
6583 u16 port = ntohs(ti->port);
6584 enum ice_status status;
6585
6586 switch (ti->type) {
6587 case UDP_TUNNEL_TYPE_VXLAN:
6588 tnl_type = TNL_VXLAN;
6589 break;
6590 case UDP_TUNNEL_TYPE_GENEVE:
6591 tnl_type = TNL_GENEVE;
6592 break;
6593 default:
6594 netdev_err(netdev, "Unknown tunnel type\n");
6595 return;
6596 }
6597
6598 status = ice_create_tunnel(&pf->hw, tnl_type, port);
6599 if (status == ICE_ERR_OUT_OF_RANGE)
6600 netdev_info(netdev, "Max tunneled UDP ports reached, port %d not added\n",
6601 port);
6602 else if (status)
6603 netdev_err(netdev, "Error adding UDP tunnel - %s\n",
6604 ice_stat_str(status));
6605}
6606
6607/**
6608 * ice_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
6609 * @netdev: This physical port's netdev
6610 * @ti: Tunnel endpoint information
6611 */
6612static void
6613ice_udp_tunnel_del(struct net_device *netdev, struct udp_tunnel_info *ti)
6614{
6615 struct ice_netdev_priv *np = netdev_priv(netdev);
6616 struct ice_vsi *vsi = np->vsi;
6617 struct ice_pf *pf = vsi->back;
6618 u16 port = ntohs(ti->port);
6619 enum ice_status status;
6620 bool retval;
6621
6622 retval = ice_tunnel_port_in_use(&pf->hw, port, NULL);
6623 if (!retval) {
6624 netdev_info(netdev, "port %d not found in UDP tunnels list\n",
6625 port);
6626 return;
6627 }
6628
6629 status = ice_destroy_tunnel(&pf->hw, port, false);
6630 if (status)
6631 netdev_err(netdev, "error deleting port %d from UDP tunnels list\n",
6632 port);
6633}
6634
6635/**
6636 * ice_open - Called when a network interface becomes active
6637 * @netdev: network interface device structure
6638 *
6639 * The open entry point is called when a network interface is made
6640 * active by the system (IFF_UP). At this point all resources needed
6641 * for transmit and receive operations are allocated, the interrupt
6642 * handler is registered with the OS, the netdev watchdog is enabled,
6643 * and the stack is notified that the interface is ready.
6644 *
6645 * Returns 0 on success, negative value on failure
6646 */
6647int ice_open(struct net_device *netdev)
6648{
6649 struct ice_netdev_priv *np = netdev_priv(netdev);
6650 struct ice_vsi *vsi = np->vsi;
6651 struct ice_pf *pf = vsi->back;
6652 struct ice_port_info *pi;
6653 int err;
6654
6655 if (test_bit(__ICE_NEEDS_RESTART, pf->state)) {
6656 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6657 return -EIO;
6658 }
6659
6660 if (test_bit(__ICE_DOWN, pf->state)) {
6661 netdev_err(netdev, "device is not ready yet\n");
6662 return -EBUSY;
6663 }
6664
6665 netif_carrier_off(netdev);
6666
6667 pi = vsi->port_info;
6668 err = ice_update_link_info(pi);
6669 if (err) {
6670 netdev_err(netdev, "Failed to get link info, error %d\n",
6671 err);
6672 return err;
6673 }
6674
6675 /* Set PHY if there is media, otherwise, turn off PHY */
6676 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6677 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6678 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state)) {
6679 err = ice_init_phy_user_cfg(pi);
6680 if (err) {
6681 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6682 err);
6683 return err;
6684 }
6685 }
6686
6687 err = ice_configure_phy(vsi);
6688 if (err) {
6689 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6690 err);
6691 return err;
6692 }
6693 } else {
6694 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6695 err = ice_aq_set_link_restart_an(pi, false, NULL);
6696 if (err) {
6697 netdev_err(netdev, "Failed to set PHY state, VSI %d error %d\n",
6698 vsi->vsi_num, err);
6699 return err;
6700 }
6701 }
6702
6703 err = ice_vsi_open(vsi);
6704 if (err)
6705 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6706 vsi->vsi_num, vsi->vsw->sw_id);
6707
6708 /* Update existing tunnels information */
6709 udp_tunnel_get_rx_info(netdev);
6710
6711 return err;
6712}
6713
6714/**
6715 * ice_stop - Disables a network interface
6716 * @netdev: network interface device structure
6717 *
6718 * The stop entry point is called when an interface is de-activated by the OS,
6719 * and the netdevice enters the DOWN state. The hardware is still under the
6720 * driver's control, but the netdev interface is disabled.
6721 *
6722 * Returns success only - not allowed to fail
6723 */
6724int ice_stop(struct net_device *netdev)
6725{
6726 struct ice_netdev_priv *np = netdev_priv(netdev);
6727 struct ice_vsi *vsi = np->vsi;
6728
6729 ice_vsi_close(vsi);
6730
6731 return 0;
6732}
6733
6734/**
6735 * ice_features_check - Validate encapsulated packet conforms to limits
6736 * @skb: skb buffer
6737 * @netdev: This port's netdev
6738 * @features: Offload features that the stack believes apply
6739 */
6740static netdev_features_t
6741ice_features_check(struct sk_buff *skb,
6742 struct net_device __always_unused *netdev,
6743 netdev_features_t features)
6744{
6745 size_t len;
6746
6747 /* No point in doing any of this if neither checksum nor GSO are
6748 * being requested for this frame. We can rule out both by just
6749 * checking for CHECKSUM_PARTIAL
6750 */
6751 if (skb->ip_summed != CHECKSUM_PARTIAL)
6752 return features;
6753
6754 /* We cannot support GSO if the MSS is going to be less than
6755 * 64 bytes. If it is then we need to drop support for GSO.
6756 */
6757 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
6758 features &= ~NETIF_F_GSO_MASK;
6759
6760 len = skb_network_header(skb) - skb->data;
6761 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
6762 goto out_rm_features;
6763
6764 len = skb_transport_header(skb) - skb_network_header(skb);
6765 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6766 goto out_rm_features;
6767
6768 if (skb->encapsulation) {
6769 len = skb_inner_network_header(skb) - skb_transport_header(skb);
6770 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
6771 goto out_rm_features;
6772
6773 len = skb_inner_transport_header(skb) -
6774 skb_inner_network_header(skb);
6775 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6776 goto out_rm_features;
6777 }
6778
6779 return features;
6780out_rm_features:
6781 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
6782}
6783
6784static const struct net_device_ops ice_netdev_safe_mode_ops = {
6785 .ndo_open = ice_open,
6786 .ndo_stop = ice_stop,
6787 .ndo_start_xmit = ice_start_xmit,
6788 .ndo_set_mac_address = ice_set_mac_address,
6789 .ndo_validate_addr = eth_validate_addr,
6790 .ndo_change_mtu = ice_change_mtu,
6791 .ndo_get_stats64 = ice_get_stats64,
6792 .ndo_tx_timeout = ice_tx_timeout,
6793};
6794
6795static const struct net_device_ops ice_netdev_ops = {
6796 .ndo_open = ice_open,
6797 .ndo_stop = ice_stop,
6798 .ndo_start_xmit = ice_start_xmit,
6799 .ndo_features_check = ice_features_check,
6800 .ndo_set_rx_mode = ice_set_rx_mode,
6801 .ndo_set_mac_address = ice_set_mac_address,
6802 .ndo_validate_addr = eth_validate_addr,
6803 .ndo_change_mtu = ice_change_mtu,
6804 .ndo_get_stats64 = ice_get_stats64,
6805 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
6806 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
6807 .ndo_set_vf_mac = ice_set_vf_mac,
6808 .ndo_get_vf_config = ice_get_vf_cfg,
6809 .ndo_set_vf_trust = ice_set_vf_trust,
6810 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
6811 .ndo_set_vf_link_state = ice_set_vf_link_state,
6812 .ndo_get_vf_stats = ice_get_vf_stats,
6813 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
6814 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
6815 .ndo_set_features = ice_set_features,
6816 .ndo_bridge_getlink = ice_bridge_getlink,
6817 .ndo_bridge_setlink = ice_bridge_setlink,
6818 .ndo_fdb_add = ice_fdb_add,
6819 .ndo_fdb_del = ice_fdb_del,
6820#ifdef CONFIG_RFS_ACCEL
6821 .ndo_rx_flow_steer = ice_rx_flow_steer,
6822#endif
6823 .ndo_tx_timeout = ice_tx_timeout,
6824 .ndo_bpf = ice_xdp,
6825 .ndo_xdp_xmit = ice_xdp_xmit,
6826 .ndo_xsk_wakeup = ice_xsk_wakeup,
6827 .ndo_udp_tunnel_add = ice_udp_tunnel_add,
6828 .ndo_udp_tunnel_del = ice_udp_tunnel_del,
6829};
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
23#define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
24static const char ice_driver_string[] = DRV_SUMMARY;
25static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
26
27/* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
28#define ICE_DDP_PKG_PATH "intel/ice/ddp/"
29#define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
30
31MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
32MODULE_DESCRIPTION(DRV_SUMMARY);
33MODULE_LICENSE("GPL v2");
34MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
35
36static int debug = -1;
37module_param(debug, int, 0644);
38#ifndef CONFIG_DYNAMIC_DEBUG
39MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
40#else
41MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
42#endif /* !CONFIG_DYNAMIC_DEBUG */
43
44static DEFINE_IDA(ice_aux_ida);
45
46static struct workqueue_struct *ice_wq;
47static const struct net_device_ops ice_netdev_safe_mode_ops;
48static const struct net_device_ops ice_netdev_ops;
49static int ice_vsi_open(struct ice_vsi *vsi);
50
51static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
52
53static void ice_vsi_release_all(struct ice_pf *pf);
54
55bool netif_is_ice(struct net_device *dev)
56{
57 return dev && (dev->netdev_ops == &ice_netdev_ops);
58}
59
60/**
61 * ice_get_tx_pending - returns number of Tx descriptors not processed
62 * @ring: the ring of descriptors
63 */
64static u16 ice_get_tx_pending(struct ice_ring *ring)
65{
66 u16 head, tail;
67
68 head = ring->next_to_clean;
69 tail = ring->next_to_use;
70
71 if (head != tail)
72 return (head < tail) ?
73 tail - head : (tail + ring->count - head);
74 return 0;
75}
76
77/**
78 * ice_check_for_hang_subtask - check for and recover hung queues
79 * @pf: pointer to PF struct
80 */
81static void ice_check_for_hang_subtask(struct ice_pf *pf)
82{
83 struct ice_vsi *vsi = NULL;
84 struct ice_hw *hw;
85 unsigned int i;
86 int packets;
87 u32 v;
88
89 ice_for_each_vsi(pf, v)
90 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
91 vsi = pf->vsi[v];
92 break;
93 }
94
95 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
96 return;
97
98 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
99 return;
100
101 hw = &vsi->back->hw;
102
103 for (i = 0; i < vsi->num_txq; i++) {
104 struct ice_ring *tx_ring = vsi->tx_rings[i];
105
106 if (tx_ring && tx_ring->desc) {
107 /* If packet counter has not changed the queue is
108 * likely stalled, so force an interrupt for this
109 * queue.
110 *
111 * prev_pkt would be negative if there was no
112 * pending work.
113 */
114 packets = tx_ring->stats.pkts & INT_MAX;
115 if (tx_ring->tx_stats.prev_pkt == packets) {
116 /* Trigger sw interrupt to revive the queue */
117 ice_trigger_sw_intr(hw, tx_ring->q_vector);
118 continue;
119 }
120
121 /* Memory barrier between read of packet count and call
122 * to ice_get_tx_pending()
123 */
124 smp_rmb();
125 tx_ring->tx_stats.prev_pkt =
126 ice_get_tx_pending(tx_ring) ? packets : -1;
127 }
128 }
129}
130
131/**
132 * ice_init_mac_fltr - Set initial MAC filters
133 * @pf: board private structure
134 *
135 * Set initial set of MAC filters for PF VSI; configure filters for permanent
136 * address and broadcast address. If an error is encountered, netdevice will be
137 * unregistered.
138 */
139static int ice_init_mac_fltr(struct ice_pf *pf)
140{
141 enum ice_status status;
142 struct ice_vsi *vsi;
143 u8 *perm_addr;
144
145 vsi = ice_get_main_vsi(pf);
146 if (!vsi)
147 return -EINVAL;
148
149 perm_addr = vsi->port_info->mac.perm_addr;
150 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
151 if (status)
152 return -EIO;
153
154 return 0;
155}
156
157/**
158 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
159 * @netdev: the net device on which the sync is happening
160 * @addr: MAC address to sync
161 *
162 * This is a callback function which is called by the in kernel device sync
163 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
164 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
165 * MAC filters from the hardware.
166 */
167static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
168{
169 struct ice_netdev_priv *np = netdev_priv(netdev);
170 struct ice_vsi *vsi = np->vsi;
171
172 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
173 ICE_FWD_TO_VSI))
174 return -EINVAL;
175
176 return 0;
177}
178
179/**
180 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
181 * @netdev: the net device on which the unsync is happening
182 * @addr: MAC address to unsync
183 *
184 * This is a callback function which is called by the in kernel device unsync
185 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
186 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
187 * delete the MAC filters from the hardware.
188 */
189static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
190{
191 struct ice_netdev_priv *np = netdev_priv(netdev);
192 struct ice_vsi *vsi = np->vsi;
193
194 /* Under some circumstances, we might receive a request to delete our
195 * own device address from our uc list. Because we store the device
196 * address in the VSI's MAC filter list, we need to ignore such
197 * requests and not delete our device address from this list.
198 */
199 if (ether_addr_equal(addr, netdev->dev_addr))
200 return 0;
201
202 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
203 ICE_FWD_TO_VSI))
204 return -EINVAL;
205
206 return 0;
207}
208
209/**
210 * ice_vsi_fltr_changed - check if filter state changed
211 * @vsi: VSI to be checked
212 *
213 * returns true if filter state has changed, false otherwise.
214 */
215static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
216{
217 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
218 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
219 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
220}
221
222/**
223 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
224 * @vsi: the VSI being configured
225 * @promisc_m: mask of promiscuous config bits
226 * @set_promisc: enable or disable promisc flag request
227 *
228 */
229static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
230{
231 struct ice_hw *hw = &vsi->back->hw;
232 enum ice_status status = 0;
233
234 if (vsi->type != ICE_VSI_PF)
235 return 0;
236
237 if (vsi->num_vlan > 1) {
238 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
239 set_promisc);
240 } else {
241 if (set_promisc)
242 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
243 0);
244 else
245 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
246 0);
247 }
248
249 if (status)
250 return -EIO;
251
252 return 0;
253}
254
255/**
256 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
257 * @vsi: ptr to the VSI
258 *
259 * Push any outstanding VSI filter changes through the AdminQ.
260 */
261static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
262{
263 struct device *dev = ice_pf_to_dev(vsi->back);
264 struct net_device *netdev = vsi->netdev;
265 bool promisc_forced_on = false;
266 struct ice_pf *pf = vsi->back;
267 struct ice_hw *hw = &pf->hw;
268 enum ice_status status = 0;
269 u32 changed_flags = 0;
270 u8 promisc_m;
271 int err = 0;
272
273 if (!vsi->netdev)
274 return -EINVAL;
275
276 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
277 usleep_range(1000, 2000);
278
279 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
280 vsi->current_netdev_flags = vsi->netdev->flags;
281
282 INIT_LIST_HEAD(&vsi->tmp_sync_list);
283 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
284
285 if (ice_vsi_fltr_changed(vsi)) {
286 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
287 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
288 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
289
290 /* grab the netdev's addr_list_lock */
291 netif_addr_lock_bh(netdev);
292 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
293 ice_add_mac_to_unsync_list);
294 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
295 ice_add_mac_to_unsync_list);
296 /* our temp lists are populated. release lock */
297 netif_addr_unlock_bh(netdev);
298 }
299
300 /* Remove MAC addresses in the unsync list */
301 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
302 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
303 if (status) {
304 netdev_err(netdev, "Failed to delete MAC filters\n");
305 /* if we failed because of alloc failures, just bail */
306 if (status == ICE_ERR_NO_MEMORY) {
307 err = -ENOMEM;
308 goto out;
309 }
310 }
311
312 /* Add MAC addresses in the sync list */
313 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
314 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
315 /* If filter is added successfully or already exists, do not go into
316 * 'if' condition and report it as error. Instead continue processing
317 * rest of the function.
318 */
319 if (status && status != ICE_ERR_ALREADY_EXISTS) {
320 netdev_err(netdev, "Failed to add MAC filters\n");
321 /* If there is no more space for new umac filters, VSI
322 * should go into promiscuous mode. There should be some
323 * space reserved for promiscuous filters.
324 */
325 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
326 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
327 vsi->state)) {
328 promisc_forced_on = true;
329 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
330 vsi->vsi_num);
331 } else {
332 err = -EIO;
333 goto out;
334 }
335 }
336 /* check for changes in promiscuous modes */
337 if (changed_flags & IFF_ALLMULTI) {
338 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
339 if (vsi->num_vlan > 1)
340 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
341 else
342 promisc_m = ICE_MCAST_PROMISC_BITS;
343
344 err = ice_cfg_promisc(vsi, promisc_m, true);
345 if (err) {
346 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
347 vsi->vsi_num);
348 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
349 goto out_promisc;
350 }
351 } else {
352 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
353 if (vsi->num_vlan > 1)
354 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
355 else
356 promisc_m = ICE_MCAST_PROMISC_BITS;
357
358 err = ice_cfg_promisc(vsi, promisc_m, false);
359 if (err) {
360 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
361 vsi->vsi_num);
362 vsi->current_netdev_flags |= IFF_ALLMULTI;
363 goto out_promisc;
364 }
365 }
366 }
367
368 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
369 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
370 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
371 if (vsi->current_netdev_flags & IFF_PROMISC) {
372 /* Apply Rx filter rule to get traffic from wire */
373 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
374 err = ice_set_dflt_vsi(pf->first_sw, vsi);
375 if (err && err != -EEXIST) {
376 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
377 err, vsi->vsi_num);
378 vsi->current_netdev_flags &=
379 ~IFF_PROMISC;
380 goto out_promisc;
381 }
382 ice_cfg_vlan_pruning(vsi, false, false);
383 }
384 } else {
385 /* Clear Rx filter to remove traffic from wire */
386 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
387 err = ice_clear_dflt_vsi(pf->first_sw);
388 if (err) {
389 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
390 err, vsi->vsi_num);
391 vsi->current_netdev_flags |=
392 IFF_PROMISC;
393 goto out_promisc;
394 }
395 if (vsi->num_vlan > 1)
396 ice_cfg_vlan_pruning(vsi, true, false);
397 }
398 }
399 }
400 goto exit;
401
402out_promisc:
403 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
404 goto exit;
405out:
406 /* if something went wrong then set the changed flag so we try again */
407 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
408 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
409exit:
410 clear_bit(ICE_CFG_BUSY, vsi->state);
411 return err;
412}
413
414/**
415 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
416 * @pf: board private structure
417 */
418static void ice_sync_fltr_subtask(struct ice_pf *pf)
419{
420 int v;
421
422 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
423 return;
424
425 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
426
427 ice_for_each_vsi(pf, v)
428 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
429 ice_vsi_sync_fltr(pf->vsi[v])) {
430 /* come back and try again later */
431 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
432 break;
433 }
434}
435
436/**
437 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
438 * @pf: the PF
439 * @locked: is the rtnl_lock already held
440 */
441static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
442{
443 int node;
444 int v;
445
446 ice_for_each_vsi(pf, v)
447 if (pf->vsi[v])
448 ice_dis_vsi(pf->vsi[v], locked);
449
450 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
451 pf->pf_agg_node[node].num_vsis = 0;
452
453 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
454 pf->vf_agg_node[node].num_vsis = 0;
455}
456
457/**
458 * ice_prepare_for_reset - prep for the core to reset
459 * @pf: board private structure
460 *
461 * Inform or close all dependent features in prep for reset.
462 */
463static void
464ice_prepare_for_reset(struct ice_pf *pf)
465{
466 struct ice_hw *hw = &pf->hw;
467 unsigned int i;
468
469 /* already prepared for reset */
470 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
471 return;
472
473 ice_unplug_aux_dev(pf);
474
475 /* Notify VFs of impending reset */
476 if (ice_check_sq_alive(hw, &hw->mailboxq))
477 ice_vc_notify_reset(pf);
478
479 /* Disable VFs until reset is completed */
480 ice_for_each_vf(pf, i)
481 ice_set_vf_state_qs_dis(&pf->vf[i]);
482
483 /* clear SW filtering DB */
484 ice_clear_hw_tbls(hw);
485 /* disable the VSIs and their queues that are not already DOWN */
486 ice_pf_dis_all_vsi(pf, false);
487
488 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
489 ice_ptp_release(pf);
490
491 if (hw->port_info)
492 ice_sched_clear_port(hw->port_info);
493
494 ice_shutdown_all_ctrlq(hw);
495
496 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
497}
498
499/**
500 * ice_do_reset - Initiate one of many types of resets
501 * @pf: board private structure
502 * @reset_type: reset type requested
503 * before this function was called.
504 */
505static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
506{
507 struct device *dev = ice_pf_to_dev(pf);
508 struct ice_hw *hw = &pf->hw;
509
510 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
511
512 ice_prepare_for_reset(pf);
513
514 /* trigger the reset */
515 if (ice_reset(hw, reset_type)) {
516 dev_err(dev, "reset %d failed\n", reset_type);
517 set_bit(ICE_RESET_FAILED, pf->state);
518 clear_bit(ICE_RESET_OICR_RECV, pf->state);
519 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
520 clear_bit(ICE_PFR_REQ, pf->state);
521 clear_bit(ICE_CORER_REQ, pf->state);
522 clear_bit(ICE_GLOBR_REQ, pf->state);
523 wake_up(&pf->reset_wait_queue);
524 return;
525 }
526
527 /* PFR is a bit of a special case because it doesn't result in an OICR
528 * interrupt. So for PFR, rebuild after the reset and clear the reset-
529 * associated state bits.
530 */
531 if (reset_type == ICE_RESET_PFR) {
532 pf->pfr_count++;
533 ice_rebuild(pf, reset_type);
534 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
535 clear_bit(ICE_PFR_REQ, pf->state);
536 wake_up(&pf->reset_wait_queue);
537 ice_reset_all_vfs(pf, true);
538 }
539}
540
541/**
542 * ice_reset_subtask - Set up for resetting the device and driver
543 * @pf: board private structure
544 */
545static void ice_reset_subtask(struct ice_pf *pf)
546{
547 enum ice_reset_req reset_type = ICE_RESET_INVAL;
548
549 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
550 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
551 * of reset is pending and sets bits in pf->state indicating the reset
552 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
553 * prepare for pending reset if not already (for PF software-initiated
554 * global resets the software should already be prepared for it as
555 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
556 * by firmware or software on other PFs, that bit is not set so prepare
557 * for the reset now), poll for reset done, rebuild and return.
558 */
559 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
560 /* Perform the largest reset requested */
561 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
562 reset_type = ICE_RESET_CORER;
563 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
564 reset_type = ICE_RESET_GLOBR;
565 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
566 reset_type = ICE_RESET_EMPR;
567 /* return if no valid reset type requested */
568 if (reset_type == ICE_RESET_INVAL)
569 return;
570 ice_prepare_for_reset(pf);
571
572 /* make sure we are ready to rebuild */
573 if (ice_check_reset(&pf->hw)) {
574 set_bit(ICE_RESET_FAILED, pf->state);
575 } else {
576 /* done with reset. start rebuild */
577 pf->hw.reset_ongoing = false;
578 ice_rebuild(pf, reset_type);
579 /* clear bit to resume normal operations, but
580 * ICE_NEEDS_RESTART bit is set in case rebuild failed
581 */
582 clear_bit(ICE_RESET_OICR_RECV, pf->state);
583 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
584 clear_bit(ICE_PFR_REQ, pf->state);
585 clear_bit(ICE_CORER_REQ, pf->state);
586 clear_bit(ICE_GLOBR_REQ, pf->state);
587 wake_up(&pf->reset_wait_queue);
588 ice_reset_all_vfs(pf, true);
589 }
590
591 return;
592 }
593
594 /* No pending resets to finish processing. Check for new resets */
595 if (test_bit(ICE_PFR_REQ, pf->state))
596 reset_type = ICE_RESET_PFR;
597 if (test_bit(ICE_CORER_REQ, pf->state))
598 reset_type = ICE_RESET_CORER;
599 if (test_bit(ICE_GLOBR_REQ, pf->state))
600 reset_type = ICE_RESET_GLOBR;
601 /* If no valid reset type requested just return */
602 if (reset_type == ICE_RESET_INVAL)
603 return;
604
605 /* reset if not already down or busy */
606 if (!test_bit(ICE_DOWN, pf->state) &&
607 !test_bit(ICE_CFG_BUSY, pf->state)) {
608 ice_do_reset(pf, reset_type);
609 }
610}
611
612/**
613 * ice_print_topo_conflict - print topology conflict message
614 * @vsi: the VSI whose topology status is being checked
615 */
616static void ice_print_topo_conflict(struct ice_vsi *vsi)
617{
618 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
619 case ICE_AQ_LINK_TOPO_CONFLICT:
620 case ICE_AQ_LINK_MEDIA_CONFLICT:
621 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
622 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
623 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
624 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");
625 break;
626 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
627 netdev_info(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");
628 break;
629 default:
630 break;
631 }
632}
633
634/**
635 * ice_print_link_msg - print link up or down message
636 * @vsi: the VSI whose link status is being queried
637 * @isup: boolean for if the link is now up or down
638 */
639void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
640{
641 struct ice_aqc_get_phy_caps_data *caps;
642 const char *an_advertised;
643 enum ice_status status;
644 const char *fec_req;
645 const char *speed;
646 const char *fec;
647 const char *fc;
648 const char *an;
649
650 if (!vsi)
651 return;
652
653 if (vsi->current_isup == isup)
654 return;
655
656 vsi->current_isup = isup;
657
658 if (!isup) {
659 netdev_info(vsi->netdev, "NIC Link is Down\n");
660 return;
661 }
662
663 switch (vsi->port_info->phy.link_info.link_speed) {
664 case ICE_AQ_LINK_SPEED_100GB:
665 speed = "100 G";
666 break;
667 case ICE_AQ_LINK_SPEED_50GB:
668 speed = "50 G";
669 break;
670 case ICE_AQ_LINK_SPEED_40GB:
671 speed = "40 G";
672 break;
673 case ICE_AQ_LINK_SPEED_25GB:
674 speed = "25 G";
675 break;
676 case ICE_AQ_LINK_SPEED_20GB:
677 speed = "20 G";
678 break;
679 case ICE_AQ_LINK_SPEED_10GB:
680 speed = "10 G";
681 break;
682 case ICE_AQ_LINK_SPEED_5GB:
683 speed = "5 G";
684 break;
685 case ICE_AQ_LINK_SPEED_2500MB:
686 speed = "2.5 G";
687 break;
688 case ICE_AQ_LINK_SPEED_1000MB:
689 speed = "1 G";
690 break;
691 case ICE_AQ_LINK_SPEED_100MB:
692 speed = "100 M";
693 break;
694 default:
695 speed = "Unknown ";
696 break;
697 }
698
699 switch (vsi->port_info->fc.current_mode) {
700 case ICE_FC_FULL:
701 fc = "Rx/Tx";
702 break;
703 case ICE_FC_TX_PAUSE:
704 fc = "Tx";
705 break;
706 case ICE_FC_RX_PAUSE:
707 fc = "Rx";
708 break;
709 case ICE_FC_NONE:
710 fc = "None";
711 break;
712 default:
713 fc = "Unknown";
714 break;
715 }
716
717 /* Get FEC mode based on negotiated link info */
718 switch (vsi->port_info->phy.link_info.fec_info) {
719 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
720 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
721 fec = "RS-FEC";
722 break;
723 case ICE_AQ_LINK_25G_KR_FEC_EN:
724 fec = "FC-FEC/BASE-R";
725 break;
726 default:
727 fec = "NONE";
728 break;
729 }
730
731 /* check if autoneg completed, might be false due to not supported */
732 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
733 an = "True";
734 else
735 an = "False";
736
737 /* Get FEC mode requested based on PHY caps last SW configuration */
738 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
739 if (!caps) {
740 fec_req = "Unknown";
741 an_advertised = "Unknown";
742 goto done;
743 }
744
745 status = ice_aq_get_phy_caps(vsi->port_info, false,
746 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
747 if (status)
748 netdev_info(vsi->netdev, "Get phy capability failed.\n");
749
750 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
751
752 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
753 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
754 fec_req = "RS-FEC";
755 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
756 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
757 fec_req = "FC-FEC/BASE-R";
758 else
759 fec_req = "NONE";
760
761 kfree(caps);
762
763done:
764 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",
765 speed, fec_req, fec, an_advertised, an, fc);
766 ice_print_topo_conflict(vsi);
767}
768
769/**
770 * ice_vsi_link_event - update the VSI's netdev
771 * @vsi: the VSI on which the link event occurred
772 * @link_up: whether or not the VSI needs to be set up or down
773 */
774static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
775{
776 if (!vsi)
777 return;
778
779 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
780 return;
781
782 if (vsi->type == ICE_VSI_PF) {
783 if (link_up == netif_carrier_ok(vsi->netdev))
784 return;
785
786 if (link_up) {
787 netif_carrier_on(vsi->netdev);
788 netif_tx_wake_all_queues(vsi->netdev);
789 } else {
790 netif_carrier_off(vsi->netdev);
791 netif_tx_stop_all_queues(vsi->netdev);
792 }
793 }
794}
795
796/**
797 * ice_set_dflt_mib - send a default config MIB to the FW
798 * @pf: private PF struct
799 *
800 * This function sends a default configuration MIB to the FW.
801 *
802 * If this function errors out at any point, the driver is still able to
803 * function. The main impact is that LFC may not operate as expected.
804 * Therefore an error state in this function should be treated with a DBG
805 * message and continue on with driver rebuild/reenable.
806 */
807static void ice_set_dflt_mib(struct ice_pf *pf)
808{
809 struct device *dev = ice_pf_to_dev(pf);
810 u8 mib_type, *buf, *lldpmib = NULL;
811 u16 len, typelen, offset = 0;
812 struct ice_lldp_org_tlv *tlv;
813 struct ice_hw *hw = &pf->hw;
814 u32 ouisubtype;
815
816 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
817 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
818 if (!lldpmib) {
819 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
820 __func__);
821 return;
822 }
823
824 /* Add ETS CFG TLV */
825 tlv = (struct ice_lldp_org_tlv *)lldpmib;
826 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
827 ICE_IEEE_ETS_TLV_LEN);
828 tlv->typelen = htons(typelen);
829 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
830 ICE_IEEE_SUBTYPE_ETS_CFG);
831 tlv->ouisubtype = htonl(ouisubtype);
832
833 buf = tlv->tlvinfo;
834 buf[0] = 0;
835
836 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
837 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
838 * Octets 13 - 20 are TSA values - leave as zeros
839 */
840 buf[5] = 0x64;
841 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
842 offset += len + 2;
843 tlv = (struct ice_lldp_org_tlv *)
844 ((char *)tlv + sizeof(tlv->typelen) + len);
845
846 /* Add ETS REC TLV */
847 buf = tlv->tlvinfo;
848 tlv->typelen = htons(typelen);
849
850 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
851 ICE_IEEE_SUBTYPE_ETS_REC);
852 tlv->ouisubtype = htonl(ouisubtype);
853
854 /* First octet of buf is reserved
855 * Octets 1 - 4 map UP to TC - all UPs map to zero
856 * Octets 5 - 12 are BW values - set TC 0 to 100%.
857 * Octets 13 - 20 are TSA value - leave as zeros
858 */
859 buf[5] = 0x64;
860 offset += len + 2;
861 tlv = (struct ice_lldp_org_tlv *)
862 ((char *)tlv + sizeof(tlv->typelen) + len);
863
864 /* Add PFC CFG TLV */
865 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
866 ICE_IEEE_PFC_TLV_LEN);
867 tlv->typelen = htons(typelen);
868
869 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
870 ICE_IEEE_SUBTYPE_PFC_CFG);
871 tlv->ouisubtype = htonl(ouisubtype);
872
873 /* Octet 1 left as all zeros - PFC disabled */
874 buf[0] = 0x08;
875 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
876 offset += len + 2;
877
878 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
879 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
880
881 kfree(lldpmib);
882}
883
884/**
885 * ice_check_module_power
886 * @pf: pointer to PF struct
887 * @link_cfg_err: bitmap from the link info structure
888 *
889 * check module power level returned by a previous call to aq_get_link_info
890 * and print error messages if module power level is not supported
891 */
892static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
893{
894 /* if module power level is supported, clear the flag */
895 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
896 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
897 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
898 return;
899 }
900
901 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
902 * above block didn't clear this bit, there's nothing to do
903 */
904 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
905 return;
906
907 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
908 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
909 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
910 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
911 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
912 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
913 }
914}
915
916/**
917 * ice_link_event - process the link event
918 * @pf: PF that the link event is associated with
919 * @pi: port_info for the port that the link event is associated with
920 * @link_up: true if the physical link is up and false if it is down
921 * @link_speed: current link speed received from the link event
922 *
923 * Returns 0 on success and negative on failure
924 */
925static int
926ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
927 u16 link_speed)
928{
929 struct device *dev = ice_pf_to_dev(pf);
930 struct ice_phy_info *phy_info;
931 enum ice_status status;
932 struct ice_vsi *vsi;
933 u16 old_link_speed;
934 bool old_link;
935
936 phy_info = &pi->phy;
937 phy_info->link_info_old = phy_info->link_info;
938
939 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
940 old_link_speed = phy_info->link_info_old.link_speed;
941
942 /* update the link info structures and re-enable link events,
943 * don't bail on failure due to other book keeping needed
944 */
945 status = ice_update_link_info(pi);
946 if (status)
947 dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
948 pi->lport, ice_stat_str(status),
949 ice_aq_str(pi->hw->adminq.sq_last_status));
950
951 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
952
953 /* Check if the link state is up after updating link info, and treat
954 * this event as an UP event since the link is actually UP now.
955 */
956 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
957 link_up = true;
958
959 vsi = ice_get_main_vsi(pf);
960 if (!vsi || !vsi->port_info)
961 return -EINVAL;
962
963 /* turn off PHY if media was removed */
964 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
965 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
966 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
967 ice_set_link(vsi, false);
968 }
969
970 /* if the old link up/down and speed is the same as the new */
971 if (link_up == old_link && link_speed == old_link_speed)
972 return 0;
973
974 if (ice_is_dcb_active(pf)) {
975 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
976 ice_dcb_rebuild(pf);
977 } else {
978 if (link_up)
979 ice_set_dflt_mib(pf);
980 }
981 ice_vsi_link_event(vsi, link_up);
982 ice_print_link_msg(vsi, link_up);
983
984 ice_vc_notify_link_state(pf);
985
986 return 0;
987}
988
989/**
990 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
991 * @pf: board private structure
992 */
993static void ice_watchdog_subtask(struct ice_pf *pf)
994{
995 int i;
996
997 /* if interface is down do nothing */
998 if (test_bit(ICE_DOWN, pf->state) ||
999 test_bit(ICE_CFG_BUSY, pf->state))
1000 return;
1001
1002 /* make sure we don't do these things too often */
1003 if (time_before(jiffies,
1004 pf->serv_tmr_prev + pf->serv_tmr_period))
1005 return;
1006
1007 pf->serv_tmr_prev = jiffies;
1008
1009 /* Update the stats for active netdevs so the network stack
1010 * can look at updated numbers whenever it cares to
1011 */
1012 ice_update_pf_stats(pf);
1013 ice_for_each_vsi(pf, i)
1014 if (pf->vsi[i] && pf->vsi[i]->netdev)
1015 ice_update_vsi_stats(pf->vsi[i]);
1016}
1017
1018/**
1019 * ice_init_link_events - enable/initialize link events
1020 * @pi: pointer to the port_info instance
1021 *
1022 * Returns -EIO on failure, 0 on success
1023 */
1024static int ice_init_link_events(struct ice_port_info *pi)
1025{
1026 u16 mask;
1027
1028 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1029 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
1030
1031 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1032 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1033 pi->lport);
1034 return -EIO;
1035 }
1036
1037 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1038 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1039 pi->lport);
1040 return -EIO;
1041 }
1042
1043 return 0;
1044}
1045
1046/**
1047 * ice_handle_link_event - handle link event via ARQ
1048 * @pf: PF that the link event is associated with
1049 * @event: event structure containing link status info
1050 */
1051static int
1052ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1053{
1054 struct ice_aqc_get_link_status_data *link_data;
1055 struct ice_port_info *port_info;
1056 int status;
1057
1058 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1059 port_info = pf->hw.port_info;
1060 if (!port_info)
1061 return -EINVAL;
1062
1063 status = ice_link_event(pf, port_info,
1064 !!(link_data->link_info & ICE_AQ_LINK_UP),
1065 le16_to_cpu(link_data->link_speed));
1066 if (status)
1067 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1068 status);
1069
1070 return status;
1071}
1072
1073enum ice_aq_task_state {
1074 ICE_AQ_TASK_WAITING = 0,
1075 ICE_AQ_TASK_COMPLETE,
1076 ICE_AQ_TASK_CANCELED,
1077};
1078
1079struct ice_aq_task {
1080 struct hlist_node entry;
1081
1082 u16 opcode;
1083 struct ice_rq_event_info *event;
1084 enum ice_aq_task_state state;
1085};
1086
1087/**
1088 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1089 * @pf: pointer to the PF private structure
1090 * @opcode: the opcode to wait for
1091 * @timeout: how long to wait, in jiffies
1092 * @event: storage for the event info
1093 *
1094 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1095 * current thread will be put to sleep until the specified event occurs or
1096 * until the given timeout is reached.
1097 *
1098 * To obtain only the descriptor contents, pass an event without an allocated
1099 * msg_buf. If the complete data buffer is desired, allocate the
1100 * event->msg_buf with enough space ahead of time.
1101 *
1102 * Returns: zero on success, or a negative error code on failure.
1103 */
1104int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1105 struct ice_rq_event_info *event)
1106{
1107 struct device *dev = ice_pf_to_dev(pf);
1108 struct ice_aq_task *task;
1109 unsigned long start;
1110 long ret;
1111 int err;
1112
1113 task = kzalloc(sizeof(*task), GFP_KERNEL);
1114 if (!task)
1115 return -ENOMEM;
1116
1117 INIT_HLIST_NODE(&task->entry);
1118 task->opcode = opcode;
1119 task->event = event;
1120 task->state = ICE_AQ_TASK_WAITING;
1121
1122 spin_lock_bh(&pf->aq_wait_lock);
1123 hlist_add_head(&task->entry, &pf->aq_wait_list);
1124 spin_unlock_bh(&pf->aq_wait_lock);
1125
1126 start = jiffies;
1127
1128 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1129 timeout);
1130 switch (task->state) {
1131 case ICE_AQ_TASK_WAITING:
1132 err = ret < 0 ? ret : -ETIMEDOUT;
1133 break;
1134 case ICE_AQ_TASK_CANCELED:
1135 err = ret < 0 ? ret : -ECANCELED;
1136 break;
1137 case ICE_AQ_TASK_COMPLETE:
1138 err = ret < 0 ? ret : 0;
1139 break;
1140 default:
1141 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1142 err = -EINVAL;
1143 break;
1144 }
1145
1146 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1147 jiffies_to_msecs(jiffies - start),
1148 jiffies_to_msecs(timeout),
1149 opcode);
1150
1151 spin_lock_bh(&pf->aq_wait_lock);
1152 hlist_del(&task->entry);
1153 spin_unlock_bh(&pf->aq_wait_lock);
1154 kfree(task);
1155
1156 return err;
1157}
1158
1159/**
1160 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1161 * @pf: pointer to the PF private structure
1162 * @opcode: the opcode of the event
1163 * @event: the event to check
1164 *
1165 * Loops over the current list of pending threads waiting for an AdminQ event.
1166 * For each matching task, copy the contents of the event into the task
1167 * structure and wake up the thread.
1168 *
1169 * If multiple threads wait for the same opcode, they will all be woken up.
1170 *
1171 * Note that event->msg_buf will only be duplicated if the event has a buffer
1172 * with enough space already allocated. Otherwise, only the descriptor and
1173 * message length will be copied.
1174 *
1175 * Returns: true if an event was found, false otherwise
1176 */
1177static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1178 struct ice_rq_event_info *event)
1179{
1180 struct ice_aq_task *task;
1181 bool found = false;
1182
1183 spin_lock_bh(&pf->aq_wait_lock);
1184 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1185 if (task->state || task->opcode != opcode)
1186 continue;
1187
1188 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1189 task->event->msg_len = event->msg_len;
1190
1191 /* Only copy the data buffer if a destination was set */
1192 if (task->event->msg_buf &&
1193 task->event->buf_len > event->buf_len) {
1194 memcpy(task->event->msg_buf, event->msg_buf,
1195 event->buf_len);
1196 task->event->buf_len = event->buf_len;
1197 }
1198
1199 task->state = ICE_AQ_TASK_COMPLETE;
1200 found = true;
1201 }
1202 spin_unlock_bh(&pf->aq_wait_lock);
1203
1204 if (found)
1205 wake_up(&pf->aq_wait_queue);
1206}
1207
1208/**
1209 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1210 * @pf: the PF private structure
1211 *
1212 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1213 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1214 */
1215static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1216{
1217 struct ice_aq_task *task;
1218
1219 spin_lock_bh(&pf->aq_wait_lock);
1220 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1221 task->state = ICE_AQ_TASK_CANCELED;
1222 spin_unlock_bh(&pf->aq_wait_lock);
1223
1224 wake_up(&pf->aq_wait_queue);
1225}
1226
1227/**
1228 * __ice_clean_ctrlq - helper function to clean controlq rings
1229 * @pf: ptr to struct ice_pf
1230 * @q_type: specific Control queue type
1231 */
1232static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1233{
1234 struct device *dev = ice_pf_to_dev(pf);
1235 struct ice_rq_event_info event;
1236 struct ice_hw *hw = &pf->hw;
1237 struct ice_ctl_q_info *cq;
1238 u16 pending, i = 0;
1239 const char *qtype;
1240 u32 oldval, val;
1241
1242 /* Do not clean control queue if/when PF reset fails */
1243 if (test_bit(ICE_RESET_FAILED, pf->state))
1244 return 0;
1245
1246 switch (q_type) {
1247 case ICE_CTL_Q_ADMIN:
1248 cq = &hw->adminq;
1249 qtype = "Admin";
1250 break;
1251 case ICE_CTL_Q_SB:
1252 cq = &hw->sbq;
1253 qtype = "Sideband";
1254 break;
1255 case ICE_CTL_Q_MAILBOX:
1256 cq = &hw->mailboxq;
1257 qtype = "Mailbox";
1258 /* we are going to try to detect a malicious VF, so set the
1259 * state to begin detection
1260 */
1261 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1262 break;
1263 default:
1264 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1265 return 0;
1266 }
1267
1268 /* check for error indications - PF_xx_AxQLEN register layout for
1269 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1270 */
1271 val = rd32(hw, cq->rq.len);
1272 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1273 PF_FW_ARQLEN_ARQCRIT_M)) {
1274 oldval = val;
1275 if (val & PF_FW_ARQLEN_ARQVFE_M)
1276 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1277 qtype);
1278 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1279 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1280 qtype);
1281 }
1282 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1283 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1284 qtype);
1285 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1286 PF_FW_ARQLEN_ARQCRIT_M);
1287 if (oldval != val)
1288 wr32(hw, cq->rq.len, val);
1289 }
1290
1291 val = rd32(hw, cq->sq.len);
1292 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1293 PF_FW_ATQLEN_ATQCRIT_M)) {
1294 oldval = val;
1295 if (val & PF_FW_ATQLEN_ATQVFE_M)
1296 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1297 qtype);
1298 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1299 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1300 qtype);
1301 }
1302 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1303 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1304 qtype);
1305 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1306 PF_FW_ATQLEN_ATQCRIT_M);
1307 if (oldval != val)
1308 wr32(hw, cq->sq.len, val);
1309 }
1310
1311 event.buf_len = cq->rq_buf_size;
1312 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1313 if (!event.msg_buf)
1314 return 0;
1315
1316 do {
1317 enum ice_status ret;
1318 u16 opcode;
1319
1320 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1321 if (ret == ICE_ERR_AQ_NO_WORK)
1322 break;
1323 if (ret) {
1324 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1325 ice_stat_str(ret));
1326 break;
1327 }
1328
1329 opcode = le16_to_cpu(event.desc.opcode);
1330
1331 /* Notify any thread that might be waiting for this event */
1332 ice_aq_check_events(pf, opcode, &event);
1333
1334 switch (opcode) {
1335 case ice_aqc_opc_get_link_status:
1336 if (ice_handle_link_event(pf, &event))
1337 dev_err(dev, "Could not handle link event\n");
1338 break;
1339 case ice_aqc_opc_event_lan_overflow:
1340 ice_vf_lan_overflow_event(pf, &event);
1341 break;
1342 case ice_mbx_opc_send_msg_to_pf:
1343 if (!ice_is_malicious_vf(pf, &event, i, pending))
1344 ice_vc_process_vf_msg(pf, &event);
1345 break;
1346 case ice_aqc_opc_fw_logging:
1347 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1348 break;
1349 case ice_aqc_opc_lldp_set_mib_change:
1350 ice_dcb_process_lldp_set_mib_change(pf, &event);
1351 break;
1352 default:
1353 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1354 qtype, opcode);
1355 break;
1356 }
1357 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1358
1359 kfree(event.msg_buf);
1360
1361 return pending && (i == ICE_DFLT_IRQ_WORK);
1362}
1363
1364/**
1365 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1366 * @hw: pointer to hardware info
1367 * @cq: control queue information
1368 *
1369 * returns true if there are pending messages in a queue, false if there aren't
1370 */
1371static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1372{
1373 u16 ntu;
1374
1375 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1376 return cq->rq.next_to_clean != ntu;
1377}
1378
1379/**
1380 * ice_clean_adminq_subtask - clean the AdminQ rings
1381 * @pf: board private structure
1382 */
1383static void ice_clean_adminq_subtask(struct ice_pf *pf)
1384{
1385 struct ice_hw *hw = &pf->hw;
1386
1387 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1388 return;
1389
1390 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1391 return;
1392
1393 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1394
1395 /* There might be a situation where new messages arrive to a control
1396 * queue between processing the last message and clearing the
1397 * EVENT_PENDING bit. So before exiting, check queue head again (using
1398 * ice_ctrlq_pending) and process new messages if any.
1399 */
1400 if (ice_ctrlq_pending(hw, &hw->adminq))
1401 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1402
1403 ice_flush(hw);
1404}
1405
1406/**
1407 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1408 * @pf: board private structure
1409 */
1410static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1411{
1412 struct ice_hw *hw = &pf->hw;
1413
1414 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1415 return;
1416
1417 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1418 return;
1419
1420 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1421
1422 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1423 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1424
1425 ice_flush(hw);
1426}
1427
1428/**
1429 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1430 * @pf: board private structure
1431 */
1432static void ice_clean_sbq_subtask(struct ice_pf *pf)
1433{
1434 struct ice_hw *hw = &pf->hw;
1435
1436 /* Nothing to do here if sideband queue is not supported */
1437 if (!ice_is_sbq_supported(hw)) {
1438 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1439 return;
1440 }
1441
1442 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1443 return;
1444
1445 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1446 return;
1447
1448 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1449
1450 if (ice_ctrlq_pending(hw, &hw->sbq))
1451 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1452
1453 ice_flush(hw);
1454}
1455
1456/**
1457 * ice_service_task_schedule - schedule the service task to wake up
1458 * @pf: board private structure
1459 *
1460 * If not already scheduled, this puts the task into the work queue.
1461 */
1462void ice_service_task_schedule(struct ice_pf *pf)
1463{
1464 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1465 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1466 !test_bit(ICE_NEEDS_RESTART, pf->state))
1467 queue_work(ice_wq, &pf->serv_task);
1468}
1469
1470/**
1471 * ice_service_task_complete - finish up the service task
1472 * @pf: board private structure
1473 */
1474static void ice_service_task_complete(struct ice_pf *pf)
1475{
1476 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1477
1478 /* force memory (pf->state) to sync before next service task */
1479 smp_mb__before_atomic();
1480 clear_bit(ICE_SERVICE_SCHED, pf->state);
1481}
1482
1483/**
1484 * ice_service_task_stop - stop service task and cancel works
1485 * @pf: board private structure
1486 *
1487 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1488 * 1 otherwise.
1489 */
1490static int ice_service_task_stop(struct ice_pf *pf)
1491{
1492 int ret;
1493
1494 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1495
1496 if (pf->serv_tmr.function)
1497 del_timer_sync(&pf->serv_tmr);
1498 if (pf->serv_task.func)
1499 cancel_work_sync(&pf->serv_task);
1500
1501 clear_bit(ICE_SERVICE_SCHED, pf->state);
1502 return ret;
1503}
1504
1505/**
1506 * ice_service_task_restart - restart service task and schedule works
1507 * @pf: board private structure
1508 *
1509 * This function is needed for suspend and resume works (e.g WoL scenario)
1510 */
1511static void ice_service_task_restart(struct ice_pf *pf)
1512{
1513 clear_bit(ICE_SERVICE_DIS, pf->state);
1514 ice_service_task_schedule(pf);
1515}
1516
1517/**
1518 * ice_service_timer - timer callback to schedule service task
1519 * @t: pointer to timer_list
1520 */
1521static void ice_service_timer(struct timer_list *t)
1522{
1523 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1524
1525 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1526 ice_service_task_schedule(pf);
1527}
1528
1529/**
1530 * ice_handle_mdd_event - handle malicious driver detect event
1531 * @pf: pointer to the PF structure
1532 *
1533 * Called from service task. OICR interrupt handler indicates MDD event.
1534 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1535 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1536 * disable the queue, the PF can be configured to reset the VF using ethtool
1537 * private flag mdd-auto-reset-vf.
1538 */
1539static void ice_handle_mdd_event(struct ice_pf *pf)
1540{
1541 struct device *dev = ice_pf_to_dev(pf);
1542 struct ice_hw *hw = &pf->hw;
1543 unsigned int i;
1544 u32 reg;
1545
1546 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1547 /* Since the VF MDD event logging is rate limited, check if
1548 * there are pending MDD events.
1549 */
1550 ice_print_vfs_mdd_events(pf);
1551 return;
1552 }
1553
1554 /* find what triggered an MDD event */
1555 reg = rd32(hw, GL_MDET_TX_PQM);
1556 if (reg & GL_MDET_TX_PQM_VALID_M) {
1557 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1558 GL_MDET_TX_PQM_PF_NUM_S;
1559 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1560 GL_MDET_TX_PQM_VF_NUM_S;
1561 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1562 GL_MDET_TX_PQM_MAL_TYPE_S;
1563 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1564 GL_MDET_TX_PQM_QNUM_S);
1565
1566 if (netif_msg_tx_err(pf))
1567 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1568 event, queue, pf_num, vf_num);
1569 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1570 }
1571
1572 reg = rd32(hw, GL_MDET_TX_TCLAN);
1573 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1574 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1575 GL_MDET_TX_TCLAN_PF_NUM_S;
1576 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1577 GL_MDET_TX_TCLAN_VF_NUM_S;
1578 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1579 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1580 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1581 GL_MDET_TX_TCLAN_QNUM_S);
1582
1583 if (netif_msg_tx_err(pf))
1584 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1585 event, queue, pf_num, vf_num);
1586 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1587 }
1588
1589 reg = rd32(hw, GL_MDET_RX);
1590 if (reg & GL_MDET_RX_VALID_M) {
1591 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1592 GL_MDET_RX_PF_NUM_S;
1593 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1594 GL_MDET_RX_VF_NUM_S;
1595 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1596 GL_MDET_RX_MAL_TYPE_S;
1597 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1598 GL_MDET_RX_QNUM_S);
1599
1600 if (netif_msg_rx_err(pf))
1601 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1602 event, queue, pf_num, vf_num);
1603 wr32(hw, GL_MDET_RX, 0xffffffff);
1604 }
1605
1606 /* check to see if this PF caused an MDD event */
1607 reg = rd32(hw, PF_MDET_TX_PQM);
1608 if (reg & PF_MDET_TX_PQM_VALID_M) {
1609 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1610 if (netif_msg_tx_err(pf))
1611 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1612 }
1613
1614 reg = rd32(hw, PF_MDET_TX_TCLAN);
1615 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1616 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1617 if (netif_msg_tx_err(pf))
1618 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1619 }
1620
1621 reg = rd32(hw, PF_MDET_RX);
1622 if (reg & PF_MDET_RX_VALID_M) {
1623 wr32(hw, PF_MDET_RX, 0xFFFF);
1624 if (netif_msg_rx_err(pf))
1625 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1626 }
1627
1628 /* Check to see if one of the VFs caused an MDD event, and then
1629 * increment counters and set print pending
1630 */
1631 ice_for_each_vf(pf, i) {
1632 struct ice_vf *vf = &pf->vf[i];
1633
1634 reg = rd32(hw, VP_MDET_TX_PQM(i));
1635 if (reg & VP_MDET_TX_PQM_VALID_M) {
1636 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1637 vf->mdd_tx_events.count++;
1638 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1639 if (netif_msg_tx_err(pf))
1640 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1641 i);
1642 }
1643
1644 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1645 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1646 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1647 vf->mdd_tx_events.count++;
1648 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1649 if (netif_msg_tx_err(pf))
1650 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1651 i);
1652 }
1653
1654 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1655 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1656 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1657 vf->mdd_tx_events.count++;
1658 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1659 if (netif_msg_tx_err(pf))
1660 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1661 i);
1662 }
1663
1664 reg = rd32(hw, VP_MDET_RX(i));
1665 if (reg & VP_MDET_RX_VALID_M) {
1666 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1667 vf->mdd_rx_events.count++;
1668 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1669 if (netif_msg_rx_err(pf))
1670 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1671 i);
1672
1673 /* Since the queue is disabled on VF Rx MDD events, the
1674 * PF can be configured to reset the VF through ethtool
1675 * private flag mdd-auto-reset-vf.
1676 */
1677 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1678 /* VF MDD event counters will be cleared by
1679 * reset, so print the event prior to reset.
1680 */
1681 ice_print_vf_rx_mdd_event(vf);
1682 ice_reset_vf(&pf->vf[i], false);
1683 }
1684 }
1685 }
1686
1687 ice_print_vfs_mdd_events(pf);
1688}
1689
1690/**
1691 * ice_force_phys_link_state - Force the physical link state
1692 * @vsi: VSI to force the physical link state to up/down
1693 * @link_up: true/false indicates to set the physical link to up/down
1694 *
1695 * Force the physical link state by getting the current PHY capabilities from
1696 * hardware and setting the PHY config based on the determined capabilities. If
1697 * link changes a link event will be triggered because both the Enable Automatic
1698 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1699 *
1700 * Returns 0 on success, negative on failure
1701 */
1702static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1703{
1704 struct ice_aqc_get_phy_caps_data *pcaps;
1705 struct ice_aqc_set_phy_cfg_data *cfg;
1706 struct ice_port_info *pi;
1707 struct device *dev;
1708 int retcode;
1709
1710 if (!vsi || !vsi->port_info || !vsi->back)
1711 return -EINVAL;
1712 if (vsi->type != ICE_VSI_PF)
1713 return 0;
1714
1715 dev = ice_pf_to_dev(vsi->back);
1716
1717 pi = vsi->port_info;
1718
1719 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1720 if (!pcaps)
1721 return -ENOMEM;
1722
1723 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1724 NULL);
1725 if (retcode) {
1726 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1727 vsi->vsi_num, retcode);
1728 retcode = -EIO;
1729 goto out;
1730 }
1731
1732 /* No change in link */
1733 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1734 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1735 goto out;
1736
1737 /* Use the current user PHY configuration. The current user PHY
1738 * configuration is initialized during probe from PHY capabilities
1739 * software mode, and updated on set PHY configuration.
1740 */
1741 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1742 if (!cfg) {
1743 retcode = -ENOMEM;
1744 goto out;
1745 }
1746
1747 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1748 if (link_up)
1749 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1750 else
1751 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1752
1753 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1754 if (retcode) {
1755 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1756 vsi->vsi_num, retcode);
1757 retcode = -EIO;
1758 }
1759
1760 kfree(cfg);
1761out:
1762 kfree(pcaps);
1763 return retcode;
1764}
1765
1766/**
1767 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1768 * @pi: port info structure
1769 *
1770 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1771 */
1772static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1773{
1774 struct ice_aqc_get_phy_caps_data *pcaps;
1775 struct ice_pf *pf = pi->hw->back;
1776 enum ice_status status;
1777 int err = 0;
1778
1779 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1780 if (!pcaps)
1781 return -ENOMEM;
1782
1783 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1784 NULL);
1785
1786 if (status) {
1787 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1788 err = -EIO;
1789 goto out;
1790 }
1791
1792 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1793 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1794
1795out:
1796 kfree(pcaps);
1797 return err;
1798}
1799
1800/**
1801 * ice_init_link_dflt_override - Initialize link default override
1802 * @pi: port info structure
1803 *
1804 * Initialize link default override and PHY total port shutdown during probe
1805 */
1806static void ice_init_link_dflt_override(struct ice_port_info *pi)
1807{
1808 struct ice_link_default_override_tlv *ldo;
1809 struct ice_pf *pf = pi->hw->back;
1810
1811 ldo = &pf->link_dflt_override;
1812 if (ice_get_link_default_override(ldo, pi))
1813 return;
1814
1815 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1816 return;
1817
1818 /* Enable Total Port Shutdown (override/replace link-down-on-close
1819 * ethtool private flag) for ports with Port Disable bit set.
1820 */
1821 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1822 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1823}
1824
1825/**
1826 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1827 * @pi: port info structure
1828 *
1829 * If default override is enabled, initialize the user PHY cfg speed and FEC
1830 * settings using the default override mask from the NVM.
1831 *
1832 * The PHY should only be configured with the default override settings the
1833 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1834 * is used to indicate that the user PHY cfg default override is initialized
1835 * and the PHY has not been configured with the default override settings. The
1836 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1837 * configured.
1838 *
1839 * This function should be called only if the FW doesn't support default
1840 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1841 */
1842static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1843{
1844 struct ice_link_default_override_tlv *ldo;
1845 struct ice_aqc_set_phy_cfg_data *cfg;
1846 struct ice_phy_info *phy = &pi->phy;
1847 struct ice_pf *pf = pi->hw->back;
1848
1849 ldo = &pf->link_dflt_override;
1850
1851 /* If link default override is enabled, use to mask NVM PHY capabilities
1852 * for speed and FEC default configuration.
1853 */
1854 cfg = &phy->curr_user_phy_cfg;
1855
1856 if (ldo->phy_type_low || ldo->phy_type_high) {
1857 cfg->phy_type_low = pf->nvm_phy_type_lo &
1858 cpu_to_le64(ldo->phy_type_low);
1859 cfg->phy_type_high = pf->nvm_phy_type_hi &
1860 cpu_to_le64(ldo->phy_type_high);
1861 }
1862 cfg->link_fec_opt = ldo->fec_options;
1863 phy->curr_user_fec_req = ICE_FEC_AUTO;
1864
1865 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1866}
1867
1868/**
1869 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1870 * @pi: port info structure
1871 *
1872 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1873 * mode to default. The PHY defaults are from get PHY capabilities topology
1874 * with media so call when media is first available. An error is returned if
1875 * called when media is not available. The PHY initialization completed state is
1876 * set here.
1877 *
1878 * These configurations are used when setting PHY
1879 * configuration. The user PHY configuration is updated on set PHY
1880 * configuration. Returns 0 on success, negative on failure
1881 */
1882static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1883{
1884 struct ice_aqc_get_phy_caps_data *pcaps;
1885 struct ice_phy_info *phy = &pi->phy;
1886 struct ice_pf *pf = pi->hw->back;
1887 enum ice_status status;
1888 int err = 0;
1889
1890 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1891 return -EIO;
1892
1893 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1894 if (!pcaps)
1895 return -ENOMEM;
1896
1897 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1898 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1899 pcaps, NULL);
1900 else
1901 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1902 pcaps, NULL);
1903 if (status) {
1904 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1905 err = -EIO;
1906 goto err_out;
1907 }
1908
1909 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1910
1911 /* check if lenient mode is supported and enabled */
1912 if (ice_fw_supports_link_override(pi->hw) &&
1913 !(pcaps->module_compliance_enforcement &
1914 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1915 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1916
1917 /* if the FW supports default PHY configuration mode, then the driver
1918 * does not have to apply link override settings. If not,
1919 * initialize user PHY configuration with link override values
1920 */
1921 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1922 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1923 ice_init_phy_cfg_dflt_override(pi);
1924 goto out;
1925 }
1926 }
1927
1928 /* if link default override is not enabled, set user flow control and
1929 * FEC settings based on what get_phy_caps returned
1930 */
1931 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1932 pcaps->link_fec_options);
1933 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1934
1935out:
1936 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1937 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
1938err_out:
1939 kfree(pcaps);
1940 return err;
1941}
1942
1943/**
1944 * ice_configure_phy - configure PHY
1945 * @vsi: VSI of PHY
1946 *
1947 * Set the PHY configuration. If the current PHY configuration is the same as
1948 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1949 * configure the based get PHY capabilities for topology with media.
1950 */
1951static int ice_configure_phy(struct ice_vsi *vsi)
1952{
1953 struct device *dev = ice_pf_to_dev(vsi->back);
1954 struct ice_port_info *pi = vsi->port_info;
1955 struct ice_aqc_get_phy_caps_data *pcaps;
1956 struct ice_aqc_set_phy_cfg_data *cfg;
1957 struct ice_phy_info *phy = &pi->phy;
1958 struct ice_pf *pf = vsi->back;
1959 enum ice_status status;
1960 int err = 0;
1961
1962 /* Ensure we have media as we cannot configure a medialess port */
1963 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1964 return -EPERM;
1965
1966 ice_print_topo_conflict(vsi);
1967
1968 if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1969 return -EPERM;
1970
1971 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1972 return ice_force_phys_link_state(vsi, true);
1973
1974 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1975 if (!pcaps)
1976 return -ENOMEM;
1977
1978 /* Get current PHY config */
1979 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1980 NULL);
1981 if (status) {
1982 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1983 vsi->vsi_num, ice_stat_str(status));
1984 err = -EIO;
1985 goto done;
1986 }
1987
1988 /* If PHY enable link is configured and configuration has not changed,
1989 * there's nothing to do
1990 */
1991 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1992 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1993 goto done;
1994
1995 /* Use PHY topology as baseline for configuration */
1996 memset(pcaps, 0, sizeof(*pcaps));
1997 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1998 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1999 pcaps, NULL);
2000 else
2001 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2002 pcaps, NULL);
2003 if (status) {
2004 dev_err(dev, "Failed to get PHY caps, VSI %d error %s\n",
2005 vsi->vsi_num, ice_stat_str(status));
2006 err = -EIO;
2007 goto done;
2008 }
2009
2010 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2011 if (!cfg) {
2012 err = -ENOMEM;
2013 goto done;
2014 }
2015
2016 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2017
2018 /* Speed - If default override pending, use curr_user_phy_cfg set in
2019 * ice_init_phy_user_cfg_ldo.
2020 */
2021 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2022 vsi->back->state)) {
2023 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2024 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2025 } else {
2026 u64 phy_low = 0, phy_high = 0;
2027
2028 ice_update_phy_type(&phy_low, &phy_high,
2029 pi->phy.curr_user_speed_req);
2030 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2031 cfg->phy_type_high = pcaps->phy_type_high &
2032 cpu_to_le64(phy_high);
2033 }
2034
2035 /* Can't provide what was requested; use PHY capabilities */
2036 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2037 cfg->phy_type_low = pcaps->phy_type_low;
2038 cfg->phy_type_high = pcaps->phy_type_high;
2039 }
2040
2041 /* FEC */
2042 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2043
2044 /* Can't provide what was requested; use PHY capabilities */
2045 if (cfg->link_fec_opt !=
2046 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2047 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2048 cfg->link_fec_opt = pcaps->link_fec_options;
2049 }
2050
2051 /* Flow Control - always supported; no need to check against
2052 * capabilities
2053 */
2054 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2055
2056 /* Enable link and link update */
2057 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2058
2059 status = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2060 if (status) {
2061 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
2062 vsi->vsi_num, ice_stat_str(status));
2063 err = -EIO;
2064 }
2065
2066 kfree(cfg);
2067done:
2068 kfree(pcaps);
2069 return err;
2070}
2071
2072/**
2073 * ice_check_media_subtask - Check for media
2074 * @pf: pointer to PF struct
2075 *
2076 * If media is available, then initialize PHY user configuration if it is not
2077 * been, and configure the PHY if the interface is up.
2078 */
2079static void ice_check_media_subtask(struct ice_pf *pf)
2080{
2081 struct ice_port_info *pi;
2082 struct ice_vsi *vsi;
2083 int err;
2084
2085 /* No need to check for media if it's already present */
2086 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2087 return;
2088
2089 vsi = ice_get_main_vsi(pf);
2090 if (!vsi)
2091 return;
2092
2093 /* Refresh link info and check if media is present */
2094 pi = vsi->port_info;
2095 err = ice_update_link_info(pi);
2096 if (err)
2097 return;
2098
2099 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
2100
2101 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2102 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2103 ice_init_phy_user_cfg(pi);
2104
2105 /* PHY settings are reset on media insertion, reconfigure
2106 * PHY to preserve settings.
2107 */
2108 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2109 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2110 return;
2111
2112 err = ice_configure_phy(vsi);
2113 if (!err)
2114 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2115
2116 /* A Link Status Event will be generated; the event handler
2117 * will complete bringing the interface up
2118 */
2119 }
2120}
2121
2122/**
2123 * ice_service_task - manage and run subtasks
2124 * @work: pointer to work_struct contained by the PF struct
2125 */
2126static void ice_service_task(struct work_struct *work)
2127{
2128 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2129 unsigned long start_time = jiffies;
2130
2131 /* subtasks */
2132
2133 /* process reset requests first */
2134 ice_reset_subtask(pf);
2135
2136 /* bail if a reset/recovery cycle is pending or rebuild failed */
2137 if (ice_is_reset_in_progress(pf->state) ||
2138 test_bit(ICE_SUSPENDED, pf->state) ||
2139 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2140 ice_service_task_complete(pf);
2141 return;
2142 }
2143
2144 ice_clean_adminq_subtask(pf);
2145 ice_check_media_subtask(pf);
2146 ice_check_for_hang_subtask(pf);
2147 ice_sync_fltr_subtask(pf);
2148 ice_handle_mdd_event(pf);
2149 ice_watchdog_subtask(pf);
2150
2151 if (ice_is_safe_mode(pf)) {
2152 ice_service_task_complete(pf);
2153 return;
2154 }
2155
2156 ice_process_vflr_event(pf);
2157 ice_clean_mailboxq_subtask(pf);
2158 ice_clean_sbq_subtask(pf);
2159 ice_sync_arfs_fltrs(pf);
2160 ice_flush_fdir_ctx(pf);
2161
2162 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2163 ice_service_task_complete(pf);
2164
2165 /* If the tasks have taken longer than one service timer period
2166 * or there is more work to be done, reset the service timer to
2167 * schedule the service task now.
2168 */
2169 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2170 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2171 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2172 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2173 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2174 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2175 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2176 mod_timer(&pf->serv_tmr, jiffies);
2177}
2178
2179/**
2180 * ice_set_ctrlq_len - helper function to set controlq length
2181 * @hw: pointer to the HW instance
2182 */
2183static void ice_set_ctrlq_len(struct ice_hw *hw)
2184{
2185 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2186 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2187 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2188 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2189 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2190 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2191 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2192 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2193 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2194 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2195 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2196 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2197}
2198
2199/**
2200 * ice_schedule_reset - schedule a reset
2201 * @pf: board private structure
2202 * @reset: reset being requested
2203 */
2204int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2205{
2206 struct device *dev = ice_pf_to_dev(pf);
2207
2208 /* bail out if earlier reset has failed */
2209 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2210 dev_dbg(dev, "earlier reset has failed\n");
2211 return -EIO;
2212 }
2213 /* bail if reset/recovery already in progress */
2214 if (ice_is_reset_in_progress(pf->state)) {
2215 dev_dbg(dev, "Reset already in progress\n");
2216 return -EBUSY;
2217 }
2218
2219 ice_unplug_aux_dev(pf);
2220
2221 switch (reset) {
2222 case ICE_RESET_PFR:
2223 set_bit(ICE_PFR_REQ, pf->state);
2224 break;
2225 case ICE_RESET_CORER:
2226 set_bit(ICE_CORER_REQ, pf->state);
2227 break;
2228 case ICE_RESET_GLOBR:
2229 set_bit(ICE_GLOBR_REQ, pf->state);
2230 break;
2231 default:
2232 return -EINVAL;
2233 }
2234
2235 ice_service_task_schedule(pf);
2236 return 0;
2237}
2238
2239/**
2240 * ice_irq_affinity_notify - Callback for affinity changes
2241 * @notify: context as to what irq was changed
2242 * @mask: the new affinity mask
2243 *
2244 * This is a callback function used by the irq_set_affinity_notifier function
2245 * so that we may register to receive changes to the irq affinity masks.
2246 */
2247static void
2248ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2249 const cpumask_t *mask)
2250{
2251 struct ice_q_vector *q_vector =
2252 container_of(notify, struct ice_q_vector, affinity_notify);
2253
2254 cpumask_copy(&q_vector->affinity_mask, mask);
2255}
2256
2257/**
2258 * ice_irq_affinity_release - Callback for affinity notifier release
2259 * @ref: internal core kernel usage
2260 *
2261 * This is a callback function used by the irq_set_affinity_notifier function
2262 * to inform the current notification subscriber that they will no longer
2263 * receive notifications.
2264 */
2265static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2266
2267/**
2268 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2269 * @vsi: the VSI being configured
2270 */
2271static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2272{
2273 struct ice_hw *hw = &vsi->back->hw;
2274 int i;
2275
2276 ice_for_each_q_vector(vsi, i)
2277 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2278
2279 ice_flush(hw);
2280 return 0;
2281}
2282
2283/**
2284 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2285 * @vsi: the VSI being configured
2286 * @basename: name for the vector
2287 */
2288static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2289{
2290 int q_vectors = vsi->num_q_vectors;
2291 struct ice_pf *pf = vsi->back;
2292 int base = vsi->base_vector;
2293 struct device *dev;
2294 int rx_int_idx = 0;
2295 int tx_int_idx = 0;
2296 int vector, err;
2297 int irq_num;
2298
2299 dev = ice_pf_to_dev(pf);
2300 for (vector = 0; vector < q_vectors; vector++) {
2301 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2302
2303 irq_num = pf->msix_entries[base + vector].vector;
2304
2305 if (q_vector->tx.ring && q_vector->rx.ring) {
2306 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2307 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2308 tx_int_idx++;
2309 } else if (q_vector->rx.ring) {
2310 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2311 "%s-%s-%d", basename, "rx", rx_int_idx++);
2312 } else if (q_vector->tx.ring) {
2313 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2314 "%s-%s-%d", basename, "tx", tx_int_idx++);
2315 } else {
2316 /* skip this unused q_vector */
2317 continue;
2318 }
2319 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2320 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2321 IRQF_SHARED, q_vector->name,
2322 q_vector);
2323 else
2324 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2325 0, q_vector->name, q_vector);
2326 if (err) {
2327 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2328 err);
2329 goto free_q_irqs;
2330 }
2331
2332 /* register for affinity change notifications */
2333 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2334 struct irq_affinity_notify *affinity_notify;
2335
2336 affinity_notify = &q_vector->affinity_notify;
2337 affinity_notify->notify = ice_irq_affinity_notify;
2338 affinity_notify->release = ice_irq_affinity_release;
2339 irq_set_affinity_notifier(irq_num, affinity_notify);
2340 }
2341
2342 /* assign the mask for this irq */
2343 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2344 }
2345
2346 vsi->irqs_ready = true;
2347 return 0;
2348
2349free_q_irqs:
2350 while (vector) {
2351 vector--;
2352 irq_num = pf->msix_entries[base + vector].vector;
2353 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2354 irq_set_affinity_notifier(irq_num, NULL);
2355 irq_set_affinity_hint(irq_num, NULL);
2356 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2357 }
2358 return err;
2359}
2360
2361/**
2362 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2363 * @vsi: VSI to setup Tx rings used by XDP
2364 *
2365 * Return 0 on success and negative value on error
2366 */
2367static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2368{
2369 struct device *dev = ice_pf_to_dev(vsi->back);
2370 int i;
2371
2372 for (i = 0; i < vsi->num_xdp_txq; i++) {
2373 u16 xdp_q_idx = vsi->alloc_txq + i;
2374 struct ice_ring *xdp_ring;
2375
2376 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2377
2378 if (!xdp_ring)
2379 goto free_xdp_rings;
2380
2381 xdp_ring->q_index = xdp_q_idx;
2382 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2383 xdp_ring->ring_active = false;
2384 xdp_ring->vsi = vsi;
2385 xdp_ring->netdev = NULL;
2386 xdp_ring->dev = dev;
2387 xdp_ring->count = vsi->num_tx_desc;
2388 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2389 if (ice_setup_tx_ring(xdp_ring))
2390 goto free_xdp_rings;
2391 ice_set_ring_xdp(xdp_ring);
2392 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2393 }
2394
2395 return 0;
2396
2397free_xdp_rings:
2398 for (; i >= 0; i--)
2399 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2400 ice_free_tx_ring(vsi->xdp_rings[i]);
2401 return -ENOMEM;
2402}
2403
2404/**
2405 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2406 * @vsi: VSI to set the bpf prog on
2407 * @prog: the bpf prog pointer
2408 */
2409static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2410{
2411 struct bpf_prog *old_prog;
2412 int i;
2413
2414 old_prog = xchg(&vsi->xdp_prog, prog);
2415 if (old_prog)
2416 bpf_prog_put(old_prog);
2417
2418 ice_for_each_rxq(vsi, i)
2419 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2420}
2421
2422/**
2423 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2424 * @vsi: VSI to bring up Tx rings used by XDP
2425 * @prog: bpf program that will be assigned to VSI
2426 *
2427 * Return 0 on success and negative value on error
2428 */
2429int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2430{
2431 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2432 int xdp_rings_rem = vsi->num_xdp_txq;
2433 struct ice_pf *pf = vsi->back;
2434 struct ice_qs_cfg xdp_qs_cfg = {
2435 .qs_mutex = &pf->avail_q_mutex,
2436 .pf_map = pf->avail_txqs,
2437 .pf_map_size = pf->max_pf_txqs,
2438 .q_count = vsi->num_xdp_txq,
2439 .scatter_count = ICE_MAX_SCATTER_TXQS,
2440 .vsi_map = vsi->txq_map,
2441 .vsi_map_offset = vsi->alloc_txq,
2442 .mapping_mode = ICE_VSI_MAP_CONTIG
2443 };
2444 enum ice_status status;
2445 struct device *dev;
2446 int i, v_idx;
2447
2448 dev = ice_pf_to_dev(pf);
2449 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2450 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2451 if (!vsi->xdp_rings)
2452 return -ENOMEM;
2453
2454 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2455 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2456 goto err_map_xdp;
2457
2458 if (ice_xdp_alloc_setup_rings(vsi))
2459 goto clear_xdp_rings;
2460
2461 /* follow the logic from ice_vsi_map_rings_to_vectors */
2462 ice_for_each_q_vector(vsi, v_idx) {
2463 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2464 int xdp_rings_per_v, q_id, q_base;
2465
2466 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2467 vsi->num_q_vectors - v_idx);
2468 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2469
2470 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2471 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2472
2473 xdp_ring->q_vector = q_vector;
2474 xdp_ring->next = q_vector->tx.ring;
2475 q_vector->tx.ring = xdp_ring;
2476 }
2477 xdp_rings_rem -= xdp_rings_per_v;
2478 }
2479
2480 /* omit the scheduler update if in reset path; XDP queues will be
2481 * taken into account at the end of ice_vsi_rebuild, where
2482 * ice_cfg_vsi_lan is being called
2483 */
2484 if (ice_is_reset_in_progress(pf->state))
2485 return 0;
2486
2487 /* tell the Tx scheduler that right now we have
2488 * additional queues
2489 */
2490 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2491 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2492
2493 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2494 max_txqs);
2495 if (status) {
2496 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2497 ice_stat_str(status));
2498 goto clear_xdp_rings;
2499 }
2500 ice_vsi_assign_bpf_prog(vsi, prog);
2501
2502 return 0;
2503clear_xdp_rings:
2504 for (i = 0; i < vsi->num_xdp_txq; i++)
2505 if (vsi->xdp_rings[i]) {
2506 kfree_rcu(vsi->xdp_rings[i], rcu);
2507 vsi->xdp_rings[i] = NULL;
2508 }
2509
2510err_map_xdp:
2511 mutex_lock(&pf->avail_q_mutex);
2512 for (i = 0; i < vsi->num_xdp_txq; i++) {
2513 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2514 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2515 }
2516 mutex_unlock(&pf->avail_q_mutex);
2517
2518 devm_kfree(dev, vsi->xdp_rings);
2519 return -ENOMEM;
2520}
2521
2522/**
2523 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2524 * @vsi: VSI to remove XDP rings
2525 *
2526 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2527 * resources
2528 */
2529int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2530{
2531 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2532 struct ice_pf *pf = vsi->back;
2533 int i, v_idx;
2534
2535 /* q_vectors are freed in reset path so there's no point in detaching
2536 * rings; in case of rebuild being triggered not from reset bits
2537 * in pf->state won't be set, so additionally check first q_vector
2538 * against NULL
2539 */
2540 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2541 goto free_qmap;
2542
2543 ice_for_each_q_vector(vsi, v_idx) {
2544 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2545 struct ice_ring *ring;
2546
2547 ice_for_each_ring(ring, q_vector->tx)
2548 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2549 break;
2550
2551 /* restore the value of last node prior to XDP setup */
2552 q_vector->tx.ring = ring;
2553 }
2554
2555free_qmap:
2556 mutex_lock(&pf->avail_q_mutex);
2557 for (i = 0; i < vsi->num_xdp_txq; i++) {
2558 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2559 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2560 }
2561 mutex_unlock(&pf->avail_q_mutex);
2562
2563 for (i = 0; i < vsi->num_xdp_txq; i++)
2564 if (vsi->xdp_rings[i]) {
2565 if (vsi->xdp_rings[i]->desc)
2566 ice_free_tx_ring(vsi->xdp_rings[i]);
2567 kfree_rcu(vsi->xdp_rings[i], rcu);
2568 vsi->xdp_rings[i] = NULL;
2569 }
2570
2571 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2572 vsi->xdp_rings = NULL;
2573
2574 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2575 return 0;
2576
2577 ice_vsi_assign_bpf_prog(vsi, NULL);
2578
2579 /* notify Tx scheduler that we destroyed XDP queues and bring
2580 * back the old number of child nodes
2581 */
2582 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2583 max_txqs[i] = vsi->num_txq;
2584
2585 /* change number of XDP Tx queues to 0 */
2586 vsi->num_xdp_txq = 0;
2587
2588 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2589 max_txqs);
2590}
2591
2592/**
2593 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2594 * @vsi: VSI to schedule napi on
2595 */
2596static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2597{
2598 int i;
2599
2600 ice_for_each_rxq(vsi, i) {
2601 struct ice_ring *rx_ring = vsi->rx_rings[i];
2602
2603 if (rx_ring->xsk_pool)
2604 napi_schedule(&rx_ring->q_vector->napi);
2605 }
2606}
2607
2608/**
2609 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2610 * @vsi: VSI to setup XDP for
2611 * @prog: XDP program
2612 * @extack: netlink extended ack
2613 */
2614static int
2615ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2616 struct netlink_ext_ack *extack)
2617{
2618 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2619 bool if_running = netif_running(vsi->netdev);
2620 int ret = 0, xdp_ring_err = 0;
2621
2622 if (frame_size > vsi->rx_buf_len) {
2623 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2624 return -EOPNOTSUPP;
2625 }
2626
2627 /* need to stop netdev while setting up the program for Rx rings */
2628 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2629 ret = ice_down(vsi);
2630 if (ret) {
2631 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2632 return ret;
2633 }
2634 }
2635
2636 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2637 vsi->num_xdp_txq = vsi->alloc_rxq;
2638 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2639 if (xdp_ring_err)
2640 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2641 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2642 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2643 if (xdp_ring_err)
2644 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2645 } else {
2646 ice_vsi_assign_bpf_prog(vsi, prog);
2647 }
2648
2649 if (if_running)
2650 ret = ice_up(vsi);
2651
2652 if (!ret && prog)
2653 ice_vsi_rx_napi_schedule(vsi);
2654
2655 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2656}
2657
2658/**
2659 * ice_xdp_safe_mode - XDP handler for safe mode
2660 * @dev: netdevice
2661 * @xdp: XDP command
2662 */
2663static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2664 struct netdev_bpf *xdp)
2665{
2666 NL_SET_ERR_MSG_MOD(xdp->extack,
2667 "Please provide working DDP firmware package in order to use XDP\n"
2668 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2669 return -EOPNOTSUPP;
2670}
2671
2672/**
2673 * ice_xdp - implements XDP handler
2674 * @dev: netdevice
2675 * @xdp: XDP command
2676 */
2677static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2678{
2679 struct ice_netdev_priv *np = netdev_priv(dev);
2680 struct ice_vsi *vsi = np->vsi;
2681
2682 if (vsi->type != ICE_VSI_PF) {
2683 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2684 return -EINVAL;
2685 }
2686
2687 switch (xdp->command) {
2688 case XDP_SETUP_PROG:
2689 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2690 case XDP_SETUP_XSK_POOL:
2691 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2692 xdp->xsk.queue_id);
2693 default:
2694 return -EINVAL;
2695 }
2696}
2697
2698/**
2699 * ice_ena_misc_vector - enable the non-queue interrupts
2700 * @pf: board private structure
2701 */
2702static void ice_ena_misc_vector(struct ice_pf *pf)
2703{
2704 struct ice_hw *hw = &pf->hw;
2705 u32 val;
2706
2707 /* Disable anti-spoof detection interrupt to prevent spurious event
2708 * interrupts during a function reset. Anti-spoof functionally is
2709 * still supported.
2710 */
2711 val = rd32(hw, GL_MDCK_TX_TDPU);
2712 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2713 wr32(hw, GL_MDCK_TX_TDPU, val);
2714
2715 /* clear things first */
2716 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2717 rd32(hw, PFINT_OICR); /* read to clear */
2718
2719 val = (PFINT_OICR_ECC_ERR_M |
2720 PFINT_OICR_MAL_DETECT_M |
2721 PFINT_OICR_GRST_M |
2722 PFINT_OICR_PCI_EXCEPTION_M |
2723 PFINT_OICR_VFLR_M |
2724 PFINT_OICR_HMC_ERR_M |
2725 PFINT_OICR_PE_PUSH_M |
2726 PFINT_OICR_PE_CRITERR_M);
2727
2728 wr32(hw, PFINT_OICR_ENA, val);
2729
2730 /* SW_ITR_IDX = 0, but don't change INTENA */
2731 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2732 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2733}
2734
2735/**
2736 * ice_misc_intr - misc interrupt handler
2737 * @irq: interrupt number
2738 * @data: pointer to a q_vector
2739 */
2740static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2741{
2742 struct ice_pf *pf = (struct ice_pf *)data;
2743 struct ice_hw *hw = &pf->hw;
2744 irqreturn_t ret = IRQ_NONE;
2745 struct device *dev;
2746 u32 oicr, ena_mask;
2747
2748 dev = ice_pf_to_dev(pf);
2749 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2750 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2751 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
2752
2753 oicr = rd32(hw, PFINT_OICR);
2754 ena_mask = rd32(hw, PFINT_OICR_ENA);
2755
2756 if (oicr & PFINT_OICR_SWINT_M) {
2757 ena_mask &= ~PFINT_OICR_SWINT_M;
2758 pf->sw_int_count++;
2759 }
2760
2761 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2762 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2763 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2764 }
2765 if (oicr & PFINT_OICR_VFLR_M) {
2766 /* disable any further VFLR event notifications */
2767 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2768 u32 reg = rd32(hw, PFINT_OICR_ENA);
2769
2770 reg &= ~PFINT_OICR_VFLR_M;
2771 wr32(hw, PFINT_OICR_ENA, reg);
2772 } else {
2773 ena_mask &= ~PFINT_OICR_VFLR_M;
2774 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2775 }
2776 }
2777
2778 if (oicr & PFINT_OICR_GRST_M) {
2779 u32 reset;
2780
2781 /* we have a reset warning */
2782 ena_mask &= ~PFINT_OICR_GRST_M;
2783 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2784 GLGEN_RSTAT_RESET_TYPE_S;
2785
2786 if (reset == ICE_RESET_CORER)
2787 pf->corer_count++;
2788 else if (reset == ICE_RESET_GLOBR)
2789 pf->globr_count++;
2790 else if (reset == ICE_RESET_EMPR)
2791 pf->empr_count++;
2792 else
2793 dev_dbg(dev, "Invalid reset type %d\n", reset);
2794
2795 /* If a reset cycle isn't already in progress, we set a bit in
2796 * pf->state so that the service task can start a reset/rebuild.
2797 */
2798 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
2799 if (reset == ICE_RESET_CORER)
2800 set_bit(ICE_CORER_RECV, pf->state);
2801 else if (reset == ICE_RESET_GLOBR)
2802 set_bit(ICE_GLOBR_RECV, pf->state);
2803 else
2804 set_bit(ICE_EMPR_RECV, pf->state);
2805
2806 /* There are couple of different bits at play here.
2807 * hw->reset_ongoing indicates whether the hardware is
2808 * in reset. This is set to true when a reset interrupt
2809 * is received and set back to false after the driver
2810 * has determined that the hardware is out of reset.
2811 *
2812 * ICE_RESET_OICR_RECV in pf->state indicates
2813 * that a post reset rebuild is required before the
2814 * driver is operational again. This is set above.
2815 *
2816 * As this is the start of the reset/rebuild cycle, set
2817 * both to indicate that.
2818 */
2819 hw->reset_ongoing = true;
2820 }
2821 }
2822
2823 if (oicr & PFINT_OICR_TSYN_TX_M) {
2824 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
2825 ice_ptp_process_ts(pf);
2826 }
2827
2828 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
2829 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
2830 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
2831
2832 /* Save EVENTs from GTSYN register */
2833 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
2834 GLTSYN_STAT_EVENT1_M |
2835 GLTSYN_STAT_EVENT2_M);
2836 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
2837 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
2838 }
2839
2840#define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
2841 if (oicr & ICE_AUX_CRIT_ERR) {
2842 struct iidc_event *event;
2843
2844 ena_mask &= ~ICE_AUX_CRIT_ERR;
2845 event = kzalloc(sizeof(*event), GFP_KERNEL);
2846 if (event) {
2847 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2848 /* report the entire OICR value to AUX driver */
2849 event->reg = oicr;
2850 ice_send_event_to_aux(pf, event);
2851 kfree(event);
2852 }
2853 }
2854
2855 /* Report any remaining unexpected interrupts */
2856 oicr &= ena_mask;
2857 if (oicr) {
2858 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2859 /* If a critical error is pending there is no choice but to
2860 * reset the device.
2861 */
2862 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
2863 PFINT_OICR_ECC_ERR_M)) {
2864 set_bit(ICE_PFR_REQ, pf->state);
2865 ice_service_task_schedule(pf);
2866 }
2867 }
2868 ret = IRQ_HANDLED;
2869
2870 ice_service_task_schedule(pf);
2871 ice_irq_dynamic_ena(hw, NULL, NULL);
2872
2873 return ret;
2874}
2875
2876/**
2877 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2878 * @hw: pointer to HW structure
2879 */
2880static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2881{
2882 /* disable Admin queue Interrupt causes */
2883 wr32(hw, PFINT_FW_CTL,
2884 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2885
2886 /* disable Mailbox queue Interrupt causes */
2887 wr32(hw, PFINT_MBX_CTL,
2888 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2889
2890 wr32(hw, PFINT_SB_CTL,
2891 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
2892
2893 /* disable Control queue Interrupt causes */
2894 wr32(hw, PFINT_OICR_CTL,
2895 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2896
2897 ice_flush(hw);
2898}
2899
2900/**
2901 * ice_free_irq_msix_misc - Unroll misc vector setup
2902 * @pf: board private structure
2903 */
2904static void ice_free_irq_msix_misc(struct ice_pf *pf)
2905{
2906 struct ice_hw *hw = &pf->hw;
2907
2908 ice_dis_ctrlq_interrupts(hw);
2909
2910 /* disable OICR interrupt */
2911 wr32(hw, PFINT_OICR_ENA, 0);
2912 ice_flush(hw);
2913
2914 if (pf->msix_entries) {
2915 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2916 devm_free_irq(ice_pf_to_dev(pf),
2917 pf->msix_entries[pf->oicr_idx].vector, pf);
2918 }
2919
2920 pf->num_avail_sw_msix += 1;
2921 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2922}
2923
2924/**
2925 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2926 * @hw: pointer to HW structure
2927 * @reg_idx: HW vector index to associate the control queue interrupts with
2928 */
2929static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2930{
2931 u32 val;
2932
2933 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2934 PFINT_OICR_CTL_CAUSE_ENA_M);
2935 wr32(hw, PFINT_OICR_CTL, val);
2936
2937 /* enable Admin queue Interrupt causes */
2938 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2939 PFINT_FW_CTL_CAUSE_ENA_M);
2940 wr32(hw, PFINT_FW_CTL, val);
2941
2942 /* enable Mailbox queue Interrupt causes */
2943 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2944 PFINT_MBX_CTL_CAUSE_ENA_M);
2945 wr32(hw, PFINT_MBX_CTL, val);
2946
2947 /* This enables Sideband queue Interrupt causes */
2948 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
2949 PFINT_SB_CTL_CAUSE_ENA_M);
2950 wr32(hw, PFINT_SB_CTL, val);
2951
2952 ice_flush(hw);
2953}
2954
2955/**
2956 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2957 * @pf: board private structure
2958 *
2959 * This sets up the handler for MSIX 0, which is used to manage the
2960 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2961 * when in MSI or Legacy interrupt mode.
2962 */
2963static int ice_req_irq_msix_misc(struct ice_pf *pf)
2964{
2965 struct device *dev = ice_pf_to_dev(pf);
2966 struct ice_hw *hw = &pf->hw;
2967 int oicr_idx, err = 0;
2968
2969 if (!pf->int_name[0])
2970 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2971 dev_driver_string(dev), dev_name(dev));
2972
2973 /* Do not request IRQ but do enable OICR interrupt since settings are
2974 * lost during reset. Note that this function is called only during
2975 * rebuild path and not while reset is in progress.
2976 */
2977 if (ice_is_reset_in_progress(pf->state))
2978 goto skip_req_irq;
2979
2980 /* reserve one vector in irq_tracker for misc interrupts */
2981 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2982 if (oicr_idx < 0)
2983 return oicr_idx;
2984
2985 pf->num_avail_sw_msix -= 1;
2986 pf->oicr_idx = (u16)oicr_idx;
2987
2988 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2989 ice_misc_intr, 0, pf->int_name, pf);
2990 if (err) {
2991 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2992 pf->int_name, err);
2993 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2994 pf->num_avail_sw_msix += 1;
2995 return err;
2996 }
2997
2998skip_req_irq:
2999 ice_ena_misc_vector(pf);
3000
3001 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3002 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3003 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3004
3005 ice_flush(hw);
3006 ice_irq_dynamic_ena(hw, NULL, NULL);
3007
3008 return 0;
3009}
3010
3011/**
3012 * ice_napi_add - register NAPI handler for the VSI
3013 * @vsi: VSI for which NAPI handler is to be registered
3014 *
3015 * This function is only called in the driver's load path. Registering the NAPI
3016 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3017 * reset/rebuild, etc.)
3018 */
3019static void ice_napi_add(struct ice_vsi *vsi)
3020{
3021 int v_idx;
3022
3023 if (!vsi->netdev)
3024 return;
3025
3026 ice_for_each_q_vector(vsi, v_idx)
3027 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3028 ice_napi_poll, NAPI_POLL_WEIGHT);
3029}
3030
3031/**
3032 * ice_set_ops - set netdev and ethtools ops for the given netdev
3033 * @netdev: netdev instance
3034 */
3035static void ice_set_ops(struct net_device *netdev)
3036{
3037 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3038
3039 if (ice_is_safe_mode(pf)) {
3040 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3041 ice_set_ethtool_safe_mode_ops(netdev);
3042 return;
3043 }
3044
3045 netdev->netdev_ops = &ice_netdev_ops;
3046 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3047 ice_set_ethtool_ops(netdev);
3048}
3049
3050/**
3051 * ice_set_netdev_features - set features for the given netdev
3052 * @netdev: netdev instance
3053 */
3054static void ice_set_netdev_features(struct net_device *netdev)
3055{
3056 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3057 netdev_features_t csumo_features;
3058 netdev_features_t vlano_features;
3059 netdev_features_t dflt_features;
3060 netdev_features_t tso_features;
3061
3062 if (ice_is_safe_mode(pf)) {
3063 /* safe mode */
3064 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3065 netdev->hw_features = netdev->features;
3066 return;
3067 }
3068
3069 dflt_features = NETIF_F_SG |
3070 NETIF_F_HIGHDMA |
3071 NETIF_F_NTUPLE |
3072 NETIF_F_RXHASH;
3073
3074 csumo_features = NETIF_F_RXCSUM |
3075 NETIF_F_IP_CSUM |
3076 NETIF_F_SCTP_CRC |
3077 NETIF_F_IPV6_CSUM;
3078
3079 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3080 NETIF_F_HW_VLAN_CTAG_TX |
3081 NETIF_F_HW_VLAN_CTAG_RX;
3082
3083 tso_features = NETIF_F_TSO |
3084 NETIF_F_TSO_ECN |
3085 NETIF_F_TSO6 |
3086 NETIF_F_GSO_GRE |
3087 NETIF_F_GSO_UDP_TUNNEL |
3088 NETIF_F_GSO_GRE_CSUM |
3089 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3090 NETIF_F_GSO_PARTIAL |
3091 NETIF_F_GSO_IPXIP4 |
3092 NETIF_F_GSO_IPXIP6 |
3093 NETIF_F_GSO_UDP_L4;
3094
3095 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3096 NETIF_F_GSO_GRE_CSUM;
3097 /* set features that user can change */
3098 netdev->hw_features = dflt_features | csumo_features |
3099 vlano_features | tso_features;
3100
3101 /* add support for HW_CSUM on packets with MPLS header */
3102 netdev->mpls_features = NETIF_F_HW_CSUM;
3103
3104 /* enable features */
3105 netdev->features |= netdev->hw_features;
3106 /* encap and VLAN devices inherit default, csumo and tso features */
3107 netdev->hw_enc_features |= dflt_features | csumo_features |
3108 tso_features;
3109 netdev->vlan_features |= dflt_features | csumo_features |
3110 tso_features;
3111}
3112
3113/**
3114 * ice_cfg_netdev - Allocate, configure and register a netdev
3115 * @vsi: the VSI associated with the new netdev
3116 *
3117 * Returns 0 on success, negative value on failure
3118 */
3119static int ice_cfg_netdev(struct ice_vsi *vsi)
3120{
3121 struct ice_netdev_priv *np;
3122 struct net_device *netdev;
3123 u8 mac_addr[ETH_ALEN];
3124
3125 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3126 vsi->alloc_rxq);
3127 if (!netdev)
3128 return -ENOMEM;
3129
3130 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3131 vsi->netdev = netdev;
3132 np = netdev_priv(netdev);
3133 np->vsi = vsi;
3134
3135 ice_set_netdev_features(netdev);
3136
3137 ice_set_ops(netdev);
3138
3139 if (vsi->type == ICE_VSI_PF) {
3140 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3141 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3142 ether_addr_copy(netdev->dev_addr, mac_addr);
3143 ether_addr_copy(netdev->perm_addr, mac_addr);
3144 }
3145
3146 netdev->priv_flags |= IFF_UNICAST_FLT;
3147
3148 /* Setup netdev TC information */
3149 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3150
3151 /* setup watchdog timeout value to be 5 second */
3152 netdev->watchdog_timeo = 5 * HZ;
3153
3154 netdev->min_mtu = ETH_MIN_MTU;
3155 netdev->max_mtu = ICE_MAX_MTU;
3156
3157 return 0;
3158}
3159
3160/**
3161 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3162 * @lut: Lookup table
3163 * @rss_table_size: Lookup table size
3164 * @rss_size: Range of queue number for hashing
3165 */
3166void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3167{
3168 u16 i;
3169
3170 for (i = 0; i < rss_table_size; i++)
3171 lut[i] = i % rss_size;
3172}
3173
3174/**
3175 * ice_pf_vsi_setup - Set up a PF VSI
3176 * @pf: board private structure
3177 * @pi: pointer to the port_info instance
3178 *
3179 * Returns pointer to the successfully allocated VSI software struct
3180 * on success, otherwise returns NULL on failure.
3181 */
3182static struct ice_vsi *
3183ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3184{
3185 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3186}
3187
3188/**
3189 * ice_ctrl_vsi_setup - Set up a control VSI
3190 * @pf: board private structure
3191 * @pi: pointer to the port_info instance
3192 *
3193 * Returns pointer to the successfully allocated VSI software struct
3194 * on success, otherwise returns NULL on failure.
3195 */
3196static struct ice_vsi *
3197ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3198{
3199 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3200}
3201
3202/**
3203 * ice_lb_vsi_setup - Set up a loopback VSI
3204 * @pf: board private structure
3205 * @pi: pointer to the port_info instance
3206 *
3207 * Returns pointer to the successfully allocated VSI software struct
3208 * on success, otherwise returns NULL on failure.
3209 */
3210struct ice_vsi *
3211ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3212{
3213 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3214}
3215
3216/**
3217 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3218 * @netdev: network interface to be adjusted
3219 * @proto: unused protocol
3220 * @vid: VLAN ID to be added
3221 *
3222 * net_device_ops implementation for adding VLAN IDs
3223 */
3224static int
3225ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3226 u16 vid)
3227{
3228 struct ice_netdev_priv *np = netdev_priv(netdev);
3229 struct ice_vsi *vsi = np->vsi;
3230 int ret;
3231
3232 /* VLAN 0 is added by default during load/reset */
3233 if (!vid)
3234 return 0;
3235
3236 /* Enable VLAN pruning when a VLAN other than 0 is added */
3237 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3238 ret = ice_cfg_vlan_pruning(vsi, true, false);
3239 if (ret)
3240 return ret;
3241 }
3242
3243 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3244 * packets aren't pruned by the device's internal switch on Rx
3245 */
3246 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3247 if (!ret)
3248 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3249
3250 return ret;
3251}
3252
3253/**
3254 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3255 * @netdev: network interface to be adjusted
3256 * @proto: unused protocol
3257 * @vid: VLAN ID to be removed
3258 *
3259 * net_device_ops implementation for removing VLAN IDs
3260 */
3261static int
3262ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3263 u16 vid)
3264{
3265 struct ice_netdev_priv *np = netdev_priv(netdev);
3266 struct ice_vsi *vsi = np->vsi;
3267 int ret;
3268
3269 /* don't allow removal of VLAN 0 */
3270 if (!vid)
3271 return 0;
3272
3273 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3274 * information
3275 */
3276 ret = ice_vsi_kill_vlan(vsi, vid);
3277 if (ret)
3278 return ret;
3279
3280 /* Disable pruning when VLAN 0 is the only VLAN rule */
3281 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3282 ret = ice_cfg_vlan_pruning(vsi, false, false);
3283
3284 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3285 return ret;
3286}
3287
3288/**
3289 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3290 * @pf: board private structure
3291 *
3292 * Returns 0 on success, negative value on failure
3293 */
3294static int ice_setup_pf_sw(struct ice_pf *pf)
3295{
3296 struct ice_vsi *vsi;
3297 int status = 0;
3298
3299 if (ice_is_reset_in_progress(pf->state))
3300 return -EBUSY;
3301
3302 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3303 if (!vsi)
3304 return -ENOMEM;
3305
3306 status = ice_cfg_netdev(vsi);
3307 if (status) {
3308 status = -ENODEV;
3309 goto unroll_vsi_setup;
3310 }
3311 /* netdev has to be configured before setting frame size */
3312 ice_vsi_cfg_frame_size(vsi);
3313
3314 /* Setup DCB netlink interface */
3315 ice_dcbnl_setup(vsi);
3316
3317 /* registering the NAPI handler requires both the queues and
3318 * netdev to be created, which are done in ice_pf_vsi_setup()
3319 * and ice_cfg_netdev() respectively
3320 */
3321 ice_napi_add(vsi);
3322
3323 status = ice_set_cpu_rx_rmap(vsi);
3324 if (status) {
3325 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3326 vsi->vsi_num, status);
3327 status = -EINVAL;
3328 goto unroll_napi_add;
3329 }
3330 status = ice_init_mac_fltr(pf);
3331 if (status)
3332 goto free_cpu_rx_map;
3333
3334 return status;
3335
3336free_cpu_rx_map:
3337 ice_free_cpu_rx_rmap(vsi);
3338
3339unroll_napi_add:
3340 if (vsi) {
3341 ice_napi_del(vsi);
3342 if (vsi->netdev) {
3343 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3344 free_netdev(vsi->netdev);
3345 vsi->netdev = NULL;
3346 }
3347 }
3348
3349unroll_vsi_setup:
3350 ice_vsi_release(vsi);
3351 return status;
3352}
3353
3354/**
3355 * ice_get_avail_q_count - Get count of queues in use
3356 * @pf_qmap: bitmap to get queue use count from
3357 * @lock: pointer to a mutex that protects access to pf_qmap
3358 * @size: size of the bitmap
3359 */
3360static u16
3361ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3362{
3363 unsigned long bit;
3364 u16 count = 0;
3365
3366 mutex_lock(lock);
3367 for_each_clear_bit(bit, pf_qmap, size)
3368 count++;
3369 mutex_unlock(lock);
3370
3371 return count;
3372}
3373
3374/**
3375 * ice_get_avail_txq_count - Get count of Tx queues in use
3376 * @pf: pointer to an ice_pf instance
3377 */
3378u16 ice_get_avail_txq_count(struct ice_pf *pf)
3379{
3380 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3381 pf->max_pf_txqs);
3382}
3383
3384/**
3385 * ice_get_avail_rxq_count - Get count of Rx queues in use
3386 * @pf: pointer to an ice_pf instance
3387 */
3388u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3389{
3390 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3391 pf->max_pf_rxqs);
3392}
3393
3394/**
3395 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3396 * @pf: board private structure to initialize
3397 */
3398static void ice_deinit_pf(struct ice_pf *pf)
3399{
3400 ice_service_task_stop(pf);
3401 mutex_destroy(&pf->sw_mutex);
3402 mutex_destroy(&pf->tc_mutex);
3403 mutex_destroy(&pf->avail_q_mutex);
3404
3405 if (pf->avail_txqs) {
3406 bitmap_free(pf->avail_txqs);
3407 pf->avail_txqs = NULL;
3408 }
3409
3410 if (pf->avail_rxqs) {
3411 bitmap_free(pf->avail_rxqs);
3412 pf->avail_rxqs = NULL;
3413 }
3414
3415 if (pf->ptp.clock)
3416 ptp_clock_unregister(pf->ptp.clock);
3417}
3418
3419/**
3420 * ice_set_pf_caps - set PFs capability flags
3421 * @pf: pointer to the PF instance
3422 */
3423static void ice_set_pf_caps(struct ice_pf *pf)
3424{
3425 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3426
3427 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3428 clear_bit(ICE_FLAG_AUX_ENA, pf->flags);
3429 if (func_caps->common_cap.rdma) {
3430 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3431 set_bit(ICE_FLAG_AUX_ENA, pf->flags);
3432 }
3433 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3434 if (func_caps->common_cap.dcb)
3435 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3436 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3437 if (func_caps->common_cap.sr_iov_1_1) {
3438 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3439 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3440 ICE_MAX_VF_COUNT);
3441 }
3442 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3443 if (func_caps->common_cap.rss_table_size)
3444 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3445
3446 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3447 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3448 u16 unused;
3449
3450 /* ctrl_vsi_idx will be set to a valid value when flow director
3451 * is setup by ice_init_fdir
3452 */
3453 pf->ctrl_vsi_idx = ICE_NO_VSI;
3454 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3455 /* force guaranteed filter pool for PF */
3456 ice_alloc_fd_guar_item(&pf->hw, &unused,
3457 func_caps->fd_fltr_guar);
3458 /* force shared filter pool for PF */
3459 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3460 func_caps->fd_fltr_best_effort);
3461 }
3462
3463 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3464 if (func_caps->common_cap.ieee_1588)
3465 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3466
3467 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3468 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3469}
3470
3471/**
3472 * ice_init_pf - Initialize general software structures (struct ice_pf)
3473 * @pf: board private structure to initialize
3474 */
3475static int ice_init_pf(struct ice_pf *pf)
3476{
3477 ice_set_pf_caps(pf);
3478
3479 mutex_init(&pf->sw_mutex);
3480 mutex_init(&pf->tc_mutex);
3481
3482 INIT_HLIST_HEAD(&pf->aq_wait_list);
3483 spin_lock_init(&pf->aq_wait_lock);
3484 init_waitqueue_head(&pf->aq_wait_queue);
3485
3486 init_waitqueue_head(&pf->reset_wait_queue);
3487
3488 /* setup service timer and periodic service task */
3489 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3490 pf->serv_tmr_period = HZ;
3491 INIT_WORK(&pf->serv_task, ice_service_task);
3492 clear_bit(ICE_SERVICE_SCHED, pf->state);
3493
3494 mutex_init(&pf->avail_q_mutex);
3495 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3496 if (!pf->avail_txqs)
3497 return -ENOMEM;
3498
3499 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3500 if (!pf->avail_rxqs) {
3501 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3502 pf->avail_txqs = NULL;
3503 return -ENOMEM;
3504 }
3505
3506 return 0;
3507}
3508
3509/**
3510 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3511 * @pf: board private structure
3512 *
3513 * compute the number of MSIX vectors required (v_budget) and request from
3514 * the OS. Return the number of vectors reserved or negative on failure
3515 */
3516static int ice_ena_msix_range(struct ice_pf *pf)
3517{
3518 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3519 struct device *dev = ice_pf_to_dev(pf);
3520 int needed, err, i;
3521
3522 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3523 num_cpus = num_online_cpus();
3524
3525 /* reserve for LAN miscellaneous handler */
3526 needed = ICE_MIN_LAN_OICR_MSIX;
3527 if (v_left < needed)
3528 goto no_hw_vecs_left_err;
3529 v_budget += needed;
3530 v_left -= needed;
3531
3532 /* reserve for flow director */
3533 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3534 needed = ICE_FDIR_MSIX;
3535 if (v_left < needed)
3536 goto no_hw_vecs_left_err;
3537 v_budget += needed;
3538 v_left -= needed;
3539 }
3540
3541 /* total used for non-traffic vectors */
3542 v_other = v_budget;
3543
3544 /* reserve vectors for LAN traffic */
3545 needed = num_cpus;
3546 if (v_left < needed)
3547 goto no_hw_vecs_left_err;
3548 pf->num_lan_msix = needed;
3549 v_budget += needed;
3550 v_left -= needed;
3551
3552 /* reserve vectors for RDMA auxiliary driver */
3553 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3554 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3555 if (v_left < needed)
3556 goto no_hw_vecs_left_err;
3557 pf->num_rdma_msix = needed;
3558 v_budget += needed;
3559 v_left -= needed;
3560 }
3561
3562 pf->msix_entries = devm_kcalloc(dev, v_budget,
3563 sizeof(*pf->msix_entries), GFP_KERNEL);
3564 if (!pf->msix_entries) {
3565 err = -ENOMEM;
3566 goto exit_err;
3567 }
3568
3569 for (i = 0; i < v_budget; i++)
3570 pf->msix_entries[i].entry = i;
3571
3572 /* actually reserve the vectors */
3573 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3574 ICE_MIN_MSIX, v_budget);
3575 if (v_actual < 0) {
3576 dev_err(dev, "unable to reserve MSI-X vectors\n");
3577 err = v_actual;
3578 goto msix_err;
3579 }
3580
3581 if (v_actual < v_budget) {
3582 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3583 v_budget, v_actual);
3584
3585 if (v_actual < ICE_MIN_MSIX) {
3586 /* error if we can't get minimum vectors */
3587 pci_disable_msix(pf->pdev);
3588 err = -ERANGE;
3589 goto msix_err;
3590 } else {
3591 int v_remain = v_actual - v_other;
3592 int v_rdma = 0, v_min_rdma = 0;
3593
3594 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3595 /* Need at least 1 interrupt in addition to
3596 * AEQ MSIX
3597 */
3598 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3599 v_min_rdma = ICE_MIN_RDMA_MSIX;
3600 }
3601
3602 if (v_actual == ICE_MIN_MSIX ||
3603 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
3604 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
3605 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3606
3607 pf->num_rdma_msix = 0;
3608 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3609 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3610 (v_remain - v_rdma < v_rdma)) {
3611 /* Support minimum RDMA and give remaining
3612 * vectors to LAN MSIX
3613 */
3614 pf->num_rdma_msix = v_min_rdma;
3615 pf->num_lan_msix = v_remain - v_min_rdma;
3616 } else {
3617 /* Split remaining MSIX with RDMA after
3618 * accounting for AEQ MSIX
3619 */
3620 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3621 ICE_RDMA_NUM_AEQ_MSIX;
3622 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3623 }
3624
3625 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3626 pf->num_lan_msix);
3627
3628 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
3629 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
3630 pf->num_rdma_msix);
3631 }
3632 }
3633
3634 return v_actual;
3635
3636msix_err:
3637 devm_kfree(dev, pf->msix_entries);
3638 goto exit_err;
3639
3640no_hw_vecs_left_err:
3641 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3642 needed, v_left);
3643 err = -ERANGE;
3644exit_err:
3645 pf->num_rdma_msix = 0;
3646 pf->num_lan_msix = 0;
3647 return err;
3648}
3649
3650/**
3651 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3652 * @pf: board private structure
3653 */
3654static void ice_dis_msix(struct ice_pf *pf)
3655{
3656 pci_disable_msix(pf->pdev);
3657 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3658 pf->msix_entries = NULL;
3659}
3660
3661/**
3662 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3663 * @pf: board private structure
3664 */
3665static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3666{
3667 ice_dis_msix(pf);
3668
3669 if (pf->irq_tracker) {
3670 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3671 pf->irq_tracker = NULL;
3672 }
3673}
3674
3675/**
3676 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3677 * @pf: board private structure to initialize
3678 */
3679static int ice_init_interrupt_scheme(struct ice_pf *pf)
3680{
3681 int vectors;
3682
3683 vectors = ice_ena_msix_range(pf);
3684
3685 if (vectors < 0)
3686 return vectors;
3687
3688 /* set up vector assignment tracking */
3689 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3690 struct_size(pf->irq_tracker, list, vectors),
3691 GFP_KERNEL);
3692 if (!pf->irq_tracker) {
3693 ice_dis_msix(pf);
3694 return -ENOMEM;
3695 }
3696
3697 /* populate SW interrupts pool with number of OS granted IRQs. */
3698 pf->num_avail_sw_msix = (u16)vectors;
3699 pf->irq_tracker->num_entries = (u16)vectors;
3700 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3701
3702 return 0;
3703}
3704
3705/**
3706 * ice_is_wol_supported - check if WoL is supported
3707 * @hw: pointer to hardware info
3708 *
3709 * Check if WoL is supported based on the HW configuration.
3710 * Returns true if NVM supports and enables WoL for this port, false otherwise
3711 */
3712bool ice_is_wol_supported(struct ice_hw *hw)
3713{
3714 u16 wol_ctrl;
3715
3716 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3717 * word) indicates WoL is not supported on the corresponding PF ID.
3718 */
3719 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3720 return false;
3721
3722 return !(BIT(hw->port_info->lport) & wol_ctrl);
3723}
3724
3725/**
3726 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3727 * @vsi: VSI being changed
3728 * @new_rx: new number of Rx queues
3729 * @new_tx: new number of Tx queues
3730 *
3731 * Only change the number of queues if new_tx, or new_rx is non-0.
3732 *
3733 * Returns 0 on success.
3734 */
3735int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3736{
3737 struct ice_pf *pf = vsi->back;
3738 int err = 0, timeout = 50;
3739
3740 if (!new_rx && !new_tx)
3741 return -EINVAL;
3742
3743 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3744 timeout--;
3745 if (!timeout)
3746 return -EBUSY;
3747 usleep_range(1000, 2000);
3748 }
3749
3750 if (new_tx)
3751 vsi->req_txq = (u16)new_tx;
3752 if (new_rx)
3753 vsi->req_rxq = (u16)new_rx;
3754
3755 /* set for the next time the netdev is started */
3756 if (!netif_running(vsi->netdev)) {
3757 ice_vsi_rebuild(vsi, false);
3758 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3759 goto done;
3760 }
3761
3762 ice_vsi_close(vsi);
3763 ice_vsi_rebuild(vsi, false);
3764 ice_pf_dcb_recfg(pf);
3765 ice_vsi_open(vsi);
3766done:
3767 clear_bit(ICE_CFG_BUSY, pf->state);
3768 return err;
3769}
3770
3771/**
3772 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3773 * @pf: PF to configure
3774 *
3775 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3776 * VSI can still Tx/Rx VLAN tagged packets.
3777 */
3778static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3779{
3780 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3781 struct ice_vsi_ctx *ctxt;
3782 enum ice_status status;
3783 struct ice_hw *hw;
3784
3785 if (!vsi)
3786 return;
3787
3788 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3789 if (!ctxt)
3790 return;
3791
3792 hw = &pf->hw;
3793 ctxt->info = vsi->info;
3794
3795 ctxt->info.valid_sections =
3796 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3797 ICE_AQ_VSI_PROP_SECURITY_VALID |
3798 ICE_AQ_VSI_PROP_SW_VALID);
3799
3800 /* disable VLAN anti-spoof */
3801 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3802 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3803
3804 /* disable VLAN pruning and keep all other settings */
3805 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3806
3807 /* allow all VLANs on Tx and don't strip on Rx */
3808 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3809 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3810
3811 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3812 if (status) {
3813 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3814 ice_stat_str(status),
3815 ice_aq_str(hw->adminq.sq_last_status));
3816 } else {
3817 vsi->info.sec_flags = ctxt->info.sec_flags;
3818 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3819 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3820 }
3821
3822 kfree(ctxt);
3823}
3824
3825/**
3826 * ice_log_pkg_init - log result of DDP package load
3827 * @hw: pointer to hardware info
3828 * @status: status of package load
3829 */
3830static void
3831ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3832{
3833 struct ice_pf *pf = (struct ice_pf *)hw->back;
3834 struct device *dev = ice_pf_to_dev(pf);
3835
3836 switch (*status) {
3837 case ICE_SUCCESS:
3838 /* The package download AdminQ command returned success because
3839 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3840 * already a package loaded on the device.
3841 */
3842 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3843 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3844 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3845 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3846 !memcmp(hw->pkg_name, hw->active_pkg_name,
3847 sizeof(hw->pkg_name))) {
3848 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3849 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3850 hw->active_pkg_name,
3851 hw->active_pkg_ver.major,
3852 hw->active_pkg_ver.minor,
3853 hw->active_pkg_ver.update,
3854 hw->active_pkg_ver.draft);
3855 else
3856 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3857 hw->active_pkg_name,
3858 hw->active_pkg_ver.major,
3859 hw->active_pkg_ver.minor,
3860 hw->active_pkg_ver.update,
3861 hw->active_pkg_ver.draft);
3862 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3863 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3864 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",
3865 hw->active_pkg_name,
3866 hw->active_pkg_ver.major,
3867 hw->active_pkg_ver.minor,
3868 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3869 *status = ICE_ERR_NOT_SUPPORTED;
3870 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3871 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3872 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",
3873 hw->active_pkg_name,
3874 hw->active_pkg_ver.major,
3875 hw->active_pkg_ver.minor,
3876 hw->active_pkg_ver.update,
3877 hw->active_pkg_ver.draft,
3878 hw->pkg_name,
3879 hw->pkg_ver.major,
3880 hw->pkg_ver.minor,
3881 hw->pkg_ver.update,
3882 hw->pkg_ver.draft);
3883 } else {
3884 dev_err(dev, "An unknown error occurred when loading the DDP package, please reboot the system. If the problem persists, update the NVM. Entering Safe Mode.\n");
3885 *status = ICE_ERR_NOT_SUPPORTED;
3886 }
3887 break;
3888 case ICE_ERR_FW_DDP_MISMATCH:
3889 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");
3890 break;
3891 case ICE_ERR_BUF_TOO_SHORT:
3892 case ICE_ERR_CFG:
3893 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3894 break;
3895 case ICE_ERR_NOT_SUPPORTED:
3896 /* Package File version not supported */
3897 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3898 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3899 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3900 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3901 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3902 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3903 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3904 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",
3905 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3906 break;
3907 case ICE_ERR_AQ_ERROR:
3908 switch (hw->pkg_dwnld_status) {
3909 case ICE_AQ_RC_ENOSEC:
3910 case ICE_AQ_RC_EBADSIG:
3911 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");
3912 return;
3913 case ICE_AQ_RC_ESVN:
3914 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");
3915 return;
3916 case ICE_AQ_RC_EBADMAN:
3917 case ICE_AQ_RC_EBADBUF:
3918 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3919 /* poll for reset to complete */
3920 if (ice_check_reset(hw))
3921 dev_err(dev, "Error resetting device. Please reload the driver\n");
3922 return;
3923 default:
3924 break;
3925 }
3926 fallthrough;
3927 default:
3928 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3929 *status);
3930 break;
3931 }
3932}
3933
3934/**
3935 * ice_load_pkg - load/reload the DDP Package file
3936 * @firmware: firmware structure when firmware requested or NULL for reload
3937 * @pf: pointer to the PF instance
3938 *
3939 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3940 * initialize HW tables.
3941 */
3942static void
3943ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3944{
3945 enum ice_status status = ICE_ERR_PARAM;
3946 struct device *dev = ice_pf_to_dev(pf);
3947 struct ice_hw *hw = &pf->hw;
3948
3949 /* Load DDP Package */
3950 if (firmware && !hw->pkg_copy) {
3951 status = ice_copy_and_init_pkg(hw, firmware->data,
3952 firmware->size);
3953 ice_log_pkg_init(hw, &status);
3954 } else if (!firmware && hw->pkg_copy) {
3955 /* Reload package during rebuild after CORER/GLOBR reset */
3956 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3957 ice_log_pkg_init(hw, &status);
3958 } else {
3959 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3960 }
3961
3962 if (status) {
3963 /* Safe Mode */
3964 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3965 return;
3966 }
3967
3968 /* Successful download package is the precondition for advanced
3969 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3970 */
3971 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3972}
3973
3974/**
3975 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3976 * @pf: pointer to the PF structure
3977 *
3978 * There is no error returned here because the driver should be able to handle
3979 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3980 * specifically with Tx.
3981 */
3982static void ice_verify_cacheline_size(struct ice_pf *pf)
3983{
3984 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3985 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3986 ICE_CACHE_LINE_BYTES);
3987}
3988
3989/**
3990 * ice_send_version - update firmware with driver version
3991 * @pf: PF struct
3992 *
3993 * Returns ICE_SUCCESS on success, else error code
3994 */
3995static enum ice_status ice_send_version(struct ice_pf *pf)
3996{
3997 struct ice_driver_ver dv;
3998
3999 dv.major_ver = 0xff;
4000 dv.minor_ver = 0xff;
4001 dv.build_ver = 0xff;
4002 dv.subbuild_ver = 0;
4003 strscpy((char *)dv.driver_string, UTS_RELEASE,
4004 sizeof(dv.driver_string));
4005 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4006}
4007
4008/**
4009 * ice_init_fdir - Initialize flow director VSI and configuration
4010 * @pf: pointer to the PF instance
4011 *
4012 * returns 0 on success, negative on error
4013 */
4014static int ice_init_fdir(struct ice_pf *pf)
4015{
4016 struct device *dev = ice_pf_to_dev(pf);
4017 struct ice_vsi *ctrl_vsi;
4018 int err;
4019
4020 /* Side Band Flow Director needs to have a control VSI.
4021 * Allocate it and store it in the PF.
4022 */
4023 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4024 if (!ctrl_vsi) {
4025 dev_dbg(dev, "could not create control VSI\n");
4026 return -ENOMEM;
4027 }
4028
4029 err = ice_vsi_open_ctrl(ctrl_vsi);
4030 if (err) {
4031 dev_dbg(dev, "could not open control VSI\n");
4032 goto err_vsi_open;
4033 }
4034
4035 mutex_init(&pf->hw.fdir_fltr_lock);
4036
4037 err = ice_fdir_create_dflt_rules(pf);
4038 if (err)
4039 goto err_fdir_rule;
4040
4041 return 0;
4042
4043err_fdir_rule:
4044 ice_fdir_release_flows(&pf->hw);
4045 ice_vsi_close(ctrl_vsi);
4046err_vsi_open:
4047 ice_vsi_release(ctrl_vsi);
4048 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4049 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4050 pf->ctrl_vsi_idx = ICE_NO_VSI;
4051 }
4052 return err;
4053}
4054
4055/**
4056 * ice_get_opt_fw_name - return optional firmware file name or NULL
4057 * @pf: pointer to the PF instance
4058 */
4059static char *ice_get_opt_fw_name(struct ice_pf *pf)
4060{
4061 /* Optional firmware name same as default with additional dash
4062 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4063 */
4064 struct pci_dev *pdev = pf->pdev;
4065 char *opt_fw_filename;
4066 u64 dsn;
4067
4068 /* Determine the name of the optional file using the DSN (two
4069 * dwords following the start of the DSN Capability).
4070 */
4071 dsn = pci_get_dsn(pdev);
4072 if (!dsn)
4073 return NULL;
4074
4075 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4076 if (!opt_fw_filename)
4077 return NULL;
4078
4079 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4080 ICE_DDP_PKG_PATH, dsn);
4081
4082 return opt_fw_filename;
4083}
4084
4085/**
4086 * ice_request_fw - Device initialization routine
4087 * @pf: pointer to the PF instance
4088 */
4089static void ice_request_fw(struct ice_pf *pf)
4090{
4091 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4092 const struct firmware *firmware = NULL;
4093 struct device *dev = ice_pf_to_dev(pf);
4094 int err = 0;
4095
4096 /* optional device-specific DDP (if present) overrides the default DDP
4097 * package file. kernel logs a debug message if the file doesn't exist,
4098 * and warning messages for other errors.
4099 */
4100 if (opt_fw_filename) {
4101 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4102 if (err) {
4103 kfree(opt_fw_filename);
4104 goto dflt_pkg_load;
4105 }
4106
4107 /* request for firmware was successful. Download to device */
4108 ice_load_pkg(firmware, pf);
4109 kfree(opt_fw_filename);
4110 release_firmware(firmware);
4111 return;
4112 }
4113
4114dflt_pkg_load:
4115 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4116 if (err) {
4117 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4118 return;
4119 }
4120
4121 /* request for firmware was successful. Download to device */
4122 ice_load_pkg(firmware, pf);
4123 release_firmware(firmware);
4124}
4125
4126/**
4127 * ice_print_wake_reason - show the wake up cause in the log
4128 * @pf: pointer to the PF struct
4129 */
4130static void ice_print_wake_reason(struct ice_pf *pf)
4131{
4132 u32 wus = pf->wakeup_reason;
4133 const char *wake_str;
4134
4135 /* if no wake event, nothing to print */
4136 if (!wus)
4137 return;
4138
4139 if (wus & PFPM_WUS_LNKC_M)
4140 wake_str = "Link\n";
4141 else if (wus & PFPM_WUS_MAG_M)
4142 wake_str = "Magic Packet\n";
4143 else if (wus & PFPM_WUS_MNG_M)
4144 wake_str = "Management\n";
4145 else if (wus & PFPM_WUS_FW_RST_WK_M)
4146 wake_str = "Firmware Reset\n";
4147 else
4148 wake_str = "Unknown\n";
4149
4150 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4151}
4152
4153/**
4154 * ice_register_netdev - register netdev and devlink port
4155 * @pf: pointer to the PF struct
4156 */
4157static int ice_register_netdev(struct ice_pf *pf)
4158{
4159 struct ice_vsi *vsi;
4160 int err = 0;
4161
4162 vsi = ice_get_main_vsi(pf);
4163 if (!vsi || !vsi->netdev)
4164 return -EIO;
4165
4166 err = register_netdev(vsi->netdev);
4167 if (err)
4168 goto err_register_netdev;
4169
4170 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4171 netif_carrier_off(vsi->netdev);
4172 netif_tx_stop_all_queues(vsi->netdev);
4173 err = ice_devlink_create_port(vsi);
4174 if (err)
4175 goto err_devlink_create;
4176
4177 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
4178
4179 return 0;
4180err_devlink_create:
4181 unregister_netdev(vsi->netdev);
4182 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4183err_register_netdev:
4184 free_netdev(vsi->netdev);
4185 vsi->netdev = NULL;
4186 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4187 return err;
4188}
4189
4190/**
4191 * ice_probe - Device initialization routine
4192 * @pdev: PCI device information struct
4193 * @ent: entry in ice_pci_tbl
4194 *
4195 * Returns 0 on success, negative on failure
4196 */
4197static int
4198ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4199{
4200 struct device *dev = &pdev->dev;
4201 struct ice_pf *pf;
4202 struct ice_hw *hw;
4203 int i, err;
4204
4205 if (pdev->is_virtfn) {
4206 dev_err(dev, "can't probe a virtual function\n");
4207 return -EINVAL;
4208 }
4209
4210 /* this driver uses devres, see
4211 * Documentation/driver-api/driver-model/devres.rst
4212 */
4213 err = pcim_enable_device(pdev);
4214 if (err)
4215 return err;
4216
4217 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4218 if (err) {
4219 dev_err(dev, "BAR0 I/O map error %d\n", err);
4220 return err;
4221 }
4222
4223 pf = ice_allocate_pf(dev);
4224 if (!pf)
4225 return -ENOMEM;
4226
4227 /* initialize Auxiliary index to invalid value */
4228 pf->aux_idx = -1;
4229
4230 /* set up for high or low DMA */
4231 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4232 if (err)
4233 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4234 if (err) {
4235 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4236 return err;
4237 }
4238
4239 pci_enable_pcie_error_reporting(pdev);
4240 pci_set_master(pdev);
4241
4242 pf->pdev = pdev;
4243 pci_set_drvdata(pdev, pf);
4244 set_bit(ICE_DOWN, pf->state);
4245 /* Disable service task until DOWN bit is cleared */
4246 set_bit(ICE_SERVICE_DIS, pf->state);
4247
4248 hw = &pf->hw;
4249 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4250 pci_save_state(pdev);
4251
4252 hw->back = pf;
4253 hw->vendor_id = pdev->vendor;
4254 hw->device_id = pdev->device;
4255 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4256 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4257 hw->subsystem_device_id = pdev->subsystem_device;
4258 hw->bus.device = PCI_SLOT(pdev->devfn);
4259 hw->bus.func = PCI_FUNC(pdev->devfn);
4260 ice_set_ctrlq_len(hw);
4261
4262 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4263
4264 err = ice_devlink_register(pf);
4265 if (err) {
4266 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4267 goto err_exit_unroll;
4268 }
4269
4270#ifndef CONFIG_DYNAMIC_DEBUG
4271 if (debug < -1)
4272 hw->debug_mask = debug;
4273#endif
4274
4275 err = ice_init_hw(hw);
4276 if (err) {
4277 dev_err(dev, "ice_init_hw failed: %d\n", err);
4278 err = -EIO;
4279 goto err_exit_unroll;
4280 }
4281
4282 ice_request_fw(pf);
4283
4284 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4285 * set in pf->state, which will cause ice_is_safe_mode to return
4286 * true
4287 */
4288 if (ice_is_safe_mode(pf)) {
4289 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4290 /* we already got function/device capabilities but these don't
4291 * reflect what the driver needs to do in safe mode. Instead of
4292 * adding conditional logic everywhere to ignore these
4293 * device/function capabilities, override them.
4294 */
4295 ice_set_safe_mode_caps(hw);
4296 }
4297
4298 err = ice_init_pf(pf);
4299 if (err) {
4300 dev_err(dev, "ice_init_pf failed: %d\n", err);
4301 goto err_init_pf_unroll;
4302 }
4303
4304 ice_devlink_init_regions(pf);
4305
4306 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4307 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4308 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4309 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4310 i = 0;
4311 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4312 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4313 pf->hw.tnl.valid_count[TNL_VXLAN];
4314 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4315 UDP_TUNNEL_TYPE_VXLAN;
4316 i++;
4317 }
4318 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4319 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4320 pf->hw.tnl.valid_count[TNL_GENEVE];
4321 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4322 UDP_TUNNEL_TYPE_GENEVE;
4323 i++;
4324 }
4325
4326 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4327 if (!pf->num_alloc_vsi) {
4328 err = -EIO;
4329 goto err_init_pf_unroll;
4330 }
4331 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4332 dev_warn(&pf->pdev->dev,
4333 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4334 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4335 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4336 }
4337
4338 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4339 GFP_KERNEL);
4340 if (!pf->vsi) {
4341 err = -ENOMEM;
4342 goto err_init_pf_unroll;
4343 }
4344
4345 err = ice_init_interrupt_scheme(pf);
4346 if (err) {
4347 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4348 err = -EIO;
4349 goto err_init_vsi_unroll;
4350 }
4351
4352 /* In case of MSIX we are going to setup the misc vector right here
4353 * to handle admin queue events etc. In case of legacy and MSI
4354 * the misc functionality and queue processing is combined in
4355 * the same vector and that gets setup at open.
4356 */
4357 err = ice_req_irq_msix_misc(pf);
4358 if (err) {
4359 dev_err(dev, "setup of misc vector failed: %d\n", err);
4360 goto err_init_interrupt_unroll;
4361 }
4362
4363 /* create switch struct for the switch element created by FW on boot */
4364 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4365 if (!pf->first_sw) {
4366 err = -ENOMEM;
4367 goto err_msix_misc_unroll;
4368 }
4369
4370 if (hw->evb_veb)
4371 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4372 else
4373 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4374
4375 pf->first_sw->pf = pf;
4376
4377 /* record the sw_id available for later use */
4378 pf->first_sw->sw_id = hw->port_info->sw_id;
4379
4380 err = ice_setup_pf_sw(pf);
4381 if (err) {
4382 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4383 goto err_alloc_sw_unroll;
4384 }
4385
4386 clear_bit(ICE_SERVICE_DIS, pf->state);
4387
4388 /* tell the firmware we are up */
4389 err = ice_send_version(pf);
4390 if (err) {
4391 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4392 UTS_RELEASE, err);
4393 goto err_send_version_unroll;
4394 }
4395
4396 /* since everything is good, start the service timer */
4397 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4398
4399 err = ice_init_link_events(pf->hw.port_info);
4400 if (err) {
4401 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4402 goto err_send_version_unroll;
4403 }
4404
4405 /* not a fatal error if this fails */
4406 err = ice_init_nvm_phy_type(pf->hw.port_info);
4407 if (err)
4408 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4409
4410 /* not a fatal error if this fails */
4411 err = ice_update_link_info(pf->hw.port_info);
4412 if (err)
4413 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4414
4415 ice_init_link_dflt_override(pf->hw.port_info);
4416
4417 ice_check_module_power(pf, pf->hw.port_info->phy.link_info.link_cfg_err);
4418
4419 /* if media available, initialize PHY settings */
4420 if (pf->hw.port_info->phy.link_info.link_info &
4421 ICE_AQ_MEDIA_AVAILABLE) {
4422 /* not a fatal error if this fails */
4423 err = ice_init_phy_user_cfg(pf->hw.port_info);
4424 if (err)
4425 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4426
4427 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4428 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4429
4430 if (vsi)
4431 ice_configure_phy(vsi);
4432 }
4433 } else {
4434 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4435 }
4436
4437 ice_verify_cacheline_size(pf);
4438
4439 /* Save wakeup reason register for later use */
4440 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4441
4442 /* check for a power management event */
4443 ice_print_wake_reason(pf);
4444
4445 /* clear wake status, all bits */
4446 wr32(hw, PFPM_WUS, U32_MAX);
4447
4448 /* Disable WoL at init, wait for user to enable */
4449 device_set_wakeup_enable(dev, false);
4450
4451 if (ice_is_safe_mode(pf)) {
4452 ice_set_safe_mode_vlan_cfg(pf);
4453 goto probe_done;
4454 }
4455
4456 /* initialize DDP driven features */
4457 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4458 ice_ptp_init(pf);
4459
4460 /* Note: Flow director init failure is non-fatal to load */
4461 if (ice_init_fdir(pf))
4462 dev_err(dev, "could not initialize flow director\n");
4463
4464 /* Note: DCB init failure is non-fatal to load */
4465 if (ice_init_pf_dcb(pf, false)) {
4466 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4467 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4468 } else {
4469 ice_cfg_lldp_mib_change(&pf->hw, true);
4470 }
4471
4472 if (ice_init_lag(pf))
4473 dev_warn(dev, "Failed to init link aggregation support\n");
4474
4475 /* print PCI link speed and width */
4476 pcie_print_link_status(pf->pdev);
4477
4478probe_done:
4479 err = ice_register_netdev(pf);
4480 if (err)
4481 goto err_netdev_reg;
4482
4483 /* ready to go, so clear down state bit */
4484 clear_bit(ICE_DOWN, pf->state);
4485 if (ice_is_aux_ena(pf)) {
4486 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4487 if (pf->aux_idx < 0) {
4488 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4489 err = -ENOMEM;
4490 goto err_netdev_reg;
4491 }
4492
4493 err = ice_init_rdma(pf);
4494 if (err) {
4495 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4496 err = -EIO;
4497 goto err_init_aux_unroll;
4498 }
4499 } else {
4500 dev_warn(dev, "RDMA is not supported on this device\n");
4501 }
4502
4503 return 0;
4504
4505err_init_aux_unroll:
4506 pf->adev = NULL;
4507 ida_free(&ice_aux_ida, pf->aux_idx);
4508err_netdev_reg:
4509err_send_version_unroll:
4510 ice_vsi_release_all(pf);
4511err_alloc_sw_unroll:
4512 set_bit(ICE_SERVICE_DIS, pf->state);
4513 set_bit(ICE_DOWN, pf->state);
4514 devm_kfree(dev, pf->first_sw);
4515err_msix_misc_unroll:
4516 ice_free_irq_msix_misc(pf);
4517err_init_interrupt_unroll:
4518 ice_clear_interrupt_scheme(pf);
4519err_init_vsi_unroll:
4520 devm_kfree(dev, pf->vsi);
4521err_init_pf_unroll:
4522 ice_deinit_pf(pf);
4523 ice_devlink_destroy_regions(pf);
4524 ice_deinit_hw(hw);
4525err_exit_unroll:
4526 ice_devlink_unregister(pf);
4527 pci_disable_pcie_error_reporting(pdev);
4528 pci_disable_device(pdev);
4529 return err;
4530}
4531
4532/**
4533 * ice_set_wake - enable or disable Wake on LAN
4534 * @pf: pointer to the PF struct
4535 *
4536 * Simple helper for WoL control
4537 */
4538static void ice_set_wake(struct ice_pf *pf)
4539{
4540 struct ice_hw *hw = &pf->hw;
4541 bool wol = pf->wol_ena;
4542
4543 /* clear wake state, otherwise new wake events won't fire */
4544 wr32(hw, PFPM_WUS, U32_MAX);
4545
4546 /* enable / disable APM wake up, no RMW needed */
4547 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4548
4549 /* set magic packet filter enabled */
4550 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4551}
4552
4553/**
4554 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4555 * @pf: pointer to the PF struct
4556 *
4557 * Issue firmware command to enable multicast magic wake, making
4558 * sure that any locally administered address (LAA) is used for
4559 * wake, and that PF reset doesn't undo the LAA.
4560 */
4561static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4562{
4563 struct device *dev = ice_pf_to_dev(pf);
4564 struct ice_hw *hw = &pf->hw;
4565 enum ice_status status;
4566 u8 mac_addr[ETH_ALEN];
4567 struct ice_vsi *vsi;
4568 u8 flags;
4569
4570 if (!pf->wol_ena)
4571 return;
4572
4573 vsi = ice_get_main_vsi(pf);
4574 if (!vsi)
4575 return;
4576
4577 /* Get current MAC address in case it's an LAA */
4578 if (vsi->netdev)
4579 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4580 else
4581 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4582
4583 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4584 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4585 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4586
4587 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4588 if (status)
4589 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4590 ice_stat_str(status),
4591 ice_aq_str(hw->adminq.sq_last_status));
4592}
4593
4594/**
4595 * ice_remove - Device removal routine
4596 * @pdev: PCI device information struct
4597 */
4598static void ice_remove(struct pci_dev *pdev)
4599{
4600 struct ice_pf *pf = pci_get_drvdata(pdev);
4601 int i;
4602
4603 if (!pf)
4604 return;
4605
4606 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4607 if (!ice_is_reset_in_progress(pf->state))
4608 break;
4609 msleep(100);
4610 }
4611
4612 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4613 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4614 ice_free_vfs(pf);
4615 }
4616
4617 ice_service_task_stop(pf);
4618
4619 ice_aq_cancel_waiting_tasks(pf);
4620 ice_unplug_aux_dev(pf);
4621 if (pf->aux_idx >= 0)
4622 ida_free(&ice_aux_ida, pf->aux_idx);
4623 set_bit(ICE_DOWN, pf->state);
4624
4625 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4626 ice_deinit_lag(pf);
4627 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4628 ice_ptp_release(pf);
4629 if (!ice_is_safe_mode(pf))
4630 ice_remove_arfs(pf);
4631 ice_setup_mc_magic_wake(pf);
4632 ice_vsi_release_all(pf);
4633 ice_set_wake(pf);
4634 ice_free_irq_msix_misc(pf);
4635 ice_for_each_vsi(pf, i) {
4636 if (!pf->vsi[i])
4637 continue;
4638 ice_vsi_free_q_vectors(pf->vsi[i]);
4639 }
4640 ice_deinit_pf(pf);
4641 ice_devlink_destroy_regions(pf);
4642 ice_deinit_hw(&pf->hw);
4643 ice_devlink_unregister(pf);
4644
4645 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4646 * do it via ice_schedule_reset() since there is no need to rebuild
4647 * and the service task is already stopped.
4648 */
4649 ice_reset(&pf->hw, ICE_RESET_PFR);
4650 pci_wait_for_pending_transaction(pdev);
4651 ice_clear_interrupt_scheme(pf);
4652 pci_disable_pcie_error_reporting(pdev);
4653 pci_disable_device(pdev);
4654}
4655
4656/**
4657 * ice_shutdown - PCI callback for shutting down device
4658 * @pdev: PCI device information struct
4659 */
4660static void ice_shutdown(struct pci_dev *pdev)
4661{
4662 struct ice_pf *pf = pci_get_drvdata(pdev);
4663
4664 ice_remove(pdev);
4665
4666 if (system_state == SYSTEM_POWER_OFF) {
4667 pci_wake_from_d3(pdev, pf->wol_ena);
4668 pci_set_power_state(pdev, PCI_D3hot);
4669 }
4670}
4671
4672#ifdef CONFIG_PM
4673/**
4674 * ice_prepare_for_shutdown - prep for PCI shutdown
4675 * @pf: board private structure
4676 *
4677 * Inform or close all dependent features in prep for PCI device shutdown
4678 */
4679static void ice_prepare_for_shutdown(struct ice_pf *pf)
4680{
4681 struct ice_hw *hw = &pf->hw;
4682 u32 v;
4683
4684 /* Notify VFs of impending reset */
4685 if (ice_check_sq_alive(hw, &hw->mailboxq))
4686 ice_vc_notify_reset(pf);
4687
4688 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4689
4690 /* disable the VSIs and their queues that are not already DOWN */
4691 ice_pf_dis_all_vsi(pf, false);
4692
4693 ice_for_each_vsi(pf, v)
4694 if (pf->vsi[v])
4695 pf->vsi[v]->vsi_num = 0;
4696
4697 ice_shutdown_all_ctrlq(hw);
4698}
4699
4700/**
4701 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4702 * @pf: board private structure to reinitialize
4703 *
4704 * This routine reinitialize interrupt scheme that was cleared during
4705 * power management suspend callback.
4706 *
4707 * This should be called during resume routine to re-allocate the q_vectors
4708 * and reacquire interrupts.
4709 */
4710static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4711{
4712 struct device *dev = ice_pf_to_dev(pf);
4713 int ret, v;
4714
4715 /* Since we clear MSIX flag during suspend, we need to
4716 * set it back during resume...
4717 */
4718
4719 ret = ice_init_interrupt_scheme(pf);
4720 if (ret) {
4721 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4722 return ret;
4723 }
4724
4725 /* Remap vectors and rings, after successful re-init interrupts */
4726 ice_for_each_vsi(pf, v) {
4727 if (!pf->vsi[v])
4728 continue;
4729
4730 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4731 if (ret)
4732 goto err_reinit;
4733 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4734 }
4735
4736 ret = ice_req_irq_msix_misc(pf);
4737 if (ret) {
4738 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4739 ret);
4740 goto err_reinit;
4741 }
4742
4743 return 0;
4744
4745err_reinit:
4746 while (v--)
4747 if (pf->vsi[v])
4748 ice_vsi_free_q_vectors(pf->vsi[v]);
4749
4750 return ret;
4751}
4752
4753/**
4754 * ice_suspend
4755 * @dev: generic device information structure
4756 *
4757 * Power Management callback to quiesce the device and prepare
4758 * for D3 transition.
4759 */
4760static int __maybe_unused ice_suspend(struct device *dev)
4761{
4762 struct pci_dev *pdev = to_pci_dev(dev);
4763 struct ice_pf *pf;
4764 int disabled, v;
4765
4766 pf = pci_get_drvdata(pdev);
4767
4768 if (!ice_pf_state_is_nominal(pf)) {
4769 dev_err(dev, "Device is not ready, no need to suspend it\n");
4770 return -EBUSY;
4771 }
4772
4773 /* Stop watchdog tasks until resume completion.
4774 * Even though it is most likely that the service task is
4775 * disabled if the device is suspended or down, the service task's
4776 * state is controlled by a different state bit, and we should
4777 * store and honor whatever state that bit is in at this point.
4778 */
4779 disabled = ice_service_task_stop(pf);
4780
4781 ice_unplug_aux_dev(pf);
4782
4783 /* Already suspended?, then there is nothing to do */
4784 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
4785 if (!disabled)
4786 ice_service_task_restart(pf);
4787 return 0;
4788 }
4789
4790 if (test_bit(ICE_DOWN, pf->state) ||
4791 ice_is_reset_in_progress(pf->state)) {
4792 dev_err(dev, "can't suspend device in reset or already down\n");
4793 if (!disabled)
4794 ice_service_task_restart(pf);
4795 return 0;
4796 }
4797
4798 ice_setup_mc_magic_wake(pf);
4799
4800 ice_prepare_for_shutdown(pf);
4801
4802 ice_set_wake(pf);
4803
4804 /* Free vectors, clear the interrupt scheme and release IRQs
4805 * for proper hibernation, especially with large number of CPUs.
4806 * Otherwise hibernation might fail when mapping all the vectors back
4807 * to CPU0.
4808 */
4809 ice_free_irq_msix_misc(pf);
4810 ice_for_each_vsi(pf, v) {
4811 if (!pf->vsi[v])
4812 continue;
4813 ice_vsi_free_q_vectors(pf->vsi[v]);
4814 }
4815 ice_free_cpu_rx_rmap(ice_get_main_vsi(pf));
4816 ice_clear_interrupt_scheme(pf);
4817
4818 pci_save_state(pdev);
4819 pci_wake_from_d3(pdev, pf->wol_ena);
4820 pci_set_power_state(pdev, PCI_D3hot);
4821 return 0;
4822}
4823
4824/**
4825 * ice_resume - PM callback for waking up from D3
4826 * @dev: generic device information structure
4827 */
4828static int __maybe_unused ice_resume(struct device *dev)
4829{
4830 struct pci_dev *pdev = to_pci_dev(dev);
4831 enum ice_reset_req reset_type;
4832 struct ice_pf *pf;
4833 struct ice_hw *hw;
4834 int ret;
4835
4836 pci_set_power_state(pdev, PCI_D0);
4837 pci_restore_state(pdev);
4838 pci_save_state(pdev);
4839
4840 if (!pci_device_is_present(pdev))
4841 return -ENODEV;
4842
4843 ret = pci_enable_device_mem(pdev);
4844 if (ret) {
4845 dev_err(dev, "Cannot enable device after suspend\n");
4846 return ret;
4847 }
4848
4849 pf = pci_get_drvdata(pdev);
4850 hw = &pf->hw;
4851
4852 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4853 ice_print_wake_reason(pf);
4854
4855 /* We cleared the interrupt scheme when we suspended, so we need to
4856 * restore it now to resume device functionality.
4857 */
4858 ret = ice_reinit_interrupt_scheme(pf);
4859 if (ret)
4860 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4861
4862 clear_bit(ICE_DOWN, pf->state);
4863 /* Now perform PF reset and rebuild */
4864 reset_type = ICE_RESET_PFR;
4865 /* re-enable service task for reset, but allow reset to schedule it */
4866 clear_bit(ICE_SERVICE_DIS, pf->state);
4867
4868 if (ice_schedule_reset(pf, reset_type))
4869 dev_err(dev, "Reset during resume failed.\n");
4870
4871 clear_bit(ICE_SUSPENDED, pf->state);
4872 ice_service_task_restart(pf);
4873
4874 /* Restart the service task */
4875 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4876
4877 return 0;
4878}
4879#endif /* CONFIG_PM */
4880
4881/**
4882 * ice_pci_err_detected - warning that PCI error has been detected
4883 * @pdev: PCI device information struct
4884 * @err: the type of PCI error
4885 *
4886 * Called to warn that something happened on the PCI bus and the error handling
4887 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4888 */
4889static pci_ers_result_t
4890ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4891{
4892 struct ice_pf *pf = pci_get_drvdata(pdev);
4893
4894 if (!pf) {
4895 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4896 __func__, err);
4897 return PCI_ERS_RESULT_DISCONNECT;
4898 }
4899
4900 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4901 ice_service_task_stop(pf);
4902
4903 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4904 set_bit(ICE_PFR_REQ, pf->state);
4905 ice_prepare_for_reset(pf);
4906 }
4907 }
4908
4909 return PCI_ERS_RESULT_NEED_RESET;
4910}
4911
4912/**
4913 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4914 * @pdev: PCI device information struct
4915 *
4916 * Called to determine if the driver can recover from the PCI slot reset by
4917 * using a register read to determine if the device is recoverable.
4918 */
4919static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4920{
4921 struct ice_pf *pf = pci_get_drvdata(pdev);
4922 pci_ers_result_t result;
4923 int err;
4924 u32 reg;
4925
4926 err = pci_enable_device_mem(pdev);
4927 if (err) {
4928 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4929 err);
4930 result = PCI_ERS_RESULT_DISCONNECT;
4931 } else {
4932 pci_set_master(pdev);
4933 pci_restore_state(pdev);
4934 pci_save_state(pdev);
4935 pci_wake_from_d3(pdev, false);
4936
4937 /* Check for life */
4938 reg = rd32(&pf->hw, GLGEN_RTRIG);
4939 if (!reg)
4940 result = PCI_ERS_RESULT_RECOVERED;
4941 else
4942 result = PCI_ERS_RESULT_DISCONNECT;
4943 }
4944
4945 err = pci_aer_clear_nonfatal_status(pdev);
4946 if (err)
4947 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4948 err);
4949 /* non-fatal, continue */
4950
4951 return result;
4952}
4953
4954/**
4955 * ice_pci_err_resume - restart operations after PCI error recovery
4956 * @pdev: PCI device information struct
4957 *
4958 * Called to allow the driver to bring things back up after PCI error and/or
4959 * reset recovery have finished
4960 */
4961static void ice_pci_err_resume(struct pci_dev *pdev)
4962{
4963 struct ice_pf *pf = pci_get_drvdata(pdev);
4964
4965 if (!pf) {
4966 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4967 __func__);
4968 return;
4969 }
4970
4971 if (test_bit(ICE_SUSPENDED, pf->state)) {
4972 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4973 __func__);
4974 return;
4975 }
4976
4977 ice_restore_all_vfs_msi_state(pdev);
4978
4979 ice_do_reset(pf, ICE_RESET_PFR);
4980 ice_service_task_restart(pf);
4981 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4982}
4983
4984/**
4985 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4986 * @pdev: PCI device information struct
4987 */
4988static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4989{
4990 struct ice_pf *pf = pci_get_drvdata(pdev);
4991
4992 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4993 ice_service_task_stop(pf);
4994
4995 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4996 set_bit(ICE_PFR_REQ, pf->state);
4997 ice_prepare_for_reset(pf);
4998 }
4999 }
5000}
5001
5002/**
5003 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5004 * @pdev: PCI device information struct
5005 */
5006static void ice_pci_err_reset_done(struct pci_dev *pdev)
5007{
5008 ice_pci_err_resume(pdev);
5009}
5010
5011/* ice_pci_tbl - PCI Device ID Table
5012 *
5013 * Wildcard entries (PCI_ANY_ID) should come last
5014 * Last entry must be all 0s
5015 *
5016 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5017 * Class, Class Mask, private data (not used) }
5018 */
5019static const struct pci_device_id ice_pci_tbl[] = {
5020 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5021 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5022 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5023 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5024 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5025 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5026 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5027 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5028 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5029 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5030 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5031 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5032 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5033 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5034 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5035 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5036 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5037 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5038 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5039 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5040 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5041 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5042 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5043 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5044 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5045 /* required last entry */
5046 { 0, }
5047};
5048MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5049
5050static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5051
5052static const struct pci_error_handlers ice_pci_err_handler = {
5053 .error_detected = ice_pci_err_detected,
5054 .slot_reset = ice_pci_err_slot_reset,
5055 .reset_prepare = ice_pci_err_reset_prepare,
5056 .reset_done = ice_pci_err_reset_done,
5057 .resume = ice_pci_err_resume
5058};
5059
5060static struct pci_driver ice_driver = {
5061 .name = KBUILD_MODNAME,
5062 .id_table = ice_pci_tbl,
5063 .probe = ice_probe,
5064 .remove = ice_remove,
5065#ifdef CONFIG_PM
5066 .driver.pm = &ice_pm_ops,
5067#endif /* CONFIG_PM */
5068 .shutdown = ice_shutdown,
5069 .sriov_configure = ice_sriov_configure,
5070 .err_handler = &ice_pci_err_handler
5071};
5072
5073/**
5074 * ice_module_init - Driver registration routine
5075 *
5076 * ice_module_init is the first routine called when the driver is
5077 * loaded. All it does is register with the PCI subsystem.
5078 */
5079static int __init ice_module_init(void)
5080{
5081 int status;
5082
5083 pr_info("%s\n", ice_driver_string);
5084 pr_info("%s\n", ice_copyright);
5085
5086 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5087 if (!ice_wq) {
5088 pr_err("Failed to create workqueue\n");
5089 return -ENOMEM;
5090 }
5091
5092 status = pci_register_driver(&ice_driver);
5093 if (status) {
5094 pr_err("failed to register PCI driver, err %d\n", status);
5095 destroy_workqueue(ice_wq);
5096 }
5097
5098 return status;
5099}
5100module_init(ice_module_init);
5101
5102/**
5103 * ice_module_exit - Driver exit cleanup routine
5104 *
5105 * ice_module_exit is called just before the driver is removed
5106 * from memory.
5107 */
5108static void __exit ice_module_exit(void)
5109{
5110 pci_unregister_driver(&ice_driver);
5111 destroy_workqueue(ice_wq);
5112 pr_info("module unloaded\n");
5113}
5114module_exit(ice_module_exit);
5115
5116/**
5117 * ice_set_mac_address - NDO callback to set MAC address
5118 * @netdev: network interface device structure
5119 * @pi: pointer to an address structure
5120 *
5121 * Returns 0 on success, negative on failure
5122 */
5123static int ice_set_mac_address(struct net_device *netdev, void *pi)
5124{
5125 struct ice_netdev_priv *np = netdev_priv(netdev);
5126 struct ice_vsi *vsi = np->vsi;
5127 struct ice_pf *pf = vsi->back;
5128 struct ice_hw *hw = &pf->hw;
5129 struct sockaddr *addr = pi;
5130 enum ice_status status;
5131 u8 old_mac[ETH_ALEN];
5132 u8 flags = 0;
5133 int err = 0;
5134 u8 *mac;
5135
5136 mac = (u8 *)addr->sa_data;
5137
5138 if (!is_valid_ether_addr(mac))
5139 return -EADDRNOTAVAIL;
5140
5141 if (ether_addr_equal(netdev->dev_addr, mac)) {
5142 netdev_dbg(netdev, "already using mac %pM\n", mac);
5143 return 0;
5144 }
5145
5146 if (test_bit(ICE_DOWN, pf->state) ||
5147 ice_is_reset_in_progress(pf->state)) {
5148 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5149 mac);
5150 return -EBUSY;
5151 }
5152
5153 netif_addr_lock_bh(netdev);
5154 ether_addr_copy(old_mac, netdev->dev_addr);
5155 /* change the netdev's MAC address */
5156 memcpy(netdev->dev_addr, mac, netdev->addr_len);
5157 netif_addr_unlock_bh(netdev);
5158
5159 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5160 status = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5161 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
5162 err = -EADDRNOTAVAIL;
5163 goto err_update_filters;
5164 }
5165
5166 /* Add filter for new MAC. If filter exists, return success */
5167 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5168 if (status == ICE_ERR_ALREADY_EXISTS)
5169 /* Although this MAC filter is already present in hardware it's
5170 * possible in some cases (e.g. bonding) that dev_addr was
5171 * modified outside of the driver and needs to be restored back
5172 * to this value.
5173 */
5174 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5175 else if (status)
5176 /* error if the new filter addition failed */
5177 err = -EADDRNOTAVAIL;
5178
5179err_update_filters:
5180 if (err) {
5181 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5182 mac);
5183 netif_addr_lock_bh(netdev);
5184 ether_addr_copy(netdev->dev_addr, old_mac);
5185 netif_addr_unlock_bh(netdev);
5186 return err;
5187 }
5188
5189 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5190 netdev->dev_addr);
5191
5192 /* write new MAC address to the firmware */
5193 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5194 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5195 if (status) {
5196 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
5197 mac, ice_stat_str(status));
5198 }
5199 return 0;
5200}
5201
5202/**
5203 * ice_set_rx_mode - NDO callback to set the netdev filters
5204 * @netdev: network interface device structure
5205 */
5206static void ice_set_rx_mode(struct net_device *netdev)
5207{
5208 struct ice_netdev_priv *np = netdev_priv(netdev);
5209 struct ice_vsi *vsi = np->vsi;
5210
5211 if (!vsi)
5212 return;
5213
5214 /* Set the flags to synchronize filters
5215 * ndo_set_rx_mode may be triggered even without a change in netdev
5216 * flags
5217 */
5218 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5219 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5220 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5221
5222 /* schedule our worker thread which will take care of
5223 * applying the new filter changes
5224 */
5225 ice_service_task_schedule(vsi->back);
5226}
5227
5228/**
5229 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5230 * @netdev: network interface device structure
5231 * @queue_index: Queue ID
5232 * @maxrate: maximum bandwidth in Mbps
5233 */
5234static int
5235ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5236{
5237 struct ice_netdev_priv *np = netdev_priv(netdev);
5238 struct ice_vsi *vsi = np->vsi;
5239 enum ice_status status;
5240 u16 q_handle;
5241 u8 tc;
5242
5243 /* Validate maxrate requested is within permitted range */
5244 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5245 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5246 maxrate, queue_index);
5247 return -EINVAL;
5248 }
5249
5250 q_handle = vsi->tx_rings[queue_index]->q_handle;
5251 tc = ice_dcb_get_tc(vsi, queue_index);
5252
5253 /* Set BW back to default, when user set maxrate to 0 */
5254 if (!maxrate)
5255 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5256 q_handle, ICE_MAX_BW);
5257 else
5258 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5259 q_handle, ICE_MAX_BW, maxrate * 1000);
5260 if (status) {
5261 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
5262 ice_stat_str(status));
5263 return -EIO;
5264 }
5265
5266 return 0;
5267}
5268
5269/**
5270 * ice_fdb_add - add an entry to the hardware database
5271 * @ndm: the input from the stack
5272 * @tb: pointer to array of nladdr (unused)
5273 * @dev: the net device pointer
5274 * @addr: the MAC address entry being added
5275 * @vid: VLAN ID
5276 * @flags: instructions from stack about fdb operation
5277 * @extack: netlink extended ack
5278 */
5279static int
5280ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5281 struct net_device *dev, const unsigned char *addr, u16 vid,
5282 u16 flags, struct netlink_ext_ack __always_unused *extack)
5283{
5284 int err;
5285
5286 if (vid) {
5287 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5288 return -EINVAL;
5289 }
5290 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5291 netdev_err(dev, "FDB only supports static addresses\n");
5292 return -EINVAL;
5293 }
5294
5295 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5296 err = dev_uc_add_excl(dev, addr);
5297 else if (is_multicast_ether_addr(addr))
5298 err = dev_mc_add_excl(dev, addr);
5299 else
5300 err = -EINVAL;
5301
5302 /* Only return duplicate errors if NLM_F_EXCL is set */
5303 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5304 err = 0;
5305
5306 return err;
5307}
5308
5309/**
5310 * ice_fdb_del - delete an entry from the hardware database
5311 * @ndm: the input from the stack
5312 * @tb: pointer to array of nladdr (unused)
5313 * @dev: the net device pointer
5314 * @addr: the MAC address entry being added
5315 * @vid: VLAN ID
5316 */
5317static int
5318ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5319 struct net_device *dev, const unsigned char *addr,
5320 __always_unused u16 vid)
5321{
5322 int err;
5323
5324 if (ndm->ndm_state & NUD_PERMANENT) {
5325 netdev_err(dev, "FDB only supports static addresses\n");
5326 return -EINVAL;
5327 }
5328
5329 if (is_unicast_ether_addr(addr))
5330 err = dev_uc_del(dev, addr);
5331 else if (is_multicast_ether_addr(addr))
5332 err = dev_mc_del(dev, addr);
5333 else
5334 err = -EINVAL;
5335
5336 return err;
5337}
5338
5339/**
5340 * ice_set_features - set the netdev feature flags
5341 * @netdev: ptr to the netdev being adjusted
5342 * @features: the feature set that the stack is suggesting
5343 */
5344static int
5345ice_set_features(struct net_device *netdev, netdev_features_t features)
5346{
5347 struct ice_netdev_priv *np = netdev_priv(netdev);
5348 struct ice_vsi *vsi = np->vsi;
5349 struct ice_pf *pf = vsi->back;
5350 int ret = 0;
5351
5352 /* Don't set any netdev advanced features with device in Safe Mode */
5353 if (ice_is_safe_mode(vsi->back)) {
5354 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5355 return ret;
5356 }
5357
5358 /* Do not change setting during reset */
5359 if (ice_is_reset_in_progress(pf->state)) {
5360 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5361 return -EBUSY;
5362 }
5363
5364 /* Multiple features can be changed in one call so keep features in
5365 * separate if/else statements to guarantee each feature is checked
5366 */
5367 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5368 ice_vsi_manage_rss_lut(vsi, true);
5369 else if (!(features & NETIF_F_RXHASH) &&
5370 netdev->features & NETIF_F_RXHASH)
5371 ice_vsi_manage_rss_lut(vsi, false);
5372
5373 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5374 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5375 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5376 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5377 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5378 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5379
5380 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5381 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5382 ret = ice_vsi_manage_vlan_insertion(vsi);
5383 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5384 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5385 ret = ice_vsi_manage_vlan_insertion(vsi);
5386
5387 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5388 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5389 ret = ice_cfg_vlan_pruning(vsi, true, false);
5390 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5391 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5392 ret = ice_cfg_vlan_pruning(vsi, false, false);
5393
5394 if ((features & NETIF_F_NTUPLE) &&
5395 !(netdev->features & NETIF_F_NTUPLE)) {
5396 ice_vsi_manage_fdir(vsi, true);
5397 ice_init_arfs(vsi);
5398 } else if (!(features & NETIF_F_NTUPLE) &&
5399 (netdev->features & NETIF_F_NTUPLE)) {
5400 ice_vsi_manage_fdir(vsi, false);
5401 ice_clear_arfs(vsi);
5402 }
5403
5404 return ret;
5405}
5406
5407/**
5408 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5409 * @vsi: VSI to setup VLAN properties for
5410 */
5411static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5412{
5413 int ret = 0;
5414
5415 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5416 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5417 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5418 ret = ice_vsi_manage_vlan_insertion(vsi);
5419
5420 return ret;
5421}
5422
5423/**
5424 * ice_vsi_cfg - Setup the VSI
5425 * @vsi: the VSI being configured
5426 *
5427 * Return 0 on success and negative value on error
5428 */
5429int ice_vsi_cfg(struct ice_vsi *vsi)
5430{
5431 int err;
5432
5433 if (vsi->netdev) {
5434 ice_set_rx_mode(vsi->netdev);
5435
5436 err = ice_vsi_vlan_setup(vsi);
5437
5438 if (err)
5439 return err;
5440 }
5441 ice_vsi_cfg_dcb_rings(vsi);
5442
5443 err = ice_vsi_cfg_lan_txqs(vsi);
5444 if (!err && ice_is_xdp_ena_vsi(vsi))
5445 err = ice_vsi_cfg_xdp_txqs(vsi);
5446 if (!err)
5447 err = ice_vsi_cfg_rxqs(vsi);
5448
5449 return err;
5450}
5451
5452/* THEORY OF MODERATION:
5453 * The below code creates custom DIM profiles for use by this driver, because
5454 * the ice driver hardware works differently than the hardware that DIMLIB was
5455 * originally made for. ice hardware doesn't have packet count limits that
5456 * can trigger an interrupt, but it *does* have interrupt rate limit support,
5457 * and this code adds that capability to be used by the driver when it's using
5458 * DIMLIB. The DIMLIB code was always designed to be a suggestion to the driver
5459 * for how to "respond" to traffic and interrupts, so this driver uses a
5460 * slightly different set of moderation parameters to get best performance.
5461 */
5462struct ice_dim {
5463 /* the throttle rate for interrupts, basically worst case delay before
5464 * an initial interrupt fires, value is stored in microseconds.
5465 */
5466 u16 itr;
5467 /* the rate limit for interrupts, which can cap a delay from a small
5468 * ITR at a certain amount of interrupts per second. f.e. a 2us ITR
5469 * could yield as much as 500,000 interrupts per second, but with a
5470 * 10us rate limit, it limits to 100,000 interrupts per second. Value
5471 * is stored in microseconds.
5472 */
5473 u16 intrl;
5474};
5475
5476/* Make a different profile for Rx that doesn't allow quite so aggressive
5477 * moderation at the high end (it maxes out at 128us or about 8k interrupts a
5478 * second. The INTRL/rate parameters here are only useful to cap small ITR
5479 * values, which is why for larger ITR's - like 128, which can only generate
5480 * 8k interrupts per second, there is no point to rate limit and the values
5481 * are set to zero. The rate limit values do affect latency, and so must
5482 * be reasonably small so to not impact latency sensitive tests.
5483 */
5484static const struct ice_dim rx_profile[] = {
5485 {2, 10},
5486 {8, 16},
5487 {32, 0},
5488 {96, 0},
5489 {128, 0}
5490};
5491
5492/* The transmit profile, which has the same sorts of values
5493 * as the previous struct
5494 */
5495static const struct ice_dim tx_profile[] = {
5496 {2, 10},
5497 {8, 16},
5498 {64, 0},
5499 {128, 0},
5500 {256, 0}
5501};
5502
5503static void ice_tx_dim_work(struct work_struct *work)
5504{
5505 struct ice_ring_container *rc;
5506 struct ice_q_vector *q_vector;
5507 struct dim *dim;
5508 u16 itr, intrl;
5509
5510 dim = container_of(work, struct dim, work);
5511 rc = container_of(dim, struct ice_ring_container, dim);
5512 q_vector = container_of(rc, struct ice_q_vector, tx);
5513
5514 if (dim->profile_ix >= ARRAY_SIZE(tx_profile))
5515 dim->profile_ix = ARRAY_SIZE(tx_profile) - 1;
5516
5517 /* look up the values in our local table */
5518 itr = tx_profile[dim->profile_ix].itr;
5519 intrl = tx_profile[dim->profile_ix].intrl;
5520
5521 ice_trace(tx_dim_work, q_vector, dim);
5522 ice_write_itr(rc, itr);
5523 ice_write_intrl(q_vector, intrl);
5524
5525 dim->state = DIM_START_MEASURE;
5526}
5527
5528static void ice_rx_dim_work(struct work_struct *work)
5529{
5530 struct ice_ring_container *rc;
5531 struct ice_q_vector *q_vector;
5532 struct dim *dim;
5533 u16 itr, intrl;
5534
5535 dim = container_of(work, struct dim, work);
5536 rc = container_of(dim, struct ice_ring_container, dim);
5537 q_vector = container_of(rc, struct ice_q_vector, rx);
5538
5539 if (dim->profile_ix >= ARRAY_SIZE(rx_profile))
5540 dim->profile_ix = ARRAY_SIZE(rx_profile) - 1;
5541
5542 /* look up the values in our local table */
5543 itr = rx_profile[dim->profile_ix].itr;
5544 intrl = rx_profile[dim->profile_ix].intrl;
5545
5546 ice_trace(rx_dim_work, q_vector, dim);
5547 ice_write_itr(rc, itr);
5548 ice_write_intrl(q_vector, intrl);
5549
5550 dim->state = DIM_START_MEASURE;
5551}
5552
5553/**
5554 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5555 * @vsi: the VSI being configured
5556 */
5557static void ice_napi_enable_all(struct ice_vsi *vsi)
5558{
5559 int q_idx;
5560
5561 if (!vsi->netdev)
5562 return;
5563
5564 ice_for_each_q_vector(vsi, q_idx) {
5565 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5566
5567 INIT_WORK(&q_vector->tx.dim.work, ice_tx_dim_work);
5568 q_vector->tx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5569
5570 INIT_WORK(&q_vector->rx.dim.work, ice_rx_dim_work);
5571 q_vector->rx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5572
5573 if (q_vector->rx.ring || q_vector->tx.ring)
5574 napi_enable(&q_vector->napi);
5575 }
5576}
5577
5578/**
5579 * ice_up_complete - Finish the last steps of bringing up a connection
5580 * @vsi: The VSI being configured
5581 *
5582 * Return 0 on success and negative value on error
5583 */
5584static int ice_up_complete(struct ice_vsi *vsi)
5585{
5586 struct ice_pf *pf = vsi->back;
5587 int err;
5588
5589 ice_vsi_cfg_msix(vsi);
5590
5591 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5592 * Tx queue group list was configured and the context bits were
5593 * programmed using ice_vsi_cfg_txqs
5594 */
5595 err = ice_vsi_start_all_rx_rings(vsi);
5596 if (err)
5597 return err;
5598
5599 clear_bit(ICE_VSI_DOWN, vsi->state);
5600 ice_napi_enable_all(vsi);
5601 ice_vsi_ena_irq(vsi);
5602
5603 if (vsi->port_info &&
5604 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5605 vsi->netdev) {
5606 ice_print_link_msg(vsi, true);
5607 netif_tx_start_all_queues(vsi->netdev);
5608 netif_carrier_on(vsi->netdev);
5609 }
5610
5611 ice_service_task_schedule(pf);
5612
5613 return 0;
5614}
5615
5616/**
5617 * ice_up - Bring the connection back up after being down
5618 * @vsi: VSI being configured
5619 */
5620int ice_up(struct ice_vsi *vsi)
5621{
5622 int err;
5623
5624 err = ice_vsi_cfg(vsi);
5625 if (!err)
5626 err = ice_up_complete(vsi);
5627
5628 return err;
5629}
5630
5631/**
5632 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5633 * @ring: Tx or Rx ring to read stats from
5634 * @pkts: packets stats counter
5635 * @bytes: bytes stats counter
5636 *
5637 * This function fetches stats from the ring considering the atomic operations
5638 * that needs to be performed to read u64 values in 32 bit machine.
5639 */
5640static void
5641ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5642{
5643 unsigned int start;
5644 *pkts = 0;
5645 *bytes = 0;
5646
5647 if (!ring)
5648 return;
5649 do {
5650 start = u64_stats_fetch_begin_irq(&ring->syncp);
5651 *pkts = ring->stats.pkts;
5652 *bytes = ring->stats.bytes;
5653 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5654}
5655
5656/**
5657 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5658 * @vsi: the VSI to be updated
5659 * @rings: rings to work on
5660 * @count: number of rings
5661 */
5662static void
5663ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5664 u16 count)
5665{
5666 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5667 u16 i;
5668
5669 for (i = 0; i < count; i++) {
5670 struct ice_ring *ring;
5671 u64 pkts, bytes;
5672
5673 ring = READ_ONCE(rings[i]);
5674 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5675 vsi_stats->tx_packets += pkts;
5676 vsi_stats->tx_bytes += bytes;
5677 vsi->tx_restart += ring->tx_stats.restart_q;
5678 vsi->tx_busy += ring->tx_stats.tx_busy;
5679 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5680 }
5681}
5682
5683/**
5684 * ice_update_vsi_ring_stats - Update VSI stats counters
5685 * @vsi: the VSI to be updated
5686 */
5687static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5688{
5689 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5690 u64 pkts, bytes;
5691 int i;
5692
5693 /* reset netdev stats */
5694 vsi_stats->tx_packets = 0;
5695 vsi_stats->tx_bytes = 0;
5696 vsi_stats->rx_packets = 0;
5697 vsi_stats->rx_bytes = 0;
5698
5699 /* reset non-netdev (extended) stats */
5700 vsi->tx_restart = 0;
5701 vsi->tx_busy = 0;
5702 vsi->tx_linearize = 0;
5703 vsi->rx_buf_failed = 0;
5704 vsi->rx_page_failed = 0;
5705
5706 rcu_read_lock();
5707
5708 /* update Tx rings counters */
5709 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5710
5711 /* update Rx rings counters */
5712 ice_for_each_rxq(vsi, i) {
5713 struct ice_ring *ring = READ_ONCE(vsi->rx_rings[i]);
5714
5715 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5716 vsi_stats->rx_packets += pkts;
5717 vsi_stats->rx_bytes += bytes;
5718 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5719 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5720 }
5721
5722 /* update XDP Tx rings counters */
5723 if (ice_is_xdp_ena_vsi(vsi))
5724 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5725 vsi->num_xdp_txq);
5726
5727 rcu_read_unlock();
5728}
5729
5730/**
5731 * ice_update_vsi_stats - Update VSI stats counters
5732 * @vsi: the VSI to be updated
5733 */
5734void ice_update_vsi_stats(struct ice_vsi *vsi)
5735{
5736 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5737 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5738 struct ice_pf *pf = vsi->back;
5739
5740 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
5741 test_bit(ICE_CFG_BUSY, pf->state))
5742 return;
5743
5744 /* get stats as recorded by Tx/Rx rings */
5745 ice_update_vsi_ring_stats(vsi);
5746
5747 /* get VSI stats as recorded by the hardware */
5748 ice_update_eth_stats(vsi);
5749
5750 cur_ns->tx_errors = cur_es->tx_errors;
5751 cur_ns->rx_dropped = cur_es->rx_discards;
5752 cur_ns->tx_dropped = cur_es->tx_discards;
5753 cur_ns->multicast = cur_es->rx_multicast;
5754
5755 /* update some more netdev stats if this is main VSI */
5756 if (vsi->type == ICE_VSI_PF) {
5757 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5758 cur_ns->rx_errors = pf->stats.crc_errors +
5759 pf->stats.illegal_bytes +
5760 pf->stats.rx_len_errors +
5761 pf->stats.rx_undersize +
5762 pf->hw_csum_rx_error +
5763 pf->stats.rx_jabber +
5764 pf->stats.rx_fragments +
5765 pf->stats.rx_oversize;
5766 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5767 /* record drops from the port level */
5768 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5769 }
5770}
5771
5772/**
5773 * ice_update_pf_stats - Update PF port stats counters
5774 * @pf: PF whose stats needs to be updated
5775 */
5776void ice_update_pf_stats(struct ice_pf *pf)
5777{
5778 struct ice_hw_port_stats *prev_ps, *cur_ps;
5779 struct ice_hw *hw = &pf->hw;
5780 u16 fd_ctr_base;
5781 u8 port;
5782
5783 port = hw->port_info->lport;
5784 prev_ps = &pf->stats_prev;
5785 cur_ps = &pf->stats;
5786
5787 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5788 &prev_ps->eth.rx_bytes,
5789 &cur_ps->eth.rx_bytes);
5790
5791 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5792 &prev_ps->eth.rx_unicast,
5793 &cur_ps->eth.rx_unicast);
5794
5795 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5796 &prev_ps->eth.rx_multicast,
5797 &cur_ps->eth.rx_multicast);
5798
5799 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5800 &prev_ps->eth.rx_broadcast,
5801 &cur_ps->eth.rx_broadcast);
5802
5803 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5804 &prev_ps->eth.rx_discards,
5805 &cur_ps->eth.rx_discards);
5806
5807 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5808 &prev_ps->eth.tx_bytes,
5809 &cur_ps->eth.tx_bytes);
5810
5811 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5812 &prev_ps->eth.tx_unicast,
5813 &cur_ps->eth.tx_unicast);
5814
5815 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5816 &prev_ps->eth.tx_multicast,
5817 &cur_ps->eth.tx_multicast);
5818
5819 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5820 &prev_ps->eth.tx_broadcast,
5821 &cur_ps->eth.tx_broadcast);
5822
5823 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5824 &prev_ps->tx_dropped_link_down,
5825 &cur_ps->tx_dropped_link_down);
5826
5827 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5828 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5829
5830 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5831 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5832
5833 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5834 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5835
5836 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5837 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5838
5839 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5840 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5841
5842 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5843 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5844
5845 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5846 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5847
5848 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5849 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5850
5851 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5852 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5853
5854 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5855 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5856
5857 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5858 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5859
5860 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5861 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5862
5863 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5864 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5865
5866 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5867 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5868
5869 fd_ctr_base = hw->fd_ctr_base;
5870
5871 ice_stat_update40(hw,
5872 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5873 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5874 &cur_ps->fd_sb_match);
5875 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5876 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5877
5878 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5879 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5880
5881 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5882 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5883
5884 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5885 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5886
5887 ice_update_dcb_stats(pf);
5888
5889 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5890 &prev_ps->crc_errors, &cur_ps->crc_errors);
5891
5892 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5893 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5894
5895 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5896 &prev_ps->mac_local_faults,
5897 &cur_ps->mac_local_faults);
5898
5899 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5900 &prev_ps->mac_remote_faults,
5901 &cur_ps->mac_remote_faults);
5902
5903 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5904 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5905
5906 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5907 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5908
5909 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5910 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5911
5912 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5913 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5914
5915 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5916 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5917
5918 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5919
5920 pf->stat_prev_loaded = true;
5921}
5922
5923/**
5924 * ice_get_stats64 - get statistics for network device structure
5925 * @netdev: network interface device structure
5926 * @stats: main device statistics structure
5927 */
5928static
5929void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5930{
5931 struct ice_netdev_priv *np = netdev_priv(netdev);
5932 struct rtnl_link_stats64 *vsi_stats;
5933 struct ice_vsi *vsi = np->vsi;
5934
5935 vsi_stats = &vsi->net_stats;
5936
5937 if (!vsi->num_txq || !vsi->num_rxq)
5938 return;
5939
5940 /* netdev packet/byte stats come from ring counter. These are obtained
5941 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5942 * But, only call the update routine and read the registers if VSI is
5943 * not down.
5944 */
5945 if (!test_bit(ICE_VSI_DOWN, vsi->state))
5946 ice_update_vsi_ring_stats(vsi);
5947 stats->tx_packets = vsi_stats->tx_packets;
5948 stats->tx_bytes = vsi_stats->tx_bytes;
5949 stats->rx_packets = vsi_stats->rx_packets;
5950 stats->rx_bytes = vsi_stats->rx_bytes;
5951
5952 /* The rest of the stats can be read from the hardware but instead we
5953 * just return values that the watchdog task has already obtained from
5954 * the hardware.
5955 */
5956 stats->multicast = vsi_stats->multicast;
5957 stats->tx_errors = vsi_stats->tx_errors;
5958 stats->tx_dropped = vsi_stats->tx_dropped;
5959 stats->rx_errors = vsi_stats->rx_errors;
5960 stats->rx_dropped = vsi_stats->rx_dropped;
5961 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5962 stats->rx_length_errors = vsi_stats->rx_length_errors;
5963}
5964
5965/**
5966 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5967 * @vsi: VSI having NAPI disabled
5968 */
5969static void ice_napi_disable_all(struct ice_vsi *vsi)
5970{
5971 int q_idx;
5972
5973 if (!vsi->netdev)
5974 return;
5975
5976 ice_for_each_q_vector(vsi, q_idx) {
5977 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5978
5979 if (q_vector->rx.ring || q_vector->tx.ring)
5980 napi_disable(&q_vector->napi);
5981
5982 cancel_work_sync(&q_vector->tx.dim.work);
5983 cancel_work_sync(&q_vector->rx.dim.work);
5984 }
5985}
5986
5987/**
5988 * ice_down - Shutdown the connection
5989 * @vsi: The VSI being stopped
5990 */
5991int ice_down(struct ice_vsi *vsi)
5992{
5993 int i, tx_err, rx_err, link_err = 0;
5994
5995 /* Caller of this function is expected to set the
5996 * vsi->state ICE_DOWN bit
5997 */
5998 if (vsi->netdev) {
5999 netif_carrier_off(vsi->netdev);
6000 netif_tx_disable(vsi->netdev);
6001 }
6002
6003 ice_vsi_dis_irq(vsi);
6004
6005 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6006 if (tx_err)
6007 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6008 vsi->vsi_num, tx_err);
6009 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6010 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6011 if (tx_err)
6012 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6013 vsi->vsi_num, tx_err);
6014 }
6015
6016 rx_err = ice_vsi_stop_all_rx_rings(vsi);
6017 if (rx_err)
6018 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6019 vsi->vsi_num, rx_err);
6020
6021 ice_napi_disable_all(vsi);
6022
6023 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6024 link_err = ice_force_phys_link_state(vsi, false);
6025 if (link_err)
6026 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6027 vsi->vsi_num, link_err);
6028 }
6029
6030 ice_for_each_txq(vsi, i)
6031 ice_clean_tx_ring(vsi->tx_rings[i]);
6032
6033 ice_for_each_rxq(vsi, i)
6034 ice_clean_rx_ring(vsi->rx_rings[i]);
6035
6036 if (tx_err || rx_err || link_err) {
6037 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6038 vsi->vsi_num, vsi->vsw->sw_id);
6039 return -EIO;
6040 }
6041
6042 return 0;
6043}
6044
6045/**
6046 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6047 * @vsi: VSI having resources allocated
6048 *
6049 * Return 0 on success, negative on failure
6050 */
6051int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6052{
6053 int i, err = 0;
6054
6055 if (!vsi->num_txq) {
6056 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6057 vsi->vsi_num);
6058 return -EINVAL;
6059 }
6060
6061 ice_for_each_txq(vsi, i) {
6062 struct ice_ring *ring = vsi->tx_rings[i];
6063
6064 if (!ring)
6065 return -EINVAL;
6066
6067 ring->netdev = vsi->netdev;
6068 err = ice_setup_tx_ring(ring);
6069 if (err)
6070 break;
6071 }
6072
6073 return err;
6074}
6075
6076/**
6077 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6078 * @vsi: VSI having resources allocated
6079 *
6080 * Return 0 on success, negative on failure
6081 */
6082int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6083{
6084 int i, err = 0;
6085
6086 if (!vsi->num_rxq) {
6087 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6088 vsi->vsi_num);
6089 return -EINVAL;
6090 }
6091
6092 ice_for_each_rxq(vsi, i) {
6093 struct ice_ring *ring = vsi->rx_rings[i];
6094
6095 if (!ring)
6096 return -EINVAL;
6097
6098 ring->netdev = vsi->netdev;
6099 err = ice_setup_rx_ring(ring);
6100 if (err)
6101 break;
6102 }
6103
6104 return err;
6105}
6106
6107/**
6108 * ice_vsi_open_ctrl - open control VSI for use
6109 * @vsi: the VSI to open
6110 *
6111 * Initialization of the Control VSI
6112 *
6113 * Returns 0 on success, negative value on error
6114 */
6115int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6116{
6117 char int_name[ICE_INT_NAME_STR_LEN];
6118 struct ice_pf *pf = vsi->back;
6119 struct device *dev;
6120 int err;
6121
6122 dev = ice_pf_to_dev(pf);
6123 /* allocate descriptors */
6124 err = ice_vsi_setup_tx_rings(vsi);
6125 if (err)
6126 goto err_setup_tx;
6127
6128 err = ice_vsi_setup_rx_rings(vsi);
6129 if (err)
6130 goto err_setup_rx;
6131
6132 err = ice_vsi_cfg(vsi);
6133 if (err)
6134 goto err_setup_rx;
6135
6136 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6137 dev_driver_string(dev), dev_name(dev));
6138 err = ice_vsi_req_irq_msix(vsi, int_name);
6139 if (err)
6140 goto err_setup_rx;
6141
6142 ice_vsi_cfg_msix(vsi);
6143
6144 err = ice_vsi_start_all_rx_rings(vsi);
6145 if (err)
6146 goto err_up_complete;
6147
6148 clear_bit(ICE_VSI_DOWN, vsi->state);
6149 ice_vsi_ena_irq(vsi);
6150
6151 return 0;
6152
6153err_up_complete:
6154 ice_down(vsi);
6155err_setup_rx:
6156 ice_vsi_free_rx_rings(vsi);
6157err_setup_tx:
6158 ice_vsi_free_tx_rings(vsi);
6159
6160 return err;
6161}
6162
6163/**
6164 * ice_vsi_open - Called when a network interface is made active
6165 * @vsi: the VSI to open
6166 *
6167 * Initialization of the VSI
6168 *
6169 * Returns 0 on success, negative value on error
6170 */
6171static int ice_vsi_open(struct ice_vsi *vsi)
6172{
6173 char int_name[ICE_INT_NAME_STR_LEN];
6174 struct ice_pf *pf = vsi->back;
6175 int err;
6176
6177 /* allocate descriptors */
6178 err = ice_vsi_setup_tx_rings(vsi);
6179 if (err)
6180 goto err_setup_tx;
6181
6182 err = ice_vsi_setup_rx_rings(vsi);
6183 if (err)
6184 goto err_setup_rx;
6185
6186 err = ice_vsi_cfg(vsi);
6187 if (err)
6188 goto err_setup_rx;
6189
6190 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6191 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6192 err = ice_vsi_req_irq_msix(vsi, int_name);
6193 if (err)
6194 goto err_setup_rx;
6195
6196 /* Notify the stack of the actual queue counts. */
6197 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6198 if (err)
6199 goto err_set_qs;
6200
6201 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6202 if (err)
6203 goto err_set_qs;
6204
6205 err = ice_up_complete(vsi);
6206 if (err)
6207 goto err_up_complete;
6208
6209 return 0;
6210
6211err_up_complete:
6212 ice_down(vsi);
6213err_set_qs:
6214 ice_vsi_free_irq(vsi);
6215err_setup_rx:
6216 ice_vsi_free_rx_rings(vsi);
6217err_setup_tx:
6218 ice_vsi_free_tx_rings(vsi);
6219
6220 return err;
6221}
6222
6223/**
6224 * ice_vsi_release_all - Delete all VSIs
6225 * @pf: PF from which all VSIs are being removed
6226 */
6227static void ice_vsi_release_all(struct ice_pf *pf)
6228{
6229 int err, i;
6230
6231 if (!pf->vsi)
6232 return;
6233
6234 ice_for_each_vsi(pf, i) {
6235 if (!pf->vsi[i])
6236 continue;
6237
6238 err = ice_vsi_release(pf->vsi[i]);
6239 if (err)
6240 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6241 i, err, pf->vsi[i]->vsi_num);
6242 }
6243}
6244
6245/**
6246 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6247 * @pf: pointer to the PF instance
6248 * @type: VSI type to rebuild
6249 *
6250 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6251 */
6252static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6253{
6254 struct device *dev = ice_pf_to_dev(pf);
6255 enum ice_status status;
6256 int i, err;
6257
6258 ice_for_each_vsi(pf, i) {
6259 struct ice_vsi *vsi = pf->vsi[i];
6260
6261 if (!vsi || vsi->type != type)
6262 continue;
6263
6264 /* rebuild the VSI */
6265 err = ice_vsi_rebuild(vsi, true);
6266 if (err) {
6267 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6268 err, vsi->idx, ice_vsi_type_str(type));
6269 return err;
6270 }
6271
6272 /* replay filters for the VSI */
6273 status = ice_replay_vsi(&pf->hw, vsi->idx);
6274 if (status) {
6275 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
6276 ice_stat_str(status), vsi->idx,
6277 ice_vsi_type_str(type));
6278 return -EIO;
6279 }
6280
6281 /* Re-map HW VSI number, using VSI handle that has been
6282 * previously validated in ice_replay_vsi() call above
6283 */
6284 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6285
6286 /* enable the VSI */
6287 err = ice_ena_vsi(vsi, false);
6288 if (err) {
6289 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6290 err, vsi->idx, ice_vsi_type_str(type));
6291 return err;
6292 }
6293
6294 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6295 ice_vsi_type_str(type));
6296 }
6297
6298 return 0;
6299}
6300
6301/**
6302 * ice_update_pf_netdev_link - Update PF netdev link status
6303 * @pf: pointer to the PF instance
6304 */
6305static void ice_update_pf_netdev_link(struct ice_pf *pf)
6306{
6307 bool link_up;
6308 int i;
6309
6310 ice_for_each_vsi(pf, i) {
6311 struct ice_vsi *vsi = pf->vsi[i];
6312
6313 if (!vsi || vsi->type != ICE_VSI_PF)
6314 return;
6315
6316 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6317 if (link_up) {
6318 netif_carrier_on(pf->vsi[i]->netdev);
6319 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6320 } else {
6321 netif_carrier_off(pf->vsi[i]->netdev);
6322 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6323 }
6324 }
6325}
6326
6327/**
6328 * ice_rebuild - rebuild after reset
6329 * @pf: PF to rebuild
6330 * @reset_type: type of reset
6331 *
6332 * Do not rebuild VF VSI in this flow because that is already handled via
6333 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6334 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6335 * to reset/rebuild all the VF VSI twice.
6336 */
6337static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6338{
6339 struct device *dev = ice_pf_to_dev(pf);
6340 struct ice_hw *hw = &pf->hw;
6341 enum ice_status ret;
6342 int err;
6343
6344 if (test_bit(ICE_DOWN, pf->state))
6345 goto clear_recovery;
6346
6347 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6348
6349 ret = ice_init_all_ctrlq(hw);
6350 if (ret) {
6351 dev_err(dev, "control queues init failed %s\n",
6352 ice_stat_str(ret));
6353 goto err_init_ctrlq;
6354 }
6355
6356 /* if DDP was previously loaded successfully */
6357 if (!ice_is_safe_mode(pf)) {
6358 /* reload the SW DB of filter tables */
6359 if (reset_type == ICE_RESET_PFR)
6360 ice_fill_blk_tbls(hw);
6361 else
6362 /* Reload DDP Package after CORER/GLOBR reset */
6363 ice_load_pkg(NULL, pf);
6364 }
6365
6366 ret = ice_clear_pf_cfg(hw);
6367 if (ret) {
6368 dev_err(dev, "clear PF configuration failed %s\n",
6369 ice_stat_str(ret));
6370 goto err_init_ctrlq;
6371 }
6372
6373 if (pf->first_sw->dflt_vsi_ena)
6374 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6375 /* clear the default VSI configuration if it exists */
6376 pf->first_sw->dflt_vsi = NULL;
6377 pf->first_sw->dflt_vsi_ena = false;
6378
6379 ice_clear_pxe_mode(hw);
6380
6381 ret = ice_init_nvm(hw);
6382 if (ret) {
6383 dev_err(dev, "ice_init_nvm failed %s\n", ice_stat_str(ret));
6384 goto err_init_ctrlq;
6385 }
6386
6387 ret = ice_get_caps(hw);
6388 if (ret) {
6389 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
6390 goto err_init_ctrlq;
6391 }
6392
6393 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6394 if (ret) {
6395 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6396 goto err_init_ctrlq;
6397 }
6398
6399 err = ice_sched_init_port(hw->port_info);
6400 if (err)
6401 goto err_sched_init_port;
6402
6403 /* start misc vector */
6404 err = ice_req_irq_msix_misc(pf);
6405 if (err) {
6406 dev_err(dev, "misc vector setup failed: %d\n", err);
6407 goto err_sched_init_port;
6408 }
6409
6410 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6411 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6412 if (!rd32(hw, PFQF_FD_SIZE)) {
6413 u16 unused, guar, b_effort;
6414
6415 guar = hw->func_caps.fd_fltr_guar;
6416 b_effort = hw->func_caps.fd_fltr_best_effort;
6417
6418 /* force guaranteed filter pool for PF */
6419 ice_alloc_fd_guar_item(hw, &unused, guar);
6420 /* force shared filter pool for PF */
6421 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6422 }
6423 }
6424
6425 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6426 ice_dcb_rebuild(pf);
6427
6428 /* If the PF previously had enabled PTP, PTP init needs to happen before
6429 * the VSI rebuild. If not, this causes the PTP link status events to
6430 * fail.
6431 */
6432 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
6433 ice_ptp_init(pf);
6434
6435 /* rebuild PF VSI */
6436 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6437 if (err) {
6438 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6439 goto err_vsi_rebuild;
6440 }
6441
6442 /* If Flow Director is active */
6443 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6444 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6445 if (err) {
6446 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6447 goto err_vsi_rebuild;
6448 }
6449
6450 /* replay HW Flow Director recipes */
6451 if (hw->fdir_prof)
6452 ice_fdir_replay_flows(hw);
6453
6454 /* replay Flow Director filters */
6455 ice_fdir_replay_fltrs(pf);
6456
6457 ice_rebuild_arfs(pf);
6458 }
6459
6460 ice_update_pf_netdev_link(pf);
6461
6462 /* tell the firmware we are up */
6463 ret = ice_send_version(pf);
6464 if (ret) {
6465 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6466 ice_stat_str(ret));
6467 goto err_vsi_rebuild;
6468 }
6469
6470 ice_replay_post(hw);
6471
6472 /* if we get here, reset flow is successful */
6473 clear_bit(ICE_RESET_FAILED, pf->state);
6474
6475 ice_plug_aux_dev(pf);
6476 return;
6477
6478err_vsi_rebuild:
6479err_sched_init_port:
6480 ice_sched_cleanup_all(hw);
6481err_init_ctrlq:
6482 ice_shutdown_all_ctrlq(hw);
6483 set_bit(ICE_RESET_FAILED, pf->state);
6484clear_recovery:
6485 /* set this bit in PF state to control service task scheduling */
6486 set_bit(ICE_NEEDS_RESTART, pf->state);
6487 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6488}
6489
6490/**
6491 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6492 * @vsi: Pointer to VSI structure
6493 */
6494static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6495{
6496 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6497 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6498 else
6499 return ICE_RXBUF_3072;
6500}
6501
6502/**
6503 * ice_change_mtu - NDO callback to change the MTU
6504 * @netdev: network interface device structure
6505 * @new_mtu: new value for maximum frame size
6506 *
6507 * Returns 0 on success, negative on failure
6508 */
6509static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6510{
6511 struct ice_netdev_priv *np = netdev_priv(netdev);
6512 struct ice_vsi *vsi = np->vsi;
6513 struct ice_pf *pf = vsi->back;
6514 struct iidc_event *event;
6515 u8 count = 0;
6516 int err = 0;
6517
6518 if (new_mtu == (int)netdev->mtu) {
6519 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6520 return 0;
6521 }
6522
6523 if (ice_is_xdp_ena_vsi(vsi)) {
6524 int frame_size = ice_max_xdp_frame_size(vsi);
6525
6526 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6527 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6528 frame_size - ICE_ETH_PKT_HDR_PAD);
6529 return -EINVAL;
6530 }
6531 }
6532
6533 /* if a reset is in progress, wait for some time for it to complete */
6534 do {
6535 if (ice_is_reset_in_progress(pf->state)) {
6536 count++;
6537 usleep_range(1000, 2000);
6538 } else {
6539 break;
6540 }
6541
6542 } while (count < 100);
6543
6544 if (count == 100) {
6545 netdev_err(netdev, "can't change MTU. Device is busy\n");
6546 return -EBUSY;
6547 }
6548
6549 event = kzalloc(sizeof(*event), GFP_KERNEL);
6550 if (!event)
6551 return -ENOMEM;
6552
6553 set_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6554 ice_send_event_to_aux(pf, event);
6555 clear_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6556
6557 netdev->mtu = (unsigned int)new_mtu;
6558
6559 /* if VSI is up, bring it down and then back up */
6560 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6561 err = ice_down(vsi);
6562 if (err) {
6563 netdev_err(netdev, "change MTU if_down err %d\n", err);
6564 goto event_after;
6565 }
6566
6567 err = ice_up(vsi);
6568 if (err) {
6569 netdev_err(netdev, "change MTU if_up err %d\n", err);
6570 goto event_after;
6571 }
6572 }
6573
6574 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6575event_after:
6576 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
6577 ice_send_event_to_aux(pf, event);
6578 kfree(event);
6579
6580 return err;
6581}
6582
6583/**
6584 * ice_do_ioctl - Access the hwtstamp interface
6585 * @netdev: network interface device structure
6586 * @ifr: interface request data
6587 * @cmd: ioctl command
6588 */
6589static int ice_do_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6590{
6591 struct ice_netdev_priv *np = netdev_priv(netdev);
6592 struct ice_pf *pf = np->vsi->back;
6593
6594 switch (cmd) {
6595 case SIOCGHWTSTAMP:
6596 return ice_ptp_get_ts_config(pf, ifr);
6597 case SIOCSHWTSTAMP:
6598 return ice_ptp_set_ts_config(pf, ifr);
6599 default:
6600 return -EOPNOTSUPP;
6601 }
6602}
6603
6604/**
6605 * ice_aq_str - convert AQ err code to a string
6606 * @aq_err: the AQ error code to convert
6607 */
6608const char *ice_aq_str(enum ice_aq_err aq_err)
6609{
6610 switch (aq_err) {
6611 case ICE_AQ_RC_OK:
6612 return "OK";
6613 case ICE_AQ_RC_EPERM:
6614 return "ICE_AQ_RC_EPERM";
6615 case ICE_AQ_RC_ENOENT:
6616 return "ICE_AQ_RC_ENOENT";
6617 case ICE_AQ_RC_ENOMEM:
6618 return "ICE_AQ_RC_ENOMEM";
6619 case ICE_AQ_RC_EBUSY:
6620 return "ICE_AQ_RC_EBUSY";
6621 case ICE_AQ_RC_EEXIST:
6622 return "ICE_AQ_RC_EEXIST";
6623 case ICE_AQ_RC_EINVAL:
6624 return "ICE_AQ_RC_EINVAL";
6625 case ICE_AQ_RC_ENOSPC:
6626 return "ICE_AQ_RC_ENOSPC";
6627 case ICE_AQ_RC_ENOSYS:
6628 return "ICE_AQ_RC_ENOSYS";
6629 case ICE_AQ_RC_EMODE:
6630 return "ICE_AQ_RC_EMODE";
6631 case ICE_AQ_RC_ENOSEC:
6632 return "ICE_AQ_RC_ENOSEC";
6633 case ICE_AQ_RC_EBADSIG:
6634 return "ICE_AQ_RC_EBADSIG";
6635 case ICE_AQ_RC_ESVN:
6636 return "ICE_AQ_RC_ESVN";
6637 case ICE_AQ_RC_EBADMAN:
6638 return "ICE_AQ_RC_EBADMAN";
6639 case ICE_AQ_RC_EBADBUF:
6640 return "ICE_AQ_RC_EBADBUF";
6641 }
6642
6643 return "ICE_AQ_RC_UNKNOWN";
6644}
6645
6646/**
6647 * ice_stat_str - convert status err code to a string
6648 * @stat_err: the status error code to convert
6649 */
6650const char *ice_stat_str(enum ice_status stat_err)
6651{
6652 switch (stat_err) {
6653 case ICE_SUCCESS:
6654 return "OK";
6655 case ICE_ERR_PARAM:
6656 return "ICE_ERR_PARAM";
6657 case ICE_ERR_NOT_IMPL:
6658 return "ICE_ERR_NOT_IMPL";
6659 case ICE_ERR_NOT_READY:
6660 return "ICE_ERR_NOT_READY";
6661 case ICE_ERR_NOT_SUPPORTED:
6662 return "ICE_ERR_NOT_SUPPORTED";
6663 case ICE_ERR_BAD_PTR:
6664 return "ICE_ERR_BAD_PTR";
6665 case ICE_ERR_INVAL_SIZE:
6666 return "ICE_ERR_INVAL_SIZE";
6667 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6668 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6669 case ICE_ERR_RESET_FAILED:
6670 return "ICE_ERR_RESET_FAILED";
6671 case ICE_ERR_FW_API_VER:
6672 return "ICE_ERR_FW_API_VER";
6673 case ICE_ERR_NO_MEMORY:
6674 return "ICE_ERR_NO_MEMORY";
6675 case ICE_ERR_CFG:
6676 return "ICE_ERR_CFG";
6677 case ICE_ERR_OUT_OF_RANGE:
6678 return "ICE_ERR_OUT_OF_RANGE";
6679 case ICE_ERR_ALREADY_EXISTS:
6680 return "ICE_ERR_ALREADY_EXISTS";
6681 case ICE_ERR_NVM:
6682 return "ICE_ERR_NVM";
6683 case ICE_ERR_NVM_CHECKSUM:
6684 return "ICE_ERR_NVM_CHECKSUM";
6685 case ICE_ERR_BUF_TOO_SHORT:
6686 return "ICE_ERR_BUF_TOO_SHORT";
6687 case ICE_ERR_NVM_BLANK_MODE:
6688 return "ICE_ERR_NVM_BLANK_MODE";
6689 case ICE_ERR_IN_USE:
6690 return "ICE_ERR_IN_USE";
6691 case ICE_ERR_MAX_LIMIT:
6692 return "ICE_ERR_MAX_LIMIT";
6693 case ICE_ERR_RESET_ONGOING:
6694 return "ICE_ERR_RESET_ONGOING";
6695 case ICE_ERR_HW_TABLE:
6696 return "ICE_ERR_HW_TABLE";
6697 case ICE_ERR_DOES_NOT_EXIST:
6698 return "ICE_ERR_DOES_NOT_EXIST";
6699 case ICE_ERR_FW_DDP_MISMATCH:
6700 return "ICE_ERR_FW_DDP_MISMATCH";
6701 case ICE_ERR_AQ_ERROR:
6702 return "ICE_ERR_AQ_ERROR";
6703 case ICE_ERR_AQ_TIMEOUT:
6704 return "ICE_ERR_AQ_TIMEOUT";
6705 case ICE_ERR_AQ_FULL:
6706 return "ICE_ERR_AQ_FULL";
6707 case ICE_ERR_AQ_NO_WORK:
6708 return "ICE_ERR_AQ_NO_WORK";
6709 case ICE_ERR_AQ_EMPTY:
6710 return "ICE_ERR_AQ_EMPTY";
6711 case ICE_ERR_AQ_FW_CRITICAL:
6712 return "ICE_ERR_AQ_FW_CRITICAL";
6713 }
6714
6715 return "ICE_ERR_UNKNOWN";
6716}
6717
6718/**
6719 * ice_set_rss_lut - Set RSS LUT
6720 * @vsi: Pointer to VSI structure
6721 * @lut: Lookup table
6722 * @lut_size: Lookup table size
6723 *
6724 * Returns 0 on success, negative on failure
6725 */
6726int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6727{
6728 struct ice_aq_get_set_rss_lut_params params = {};
6729 struct ice_hw *hw = &vsi->back->hw;
6730 enum ice_status status;
6731
6732 if (!lut)
6733 return -EINVAL;
6734
6735 params.vsi_handle = vsi->idx;
6736 params.lut_size = lut_size;
6737 params.lut_type = vsi->rss_lut_type;
6738 params.lut = lut;
6739
6740 status = ice_aq_set_rss_lut(hw, ¶ms);
6741 if (status) {
6742 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %s aq_err %s\n",
6743 ice_stat_str(status),
6744 ice_aq_str(hw->adminq.sq_last_status));
6745 return -EIO;
6746 }
6747
6748 return 0;
6749}
6750
6751/**
6752 * ice_set_rss_key - Set RSS key
6753 * @vsi: Pointer to the VSI structure
6754 * @seed: RSS hash seed
6755 *
6756 * Returns 0 on success, negative on failure
6757 */
6758int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
6759{
6760 struct ice_hw *hw = &vsi->back->hw;
6761 enum ice_status status;
6762
6763 if (!seed)
6764 return -EINVAL;
6765
6766 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6767 if (status) {
6768 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %s aq_err %s\n",
6769 ice_stat_str(status),
6770 ice_aq_str(hw->adminq.sq_last_status));
6771 return -EIO;
6772 }
6773
6774 return 0;
6775}
6776
6777/**
6778 * ice_get_rss_lut - Get RSS LUT
6779 * @vsi: Pointer to VSI structure
6780 * @lut: Buffer to store the lookup table entries
6781 * @lut_size: Size of buffer to store the lookup table entries
6782 *
6783 * Returns 0 on success, negative on failure
6784 */
6785int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6786{
6787 struct ice_aq_get_set_rss_lut_params params = {};
6788 struct ice_hw *hw = &vsi->back->hw;
6789 enum ice_status status;
6790
6791 if (!lut)
6792 return -EINVAL;
6793
6794 params.vsi_handle = vsi->idx;
6795 params.lut_size = lut_size;
6796 params.lut_type = vsi->rss_lut_type;
6797 params.lut = lut;
6798
6799 status = ice_aq_get_rss_lut(hw, ¶ms);
6800 if (status) {
6801 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %s aq_err %s\n",
6802 ice_stat_str(status),
6803 ice_aq_str(hw->adminq.sq_last_status));
6804 return -EIO;
6805 }
6806
6807 return 0;
6808}
6809
6810/**
6811 * ice_get_rss_key - Get RSS key
6812 * @vsi: Pointer to VSI structure
6813 * @seed: Buffer to store the key in
6814 *
6815 * Returns 0 on success, negative on failure
6816 */
6817int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
6818{
6819 struct ice_hw *hw = &vsi->back->hw;
6820 enum ice_status status;
6821
6822 if (!seed)
6823 return -EINVAL;
6824
6825 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6826 if (status) {
6827 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %s aq_err %s\n",
6828 ice_stat_str(status),
6829 ice_aq_str(hw->adminq.sq_last_status));
6830 return -EIO;
6831 }
6832
6833 return 0;
6834}
6835
6836/**
6837 * ice_bridge_getlink - Get the hardware bridge mode
6838 * @skb: skb buff
6839 * @pid: process ID
6840 * @seq: RTNL message seq
6841 * @dev: the netdev being configured
6842 * @filter_mask: filter mask passed in
6843 * @nlflags: netlink flags passed in
6844 *
6845 * Return the bridge mode (VEB/VEPA)
6846 */
6847static int
6848ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6849 struct net_device *dev, u32 filter_mask, int nlflags)
6850{
6851 struct ice_netdev_priv *np = netdev_priv(dev);
6852 struct ice_vsi *vsi = np->vsi;
6853 struct ice_pf *pf = vsi->back;
6854 u16 bmode;
6855
6856 bmode = pf->first_sw->bridge_mode;
6857
6858 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6859 filter_mask, NULL);
6860}
6861
6862/**
6863 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6864 * @vsi: Pointer to VSI structure
6865 * @bmode: Hardware bridge mode (VEB/VEPA)
6866 *
6867 * Returns 0 on success, negative on failure
6868 */
6869static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6870{
6871 struct ice_aqc_vsi_props *vsi_props;
6872 struct ice_hw *hw = &vsi->back->hw;
6873 struct ice_vsi_ctx *ctxt;
6874 enum ice_status status;
6875 int ret = 0;
6876
6877 vsi_props = &vsi->info;
6878
6879 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6880 if (!ctxt)
6881 return -ENOMEM;
6882
6883 ctxt->info = vsi->info;
6884
6885 if (bmode == BRIDGE_MODE_VEB)
6886 /* change from VEPA to VEB mode */
6887 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6888 else
6889 /* change from VEB to VEPA mode */
6890 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6891 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6892
6893 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6894 if (status) {
6895 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6896 bmode, ice_stat_str(status),
6897 ice_aq_str(hw->adminq.sq_last_status));
6898 ret = -EIO;
6899 goto out;
6900 }
6901 /* Update sw flags for book keeping */
6902 vsi_props->sw_flags = ctxt->info.sw_flags;
6903
6904out:
6905 kfree(ctxt);
6906 return ret;
6907}
6908
6909/**
6910 * ice_bridge_setlink - Set the hardware bridge mode
6911 * @dev: the netdev being configured
6912 * @nlh: RTNL message
6913 * @flags: bridge setlink flags
6914 * @extack: netlink extended ack
6915 *
6916 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6917 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6918 * not already set for all VSIs connected to this switch. And also update the
6919 * unicast switch filter rules for the corresponding switch of the netdev.
6920 */
6921static int
6922ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6923 u16 __always_unused flags,
6924 struct netlink_ext_ack __always_unused *extack)
6925{
6926 struct ice_netdev_priv *np = netdev_priv(dev);
6927 struct ice_pf *pf = np->vsi->back;
6928 struct nlattr *attr, *br_spec;
6929 struct ice_hw *hw = &pf->hw;
6930 enum ice_status status;
6931 struct ice_sw *pf_sw;
6932 int rem, v, err = 0;
6933
6934 pf_sw = pf->first_sw;
6935 /* find the attribute in the netlink message */
6936 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6937
6938 nla_for_each_nested(attr, br_spec, rem) {
6939 __u16 mode;
6940
6941 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6942 continue;
6943 mode = nla_get_u16(attr);
6944 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6945 return -EINVAL;
6946 /* Continue if bridge mode is not being flipped */
6947 if (mode == pf_sw->bridge_mode)
6948 continue;
6949 /* Iterates through the PF VSI list and update the loopback
6950 * mode of the VSI
6951 */
6952 ice_for_each_vsi(pf, v) {
6953 if (!pf->vsi[v])
6954 continue;
6955 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6956 if (err)
6957 return err;
6958 }
6959
6960 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6961 /* Update the unicast switch filter rules for the corresponding
6962 * switch of the netdev
6963 */
6964 status = ice_update_sw_rule_bridge_mode(hw);
6965 if (status) {
6966 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6967 mode, ice_stat_str(status),
6968 ice_aq_str(hw->adminq.sq_last_status));
6969 /* revert hw->evb_veb */
6970 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6971 return -EIO;
6972 }
6973
6974 pf_sw->bridge_mode = mode;
6975 }
6976
6977 return 0;
6978}
6979
6980/**
6981 * ice_tx_timeout - Respond to a Tx Hang
6982 * @netdev: network interface device structure
6983 * @txqueue: Tx queue
6984 */
6985static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6986{
6987 struct ice_netdev_priv *np = netdev_priv(netdev);
6988 struct ice_ring *tx_ring = NULL;
6989 struct ice_vsi *vsi = np->vsi;
6990 struct ice_pf *pf = vsi->back;
6991 u32 i;
6992
6993 pf->tx_timeout_count++;
6994
6995 /* Check if PFC is enabled for the TC to which the queue belongs
6996 * to. If yes then Tx timeout is not caused by a hung queue, no
6997 * need to reset and rebuild
6998 */
6999 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7000 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7001 txqueue);
7002 return;
7003 }
7004
7005 /* now that we have an index, find the tx_ring struct */
7006 for (i = 0; i < vsi->num_txq; i++)
7007 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7008 if (txqueue == vsi->tx_rings[i]->q_index) {
7009 tx_ring = vsi->tx_rings[i];
7010 break;
7011 }
7012
7013 /* Reset recovery level if enough time has elapsed after last timeout.
7014 * Also ensure no new reset action happens before next timeout period.
7015 */
7016 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7017 pf->tx_timeout_recovery_level = 1;
7018 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7019 netdev->watchdog_timeo)))
7020 return;
7021
7022 if (tx_ring) {
7023 struct ice_hw *hw = &pf->hw;
7024 u32 head, val = 0;
7025
7026 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7027 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7028 /* Read interrupt register */
7029 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7030
7031 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7032 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7033 head, tx_ring->next_to_use, val);
7034 }
7035
7036 pf->tx_timeout_last_recovery = jiffies;
7037 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7038 pf->tx_timeout_recovery_level, txqueue);
7039
7040 switch (pf->tx_timeout_recovery_level) {
7041 case 1:
7042 set_bit(ICE_PFR_REQ, pf->state);
7043 break;
7044 case 2:
7045 set_bit(ICE_CORER_REQ, pf->state);
7046 break;
7047 case 3:
7048 set_bit(ICE_GLOBR_REQ, pf->state);
7049 break;
7050 default:
7051 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7052 set_bit(ICE_DOWN, pf->state);
7053 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7054 set_bit(ICE_SERVICE_DIS, pf->state);
7055 break;
7056 }
7057
7058 ice_service_task_schedule(pf);
7059 pf->tx_timeout_recovery_level++;
7060}
7061
7062/**
7063 * ice_open - Called when a network interface becomes active
7064 * @netdev: network interface device structure
7065 *
7066 * The open entry point is called when a network interface is made
7067 * active by the system (IFF_UP). At this point all resources needed
7068 * for transmit and receive operations are allocated, the interrupt
7069 * handler is registered with the OS, the netdev watchdog is enabled,
7070 * and the stack is notified that the interface is ready.
7071 *
7072 * Returns 0 on success, negative value on failure
7073 */
7074int ice_open(struct net_device *netdev)
7075{
7076 struct ice_netdev_priv *np = netdev_priv(netdev);
7077 struct ice_pf *pf = np->vsi->back;
7078
7079 if (ice_is_reset_in_progress(pf->state)) {
7080 netdev_err(netdev, "can't open net device while reset is in progress");
7081 return -EBUSY;
7082 }
7083
7084 return ice_open_internal(netdev);
7085}
7086
7087/**
7088 * ice_open_internal - Called when a network interface becomes active
7089 * @netdev: network interface device structure
7090 *
7091 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
7092 * handling routine
7093 *
7094 * Returns 0 on success, negative value on failure
7095 */
7096int ice_open_internal(struct net_device *netdev)
7097{
7098 struct ice_netdev_priv *np = netdev_priv(netdev);
7099 struct ice_vsi *vsi = np->vsi;
7100 struct ice_pf *pf = vsi->back;
7101 struct ice_port_info *pi;
7102 enum ice_status status;
7103 int err;
7104
7105 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
7106 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
7107 return -EIO;
7108 }
7109
7110 netif_carrier_off(netdev);
7111
7112 pi = vsi->port_info;
7113 status = ice_update_link_info(pi);
7114 if (status) {
7115 netdev_err(netdev, "Failed to get link info, error %s\n",
7116 ice_stat_str(status));
7117 return -EIO;
7118 }
7119
7120 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
7121
7122 /* Set PHY if there is media, otherwise, turn off PHY */
7123 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
7124 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
7125 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
7126 err = ice_init_phy_user_cfg(pi);
7127 if (err) {
7128 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
7129 err);
7130 return err;
7131 }
7132 }
7133
7134 err = ice_configure_phy(vsi);
7135 if (err) {
7136 netdev_err(netdev, "Failed to set physical link up, error %d\n",
7137 err);
7138 return err;
7139 }
7140 } else {
7141 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
7142 ice_set_link(vsi, false);
7143 }
7144
7145 err = ice_vsi_open(vsi);
7146 if (err)
7147 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
7148 vsi->vsi_num, vsi->vsw->sw_id);
7149
7150 /* Update existing tunnels information */
7151 udp_tunnel_get_rx_info(netdev);
7152
7153 return err;
7154}
7155
7156/**
7157 * ice_stop - Disables a network interface
7158 * @netdev: network interface device structure
7159 *
7160 * The stop entry point is called when an interface is de-activated by the OS,
7161 * and the netdevice enters the DOWN state. The hardware is still under the
7162 * driver's control, but the netdev interface is disabled.
7163 *
7164 * Returns success only - not allowed to fail
7165 */
7166int ice_stop(struct net_device *netdev)
7167{
7168 struct ice_netdev_priv *np = netdev_priv(netdev);
7169 struct ice_vsi *vsi = np->vsi;
7170 struct ice_pf *pf = vsi->back;
7171
7172 if (ice_is_reset_in_progress(pf->state)) {
7173 netdev_err(netdev, "can't stop net device while reset is in progress");
7174 return -EBUSY;
7175 }
7176
7177 ice_vsi_close(vsi);
7178
7179 return 0;
7180}
7181
7182/**
7183 * ice_features_check - Validate encapsulated packet conforms to limits
7184 * @skb: skb buffer
7185 * @netdev: This port's netdev
7186 * @features: Offload features that the stack believes apply
7187 */
7188static netdev_features_t
7189ice_features_check(struct sk_buff *skb,
7190 struct net_device __always_unused *netdev,
7191 netdev_features_t features)
7192{
7193 size_t len;
7194
7195 /* No point in doing any of this if neither checksum nor GSO are
7196 * being requested for this frame. We can rule out both by just
7197 * checking for CHECKSUM_PARTIAL
7198 */
7199 if (skb->ip_summed != CHECKSUM_PARTIAL)
7200 return features;
7201
7202 /* We cannot support GSO if the MSS is going to be less than
7203 * 64 bytes. If it is then we need to drop support for GSO.
7204 */
7205 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
7206 features &= ~NETIF_F_GSO_MASK;
7207
7208 len = skb_network_header(skb) - skb->data;
7209 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
7210 goto out_rm_features;
7211
7212 len = skb_transport_header(skb) - skb_network_header(skb);
7213 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7214 goto out_rm_features;
7215
7216 if (skb->encapsulation) {
7217 len = skb_inner_network_header(skb) - skb_transport_header(skb);
7218 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
7219 goto out_rm_features;
7220
7221 len = skb_inner_transport_header(skb) -
7222 skb_inner_network_header(skb);
7223 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7224 goto out_rm_features;
7225 }
7226
7227 return features;
7228out_rm_features:
7229 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
7230}
7231
7232static const struct net_device_ops ice_netdev_safe_mode_ops = {
7233 .ndo_open = ice_open,
7234 .ndo_stop = ice_stop,
7235 .ndo_start_xmit = ice_start_xmit,
7236 .ndo_set_mac_address = ice_set_mac_address,
7237 .ndo_validate_addr = eth_validate_addr,
7238 .ndo_change_mtu = ice_change_mtu,
7239 .ndo_get_stats64 = ice_get_stats64,
7240 .ndo_tx_timeout = ice_tx_timeout,
7241 .ndo_bpf = ice_xdp_safe_mode,
7242};
7243
7244static const struct net_device_ops ice_netdev_ops = {
7245 .ndo_open = ice_open,
7246 .ndo_stop = ice_stop,
7247 .ndo_start_xmit = ice_start_xmit,
7248 .ndo_features_check = ice_features_check,
7249 .ndo_set_rx_mode = ice_set_rx_mode,
7250 .ndo_set_mac_address = ice_set_mac_address,
7251 .ndo_validate_addr = eth_validate_addr,
7252 .ndo_change_mtu = ice_change_mtu,
7253 .ndo_get_stats64 = ice_get_stats64,
7254 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
7255 .ndo_do_ioctl = ice_do_ioctl,
7256 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
7257 .ndo_set_vf_mac = ice_set_vf_mac,
7258 .ndo_get_vf_config = ice_get_vf_cfg,
7259 .ndo_set_vf_trust = ice_set_vf_trust,
7260 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
7261 .ndo_set_vf_link_state = ice_set_vf_link_state,
7262 .ndo_get_vf_stats = ice_get_vf_stats,
7263 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
7264 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
7265 .ndo_set_features = ice_set_features,
7266 .ndo_bridge_getlink = ice_bridge_getlink,
7267 .ndo_bridge_setlink = ice_bridge_setlink,
7268 .ndo_fdb_add = ice_fdb_add,
7269 .ndo_fdb_del = ice_fdb_del,
7270#ifdef CONFIG_RFS_ACCEL
7271 .ndo_rx_flow_steer = ice_rx_flow_steer,
7272#endif
7273 .ndo_tx_timeout = ice_tx_timeout,
7274 .ndo_bpf = ice_xdp,
7275 .ndo_xdp_xmit = ice_xdp_xmit,
7276 .ndo_xsk_wakeup = ice_xsk_wakeup,
7277};