Loading...
Note: File does not exist in v4.6.
1// SPDX-License-Identifier: GPL-2.0
2/* Copyright (c) 2018-2023, Intel Corporation. */
3
4/* Intel(R) Ethernet Connection E800 Series Linux Driver */
5
6#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
7
8#include <generated/utsrelease.h>
9#include <linux/crash_dump.h>
10#include "ice.h"
11#include "ice_base.h"
12#include "ice_lib.h"
13#include "ice_fltr.h"
14#include "ice_dcb_lib.h"
15#include "ice_dcb_nl.h"
16#include "ice_devlink.h"
17#include "ice_hwmon.h"
18/* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
19 * ice tracepoint functions. This must be done exactly once across the
20 * ice driver.
21 */
22#define CREATE_TRACE_POINTS
23#include "ice_trace.h"
24#include "ice_eswitch.h"
25#include "ice_tc_lib.h"
26#include "ice_vsi_vlan_ops.h"
27#include <net/xdp_sock_drv.h>
28
29#define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
30static const char ice_driver_string[] = DRV_SUMMARY;
31static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
32
33/* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
34#define ICE_DDP_PKG_PATH "intel/ice/ddp/"
35#define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
36
37MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
38MODULE_DESCRIPTION(DRV_SUMMARY);
39MODULE_LICENSE("GPL v2");
40MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
41
42static int debug = -1;
43module_param(debug, int, 0644);
44#ifndef CONFIG_DYNAMIC_DEBUG
45MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
46#else
47MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
48#endif /* !CONFIG_DYNAMIC_DEBUG */
49
50DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key);
51EXPORT_SYMBOL(ice_xdp_locking_key);
52
53/**
54 * ice_hw_to_dev - Get device pointer from the hardware structure
55 * @hw: pointer to the device HW structure
56 *
57 * Used to access the device pointer from compilation units which can't easily
58 * include the definition of struct ice_pf without leading to circular header
59 * dependencies.
60 */
61struct device *ice_hw_to_dev(struct ice_hw *hw)
62{
63 struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
64
65 return &pf->pdev->dev;
66}
67
68static struct workqueue_struct *ice_wq;
69struct workqueue_struct *ice_lag_wq;
70static const struct net_device_ops ice_netdev_safe_mode_ops;
71static const struct net_device_ops ice_netdev_ops;
72
73static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
74
75static void ice_vsi_release_all(struct ice_pf *pf);
76
77static int ice_rebuild_channels(struct ice_pf *pf);
78static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr);
79
80static int
81ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
82 void *cb_priv, enum tc_setup_type type, void *type_data,
83 void *data,
84 void (*cleanup)(struct flow_block_cb *block_cb));
85
86bool netif_is_ice(const struct net_device *dev)
87{
88 return dev && (dev->netdev_ops == &ice_netdev_ops);
89}
90
91/**
92 * ice_get_tx_pending - returns number of Tx descriptors not processed
93 * @ring: the ring of descriptors
94 */
95static u16 ice_get_tx_pending(struct ice_tx_ring *ring)
96{
97 u16 head, tail;
98
99 head = ring->next_to_clean;
100 tail = ring->next_to_use;
101
102 if (head != tail)
103 return (head < tail) ?
104 tail - head : (tail + ring->count - head);
105 return 0;
106}
107
108/**
109 * ice_check_for_hang_subtask - check for and recover hung queues
110 * @pf: pointer to PF struct
111 */
112static void ice_check_for_hang_subtask(struct ice_pf *pf)
113{
114 struct ice_vsi *vsi = NULL;
115 struct ice_hw *hw;
116 unsigned int i;
117 int packets;
118 u32 v;
119
120 ice_for_each_vsi(pf, v)
121 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
122 vsi = pf->vsi[v];
123 break;
124 }
125
126 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
127 return;
128
129 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
130 return;
131
132 hw = &vsi->back->hw;
133
134 ice_for_each_txq(vsi, i) {
135 struct ice_tx_ring *tx_ring = vsi->tx_rings[i];
136 struct ice_ring_stats *ring_stats;
137
138 if (!tx_ring)
139 continue;
140 if (ice_ring_ch_enabled(tx_ring))
141 continue;
142
143 ring_stats = tx_ring->ring_stats;
144 if (!ring_stats)
145 continue;
146
147 if (tx_ring->desc) {
148 /* If packet counter has not changed the queue is
149 * likely stalled, so force an interrupt for this
150 * queue.
151 *
152 * prev_pkt would be negative if there was no
153 * pending work.
154 */
155 packets = ring_stats->stats.pkts & INT_MAX;
156 if (ring_stats->tx_stats.prev_pkt == packets) {
157 /* Trigger sw interrupt to revive the queue */
158 ice_trigger_sw_intr(hw, tx_ring->q_vector);
159 continue;
160 }
161
162 /* Memory barrier between read of packet count and call
163 * to ice_get_tx_pending()
164 */
165 smp_rmb();
166 ring_stats->tx_stats.prev_pkt =
167 ice_get_tx_pending(tx_ring) ? packets : -1;
168 }
169 }
170}
171
172/**
173 * ice_init_mac_fltr - Set initial MAC filters
174 * @pf: board private structure
175 *
176 * Set initial set of MAC filters for PF VSI; configure filters for permanent
177 * address and broadcast address. If an error is encountered, netdevice will be
178 * unregistered.
179 */
180static int ice_init_mac_fltr(struct ice_pf *pf)
181{
182 struct ice_vsi *vsi;
183 u8 *perm_addr;
184
185 vsi = ice_get_main_vsi(pf);
186 if (!vsi)
187 return -EINVAL;
188
189 perm_addr = vsi->port_info->mac.perm_addr;
190 return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
191}
192
193/**
194 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
195 * @netdev: the net device on which the sync is happening
196 * @addr: MAC address to sync
197 *
198 * This is a callback function which is called by the in kernel device sync
199 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
200 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
201 * MAC filters from the hardware.
202 */
203static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
204{
205 struct ice_netdev_priv *np = netdev_priv(netdev);
206 struct ice_vsi *vsi = np->vsi;
207
208 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
209 ICE_FWD_TO_VSI))
210 return -EINVAL;
211
212 return 0;
213}
214
215/**
216 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
217 * @netdev: the net device on which the unsync is happening
218 * @addr: MAC address to unsync
219 *
220 * This is a callback function which is called by the in kernel device unsync
221 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
222 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
223 * delete the MAC filters from the hardware.
224 */
225static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
226{
227 struct ice_netdev_priv *np = netdev_priv(netdev);
228 struct ice_vsi *vsi = np->vsi;
229
230 /* Under some circumstances, we might receive a request to delete our
231 * own device address from our uc list. Because we store the device
232 * address in the VSI's MAC filter list, we need to ignore such
233 * requests and not delete our device address from this list.
234 */
235 if (ether_addr_equal(addr, netdev->dev_addr))
236 return 0;
237
238 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
239 ICE_FWD_TO_VSI))
240 return -EINVAL;
241
242 return 0;
243}
244
245/**
246 * ice_vsi_fltr_changed - check if filter state changed
247 * @vsi: VSI to be checked
248 *
249 * returns true if filter state has changed, false otherwise.
250 */
251static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
252{
253 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
254 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
255}
256
257/**
258 * ice_set_promisc - Enable promiscuous mode for a given PF
259 * @vsi: the VSI being configured
260 * @promisc_m: mask of promiscuous config bits
261 *
262 */
263static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
264{
265 int status;
266
267 if (vsi->type != ICE_VSI_PF)
268 return 0;
269
270 if (ice_vsi_has_non_zero_vlans(vsi)) {
271 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
272 status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi,
273 promisc_m);
274 } else {
275 status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
276 promisc_m, 0);
277 }
278 if (status && status != -EEXIST)
279 return status;
280
281 netdev_dbg(vsi->netdev, "set promisc filter bits for VSI %i: 0x%x\n",
282 vsi->vsi_num, promisc_m);
283 return 0;
284}
285
286/**
287 * ice_clear_promisc - Disable promiscuous mode for a given PF
288 * @vsi: the VSI being configured
289 * @promisc_m: mask of promiscuous config bits
290 *
291 */
292static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
293{
294 int status;
295
296 if (vsi->type != ICE_VSI_PF)
297 return 0;
298
299 if (ice_vsi_has_non_zero_vlans(vsi)) {
300 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
301 status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi,
302 promisc_m);
303 } else {
304 status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
305 promisc_m, 0);
306 }
307
308 netdev_dbg(vsi->netdev, "clear promisc filter bits for VSI %i: 0x%x\n",
309 vsi->vsi_num, promisc_m);
310 return status;
311}
312
313/**
314 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
315 * @vsi: ptr to the VSI
316 *
317 * Push any outstanding VSI filter changes through the AdminQ.
318 */
319static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
320{
321 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
322 struct device *dev = ice_pf_to_dev(vsi->back);
323 struct net_device *netdev = vsi->netdev;
324 bool promisc_forced_on = false;
325 struct ice_pf *pf = vsi->back;
326 struct ice_hw *hw = &pf->hw;
327 u32 changed_flags = 0;
328 int err;
329
330 if (!vsi->netdev)
331 return -EINVAL;
332
333 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
334 usleep_range(1000, 2000);
335
336 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
337 vsi->current_netdev_flags = vsi->netdev->flags;
338
339 INIT_LIST_HEAD(&vsi->tmp_sync_list);
340 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
341
342 if (ice_vsi_fltr_changed(vsi)) {
343 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
344 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
345
346 /* grab the netdev's addr_list_lock */
347 netif_addr_lock_bh(netdev);
348 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
349 ice_add_mac_to_unsync_list);
350 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
351 ice_add_mac_to_unsync_list);
352 /* our temp lists are populated. release lock */
353 netif_addr_unlock_bh(netdev);
354 }
355
356 /* Remove MAC addresses in the unsync list */
357 err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
358 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
359 if (err) {
360 netdev_err(netdev, "Failed to delete MAC filters\n");
361 /* if we failed because of alloc failures, just bail */
362 if (err == -ENOMEM)
363 goto out;
364 }
365
366 /* Add MAC addresses in the sync list */
367 err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
368 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
369 /* If filter is added successfully or already exists, do not go into
370 * 'if' condition and report it as error. Instead continue processing
371 * rest of the function.
372 */
373 if (err && err != -EEXIST) {
374 netdev_err(netdev, "Failed to add MAC filters\n");
375 /* If there is no more space for new umac filters, VSI
376 * should go into promiscuous mode. There should be some
377 * space reserved for promiscuous filters.
378 */
379 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
380 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
381 vsi->state)) {
382 promisc_forced_on = true;
383 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
384 vsi->vsi_num);
385 } else {
386 goto out;
387 }
388 }
389 err = 0;
390 /* check for changes in promiscuous modes */
391 if (changed_flags & IFF_ALLMULTI) {
392 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
393 err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS);
394 if (err) {
395 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
396 goto out_promisc;
397 }
398 } else {
399 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
400 err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS);
401 if (err) {
402 vsi->current_netdev_flags |= IFF_ALLMULTI;
403 goto out_promisc;
404 }
405 }
406 }
407
408 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
409 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
410 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
411 if (vsi->current_netdev_flags & IFF_PROMISC) {
412 /* Apply Rx filter rule to get traffic from wire */
413 if (!ice_is_dflt_vsi_in_use(vsi->port_info)) {
414 err = ice_set_dflt_vsi(vsi);
415 if (err && err != -EEXIST) {
416 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
417 err, vsi->vsi_num);
418 vsi->current_netdev_flags &=
419 ~IFF_PROMISC;
420 goto out_promisc;
421 }
422 err = 0;
423 vlan_ops->dis_rx_filtering(vsi);
424
425 /* promiscuous mode implies allmulticast so
426 * that VSIs that are in promiscuous mode are
427 * subscribed to multicast packets coming to
428 * the port
429 */
430 err = ice_set_promisc(vsi,
431 ICE_MCAST_PROMISC_BITS);
432 if (err)
433 goto out_promisc;
434 }
435 } else {
436 /* Clear Rx filter to remove traffic from wire */
437 if (ice_is_vsi_dflt_vsi(vsi)) {
438 err = ice_clear_dflt_vsi(vsi);
439 if (err) {
440 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
441 err, vsi->vsi_num);
442 vsi->current_netdev_flags |=
443 IFF_PROMISC;
444 goto out_promisc;
445 }
446 if (vsi->netdev->features &
447 NETIF_F_HW_VLAN_CTAG_FILTER)
448 vlan_ops->ena_rx_filtering(vsi);
449 }
450
451 /* disable allmulti here, but only if allmulti is not
452 * still enabled for the netdev
453 */
454 if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) {
455 err = ice_clear_promisc(vsi,
456 ICE_MCAST_PROMISC_BITS);
457 if (err) {
458 netdev_err(netdev, "Error %d clearing multicast promiscuous on VSI %i\n",
459 err, vsi->vsi_num);
460 }
461 }
462 }
463 }
464 goto exit;
465
466out_promisc:
467 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
468 goto exit;
469out:
470 /* if something went wrong then set the changed flag so we try again */
471 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
472 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
473exit:
474 clear_bit(ICE_CFG_BUSY, vsi->state);
475 return err;
476}
477
478/**
479 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
480 * @pf: board private structure
481 */
482static void ice_sync_fltr_subtask(struct ice_pf *pf)
483{
484 int v;
485
486 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
487 return;
488
489 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
490
491 ice_for_each_vsi(pf, v)
492 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
493 ice_vsi_sync_fltr(pf->vsi[v])) {
494 /* come back and try again later */
495 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
496 break;
497 }
498}
499
500/**
501 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
502 * @pf: the PF
503 * @locked: is the rtnl_lock already held
504 */
505static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
506{
507 int node;
508 int v;
509
510 ice_for_each_vsi(pf, v)
511 if (pf->vsi[v])
512 ice_dis_vsi(pf->vsi[v], locked);
513
514 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
515 pf->pf_agg_node[node].num_vsis = 0;
516
517 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
518 pf->vf_agg_node[node].num_vsis = 0;
519}
520
521/**
522 * ice_clear_sw_switch_recipes - clear switch recipes
523 * @pf: board private structure
524 *
525 * Mark switch recipes as not created in sw structures. There are cases where
526 * rules (especially advanced rules) need to be restored, either re-read from
527 * hardware or added again. For example after the reset. 'recp_created' flag
528 * prevents from doing that and need to be cleared upfront.
529 */
530static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
531{
532 struct ice_sw_recipe *recp;
533 u8 i;
534
535 recp = pf->hw.switch_info->recp_list;
536 for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
537 recp[i].recp_created = false;
538}
539
540/**
541 * ice_prepare_for_reset - prep for reset
542 * @pf: board private structure
543 * @reset_type: reset type requested
544 *
545 * Inform or close all dependent features in prep for reset.
546 */
547static void
548ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
549{
550 struct ice_hw *hw = &pf->hw;
551 struct ice_vsi *vsi;
552 struct ice_vf *vf;
553 unsigned int bkt;
554
555 dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
556
557 /* already prepared for reset */
558 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
559 return;
560
561 ice_unplug_aux_dev(pf);
562
563 /* Notify VFs of impending reset */
564 if (ice_check_sq_alive(hw, &hw->mailboxq))
565 ice_vc_notify_reset(pf);
566
567 /* Disable VFs until reset is completed */
568 mutex_lock(&pf->vfs.table_lock);
569 ice_for_each_vf(pf, bkt, vf)
570 ice_set_vf_state_dis(vf);
571 mutex_unlock(&pf->vfs.table_lock);
572
573 if (ice_is_eswitch_mode_switchdev(pf)) {
574 if (reset_type != ICE_RESET_PFR)
575 ice_clear_sw_switch_recipes(pf);
576 }
577
578 /* release ADQ specific HW and SW resources */
579 vsi = ice_get_main_vsi(pf);
580 if (!vsi)
581 goto skip;
582
583 /* to be on safe side, reset orig_rss_size so that normal flow
584 * of deciding rss_size can take precedence
585 */
586 vsi->orig_rss_size = 0;
587
588 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
589 if (reset_type == ICE_RESET_PFR) {
590 vsi->old_ena_tc = vsi->all_enatc;
591 vsi->old_numtc = vsi->all_numtc;
592 } else {
593 ice_remove_q_channels(vsi, true);
594
595 /* for other reset type, do not support channel rebuild
596 * hence reset needed info
597 */
598 vsi->old_ena_tc = 0;
599 vsi->all_enatc = 0;
600 vsi->old_numtc = 0;
601 vsi->all_numtc = 0;
602 vsi->req_txq = 0;
603 vsi->req_rxq = 0;
604 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
605 memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
606 }
607 }
608skip:
609
610 /* clear SW filtering DB */
611 ice_clear_hw_tbls(hw);
612 /* disable the VSIs and their queues that are not already DOWN */
613 ice_pf_dis_all_vsi(pf, false);
614
615 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
616 ice_ptp_prepare_for_reset(pf);
617
618 if (ice_is_feature_supported(pf, ICE_F_GNSS))
619 ice_gnss_exit(pf);
620
621 if (hw->port_info)
622 ice_sched_clear_port(hw->port_info);
623
624 ice_shutdown_all_ctrlq(hw);
625
626 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
627}
628
629/**
630 * ice_do_reset - Initiate one of many types of resets
631 * @pf: board private structure
632 * @reset_type: reset type requested before this function was called.
633 */
634static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
635{
636 struct device *dev = ice_pf_to_dev(pf);
637 struct ice_hw *hw = &pf->hw;
638
639 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
640
641 if (pf->lag && pf->lag->bonded && reset_type == ICE_RESET_PFR) {
642 dev_dbg(dev, "PFR on a bonded interface, promoting to CORER\n");
643 reset_type = ICE_RESET_CORER;
644 }
645
646 ice_prepare_for_reset(pf, reset_type);
647
648 /* trigger the reset */
649 if (ice_reset(hw, reset_type)) {
650 dev_err(dev, "reset %d failed\n", reset_type);
651 set_bit(ICE_RESET_FAILED, pf->state);
652 clear_bit(ICE_RESET_OICR_RECV, pf->state);
653 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
654 clear_bit(ICE_PFR_REQ, pf->state);
655 clear_bit(ICE_CORER_REQ, pf->state);
656 clear_bit(ICE_GLOBR_REQ, pf->state);
657 wake_up(&pf->reset_wait_queue);
658 return;
659 }
660
661 /* PFR is a bit of a special case because it doesn't result in an OICR
662 * interrupt. So for PFR, rebuild after the reset and clear the reset-
663 * associated state bits.
664 */
665 if (reset_type == ICE_RESET_PFR) {
666 pf->pfr_count++;
667 ice_rebuild(pf, reset_type);
668 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
669 clear_bit(ICE_PFR_REQ, pf->state);
670 wake_up(&pf->reset_wait_queue);
671 ice_reset_all_vfs(pf);
672 }
673}
674
675/**
676 * ice_reset_subtask - Set up for resetting the device and driver
677 * @pf: board private structure
678 */
679static void ice_reset_subtask(struct ice_pf *pf)
680{
681 enum ice_reset_req reset_type = ICE_RESET_INVAL;
682
683 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
684 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
685 * of reset is pending and sets bits in pf->state indicating the reset
686 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
687 * prepare for pending reset if not already (for PF software-initiated
688 * global resets the software should already be prepared for it as
689 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
690 * by firmware or software on other PFs, that bit is not set so prepare
691 * for the reset now), poll for reset done, rebuild and return.
692 */
693 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
694 /* Perform the largest reset requested */
695 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
696 reset_type = ICE_RESET_CORER;
697 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
698 reset_type = ICE_RESET_GLOBR;
699 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
700 reset_type = ICE_RESET_EMPR;
701 /* return if no valid reset type requested */
702 if (reset_type == ICE_RESET_INVAL)
703 return;
704 ice_prepare_for_reset(pf, reset_type);
705
706 /* make sure we are ready to rebuild */
707 if (ice_check_reset(&pf->hw)) {
708 set_bit(ICE_RESET_FAILED, pf->state);
709 } else {
710 /* done with reset. start rebuild */
711 pf->hw.reset_ongoing = false;
712 ice_rebuild(pf, reset_type);
713 /* clear bit to resume normal operations, but
714 * ICE_NEEDS_RESTART bit is set in case rebuild failed
715 */
716 clear_bit(ICE_RESET_OICR_RECV, pf->state);
717 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
718 clear_bit(ICE_PFR_REQ, pf->state);
719 clear_bit(ICE_CORER_REQ, pf->state);
720 clear_bit(ICE_GLOBR_REQ, pf->state);
721 wake_up(&pf->reset_wait_queue);
722 ice_reset_all_vfs(pf);
723 }
724
725 return;
726 }
727
728 /* No pending resets to finish processing. Check for new resets */
729 if (test_bit(ICE_PFR_REQ, pf->state)) {
730 reset_type = ICE_RESET_PFR;
731 if (pf->lag && pf->lag->bonded) {
732 dev_dbg(ice_pf_to_dev(pf), "PFR on a bonded interface, promoting to CORER\n");
733 reset_type = ICE_RESET_CORER;
734 }
735 }
736 if (test_bit(ICE_CORER_REQ, pf->state))
737 reset_type = ICE_RESET_CORER;
738 if (test_bit(ICE_GLOBR_REQ, pf->state))
739 reset_type = ICE_RESET_GLOBR;
740 /* If no valid reset type requested just return */
741 if (reset_type == ICE_RESET_INVAL)
742 return;
743
744 /* reset if not already down or busy */
745 if (!test_bit(ICE_DOWN, pf->state) &&
746 !test_bit(ICE_CFG_BUSY, pf->state)) {
747 ice_do_reset(pf, reset_type);
748 }
749}
750
751/**
752 * ice_print_topo_conflict - print topology conflict message
753 * @vsi: the VSI whose topology status is being checked
754 */
755static void ice_print_topo_conflict(struct ice_vsi *vsi)
756{
757 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
758 case ICE_AQ_LINK_TOPO_CONFLICT:
759 case ICE_AQ_LINK_MEDIA_CONFLICT:
760 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
761 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
762 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
763 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");
764 break;
765 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
766 if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
767 netdev_warn(vsi->netdev, "An unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules\n");
768 else
769 netdev_err(vsi->netdev, "Rx/Tx is disabled on this device because an unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
770 break;
771 default:
772 break;
773 }
774}
775
776/**
777 * ice_print_link_msg - print link up or down message
778 * @vsi: the VSI whose link status is being queried
779 * @isup: boolean for if the link is now up or down
780 */
781void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
782{
783 struct ice_aqc_get_phy_caps_data *caps;
784 const char *an_advertised;
785 const char *fec_req;
786 const char *speed;
787 const char *fec;
788 const char *fc;
789 const char *an;
790 int status;
791
792 if (!vsi)
793 return;
794
795 if (vsi->current_isup == isup)
796 return;
797
798 vsi->current_isup = isup;
799
800 if (!isup) {
801 netdev_info(vsi->netdev, "NIC Link is Down\n");
802 return;
803 }
804
805 switch (vsi->port_info->phy.link_info.link_speed) {
806 case ICE_AQ_LINK_SPEED_100GB:
807 speed = "100 G";
808 break;
809 case ICE_AQ_LINK_SPEED_50GB:
810 speed = "50 G";
811 break;
812 case ICE_AQ_LINK_SPEED_40GB:
813 speed = "40 G";
814 break;
815 case ICE_AQ_LINK_SPEED_25GB:
816 speed = "25 G";
817 break;
818 case ICE_AQ_LINK_SPEED_20GB:
819 speed = "20 G";
820 break;
821 case ICE_AQ_LINK_SPEED_10GB:
822 speed = "10 G";
823 break;
824 case ICE_AQ_LINK_SPEED_5GB:
825 speed = "5 G";
826 break;
827 case ICE_AQ_LINK_SPEED_2500MB:
828 speed = "2.5 G";
829 break;
830 case ICE_AQ_LINK_SPEED_1000MB:
831 speed = "1 G";
832 break;
833 case ICE_AQ_LINK_SPEED_100MB:
834 speed = "100 M";
835 break;
836 default:
837 speed = "Unknown ";
838 break;
839 }
840
841 switch (vsi->port_info->fc.current_mode) {
842 case ICE_FC_FULL:
843 fc = "Rx/Tx";
844 break;
845 case ICE_FC_TX_PAUSE:
846 fc = "Tx";
847 break;
848 case ICE_FC_RX_PAUSE:
849 fc = "Rx";
850 break;
851 case ICE_FC_NONE:
852 fc = "None";
853 break;
854 default:
855 fc = "Unknown";
856 break;
857 }
858
859 /* Get FEC mode based on negotiated link info */
860 switch (vsi->port_info->phy.link_info.fec_info) {
861 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
862 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
863 fec = "RS-FEC";
864 break;
865 case ICE_AQ_LINK_25G_KR_FEC_EN:
866 fec = "FC-FEC/BASE-R";
867 break;
868 default:
869 fec = "NONE";
870 break;
871 }
872
873 /* check if autoneg completed, might be false due to not supported */
874 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
875 an = "True";
876 else
877 an = "False";
878
879 /* Get FEC mode requested based on PHY caps last SW configuration */
880 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
881 if (!caps) {
882 fec_req = "Unknown";
883 an_advertised = "Unknown";
884 goto done;
885 }
886
887 status = ice_aq_get_phy_caps(vsi->port_info, false,
888 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
889 if (status)
890 netdev_info(vsi->netdev, "Get phy capability failed.\n");
891
892 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
893
894 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
895 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
896 fec_req = "RS-FEC";
897 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
898 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
899 fec_req = "FC-FEC/BASE-R";
900 else
901 fec_req = "NONE";
902
903 kfree(caps);
904
905done:
906 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",
907 speed, fec_req, fec, an_advertised, an, fc);
908 ice_print_topo_conflict(vsi);
909}
910
911/**
912 * ice_vsi_link_event - update the VSI's netdev
913 * @vsi: the VSI on which the link event occurred
914 * @link_up: whether or not the VSI needs to be set up or down
915 */
916static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
917{
918 if (!vsi)
919 return;
920
921 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
922 return;
923
924 if (vsi->type == ICE_VSI_PF) {
925 if (link_up == netif_carrier_ok(vsi->netdev))
926 return;
927
928 if (link_up) {
929 netif_carrier_on(vsi->netdev);
930 netif_tx_wake_all_queues(vsi->netdev);
931 } else {
932 netif_carrier_off(vsi->netdev);
933 netif_tx_stop_all_queues(vsi->netdev);
934 }
935 }
936}
937
938/**
939 * ice_set_dflt_mib - send a default config MIB to the FW
940 * @pf: private PF struct
941 *
942 * This function sends a default configuration MIB to the FW.
943 *
944 * If this function errors out at any point, the driver is still able to
945 * function. The main impact is that LFC may not operate as expected.
946 * Therefore an error state in this function should be treated with a DBG
947 * message and continue on with driver rebuild/reenable.
948 */
949static void ice_set_dflt_mib(struct ice_pf *pf)
950{
951 struct device *dev = ice_pf_to_dev(pf);
952 u8 mib_type, *buf, *lldpmib = NULL;
953 u16 len, typelen, offset = 0;
954 struct ice_lldp_org_tlv *tlv;
955 struct ice_hw *hw = &pf->hw;
956 u32 ouisubtype;
957
958 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
959 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
960 if (!lldpmib) {
961 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
962 __func__);
963 return;
964 }
965
966 /* Add ETS CFG TLV */
967 tlv = (struct ice_lldp_org_tlv *)lldpmib;
968 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
969 ICE_IEEE_ETS_TLV_LEN);
970 tlv->typelen = htons(typelen);
971 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
972 ICE_IEEE_SUBTYPE_ETS_CFG);
973 tlv->ouisubtype = htonl(ouisubtype);
974
975 buf = tlv->tlvinfo;
976 buf[0] = 0;
977
978 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
979 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
980 * Octets 13 - 20 are TSA values - leave as zeros
981 */
982 buf[5] = 0x64;
983 len = FIELD_GET(ICE_LLDP_TLV_LEN_M, typelen);
984 offset += len + 2;
985 tlv = (struct ice_lldp_org_tlv *)
986 ((char *)tlv + sizeof(tlv->typelen) + len);
987
988 /* Add ETS REC TLV */
989 buf = tlv->tlvinfo;
990 tlv->typelen = htons(typelen);
991
992 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
993 ICE_IEEE_SUBTYPE_ETS_REC);
994 tlv->ouisubtype = htonl(ouisubtype);
995
996 /* First octet of buf is reserved
997 * Octets 1 - 4 map UP to TC - all UPs map to zero
998 * Octets 5 - 12 are BW values - set TC 0 to 100%.
999 * Octets 13 - 20 are TSA value - leave as zeros
1000 */
1001 buf[5] = 0x64;
1002 offset += len + 2;
1003 tlv = (struct ice_lldp_org_tlv *)
1004 ((char *)tlv + sizeof(tlv->typelen) + len);
1005
1006 /* Add PFC CFG TLV */
1007 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
1008 ICE_IEEE_PFC_TLV_LEN);
1009 tlv->typelen = htons(typelen);
1010
1011 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
1012 ICE_IEEE_SUBTYPE_PFC_CFG);
1013 tlv->ouisubtype = htonl(ouisubtype);
1014
1015 /* Octet 1 left as all zeros - PFC disabled */
1016 buf[0] = 0x08;
1017 len = FIELD_GET(ICE_LLDP_TLV_LEN_M, typelen);
1018 offset += len + 2;
1019
1020 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
1021 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
1022
1023 kfree(lldpmib);
1024}
1025
1026/**
1027 * ice_check_phy_fw_load - check if PHY FW load failed
1028 * @pf: pointer to PF struct
1029 * @link_cfg_err: bitmap from the link info structure
1030 *
1031 * check if external PHY FW load failed and print an error message if it did
1032 */
1033static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
1034{
1035 if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
1036 clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1037 return;
1038 }
1039
1040 if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
1041 return;
1042
1043 if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
1044 dev_err(ice_pf_to_dev(pf), "Device failed to load the FW for the external PHY. Please download and install the latest NVM for your device and try again\n");
1045 set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1046 }
1047}
1048
1049/**
1050 * ice_check_module_power
1051 * @pf: pointer to PF struct
1052 * @link_cfg_err: bitmap from the link info structure
1053 *
1054 * check module power level returned by a previous call to aq_get_link_info
1055 * and print error messages if module power level is not supported
1056 */
1057static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
1058{
1059 /* if module power level is supported, clear the flag */
1060 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
1061 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
1062 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1063 return;
1064 }
1065
1066 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1067 * above block didn't clear this bit, there's nothing to do
1068 */
1069 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1070 return;
1071
1072 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1073 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1074 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1075 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1076 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1077 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1078 }
1079}
1080
1081/**
1082 * ice_check_link_cfg_err - check if link configuration failed
1083 * @pf: pointer to the PF struct
1084 * @link_cfg_err: bitmap from the link info structure
1085 *
1086 * print if any link configuration failure happens due to the value in the
1087 * link_cfg_err parameter in the link info structure
1088 */
1089static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1090{
1091 ice_check_module_power(pf, link_cfg_err);
1092 ice_check_phy_fw_load(pf, link_cfg_err);
1093}
1094
1095/**
1096 * ice_link_event - process the link event
1097 * @pf: PF that the link event is associated with
1098 * @pi: port_info for the port that the link event is associated with
1099 * @link_up: true if the physical link is up and false if it is down
1100 * @link_speed: current link speed received from the link event
1101 *
1102 * Returns 0 on success and negative on failure
1103 */
1104static int
1105ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1106 u16 link_speed)
1107{
1108 struct device *dev = ice_pf_to_dev(pf);
1109 struct ice_phy_info *phy_info;
1110 struct ice_vsi *vsi;
1111 u16 old_link_speed;
1112 bool old_link;
1113 int status;
1114
1115 phy_info = &pi->phy;
1116 phy_info->link_info_old = phy_info->link_info;
1117
1118 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1119 old_link_speed = phy_info->link_info_old.link_speed;
1120
1121 /* update the link info structures and re-enable link events,
1122 * don't bail on failure due to other book keeping needed
1123 */
1124 status = ice_update_link_info(pi);
1125 if (status)
1126 dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1127 pi->lport, status,
1128 ice_aq_str(pi->hw->adminq.sq_last_status));
1129
1130 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1131
1132 /* Check if the link state is up after updating link info, and treat
1133 * this event as an UP event since the link is actually UP now.
1134 */
1135 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1136 link_up = true;
1137
1138 vsi = ice_get_main_vsi(pf);
1139 if (!vsi || !vsi->port_info)
1140 return -EINVAL;
1141
1142 /* turn off PHY if media was removed */
1143 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1144 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1145 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1146 ice_set_link(vsi, false);
1147 }
1148
1149 /* if the old link up/down and speed is the same as the new */
1150 if (link_up == old_link && link_speed == old_link_speed)
1151 return 0;
1152
1153 ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1154
1155 if (ice_is_dcb_active(pf)) {
1156 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1157 ice_dcb_rebuild(pf);
1158 } else {
1159 if (link_up)
1160 ice_set_dflt_mib(pf);
1161 }
1162 ice_vsi_link_event(vsi, link_up);
1163 ice_print_link_msg(vsi, link_up);
1164
1165 ice_vc_notify_link_state(pf);
1166
1167 return 0;
1168}
1169
1170/**
1171 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1172 * @pf: board private structure
1173 */
1174static void ice_watchdog_subtask(struct ice_pf *pf)
1175{
1176 int i;
1177
1178 /* if interface is down do nothing */
1179 if (test_bit(ICE_DOWN, pf->state) ||
1180 test_bit(ICE_CFG_BUSY, pf->state))
1181 return;
1182
1183 /* make sure we don't do these things too often */
1184 if (time_before(jiffies,
1185 pf->serv_tmr_prev + pf->serv_tmr_period))
1186 return;
1187
1188 pf->serv_tmr_prev = jiffies;
1189
1190 /* Update the stats for active netdevs so the network stack
1191 * can look at updated numbers whenever it cares to
1192 */
1193 ice_update_pf_stats(pf);
1194 ice_for_each_vsi(pf, i)
1195 if (pf->vsi[i] && pf->vsi[i]->netdev)
1196 ice_update_vsi_stats(pf->vsi[i]);
1197}
1198
1199/**
1200 * ice_init_link_events - enable/initialize link events
1201 * @pi: pointer to the port_info instance
1202 *
1203 * Returns -EIO on failure, 0 on success
1204 */
1205static int ice_init_link_events(struct ice_port_info *pi)
1206{
1207 u16 mask;
1208
1209 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1210 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1211 ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1212
1213 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1214 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1215 pi->lport);
1216 return -EIO;
1217 }
1218
1219 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1220 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1221 pi->lport);
1222 return -EIO;
1223 }
1224
1225 return 0;
1226}
1227
1228/**
1229 * ice_handle_link_event - handle link event via ARQ
1230 * @pf: PF that the link event is associated with
1231 * @event: event structure containing link status info
1232 */
1233static int
1234ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1235{
1236 struct ice_aqc_get_link_status_data *link_data;
1237 struct ice_port_info *port_info;
1238 int status;
1239
1240 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1241 port_info = pf->hw.port_info;
1242 if (!port_info)
1243 return -EINVAL;
1244
1245 status = ice_link_event(pf, port_info,
1246 !!(link_data->link_info & ICE_AQ_LINK_UP),
1247 le16_to_cpu(link_data->link_speed));
1248 if (status)
1249 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1250 status);
1251
1252 return status;
1253}
1254
1255/**
1256 * ice_get_fwlog_data - copy the FW log data from ARQ event
1257 * @pf: PF that the FW log event is associated with
1258 * @event: event structure containing FW log data
1259 */
1260static void
1261ice_get_fwlog_data(struct ice_pf *pf, struct ice_rq_event_info *event)
1262{
1263 struct ice_fwlog_data *fwlog;
1264 struct ice_hw *hw = &pf->hw;
1265
1266 fwlog = &hw->fwlog_ring.rings[hw->fwlog_ring.tail];
1267
1268 memset(fwlog->data, 0, PAGE_SIZE);
1269 fwlog->data_size = le16_to_cpu(event->desc.datalen);
1270
1271 memcpy(fwlog->data, event->msg_buf, fwlog->data_size);
1272 ice_fwlog_ring_increment(&hw->fwlog_ring.tail, hw->fwlog_ring.size);
1273
1274 if (ice_fwlog_ring_full(&hw->fwlog_ring)) {
1275 /* the rings are full so bump the head to create room */
1276 ice_fwlog_ring_increment(&hw->fwlog_ring.head,
1277 hw->fwlog_ring.size);
1278 }
1279}
1280
1281/**
1282 * ice_aq_prep_for_event - Prepare to wait for an AdminQ event from firmware
1283 * @pf: pointer to the PF private structure
1284 * @task: intermediate helper storage and identifier for waiting
1285 * @opcode: the opcode to wait for
1286 *
1287 * Prepares to wait for a specific AdminQ completion event on the ARQ for
1288 * a given PF. Actual wait would be done by a call to ice_aq_wait_for_event().
1289 *
1290 * Calls are separated to allow caller registering for event before sending
1291 * the command, which mitigates a race between registering and FW responding.
1292 *
1293 * To obtain only the descriptor contents, pass an task->event with null
1294 * msg_buf. If the complete data buffer is desired, allocate the
1295 * task->event.msg_buf with enough space ahead of time.
1296 */
1297void ice_aq_prep_for_event(struct ice_pf *pf, struct ice_aq_task *task,
1298 u16 opcode)
1299{
1300 INIT_HLIST_NODE(&task->entry);
1301 task->opcode = opcode;
1302 task->state = ICE_AQ_TASK_WAITING;
1303
1304 spin_lock_bh(&pf->aq_wait_lock);
1305 hlist_add_head(&task->entry, &pf->aq_wait_list);
1306 spin_unlock_bh(&pf->aq_wait_lock);
1307}
1308
1309/**
1310 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1311 * @pf: pointer to the PF private structure
1312 * @task: ptr prepared by ice_aq_prep_for_event()
1313 * @timeout: how long to wait, in jiffies
1314 *
1315 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1316 * current thread will be put to sleep until the specified event occurs or
1317 * until the given timeout is reached.
1318 *
1319 * Returns: zero on success, or a negative error code on failure.
1320 */
1321int ice_aq_wait_for_event(struct ice_pf *pf, struct ice_aq_task *task,
1322 unsigned long timeout)
1323{
1324 enum ice_aq_task_state *state = &task->state;
1325 struct device *dev = ice_pf_to_dev(pf);
1326 unsigned long start = jiffies;
1327 long ret;
1328 int err;
1329
1330 ret = wait_event_interruptible_timeout(pf->aq_wait_queue,
1331 *state != ICE_AQ_TASK_WAITING,
1332 timeout);
1333 switch (*state) {
1334 case ICE_AQ_TASK_NOT_PREPARED:
1335 WARN(1, "call to %s without ice_aq_prep_for_event()", __func__);
1336 err = -EINVAL;
1337 break;
1338 case ICE_AQ_TASK_WAITING:
1339 err = ret < 0 ? ret : -ETIMEDOUT;
1340 break;
1341 case ICE_AQ_TASK_CANCELED:
1342 err = ret < 0 ? ret : -ECANCELED;
1343 break;
1344 case ICE_AQ_TASK_COMPLETE:
1345 err = ret < 0 ? ret : 0;
1346 break;
1347 default:
1348 WARN(1, "Unexpected AdminQ wait task state %u", *state);
1349 err = -EINVAL;
1350 break;
1351 }
1352
1353 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1354 jiffies_to_msecs(jiffies - start),
1355 jiffies_to_msecs(timeout),
1356 task->opcode);
1357
1358 spin_lock_bh(&pf->aq_wait_lock);
1359 hlist_del(&task->entry);
1360 spin_unlock_bh(&pf->aq_wait_lock);
1361
1362 return err;
1363}
1364
1365/**
1366 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1367 * @pf: pointer to the PF private structure
1368 * @opcode: the opcode of the event
1369 * @event: the event to check
1370 *
1371 * Loops over the current list of pending threads waiting for an AdminQ event.
1372 * For each matching task, copy the contents of the event into the task
1373 * structure and wake up the thread.
1374 *
1375 * If multiple threads wait for the same opcode, they will all be woken up.
1376 *
1377 * Note that event->msg_buf will only be duplicated if the event has a buffer
1378 * with enough space already allocated. Otherwise, only the descriptor and
1379 * message length will be copied.
1380 *
1381 * Returns: true if an event was found, false otherwise
1382 */
1383static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1384 struct ice_rq_event_info *event)
1385{
1386 struct ice_rq_event_info *task_ev;
1387 struct ice_aq_task *task;
1388 bool found = false;
1389
1390 spin_lock_bh(&pf->aq_wait_lock);
1391 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1392 if (task->state != ICE_AQ_TASK_WAITING)
1393 continue;
1394 if (task->opcode != opcode)
1395 continue;
1396
1397 task_ev = &task->event;
1398 memcpy(&task_ev->desc, &event->desc, sizeof(event->desc));
1399 task_ev->msg_len = event->msg_len;
1400
1401 /* Only copy the data buffer if a destination was set */
1402 if (task_ev->msg_buf && task_ev->buf_len >= event->buf_len) {
1403 memcpy(task_ev->msg_buf, event->msg_buf,
1404 event->buf_len);
1405 task_ev->buf_len = event->buf_len;
1406 }
1407
1408 task->state = ICE_AQ_TASK_COMPLETE;
1409 found = true;
1410 }
1411 spin_unlock_bh(&pf->aq_wait_lock);
1412
1413 if (found)
1414 wake_up(&pf->aq_wait_queue);
1415}
1416
1417/**
1418 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1419 * @pf: the PF private structure
1420 *
1421 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1422 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1423 */
1424static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1425{
1426 struct ice_aq_task *task;
1427
1428 spin_lock_bh(&pf->aq_wait_lock);
1429 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1430 task->state = ICE_AQ_TASK_CANCELED;
1431 spin_unlock_bh(&pf->aq_wait_lock);
1432
1433 wake_up(&pf->aq_wait_queue);
1434}
1435
1436#define ICE_MBX_OVERFLOW_WATERMARK 64
1437
1438/**
1439 * __ice_clean_ctrlq - helper function to clean controlq rings
1440 * @pf: ptr to struct ice_pf
1441 * @q_type: specific Control queue type
1442 */
1443static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1444{
1445 struct device *dev = ice_pf_to_dev(pf);
1446 struct ice_rq_event_info event;
1447 struct ice_hw *hw = &pf->hw;
1448 struct ice_ctl_q_info *cq;
1449 u16 pending, i = 0;
1450 const char *qtype;
1451 u32 oldval, val;
1452
1453 /* Do not clean control queue if/when PF reset fails */
1454 if (test_bit(ICE_RESET_FAILED, pf->state))
1455 return 0;
1456
1457 switch (q_type) {
1458 case ICE_CTL_Q_ADMIN:
1459 cq = &hw->adminq;
1460 qtype = "Admin";
1461 break;
1462 case ICE_CTL_Q_SB:
1463 cq = &hw->sbq;
1464 qtype = "Sideband";
1465 break;
1466 case ICE_CTL_Q_MAILBOX:
1467 cq = &hw->mailboxq;
1468 qtype = "Mailbox";
1469 /* we are going to try to detect a malicious VF, so set the
1470 * state to begin detection
1471 */
1472 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1473 break;
1474 default:
1475 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1476 return 0;
1477 }
1478
1479 /* check for error indications - PF_xx_AxQLEN register layout for
1480 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1481 */
1482 val = rd32(hw, cq->rq.len);
1483 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1484 PF_FW_ARQLEN_ARQCRIT_M)) {
1485 oldval = val;
1486 if (val & PF_FW_ARQLEN_ARQVFE_M)
1487 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1488 qtype);
1489 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1490 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1491 qtype);
1492 }
1493 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1494 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1495 qtype);
1496 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1497 PF_FW_ARQLEN_ARQCRIT_M);
1498 if (oldval != val)
1499 wr32(hw, cq->rq.len, val);
1500 }
1501
1502 val = rd32(hw, cq->sq.len);
1503 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1504 PF_FW_ATQLEN_ATQCRIT_M)) {
1505 oldval = val;
1506 if (val & PF_FW_ATQLEN_ATQVFE_M)
1507 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1508 qtype);
1509 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1510 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1511 qtype);
1512 }
1513 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1514 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1515 qtype);
1516 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1517 PF_FW_ATQLEN_ATQCRIT_M);
1518 if (oldval != val)
1519 wr32(hw, cq->sq.len, val);
1520 }
1521
1522 event.buf_len = cq->rq_buf_size;
1523 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1524 if (!event.msg_buf)
1525 return 0;
1526
1527 do {
1528 struct ice_mbx_data data = {};
1529 u16 opcode;
1530 int ret;
1531
1532 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1533 if (ret == -EALREADY)
1534 break;
1535 if (ret) {
1536 dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1537 ret);
1538 break;
1539 }
1540
1541 opcode = le16_to_cpu(event.desc.opcode);
1542
1543 /* Notify any thread that might be waiting for this event */
1544 ice_aq_check_events(pf, opcode, &event);
1545
1546 switch (opcode) {
1547 case ice_aqc_opc_get_link_status:
1548 if (ice_handle_link_event(pf, &event))
1549 dev_err(dev, "Could not handle link event\n");
1550 break;
1551 case ice_aqc_opc_event_lan_overflow:
1552 ice_vf_lan_overflow_event(pf, &event);
1553 break;
1554 case ice_mbx_opc_send_msg_to_pf:
1555 data.num_msg_proc = i;
1556 data.num_pending_arq = pending;
1557 data.max_num_msgs_mbx = hw->mailboxq.num_rq_entries;
1558 data.async_watermark_val = ICE_MBX_OVERFLOW_WATERMARK;
1559
1560 ice_vc_process_vf_msg(pf, &event, &data);
1561 break;
1562 case ice_aqc_opc_fw_logs_event:
1563 ice_get_fwlog_data(pf, &event);
1564 break;
1565 case ice_aqc_opc_lldp_set_mib_change:
1566 ice_dcb_process_lldp_set_mib_change(pf, &event);
1567 break;
1568 default:
1569 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1570 qtype, opcode);
1571 break;
1572 }
1573 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1574
1575 kfree(event.msg_buf);
1576
1577 return pending && (i == ICE_DFLT_IRQ_WORK);
1578}
1579
1580/**
1581 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1582 * @hw: pointer to hardware info
1583 * @cq: control queue information
1584 *
1585 * returns true if there are pending messages in a queue, false if there aren't
1586 */
1587static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1588{
1589 u16 ntu;
1590
1591 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1592 return cq->rq.next_to_clean != ntu;
1593}
1594
1595/**
1596 * ice_clean_adminq_subtask - clean the AdminQ rings
1597 * @pf: board private structure
1598 */
1599static void ice_clean_adminq_subtask(struct ice_pf *pf)
1600{
1601 struct ice_hw *hw = &pf->hw;
1602
1603 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1604 return;
1605
1606 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1607 return;
1608
1609 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1610
1611 /* There might be a situation where new messages arrive to a control
1612 * queue between processing the last message and clearing the
1613 * EVENT_PENDING bit. So before exiting, check queue head again (using
1614 * ice_ctrlq_pending) and process new messages if any.
1615 */
1616 if (ice_ctrlq_pending(hw, &hw->adminq))
1617 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1618
1619 ice_flush(hw);
1620}
1621
1622/**
1623 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1624 * @pf: board private structure
1625 */
1626static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1627{
1628 struct ice_hw *hw = &pf->hw;
1629
1630 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1631 return;
1632
1633 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1634 return;
1635
1636 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1637
1638 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1639 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1640
1641 ice_flush(hw);
1642}
1643
1644/**
1645 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1646 * @pf: board private structure
1647 */
1648static void ice_clean_sbq_subtask(struct ice_pf *pf)
1649{
1650 struct ice_hw *hw = &pf->hw;
1651
1652 /* Nothing to do here if sideband queue is not supported */
1653 if (!ice_is_sbq_supported(hw)) {
1654 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1655 return;
1656 }
1657
1658 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1659 return;
1660
1661 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1662 return;
1663
1664 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1665
1666 if (ice_ctrlq_pending(hw, &hw->sbq))
1667 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1668
1669 ice_flush(hw);
1670}
1671
1672/**
1673 * ice_service_task_schedule - schedule the service task to wake up
1674 * @pf: board private structure
1675 *
1676 * If not already scheduled, this puts the task into the work queue.
1677 */
1678void ice_service_task_schedule(struct ice_pf *pf)
1679{
1680 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1681 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1682 !test_bit(ICE_NEEDS_RESTART, pf->state))
1683 queue_work(ice_wq, &pf->serv_task);
1684}
1685
1686/**
1687 * ice_service_task_complete - finish up the service task
1688 * @pf: board private structure
1689 */
1690static void ice_service_task_complete(struct ice_pf *pf)
1691{
1692 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1693
1694 /* force memory (pf->state) to sync before next service task */
1695 smp_mb__before_atomic();
1696 clear_bit(ICE_SERVICE_SCHED, pf->state);
1697}
1698
1699/**
1700 * ice_service_task_stop - stop service task and cancel works
1701 * @pf: board private structure
1702 *
1703 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1704 * 1 otherwise.
1705 */
1706static int ice_service_task_stop(struct ice_pf *pf)
1707{
1708 int ret;
1709
1710 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1711
1712 if (pf->serv_tmr.function)
1713 del_timer_sync(&pf->serv_tmr);
1714 if (pf->serv_task.func)
1715 cancel_work_sync(&pf->serv_task);
1716
1717 clear_bit(ICE_SERVICE_SCHED, pf->state);
1718 return ret;
1719}
1720
1721/**
1722 * ice_service_task_restart - restart service task and schedule works
1723 * @pf: board private structure
1724 *
1725 * This function is needed for suspend and resume works (e.g WoL scenario)
1726 */
1727static void ice_service_task_restart(struct ice_pf *pf)
1728{
1729 clear_bit(ICE_SERVICE_DIS, pf->state);
1730 ice_service_task_schedule(pf);
1731}
1732
1733/**
1734 * ice_service_timer - timer callback to schedule service task
1735 * @t: pointer to timer_list
1736 */
1737static void ice_service_timer(struct timer_list *t)
1738{
1739 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1740
1741 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1742 ice_service_task_schedule(pf);
1743}
1744
1745/**
1746 * ice_handle_mdd_event - handle malicious driver detect event
1747 * @pf: pointer to the PF structure
1748 *
1749 * Called from service task. OICR interrupt handler indicates MDD event.
1750 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1751 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1752 * disable the queue, the PF can be configured to reset the VF using ethtool
1753 * private flag mdd-auto-reset-vf.
1754 */
1755static void ice_handle_mdd_event(struct ice_pf *pf)
1756{
1757 struct device *dev = ice_pf_to_dev(pf);
1758 struct ice_hw *hw = &pf->hw;
1759 struct ice_vf *vf;
1760 unsigned int bkt;
1761 u32 reg;
1762
1763 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1764 /* Since the VF MDD event logging is rate limited, check if
1765 * there are pending MDD events.
1766 */
1767 ice_print_vfs_mdd_events(pf);
1768 return;
1769 }
1770
1771 /* find what triggered an MDD event */
1772 reg = rd32(hw, GL_MDET_TX_PQM);
1773 if (reg & GL_MDET_TX_PQM_VALID_M) {
1774 u8 pf_num = FIELD_GET(GL_MDET_TX_PQM_PF_NUM_M, reg);
1775 u16 vf_num = FIELD_GET(GL_MDET_TX_PQM_VF_NUM_M, reg);
1776 u8 event = FIELD_GET(GL_MDET_TX_PQM_MAL_TYPE_M, reg);
1777 u16 queue = FIELD_GET(GL_MDET_TX_PQM_QNUM_M, reg);
1778
1779 if (netif_msg_tx_err(pf))
1780 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1781 event, queue, pf_num, vf_num);
1782 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1783 }
1784
1785 reg = rd32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw));
1786 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1787 u8 pf_num = FIELD_GET(GL_MDET_TX_TCLAN_PF_NUM_M, reg);
1788 u16 vf_num = FIELD_GET(GL_MDET_TX_TCLAN_VF_NUM_M, reg);
1789 u8 event = FIELD_GET(GL_MDET_TX_TCLAN_MAL_TYPE_M, reg);
1790 u16 queue = FIELD_GET(GL_MDET_TX_TCLAN_QNUM_M, reg);
1791
1792 if (netif_msg_tx_err(pf))
1793 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1794 event, queue, pf_num, vf_num);
1795 wr32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw), U32_MAX);
1796 }
1797
1798 reg = rd32(hw, GL_MDET_RX);
1799 if (reg & GL_MDET_RX_VALID_M) {
1800 u8 pf_num = FIELD_GET(GL_MDET_RX_PF_NUM_M, reg);
1801 u16 vf_num = FIELD_GET(GL_MDET_RX_VF_NUM_M, reg);
1802 u8 event = FIELD_GET(GL_MDET_RX_MAL_TYPE_M, reg);
1803 u16 queue = FIELD_GET(GL_MDET_RX_QNUM_M, reg);
1804
1805 if (netif_msg_rx_err(pf))
1806 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1807 event, queue, pf_num, vf_num);
1808 wr32(hw, GL_MDET_RX, 0xffffffff);
1809 }
1810
1811 /* check to see if this PF caused an MDD event */
1812 reg = rd32(hw, PF_MDET_TX_PQM);
1813 if (reg & PF_MDET_TX_PQM_VALID_M) {
1814 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1815 if (netif_msg_tx_err(pf))
1816 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1817 }
1818
1819 reg = rd32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw));
1820 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1821 wr32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw), 0xffff);
1822 if (netif_msg_tx_err(pf))
1823 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1824 }
1825
1826 reg = rd32(hw, PF_MDET_RX);
1827 if (reg & PF_MDET_RX_VALID_M) {
1828 wr32(hw, PF_MDET_RX, 0xFFFF);
1829 if (netif_msg_rx_err(pf))
1830 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1831 }
1832
1833 /* Check to see if one of the VFs caused an MDD event, and then
1834 * increment counters and set print pending
1835 */
1836 mutex_lock(&pf->vfs.table_lock);
1837 ice_for_each_vf(pf, bkt, vf) {
1838 reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1839 if (reg & VP_MDET_TX_PQM_VALID_M) {
1840 wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1841 vf->mdd_tx_events.count++;
1842 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1843 if (netif_msg_tx_err(pf))
1844 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1845 vf->vf_id);
1846 }
1847
1848 reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1849 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1850 wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1851 vf->mdd_tx_events.count++;
1852 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1853 if (netif_msg_tx_err(pf))
1854 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1855 vf->vf_id);
1856 }
1857
1858 reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1859 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1860 wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1861 vf->mdd_tx_events.count++;
1862 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1863 if (netif_msg_tx_err(pf))
1864 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1865 vf->vf_id);
1866 }
1867
1868 reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1869 if (reg & VP_MDET_RX_VALID_M) {
1870 wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1871 vf->mdd_rx_events.count++;
1872 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1873 if (netif_msg_rx_err(pf))
1874 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1875 vf->vf_id);
1876
1877 /* Since the queue is disabled on VF Rx MDD events, the
1878 * PF can be configured to reset the VF through ethtool
1879 * private flag mdd-auto-reset-vf.
1880 */
1881 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1882 /* VF MDD event counters will be cleared by
1883 * reset, so print the event prior to reset.
1884 */
1885 ice_print_vf_rx_mdd_event(vf);
1886 ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1887 }
1888 }
1889 }
1890 mutex_unlock(&pf->vfs.table_lock);
1891
1892 ice_print_vfs_mdd_events(pf);
1893}
1894
1895/**
1896 * ice_force_phys_link_state - Force the physical link state
1897 * @vsi: VSI to force the physical link state to up/down
1898 * @link_up: true/false indicates to set the physical link to up/down
1899 *
1900 * Force the physical link state by getting the current PHY capabilities from
1901 * hardware and setting the PHY config based on the determined capabilities. If
1902 * link changes a link event will be triggered because both the Enable Automatic
1903 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1904 *
1905 * Returns 0 on success, negative on failure
1906 */
1907static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1908{
1909 struct ice_aqc_get_phy_caps_data *pcaps;
1910 struct ice_aqc_set_phy_cfg_data *cfg;
1911 struct ice_port_info *pi;
1912 struct device *dev;
1913 int retcode;
1914
1915 if (!vsi || !vsi->port_info || !vsi->back)
1916 return -EINVAL;
1917 if (vsi->type != ICE_VSI_PF)
1918 return 0;
1919
1920 dev = ice_pf_to_dev(vsi->back);
1921
1922 pi = vsi->port_info;
1923
1924 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1925 if (!pcaps)
1926 return -ENOMEM;
1927
1928 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1929 NULL);
1930 if (retcode) {
1931 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1932 vsi->vsi_num, retcode);
1933 retcode = -EIO;
1934 goto out;
1935 }
1936
1937 /* No change in link */
1938 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1939 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1940 goto out;
1941
1942 /* Use the current user PHY configuration. The current user PHY
1943 * configuration is initialized during probe from PHY capabilities
1944 * software mode, and updated on set PHY configuration.
1945 */
1946 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1947 if (!cfg) {
1948 retcode = -ENOMEM;
1949 goto out;
1950 }
1951
1952 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1953 if (link_up)
1954 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1955 else
1956 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1957
1958 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1959 if (retcode) {
1960 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1961 vsi->vsi_num, retcode);
1962 retcode = -EIO;
1963 }
1964
1965 kfree(cfg);
1966out:
1967 kfree(pcaps);
1968 return retcode;
1969}
1970
1971/**
1972 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1973 * @pi: port info structure
1974 *
1975 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1976 */
1977static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1978{
1979 struct ice_aqc_get_phy_caps_data *pcaps;
1980 struct ice_pf *pf = pi->hw->back;
1981 int err;
1982
1983 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1984 if (!pcaps)
1985 return -ENOMEM;
1986
1987 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1988 pcaps, NULL);
1989
1990 if (err) {
1991 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1992 goto out;
1993 }
1994
1995 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1996 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1997
1998out:
1999 kfree(pcaps);
2000 return err;
2001}
2002
2003/**
2004 * ice_init_link_dflt_override - Initialize link default override
2005 * @pi: port info structure
2006 *
2007 * Initialize link default override and PHY total port shutdown during probe
2008 */
2009static void ice_init_link_dflt_override(struct ice_port_info *pi)
2010{
2011 struct ice_link_default_override_tlv *ldo;
2012 struct ice_pf *pf = pi->hw->back;
2013
2014 ldo = &pf->link_dflt_override;
2015 if (ice_get_link_default_override(ldo, pi))
2016 return;
2017
2018 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
2019 return;
2020
2021 /* Enable Total Port Shutdown (override/replace link-down-on-close
2022 * ethtool private flag) for ports with Port Disable bit set.
2023 */
2024 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
2025 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
2026}
2027
2028/**
2029 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
2030 * @pi: port info structure
2031 *
2032 * If default override is enabled, initialize the user PHY cfg speed and FEC
2033 * settings using the default override mask from the NVM.
2034 *
2035 * The PHY should only be configured with the default override settings the
2036 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
2037 * is used to indicate that the user PHY cfg default override is initialized
2038 * and the PHY has not been configured with the default override settings. The
2039 * state is set here, and cleared in ice_configure_phy the first time the PHY is
2040 * configured.
2041 *
2042 * This function should be called only if the FW doesn't support default
2043 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
2044 */
2045static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
2046{
2047 struct ice_link_default_override_tlv *ldo;
2048 struct ice_aqc_set_phy_cfg_data *cfg;
2049 struct ice_phy_info *phy = &pi->phy;
2050 struct ice_pf *pf = pi->hw->back;
2051
2052 ldo = &pf->link_dflt_override;
2053
2054 /* If link default override is enabled, use to mask NVM PHY capabilities
2055 * for speed and FEC default configuration.
2056 */
2057 cfg = &phy->curr_user_phy_cfg;
2058
2059 if (ldo->phy_type_low || ldo->phy_type_high) {
2060 cfg->phy_type_low = pf->nvm_phy_type_lo &
2061 cpu_to_le64(ldo->phy_type_low);
2062 cfg->phy_type_high = pf->nvm_phy_type_hi &
2063 cpu_to_le64(ldo->phy_type_high);
2064 }
2065 cfg->link_fec_opt = ldo->fec_options;
2066 phy->curr_user_fec_req = ICE_FEC_AUTO;
2067
2068 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2069}
2070
2071/**
2072 * ice_init_phy_user_cfg - Initialize the PHY user configuration
2073 * @pi: port info structure
2074 *
2075 * Initialize the current user PHY configuration, speed, FEC, and FC requested
2076 * mode to default. The PHY defaults are from get PHY capabilities topology
2077 * with media so call when media is first available. An error is returned if
2078 * called when media is not available. The PHY initialization completed state is
2079 * set here.
2080 *
2081 * These configurations are used when setting PHY
2082 * configuration. The user PHY configuration is updated on set PHY
2083 * configuration. Returns 0 on success, negative on failure
2084 */
2085static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2086{
2087 struct ice_aqc_get_phy_caps_data *pcaps;
2088 struct ice_phy_info *phy = &pi->phy;
2089 struct ice_pf *pf = pi->hw->back;
2090 int err;
2091
2092 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2093 return -EIO;
2094
2095 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2096 if (!pcaps)
2097 return -ENOMEM;
2098
2099 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2100 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2101 pcaps, NULL);
2102 else
2103 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2104 pcaps, NULL);
2105 if (err) {
2106 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2107 goto err_out;
2108 }
2109
2110 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2111
2112 /* check if lenient mode is supported and enabled */
2113 if (ice_fw_supports_link_override(pi->hw) &&
2114 !(pcaps->module_compliance_enforcement &
2115 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2116 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2117
2118 /* if the FW supports default PHY configuration mode, then the driver
2119 * does not have to apply link override settings. If not,
2120 * initialize user PHY configuration with link override values
2121 */
2122 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2123 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2124 ice_init_phy_cfg_dflt_override(pi);
2125 goto out;
2126 }
2127 }
2128
2129 /* if link default override is not enabled, set user flow control and
2130 * FEC settings based on what get_phy_caps returned
2131 */
2132 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2133 pcaps->link_fec_options);
2134 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2135
2136out:
2137 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2138 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2139err_out:
2140 kfree(pcaps);
2141 return err;
2142}
2143
2144/**
2145 * ice_configure_phy - configure PHY
2146 * @vsi: VSI of PHY
2147 *
2148 * Set the PHY configuration. If the current PHY configuration is the same as
2149 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2150 * configure the based get PHY capabilities for topology with media.
2151 */
2152static int ice_configure_phy(struct ice_vsi *vsi)
2153{
2154 struct device *dev = ice_pf_to_dev(vsi->back);
2155 struct ice_port_info *pi = vsi->port_info;
2156 struct ice_aqc_get_phy_caps_data *pcaps;
2157 struct ice_aqc_set_phy_cfg_data *cfg;
2158 struct ice_phy_info *phy = &pi->phy;
2159 struct ice_pf *pf = vsi->back;
2160 int err;
2161
2162 /* Ensure we have media as we cannot configure a medialess port */
2163 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2164 return -ENOMEDIUM;
2165
2166 ice_print_topo_conflict(vsi);
2167
2168 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2169 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2170 return -EPERM;
2171
2172 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2173 return ice_force_phys_link_state(vsi, true);
2174
2175 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2176 if (!pcaps)
2177 return -ENOMEM;
2178
2179 /* Get current PHY config */
2180 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2181 NULL);
2182 if (err) {
2183 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2184 vsi->vsi_num, err);
2185 goto done;
2186 }
2187
2188 /* If PHY enable link is configured and configuration has not changed,
2189 * there's nothing to do
2190 */
2191 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2192 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2193 goto done;
2194
2195 /* Use PHY topology as baseline for configuration */
2196 memset(pcaps, 0, sizeof(*pcaps));
2197 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2198 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2199 pcaps, NULL);
2200 else
2201 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2202 pcaps, NULL);
2203 if (err) {
2204 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2205 vsi->vsi_num, err);
2206 goto done;
2207 }
2208
2209 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2210 if (!cfg) {
2211 err = -ENOMEM;
2212 goto done;
2213 }
2214
2215 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2216
2217 /* Speed - If default override pending, use curr_user_phy_cfg set in
2218 * ice_init_phy_user_cfg_ldo.
2219 */
2220 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2221 vsi->back->state)) {
2222 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2223 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2224 } else {
2225 u64 phy_low = 0, phy_high = 0;
2226
2227 ice_update_phy_type(&phy_low, &phy_high,
2228 pi->phy.curr_user_speed_req);
2229 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2230 cfg->phy_type_high = pcaps->phy_type_high &
2231 cpu_to_le64(phy_high);
2232 }
2233
2234 /* Can't provide what was requested; use PHY capabilities */
2235 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2236 cfg->phy_type_low = pcaps->phy_type_low;
2237 cfg->phy_type_high = pcaps->phy_type_high;
2238 }
2239
2240 /* FEC */
2241 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2242
2243 /* Can't provide what was requested; use PHY capabilities */
2244 if (cfg->link_fec_opt !=
2245 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2246 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2247 cfg->link_fec_opt = pcaps->link_fec_options;
2248 }
2249
2250 /* Flow Control - always supported; no need to check against
2251 * capabilities
2252 */
2253 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2254
2255 /* Enable link and link update */
2256 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2257
2258 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2259 if (err)
2260 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2261 vsi->vsi_num, err);
2262
2263 kfree(cfg);
2264done:
2265 kfree(pcaps);
2266 return err;
2267}
2268
2269/**
2270 * ice_check_media_subtask - Check for media
2271 * @pf: pointer to PF struct
2272 *
2273 * If media is available, then initialize PHY user configuration if it is not
2274 * been, and configure the PHY if the interface is up.
2275 */
2276static void ice_check_media_subtask(struct ice_pf *pf)
2277{
2278 struct ice_port_info *pi;
2279 struct ice_vsi *vsi;
2280 int err;
2281
2282 /* No need to check for media if it's already present */
2283 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2284 return;
2285
2286 vsi = ice_get_main_vsi(pf);
2287 if (!vsi)
2288 return;
2289
2290 /* Refresh link info and check if media is present */
2291 pi = vsi->port_info;
2292 err = ice_update_link_info(pi);
2293 if (err)
2294 return;
2295
2296 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2297
2298 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2299 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2300 ice_init_phy_user_cfg(pi);
2301
2302 /* PHY settings are reset on media insertion, reconfigure
2303 * PHY to preserve settings.
2304 */
2305 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2306 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2307 return;
2308
2309 err = ice_configure_phy(vsi);
2310 if (!err)
2311 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2312
2313 /* A Link Status Event will be generated; the event handler
2314 * will complete bringing the interface up
2315 */
2316 }
2317}
2318
2319/**
2320 * ice_service_task - manage and run subtasks
2321 * @work: pointer to work_struct contained by the PF struct
2322 */
2323static void ice_service_task(struct work_struct *work)
2324{
2325 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2326 unsigned long start_time = jiffies;
2327
2328 /* subtasks */
2329
2330 /* process reset requests first */
2331 ice_reset_subtask(pf);
2332
2333 /* bail if a reset/recovery cycle is pending or rebuild failed */
2334 if (ice_is_reset_in_progress(pf->state) ||
2335 test_bit(ICE_SUSPENDED, pf->state) ||
2336 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2337 ice_service_task_complete(pf);
2338 return;
2339 }
2340
2341 if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2342 struct iidc_event *event;
2343
2344 event = kzalloc(sizeof(*event), GFP_KERNEL);
2345 if (event) {
2346 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2347 /* report the entire OICR value to AUX driver */
2348 swap(event->reg, pf->oicr_err_reg);
2349 ice_send_event_to_aux(pf, event);
2350 kfree(event);
2351 }
2352 }
2353
2354 /* unplug aux dev per request, if an unplug request came in
2355 * while processing a plug request, this will handle it
2356 */
2357 if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags))
2358 ice_unplug_aux_dev(pf);
2359
2360 /* Plug aux device per request */
2361 if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2362 ice_plug_aux_dev(pf);
2363
2364 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2365 struct iidc_event *event;
2366
2367 event = kzalloc(sizeof(*event), GFP_KERNEL);
2368 if (event) {
2369 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2370 ice_send_event_to_aux(pf, event);
2371 kfree(event);
2372 }
2373 }
2374
2375 ice_clean_adminq_subtask(pf);
2376 ice_check_media_subtask(pf);
2377 ice_check_for_hang_subtask(pf);
2378 ice_sync_fltr_subtask(pf);
2379 ice_handle_mdd_event(pf);
2380 ice_watchdog_subtask(pf);
2381
2382 if (ice_is_safe_mode(pf)) {
2383 ice_service_task_complete(pf);
2384 return;
2385 }
2386
2387 ice_process_vflr_event(pf);
2388 ice_clean_mailboxq_subtask(pf);
2389 ice_clean_sbq_subtask(pf);
2390 ice_sync_arfs_fltrs(pf);
2391 ice_flush_fdir_ctx(pf);
2392
2393 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2394 ice_service_task_complete(pf);
2395
2396 /* If the tasks have taken longer than one service timer period
2397 * or there is more work to be done, reset the service timer to
2398 * schedule the service task now.
2399 */
2400 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2401 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2402 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2403 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2404 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2405 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2406 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2407 mod_timer(&pf->serv_tmr, jiffies);
2408}
2409
2410/**
2411 * ice_set_ctrlq_len - helper function to set controlq length
2412 * @hw: pointer to the HW instance
2413 */
2414static void ice_set_ctrlq_len(struct ice_hw *hw)
2415{
2416 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2417 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2418 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2419 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2420 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2421 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2422 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2423 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2424 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2425 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2426 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2427 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2428}
2429
2430/**
2431 * ice_schedule_reset - schedule a reset
2432 * @pf: board private structure
2433 * @reset: reset being requested
2434 */
2435int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2436{
2437 struct device *dev = ice_pf_to_dev(pf);
2438
2439 /* bail out if earlier reset has failed */
2440 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2441 dev_dbg(dev, "earlier reset has failed\n");
2442 return -EIO;
2443 }
2444 /* bail if reset/recovery already in progress */
2445 if (ice_is_reset_in_progress(pf->state)) {
2446 dev_dbg(dev, "Reset already in progress\n");
2447 return -EBUSY;
2448 }
2449
2450 switch (reset) {
2451 case ICE_RESET_PFR:
2452 set_bit(ICE_PFR_REQ, pf->state);
2453 break;
2454 case ICE_RESET_CORER:
2455 set_bit(ICE_CORER_REQ, pf->state);
2456 break;
2457 case ICE_RESET_GLOBR:
2458 set_bit(ICE_GLOBR_REQ, pf->state);
2459 break;
2460 default:
2461 return -EINVAL;
2462 }
2463
2464 ice_service_task_schedule(pf);
2465 return 0;
2466}
2467
2468/**
2469 * ice_irq_affinity_notify - Callback for affinity changes
2470 * @notify: context as to what irq was changed
2471 * @mask: the new affinity mask
2472 *
2473 * This is a callback function used by the irq_set_affinity_notifier function
2474 * so that we may register to receive changes to the irq affinity masks.
2475 */
2476static void
2477ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2478 const cpumask_t *mask)
2479{
2480 struct ice_q_vector *q_vector =
2481 container_of(notify, struct ice_q_vector, affinity_notify);
2482
2483 cpumask_copy(&q_vector->affinity_mask, mask);
2484}
2485
2486/**
2487 * ice_irq_affinity_release - Callback for affinity notifier release
2488 * @ref: internal core kernel usage
2489 *
2490 * This is a callback function used by the irq_set_affinity_notifier function
2491 * to inform the current notification subscriber that they will no longer
2492 * receive notifications.
2493 */
2494static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2495
2496/**
2497 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2498 * @vsi: the VSI being configured
2499 */
2500static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2501{
2502 struct ice_hw *hw = &vsi->back->hw;
2503 int i;
2504
2505 ice_for_each_q_vector(vsi, i)
2506 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2507
2508 ice_flush(hw);
2509 return 0;
2510}
2511
2512/**
2513 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2514 * @vsi: the VSI being configured
2515 * @basename: name for the vector
2516 */
2517static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2518{
2519 int q_vectors = vsi->num_q_vectors;
2520 struct ice_pf *pf = vsi->back;
2521 struct device *dev;
2522 int rx_int_idx = 0;
2523 int tx_int_idx = 0;
2524 int vector, err;
2525 int irq_num;
2526
2527 dev = ice_pf_to_dev(pf);
2528 for (vector = 0; vector < q_vectors; vector++) {
2529 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2530
2531 irq_num = q_vector->irq.virq;
2532
2533 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2534 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2535 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2536 tx_int_idx++;
2537 } else if (q_vector->rx.rx_ring) {
2538 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2539 "%s-%s-%d", basename, "rx", rx_int_idx++);
2540 } else if (q_vector->tx.tx_ring) {
2541 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2542 "%s-%s-%d", basename, "tx", tx_int_idx++);
2543 } else {
2544 /* skip this unused q_vector */
2545 continue;
2546 }
2547 if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2548 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2549 IRQF_SHARED, q_vector->name,
2550 q_vector);
2551 else
2552 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2553 0, q_vector->name, q_vector);
2554 if (err) {
2555 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2556 err);
2557 goto free_q_irqs;
2558 }
2559
2560 /* register for affinity change notifications */
2561 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2562 struct irq_affinity_notify *affinity_notify;
2563
2564 affinity_notify = &q_vector->affinity_notify;
2565 affinity_notify->notify = ice_irq_affinity_notify;
2566 affinity_notify->release = ice_irq_affinity_release;
2567 irq_set_affinity_notifier(irq_num, affinity_notify);
2568 }
2569
2570 /* assign the mask for this irq */
2571 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2572 }
2573
2574 err = ice_set_cpu_rx_rmap(vsi);
2575 if (err) {
2576 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2577 vsi->vsi_num, ERR_PTR(err));
2578 goto free_q_irqs;
2579 }
2580
2581 vsi->irqs_ready = true;
2582 return 0;
2583
2584free_q_irqs:
2585 while (vector--) {
2586 irq_num = vsi->q_vectors[vector]->irq.virq;
2587 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2588 irq_set_affinity_notifier(irq_num, NULL);
2589 irq_set_affinity_hint(irq_num, NULL);
2590 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2591 }
2592 return err;
2593}
2594
2595/**
2596 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2597 * @vsi: VSI to setup Tx rings used by XDP
2598 *
2599 * Return 0 on success and negative value on error
2600 */
2601static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2602{
2603 struct device *dev = ice_pf_to_dev(vsi->back);
2604 struct ice_tx_desc *tx_desc;
2605 int i, j;
2606
2607 ice_for_each_xdp_txq(vsi, i) {
2608 u16 xdp_q_idx = vsi->alloc_txq + i;
2609 struct ice_ring_stats *ring_stats;
2610 struct ice_tx_ring *xdp_ring;
2611
2612 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2613 if (!xdp_ring)
2614 goto free_xdp_rings;
2615
2616 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
2617 if (!ring_stats) {
2618 ice_free_tx_ring(xdp_ring);
2619 goto free_xdp_rings;
2620 }
2621
2622 xdp_ring->ring_stats = ring_stats;
2623 xdp_ring->q_index = xdp_q_idx;
2624 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2625 xdp_ring->vsi = vsi;
2626 xdp_ring->netdev = NULL;
2627 xdp_ring->dev = dev;
2628 xdp_ring->count = vsi->num_tx_desc;
2629 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2630 if (ice_setup_tx_ring(xdp_ring))
2631 goto free_xdp_rings;
2632 ice_set_ring_xdp(xdp_ring);
2633 spin_lock_init(&xdp_ring->tx_lock);
2634 for (j = 0; j < xdp_ring->count; j++) {
2635 tx_desc = ICE_TX_DESC(xdp_ring, j);
2636 tx_desc->cmd_type_offset_bsz = 0;
2637 }
2638 }
2639
2640 return 0;
2641
2642free_xdp_rings:
2643 for (; i >= 0; i--) {
2644 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc) {
2645 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2646 vsi->xdp_rings[i]->ring_stats = NULL;
2647 ice_free_tx_ring(vsi->xdp_rings[i]);
2648 }
2649 }
2650 return -ENOMEM;
2651}
2652
2653/**
2654 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2655 * @vsi: VSI to set the bpf prog on
2656 * @prog: the bpf prog pointer
2657 */
2658static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2659{
2660 struct bpf_prog *old_prog;
2661 int i;
2662
2663 old_prog = xchg(&vsi->xdp_prog, prog);
2664 ice_for_each_rxq(vsi, i)
2665 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2666
2667 if (old_prog)
2668 bpf_prog_put(old_prog);
2669}
2670
2671/**
2672 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2673 * @vsi: VSI to bring up Tx rings used by XDP
2674 * @prog: bpf program that will be assigned to VSI
2675 *
2676 * Return 0 on success and negative value on error
2677 */
2678int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2679{
2680 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2681 int xdp_rings_rem = vsi->num_xdp_txq;
2682 struct ice_pf *pf = vsi->back;
2683 struct ice_qs_cfg xdp_qs_cfg = {
2684 .qs_mutex = &pf->avail_q_mutex,
2685 .pf_map = pf->avail_txqs,
2686 .pf_map_size = pf->max_pf_txqs,
2687 .q_count = vsi->num_xdp_txq,
2688 .scatter_count = ICE_MAX_SCATTER_TXQS,
2689 .vsi_map = vsi->txq_map,
2690 .vsi_map_offset = vsi->alloc_txq,
2691 .mapping_mode = ICE_VSI_MAP_CONTIG
2692 };
2693 struct device *dev;
2694 int i, v_idx;
2695 int status;
2696
2697 dev = ice_pf_to_dev(pf);
2698 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2699 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2700 if (!vsi->xdp_rings)
2701 return -ENOMEM;
2702
2703 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2704 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2705 goto err_map_xdp;
2706
2707 if (static_key_enabled(&ice_xdp_locking_key))
2708 netdev_warn(vsi->netdev,
2709 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2710
2711 if (ice_xdp_alloc_setup_rings(vsi))
2712 goto clear_xdp_rings;
2713
2714 /* follow the logic from ice_vsi_map_rings_to_vectors */
2715 ice_for_each_q_vector(vsi, v_idx) {
2716 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2717 int xdp_rings_per_v, q_id, q_base;
2718
2719 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2720 vsi->num_q_vectors - v_idx);
2721 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2722
2723 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2724 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2725
2726 xdp_ring->q_vector = q_vector;
2727 xdp_ring->next = q_vector->tx.tx_ring;
2728 q_vector->tx.tx_ring = xdp_ring;
2729 }
2730 xdp_rings_rem -= xdp_rings_per_v;
2731 }
2732
2733 ice_for_each_rxq(vsi, i) {
2734 if (static_key_enabled(&ice_xdp_locking_key)) {
2735 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2736 } else {
2737 struct ice_q_vector *q_vector = vsi->rx_rings[i]->q_vector;
2738 struct ice_tx_ring *ring;
2739
2740 ice_for_each_tx_ring(ring, q_vector->tx) {
2741 if (ice_ring_is_xdp(ring)) {
2742 vsi->rx_rings[i]->xdp_ring = ring;
2743 break;
2744 }
2745 }
2746 }
2747 ice_tx_xsk_pool(vsi, i);
2748 }
2749
2750 /* omit the scheduler update if in reset path; XDP queues will be
2751 * taken into account at the end of ice_vsi_rebuild, where
2752 * ice_cfg_vsi_lan is being called
2753 */
2754 if (ice_is_reset_in_progress(pf->state))
2755 return 0;
2756
2757 /* tell the Tx scheduler that right now we have
2758 * additional queues
2759 */
2760 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2761 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2762
2763 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2764 max_txqs);
2765 if (status) {
2766 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2767 status);
2768 goto clear_xdp_rings;
2769 }
2770
2771 /* assign the prog only when it's not already present on VSI;
2772 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2773 * VSI rebuild that happens under ethtool -L can expose us to
2774 * the bpf_prog refcount issues as we would be swapping same
2775 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2776 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2777 * this is not harmful as dev_xdp_install bumps the refcount
2778 * before calling the op exposed by the driver;
2779 */
2780 if (!ice_is_xdp_ena_vsi(vsi))
2781 ice_vsi_assign_bpf_prog(vsi, prog);
2782
2783 return 0;
2784clear_xdp_rings:
2785 ice_for_each_xdp_txq(vsi, i)
2786 if (vsi->xdp_rings[i]) {
2787 kfree_rcu(vsi->xdp_rings[i], rcu);
2788 vsi->xdp_rings[i] = NULL;
2789 }
2790
2791err_map_xdp:
2792 mutex_lock(&pf->avail_q_mutex);
2793 ice_for_each_xdp_txq(vsi, i) {
2794 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2795 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2796 }
2797 mutex_unlock(&pf->avail_q_mutex);
2798
2799 devm_kfree(dev, vsi->xdp_rings);
2800 return -ENOMEM;
2801}
2802
2803/**
2804 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2805 * @vsi: VSI to remove XDP rings
2806 *
2807 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2808 * resources
2809 */
2810int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2811{
2812 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2813 struct ice_pf *pf = vsi->back;
2814 int i, v_idx;
2815
2816 /* q_vectors are freed in reset path so there's no point in detaching
2817 * rings; in case of rebuild being triggered not from reset bits
2818 * in pf->state won't be set, so additionally check first q_vector
2819 * against NULL
2820 */
2821 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2822 goto free_qmap;
2823
2824 ice_for_each_q_vector(vsi, v_idx) {
2825 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2826 struct ice_tx_ring *ring;
2827
2828 ice_for_each_tx_ring(ring, q_vector->tx)
2829 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2830 break;
2831
2832 /* restore the value of last node prior to XDP setup */
2833 q_vector->tx.tx_ring = ring;
2834 }
2835
2836free_qmap:
2837 mutex_lock(&pf->avail_q_mutex);
2838 ice_for_each_xdp_txq(vsi, i) {
2839 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2840 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2841 }
2842 mutex_unlock(&pf->avail_q_mutex);
2843
2844 ice_for_each_xdp_txq(vsi, i)
2845 if (vsi->xdp_rings[i]) {
2846 if (vsi->xdp_rings[i]->desc) {
2847 synchronize_rcu();
2848 ice_free_tx_ring(vsi->xdp_rings[i]);
2849 }
2850 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2851 vsi->xdp_rings[i]->ring_stats = NULL;
2852 kfree_rcu(vsi->xdp_rings[i], rcu);
2853 vsi->xdp_rings[i] = NULL;
2854 }
2855
2856 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2857 vsi->xdp_rings = NULL;
2858
2859 if (static_key_enabled(&ice_xdp_locking_key))
2860 static_branch_dec(&ice_xdp_locking_key);
2861
2862 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2863 return 0;
2864
2865 ice_vsi_assign_bpf_prog(vsi, NULL);
2866
2867 /* notify Tx scheduler that we destroyed XDP queues and bring
2868 * back the old number of child nodes
2869 */
2870 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2871 max_txqs[i] = vsi->num_txq;
2872
2873 /* change number of XDP Tx queues to 0 */
2874 vsi->num_xdp_txq = 0;
2875
2876 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2877 max_txqs);
2878}
2879
2880/**
2881 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2882 * @vsi: VSI to schedule napi on
2883 */
2884static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2885{
2886 int i;
2887
2888 ice_for_each_rxq(vsi, i) {
2889 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2890
2891 if (rx_ring->xsk_pool)
2892 napi_schedule(&rx_ring->q_vector->napi);
2893 }
2894}
2895
2896/**
2897 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2898 * @vsi: VSI to determine the count of XDP Tx qs
2899 *
2900 * returns 0 if Tx qs count is higher than at least half of CPU count,
2901 * -ENOMEM otherwise
2902 */
2903int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2904{
2905 u16 avail = ice_get_avail_txq_count(vsi->back);
2906 u16 cpus = num_possible_cpus();
2907
2908 if (avail < cpus / 2)
2909 return -ENOMEM;
2910
2911 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2912
2913 if (vsi->num_xdp_txq < cpus)
2914 static_branch_inc(&ice_xdp_locking_key);
2915
2916 return 0;
2917}
2918
2919/**
2920 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2921 * @vsi: Pointer to VSI structure
2922 */
2923static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
2924{
2925 if (test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
2926 return ICE_RXBUF_1664;
2927 else
2928 return ICE_RXBUF_3072;
2929}
2930
2931/**
2932 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2933 * @vsi: VSI to setup XDP for
2934 * @prog: XDP program
2935 * @extack: netlink extended ack
2936 */
2937static int
2938ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2939 struct netlink_ext_ack *extack)
2940{
2941 unsigned int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2942 bool if_running = netif_running(vsi->netdev);
2943 int ret = 0, xdp_ring_err = 0;
2944
2945 if (prog && !prog->aux->xdp_has_frags) {
2946 if (frame_size > ice_max_xdp_frame_size(vsi)) {
2947 NL_SET_ERR_MSG_MOD(extack,
2948 "MTU is too large for linear frames and XDP prog does not support frags");
2949 return -EOPNOTSUPP;
2950 }
2951 }
2952
2953 /* hot swap progs and avoid toggling link */
2954 if (ice_is_xdp_ena_vsi(vsi) == !!prog) {
2955 ice_vsi_assign_bpf_prog(vsi, prog);
2956 return 0;
2957 }
2958
2959 /* need to stop netdev while setting up the program for Rx rings */
2960 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2961 ret = ice_down(vsi);
2962 if (ret) {
2963 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2964 return ret;
2965 }
2966 }
2967
2968 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2969 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2970 if (xdp_ring_err) {
2971 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2972 } else {
2973 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2974 if (xdp_ring_err)
2975 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2976 }
2977 xdp_features_set_redirect_target(vsi->netdev, true);
2978 /* reallocate Rx queues that are used for zero-copy */
2979 xdp_ring_err = ice_realloc_zc_buf(vsi, true);
2980 if (xdp_ring_err)
2981 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed");
2982 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2983 xdp_features_clear_redirect_target(vsi->netdev);
2984 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2985 if (xdp_ring_err)
2986 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2987 /* reallocate Rx queues that were used for zero-copy */
2988 xdp_ring_err = ice_realloc_zc_buf(vsi, false);
2989 if (xdp_ring_err)
2990 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed");
2991 }
2992
2993 if (if_running)
2994 ret = ice_up(vsi);
2995
2996 if (!ret && prog)
2997 ice_vsi_rx_napi_schedule(vsi);
2998
2999 return (ret || xdp_ring_err) ? -ENOMEM : 0;
3000}
3001
3002/**
3003 * ice_xdp_safe_mode - XDP handler for safe mode
3004 * @dev: netdevice
3005 * @xdp: XDP command
3006 */
3007static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
3008 struct netdev_bpf *xdp)
3009{
3010 NL_SET_ERR_MSG_MOD(xdp->extack,
3011 "Please provide working DDP firmware package in order to use XDP\n"
3012 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
3013 return -EOPNOTSUPP;
3014}
3015
3016/**
3017 * ice_xdp - implements XDP handler
3018 * @dev: netdevice
3019 * @xdp: XDP command
3020 */
3021static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
3022{
3023 struct ice_netdev_priv *np = netdev_priv(dev);
3024 struct ice_vsi *vsi = np->vsi;
3025
3026 if (vsi->type != ICE_VSI_PF) {
3027 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
3028 return -EINVAL;
3029 }
3030
3031 switch (xdp->command) {
3032 case XDP_SETUP_PROG:
3033 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
3034 case XDP_SETUP_XSK_POOL:
3035 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
3036 xdp->xsk.queue_id);
3037 default:
3038 return -EINVAL;
3039 }
3040}
3041
3042/**
3043 * ice_ena_misc_vector - enable the non-queue interrupts
3044 * @pf: board private structure
3045 */
3046static void ice_ena_misc_vector(struct ice_pf *pf)
3047{
3048 struct ice_hw *hw = &pf->hw;
3049 u32 pf_intr_start_offset;
3050 u32 val;
3051
3052 /* Disable anti-spoof detection interrupt to prevent spurious event
3053 * interrupts during a function reset. Anti-spoof functionally is
3054 * still supported.
3055 */
3056 val = rd32(hw, GL_MDCK_TX_TDPU);
3057 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
3058 wr32(hw, GL_MDCK_TX_TDPU, val);
3059
3060 /* clear things first */
3061 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
3062 rd32(hw, PFINT_OICR); /* read to clear */
3063
3064 val = (PFINT_OICR_ECC_ERR_M |
3065 PFINT_OICR_MAL_DETECT_M |
3066 PFINT_OICR_GRST_M |
3067 PFINT_OICR_PCI_EXCEPTION_M |
3068 PFINT_OICR_VFLR_M |
3069 PFINT_OICR_HMC_ERR_M |
3070 PFINT_OICR_PE_PUSH_M |
3071 PFINT_OICR_PE_CRITERR_M);
3072
3073 wr32(hw, PFINT_OICR_ENA, val);
3074
3075 /* SW_ITR_IDX = 0, but don't change INTENA */
3076 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
3077 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3078
3079 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3080 return;
3081 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3082 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset),
3083 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3084}
3085
3086/**
3087 * ice_ll_ts_intr - ll_ts interrupt handler
3088 * @irq: interrupt number
3089 * @data: pointer to a q_vector
3090 */
3091static irqreturn_t ice_ll_ts_intr(int __always_unused irq, void *data)
3092{
3093 struct ice_pf *pf = data;
3094 u32 pf_intr_start_offset;
3095 struct ice_ptp_tx *tx;
3096 unsigned long flags;
3097 struct ice_hw *hw;
3098 u32 val;
3099 u8 idx;
3100
3101 hw = &pf->hw;
3102 tx = &pf->ptp.port.tx;
3103 spin_lock_irqsave(&tx->lock, flags);
3104 ice_ptp_complete_tx_single_tstamp(tx);
3105
3106 idx = find_next_bit_wrap(tx->in_use, tx->len,
3107 tx->last_ll_ts_idx_read + 1);
3108 if (idx != tx->len)
3109 ice_ptp_req_tx_single_tstamp(tx, idx);
3110 spin_unlock_irqrestore(&tx->lock, flags);
3111
3112 val = GLINT_DYN_CTL_INTENA_M | GLINT_DYN_CTL_CLEARPBA_M |
3113 (ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
3114 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3115 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset),
3116 val);
3117
3118 return IRQ_HANDLED;
3119}
3120
3121/**
3122 * ice_misc_intr - misc interrupt handler
3123 * @irq: interrupt number
3124 * @data: pointer to a q_vector
3125 */
3126static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
3127{
3128 struct ice_pf *pf = (struct ice_pf *)data;
3129 irqreturn_t ret = IRQ_HANDLED;
3130 struct ice_hw *hw = &pf->hw;
3131 struct device *dev;
3132 u32 oicr, ena_mask;
3133
3134 dev = ice_pf_to_dev(pf);
3135 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3136 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3137 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3138
3139 oicr = rd32(hw, PFINT_OICR);
3140 ena_mask = rd32(hw, PFINT_OICR_ENA);
3141
3142 if (oicr & PFINT_OICR_SWINT_M) {
3143 ena_mask &= ~PFINT_OICR_SWINT_M;
3144 pf->sw_int_count++;
3145 }
3146
3147 if (oicr & PFINT_OICR_MAL_DETECT_M) {
3148 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3149 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3150 }
3151 if (oicr & PFINT_OICR_VFLR_M) {
3152 /* disable any further VFLR event notifications */
3153 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3154 u32 reg = rd32(hw, PFINT_OICR_ENA);
3155
3156 reg &= ~PFINT_OICR_VFLR_M;
3157 wr32(hw, PFINT_OICR_ENA, reg);
3158 } else {
3159 ena_mask &= ~PFINT_OICR_VFLR_M;
3160 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3161 }
3162 }
3163
3164 if (oicr & PFINT_OICR_GRST_M) {
3165 u32 reset;
3166
3167 /* we have a reset warning */
3168 ena_mask &= ~PFINT_OICR_GRST_M;
3169 reset = FIELD_GET(GLGEN_RSTAT_RESET_TYPE_M,
3170 rd32(hw, GLGEN_RSTAT));
3171
3172 if (reset == ICE_RESET_CORER)
3173 pf->corer_count++;
3174 else if (reset == ICE_RESET_GLOBR)
3175 pf->globr_count++;
3176 else if (reset == ICE_RESET_EMPR)
3177 pf->empr_count++;
3178 else
3179 dev_dbg(dev, "Invalid reset type %d\n", reset);
3180
3181 /* If a reset cycle isn't already in progress, we set a bit in
3182 * pf->state so that the service task can start a reset/rebuild.
3183 */
3184 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3185 if (reset == ICE_RESET_CORER)
3186 set_bit(ICE_CORER_RECV, pf->state);
3187 else if (reset == ICE_RESET_GLOBR)
3188 set_bit(ICE_GLOBR_RECV, pf->state);
3189 else
3190 set_bit(ICE_EMPR_RECV, pf->state);
3191
3192 /* There are couple of different bits at play here.
3193 * hw->reset_ongoing indicates whether the hardware is
3194 * in reset. This is set to true when a reset interrupt
3195 * is received and set back to false after the driver
3196 * has determined that the hardware is out of reset.
3197 *
3198 * ICE_RESET_OICR_RECV in pf->state indicates
3199 * that a post reset rebuild is required before the
3200 * driver is operational again. This is set above.
3201 *
3202 * As this is the start of the reset/rebuild cycle, set
3203 * both to indicate that.
3204 */
3205 hw->reset_ongoing = true;
3206 }
3207 }
3208
3209 if (oicr & PFINT_OICR_TSYN_TX_M) {
3210 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3211 if (ice_pf_state_is_nominal(pf) &&
3212 pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) {
3213 struct ice_ptp_tx *tx = &pf->ptp.port.tx;
3214 unsigned long flags;
3215 u8 idx;
3216
3217 spin_lock_irqsave(&tx->lock, flags);
3218 idx = find_next_bit_wrap(tx->in_use, tx->len,
3219 tx->last_ll_ts_idx_read + 1);
3220 if (idx != tx->len)
3221 ice_ptp_req_tx_single_tstamp(tx, idx);
3222 spin_unlock_irqrestore(&tx->lock, flags);
3223 } else if (ice_ptp_pf_handles_tx_interrupt(pf)) {
3224 set_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread);
3225 ret = IRQ_WAKE_THREAD;
3226 }
3227 }
3228
3229 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3230 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3231 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3232
3233 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3234
3235 if (ice_pf_src_tmr_owned(pf)) {
3236 /* Save EVENTs from GLTSYN register */
3237 pf->ptp.ext_ts_irq |= gltsyn_stat &
3238 (GLTSYN_STAT_EVENT0_M |
3239 GLTSYN_STAT_EVENT1_M |
3240 GLTSYN_STAT_EVENT2_M);
3241
3242 ice_ptp_extts_event(pf);
3243 }
3244 }
3245
3246#define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3247 if (oicr & ICE_AUX_CRIT_ERR) {
3248 pf->oicr_err_reg |= oicr;
3249 set_bit(ICE_AUX_ERR_PENDING, pf->state);
3250 ena_mask &= ~ICE_AUX_CRIT_ERR;
3251 }
3252
3253 /* Report any remaining unexpected interrupts */
3254 oicr &= ena_mask;
3255 if (oicr) {
3256 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3257 /* If a critical error is pending there is no choice but to
3258 * reset the device.
3259 */
3260 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3261 PFINT_OICR_ECC_ERR_M)) {
3262 set_bit(ICE_PFR_REQ, pf->state);
3263 }
3264 }
3265 ice_service_task_schedule(pf);
3266 if (ret == IRQ_HANDLED)
3267 ice_irq_dynamic_ena(hw, NULL, NULL);
3268
3269 return ret;
3270}
3271
3272/**
3273 * ice_misc_intr_thread_fn - misc interrupt thread function
3274 * @irq: interrupt number
3275 * @data: pointer to a q_vector
3276 */
3277static irqreturn_t ice_misc_intr_thread_fn(int __always_unused irq, void *data)
3278{
3279 struct ice_pf *pf = data;
3280 struct ice_hw *hw;
3281
3282 hw = &pf->hw;
3283
3284 if (ice_is_reset_in_progress(pf->state))
3285 goto skip_irq;
3286
3287 if (test_and_clear_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread)) {
3288 /* Process outstanding Tx timestamps. If there is more work,
3289 * re-arm the interrupt to trigger again.
3290 */
3291 if (ice_ptp_process_ts(pf) == ICE_TX_TSTAMP_WORK_PENDING) {
3292 wr32(hw, PFINT_OICR, PFINT_OICR_TSYN_TX_M);
3293 ice_flush(hw);
3294 }
3295 }
3296
3297skip_irq:
3298 ice_irq_dynamic_ena(hw, NULL, NULL);
3299
3300 return IRQ_HANDLED;
3301}
3302
3303/**
3304 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3305 * @hw: pointer to HW structure
3306 */
3307static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3308{
3309 /* disable Admin queue Interrupt causes */
3310 wr32(hw, PFINT_FW_CTL,
3311 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3312
3313 /* disable Mailbox queue Interrupt causes */
3314 wr32(hw, PFINT_MBX_CTL,
3315 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3316
3317 wr32(hw, PFINT_SB_CTL,
3318 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3319
3320 /* disable Control queue Interrupt causes */
3321 wr32(hw, PFINT_OICR_CTL,
3322 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3323
3324 ice_flush(hw);
3325}
3326
3327/**
3328 * ice_free_irq_msix_ll_ts- Unroll ll_ts vector setup
3329 * @pf: board private structure
3330 */
3331static void ice_free_irq_msix_ll_ts(struct ice_pf *pf)
3332{
3333 int irq_num = pf->ll_ts_irq.virq;
3334
3335 synchronize_irq(irq_num);
3336 devm_free_irq(ice_pf_to_dev(pf), irq_num, pf);
3337
3338 ice_free_irq(pf, pf->ll_ts_irq);
3339}
3340
3341/**
3342 * ice_free_irq_msix_misc - Unroll misc vector setup
3343 * @pf: board private structure
3344 */
3345static void ice_free_irq_msix_misc(struct ice_pf *pf)
3346{
3347 int misc_irq_num = pf->oicr_irq.virq;
3348 struct ice_hw *hw = &pf->hw;
3349
3350 ice_dis_ctrlq_interrupts(hw);
3351
3352 /* disable OICR interrupt */
3353 wr32(hw, PFINT_OICR_ENA, 0);
3354 ice_flush(hw);
3355
3356 synchronize_irq(misc_irq_num);
3357 devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
3358
3359 ice_free_irq(pf, pf->oicr_irq);
3360 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3361 ice_free_irq_msix_ll_ts(pf);
3362}
3363
3364/**
3365 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3366 * @hw: pointer to HW structure
3367 * @reg_idx: HW vector index to associate the control queue interrupts with
3368 */
3369static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3370{
3371 u32 val;
3372
3373 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3374 PFINT_OICR_CTL_CAUSE_ENA_M);
3375 wr32(hw, PFINT_OICR_CTL, val);
3376
3377 /* enable Admin queue Interrupt causes */
3378 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3379 PFINT_FW_CTL_CAUSE_ENA_M);
3380 wr32(hw, PFINT_FW_CTL, val);
3381
3382 /* enable Mailbox queue Interrupt causes */
3383 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3384 PFINT_MBX_CTL_CAUSE_ENA_M);
3385 wr32(hw, PFINT_MBX_CTL, val);
3386
3387 if (!hw->dev_caps.ts_dev_info.ts_ll_int_read) {
3388 /* enable Sideband queue Interrupt causes */
3389 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3390 PFINT_SB_CTL_CAUSE_ENA_M);
3391 wr32(hw, PFINT_SB_CTL, val);
3392 }
3393
3394 ice_flush(hw);
3395}
3396
3397/**
3398 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3399 * @pf: board private structure
3400 *
3401 * This sets up the handler for MSIX 0, which is used to manage the
3402 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3403 * when in MSI or Legacy interrupt mode.
3404 */
3405static int ice_req_irq_msix_misc(struct ice_pf *pf)
3406{
3407 struct device *dev = ice_pf_to_dev(pf);
3408 struct ice_hw *hw = &pf->hw;
3409 u32 pf_intr_start_offset;
3410 struct msi_map irq;
3411 int err = 0;
3412
3413 if (!pf->int_name[0])
3414 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3415 dev_driver_string(dev), dev_name(dev));
3416
3417 if (!pf->int_name_ll_ts[0])
3418 snprintf(pf->int_name_ll_ts, sizeof(pf->int_name_ll_ts) - 1,
3419 "%s-%s:ll_ts", dev_driver_string(dev), dev_name(dev));
3420 /* Do not request IRQ but do enable OICR interrupt since settings are
3421 * lost during reset. Note that this function is called only during
3422 * rebuild path and not while reset is in progress.
3423 */
3424 if (ice_is_reset_in_progress(pf->state))
3425 goto skip_req_irq;
3426
3427 /* reserve one vector in irq_tracker for misc interrupts */
3428 irq = ice_alloc_irq(pf, false);
3429 if (irq.index < 0)
3430 return irq.index;
3431
3432 pf->oicr_irq = irq;
3433 err = devm_request_threaded_irq(dev, pf->oicr_irq.virq, ice_misc_intr,
3434 ice_misc_intr_thread_fn, 0,
3435 pf->int_name, pf);
3436 if (err) {
3437 dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n",
3438 pf->int_name, err);
3439 ice_free_irq(pf, pf->oicr_irq);
3440 return err;
3441 }
3442
3443 /* reserve one vector in irq_tracker for ll_ts interrupt */
3444 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3445 goto skip_req_irq;
3446
3447 irq = ice_alloc_irq(pf, false);
3448 if (irq.index < 0)
3449 return irq.index;
3450
3451 pf->ll_ts_irq = irq;
3452 err = devm_request_irq(dev, pf->ll_ts_irq.virq, ice_ll_ts_intr, 0,
3453 pf->int_name_ll_ts, pf);
3454 if (err) {
3455 dev_err(dev, "devm_request_irq for %s failed: %d\n",
3456 pf->int_name_ll_ts, err);
3457 ice_free_irq(pf, pf->ll_ts_irq);
3458 return err;
3459 }
3460
3461skip_req_irq:
3462 ice_ena_misc_vector(pf);
3463
3464 ice_ena_ctrlq_interrupts(hw, pf->oicr_irq.index);
3465 /* This enables LL TS interrupt */
3466 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3467 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3468 wr32(hw, PFINT_SB_CTL,
3469 ((pf->ll_ts_irq.index + pf_intr_start_offset) &
3470 PFINT_SB_CTL_MSIX_INDX_M) | PFINT_SB_CTL_CAUSE_ENA_M);
3471 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_irq.index),
3472 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3473
3474 ice_flush(hw);
3475 ice_irq_dynamic_ena(hw, NULL, NULL);
3476
3477 return 0;
3478}
3479
3480/**
3481 * ice_napi_add - register NAPI handler for the VSI
3482 * @vsi: VSI for which NAPI handler is to be registered
3483 *
3484 * This function is only called in the driver's load path. Registering the NAPI
3485 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3486 * reset/rebuild, etc.)
3487 */
3488static void ice_napi_add(struct ice_vsi *vsi)
3489{
3490 int v_idx;
3491
3492 if (!vsi->netdev)
3493 return;
3494
3495 ice_for_each_q_vector(vsi, v_idx) {
3496 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3497 ice_napi_poll);
3498 __ice_q_vector_set_napi_queues(vsi->q_vectors[v_idx], false);
3499 }
3500}
3501
3502/**
3503 * ice_set_ops - set netdev and ethtools ops for the given netdev
3504 * @vsi: the VSI associated with the new netdev
3505 */
3506static void ice_set_ops(struct ice_vsi *vsi)
3507{
3508 struct net_device *netdev = vsi->netdev;
3509 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3510
3511 if (ice_is_safe_mode(pf)) {
3512 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3513 ice_set_ethtool_safe_mode_ops(netdev);
3514 return;
3515 }
3516
3517 netdev->netdev_ops = &ice_netdev_ops;
3518 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3519 netdev->xdp_metadata_ops = &ice_xdp_md_ops;
3520 ice_set_ethtool_ops(netdev);
3521
3522 if (vsi->type != ICE_VSI_PF)
3523 return;
3524
3525 netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
3526 NETDEV_XDP_ACT_XSK_ZEROCOPY |
3527 NETDEV_XDP_ACT_RX_SG;
3528 netdev->xdp_zc_max_segs = ICE_MAX_BUF_TXD;
3529}
3530
3531/**
3532 * ice_set_netdev_features - set features for the given netdev
3533 * @netdev: netdev instance
3534 */
3535static void ice_set_netdev_features(struct net_device *netdev)
3536{
3537 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3538 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3539 netdev_features_t csumo_features;
3540 netdev_features_t vlano_features;
3541 netdev_features_t dflt_features;
3542 netdev_features_t tso_features;
3543
3544 if (ice_is_safe_mode(pf)) {
3545 /* safe mode */
3546 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3547 netdev->hw_features = netdev->features;
3548 return;
3549 }
3550
3551 dflt_features = NETIF_F_SG |
3552 NETIF_F_HIGHDMA |
3553 NETIF_F_NTUPLE |
3554 NETIF_F_RXHASH;
3555
3556 csumo_features = NETIF_F_RXCSUM |
3557 NETIF_F_IP_CSUM |
3558 NETIF_F_SCTP_CRC |
3559 NETIF_F_IPV6_CSUM;
3560
3561 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3562 NETIF_F_HW_VLAN_CTAG_TX |
3563 NETIF_F_HW_VLAN_CTAG_RX;
3564
3565 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3566 if (is_dvm_ena)
3567 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3568
3569 tso_features = NETIF_F_TSO |
3570 NETIF_F_TSO_ECN |
3571 NETIF_F_TSO6 |
3572 NETIF_F_GSO_GRE |
3573 NETIF_F_GSO_UDP_TUNNEL |
3574 NETIF_F_GSO_GRE_CSUM |
3575 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3576 NETIF_F_GSO_PARTIAL |
3577 NETIF_F_GSO_IPXIP4 |
3578 NETIF_F_GSO_IPXIP6 |
3579 NETIF_F_GSO_UDP_L4;
3580
3581 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3582 NETIF_F_GSO_GRE_CSUM;
3583 /* set features that user can change */
3584 netdev->hw_features = dflt_features | csumo_features |
3585 vlano_features | tso_features;
3586
3587 /* add support for HW_CSUM on packets with MPLS header */
3588 netdev->mpls_features = NETIF_F_HW_CSUM |
3589 NETIF_F_TSO |
3590 NETIF_F_TSO6;
3591
3592 /* enable features */
3593 netdev->features |= netdev->hw_features;
3594
3595 netdev->hw_features |= NETIF_F_HW_TC;
3596 netdev->hw_features |= NETIF_F_LOOPBACK;
3597
3598 /* encap and VLAN devices inherit default, csumo and tso features */
3599 netdev->hw_enc_features |= dflt_features | csumo_features |
3600 tso_features;
3601 netdev->vlan_features |= dflt_features | csumo_features |
3602 tso_features;
3603
3604 /* advertise support but don't enable by default since only one type of
3605 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3606 * type turns on the other has to be turned off. This is enforced by the
3607 * ice_fix_features() ndo callback.
3608 */
3609 if (is_dvm_ena)
3610 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3611 NETIF_F_HW_VLAN_STAG_TX;
3612
3613 /* Leave CRC / FCS stripping enabled by default, but allow the value to
3614 * be changed at runtime
3615 */
3616 netdev->hw_features |= NETIF_F_RXFCS;
3617
3618 netif_set_tso_max_size(netdev, ICE_MAX_TSO_SIZE);
3619}
3620
3621/**
3622 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3623 * @lut: Lookup table
3624 * @rss_table_size: Lookup table size
3625 * @rss_size: Range of queue number for hashing
3626 */
3627void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3628{
3629 u16 i;
3630
3631 for (i = 0; i < rss_table_size; i++)
3632 lut[i] = i % rss_size;
3633}
3634
3635/**
3636 * ice_pf_vsi_setup - Set up a PF VSI
3637 * @pf: board private structure
3638 * @pi: pointer to the port_info instance
3639 *
3640 * Returns pointer to the successfully allocated VSI software struct
3641 * on success, otherwise returns NULL on failure.
3642 */
3643static struct ice_vsi *
3644ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3645{
3646 struct ice_vsi_cfg_params params = {};
3647
3648 params.type = ICE_VSI_PF;
3649 params.pi = pi;
3650 params.flags = ICE_VSI_FLAG_INIT;
3651
3652 return ice_vsi_setup(pf, ¶ms);
3653}
3654
3655static struct ice_vsi *
3656ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3657 struct ice_channel *ch)
3658{
3659 struct ice_vsi_cfg_params params = {};
3660
3661 params.type = ICE_VSI_CHNL;
3662 params.pi = pi;
3663 params.ch = ch;
3664 params.flags = ICE_VSI_FLAG_INIT;
3665
3666 return ice_vsi_setup(pf, ¶ms);
3667}
3668
3669/**
3670 * ice_ctrl_vsi_setup - Set up a control VSI
3671 * @pf: board private structure
3672 * @pi: pointer to the port_info instance
3673 *
3674 * Returns pointer to the successfully allocated VSI software struct
3675 * on success, otherwise returns NULL on failure.
3676 */
3677static struct ice_vsi *
3678ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3679{
3680 struct ice_vsi_cfg_params params = {};
3681
3682 params.type = ICE_VSI_CTRL;
3683 params.pi = pi;
3684 params.flags = ICE_VSI_FLAG_INIT;
3685
3686 return ice_vsi_setup(pf, ¶ms);
3687}
3688
3689/**
3690 * ice_lb_vsi_setup - Set up a loopback VSI
3691 * @pf: board private structure
3692 * @pi: pointer to the port_info instance
3693 *
3694 * Returns pointer to the successfully allocated VSI software struct
3695 * on success, otherwise returns NULL on failure.
3696 */
3697struct ice_vsi *
3698ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3699{
3700 struct ice_vsi_cfg_params params = {};
3701
3702 params.type = ICE_VSI_LB;
3703 params.pi = pi;
3704 params.flags = ICE_VSI_FLAG_INIT;
3705
3706 return ice_vsi_setup(pf, ¶ms);
3707}
3708
3709/**
3710 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3711 * @netdev: network interface to be adjusted
3712 * @proto: VLAN TPID
3713 * @vid: VLAN ID to be added
3714 *
3715 * net_device_ops implementation for adding VLAN IDs
3716 */
3717static int
3718ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3719{
3720 struct ice_netdev_priv *np = netdev_priv(netdev);
3721 struct ice_vsi_vlan_ops *vlan_ops;
3722 struct ice_vsi *vsi = np->vsi;
3723 struct ice_vlan vlan;
3724 int ret;
3725
3726 /* VLAN 0 is added by default during load/reset */
3727 if (!vid)
3728 return 0;
3729
3730 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3731 usleep_range(1000, 2000);
3732
3733 /* Add multicast promisc rule for the VLAN ID to be added if
3734 * all-multicast is currently enabled.
3735 */
3736 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3737 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3738 ICE_MCAST_VLAN_PROMISC_BITS,
3739 vid);
3740 if (ret)
3741 goto finish;
3742 }
3743
3744 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3745
3746 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3747 * packets aren't pruned by the device's internal switch on Rx
3748 */
3749 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3750 ret = vlan_ops->add_vlan(vsi, &vlan);
3751 if (ret)
3752 goto finish;
3753
3754 /* If all-multicast is currently enabled and this VLAN ID is only one
3755 * besides VLAN-0 we have to update look-up type of multicast promisc
3756 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3757 */
3758 if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3759 ice_vsi_num_non_zero_vlans(vsi) == 1) {
3760 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3761 ICE_MCAST_PROMISC_BITS, 0);
3762 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3763 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3764 }
3765
3766finish:
3767 clear_bit(ICE_CFG_BUSY, vsi->state);
3768
3769 return ret;
3770}
3771
3772/**
3773 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3774 * @netdev: network interface to be adjusted
3775 * @proto: VLAN TPID
3776 * @vid: VLAN ID to be removed
3777 *
3778 * net_device_ops implementation for removing VLAN IDs
3779 */
3780static int
3781ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3782{
3783 struct ice_netdev_priv *np = netdev_priv(netdev);
3784 struct ice_vsi_vlan_ops *vlan_ops;
3785 struct ice_vsi *vsi = np->vsi;
3786 struct ice_vlan vlan;
3787 int ret;
3788
3789 /* don't allow removal of VLAN 0 */
3790 if (!vid)
3791 return 0;
3792
3793 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3794 usleep_range(1000, 2000);
3795
3796 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3797 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3798 if (ret) {
3799 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3800 vsi->vsi_num);
3801 vsi->current_netdev_flags |= IFF_ALLMULTI;
3802 }
3803
3804 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3805
3806 /* Make sure VLAN delete is successful before updating VLAN
3807 * information
3808 */
3809 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3810 ret = vlan_ops->del_vlan(vsi, &vlan);
3811 if (ret)
3812 goto finish;
3813
3814 /* Remove multicast promisc rule for the removed VLAN ID if
3815 * all-multicast is enabled.
3816 */
3817 if (vsi->current_netdev_flags & IFF_ALLMULTI)
3818 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3819 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3820
3821 if (!ice_vsi_has_non_zero_vlans(vsi)) {
3822 /* Update look-up type of multicast promisc rule for VLAN 0
3823 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3824 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3825 */
3826 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3827 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3828 ICE_MCAST_VLAN_PROMISC_BITS,
3829 0);
3830 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3831 ICE_MCAST_PROMISC_BITS, 0);
3832 }
3833 }
3834
3835finish:
3836 clear_bit(ICE_CFG_BUSY, vsi->state);
3837
3838 return ret;
3839}
3840
3841/**
3842 * ice_rep_indr_tc_block_unbind
3843 * @cb_priv: indirection block private data
3844 */
3845static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3846{
3847 struct ice_indr_block_priv *indr_priv = cb_priv;
3848
3849 list_del(&indr_priv->list);
3850 kfree(indr_priv);
3851}
3852
3853/**
3854 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3855 * @vsi: VSI struct which has the netdev
3856 */
3857static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3858{
3859 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3860
3861 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3862 ice_rep_indr_tc_block_unbind);
3863}
3864
3865/**
3866 * ice_tc_indir_block_register - Register TC indirect block notifications
3867 * @vsi: VSI struct which has the netdev
3868 *
3869 * Returns 0 on success, negative value on failure
3870 */
3871static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3872{
3873 struct ice_netdev_priv *np;
3874
3875 if (!vsi || !vsi->netdev)
3876 return -EINVAL;
3877
3878 np = netdev_priv(vsi->netdev);
3879
3880 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3881 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3882}
3883
3884/**
3885 * ice_get_avail_q_count - Get count of queues in use
3886 * @pf_qmap: bitmap to get queue use count from
3887 * @lock: pointer to a mutex that protects access to pf_qmap
3888 * @size: size of the bitmap
3889 */
3890static u16
3891ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3892{
3893 unsigned long bit;
3894 u16 count = 0;
3895
3896 mutex_lock(lock);
3897 for_each_clear_bit(bit, pf_qmap, size)
3898 count++;
3899 mutex_unlock(lock);
3900
3901 return count;
3902}
3903
3904/**
3905 * ice_get_avail_txq_count - Get count of Tx queues in use
3906 * @pf: pointer to an ice_pf instance
3907 */
3908u16 ice_get_avail_txq_count(struct ice_pf *pf)
3909{
3910 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3911 pf->max_pf_txqs);
3912}
3913
3914/**
3915 * ice_get_avail_rxq_count - Get count of Rx queues in use
3916 * @pf: pointer to an ice_pf instance
3917 */
3918u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3919{
3920 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3921 pf->max_pf_rxqs);
3922}
3923
3924/**
3925 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3926 * @pf: board private structure to initialize
3927 */
3928static void ice_deinit_pf(struct ice_pf *pf)
3929{
3930 ice_service_task_stop(pf);
3931 mutex_destroy(&pf->lag_mutex);
3932 mutex_destroy(&pf->adev_mutex);
3933 mutex_destroy(&pf->sw_mutex);
3934 mutex_destroy(&pf->tc_mutex);
3935 mutex_destroy(&pf->avail_q_mutex);
3936 mutex_destroy(&pf->vfs.table_lock);
3937
3938 if (pf->avail_txqs) {
3939 bitmap_free(pf->avail_txqs);
3940 pf->avail_txqs = NULL;
3941 }
3942
3943 if (pf->avail_rxqs) {
3944 bitmap_free(pf->avail_rxqs);
3945 pf->avail_rxqs = NULL;
3946 }
3947
3948 if (pf->ptp.clock)
3949 ptp_clock_unregister(pf->ptp.clock);
3950}
3951
3952/**
3953 * ice_set_pf_caps - set PFs capability flags
3954 * @pf: pointer to the PF instance
3955 */
3956static void ice_set_pf_caps(struct ice_pf *pf)
3957{
3958 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3959
3960 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3961 if (func_caps->common_cap.rdma)
3962 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3963 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3964 if (func_caps->common_cap.dcb)
3965 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3966 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3967 if (func_caps->common_cap.sr_iov_1_1) {
3968 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3969 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3970 ICE_MAX_SRIOV_VFS);
3971 }
3972 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3973 if (func_caps->common_cap.rss_table_size)
3974 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3975
3976 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3977 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3978 u16 unused;
3979
3980 /* ctrl_vsi_idx will be set to a valid value when flow director
3981 * is setup by ice_init_fdir
3982 */
3983 pf->ctrl_vsi_idx = ICE_NO_VSI;
3984 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3985 /* force guaranteed filter pool for PF */
3986 ice_alloc_fd_guar_item(&pf->hw, &unused,
3987 func_caps->fd_fltr_guar);
3988 /* force shared filter pool for PF */
3989 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3990 func_caps->fd_fltr_best_effort);
3991 }
3992
3993 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3994 if (func_caps->common_cap.ieee_1588 &&
3995 !(pf->hw.mac_type == ICE_MAC_E830))
3996 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3997
3998 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3999 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
4000}
4001
4002/**
4003 * ice_init_pf - Initialize general software structures (struct ice_pf)
4004 * @pf: board private structure to initialize
4005 */
4006static int ice_init_pf(struct ice_pf *pf)
4007{
4008 ice_set_pf_caps(pf);
4009
4010 mutex_init(&pf->sw_mutex);
4011 mutex_init(&pf->tc_mutex);
4012 mutex_init(&pf->adev_mutex);
4013 mutex_init(&pf->lag_mutex);
4014
4015 INIT_HLIST_HEAD(&pf->aq_wait_list);
4016 spin_lock_init(&pf->aq_wait_lock);
4017 init_waitqueue_head(&pf->aq_wait_queue);
4018
4019 init_waitqueue_head(&pf->reset_wait_queue);
4020
4021 /* setup service timer and periodic service task */
4022 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
4023 pf->serv_tmr_period = HZ;
4024 INIT_WORK(&pf->serv_task, ice_service_task);
4025 clear_bit(ICE_SERVICE_SCHED, pf->state);
4026
4027 mutex_init(&pf->avail_q_mutex);
4028 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
4029 if (!pf->avail_txqs)
4030 return -ENOMEM;
4031
4032 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
4033 if (!pf->avail_rxqs) {
4034 bitmap_free(pf->avail_txqs);
4035 pf->avail_txqs = NULL;
4036 return -ENOMEM;
4037 }
4038
4039 mutex_init(&pf->vfs.table_lock);
4040 hash_init(pf->vfs.table);
4041 ice_mbx_init_snapshot(&pf->hw);
4042
4043 return 0;
4044}
4045
4046/**
4047 * ice_is_wol_supported - check if WoL is supported
4048 * @hw: pointer to hardware info
4049 *
4050 * Check if WoL is supported based on the HW configuration.
4051 * Returns true if NVM supports and enables WoL for this port, false otherwise
4052 */
4053bool ice_is_wol_supported(struct ice_hw *hw)
4054{
4055 u16 wol_ctrl;
4056
4057 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4058 * word) indicates WoL is not supported on the corresponding PF ID.
4059 */
4060 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4061 return false;
4062
4063 return !(BIT(hw->port_info->lport) & wol_ctrl);
4064}
4065
4066/**
4067 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4068 * @vsi: VSI being changed
4069 * @new_rx: new number of Rx queues
4070 * @new_tx: new number of Tx queues
4071 * @locked: is adev device_lock held
4072 *
4073 * Only change the number of queues if new_tx, or new_rx is non-0.
4074 *
4075 * Returns 0 on success.
4076 */
4077int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx, bool locked)
4078{
4079 struct ice_pf *pf = vsi->back;
4080 int err = 0, timeout = 50;
4081
4082 if (!new_rx && !new_tx)
4083 return -EINVAL;
4084
4085 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4086 timeout--;
4087 if (!timeout)
4088 return -EBUSY;
4089 usleep_range(1000, 2000);
4090 }
4091
4092 if (new_tx)
4093 vsi->req_txq = (u16)new_tx;
4094 if (new_rx)
4095 vsi->req_rxq = (u16)new_rx;
4096
4097 /* set for the next time the netdev is started */
4098 if (!netif_running(vsi->netdev)) {
4099 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4100 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4101 goto done;
4102 }
4103
4104 ice_vsi_close(vsi);
4105 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4106 ice_pf_dcb_recfg(pf, locked);
4107 ice_vsi_open(vsi);
4108done:
4109 clear_bit(ICE_CFG_BUSY, pf->state);
4110 return err;
4111}
4112
4113/**
4114 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4115 * @pf: PF to configure
4116 *
4117 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4118 * VSI can still Tx/Rx VLAN tagged packets.
4119 */
4120static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4121{
4122 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4123 struct ice_vsi_ctx *ctxt;
4124 struct ice_hw *hw;
4125 int status;
4126
4127 if (!vsi)
4128 return;
4129
4130 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4131 if (!ctxt)
4132 return;
4133
4134 hw = &pf->hw;
4135 ctxt->info = vsi->info;
4136
4137 ctxt->info.valid_sections =
4138 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4139 ICE_AQ_VSI_PROP_SECURITY_VALID |
4140 ICE_AQ_VSI_PROP_SW_VALID);
4141
4142 /* disable VLAN anti-spoof */
4143 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4144 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4145
4146 /* disable VLAN pruning and keep all other settings */
4147 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4148
4149 /* allow all VLANs on Tx and don't strip on Rx */
4150 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4151 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4152
4153 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4154 if (status) {
4155 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4156 status, ice_aq_str(hw->adminq.sq_last_status));
4157 } else {
4158 vsi->info.sec_flags = ctxt->info.sec_flags;
4159 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4160 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4161 }
4162
4163 kfree(ctxt);
4164}
4165
4166/**
4167 * ice_log_pkg_init - log result of DDP package load
4168 * @hw: pointer to hardware info
4169 * @state: state of package load
4170 */
4171static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4172{
4173 struct ice_pf *pf = hw->back;
4174 struct device *dev;
4175
4176 dev = ice_pf_to_dev(pf);
4177
4178 switch (state) {
4179 case ICE_DDP_PKG_SUCCESS:
4180 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4181 hw->active_pkg_name,
4182 hw->active_pkg_ver.major,
4183 hw->active_pkg_ver.minor,
4184 hw->active_pkg_ver.update,
4185 hw->active_pkg_ver.draft);
4186 break;
4187 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4188 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4189 hw->active_pkg_name,
4190 hw->active_pkg_ver.major,
4191 hw->active_pkg_ver.minor,
4192 hw->active_pkg_ver.update,
4193 hw->active_pkg_ver.draft);
4194 break;
4195 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4196 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",
4197 hw->active_pkg_name,
4198 hw->active_pkg_ver.major,
4199 hw->active_pkg_ver.minor,
4200 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4201 break;
4202 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4203 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",
4204 hw->active_pkg_name,
4205 hw->active_pkg_ver.major,
4206 hw->active_pkg_ver.minor,
4207 hw->active_pkg_ver.update,
4208 hw->active_pkg_ver.draft,
4209 hw->pkg_name,
4210 hw->pkg_ver.major,
4211 hw->pkg_ver.minor,
4212 hw->pkg_ver.update,
4213 hw->pkg_ver.draft);
4214 break;
4215 case ICE_DDP_PKG_FW_MISMATCH:
4216 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");
4217 break;
4218 case ICE_DDP_PKG_INVALID_FILE:
4219 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4220 break;
4221 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4222 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4223 break;
4224 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4225 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",
4226 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4227 break;
4228 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4229 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");
4230 break;
4231 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4232 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");
4233 break;
4234 case ICE_DDP_PKG_LOAD_ERROR:
4235 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4236 /* poll for reset to complete */
4237 if (ice_check_reset(hw))
4238 dev_err(dev, "Error resetting device. Please reload the driver\n");
4239 break;
4240 case ICE_DDP_PKG_ERR:
4241 default:
4242 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4243 break;
4244 }
4245}
4246
4247/**
4248 * ice_load_pkg - load/reload the DDP Package file
4249 * @firmware: firmware structure when firmware requested or NULL for reload
4250 * @pf: pointer to the PF instance
4251 *
4252 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4253 * initialize HW tables.
4254 */
4255static void
4256ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4257{
4258 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4259 struct device *dev = ice_pf_to_dev(pf);
4260 struct ice_hw *hw = &pf->hw;
4261
4262 /* Load DDP Package */
4263 if (firmware && !hw->pkg_copy) {
4264 state = ice_copy_and_init_pkg(hw, firmware->data,
4265 firmware->size);
4266 ice_log_pkg_init(hw, state);
4267 } else if (!firmware && hw->pkg_copy) {
4268 /* Reload package during rebuild after CORER/GLOBR reset */
4269 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4270 ice_log_pkg_init(hw, state);
4271 } else {
4272 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4273 }
4274
4275 if (!ice_is_init_pkg_successful(state)) {
4276 /* Safe Mode */
4277 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4278 return;
4279 }
4280
4281 /* Successful download package is the precondition for advanced
4282 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4283 */
4284 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4285}
4286
4287/**
4288 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4289 * @pf: pointer to the PF structure
4290 *
4291 * There is no error returned here because the driver should be able to handle
4292 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4293 * specifically with Tx.
4294 */
4295static void ice_verify_cacheline_size(struct ice_pf *pf)
4296{
4297 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4298 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4299 ICE_CACHE_LINE_BYTES);
4300}
4301
4302/**
4303 * ice_send_version - update firmware with driver version
4304 * @pf: PF struct
4305 *
4306 * Returns 0 on success, else error code
4307 */
4308static int ice_send_version(struct ice_pf *pf)
4309{
4310 struct ice_driver_ver dv;
4311
4312 dv.major_ver = 0xff;
4313 dv.minor_ver = 0xff;
4314 dv.build_ver = 0xff;
4315 dv.subbuild_ver = 0;
4316 strscpy((char *)dv.driver_string, UTS_RELEASE,
4317 sizeof(dv.driver_string));
4318 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4319}
4320
4321/**
4322 * ice_init_fdir - Initialize flow director VSI and configuration
4323 * @pf: pointer to the PF instance
4324 *
4325 * returns 0 on success, negative on error
4326 */
4327static int ice_init_fdir(struct ice_pf *pf)
4328{
4329 struct device *dev = ice_pf_to_dev(pf);
4330 struct ice_vsi *ctrl_vsi;
4331 int err;
4332
4333 /* Side Band Flow Director needs to have a control VSI.
4334 * Allocate it and store it in the PF.
4335 */
4336 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4337 if (!ctrl_vsi) {
4338 dev_dbg(dev, "could not create control VSI\n");
4339 return -ENOMEM;
4340 }
4341
4342 err = ice_vsi_open_ctrl(ctrl_vsi);
4343 if (err) {
4344 dev_dbg(dev, "could not open control VSI\n");
4345 goto err_vsi_open;
4346 }
4347
4348 mutex_init(&pf->hw.fdir_fltr_lock);
4349
4350 err = ice_fdir_create_dflt_rules(pf);
4351 if (err)
4352 goto err_fdir_rule;
4353
4354 return 0;
4355
4356err_fdir_rule:
4357 ice_fdir_release_flows(&pf->hw);
4358 ice_vsi_close(ctrl_vsi);
4359err_vsi_open:
4360 ice_vsi_release(ctrl_vsi);
4361 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4362 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4363 pf->ctrl_vsi_idx = ICE_NO_VSI;
4364 }
4365 return err;
4366}
4367
4368static void ice_deinit_fdir(struct ice_pf *pf)
4369{
4370 struct ice_vsi *vsi = ice_get_ctrl_vsi(pf);
4371
4372 if (!vsi)
4373 return;
4374
4375 ice_vsi_manage_fdir(vsi, false);
4376 ice_vsi_release(vsi);
4377 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4378 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4379 pf->ctrl_vsi_idx = ICE_NO_VSI;
4380 }
4381
4382 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4383}
4384
4385/**
4386 * ice_get_opt_fw_name - return optional firmware file name or NULL
4387 * @pf: pointer to the PF instance
4388 */
4389static char *ice_get_opt_fw_name(struct ice_pf *pf)
4390{
4391 /* Optional firmware name same as default with additional dash
4392 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4393 */
4394 struct pci_dev *pdev = pf->pdev;
4395 char *opt_fw_filename;
4396 u64 dsn;
4397
4398 /* Determine the name of the optional file using the DSN (two
4399 * dwords following the start of the DSN Capability).
4400 */
4401 dsn = pci_get_dsn(pdev);
4402 if (!dsn)
4403 return NULL;
4404
4405 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4406 if (!opt_fw_filename)
4407 return NULL;
4408
4409 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4410 ICE_DDP_PKG_PATH, dsn);
4411
4412 return opt_fw_filename;
4413}
4414
4415/**
4416 * ice_request_fw - Device initialization routine
4417 * @pf: pointer to the PF instance
4418 */
4419static void ice_request_fw(struct ice_pf *pf)
4420{
4421 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4422 const struct firmware *firmware = NULL;
4423 struct device *dev = ice_pf_to_dev(pf);
4424 int err = 0;
4425
4426 /* optional device-specific DDP (if present) overrides the default DDP
4427 * package file. kernel logs a debug message if the file doesn't exist,
4428 * and warning messages for other errors.
4429 */
4430 if (opt_fw_filename) {
4431 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4432 if (err) {
4433 kfree(opt_fw_filename);
4434 goto dflt_pkg_load;
4435 }
4436
4437 /* request for firmware was successful. Download to device */
4438 ice_load_pkg(firmware, pf);
4439 kfree(opt_fw_filename);
4440 release_firmware(firmware);
4441 return;
4442 }
4443
4444dflt_pkg_load:
4445 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4446 if (err) {
4447 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4448 return;
4449 }
4450
4451 /* request for firmware was successful. Download to device */
4452 ice_load_pkg(firmware, pf);
4453 release_firmware(firmware);
4454}
4455
4456/**
4457 * ice_print_wake_reason - show the wake up cause in the log
4458 * @pf: pointer to the PF struct
4459 */
4460static void ice_print_wake_reason(struct ice_pf *pf)
4461{
4462 u32 wus = pf->wakeup_reason;
4463 const char *wake_str;
4464
4465 /* if no wake event, nothing to print */
4466 if (!wus)
4467 return;
4468
4469 if (wus & PFPM_WUS_LNKC_M)
4470 wake_str = "Link\n";
4471 else if (wus & PFPM_WUS_MAG_M)
4472 wake_str = "Magic Packet\n";
4473 else if (wus & PFPM_WUS_MNG_M)
4474 wake_str = "Management\n";
4475 else if (wus & PFPM_WUS_FW_RST_WK_M)
4476 wake_str = "Firmware Reset\n";
4477 else
4478 wake_str = "Unknown\n";
4479
4480 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4481}
4482
4483/**
4484 * ice_pf_fwlog_update_module - update 1 module
4485 * @pf: pointer to the PF struct
4486 * @log_level: log_level to use for the @module
4487 * @module: module to update
4488 */
4489void ice_pf_fwlog_update_module(struct ice_pf *pf, int log_level, int module)
4490{
4491 struct ice_hw *hw = &pf->hw;
4492
4493 hw->fwlog_cfg.module_entries[module].log_level = log_level;
4494}
4495
4496/**
4497 * ice_register_netdev - register netdev
4498 * @vsi: pointer to the VSI struct
4499 */
4500static int ice_register_netdev(struct ice_vsi *vsi)
4501{
4502 int err;
4503
4504 if (!vsi || !vsi->netdev)
4505 return -EIO;
4506
4507 err = register_netdev(vsi->netdev);
4508 if (err)
4509 return err;
4510
4511 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4512 netif_carrier_off(vsi->netdev);
4513 netif_tx_stop_all_queues(vsi->netdev);
4514
4515 return 0;
4516}
4517
4518static void ice_unregister_netdev(struct ice_vsi *vsi)
4519{
4520 if (!vsi || !vsi->netdev)
4521 return;
4522
4523 unregister_netdev(vsi->netdev);
4524 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4525}
4526
4527/**
4528 * ice_cfg_netdev - Allocate, configure and register a netdev
4529 * @vsi: the VSI associated with the new netdev
4530 *
4531 * Returns 0 on success, negative value on failure
4532 */
4533static int ice_cfg_netdev(struct ice_vsi *vsi)
4534{
4535 struct ice_netdev_priv *np;
4536 struct net_device *netdev;
4537 u8 mac_addr[ETH_ALEN];
4538
4539 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
4540 vsi->alloc_rxq);
4541 if (!netdev)
4542 return -ENOMEM;
4543
4544 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4545 vsi->netdev = netdev;
4546 np = netdev_priv(netdev);
4547 np->vsi = vsi;
4548
4549 ice_set_netdev_features(netdev);
4550 ice_set_ops(vsi);
4551
4552 if (vsi->type == ICE_VSI_PF) {
4553 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
4554 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4555 eth_hw_addr_set(netdev, mac_addr);
4556 }
4557
4558 netdev->priv_flags |= IFF_UNICAST_FLT;
4559
4560 /* Setup netdev TC information */
4561 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
4562
4563 netdev->max_mtu = ICE_MAX_MTU;
4564
4565 return 0;
4566}
4567
4568static void ice_decfg_netdev(struct ice_vsi *vsi)
4569{
4570 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4571 free_netdev(vsi->netdev);
4572 vsi->netdev = NULL;
4573}
4574
4575static int ice_start_eth(struct ice_vsi *vsi)
4576{
4577 int err;
4578
4579 err = ice_init_mac_fltr(vsi->back);
4580 if (err)
4581 return err;
4582
4583 err = ice_vsi_open(vsi);
4584 if (err)
4585 ice_fltr_remove_all(vsi);
4586
4587 return err;
4588}
4589
4590static void ice_stop_eth(struct ice_vsi *vsi)
4591{
4592 ice_fltr_remove_all(vsi);
4593 ice_vsi_close(vsi);
4594}
4595
4596static int ice_init_eth(struct ice_pf *pf)
4597{
4598 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4599 int err;
4600
4601 if (!vsi)
4602 return -EINVAL;
4603
4604 /* init channel list */
4605 INIT_LIST_HEAD(&vsi->ch_list);
4606
4607 err = ice_cfg_netdev(vsi);
4608 if (err)
4609 return err;
4610 /* Setup DCB netlink interface */
4611 ice_dcbnl_setup(vsi);
4612
4613 err = ice_init_mac_fltr(pf);
4614 if (err)
4615 goto err_init_mac_fltr;
4616
4617 err = ice_devlink_create_pf_port(pf);
4618 if (err)
4619 goto err_devlink_create_pf_port;
4620
4621 SET_NETDEV_DEVLINK_PORT(vsi->netdev, &pf->devlink_port);
4622
4623 err = ice_register_netdev(vsi);
4624 if (err)
4625 goto err_register_netdev;
4626
4627 err = ice_tc_indir_block_register(vsi);
4628 if (err)
4629 goto err_tc_indir_block_register;
4630
4631 ice_napi_add(vsi);
4632
4633 return 0;
4634
4635err_tc_indir_block_register:
4636 ice_unregister_netdev(vsi);
4637err_register_netdev:
4638 ice_devlink_destroy_pf_port(pf);
4639err_devlink_create_pf_port:
4640err_init_mac_fltr:
4641 ice_decfg_netdev(vsi);
4642 return err;
4643}
4644
4645static void ice_deinit_eth(struct ice_pf *pf)
4646{
4647 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4648
4649 if (!vsi)
4650 return;
4651
4652 ice_vsi_close(vsi);
4653 ice_unregister_netdev(vsi);
4654 ice_devlink_destroy_pf_port(pf);
4655 ice_tc_indir_block_unregister(vsi);
4656 ice_decfg_netdev(vsi);
4657}
4658
4659/**
4660 * ice_wait_for_fw - wait for full FW readiness
4661 * @hw: pointer to the hardware structure
4662 * @timeout: milliseconds that can elapse before timing out
4663 */
4664static int ice_wait_for_fw(struct ice_hw *hw, u32 timeout)
4665{
4666 int fw_loading;
4667 u32 elapsed = 0;
4668
4669 while (elapsed <= timeout) {
4670 fw_loading = rd32(hw, GL_MNG_FWSM) & GL_MNG_FWSM_FW_LOADING_M;
4671
4672 /* firmware was not yet loaded, we have to wait more */
4673 if (fw_loading) {
4674 elapsed += 100;
4675 msleep(100);
4676 continue;
4677 }
4678 return 0;
4679 }
4680
4681 return -ETIMEDOUT;
4682}
4683
4684static int ice_init_dev(struct ice_pf *pf)
4685{
4686 struct device *dev = ice_pf_to_dev(pf);
4687 struct ice_hw *hw = &pf->hw;
4688 int err;
4689
4690 err = ice_init_hw(hw);
4691 if (err) {
4692 dev_err(dev, "ice_init_hw failed: %d\n", err);
4693 return err;
4694 }
4695
4696 /* Some cards require longer initialization times
4697 * due to necessity of loading FW from an external source.
4698 * This can take even half a minute.
4699 */
4700 if (ice_is_pf_c827(hw)) {
4701 err = ice_wait_for_fw(hw, 30000);
4702 if (err) {
4703 dev_err(dev, "ice_wait_for_fw timed out");
4704 return err;
4705 }
4706 }
4707
4708 ice_init_feature_support(pf);
4709
4710 ice_request_fw(pf);
4711
4712 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4713 * set in pf->state, which will cause ice_is_safe_mode to return
4714 * true
4715 */
4716 if (ice_is_safe_mode(pf)) {
4717 /* we already got function/device capabilities but these don't
4718 * reflect what the driver needs to do in safe mode. Instead of
4719 * adding conditional logic everywhere to ignore these
4720 * device/function capabilities, override them.
4721 */
4722 ice_set_safe_mode_caps(hw);
4723 }
4724
4725 err = ice_init_pf(pf);
4726 if (err) {
4727 dev_err(dev, "ice_init_pf failed: %d\n", err);
4728 goto err_init_pf;
4729 }
4730
4731 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4732 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4733 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4734 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4735 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4736 pf->hw.udp_tunnel_nic.tables[0].n_entries =
4737 pf->hw.tnl.valid_count[TNL_VXLAN];
4738 pf->hw.udp_tunnel_nic.tables[0].tunnel_types =
4739 UDP_TUNNEL_TYPE_VXLAN;
4740 }
4741 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4742 pf->hw.udp_tunnel_nic.tables[1].n_entries =
4743 pf->hw.tnl.valid_count[TNL_GENEVE];
4744 pf->hw.udp_tunnel_nic.tables[1].tunnel_types =
4745 UDP_TUNNEL_TYPE_GENEVE;
4746 }
4747
4748 err = ice_init_interrupt_scheme(pf);
4749 if (err) {
4750 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4751 err = -EIO;
4752 goto err_init_interrupt_scheme;
4753 }
4754
4755 /* In case of MSIX we are going to setup the misc vector right here
4756 * to handle admin queue events etc. In case of legacy and MSI
4757 * the misc functionality and queue processing is combined in
4758 * the same vector and that gets setup at open.
4759 */
4760 err = ice_req_irq_msix_misc(pf);
4761 if (err) {
4762 dev_err(dev, "setup of misc vector failed: %d\n", err);
4763 goto err_req_irq_msix_misc;
4764 }
4765
4766 return 0;
4767
4768err_req_irq_msix_misc:
4769 ice_clear_interrupt_scheme(pf);
4770err_init_interrupt_scheme:
4771 ice_deinit_pf(pf);
4772err_init_pf:
4773 ice_deinit_hw(hw);
4774 return err;
4775}
4776
4777static void ice_deinit_dev(struct ice_pf *pf)
4778{
4779 ice_free_irq_msix_misc(pf);
4780 ice_deinit_pf(pf);
4781 ice_deinit_hw(&pf->hw);
4782
4783 /* Service task is already stopped, so call reset directly. */
4784 ice_reset(&pf->hw, ICE_RESET_PFR);
4785 pci_wait_for_pending_transaction(pf->pdev);
4786 ice_clear_interrupt_scheme(pf);
4787}
4788
4789static void ice_init_features(struct ice_pf *pf)
4790{
4791 struct device *dev = ice_pf_to_dev(pf);
4792
4793 if (ice_is_safe_mode(pf))
4794 return;
4795
4796 /* initialize DDP driven features */
4797 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4798 ice_ptp_init(pf);
4799
4800 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4801 ice_gnss_init(pf);
4802
4803 if (ice_is_feature_supported(pf, ICE_F_CGU) ||
4804 ice_is_feature_supported(pf, ICE_F_PHY_RCLK))
4805 ice_dpll_init(pf);
4806
4807 /* Note: Flow director init failure is non-fatal to load */
4808 if (ice_init_fdir(pf))
4809 dev_err(dev, "could not initialize flow director\n");
4810
4811 /* Note: DCB init failure is non-fatal to load */
4812 if (ice_init_pf_dcb(pf, false)) {
4813 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4814 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4815 } else {
4816 ice_cfg_lldp_mib_change(&pf->hw, true);
4817 }
4818
4819 if (ice_init_lag(pf))
4820 dev_warn(dev, "Failed to init link aggregation support\n");
4821
4822 ice_hwmon_init(pf);
4823}
4824
4825static void ice_deinit_features(struct ice_pf *pf)
4826{
4827 if (ice_is_safe_mode(pf))
4828 return;
4829
4830 ice_deinit_lag(pf);
4831 if (test_bit(ICE_FLAG_DCB_CAPABLE, pf->flags))
4832 ice_cfg_lldp_mib_change(&pf->hw, false);
4833 ice_deinit_fdir(pf);
4834 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4835 ice_gnss_exit(pf);
4836 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4837 ice_ptp_release(pf);
4838 if (test_bit(ICE_FLAG_DPLL, pf->flags))
4839 ice_dpll_deinit(pf);
4840 if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_SWITCHDEV)
4841 xa_destroy(&pf->eswitch.reprs);
4842}
4843
4844static void ice_init_wakeup(struct ice_pf *pf)
4845{
4846 /* Save wakeup reason register for later use */
4847 pf->wakeup_reason = rd32(&pf->hw, PFPM_WUS);
4848
4849 /* check for a power management event */
4850 ice_print_wake_reason(pf);
4851
4852 /* clear wake status, all bits */
4853 wr32(&pf->hw, PFPM_WUS, U32_MAX);
4854
4855 /* Disable WoL at init, wait for user to enable */
4856 device_set_wakeup_enable(ice_pf_to_dev(pf), false);
4857}
4858
4859static int ice_init_link(struct ice_pf *pf)
4860{
4861 struct device *dev = ice_pf_to_dev(pf);
4862 int err;
4863
4864 err = ice_init_link_events(pf->hw.port_info);
4865 if (err) {
4866 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4867 return err;
4868 }
4869
4870 /* not a fatal error if this fails */
4871 err = ice_init_nvm_phy_type(pf->hw.port_info);
4872 if (err)
4873 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4874
4875 /* not a fatal error if this fails */
4876 err = ice_update_link_info(pf->hw.port_info);
4877 if (err)
4878 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4879
4880 ice_init_link_dflt_override(pf->hw.port_info);
4881
4882 ice_check_link_cfg_err(pf,
4883 pf->hw.port_info->phy.link_info.link_cfg_err);
4884
4885 /* if media available, initialize PHY settings */
4886 if (pf->hw.port_info->phy.link_info.link_info &
4887 ICE_AQ_MEDIA_AVAILABLE) {
4888 /* not a fatal error if this fails */
4889 err = ice_init_phy_user_cfg(pf->hw.port_info);
4890 if (err)
4891 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4892
4893 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4894 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4895
4896 if (vsi)
4897 ice_configure_phy(vsi);
4898 }
4899 } else {
4900 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4901 }
4902
4903 return err;
4904}
4905
4906static int ice_init_pf_sw(struct ice_pf *pf)
4907{
4908 bool dvm = ice_is_dvm_ena(&pf->hw);
4909 struct ice_vsi *vsi;
4910 int err;
4911
4912 /* create switch struct for the switch element created by FW on boot */
4913 pf->first_sw = kzalloc(sizeof(*pf->first_sw), GFP_KERNEL);
4914 if (!pf->first_sw)
4915 return -ENOMEM;
4916
4917 if (pf->hw.evb_veb)
4918 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4919 else
4920 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4921
4922 pf->first_sw->pf = pf;
4923
4924 /* record the sw_id available for later use */
4925 pf->first_sw->sw_id = pf->hw.port_info->sw_id;
4926
4927 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
4928 if (err)
4929 goto err_aq_set_port_params;
4930
4931 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
4932 if (!vsi) {
4933 err = -ENOMEM;
4934 goto err_pf_vsi_setup;
4935 }
4936
4937 return 0;
4938
4939err_pf_vsi_setup:
4940err_aq_set_port_params:
4941 kfree(pf->first_sw);
4942 return err;
4943}
4944
4945static void ice_deinit_pf_sw(struct ice_pf *pf)
4946{
4947 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4948
4949 if (!vsi)
4950 return;
4951
4952 ice_vsi_release(vsi);
4953 kfree(pf->first_sw);
4954}
4955
4956static int ice_alloc_vsis(struct ice_pf *pf)
4957{
4958 struct device *dev = ice_pf_to_dev(pf);
4959
4960 pf->num_alloc_vsi = pf->hw.func_caps.guar_num_vsi;
4961 if (!pf->num_alloc_vsi)
4962 return -EIO;
4963
4964 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4965 dev_warn(dev,
4966 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4967 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4968 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4969 }
4970
4971 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4972 GFP_KERNEL);
4973 if (!pf->vsi)
4974 return -ENOMEM;
4975
4976 pf->vsi_stats = devm_kcalloc(dev, pf->num_alloc_vsi,
4977 sizeof(*pf->vsi_stats), GFP_KERNEL);
4978 if (!pf->vsi_stats) {
4979 devm_kfree(dev, pf->vsi);
4980 return -ENOMEM;
4981 }
4982
4983 return 0;
4984}
4985
4986static void ice_dealloc_vsis(struct ice_pf *pf)
4987{
4988 devm_kfree(ice_pf_to_dev(pf), pf->vsi_stats);
4989 pf->vsi_stats = NULL;
4990
4991 pf->num_alloc_vsi = 0;
4992 devm_kfree(ice_pf_to_dev(pf), pf->vsi);
4993 pf->vsi = NULL;
4994}
4995
4996static int ice_init_devlink(struct ice_pf *pf)
4997{
4998 int err;
4999
5000 err = ice_devlink_register_params(pf);
5001 if (err)
5002 return err;
5003
5004 ice_devlink_init_regions(pf);
5005 ice_devlink_register(pf);
5006
5007 return 0;
5008}
5009
5010static void ice_deinit_devlink(struct ice_pf *pf)
5011{
5012 ice_devlink_unregister(pf);
5013 ice_devlink_destroy_regions(pf);
5014 ice_devlink_unregister_params(pf);
5015}
5016
5017static int ice_init(struct ice_pf *pf)
5018{
5019 int err;
5020
5021 err = ice_init_dev(pf);
5022 if (err)
5023 return err;
5024
5025 err = ice_alloc_vsis(pf);
5026 if (err)
5027 goto err_alloc_vsis;
5028
5029 err = ice_init_pf_sw(pf);
5030 if (err)
5031 goto err_init_pf_sw;
5032
5033 ice_init_wakeup(pf);
5034
5035 err = ice_init_link(pf);
5036 if (err)
5037 goto err_init_link;
5038
5039 err = ice_send_version(pf);
5040 if (err)
5041 goto err_init_link;
5042
5043 ice_verify_cacheline_size(pf);
5044
5045 if (ice_is_safe_mode(pf))
5046 ice_set_safe_mode_vlan_cfg(pf);
5047 else
5048 /* print PCI link speed and width */
5049 pcie_print_link_status(pf->pdev);
5050
5051 /* ready to go, so clear down state bit */
5052 clear_bit(ICE_DOWN, pf->state);
5053 clear_bit(ICE_SERVICE_DIS, pf->state);
5054
5055 /* since everything is good, start the service timer */
5056 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5057
5058 return 0;
5059
5060err_init_link:
5061 ice_deinit_pf_sw(pf);
5062err_init_pf_sw:
5063 ice_dealloc_vsis(pf);
5064err_alloc_vsis:
5065 ice_deinit_dev(pf);
5066 return err;
5067}
5068
5069static void ice_deinit(struct ice_pf *pf)
5070{
5071 set_bit(ICE_SERVICE_DIS, pf->state);
5072 set_bit(ICE_DOWN, pf->state);
5073
5074 ice_deinit_pf_sw(pf);
5075 ice_dealloc_vsis(pf);
5076 ice_deinit_dev(pf);
5077}
5078
5079/**
5080 * ice_load - load pf by init hw and starting VSI
5081 * @pf: pointer to the pf instance
5082 */
5083int ice_load(struct ice_pf *pf)
5084{
5085 struct ice_vsi_cfg_params params = {};
5086 struct ice_vsi *vsi;
5087 int err;
5088
5089 err = ice_init_dev(pf);
5090 if (err)
5091 return err;
5092
5093 vsi = ice_get_main_vsi(pf);
5094
5095 params = ice_vsi_to_params(vsi);
5096 params.flags = ICE_VSI_FLAG_INIT;
5097
5098 rtnl_lock();
5099 err = ice_vsi_cfg(vsi, ¶ms);
5100 if (err)
5101 goto err_vsi_cfg;
5102
5103 err = ice_start_eth(ice_get_main_vsi(pf));
5104 if (err)
5105 goto err_start_eth;
5106 rtnl_unlock();
5107
5108 err = ice_init_rdma(pf);
5109 if (err)
5110 goto err_init_rdma;
5111
5112 ice_init_features(pf);
5113 ice_service_task_restart(pf);
5114
5115 clear_bit(ICE_DOWN, pf->state);
5116
5117 return 0;
5118
5119err_init_rdma:
5120 ice_vsi_close(ice_get_main_vsi(pf));
5121 rtnl_lock();
5122err_start_eth:
5123 ice_vsi_decfg(ice_get_main_vsi(pf));
5124err_vsi_cfg:
5125 rtnl_unlock();
5126 ice_deinit_dev(pf);
5127 return err;
5128}
5129
5130/**
5131 * ice_unload - unload pf by stopping VSI and deinit hw
5132 * @pf: pointer to the pf instance
5133 */
5134void ice_unload(struct ice_pf *pf)
5135{
5136 ice_deinit_features(pf);
5137 ice_deinit_rdma(pf);
5138 rtnl_lock();
5139 ice_stop_eth(ice_get_main_vsi(pf));
5140 ice_vsi_decfg(ice_get_main_vsi(pf));
5141 rtnl_unlock();
5142 ice_deinit_dev(pf);
5143}
5144
5145/**
5146 * ice_probe - Device initialization routine
5147 * @pdev: PCI device information struct
5148 * @ent: entry in ice_pci_tbl
5149 *
5150 * Returns 0 on success, negative on failure
5151 */
5152static int
5153ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
5154{
5155 struct device *dev = &pdev->dev;
5156 struct ice_pf *pf;
5157 struct ice_hw *hw;
5158 int err;
5159
5160 if (pdev->is_virtfn) {
5161 dev_err(dev, "can't probe a virtual function\n");
5162 return -EINVAL;
5163 }
5164
5165 /* when under a kdump kernel initiate a reset before enabling the
5166 * device in order to clear out any pending DMA transactions. These
5167 * transactions can cause some systems to machine check when doing
5168 * the pcim_enable_device() below.
5169 */
5170 if (is_kdump_kernel()) {
5171 pci_save_state(pdev);
5172 pci_clear_master(pdev);
5173 err = pcie_flr(pdev);
5174 if (err)
5175 return err;
5176 pci_restore_state(pdev);
5177 }
5178
5179 /* this driver uses devres, see
5180 * Documentation/driver-api/driver-model/devres.rst
5181 */
5182 err = pcim_enable_device(pdev);
5183 if (err)
5184 return err;
5185
5186 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
5187 if (err) {
5188 dev_err(dev, "BAR0 I/O map error %d\n", err);
5189 return err;
5190 }
5191
5192 pf = ice_allocate_pf(dev);
5193 if (!pf)
5194 return -ENOMEM;
5195
5196 /* initialize Auxiliary index to invalid value */
5197 pf->aux_idx = -1;
5198
5199 /* set up for high or low DMA */
5200 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
5201 if (err) {
5202 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
5203 return err;
5204 }
5205
5206 pci_set_master(pdev);
5207
5208 pf->pdev = pdev;
5209 pci_set_drvdata(pdev, pf);
5210 set_bit(ICE_DOWN, pf->state);
5211 /* Disable service task until DOWN bit is cleared */
5212 set_bit(ICE_SERVICE_DIS, pf->state);
5213
5214 hw = &pf->hw;
5215 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
5216 pci_save_state(pdev);
5217
5218 hw->back = pf;
5219 hw->port_info = NULL;
5220 hw->vendor_id = pdev->vendor;
5221 hw->device_id = pdev->device;
5222 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
5223 hw->subsystem_vendor_id = pdev->subsystem_vendor;
5224 hw->subsystem_device_id = pdev->subsystem_device;
5225 hw->bus.device = PCI_SLOT(pdev->devfn);
5226 hw->bus.func = PCI_FUNC(pdev->devfn);
5227 ice_set_ctrlq_len(hw);
5228
5229 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
5230
5231#ifndef CONFIG_DYNAMIC_DEBUG
5232 if (debug < -1)
5233 hw->debug_mask = debug;
5234#endif
5235
5236 err = ice_init(pf);
5237 if (err)
5238 goto err_init;
5239
5240 err = ice_init_eth(pf);
5241 if (err)
5242 goto err_init_eth;
5243
5244 err = ice_init_rdma(pf);
5245 if (err)
5246 goto err_init_rdma;
5247
5248 err = ice_init_devlink(pf);
5249 if (err)
5250 goto err_init_devlink;
5251
5252 ice_init_features(pf);
5253
5254 return 0;
5255
5256err_init_devlink:
5257 ice_deinit_rdma(pf);
5258err_init_rdma:
5259 ice_deinit_eth(pf);
5260err_init_eth:
5261 ice_deinit(pf);
5262err_init:
5263 pci_disable_device(pdev);
5264 return err;
5265}
5266
5267/**
5268 * ice_set_wake - enable or disable Wake on LAN
5269 * @pf: pointer to the PF struct
5270 *
5271 * Simple helper for WoL control
5272 */
5273static void ice_set_wake(struct ice_pf *pf)
5274{
5275 struct ice_hw *hw = &pf->hw;
5276 bool wol = pf->wol_ena;
5277
5278 /* clear wake state, otherwise new wake events won't fire */
5279 wr32(hw, PFPM_WUS, U32_MAX);
5280
5281 /* enable / disable APM wake up, no RMW needed */
5282 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
5283
5284 /* set magic packet filter enabled */
5285 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
5286}
5287
5288/**
5289 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
5290 * @pf: pointer to the PF struct
5291 *
5292 * Issue firmware command to enable multicast magic wake, making
5293 * sure that any locally administered address (LAA) is used for
5294 * wake, and that PF reset doesn't undo the LAA.
5295 */
5296static void ice_setup_mc_magic_wake(struct ice_pf *pf)
5297{
5298 struct device *dev = ice_pf_to_dev(pf);
5299 struct ice_hw *hw = &pf->hw;
5300 u8 mac_addr[ETH_ALEN];
5301 struct ice_vsi *vsi;
5302 int status;
5303 u8 flags;
5304
5305 if (!pf->wol_ena)
5306 return;
5307
5308 vsi = ice_get_main_vsi(pf);
5309 if (!vsi)
5310 return;
5311
5312 /* Get current MAC address in case it's an LAA */
5313 if (vsi->netdev)
5314 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
5315 else
5316 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
5317
5318 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
5319 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
5320 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
5321
5322 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
5323 if (status)
5324 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
5325 status, ice_aq_str(hw->adminq.sq_last_status));
5326}
5327
5328/**
5329 * ice_remove - Device removal routine
5330 * @pdev: PCI device information struct
5331 */
5332static void ice_remove(struct pci_dev *pdev)
5333{
5334 struct ice_pf *pf = pci_get_drvdata(pdev);
5335 int i;
5336
5337 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
5338 if (!ice_is_reset_in_progress(pf->state))
5339 break;
5340 msleep(100);
5341 }
5342
5343 ice_debugfs_exit();
5344
5345 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
5346 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5347 ice_free_vfs(pf);
5348 }
5349
5350 ice_hwmon_exit(pf);
5351
5352 ice_service_task_stop(pf);
5353 ice_aq_cancel_waiting_tasks(pf);
5354 set_bit(ICE_DOWN, pf->state);
5355
5356 if (!ice_is_safe_mode(pf))
5357 ice_remove_arfs(pf);
5358 ice_deinit_features(pf);
5359 ice_deinit_devlink(pf);
5360 ice_deinit_rdma(pf);
5361 ice_deinit_eth(pf);
5362 ice_deinit(pf);
5363
5364 ice_vsi_release_all(pf);
5365
5366 ice_setup_mc_magic_wake(pf);
5367 ice_set_wake(pf);
5368
5369 pci_disable_device(pdev);
5370}
5371
5372/**
5373 * ice_shutdown - PCI callback for shutting down device
5374 * @pdev: PCI device information struct
5375 */
5376static void ice_shutdown(struct pci_dev *pdev)
5377{
5378 struct ice_pf *pf = pci_get_drvdata(pdev);
5379
5380 ice_remove(pdev);
5381
5382 if (system_state == SYSTEM_POWER_OFF) {
5383 pci_wake_from_d3(pdev, pf->wol_ena);
5384 pci_set_power_state(pdev, PCI_D3hot);
5385 }
5386}
5387
5388#ifdef CONFIG_PM
5389/**
5390 * ice_prepare_for_shutdown - prep for PCI shutdown
5391 * @pf: board private structure
5392 *
5393 * Inform or close all dependent features in prep for PCI device shutdown
5394 */
5395static void ice_prepare_for_shutdown(struct ice_pf *pf)
5396{
5397 struct ice_hw *hw = &pf->hw;
5398 u32 v;
5399
5400 /* Notify VFs of impending reset */
5401 if (ice_check_sq_alive(hw, &hw->mailboxq))
5402 ice_vc_notify_reset(pf);
5403
5404 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5405
5406 /* disable the VSIs and their queues that are not already DOWN */
5407 ice_pf_dis_all_vsi(pf, false);
5408
5409 ice_for_each_vsi(pf, v)
5410 if (pf->vsi[v])
5411 pf->vsi[v]->vsi_num = 0;
5412
5413 ice_shutdown_all_ctrlq(hw);
5414}
5415
5416/**
5417 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5418 * @pf: board private structure to reinitialize
5419 *
5420 * This routine reinitialize interrupt scheme that was cleared during
5421 * power management suspend callback.
5422 *
5423 * This should be called during resume routine to re-allocate the q_vectors
5424 * and reacquire interrupts.
5425 */
5426static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5427{
5428 struct device *dev = ice_pf_to_dev(pf);
5429 int ret, v;
5430
5431 /* Since we clear MSIX flag during suspend, we need to
5432 * set it back during resume...
5433 */
5434
5435 ret = ice_init_interrupt_scheme(pf);
5436 if (ret) {
5437 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5438 return ret;
5439 }
5440
5441 /* Remap vectors and rings, after successful re-init interrupts */
5442 ice_for_each_vsi(pf, v) {
5443 if (!pf->vsi[v])
5444 continue;
5445
5446 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5447 if (ret)
5448 goto err_reinit;
5449 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5450 ice_vsi_set_napi_queues(pf->vsi[v]);
5451 }
5452
5453 ret = ice_req_irq_msix_misc(pf);
5454 if (ret) {
5455 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5456 ret);
5457 goto err_reinit;
5458 }
5459
5460 return 0;
5461
5462err_reinit:
5463 while (v--)
5464 if (pf->vsi[v])
5465 ice_vsi_free_q_vectors(pf->vsi[v]);
5466
5467 return ret;
5468}
5469
5470/**
5471 * ice_suspend
5472 * @dev: generic device information structure
5473 *
5474 * Power Management callback to quiesce the device and prepare
5475 * for D3 transition.
5476 */
5477static int __maybe_unused ice_suspend(struct device *dev)
5478{
5479 struct pci_dev *pdev = to_pci_dev(dev);
5480 struct ice_pf *pf;
5481 int disabled, v;
5482
5483 pf = pci_get_drvdata(pdev);
5484
5485 if (!ice_pf_state_is_nominal(pf)) {
5486 dev_err(dev, "Device is not ready, no need to suspend it\n");
5487 return -EBUSY;
5488 }
5489
5490 /* Stop watchdog tasks until resume completion.
5491 * Even though it is most likely that the service task is
5492 * disabled if the device is suspended or down, the service task's
5493 * state is controlled by a different state bit, and we should
5494 * store and honor whatever state that bit is in at this point.
5495 */
5496 disabled = ice_service_task_stop(pf);
5497
5498 ice_unplug_aux_dev(pf);
5499
5500 /* Already suspended?, then there is nothing to do */
5501 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5502 if (!disabled)
5503 ice_service_task_restart(pf);
5504 return 0;
5505 }
5506
5507 if (test_bit(ICE_DOWN, pf->state) ||
5508 ice_is_reset_in_progress(pf->state)) {
5509 dev_err(dev, "can't suspend device in reset or already down\n");
5510 if (!disabled)
5511 ice_service_task_restart(pf);
5512 return 0;
5513 }
5514
5515 ice_setup_mc_magic_wake(pf);
5516
5517 ice_prepare_for_shutdown(pf);
5518
5519 ice_set_wake(pf);
5520
5521 /* Free vectors, clear the interrupt scheme and release IRQs
5522 * for proper hibernation, especially with large number of CPUs.
5523 * Otherwise hibernation might fail when mapping all the vectors back
5524 * to CPU0.
5525 */
5526 ice_free_irq_msix_misc(pf);
5527 ice_for_each_vsi(pf, v) {
5528 if (!pf->vsi[v])
5529 continue;
5530 ice_vsi_free_q_vectors(pf->vsi[v]);
5531 }
5532 ice_clear_interrupt_scheme(pf);
5533
5534 pci_save_state(pdev);
5535 pci_wake_from_d3(pdev, pf->wol_ena);
5536 pci_set_power_state(pdev, PCI_D3hot);
5537 return 0;
5538}
5539
5540/**
5541 * ice_resume - PM callback for waking up from D3
5542 * @dev: generic device information structure
5543 */
5544static int __maybe_unused ice_resume(struct device *dev)
5545{
5546 struct pci_dev *pdev = to_pci_dev(dev);
5547 enum ice_reset_req reset_type;
5548 struct ice_pf *pf;
5549 struct ice_hw *hw;
5550 int ret;
5551
5552 pci_set_power_state(pdev, PCI_D0);
5553 pci_restore_state(pdev);
5554 pci_save_state(pdev);
5555
5556 if (!pci_device_is_present(pdev))
5557 return -ENODEV;
5558
5559 ret = pci_enable_device_mem(pdev);
5560 if (ret) {
5561 dev_err(dev, "Cannot enable device after suspend\n");
5562 return ret;
5563 }
5564
5565 pf = pci_get_drvdata(pdev);
5566 hw = &pf->hw;
5567
5568 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5569 ice_print_wake_reason(pf);
5570
5571 /* We cleared the interrupt scheme when we suspended, so we need to
5572 * restore it now to resume device functionality.
5573 */
5574 ret = ice_reinit_interrupt_scheme(pf);
5575 if (ret)
5576 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5577
5578 clear_bit(ICE_DOWN, pf->state);
5579 /* Now perform PF reset and rebuild */
5580 reset_type = ICE_RESET_PFR;
5581 /* re-enable service task for reset, but allow reset to schedule it */
5582 clear_bit(ICE_SERVICE_DIS, pf->state);
5583
5584 if (ice_schedule_reset(pf, reset_type))
5585 dev_err(dev, "Reset during resume failed.\n");
5586
5587 clear_bit(ICE_SUSPENDED, pf->state);
5588 ice_service_task_restart(pf);
5589
5590 /* Restart the service task */
5591 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5592
5593 return 0;
5594}
5595#endif /* CONFIG_PM */
5596
5597/**
5598 * ice_pci_err_detected - warning that PCI error has been detected
5599 * @pdev: PCI device information struct
5600 * @err: the type of PCI error
5601 *
5602 * Called to warn that something happened on the PCI bus and the error handling
5603 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5604 */
5605static pci_ers_result_t
5606ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5607{
5608 struct ice_pf *pf = pci_get_drvdata(pdev);
5609
5610 if (!pf) {
5611 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5612 __func__, err);
5613 return PCI_ERS_RESULT_DISCONNECT;
5614 }
5615
5616 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5617 ice_service_task_stop(pf);
5618
5619 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5620 set_bit(ICE_PFR_REQ, pf->state);
5621 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5622 }
5623 }
5624
5625 return PCI_ERS_RESULT_NEED_RESET;
5626}
5627
5628/**
5629 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5630 * @pdev: PCI device information struct
5631 *
5632 * Called to determine if the driver can recover from the PCI slot reset by
5633 * using a register read to determine if the device is recoverable.
5634 */
5635static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5636{
5637 struct ice_pf *pf = pci_get_drvdata(pdev);
5638 pci_ers_result_t result;
5639 int err;
5640 u32 reg;
5641
5642 err = pci_enable_device_mem(pdev);
5643 if (err) {
5644 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5645 err);
5646 result = PCI_ERS_RESULT_DISCONNECT;
5647 } else {
5648 pci_set_master(pdev);
5649 pci_restore_state(pdev);
5650 pci_save_state(pdev);
5651 pci_wake_from_d3(pdev, false);
5652
5653 /* Check for life */
5654 reg = rd32(&pf->hw, GLGEN_RTRIG);
5655 if (!reg)
5656 result = PCI_ERS_RESULT_RECOVERED;
5657 else
5658 result = PCI_ERS_RESULT_DISCONNECT;
5659 }
5660
5661 return result;
5662}
5663
5664/**
5665 * ice_pci_err_resume - restart operations after PCI error recovery
5666 * @pdev: PCI device information struct
5667 *
5668 * Called to allow the driver to bring things back up after PCI error and/or
5669 * reset recovery have finished
5670 */
5671static void ice_pci_err_resume(struct pci_dev *pdev)
5672{
5673 struct ice_pf *pf = pci_get_drvdata(pdev);
5674
5675 if (!pf) {
5676 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5677 __func__);
5678 return;
5679 }
5680
5681 if (test_bit(ICE_SUSPENDED, pf->state)) {
5682 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5683 __func__);
5684 return;
5685 }
5686
5687 ice_restore_all_vfs_msi_state(pf);
5688
5689 ice_do_reset(pf, ICE_RESET_PFR);
5690 ice_service_task_restart(pf);
5691 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5692}
5693
5694/**
5695 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5696 * @pdev: PCI device information struct
5697 */
5698static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5699{
5700 struct ice_pf *pf = pci_get_drvdata(pdev);
5701
5702 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5703 ice_service_task_stop(pf);
5704
5705 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5706 set_bit(ICE_PFR_REQ, pf->state);
5707 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5708 }
5709 }
5710}
5711
5712/**
5713 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5714 * @pdev: PCI device information struct
5715 */
5716static void ice_pci_err_reset_done(struct pci_dev *pdev)
5717{
5718 ice_pci_err_resume(pdev);
5719}
5720
5721/* ice_pci_tbl - PCI Device ID Table
5722 *
5723 * Wildcard entries (PCI_ANY_ID) should come last
5724 * Last entry must be all 0s
5725 *
5726 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5727 * Class, Class Mask, private data (not used) }
5728 */
5729static const struct pci_device_id ice_pci_tbl[] = {
5730 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE) },
5731 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP) },
5732 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP) },
5733 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE) },
5734 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP) },
5735 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP) },
5736 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE) },
5737 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP) },
5738 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP) },
5739 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T) },
5740 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII) },
5741 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE) },
5742 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP) },
5743 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP) },
5744 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T) },
5745 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII) },
5746 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE) },
5747 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP) },
5748 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T) },
5749 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII) },
5750 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE) },
5751 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP) },
5752 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T) },
5753 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE) },
5754 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP) },
5755 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT) },
5756 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_BACKPLANE) },
5757 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_QSFP56) },
5758 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP) },
5759 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP_DD) },
5760 /* required last entry */
5761 {}
5762};
5763MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5764
5765static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5766
5767static const struct pci_error_handlers ice_pci_err_handler = {
5768 .error_detected = ice_pci_err_detected,
5769 .slot_reset = ice_pci_err_slot_reset,
5770 .reset_prepare = ice_pci_err_reset_prepare,
5771 .reset_done = ice_pci_err_reset_done,
5772 .resume = ice_pci_err_resume
5773};
5774
5775static struct pci_driver ice_driver = {
5776 .name = KBUILD_MODNAME,
5777 .id_table = ice_pci_tbl,
5778 .probe = ice_probe,
5779 .remove = ice_remove,
5780#ifdef CONFIG_PM
5781 .driver.pm = &ice_pm_ops,
5782#endif /* CONFIG_PM */
5783 .shutdown = ice_shutdown,
5784 .sriov_configure = ice_sriov_configure,
5785 .sriov_get_vf_total_msix = ice_sriov_get_vf_total_msix,
5786 .sriov_set_msix_vec_count = ice_sriov_set_msix_vec_count,
5787 .err_handler = &ice_pci_err_handler
5788};
5789
5790/**
5791 * ice_module_init - Driver registration routine
5792 *
5793 * ice_module_init is the first routine called when the driver is
5794 * loaded. All it does is register with the PCI subsystem.
5795 */
5796static int __init ice_module_init(void)
5797{
5798 int status = -ENOMEM;
5799
5800 pr_info("%s\n", ice_driver_string);
5801 pr_info("%s\n", ice_copyright);
5802
5803 ice_adv_lnk_speed_maps_init();
5804
5805 ice_wq = alloc_workqueue("%s", 0, 0, KBUILD_MODNAME);
5806 if (!ice_wq) {
5807 pr_err("Failed to create workqueue\n");
5808 return status;
5809 }
5810
5811 ice_lag_wq = alloc_ordered_workqueue("ice_lag_wq", 0);
5812 if (!ice_lag_wq) {
5813 pr_err("Failed to create LAG workqueue\n");
5814 goto err_dest_wq;
5815 }
5816
5817 ice_debugfs_init();
5818
5819 status = pci_register_driver(&ice_driver);
5820 if (status) {
5821 pr_err("failed to register PCI driver, err %d\n", status);
5822 goto err_dest_lag_wq;
5823 }
5824
5825 return 0;
5826
5827err_dest_lag_wq:
5828 destroy_workqueue(ice_lag_wq);
5829 ice_debugfs_exit();
5830err_dest_wq:
5831 destroy_workqueue(ice_wq);
5832 return status;
5833}
5834module_init(ice_module_init);
5835
5836/**
5837 * ice_module_exit - Driver exit cleanup routine
5838 *
5839 * ice_module_exit is called just before the driver is removed
5840 * from memory.
5841 */
5842static void __exit ice_module_exit(void)
5843{
5844 pci_unregister_driver(&ice_driver);
5845 destroy_workqueue(ice_wq);
5846 destroy_workqueue(ice_lag_wq);
5847 pr_info("module unloaded\n");
5848}
5849module_exit(ice_module_exit);
5850
5851/**
5852 * ice_set_mac_address - NDO callback to set MAC address
5853 * @netdev: network interface device structure
5854 * @pi: pointer to an address structure
5855 *
5856 * Returns 0 on success, negative on failure
5857 */
5858static int ice_set_mac_address(struct net_device *netdev, void *pi)
5859{
5860 struct ice_netdev_priv *np = netdev_priv(netdev);
5861 struct ice_vsi *vsi = np->vsi;
5862 struct ice_pf *pf = vsi->back;
5863 struct ice_hw *hw = &pf->hw;
5864 struct sockaddr *addr = pi;
5865 u8 old_mac[ETH_ALEN];
5866 u8 flags = 0;
5867 u8 *mac;
5868 int err;
5869
5870 mac = (u8 *)addr->sa_data;
5871
5872 if (!is_valid_ether_addr(mac))
5873 return -EADDRNOTAVAIL;
5874
5875 if (test_bit(ICE_DOWN, pf->state) ||
5876 ice_is_reset_in_progress(pf->state)) {
5877 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5878 mac);
5879 return -EBUSY;
5880 }
5881
5882 if (ice_chnl_dmac_fltr_cnt(pf)) {
5883 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5884 mac);
5885 return -EAGAIN;
5886 }
5887
5888 netif_addr_lock_bh(netdev);
5889 ether_addr_copy(old_mac, netdev->dev_addr);
5890 /* change the netdev's MAC address */
5891 eth_hw_addr_set(netdev, mac);
5892 netif_addr_unlock_bh(netdev);
5893
5894 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5895 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5896 if (err && err != -ENOENT) {
5897 err = -EADDRNOTAVAIL;
5898 goto err_update_filters;
5899 }
5900
5901 /* Add filter for new MAC. If filter exists, return success */
5902 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5903 if (err == -EEXIST) {
5904 /* Although this MAC filter is already present in hardware it's
5905 * possible in some cases (e.g. bonding) that dev_addr was
5906 * modified outside of the driver and needs to be restored back
5907 * to this value.
5908 */
5909 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5910
5911 return 0;
5912 } else if (err) {
5913 /* error if the new filter addition failed */
5914 err = -EADDRNOTAVAIL;
5915 }
5916
5917err_update_filters:
5918 if (err) {
5919 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5920 mac);
5921 netif_addr_lock_bh(netdev);
5922 eth_hw_addr_set(netdev, old_mac);
5923 netif_addr_unlock_bh(netdev);
5924 return err;
5925 }
5926
5927 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5928 netdev->dev_addr);
5929
5930 /* write new MAC address to the firmware */
5931 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5932 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5933 if (err) {
5934 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5935 mac, err);
5936 }
5937 return 0;
5938}
5939
5940/**
5941 * ice_set_rx_mode - NDO callback to set the netdev filters
5942 * @netdev: network interface device structure
5943 */
5944static void ice_set_rx_mode(struct net_device *netdev)
5945{
5946 struct ice_netdev_priv *np = netdev_priv(netdev);
5947 struct ice_vsi *vsi = np->vsi;
5948
5949 if (!vsi || ice_is_switchdev_running(vsi->back))
5950 return;
5951
5952 /* Set the flags to synchronize filters
5953 * ndo_set_rx_mode may be triggered even without a change in netdev
5954 * flags
5955 */
5956 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5957 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5958 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5959
5960 /* schedule our worker thread which will take care of
5961 * applying the new filter changes
5962 */
5963 ice_service_task_schedule(vsi->back);
5964}
5965
5966/**
5967 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5968 * @netdev: network interface device structure
5969 * @queue_index: Queue ID
5970 * @maxrate: maximum bandwidth in Mbps
5971 */
5972static int
5973ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5974{
5975 struct ice_netdev_priv *np = netdev_priv(netdev);
5976 struct ice_vsi *vsi = np->vsi;
5977 u16 q_handle;
5978 int status;
5979 u8 tc;
5980
5981 /* Validate maxrate requested is within permitted range */
5982 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5983 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5984 maxrate, queue_index);
5985 return -EINVAL;
5986 }
5987
5988 q_handle = vsi->tx_rings[queue_index]->q_handle;
5989 tc = ice_dcb_get_tc(vsi, queue_index);
5990
5991 vsi = ice_locate_vsi_using_queue(vsi, queue_index);
5992 if (!vsi) {
5993 netdev_err(netdev, "Invalid VSI for given queue %d\n",
5994 queue_index);
5995 return -EINVAL;
5996 }
5997
5998 /* Set BW back to default, when user set maxrate to 0 */
5999 if (!maxrate)
6000 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
6001 q_handle, ICE_MAX_BW);
6002 else
6003 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
6004 q_handle, ICE_MAX_BW, maxrate * 1000);
6005 if (status)
6006 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
6007 status);
6008
6009 return status;
6010}
6011
6012/**
6013 * ice_fdb_add - add an entry to the hardware database
6014 * @ndm: the input from the stack
6015 * @tb: pointer to array of nladdr (unused)
6016 * @dev: the net device pointer
6017 * @addr: the MAC address entry being added
6018 * @vid: VLAN ID
6019 * @flags: instructions from stack about fdb operation
6020 * @extack: netlink extended ack
6021 */
6022static int
6023ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
6024 struct net_device *dev, const unsigned char *addr, u16 vid,
6025 u16 flags, struct netlink_ext_ack __always_unused *extack)
6026{
6027 int err;
6028
6029 if (vid) {
6030 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
6031 return -EINVAL;
6032 }
6033 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
6034 netdev_err(dev, "FDB only supports static addresses\n");
6035 return -EINVAL;
6036 }
6037
6038 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
6039 err = dev_uc_add_excl(dev, addr);
6040 else if (is_multicast_ether_addr(addr))
6041 err = dev_mc_add_excl(dev, addr);
6042 else
6043 err = -EINVAL;
6044
6045 /* Only return duplicate errors if NLM_F_EXCL is set */
6046 if (err == -EEXIST && !(flags & NLM_F_EXCL))
6047 err = 0;
6048
6049 return err;
6050}
6051
6052/**
6053 * ice_fdb_del - delete an entry from the hardware database
6054 * @ndm: the input from the stack
6055 * @tb: pointer to array of nladdr (unused)
6056 * @dev: the net device pointer
6057 * @addr: the MAC address entry being added
6058 * @vid: VLAN ID
6059 * @extack: netlink extended ack
6060 */
6061static int
6062ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
6063 struct net_device *dev, const unsigned char *addr,
6064 __always_unused u16 vid, struct netlink_ext_ack *extack)
6065{
6066 int err;
6067
6068 if (ndm->ndm_state & NUD_PERMANENT) {
6069 netdev_err(dev, "FDB only supports static addresses\n");
6070 return -EINVAL;
6071 }
6072
6073 if (is_unicast_ether_addr(addr))
6074 err = dev_uc_del(dev, addr);
6075 else if (is_multicast_ether_addr(addr))
6076 err = dev_mc_del(dev, addr);
6077 else
6078 err = -EINVAL;
6079
6080 return err;
6081}
6082
6083#define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6084 NETIF_F_HW_VLAN_CTAG_TX | \
6085 NETIF_F_HW_VLAN_STAG_RX | \
6086 NETIF_F_HW_VLAN_STAG_TX)
6087
6088#define NETIF_VLAN_STRIPPING_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6089 NETIF_F_HW_VLAN_STAG_RX)
6090
6091#define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
6092 NETIF_F_HW_VLAN_STAG_FILTER)
6093
6094/**
6095 * ice_fix_features - fix the netdev features flags based on device limitations
6096 * @netdev: ptr to the netdev that flags are being fixed on
6097 * @features: features that need to be checked and possibly fixed
6098 *
6099 * Make sure any fixups are made to features in this callback. This enables the
6100 * driver to not have to check unsupported configurations throughout the driver
6101 * because that's the responsiblity of this callback.
6102 *
6103 * Single VLAN Mode (SVM) Supported Features:
6104 * NETIF_F_HW_VLAN_CTAG_FILTER
6105 * NETIF_F_HW_VLAN_CTAG_RX
6106 * NETIF_F_HW_VLAN_CTAG_TX
6107 *
6108 * Double VLAN Mode (DVM) Supported Features:
6109 * NETIF_F_HW_VLAN_CTAG_FILTER
6110 * NETIF_F_HW_VLAN_CTAG_RX
6111 * NETIF_F_HW_VLAN_CTAG_TX
6112 *
6113 * NETIF_F_HW_VLAN_STAG_FILTER
6114 * NETIF_HW_VLAN_STAG_RX
6115 * NETIF_HW_VLAN_STAG_TX
6116 *
6117 * Features that need fixing:
6118 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
6119 * These are mutually exlusive as the VSI context cannot support multiple
6120 * VLAN ethertypes simultaneously for stripping and/or insertion. If this
6121 * is not done, then default to clearing the requested STAG offload
6122 * settings.
6123 *
6124 * All supported filtering has to be enabled or disabled together. For
6125 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled
6126 * together. If this is not done, then default to VLAN filtering disabled.
6127 * These are mutually exclusive as there is currently no way to
6128 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN
6129 * prune rules.
6130 */
6131static netdev_features_t
6132ice_fix_features(struct net_device *netdev, netdev_features_t features)
6133{
6134 struct ice_netdev_priv *np = netdev_priv(netdev);
6135 netdev_features_t req_vlan_fltr, cur_vlan_fltr;
6136 bool cur_ctag, cur_stag, req_ctag, req_stag;
6137
6138 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
6139 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6140 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6141
6142 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
6143 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6144 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6145
6146 if (req_vlan_fltr != cur_vlan_fltr) {
6147 if (ice_is_dvm_ena(&np->vsi->back->hw)) {
6148 if (req_ctag && req_stag) {
6149 features |= NETIF_VLAN_FILTERING_FEATURES;
6150 } else if (!req_ctag && !req_stag) {
6151 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6152 } else if ((!cur_ctag && req_ctag && !cur_stag) ||
6153 (!cur_stag && req_stag && !cur_ctag)) {
6154 features |= NETIF_VLAN_FILTERING_FEATURES;
6155 netdev_warn(netdev, "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been enabled for both types.\n");
6156 } else if ((cur_ctag && !req_ctag && cur_stag) ||
6157 (cur_stag && !req_stag && cur_ctag)) {
6158 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6159 netdev_warn(netdev, "802.1Q and 802.1ad VLAN filtering must be either both on or both off. VLAN filtering has been disabled for both types.\n");
6160 }
6161 } else {
6162 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
6163 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
6164
6165 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
6166 features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6167 }
6168 }
6169
6170 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
6171 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
6172 netdev_warn(netdev, "cannot support CTAG and STAG VLAN stripping and/or insertion simultaneously since CTAG and STAG offloads are mutually exclusive, clearing STAG offload settings\n");
6173 features &= ~(NETIF_F_HW_VLAN_STAG_RX |
6174 NETIF_F_HW_VLAN_STAG_TX);
6175 }
6176
6177 if (!(netdev->features & NETIF_F_RXFCS) &&
6178 (features & NETIF_F_RXFCS) &&
6179 (features & NETIF_VLAN_STRIPPING_FEATURES) &&
6180 !ice_vsi_has_non_zero_vlans(np->vsi)) {
6181 netdev_warn(netdev, "Disabling VLAN stripping as FCS/CRC stripping is also disabled and there is no VLAN configured\n");
6182 features &= ~NETIF_VLAN_STRIPPING_FEATURES;
6183 }
6184
6185 return features;
6186}
6187
6188/**
6189 * ice_set_rx_rings_vlan_proto - update rings with new stripped VLAN proto
6190 * @vsi: PF's VSI
6191 * @vlan_ethertype: VLAN ethertype (802.1Q or 802.1ad) in network byte order
6192 *
6193 * Store current stripped VLAN proto in ring packet context,
6194 * so it can be accessed more efficiently by packet processing code.
6195 */
6196static void
6197ice_set_rx_rings_vlan_proto(struct ice_vsi *vsi, __be16 vlan_ethertype)
6198{
6199 u16 i;
6200
6201 ice_for_each_alloc_rxq(vsi, i)
6202 vsi->rx_rings[i]->pkt_ctx.vlan_proto = vlan_ethertype;
6203}
6204
6205/**
6206 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
6207 * @vsi: PF's VSI
6208 * @features: features used to determine VLAN offload settings
6209 *
6210 * First, determine the vlan_ethertype based on the VLAN offload bits in
6211 * features. Then determine if stripping and insertion should be enabled or
6212 * disabled. Finally enable or disable VLAN stripping and insertion.
6213 */
6214static int
6215ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
6216{
6217 bool enable_stripping = true, enable_insertion = true;
6218 struct ice_vsi_vlan_ops *vlan_ops;
6219 int strip_err = 0, insert_err = 0;
6220 u16 vlan_ethertype = 0;
6221
6222 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6223
6224 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
6225 vlan_ethertype = ETH_P_8021AD;
6226 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
6227 vlan_ethertype = ETH_P_8021Q;
6228
6229 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
6230 enable_stripping = false;
6231 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
6232 enable_insertion = false;
6233
6234 if (enable_stripping)
6235 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
6236 else
6237 strip_err = vlan_ops->dis_stripping(vsi);
6238
6239 if (enable_insertion)
6240 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
6241 else
6242 insert_err = vlan_ops->dis_insertion(vsi);
6243
6244 if (strip_err || insert_err)
6245 return -EIO;
6246
6247 ice_set_rx_rings_vlan_proto(vsi, enable_stripping ?
6248 htons(vlan_ethertype) : 0);
6249
6250 return 0;
6251}
6252
6253/**
6254 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
6255 * @vsi: PF's VSI
6256 * @features: features used to determine VLAN filtering settings
6257 *
6258 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
6259 * features.
6260 */
6261static int
6262ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
6263{
6264 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6265 int err = 0;
6266
6267 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
6268 * if either bit is set
6269 */
6270 if (features &
6271 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
6272 err = vlan_ops->ena_rx_filtering(vsi);
6273 else
6274 err = vlan_ops->dis_rx_filtering(vsi);
6275
6276 return err;
6277}
6278
6279/**
6280 * ice_set_vlan_features - set VLAN settings based on suggested feature set
6281 * @netdev: ptr to the netdev being adjusted
6282 * @features: the feature set that the stack is suggesting
6283 *
6284 * Only update VLAN settings if the requested_vlan_features are different than
6285 * the current_vlan_features.
6286 */
6287static int
6288ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
6289{
6290 netdev_features_t current_vlan_features, requested_vlan_features;
6291 struct ice_netdev_priv *np = netdev_priv(netdev);
6292 struct ice_vsi *vsi = np->vsi;
6293 int err;
6294
6295 current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
6296 requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
6297 if (current_vlan_features ^ requested_vlan_features) {
6298 if ((features & NETIF_F_RXFCS) &&
6299 (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6300 dev_err(ice_pf_to_dev(vsi->back),
6301 "To enable VLAN stripping, you must first enable FCS/CRC stripping\n");
6302 return -EIO;
6303 }
6304
6305 err = ice_set_vlan_offload_features(vsi, features);
6306 if (err)
6307 return err;
6308 }
6309
6310 current_vlan_features = netdev->features &
6311 NETIF_VLAN_FILTERING_FEATURES;
6312 requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
6313 if (current_vlan_features ^ requested_vlan_features) {
6314 err = ice_set_vlan_filtering_features(vsi, features);
6315 if (err)
6316 return err;
6317 }
6318
6319 return 0;
6320}
6321
6322/**
6323 * ice_set_loopback - turn on/off loopback mode on underlying PF
6324 * @vsi: ptr to VSI
6325 * @ena: flag to indicate the on/off setting
6326 */
6327static int ice_set_loopback(struct ice_vsi *vsi, bool ena)
6328{
6329 bool if_running = netif_running(vsi->netdev);
6330 int ret;
6331
6332 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6333 ret = ice_down(vsi);
6334 if (ret) {
6335 netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
6336 return ret;
6337 }
6338 }
6339 ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
6340 if (ret)
6341 netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
6342 if (if_running)
6343 ret = ice_up(vsi);
6344
6345 return ret;
6346}
6347
6348/**
6349 * ice_set_features - set the netdev feature flags
6350 * @netdev: ptr to the netdev being adjusted
6351 * @features: the feature set that the stack is suggesting
6352 */
6353static int
6354ice_set_features(struct net_device *netdev, netdev_features_t features)
6355{
6356 netdev_features_t changed = netdev->features ^ features;
6357 struct ice_netdev_priv *np = netdev_priv(netdev);
6358 struct ice_vsi *vsi = np->vsi;
6359 struct ice_pf *pf = vsi->back;
6360 int ret = 0;
6361
6362 /* Don't set any netdev advanced features with device in Safe Mode */
6363 if (ice_is_safe_mode(pf)) {
6364 dev_err(ice_pf_to_dev(pf),
6365 "Device is in Safe Mode - not enabling advanced netdev features\n");
6366 return ret;
6367 }
6368
6369 /* Do not change setting during reset */
6370 if (ice_is_reset_in_progress(pf->state)) {
6371 dev_err(ice_pf_to_dev(pf),
6372 "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
6373 return -EBUSY;
6374 }
6375
6376 /* Multiple features can be changed in one call so keep features in
6377 * separate if/else statements to guarantee each feature is checked
6378 */
6379 if (changed & NETIF_F_RXHASH)
6380 ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH));
6381
6382 ret = ice_set_vlan_features(netdev, features);
6383 if (ret)
6384 return ret;
6385
6386 /* Turn on receive of FCS aka CRC, and after setting this
6387 * flag the packet data will have the 4 byte CRC appended
6388 */
6389 if (changed & NETIF_F_RXFCS) {
6390 if ((features & NETIF_F_RXFCS) &&
6391 (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6392 dev_err(ice_pf_to_dev(vsi->back),
6393 "To disable FCS/CRC stripping, you must first disable VLAN stripping\n");
6394 return -EIO;
6395 }
6396
6397 ice_vsi_cfg_crc_strip(vsi, !!(features & NETIF_F_RXFCS));
6398 ret = ice_down_up(vsi);
6399 if (ret)
6400 return ret;
6401 }
6402
6403 if (changed & NETIF_F_NTUPLE) {
6404 bool ena = !!(features & NETIF_F_NTUPLE);
6405
6406 ice_vsi_manage_fdir(vsi, ena);
6407 ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi);
6408 }
6409
6410 /* don't turn off hw_tc_offload when ADQ is already enabled */
6411 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
6412 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
6413 return -EACCES;
6414 }
6415
6416 if (changed & NETIF_F_HW_TC) {
6417 bool ena = !!(features & NETIF_F_HW_TC);
6418
6419 ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) :
6420 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6421 }
6422
6423 if (changed & NETIF_F_LOOPBACK)
6424 ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
6425
6426 return ret;
6427}
6428
6429/**
6430 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6431 * @vsi: VSI to setup VLAN properties for
6432 */
6433static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6434{
6435 int err;
6436
6437 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6438 if (err)
6439 return err;
6440
6441 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6442 if (err)
6443 return err;
6444
6445 return ice_vsi_add_vlan_zero(vsi);
6446}
6447
6448/**
6449 * ice_vsi_cfg_lan - Setup the VSI lan related config
6450 * @vsi: the VSI being configured
6451 *
6452 * Return 0 on success and negative value on error
6453 */
6454int ice_vsi_cfg_lan(struct ice_vsi *vsi)
6455{
6456 int err;
6457
6458 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6459 ice_set_rx_mode(vsi->netdev);
6460
6461 err = ice_vsi_vlan_setup(vsi);
6462 if (err)
6463 return err;
6464 }
6465 ice_vsi_cfg_dcb_rings(vsi);
6466
6467 err = ice_vsi_cfg_lan_txqs(vsi);
6468 if (!err && ice_is_xdp_ena_vsi(vsi))
6469 err = ice_vsi_cfg_xdp_txqs(vsi);
6470 if (!err)
6471 err = ice_vsi_cfg_rxqs(vsi);
6472
6473 return err;
6474}
6475
6476/* THEORY OF MODERATION:
6477 * The ice driver hardware works differently than the hardware that DIMLIB was
6478 * originally made for. ice hardware doesn't have packet count limits that
6479 * can trigger an interrupt, but it *does* have interrupt rate limit support,
6480 * which is hard-coded to a limit of 250,000 ints/second.
6481 * If not using dynamic moderation, the INTRL value can be modified
6482 * by ethtool rx-usecs-high.
6483 */
6484struct ice_dim {
6485 /* the throttle rate for interrupts, basically worst case delay before
6486 * an initial interrupt fires, value is stored in microseconds.
6487 */
6488 u16 itr;
6489};
6490
6491/* Make a different profile for Rx that doesn't allow quite so aggressive
6492 * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6493 * second.
6494 */
6495static const struct ice_dim rx_profile[] = {
6496 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6497 {8}, /* 125,000 ints/s */
6498 {16}, /* 62,500 ints/s */
6499 {62}, /* 16,129 ints/s */
6500 {126} /* 7,936 ints/s */
6501};
6502
6503/* The transmit profile, which has the same sorts of values
6504 * as the previous struct
6505 */
6506static const struct ice_dim tx_profile[] = {
6507 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6508 {8}, /* 125,000 ints/s */
6509 {40}, /* 16,125 ints/s */
6510 {128}, /* 7,812 ints/s */
6511 {256} /* 3,906 ints/s */
6512};
6513
6514static void ice_tx_dim_work(struct work_struct *work)
6515{
6516 struct ice_ring_container *rc;
6517 struct dim *dim;
6518 u16 itr;
6519
6520 dim = container_of(work, struct dim, work);
6521 rc = dim->priv;
6522
6523 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6524
6525 /* look up the values in our local table */
6526 itr = tx_profile[dim->profile_ix].itr;
6527
6528 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6529 ice_write_itr(rc, itr);
6530
6531 dim->state = DIM_START_MEASURE;
6532}
6533
6534static void ice_rx_dim_work(struct work_struct *work)
6535{
6536 struct ice_ring_container *rc;
6537 struct dim *dim;
6538 u16 itr;
6539
6540 dim = container_of(work, struct dim, work);
6541 rc = dim->priv;
6542
6543 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6544
6545 /* look up the values in our local table */
6546 itr = rx_profile[dim->profile_ix].itr;
6547
6548 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6549 ice_write_itr(rc, itr);
6550
6551 dim->state = DIM_START_MEASURE;
6552}
6553
6554#define ICE_DIM_DEFAULT_PROFILE_IX 1
6555
6556/**
6557 * ice_init_moderation - set up interrupt moderation
6558 * @q_vector: the vector containing rings to be configured
6559 *
6560 * Set up interrupt moderation registers, with the intent to do the right thing
6561 * when called from reset or from probe, and whether or not dynamic moderation
6562 * is enabled or not. Take special care to write all the registers in both
6563 * dynamic moderation mode or not in order to make sure hardware is in a known
6564 * state.
6565 */
6566static void ice_init_moderation(struct ice_q_vector *q_vector)
6567{
6568 struct ice_ring_container *rc;
6569 bool tx_dynamic, rx_dynamic;
6570
6571 rc = &q_vector->tx;
6572 INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6573 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6574 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6575 rc->dim.priv = rc;
6576 tx_dynamic = ITR_IS_DYNAMIC(rc);
6577
6578 /* set the initial TX ITR to match the above */
6579 ice_write_itr(rc, tx_dynamic ?
6580 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6581
6582 rc = &q_vector->rx;
6583 INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6584 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6585 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6586 rc->dim.priv = rc;
6587 rx_dynamic = ITR_IS_DYNAMIC(rc);
6588
6589 /* set the initial RX ITR to match the above */
6590 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6591 rc->itr_setting);
6592
6593 ice_set_q_vector_intrl(q_vector);
6594}
6595
6596/**
6597 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6598 * @vsi: the VSI being configured
6599 */
6600static void ice_napi_enable_all(struct ice_vsi *vsi)
6601{
6602 int q_idx;
6603
6604 if (!vsi->netdev)
6605 return;
6606
6607 ice_for_each_q_vector(vsi, q_idx) {
6608 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6609
6610 ice_init_moderation(q_vector);
6611
6612 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6613 napi_enable(&q_vector->napi);
6614 }
6615}
6616
6617/**
6618 * ice_up_complete - Finish the last steps of bringing up a connection
6619 * @vsi: The VSI being configured
6620 *
6621 * Return 0 on success and negative value on error
6622 */
6623static int ice_up_complete(struct ice_vsi *vsi)
6624{
6625 struct ice_pf *pf = vsi->back;
6626 int err;
6627
6628 ice_vsi_cfg_msix(vsi);
6629
6630 /* Enable only Rx rings, Tx rings were enabled by the FW when the
6631 * Tx queue group list was configured and the context bits were
6632 * programmed using ice_vsi_cfg_txqs
6633 */
6634 err = ice_vsi_start_all_rx_rings(vsi);
6635 if (err)
6636 return err;
6637
6638 clear_bit(ICE_VSI_DOWN, vsi->state);
6639 ice_napi_enable_all(vsi);
6640 ice_vsi_ena_irq(vsi);
6641
6642 if (vsi->port_info &&
6643 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6644 vsi->netdev && vsi->type == ICE_VSI_PF) {
6645 ice_print_link_msg(vsi, true);
6646 netif_tx_start_all_queues(vsi->netdev);
6647 netif_carrier_on(vsi->netdev);
6648 ice_ptp_link_change(pf, pf->hw.pf_id, true);
6649 }
6650
6651 /* Perform an initial read of the statistics registers now to
6652 * set the baseline so counters are ready when interface is up
6653 */
6654 ice_update_eth_stats(vsi);
6655
6656 if (vsi->type == ICE_VSI_PF)
6657 ice_service_task_schedule(pf);
6658
6659 return 0;
6660}
6661
6662/**
6663 * ice_up - Bring the connection back up after being down
6664 * @vsi: VSI being configured
6665 */
6666int ice_up(struct ice_vsi *vsi)
6667{
6668 int err;
6669
6670 err = ice_vsi_cfg_lan(vsi);
6671 if (!err)
6672 err = ice_up_complete(vsi);
6673
6674 return err;
6675}
6676
6677/**
6678 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6679 * @syncp: pointer to u64_stats_sync
6680 * @stats: stats that pkts and bytes count will be taken from
6681 * @pkts: packets stats counter
6682 * @bytes: bytes stats counter
6683 *
6684 * This function fetches stats from the ring considering the atomic operations
6685 * that needs to be performed to read u64 values in 32 bit machine.
6686 */
6687void
6688ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6689 struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6690{
6691 unsigned int start;
6692
6693 do {
6694 start = u64_stats_fetch_begin(syncp);
6695 *pkts = stats.pkts;
6696 *bytes = stats.bytes;
6697 } while (u64_stats_fetch_retry(syncp, start));
6698}
6699
6700/**
6701 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6702 * @vsi: the VSI to be updated
6703 * @vsi_stats: the stats struct to be updated
6704 * @rings: rings to work on
6705 * @count: number of rings
6706 */
6707static void
6708ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6709 struct rtnl_link_stats64 *vsi_stats,
6710 struct ice_tx_ring **rings, u16 count)
6711{
6712 u16 i;
6713
6714 for (i = 0; i < count; i++) {
6715 struct ice_tx_ring *ring;
6716 u64 pkts = 0, bytes = 0;
6717
6718 ring = READ_ONCE(rings[i]);
6719 if (!ring || !ring->ring_stats)
6720 continue;
6721 ice_fetch_u64_stats_per_ring(&ring->ring_stats->syncp,
6722 ring->ring_stats->stats, &pkts,
6723 &bytes);
6724 vsi_stats->tx_packets += pkts;
6725 vsi_stats->tx_bytes += bytes;
6726 vsi->tx_restart += ring->ring_stats->tx_stats.restart_q;
6727 vsi->tx_busy += ring->ring_stats->tx_stats.tx_busy;
6728 vsi->tx_linearize += ring->ring_stats->tx_stats.tx_linearize;
6729 }
6730}
6731
6732/**
6733 * ice_update_vsi_ring_stats - Update VSI stats counters
6734 * @vsi: the VSI to be updated
6735 */
6736static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6737{
6738 struct rtnl_link_stats64 *net_stats, *stats_prev;
6739 struct rtnl_link_stats64 *vsi_stats;
6740 u64 pkts, bytes;
6741 int i;
6742
6743 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6744 if (!vsi_stats)
6745 return;
6746
6747 /* reset non-netdev (extended) stats */
6748 vsi->tx_restart = 0;
6749 vsi->tx_busy = 0;
6750 vsi->tx_linearize = 0;
6751 vsi->rx_buf_failed = 0;
6752 vsi->rx_page_failed = 0;
6753
6754 rcu_read_lock();
6755
6756 /* update Tx rings counters */
6757 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6758 vsi->num_txq);
6759
6760 /* update Rx rings counters */
6761 ice_for_each_rxq(vsi, i) {
6762 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6763 struct ice_ring_stats *ring_stats;
6764
6765 ring_stats = ring->ring_stats;
6766 ice_fetch_u64_stats_per_ring(&ring_stats->syncp,
6767 ring_stats->stats, &pkts,
6768 &bytes);
6769 vsi_stats->rx_packets += pkts;
6770 vsi_stats->rx_bytes += bytes;
6771 vsi->rx_buf_failed += ring_stats->rx_stats.alloc_buf_failed;
6772 vsi->rx_page_failed += ring_stats->rx_stats.alloc_page_failed;
6773 }
6774
6775 /* update XDP Tx rings counters */
6776 if (ice_is_xdp_ena_vsi(vsi))
6777 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6778 vsi->num_xdp_txq);
6779
6780 rcu_read_unlock();
6781
6782 net_stats = &vsi->net_stats;
6783 stats_prev = &vsi->net_stats_prev;
6784
6785 /* clear prev counters after reset */
6786 if (vsi_stats->tx_packets < stats_prev->tx_packets ||
6787 vsi_stats->rx_packets < stats_prev->rx_packets) {
6788 stats_prev->tx_packets = 0;
6789 stats_prev->tx_bytes = 0;
6790 stats_prev->rx_packets = 0;
6791 stats_prev->rx_bytes = 0;
6792 }
6793
6794 /* update netdev counters */
6795 net_stats->tx_packets += vsi_stats->tx_packets - stats_prev->tx_packets;
6796 net_stats->tx_bytes += vsi_stats->tx_bytes - stats_prev->tx_bytes;
6797 net_stats->rx_packets += vsi_stats->rx_packets - stats_prev->rx_packets;
6798 net_stats->rx_bytes += vsi_stats->rx_bytes - stats_prev->rx_bytes;
6799
6800 stats_prev->tx_packets = vsi_stats->tx_packets;
6801 stats_prev->tx_bytes = vsi_stats->tx_bytes;
6802 stats_prev->rx_packets = vsi_stats->rx_packets;
6803 stats_prev->rx_bytes = vsi_stats->rx_bytes;
6804
6805 kfree(vsi_stats);
6806}
6807
6808/**
6809 * ice_update_vsi_stats - Update VSI stats counters
6810 * @vsi: the VSI to be updated
6811 */
6812void ice_update_vsi_stats(struct ice_vsi *vsi)
6813{
6814 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6815 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6816 struct ice_pf *pf = vsi->back;
6817
6818 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6819 test_bit(ICE_CFG_BUSY, pf->state))
6820 return;
6821
6822 /* get stats as recorded by Tx/Rx rings */
6823 ice_update_vsi_ring_stats(vsi);
6824
6825 /* get VSI stats as recorded by the hardware */
6826 ice_update_eth_stats(vsi);
6827
6828 cur_ns->tx_errors = cur_es->tx_errors;
6829 cur_ns->rx_dropped = cur_es->rx_discards;
6830 cur_ns->tx_dropped = cur_es->tx_discards;
6831 cur_ns->multicast = cur_es->rx_multicast;
6832
6833 /* update some more netdev stats if this is main VSI */
6834 if (vsi->type == ICE_VSI_PF) {
6835 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6836 cur_ns->rx_errors = pf->stats.crc_errors +
6837 pf->stats.illegal_bytes +
6838 pf->stats.rx_undersize +
6839 pf->hw_csum_rx_error +
6840 pf->stats.rx_jabber +
6841 pf->stats.rx_fragments +
6842 pf->stats.rx_oversize;
6843 /* record drops from the port level */
6844 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6845 }
6846}
6847
6848/**
6849 * ice_update_pf_stats - Update PF port stats counters
6850 * @pf: PF whose stats needs to be updated
6851 */
6852void ice_update_pf_stats(struct ice_pf *pf)
6853{
6854 struct ice_hw_port_stats *prev_ps, *cur_ps;
6855 struct ice_hw *hw = &pf->hw;
6856 u16 fd_ctr_base;
6857 u8 port;
6858
6859 port = hw->port_info->lport;
6860 prev_ps = &pf->stats_prev;
6861 cur_ps = &pf->stats;
6862
6863 if (ice_is_reset_in_progress(pf->state))
6864 pf->stat_prev_loaded = false;
6865
6866 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6867 &prev_ps->eth.rx_bytes,
6868 &cur_ps->eth.rx_bytes);
6869
6870 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6871 &prev_ps->eth.rx_unicast,
6872 &cur_ps->eth.rx_unicast);
6873
6874 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6875 &prev_ps->eth.rx_multicast,
6876 &cur_ps->eth.rx_multicast);
6877
6878 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6879 &prev_ps->eth.rx_broadcast,
6880 &cur_ps->eth.rx_broadcast);
6881
6882 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6883 &prev_ps->eth.rx_discards,
6884 &cur_ps->eth.rx_discards);
6885
6886 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6887 &prev_ps->eth.tx_bytes,
6888 &cur_ps->eth.tx_bytes);
6889
6890 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6891 &prev_ps->eth.tx_unicast,
6892 &cur_ps->eth.tx_unicast);
6893
6894 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6895 &prev_ps->eth.tx_multicast,
6896 &cur_ps->eth.tx_multicast);
6897
6898 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6899 &prev_ps->eth.tx_broadcast,
6900 &cur_ps->eth.tx_broadcast);
6901
6902 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6903 &prev_ps->tx_dropped_link_down,
6904 &cur_ps->tx_dropped_link_down);
6905
6906 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6907 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6908
6909 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6910 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6911
6912 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6913 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6914
6915 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6916 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6917
6918 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6919 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6920
6921 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6922 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6923
6924 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6925 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6926
6927 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6928 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6929
6930 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6931 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6932
6933 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6934 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6935
6936 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6937 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6938
6939 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6940 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6941
6942 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6943 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6944
6945 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6946 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6947
6948 fd_ctr_base = hw->fd_ctr_base;
6949
6950 ice_stat_update40(hw,
6951 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6952 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6953 &cur_ps->fd_sb_match);
6954 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6955 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6956
6957 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6958 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6959
6960 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6961 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6962
6963 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6964 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6965
6966 ice_update_dcb_stats(pf);
6967
6968 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6969 &prev_ps->crc_errors, &cur_ps->crc_errors);
6970
6971 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6972 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6973
6974 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6975 &prev_ps->mac_local_faults,
6976 &cur_ps->mac_local_faults);
6977
6978 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6979 &prev_ps->mac_remote_faults,
6980 &cur_ps->mac_remote_faults);
6981
6982 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6983 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6984
6985 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6986 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6987
6988 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6989 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6990
6991 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6992 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6993
6994 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6995
6996 pf->stat_prev_loaded = true;
6997}
6998
6999/**
7000 * ice_get_stats64 - get statistics for network device structure
7001 * @netdev: network interface device structure
7002 * @stats: main device statistics structure
7003 */
7004static
7005void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
7006{
7007 struct ice_netdev_priv *np = netdev_priv(netdev);
7008 struct rtnl_link_stats64 *vsi_stats;
7009 struct ice_vsi *vsi = np->vsi;
7010
7011 vsi_stats = &vsi->net_stats;
7012
7013 if (!vsi->num_txq || !vsi->num_rxq)
7014 return;
7015
7016 /* netdev packet/byte stats come from ring counter. These are obtained
7017 * by summing up ring counters (done by ice_update_vsi_ring_stats).
7018 * But, only call the update routine and read the registers if VSI is
7019 * not down.
7020 */
7021 if (!test_bit(ICE_VSI_DOWN, vsi->state))
7022 ice_update_vsi_ring_stats(vsi);
7023 stats->tx_packets = vsi_stats->tx_packets;
7024 stats->tx_bytes = vsi_stats->tx_bytes;
7025 stats->rx_packets = vsi_stats->rx_packets;
7026 stats->rx_bytes = vsi_stats->rx_bytes;
7027
7028 /* The rest of the stats can be read from the hardware but instead we
7029 * just return values that the watchdog task has already obtained from
7030 * the hardware.
7031 */
7032 stats->multicast = vsi_stats->multicast;
7033 stats->tx_errors = vsi_stats->tx_errors;
7034 stats->tx_dropped = vsi_stats->tx_dropped;
7035 stats->rx_errors = vsi_stats->rx_errors;
7036 stats->rx_dropped = vsi_stats->rx_dropped;
7037 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
7038 stats->rx_length_errors = vsi_stats->rx_length_errors;
7039}
7040
7041/**
7042 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
7043 * @vsi: VSI having NAPI disabled
7044 */
7045static void ice_napi_disable_all(struct ice_vsi *vsi)
7046{
7047 int q_idx;
7048
7049 if (!vsi->netdev)
7050 return;
7051
7052 ice_for_each_q_vector(vsi, q_idx) {
7053 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
7054
7055 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
7056 napi_disable(&q_vector->napi);
7057
7058 cancel_work_sync(&q_vector->tx.dim.work);
7059 cancel_work_sync(&q_vector->rx.dim.work);
7060 }
7061}
7062
7063/**
7064 * ice_down - Shutdown the connection
7065 * @vsi: The VSI being stopped
7066 *
7067 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
7068 */
7069int ice_down(struct ice_vsi *vsi)
7070{
7071 int i, tx_err, rx_err, vlan_err = 0;
7072
7073 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
7074
7075 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
7076 vlan_err = ice_vsi_del_vlan_zero(vsi);
7077 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
7078 netif_carrier_off(vsi->netdev);
7079 netif_tx_disable(vsi->netdev);
7080 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
7081 ice_eswitch_stop_all_tx_queues(vsi->back);
7082 }
7083
7084 ice_vsi_dis_irq(vsi);
7085
7086 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
7087 if (tx_err)
7088 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
7089 vsi->vsi_num, tx_err);
7090 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
7091 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
7092 if (tx_err)
7093 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
7094 vsi->vsi_num, tx_err);
7095 }
7096
7097 rx_err = ice_vsi_stop_all_rx_rings(vsi);
7098 if (rx_err)
7099 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
7100 vsi->vsi_num, rx_err);
7101
7102 ice_napi_disable_all(vsi);
7103
7104 ice_for_each_txq(vsi, i)
7105 ice_clean_tx_ring(vsi->tx_rings[i]);
7106
7107 if (ice_is_xdp_ena_vsi(vsi))
7108 ice_for_each_xdp_txq(vsi, i)
7109 ice_clean_tx_ring(vsi->xdp_rings[i]);
7110
7111 ice_for_each_rxq(vsi, i)
7112 ice_clean_rx_ring(vsi->rx_rings[i]);
7113
7114 if (tx_err || rx_err || vlan_err) {
7115 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
7116 vsi->vsi_num, vsi->vsw->sw_id);
7117 return -EIO;
7118 }
7119
7120 return 0;
7121}
7122
7123/**
7124 * ice_down_up - shutdown the VSI connection and bring it up
7125 * @vsi: the VSI to be reconnected
7126 */
7127int ice_down_up(struct ice_vsi *vsi)
7128{
7129 int ret;
7130
7131 /* if DOWN already set, nothing to do */
7132 if (test_and_set_bit(ICE_VSI_DOWN, vsi->state))
7133 return 0;
7134
7135 ret = ice_down(vsi);
7136 if (ret)
7137 return ret;
7138
7139 ret = ice_up(vsi);
7140 if (ret) {
7141 netdev_err(vsi->netdev, "reallocating resources failed during netdev features change, may need to reload driver\n");
7142 return ret;
7143 }
7144
7145 return 0;
7146}
7147
7148/**
7149 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
7150 * @vsi: VSI having resources allocated
7151 *
7152 * Return 0 on success, negative on failure
7153 */
7154int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
7155{
7156 int i, err = 0;
7157
7158 if (!vsi->num_txq) {
7159 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
7160 vsi->vsi_num);
7161 return -EINVAL;
7162 }
7163
7164 ice_for_each_txq(vsi, i) {
7165 struct ice_tx_ring *ring = vsi->tx_rings[i];
7166
7167 if (!ring)
7168 return -EINVAL;
7169
7170 if (vsi->netdev)
7171 ring->netdev = vsi->netdev;
7172 err = ice_setup_tx_ring(ring);
7173 if (err)
7174 break;
7175 }
7176
7177 return err;
7178}
7179
7180/**
7181 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
7182 * @vsi: VSI having resources allocated
7183 *
7184 * Return 0 on success, negative on failure
7185 */
7186int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
7187{
7188 int i, err = 0;
7189
7190 if (!vsi->num_rxq) {
7191 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
7192 vsi->vsi_num);
7193 return -EINVAL;
7194 }
7195
7196 ice_for_each_rxq(vsi, i) {
7197 struct ice_rx_ring *ring = vsi->rx_rings[i];
7198
7199 if (!ring)
7200 return -EINVAL;
7201
7202 if (vsi->netdev)
7203 ring->netdev = vsi->netdev;
7204 err = ice_setup_rx_ring(ring);
7205 if (err)
7206 break;
7207 }
7208
7209 return err;
7210}
7211
7212/**
7213 * ice_vsi_open_ctrl - open control VSI for use
7214 * @vsi: the VSI to open
7215 *
7216 * Initialization of the Control VSI
7217 *
7218 * Returns 0 on success, negative value on error
7219 */
7220int ice_vsi_open_ctrl(struct ice_vsi *vsi)
7221{
7222 char int_name[ICE_INT_NAME_STR_LEN];
7223 struct ice_pf *pf = vsi->back;
7224 struct device *dev;
7225 int err;
7226
7227 dev = ice_pf_to_dev(pf);
7228 /* allocate descriptors */
7229 err = ice_vsi_setup_tx_rings(vsi);
7230 if (err)
7231 goto err_setup_tx;
7232
7233 err = ice_vsi_setup_rx_rings(vsi);
7234 if (err)
7235 goto err_setup_rx;
7236
7237 err = ice_vsi_cfg_lan(vsi);
7238 if (err)
7239 goto err_setup_rx;
7240
7241 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
7242 dev_driver_string(dev), dev_name(dev));
7243 err = ice_vsi_req_irq_msix(vsi, int_name);
7244 if (err)
7245 goto err_setup_rx;
7246
7247 ice_vsi_cfg_msix(vsi);
7248
7249 err = ice_vsi_start_all_rx_rings(vsi);
7250 if (err)
7251 goto err_up_complete;
7252
7253 clear_bit(ICE_VSI_DOWN, vsi->state);
7254 ice_vsi_ena_irq(vsi);
7255
7256 return 0;
7257
7258err_up_complete:
7259 ice_down(vsi);
7260err_setup_rx:
7261 ice_vsi_free_rx_rings(vsi);
7262err_setup_tx:
7263 ice_vsi_free_tx_rings(vsi);
7264
7265 return err;
7266}
7267
7268/**
7269 * ice_vsi_open - Called when a network interface is made active
7270 * @vsi: the VSI to open
7271 *
7272 * Initialization of the VSI
7273 *
7274 * Returns 0 on success, negative value on error
7275 */
7276int ice_vsi_open(struct ice_vsi *vsi)
7277{
7278 char int_name[ICE_INT_NAME_STR_LEN];
7279 struct ice_pf *pf = vsi->back;
7280 int err;
7281
7282 /* allocate descriptors */
7283 err = ice_vsi_setup_tx_rings(vsi);
7284 if (err)
7285 goto err_setup_tx;
7286
7287 err = ice_vsi_setup_rx_rings(vsi);
7288 if (err)
7289 goto err_setup_rx;
7290
7291 err = ice_vsi_cfg_lan(vsi);
7292 if (err)
7293 goto err_setup_rx;
7294
7295 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
7296 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
7297 err = ice_vsi_req_irq_msix(vsi, int_name);
7298 if (err)
7299 goto err_setup_rx;
7300
7301 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
7302
7303 if (vsi->type == ICE_VSI_PF) {
7304 /* Notify the stack of the actual queue counts. */
7305 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
7306 if (err)
7307 goto err_set_qs;
7308
7309 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
7310 if (err)
7311 goto err_set_qs;
7312 }
7313
7314 err = ice_up_complete(vsi);
7315 if (err)
7316 goto err_up_complete;
7317
7318 return 0;
7319
7320err_up_complete:
7321 ice_down(vsi);
7322err_set_qs:
7323 ice_vsi_free_irq(vsi);
7324err_setup_rx:
7325 ice_vsi_free_rx_rings(vsi);
7326err_setup_tx:
7327 ice_vsi_free_tx_rings(vsi);
7328
7329 return err;
7330}
7331
7332/**
7333 * ice_vsi_release_all - Delete all VSIs
7334 * @pf: PF from which all VSIs are being removed
7335 */
7336static void ice_vsi_release_all(struct ice_pf *pf)
7337{
7338 int err, i;
7339
7340 if (!pf->vsi)
7341 return;
7342
7343 ice_for_each_vsi(pf, i) {
7344 if (!pf->vsi[i])
7345 continue;
7346
7347 if (pf->vsi[i]->type == ICE_VSI_CHNL)
7348 continue;
7349
7350 err = ice_vsi_release(pf->vsi[i]);
7351 if (err)
7352 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
7353 i, err, pf->vsi[i]->vsi_num);
7354 }
7355}
7356
7357/**
7358 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
7359 * @pf: pointer to the PF instance
7360 * @type: VSI type to rebuild
7361 *
7362 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
7363 */
7364static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
7365{
7366 struct device *dev = ice_pf_to_dev(pf);
7367 int i, err;
7368
7369 ice_for_each_vsi(pf, i) {
7370 struct ice_vsi *vsi = pf->vsi[i];
7371
7372 if (!vsi || vsi->type != type)
7373 continue;
7374
7375 /* rebuild the VSI */
7376 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
7377 if (err) {
7378 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
7379 err, vsi->idx, ice_vsi_type_str(type));
7380 return err;
7381 }
7382
7383 /* replay filters for the VSI */
7384 err = ice_replay_vsi(&pf->hw, vsi->idx);
7385 if (err) {
7386 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
7387 err, vsi->idx, ice_vsi_type_str(type));
7388 return err;
7389 }
7390
7391 /* Re-map HW VSI number, using VSI handle that has been
7392 * previously validated in ice_replay_vsi() call above
7393 */
7394 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
7395
7396 /* enable the VSI */
7397 err = ice_ena_vsi(vsi, false);
7398 if (err) {
7399 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
7400 err, vsi->idx, ice_vsi_type_str(type));
7401 return err;
7402 }
7403
7404 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
7405 ice_vsi_type_str(type));
7406 }
7407
7408 return 0;
7409}
7410
7411/**
7412 * ice_update_pf_netdev_link - Update PF netdev link status
7413 * @pf: pointer to the PF instance
7414 */
7415static void ice_update_pf_netdev_link(struct ice_pf *pf)
7416{
7417 bool link_up;
7418 int i;
7419
7420 ice_for_each_vsi(pf, i) {
7421 struct ice_vsi *vsi = pf->vsi[i];
7422
7423 if (!vsi || vsi->type != ICE_VSI_PF)
7424 return;
7425
7426 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
7427 if (link_up) {
7428 netif_carrier_on(pf->vsi[i]->netdev);
7429 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
7430 } else {
7431 netif_carrier_off(pf->vsi[i]->netdev);
7432 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
7433 }
7434 }
7435}
7436
7437/**
7438 * ice_rebuild - rebuild after reset
7439 * @pf: PF to rebuild
7440 * @reset_type: type of reset
7441 *
7442 * Do not rebuild VF VSI in this flow because that is already handled via
7443 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
7444 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
7445 * to reset/rebuild all the VF VSI twice.
7446 */
7447static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
7448{
7449 struct device *dev = ice_pf_to_dev(pf);
7450 struct ice_hw *hw = &pf->hw;
7451 bool dvm;
7452 int err;
7453
7454 if (test_bit(ICE_DOWN, pf->state))
7455 goto clear_recovery;
7456
7457 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
7458
7459#define ICE_EMP_RESET_SLEEP_MS 5000
7460 if (reset_type == ICE_RESET_EMPR) {
7461 /* If an EMP reset has occurred, any previously pending flash
7462 * update will have completed. We no longer know whether or
7463 * not the NVM update EMP reset is restricted.
7464 */
7465 pf->fw_emp_reset_disabled = false;
7466
7467 msleep(ICE_EMP_RESET_SLEEP_MS);
7468 }
7469
7470 err = ice_init_all_ctrlq(hw);
7471 if (err) {
7472 dev_err(dev, "control queues init failed %d\n", err);
7473 goto err_init_ctrlq;
7474 }
7475
7476 /* if DDP was previously loaded successfully */
7477 if (!ice_is_safe_mode(pf)) {
7478 /* reload the SW DB of filter tables */
7479 if (reset_type == ICE_RESET_PFR)
7480 ice_fill_blk_tbls(hw);
7481 else
7482 /* Reload DDP Package after CORER/GLOBR reset */
7483 ice_load_pkg(NULL, pf);
7484 }
7485
7486 err = ice_clear_pf_cfg(hw);
7487 if (err) {
7488 dev_err(dev, "clear PF configuration failed %d\n", err);
7489 goto err_init_ctrlq;
7490 }
7491
7492 ice_clear_pxe_mode(hw);
7493
7494 err = ice_init_nvm(hw);
7495 if (err) {
7496 dev_err(dev, "ice_init_nvm failed %d\n", err);
7497 goto err_init_ctrlq;
7498 }
7499
7500 err = ice_get_caps(hw);
7501 if (err) {
7502 dev_err(dev, "ice_get_caps failed %d\n", err);
7503 goto err_init_ctrlq;
7504 }
7505
7506 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7507 if (err) {
7508 dev_err(dev, "set_mac_cfg failed %d\n", err);
7509 goto err_init_ctrlq;
7510 }
7511
7512 dvm = ice_is_dvm_ena(hw);
7513
7514 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7515 if (err)
7516 goto err_init_ctrlq;
7517
7518 err = ice_sched_init_port(hw->port_info);
7519 if (err)
7520 goto err_sched_init_port;
7521
7522 /* start misc vector */
7523 err = ice_req_irq_msix_misc(pf);
7524 if (err) {
7525 dev_err(dev, "misc vector setup failed: %d\n", err);
7526 goto err_sched_init_port;
7527 }
7528
7529 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7530 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7531 if (!rd32(hw, PFQF_FD_SIZE)) {
7532 u16 unused, guar, b_effort;
7533
7534 guar = hw->func_caps.fd_fltr_guar;
7535 b_effort = hw->func_caps.fd_fltr_best_effort;
7536
7537 /* force guaranteed filter pool for PF */
7538 ice_alloc_fd_guar_item(hw, &unused, guar);
7539 /* force shared filter pool for PF */
7540 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7541 }
7542 }
7543
7544 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7545 ice_dcb_rebuild(pf);
7546
7547 /* If the PF previously had enabled PTP, PTP init needs to happen before
7548 * the VSI rebuild. If not, this causes the PTP link status events to
7549 * fail.
7550 */
7551 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7552 ice_ptp_reset(pf);
7553
7554 if (ice_is_feature_supported(pf, ICE_F_GNSS))
7555 ice_gnss_init(pf);
7556
7557 /* rebuild PF VSI */
7558 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7559 if (err) {
7560 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7561 goto err_vsi_rebuild;
7562 }
7563
7564 err = ice_eswitch_rebuild(pf);
7565 if (err) {
7566 dev_err(dev, "Switchdev rebuild failed: %d\n", err);
7567 goto err_vsi_rebuild;
7568 }
7569
7570 if (reset_type == ICE_RESET_PFR) {
7571 err = ice_rebuild_channels(pf);
7572 if (err) {
7573 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7574 err);
7575 goto err_vsi_rebuild;
7576 }
7577 }
7578
7579 /* If Flow Director is active */
7580 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7581 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7582 if (err) {
7583 dev_err(dev, "control VSI rebuild failed: %d\n", err);
7584 goto err_vsi_rebuild;
7585 }
7586
7587 /* replay HW Flow Director recipes */
7588 if (hw->fdir_prof)
7589 ice_fdir_replay_flows(hw);
7590
7591 /* replay Flow Director filters */
7592 ice_fdir_replay_fltrs(pf);
7593
7594 ice_rebuild_arfs(pf);
7595 }
7596
7597 ice_update_pf_netdev_link(pf);
7598
7599 /* tell the firmware we are up */
7600 err = ice_send_version(pf);
7601 if (err) {
7602 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7603 err);
7604 goto err_vsi_rebuild;
7605 }
7606
7607 ice_replay_post(hw);
7608
7609 /* if we get here, reset flow is successful */
7610 clear_bit(ICE_RESET_FAILED, pf->state);
7611
7612 ice_plug_aux_dev(pf);
7613 if (ice_is_feature_supported(pf, ICE_F_SRIOV_LAG))
7614 ice_lag_rebuild(pf);
7615
7616 /* Restore timestamp mode settings after VSI rebuild */
7617 ice_ptp_restore_timestamp_mode(pf);
7618 return;
7619
7620err_vsi_rebuild:
7621err_sched_init_port:
7622 ice_sched_cleanup_all(hw);
7623err_init_ctrlq:
7624 ice_shutdown_all_ctrlq(hw);
7625 set_bit(ICE_RESET_FAILED, pf->state);
7626clear_recovery:
7627 /* set this bit in PF state to control service task scheduling */
7628 set_bit(ICE_NEEDS_RESTART, pf->state);
7629 dev_err(dev, "Rebuild failed, unload and reload driver\n");
7630}
7631
7632/**
7633 * ice_change_mtu - NDO callback to change the MTU
7634 * @netdev: network interface device structure
7635 * @new_mtu: new value for maximum frame size
7636 *
7637 * Returns 0 on success, negative on failure
7638 */
7639static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7640{
7641 struct ice_netdev_priv *np = netdev_priv(netdev);
7642 struct ice_vsi *vsi = np->vsi;
7643 struct ice_pf *pf = vsi->back;
7644 struct bpf_prog *prog;
7645 u8 count = 0;
7646 int err = 0;
7647
7648 if (new_mtu == (int)netdev->mtu) {
7649 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7650 return 0;
7651 }
7652
7653 prog = vsi->xdp_prog;
7654 if (prog && !prog->aux->xdp_has_frags) {
7655 int frame_size = ice_max_xdp_frame_size(vsi);
7656
7657 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7658 netdev_err(netdev, "max MTU for XDP usage is %d\n",
7659 frame_size - ICE_ETH_PKT_HDR_PAD);
7660 return -EINVAL;
7661 }
7662 } else if (test_bit(ICE_FLAG_LEGACY_RX, pf->flags)) {
7663 if (new_mtu + ICE_ETH_PKT_HDR_PAD > ICE_MAX_FRAME_LEGACY_RX) {
7664 netdev_err(netdev, "Too big MTU for legacy-rx; Max is %d\n",
7665 ICE_MAX_FRAME_LEGACY_RX - ICE_ETH_PKT_HDR_PAD);
7666 return -EINVAL;
7667 }
7668 }
7669
7670 /* if a reset is in progress, wait for some time for it to complete */
7671 do {
7672 if (ice_is_reset_in_progress(pf->state)) {
7673 count++;
7674 usleep_range(1000, 2000);
7675 } else {
7676 break;
7677 }
7678
7679 } while (count < 100);
7680
7681 if (count == 100) {
7682 netdev_err(netdev, "can't change MTU. Device is busy\n");
7683 return -EBUSY;
7684 }
7685
7686 netdev->mtu = (unsigned int)new_mtu;
7687 err = ice_down_up(vsi);
7688 if (err)
7689 return err;
7690
7691 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7692 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7693
7694 return err;
7695}
7696
7697/**
7698 * ice_eth_ioctl - Access the hwtstamp interface
7699 * @netdev: network interface device structure
7700 * @ifr: interface request data
7701 * @cmd: ioctl command
7702 */
7703static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7704{
7705 struct ice_netdev_priv *np = netdev_priv(netdev);
7706 struct ice_pf *pf = np->vsi->back;
7707
7708 switch (cmd) {
7709 case SIOCGHWTSTAMP:
7710 return ice_ptp_get_ts_config(pf, ifr);
7711 case SIOCSHWTSTAMP:
7712 return ice_ptp_set_ts_config(pf, ifr);
7713 default:
7714 return -EOPNOTSUPP;
7715 }
7716}
7717
7718/**
7719 * ice_aq_str - convert AQ err code to a string
7720 * @aq_err: the AQ error code to convert
7721 */
7722const char *ice_aq_str(enum ice_aq_err aq_err)
7723{
7724 switch (aq_err) {
7725 case ICE_AQ_RC_OK:
7726 return "OK";
7727 case ICE_AQ_RC_EPERM:
7728 return "ICE_AQ_RC_EPERM";
7729 case ICE_AQ_RC_ENOENT:
7730 return "ICE_AQ_RC_ENOENT";
7731 case ICE_AQ_RC_ENOMEM:
7732 return "ICE_AQ_RC_ENOMEM";
7733 case ICE_AQ_RC_EBUSY:
7734 return "ICE_AQ_RC_EBUSY";
7735 case ICE_AQ_RC_EEXIST:
7736 return "ICE_AQ_RC_EEXIST";
7737 case ICE_AQ_RC_EINVAL:
7738 return "ICE_AQ_RC_EINVAL";
7739 case ICE_AQ_RC_ENOSPC:
7740 return "ICE_AQ_RC_ENOSPC";
7741 case ICE_AQ_RC_ENOSYS:
7742 return "ICE_AQ_RC_ENOSYS";
7743 case ICE_AQ_RC_EMODE:
7744 return "ICE_AQ_RC_EMODE";
7745 case ICE_AQ_RC_ENOSEC:
7746 return "ICE_AQ_RC_ENOSEC";
7747 case ICE_AQ_RC_EBADSIG:
7748 return "ICE_AQ_RC_EBADSIG";
7749 case ICE_AQ_RC_ESVN:
7750 return "ICE_AQ_RC_ESVN";
7751 case ICE_AQ_RC_EBADMAN:
7752 return "ICE_AQ_RC_EBADMAN";
7753 case ICE_AQ_RC_EBADBUF:
7754 return "ICE_AQ_RC_EBADBUF";
7755 }
7756
7757 return "ICE_AQ_RC_UNKNOWN";
7758}
7759
7760/**
7761 * ice_set_rss_lut - Set RSS LUT
7762 * @vsi: Pointer to VSI structure
7763 * @lut: Lookup table
7764 * @lut_size: Lookup table size
7765 *
7766 * Returns 0 on success, negative on failure
7767 */
7768int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7769{
7770 struct ice_aq_get_set_rss_lut_params params = {};
7771 struct ice_hw *hw = &vsi->back->hw;
7772 int status;
7773
7774 if (!lut)
7775 return -EINVAL;
7776
7777 params.vsi_handle = vsi->idx;
7778 params.lut_size = lut_size;
7779 params.lut_type = vsi->rss_lut_type;
7780 params.lut = lut;
7781
7782 status = ice_aq_set_rss_lut(hw, ¶ms);
7783 if (status)
7784 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7785 status, ice_aq_str(hw->adminq.sq_last_status));
7786
7787 return status;
7788}
7789
7790/**
7791 * ice_set_rss_key - Set RSS key
7792 * @vsi: Pointer to the VSI structure
7793 * @seed: RSS hash seed
7794 *
7795 * Returns 0 on success, negative on failure
7796 */
7797int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7798{
7799 struct ice_hw *hw = &vsi->back->hw;
7800 int status;
7801
7802 if (!seed)
7803 return -EINVAL;
7804
7805 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7806 if (status)
7807 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7808 status, ice_aq_str(hw->adminq.sq_last_status));
7809
7810 return status;
7811}
7812
7813/**
7814 * ice_get_rss_lut - Get RSS LUT
7815 * @vsi: Pointer to VSI structure
7816 * @lut: Buffer to store the lookup table entries
7817 * @lut_size: Size of buffer to store the lookup table entries
7818 *
7819 * Returns 0 on success, negative on failure
7820 */
7821int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7822{
7823 struct ice_aq_get_set_rss_lut_params params = {};
7824 struct ice_hw *hw = &vsi->back->hw;
7825 int status;
7826
7827 if (!lut)
7828 return -EINVAL;
7829
7830 params.vsi_handle = vsi->idx;
7831 params.lut_size = lut_size;
7832 params.lut_type = vsi->rss_lut_type;
7833 params.lut = lut;
7834
7835 status = ice_aq_get_rss_lut(hw, ¶ms);
7836 if (status)
7837 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7838 status, ice_aq_str(hw->adminq.sq_last_status));
7839
7840 return status;
7841}
7842
7843/**
7844 * ice_get_rss_key - Get RSS key
7845 * @vsi: Pointer to VSI structure
7846 * @seed: Buffer to store the key in
7847 *
7848 * Returns 0 on success, negative on failure
7849 */
7850int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7851{
7852 struct ice_hw *hw = &vsi->back->hw;
7853 int status;
7854
7855 if (!seed)
7856 return -EINVAL;
7857
7858 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7859 if (status)
7860 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7861 status, ice_aq_str(hw->adminq.sq_last_status));
7862
7863 return status;
7864}
7865
7866/**
7867 * ice_set_rss_hfunc - Set RSS HASH function
7868 * @vsi: Pointer to VSI structure
7869 * @hfunc: hash function (ICE_AQ_VSI_Q_OPT_RSS_*)
7870 *
7871 * Returns 0 on success, negative on failure
7872 */
7873int ice_set_rss_hfunc(struct ice_vsi *vsi, u8 hfunc)
7874{
7875 struct ice_hw *hw = &vsi->back->hw;
7876 struct ice_vsi_ctx *ctx;
7877 bool symm;
7878 int err;
7879
7880 if (hfunc == vsi->rss_hfunc)
7881 return 0;
7882
7883 if (hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ &&
7884 hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ)
7885 return -EOPNOTSUPP;
7886
7887 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
7888 if (!ctx)
7889 return -ENOMEM;
7890
7891 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
7892 ctx->info.q_opt_rss = vsi->info.q_opt_rss;
7893 ctx->info.q_opt_rss &= ~ICE_AQ_VSI_Q_OPT_RSS_HASH_M;
7894 ctx->info.q_opt_rss |=
7895 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hfunc);
7896 ctx->info.q_opt_tc = vsi->info.q_opt_tc;
7897 ctx->info.q_opt_flags = vsi->info.q_opt_rss;
7898
7899 err = ice_update_vsi(hw, vsi->idx, ctx, NULL);
7900 if (err) {
7901 dev_err(ice_pf_to_dev(vsi->back), "Failed to configure RSS hash for VSI %d, error %d\n",
7902 vsi->vsi_num, err);
7903 } else {
7904 vsi->info.q_opt_rss = ctx->info.q_opt_rss;
7905 vsi->rss_hfunc = hfunc;
7906 netdev_info(vsi->netdev, "Hash function set to: %sToeplitz\n",
7907 hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ ?
7908 "Symmetric " : "");
7909 }
7910 kfree(ctx);
7911 if (err)
7912 return err;
7913
7914 /* Fix the symmetry setting for all existing RSS configurations */
7915 symm = !!(hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ);
7916 return ice_set_rss_cfg_symm(hw, vsi, symm);
7917}
7918
7919/**
7920 * ice_bridge_getlink - Get the hardware bridge mode
7921 * @skb: skb buff
7922 * @pid: process ID
7923 * @seq: RTNL message seq
7924 * @dev: the netdev being configured
7925 * @filter_mask: filter mask passed in
7926 * @nlflags: netlink flags passed in
7927 *
7928 * Return the bridge mode (VEB/VEPA)
7929 */
7930static int
7931ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7932 struct net_device *dev, u32 filter_mask, int nlflags)
7933{
7934 struct ice_netdev_priv *np = netdev_priv(dev);
7935 struct ice_vsi *vsi = np->vsi;
7936 struct ice_pf *pf = vsi->back;
7937 u16 bmode;
7938
7939 bmode = pf->first_sw->bridge_mode;
7940
7941 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7942 filter_mask, NULL);
7943}
7944
7945/**
7946 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7947 * @vsi: Pointer to VSI structure
7948 * @bmode: Hardware bridge mode (VEB/VEPA)
7949 *
7950 * Returns 0 on success, negative on failure
7951 */
7952static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7953{
7954 struct ice_aqc_vsi_props *vsi_props;
7955 struct ice_hw *hw = &vsi->back->hw;
7956 struct ice_vsi_ctx *ctxt;
7957 int ret;
7958
7959 vsi_props = &vsi->info;
7960
7961 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7962 if (!ctxt)
7963 return -ENOMEM;
7964
7965 ctxt->info = vsi->info;
7966
7967 if (bmode == BRIDGE_MODE_VEB)
7968 /* change from VEPA to VEB mode */
7969 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7970 else
7971 /* change from VEB to VEPA mode */
7972 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7973 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7974
7975 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7976 if (ret) {
7977 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7978 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7979 goto out;
7980 }
7981 /* Update sw flags for book keeping */
7982 vsi_props->sw_flags = ctxt->info.sw_flags;
7983
7984out:
7985 kfree(ctxt);
7986 return ret;
7987}
7988
7989/**
7990 * ice_bridge_setlink - Set the hardware bridge mode
7991 * @dev: the netdev being configured
7992 * @nlh: RTNL message
7993 * @flags: bridge setlink flags
7994 * @extack: netlink extended ack
7995 *
7996 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7997 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7998 * not already set for all VSIs connected to this switch. And also update the
7999 * unicast switch filter rules for the corresponding switch of the netdev.
8000 */
8001static int
8002ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
8003 u16 __always_unused flags,
8004 struct netlink_ext_ack __always_unused *extack)
8005{
8006 struct ice_netdev_priv *np = netdev_priv(dev);
8007 struct ice_pf *pf = np->vsi->back;
8008 struct nlattr *attr, *br_spec;
8009 struct ice_hw *hw = &pf->hw;
8010 struct ice_sw *pf_sw;
8011 int rem, v, err = 0;
8012
8013 pf_sw = pf->first_sw;
8014 /* find the attribute in the netlink message */
8015 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8016 if (!br_spec)
8017 return -EINVAL;
8018
8019 nla_for_each_nested(attr, br_spec, rem) {
8020 __u16 mode;
8021
8022 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8023 continue;
8024 mode = nla_get_u16(attr);
8025 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
8026 return -EINVAL;
8027 /* Continue if bridge mode is not being flipped */
8028 if (mode == pf_sw->bridge_mode)
8029 continue;
8030 /* Iterates through the PF VSI list and update the loopback
8031 * mode of the VSI
8032 */
8033 ice_for_each_vsi(pf, v) {
8034 if (!pf->vsi[v])
8035 continue;
8036 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
8037 if (err)
8038 return err;
8039 }
8040
8041 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
8042 /* Update the unicast switch filter rules for the corresponding
8043 * switch of the netdev
8044 */
8045 err = ice_update_sw_rule_bridge_mode(hw);
8046 if (err) {
8047 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
8048 mode, err,
8049 ice_aq_str(hw->adminq.sq_last_status));
8050 /* revert hw->evb_veb */
8051 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
8052 return err;
8053 }
8054
8055 pf_sw->bridge_mode = mode;
8056 }
8057
8058 return 0;
8059}
8060
8061/**
8062 * ice_tx_timeout - Respond to a Tx Hang
8063 * @netdev: network interface device structure
8064 * @txqueue: Tx queue
8065 */
8066static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
8067{
8068 struct ice_netdev_priv *np = netdev_priv(netdev);
8069 struct ice_tx_ring *tx_ring = NULL;
8070 struct ice_vsi *vsi = np->vsi;
8071 struct ice_pf *pf = vsi->back;
8072 u32 i;
8073
8074 pf->tx_timeout_count++;
8075
8076 /* Check if PFC is enabled for the TC to which the queue belongs
8077 * to. If yes then Tx timeout is not caused by a hung queue, no
8078 * need to reset and rebuild
8079 */
8080 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
8081 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
8082 txqueue);
8083 return;
8084 }
8085
8086 /* now that we have an index, find the tx_ring struct */
8087 ice_for_each_txq(vsi, i)
8088 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
8089 if (txqueue == vsi->tx_rings[i]->q_index) {
8090 tx_ring = vsi->tx_rings[i];
8091 break;
8092 }
8093
8094 /* Reset recovery level if enough time has elapsed after last timeout.
8095 * Also ensure no new reset action happens before next timeout period.
8096 */
8097 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
8098 pf->tx_timeout_recovery_level = 1;
8099 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
8100 netdev->watchdog_timeo)))
8101 return;
8102
8103 if (tx_ring) {
8104 struct ice_hw *hw = &pf->hw;
8105 u32 head, val = 0;
8106
8107 head = FIELD_GET(QTX_COMM_HEAD_HEAD_M,
8108 rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])));
8109 /* Read interrupt register */
8110 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
8111
8112 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
8113 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
8114 head, tx_ring->next_to_use, val);
8115 }
8116
8117 pf->tx_timeout_last_recovery = jiffies;
8118 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
8119 pf->tx_timeout_recovery_level, txqueue);
8120
8121 switch (pf->tx_timeout_recovery_level) {
8122 case 1:
8123 set_bit(ICE_PFR_REQ, pf->state);
8124 break;
8125 case 2:
8126 set_bit(ICE_CORER_REQ, pf->state);
8127 break;
8128 case 3:
8129 set_bit(ICE_GLOBR_REQ, pf->state);
8130 break;
8131 default:
8132 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
8133 set_bit(ICE_DOWN, pf->state);
8134 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
8135 set_bit(ICE_SERVICE_DIS, pf->state);
8136 break;
8137 }
8138
8139 ice_service_task_schedule(pf);
8140 pf->tx_timeout_recovery_level++;
8141}
8142
8143/**
8144 * ice_setup_tc_cls_flower - flower classifier offloads
8145 * @np: net device to configure
8146 * @filter_dev: device on which filter is added
8147 * @cls_flower: offload data
8148 */
8149static int
8150ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
8151 struct net_device *filter_dev,
8152 struct flow_cls_offload *cls_flower)
8153{
8154 struct ice_vsi *vsi = np->vsi;
8155
8156 if (cls_flower->common.chain_index)
8157 return -EOPNOTSUPP;
8158
8159 switch (cls_flower->command) {
8160 case FLOW_CLS_REPLACE:
8161 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
8162 case FLOW_CLS_DESTROY:
8163 return ice_del_cls_flower(vsi, cls_flower);
8164 default:
8165 return -EINVAL;
8166 }
8167}
8168
8169/**
8170 * ice_setup_tc_block_cb - callback handler registered for TC block
8171 * @type: TC SETUP type
8172 * @type_data: TC flower offload data that contains user input
8173 * @cb_priv: netdev private data
8174 */
8175static int
8176ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
8177{
8178 struct ice_netdev_priv *np = cb_priv;
8179
8180 switch (type) {
8181 case TC_SETUP_CLSFLOWER:
8182 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
8183 type_data);
8184 default:
8185 return -EOPNOTSUPP;
8186 }
8187}
8188
8189/**
8190 * ice_validate_mqprio_qopt - Validate TCF input parameters
8191 * @vsi: Pointer to VSI
8192 * @mqprio_qopt: input parameters for mqprio queue configuration
8193 *
8194 * This function validates MQPRIO params, such as qcount (power of 2 wherever
8195 * needed), and make sure user doesn't specify qcount and BW rate limit
8196 * for TCs, which are more than "num_tc"
8197 */
8198static int
8199ice_validate_mqprio_qopt(struct ice_vsi *vsi,
8200 struct tc_mqprio_qopt_offload *mqprio_qopt)
8201{
8202 int non_power_of_2_qcount = 0;
8203 struct ice_pf *pf = vsi->back;
8204 int max_rss_q_cnt = 0;
8205 u64 sum_min_rate = 0;
8206 struct device *dev;
8207 int i, speed;
8208 u8 num_tc;
8209
8210 if (vsi->type != ICE_VSI_PF)
8211 return -EINVAL;
8212
8213 if (mqprio_qopt->qopt.offset[0] != 0 ||
8214 mqprio_qopt->qopt.num_tc < 1 ||
8215 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
8216 return -EINVAL;
8217
8218 dev = ice_pf_to_dev(pf);
8219 vsi->ch_rss_size = 0;
8220 num_tc = mqprio_qopt->qopt.num_tc;
8221 speed = ice_get_link_speed_kbps(vsi);
8222
8223 for (i = 0; num_tc; i++) {
8224 int qcount = mqprio_qopt->qopt.count[i];
8225 u64 max_rate, min_rate, rem;
8226
8227 if (!qcount)
8228 return -EINVAL;
8229
8230 if (is_power_of_2(qcount)) {
8231 if (non_power_of_2_qcount &&
8232 qcount > non_power_of_2_qcount) {
8233 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
8234 qcount, non_power_of_2_qcount);
8235 return -EINVAL;
8236 }
8237 if (qcount > max_rss_q_cnt)
8238 max_rss_q_cnt = qcount;
8239 } else {
8240 if (non_power_of_2_qcount &&
8241 qcount != non_power_of_2_qcount) {
8242 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
8243 qcount, non_power_of_2_qcount);
8244 return -EINVAL;
8245 }
8246 if (qcount < max_rss_q_cnt) {
8247 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
8248 qcount, max_rss_q_cnt);
8249 return -EINVAL;
8250 }
8251 max_rss_q_cnt = qcount;
8252 non_power_of_2_qcount = qcount;
8253 }
8254
8255 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
8256 * converts the bandwidth rate limit into Bytes/s when
8257 * passing it down to the driver. So convert input bandwidth
8258 * from Bytes/s to Kbps
8259 */
8260 max_rate = mqprio_qopt->max_rate[i];
8261 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
8262
8263 /* min_rate is minimum guaranteed rate and it can't be zero */
8264 min_rate = mqprio_qopt->min_rate[i];
8265 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
8266 sum_min_rate += min_rate;
8267
8268 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
8269 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
8270 min_rate, ICE_MIN_BW_LIMIT);
8271 return -EINVAL;
8272 }
8273
8274 if (max_rate && max_rate > speed) {
8275 dev_err(dev, "TC%d: max_rate(%llu Kbps) > link speed of %u Kbps\n",
8276 i, max_rate, speed);
8277 return -EINVAL;
8278 }
8279
8280 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
8281 if (rem) {
8282 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
8283 i, ICE_MIN_BW_LIMIT);
8284 return -EINVAL;
8285 }
8286
8287 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
8288 if (rem) {
8289 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
8290 i, ICE_MIN_BW_LIMIT);
8291 return -EINVAL;
8292 }
8293
8294 /* min_rate can't be more than max_rate, except when max_rate
8295 * is zero (implies max_rate sought is max line rate). In such
8296 * a case min_rate can be more than max.
8297 */
8298 if (max_rate && min_rate > max_rate) {
8299 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
8300 min_rate, max_rate);
8301 return -EINVAL;
8302 }
8303
8304 if (i >= mqprio_qopt->qopt.num_tc - 1)
8305 break;
8306 if (mqprio_qopt->qopt.offset[i + 1] !=
8307 (mqprio_qopt->qopt.offset[i] + qcount))
8308 return -EINVAL;
8309 }
8310 if (vsi->num_rxq <
8311 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8312 return -EINVAL;
8313 if (vsi->num_txq <
8314 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8315 return -EINVAL;
8316
8317 if (sum_min_rate && sum_min_rate > (u64)speed) {
8318 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
8319 sum_min_rate, speed);
8320 return -EINVAL;
8321 }
8322
8323 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
8324 vsi->ch_rss_size = max_rss_q_cnt;
8325
8326 return 0;
8327}
8328
8329/**
8330 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
8331 * @pf: ptr to PF device
8332 * @vsi: ptr to VSI
8333 */
8334static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
8335{
8336 struct device *dev = ice_pf_to_dev(pf);
8337 bool added = false;
8338 struct ice_hw *hw;
8339 int flow;
8340
8341 if (!(vsi->num_gfltr || vsi->num_bfltr))
8342 return -EINVAL;
8343
8344 hw = &pf->hw;
8345 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
8346 struct ice_fd_hw_prof *prof;
8347 int tun, status;
8348 u64 entry_h;
8349
8350 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
8351 hw->fdir_prof[flow]->cnt))
8352 continue;
8353
8354 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
8355 enum ice_flow_priority prio;
8356
8357 /* add this VSI to FDir profile for this flow */
8358 prio = ICE_FLOW_PRIO_NORMAL;
8359 prof = hw->fdir_prof[flow];
8360 status = ice_flow_add_entry(hw, ICE_BLK_FD,
8361 prof->prof_id[tun],
8362 prof->vsi_h[0], vsi->idx,
8363 prio, prof->fdir_seg[tun],
8364 &entry_h);
8365 if (status) {
8366 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
8367 vsi->idx, flow);
8368 continue;
8369 }
8370
8371 prof->entry_h[prof->cnt][tun] = entry_h;
8372 }
8373
8374 /* store VSI for filter replay and delete */
8375 prof->vsi_h[prof->cnt] = vsi->idx;
8376 prof->cnt++;
8377
8378 added = true;
8379 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
8380 flow);
8381 }
8382
8383 if (!added)
8384 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
8385
8386 return 0;
8387}
8388
8389/**
8390 * ice_add_channel - add a channel by adding VSI
8391 * @pf: ptr to PF device
8392 * @sw_id: underlying HW switching element ID
8393 * @ch: ptr to channel structure
8394 *
8395 * Add a channel (VSI) using add_vsi and queue_map
8396 */
8397static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
8398{
8399 struct device *dev = ice_pf_to_dev(pf);
8400 struct ice_vsi *vsi;
8401
8402 if (ch->type != ICE_VSI_CHNL) {
8403 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
8404 return -EINVAL;
8405 }
8406
8407 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
8408 if (!vsi || vsi->type != ICE_VSI_CHNL) {
8409 dev_err(dev, "create chnl VSI failure\n");
8410 return -EINVAL;
8411 }
8412
8413 ice_add_vsi_to_fdir(pf, vsi);
8414
8415 ch->sw_id = sw_id;
8416 ch->vsi_num = vsi->vsi_num;
8417 ch->info.mapping_flags = vsi->info.mapping_flags;
8418 ch->ch_vsi = vsi;
8419 /* set the back pointer of channel for newly created VSI */
8420 vsi->ch = ch;
8421
8422 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
8423 sizeof(vsi->info.q_mapping));
8424 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
8425 sizeof(vsi->info.tc_mapping));
8426
8427 return 0;
8428}
8429
8430/**
8431 * ice_chnl_cfg_res
8432 * @vsi: the VSI being setup
8433 * @ch: ptr to channel structure
8434 *
8435 * Configure channel specific resources such as rings, vector.
8436 */
8437static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
8438{
8439 int i;
8440
8441 for (i = 0; i < ch->num_txq; i++) {
8442 struct ice_q_vector *tx_q_vector, *rx_q_vector;
8443 struct ice_ring_container *rc;
8444 struct ice_tx_ring *tx_ring;
8445 struct ice_rx_ring *rx_ring;
8446
8447 tx_ring = vsi->tx_rings[ch->base_q + i];
8448 rx_ring = vsi->rx_rings[ch->base_q + i];
8449 if (!tx_ring || !rx_ring)
8450 continue;
8451
8452 /* setup ring being channel enabled */
8453 tx_ring->ch = ch;
8454 rx_ring->ch = ch;
8455
8456 /* following code block sets up vector specific attributes */
8457 tx_q_vector = tx_ring->q_vector;
8458 rx_q_vector = rx_ring->q_vector;
8459 if (!tx_q_vector && !rx_q_vector)
8460 continue;
8461
8462 if (tx_q_vector) {
8463 tx_q_vector->ch = ch;
8464 /* setup Tx and Rx ITR setting if DIM is off */
8465 rc = &tx_q_vector->tx;
8466 if (!ITR_IS_DYNAMIC(rc))
8467 ice_write_itr(rc, rc->itr_setting);
8468 }
8469 if (rx_q_vector) {
8470 rx_q_vector->ch = ch;
8471 /* setup Tx and Rx ITR setting if DIM is off */
8472 rc = &rx_q_vector->rx;
8473 if (!ITR_IS_DYNAMIC(rc))
8474 ice_write_itr(rc, rc->itr_setting);
8475 }
8476 }
8477
8478 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
8479 * GLINT_ITR register would have written to perform in-context
8480 * update, hence perform flush
8481 */
8482 if (ch->num_txq || ch->num_rxq)
8483 ice_flush(&vsi->back->hw);
8484}
8485
8486/**
8487 * ice_cfg_chnl_all_res - configure channel resources
8488 * @vsi: pte to main_vsi
8489 * @ch: ptr to channel structure
8490 *
8491 * This function configures channel specific resources such as flow-director
8492 * counter index, and other resources such as queues, vectors, ITR settings
8493 */
8494static void
8495ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8496{
8497 /* configure channel (aka ADQ) resources such as queues, vectors,
8498 * ITR settings for channel specific vectors and anything else
8499 */
8500 ice_chnl_cfg_res(vsi, ch);
8501}
8502
8503/**
8504 * ice_setup_hw_channel - setup new channel
8505 * @pf: ptr to PF device
8506 * @vsi: the VSI being setup
8507 * @ch: ptr to channel structure
8508 * @sw_id: underlying HW switching element ID
8509 * @type: type of channel to be created (VMDq2/VF)
8510 *
8511 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8512 * and configures Tx rings accordingly
8513 */
8514static int
8515ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8516 struct ice_channel *ch, u16 sw_id, u8 type)
8517{
8518 struct device *dev = ice_pf_to_dev(pf);
8519 int ret;
8520
8521 ch->base_q = vsi->next_base_q;
8522 ch->type = type;
8523
8524 ret = ice_add_channel(pf, sw_id, ch);
8525 if (ret) {
8526 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8527 return ret;
8528 }
8529
8530 /* configure/setup ADQ specific resources */
8531 ice_cfg_chnl_all_res(vsi, ch);
8532
8533 /* make sure to update the next_base_q so that subsequent channel's
8534 * (aka ADQ) VSI queue map is correct
8535 */
8536 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8537 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8538 ch->num_rxq);
8539
8540 return 0;
8541}
8542
8543/**
8544 * ice_setup_channel - setup new channel using uplink element
8545 * @pf: ptr to PF device
8546 * @vsi: the VSI being setup
8547 * @ch: ptr to channel structure
8548 *
8549 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8550 * and uplink switching element
8551 */
8552static bool
8553ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8554 struct ice_channel *ch)
8555{
8556 struct device *dev = ice_pf_to_dev(pf);
8557 u16 sw_id;
8558 int ret;
8559
8560 if (vsi->type != ICE_VSI_PF) {
8561 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8562 return false;
8563 }
8564
8565 sw_id = pf->first_sw->sw_id;
8566
8567 /* create channel (VSI) */
8568 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8569 if (ret) {
8570 dev_err(dev, "failed to setup hw_channel\n");
8571 return false;
8572 }
8573 dev_dbg(dev, "successfully created channel()\n");
8574
8575 return ch->ch_vsi ? true : false;
8576}
8577
8578/**
8579 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8580 * @vsi: VSI to be configured
8581 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8582 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8583 */
8584static int
8585ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8586{
8587 int err;
8588
8589 err = ice_set_min_bw_limit(vsi, min_tx_rate);
8590 if (err)
8591 return err;
8592
8593 return ice_set_max_bw_limit(vsi, max_tx_rate);
8594}
8595
8596/**
8597 * ice_create_q_channel - function to create channel
8598 * @vsi: VSI to be configured
8599 * @ch: ptr to channel (it contains channel specific params)
8600 *
8601 * This function creates channel (VSI) using num_queues specified by user,
8602 * reconfigs RSS if needed.
8603 */
8604static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8605{
8606 struct ice_pf *pf = vsi->back;
8607 struct device *dev;
8608
8609 if (!ch)
8610 return -EINVAL;
8611
8612 dev = ice_pf_to_dev(pf);
8613 if (!ch->num_txq || !ch->num_rxq) {
8614 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8615 return -EINVAL;
8616 }
8617
8618 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8619 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8620 vsi->cnt_q_avail, ch->num_txq);
8621 return -EINVAL;
8622 }
8623
8624 if (!ice_setup_channel(pf, vsi, ch)) {
8625 dev_info(dev, "Failed to setup channel\n");
8626 return -EINVAL;
8627 }
8628 /* configure BW rate limit */
8629 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8630 int ret;
8631
8632 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8633 ch->min_tx_rate);
8634 if (ret)
8635 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8636 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8637 else
8638 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8639 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8640 }
8641
8642 vsi->cnt_q_avail -= ch->num_txq;
8643
8644 return 0;
8645}
8646
8647/**
8648 * ice_rem_all_chnl_fltrs - removes all channel filters
8649 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8650 *
8651 * Remove all advanced switch filters only if they are channel specific
8652 * tc-flower based filter
8653 */
8654static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8655{
8656 struct ice_tc_flower_fltr *fltr;
8657 struct hlist_node *node;
8658
8659 /* to remove all channel filters, iterate an ordered list of filters */
8660 hlist_for_each_entry_safe(fltr, node,
8661 &pf->tc_flower_fltr_list,
8662 tc_flower_node) {
8663 struct ice_rule_query_data rule;
8664 int status;
8665
8666 /* for now process only channel specific filters */
8667 if (!ice_is_chnl_fltr(fltr))
8668 continue;
8669
8670 rule.rid = fltr->rid;
8671 rule.rule_id = fltr->rule_id;
8672 rule.vsi_handle = fltr->dest_vsi_handle;
8673 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8674 if (status) {
8675 if (status == -ENOENT)
8676 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8677 rule.rule_id);
8678 else
8679 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8680 status);
8681 } else if (fltr->dest_vsi) {
8682 /* update advanced switch filter count */
8683 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8684 u32 flags = fltr->flags;
8685
8686 fltr->dest_vsi->num_chnl_fltr--;
8687 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8688 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8689 pf->num_dmac_chnl_fltrs--;
8690 }
8691 }
8692
8693 hlist_del(&fltr->tc_flower_node);
8694 kfree(fltr);
8695 }
8696}
8697
8698/**
8699 * ice_remove_q_channels - Remove queue channels for the TCs
8700 * @vsi: VSI to be configured
8701 * @rem_fltr: delete advanced switch filter or not
8702 *
8703 * Remove queue channels for the TCs
8704 */
8705static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8706{
8707 struct ice_channel *ch, *ch_tmp;
8708 struct ice_pf *pf = vsi->back;
8709 int i;
8710
8711 /* remove all tc-flower based filter if they are channel filters only */
8712 if (rem_fltr)
8713 ice_rem_all_chnl_fltrs(pf);
8714
8715 /* remove ntuple filters since queue configuration is being changed */
8716 if (vsi->netdev->features & NETIF_F_NTUPLE) {
8717 struct ice_hw *hw = &pf->hw;
8718
8719 mutex_lock(&hw->fdir_fltr_lock);
8720 ice_fdir_del_all_fltrs(vsi);
8721 mutex_unlock(&hw->fdir_fltr_lock);
8722 }
8723
8724 /* perform cleanup for channels if they exist */
8725 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8726 struct ice_vsi *ch_vsi;
8727
8728 list_del(&ch->list);
8729 ch_vsi = ch->ch_vsi;
8730 if (!ch_vsi) {
8731 kfree(ch);
8732 continue;
8733 }
8734
8735 /* Reset queue contexts */
8736 for (i = 0; i < ch->num_rxq; i++) {
8737 struct ice_tx_ring *tx_ring;
8738 struct ice_rx_ring *rx_ring;
8739
8740 tx_ring = vsi->tx_rings[ch->base_q + i];
8741 rx_ring = vsi->rx_rings[ch->base_q + i];
8742 if (tx_ring) {
8743 tx_ring->ch = NULL;
8744 if (tx_ring->q_vector)
8745 tx_ring->q_vector->ch = NULL;
8746 }
8747 if (rx_ring) {
8748 rx_ring->ch = NULL;
8749 if (rx_ring->q_vector)
8750 rx_ring->q_vector->ch = NULL;
8751 }
8752 }
8753
8754 /* Release FD resources for the channel VSI */
8755 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8756
8757 /* clear the VSI from scheduler tree */
8758 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8759
8760 /* Delete VSI from FW, PF and HW VSI arrays */
8761 ice_vsi_delete(ch->ch_vsi);
8762
8763 /* free the channel */
8764 kfree(ch);
8765 }
8766
8767 /* clear the channel VSI map which is stored in main VSI */
8768 ice_for_each_chnl_tc(i)
8769 vsi->tc_map_vsi[i] = NULL;
8770
8771 /* reset main VSI's all TC information */
8772 vsi->all_enatc = 0;
8773 vsi->all_numtc = 0;
8774}
8775
8776/**
8777 * ice_rebuild_channels - rebuild channel
8778 * @pf: ptr to PF
8779 *
8780 * Recreate channel VSIs and replay filters
8781 */
8782static int ice_rebuild_channels(struct ice_pf *pf)
8783{
8784 struct device *dev = ice_pf_to_dev(pf);
8785 struct ice_vsi *main_vsi;
8786 bool rem_adv_fltr = true;
8787 struct ice_channel *ch;
8788 struct ice_vsi *vsi;
8789 int tc_idx = 1;
8790 int i, err;
8791
8792 main_vsi = ice_get_main_vsi(pf);
8793 if (!main_vsi)
8794 return 0;
8795
8796 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8797 main_vsi->old_numtc == 1)
8798 return 0; /* nothing to be done */
8799
8800 /* reconfigure main VSI based on old value of TC and cached values
8801 * for MQPRIO opts
8802 */
8803 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8804 if (err) {
8805 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8806 main_vsi->old_ena_tc, main_vsi->vsi_num);
8807 return err;
8808 }
8809
8810 /* rebuild ADQ VSIs */
8811 ice_for_each_vsi(pf, i) {
8812 enum ice_vsi_type type;
8813
8814 vsi = pf->vsi[i];
8815 if (!vsi || vsi->type != ICE_VSI_CHNL)
8816 continue;
8817
8818 type = vsi->type;
8819
8820 /* rebuild ADQ VSI */
8821 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
8822 if (err) {
8823 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8824 ice_vsi_type_str(type), vsi->idx, err);
8825 goto cleanup;
8826 }
8827
8828 /* Re-map HW VSI number, using VSI handle that has been
8829 * previously validated in ice_replay_vsi() call above
8830 */
8831 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8832
8833 /* replay filters for the VSI */
8834 err = ice_replay_vsi(&pf->hw, vsi->idx);
8835 if (err) {
8836 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8837 ice_vsi_type_str(type), err, vsi->idx);
8838 rem_adv_fltr = false;
8839 goto cleanup;
8840 }
8841 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8842 ice_vsi_type_str(type), vsi->idx);
8843
8844 /* store ADQ VSI at correct TC index in main VSI's
8845 * map of TC to VSI
8846 */
8847 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8848 }
8849
8850 /* ADQ VSI(s) has been rebuilt successfully, so setup
8851 * channel for main VSI's Tx and Rx rings
8852 */
8853 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8854 struct ice_vsi *ch_vsi;
8855
8856 ch_vsi = ch->ch_vsi;
8857 if (!ch_vsi)
8858 continue;
8859
8860 /* reconfig channel resources */
8861 ice_cfg_chnl_all_res(main_vsi, ch);
8862
8863 /* replay BW rate limit if it is non-zero */
8864 if (!ch->max_tx_rate && !ch->min_tx_rate)
8865 continue;
8866
8867 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8868 ch->min_tx_rate);
8869 if (err)
8870 dev_err(dev, "failed (err:%d) to rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8871 err, ch->max_tx_rate, ch->min_tx_rate,
8872 ch_vsi->vsi_num);
8873 else
8874 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8875 ch->max_tx_rate, ch->min_tx_rate,
8876 ch_vsi->vsi_num);
8877 }
8878
8879 /* reconfig RSS for main VSI */
8880 if (main_vsi->ch_rss_size)
8881 ice_vsi_cfg_rss_lut_key(main_vsi);
8882
8883 return 0;
8884
8885cleanup:
8886 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8887 return err;
8888}
8889
8890/**
8891 * ice_create_q_channels - Add queue channel for the given TCs
8892 * @vsi: VSI to be configured
8893 *
8894 * Configures queue channel mapping to the given TCs
8895 */
8896static int ice_create_q_channels(struct ice_vsi *vsi)
8897{
8898 struct ice_pf *pf = vsi->back;
8899 struct ice_channel *ch;
8900 int ret = 0, i;
8901
8902 ice_for_each_chnl_tc(i) {
8903 if (!(vsi->all_enatc & BIT(i)))
8904 continue;
8905
8906 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8907 if (!ch) {
8908 ret = -ENOMEM;
8909 goto err_free;
8910 }
8911 INIT_LIST_HEAD(&ch->list);
8912 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8913 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8914 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8915 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8916 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8917
8918 /* convert to Kbits/s */
8919 if (ch->max_tx_rate)
8920 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8921 ICE_BW_KBPS_DIVISOR);
8922 if (ch->min_tx_rate)
8923 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8924 ICE_BW_KBPS_DIVISOR);
8925
8926 ret = ice_create_q_channel(vsi, ch);
8927 if (ret) {
8928 dev_err(ice_pf_to_dev(pf),
8929 "failed creating channel TC:%d\n", i);
8930 kfree(ch);
8931 goto err_free;
8932 }
8933 list_add_tail(&ch->list, &vsi->ch_list);
8934 vsi->tc_map_vsi[i] = ch->ch_vsi;
8935 dev_dbg(ice_pf_to_dev(pf),
8936 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8937 }
8938 return 0;
8939
8940err_free:
8941 ice_remove_q_channels(vsi, false);
8942
8943 return ret;
8944}
8945
8946/**
8947 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8948 * @netdev: net device to configure
8949 * @type_data: TC offload data
8950 */
8951static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8952{
8953 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8954 struct ice_netdev_priv *np = netdev_priv(netdev);
8955 struct ice_vsi *vsi = np->vsi;
8956 struct ice_pf *pf = vsi->back;
8957 u16 mode, ena_tc_qdisc = 0;
8958 int cur_txq, cur_rxq;
8959 u8 hw = 0, num_tcf;
8960 struct device *dev;
8961 int ret, i;
8962
8963 dev = ice_pf_to_dev(pf);
8964 num_tcf = mqprio_qopt->qopt.num_tc;
8965 hw = mqprio_qopt->qopt.hw;
8966 mode = mqprio_qopt->mode;
8967 if (!hw) {
8968 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8969 vsi->ch_rss_size = 0;
8970 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8971 goto config_tcf;
8972 }
8973
8974 /* Generate queue region map for number of TCF requested */
8975 for (i = 0; i < num_tcf; i++)
8976 ena_tc_qdisc |= BIT(i);
8977
8978 switch (mode) {
8979 case TC_MQPRIO_MODE_CHANNEL:
8980
8981 if (pf->hw.port_info->is_custom_tx_enabled) {
8982 dev_err(dev, "Custom Tx scheduler feature enabled, can't configure ADQ\n");
8983 return -EBUSY;
8984 }
8985 ice_tear_down_devlink_rate_tree(pf);
8986
8987 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8988 if (ret) {
8989 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8990 ret);
8991 return ret;
8992 }
8993 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8994 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8995 /* don't assume state of hw_tc_offload during driver load
8996 * and set the flag for TC flower filter if hw_tc_offload
8997 * already ON
8998 */
8999 if (vsi->netdev->features & NETIF_F_HW_TC)
9000 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
9001 break;
9002 default:
9003 return -EINVAL;
9004 }
9005
9006config_tcf:
9007
9008 /* Requesting same TCF configuration as already enabled */
9009 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
9010 mode != TC_MQPRIO_MODE_CHANNEL)
9011 return 0;
9012
9013 /* Pause VSI queues */
9014 ice_dis_vsi(vsi, true);
9015
9016 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
9017 ice_remove_q_channels(vsi, true);
9018
9019 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
9020 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
9021 num_online_cpus());
9022 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
9023 num_online_cpus());
9024 } else {
9025 /* logic to rebuild VSI, same like ethtool -L */
9026 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
9027
9028 for (i = 0; i < num_tcf; i++) {
9029 if (!(ena_tc_qdisc & BIT(i)))
9030 continue;
9031
9032 offset = vsi->mqprio_qopt.qopt.offset[i];
9033 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
9034 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
9035 }
9036 vsi->req_txq = offset + qcount_tx;
9037 vsi->req_rxq = offset + qcount_rx;
9038
9039 /* store away original rss_size info, so that it gets reused
9040 * form ice_vsi_rebuild during tc-qdisc delete stage - to
9041 * determine, what should be the rss_sizefor main VSI
9042 */
9043 vsi->orig_rss_size = vsi->rss_size;
9044 }
9045
9046 /* save current values of Tx and Rx queues before calling VSI rebuild
9047 * for fallback option
9048 */
9049 cur_txq = vsi->num_txq;
9050 cur_rxq = vsi->num_rxq;
9051
9052 /* proceed with rebuild main VSI using correct number of queues */
9053 ret = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
9054 if (ret) {
9055 /* fallback to current number of queues */
9056 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
9057 vsi->req_txq = cur_txq;
9058 vsi->req_rxq = cur_rxq;
9059 clear_bit(ICE_RESET_FAILED, pf->state);
9060 if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) {
9061 dev_err(dev, "Rebuild of main VSI failed again\n");
9062 return ret;
9063 }
9064 }
9065
9066 vsi->all_numtc = num_tcf;
9067 vsi->all_enatc = ena_tc_qdisc;
9068 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
9069 if (ret) {
9070 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
9071 vsi->vsi_num);
9072 goto exit;
9073 }
9074
9075 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
9076 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
9077 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
9078
9079 /* set TC0 rate limit if specified */
9080 if (max_tx_rate || min_tx_rate) {
9081 /* convert to Kbits/s */
9082 if (max_tx_rate)
9083 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
9084 if (min_tx_rate)
9085 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
9086
9087 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
9088 if (!ret) {
9089 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
9090 max_tx_rate, min_tx_rate, vsi->vsi_num);
9091 } else {
9092 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
9093 max_tx_rate, min_tx_rate, vsi->vsi_num);
9094 goto exit;
9095 }
9096 }
9097 ret = ice_create_q_channels(vsi);
9098 if (ret) {
9099 netdev_err(netdev, "failed configuring queue channels\n");
9100 goto exit;
9101 } else {
9102 netdev_dbg(netdev, "successfully configured channels\n");
9103 }
9104 }
9105
9106 if (vsi->ch_rss_size)
9107 ice_vsi_cfg_rss_lut_key(vsi);
9108
9109exit:
9110 /* if error, reset the all_numtc and all_enatc */
9111 if (ret) {
9112 vsi->all_numtc = 0;
9113 vsi->all_enatc = 0;
9114 }
9115 /* resume VSI */
9116 ice_ena_vsi(vsi, true);
9117
9118 return ret;
9119}
9120
9121static LIST_HEAD(ice_block_cb_list);
9122
9123static int
9124ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
9125 void *type_data)
9126{
9127 struct ice_netdev_priv *np = netdev_priv(netdev);
9128 struct ice_pf *pf = np->vsi->back;
9129 bool locked = false;
9130 int err;
9131
9132 switch (type) {
9133 case TC_SETUP_BLOCK:
9134 return flow_block_cb_setup_simple(type_data,
9135 &ice_block_cb_list,
9136 ice_setup_tc_block_cb,
9137 np, np, true);
9138 case TC_SETUP_QDISC_MQPRIO:
9139 if (ice_is_eswitch_mode_switchdev(pf)) {
9140 netdev_err(netdev, "TC MQPRIO offload not supported, switchdev is enabled\n");
9141 return -EOPNOTSUPP;
9142 }
9143
9144 if (pf->adev) {
9145 mutex_lock(&pf->adev_mutex);
9146 device_lock(&pf->adev->dev);
9147 locked = true;
9148 if (pf->adev->dev.driver) {
9149 netdev_err(netdev, "Cannot change qdisc when RDMA is active\n");
9150 err = -EBUSY;
9151 goto adev_unlock;
9152 }
9153 }
9154
9155 /* setup traffic classifier for receive side */
9156 mutex_lock(&pf->tc_mutex);
9157 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
9158 mutex_unlock(&pf->tc_mutex);
9159
9160adev_unlock:
9161 if (locked) {
9162 device_unlock(&pf->adev->dev);
9163 mutex_unlock(&pf->adev_mutex);
9164 }
9165 return err;
9166 default:
9167 return -EOPNOTSUPP;
9168 }
9169 return -EOPNOTSUPP;
9170}
9171
9172static struct ice_indr_block_priv *
9173ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
9174 struct net_device *netdev)
9175{
9176 struct ice_indr_block_priv *cb_priv;
9177
9178 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
9179 if (!cb_priv->netdev)
9180 return NULL;
9181 if (cb_priv->netdev == netdev)
9182 return cb_priv;
9183 }
9184 return NULL;
9185}
9186
9187static int
9188ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
9189 void *indr_priv)
9190{
9191 struct ice_indr_block_priv *priv = indr_priv;
9192 struct ice_netdev_priv *np = priv->np;
9193
9194 switch (type) {
9195 case TC_SETUP_CLSFLOWER:
9196 return ice_setup_tc_cls_flower(np, priv->netdev,
9197 (struct flow_cls_offload *)
9198 type_data);
9199 default:
9200 return -EOPNOTSUPP;
9201 }
9202}
9203
9204static int
9205ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
9206 struct ice_netdev_priv *np,
9207 struct flow_block_offload *f, void *data,
9208 void (*cleanup)(struct flow_block_cb *block_cb))
9209{
9210 struct ice_indr_block_priv *indr_priv;
9211 struct flow_block_cb *block_cb;
9212
9213 if (!ice_is_tunnel_supported(netdev) &&
9214 !(is_vlan_dev(netdev) &&
9215 vlan_dev_real_dev(netdev) == np->vsi->netdev))
9216 return -EOPNOTSUPP;
9217
9218 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9219 return -EOPNOTSUPP;
9220
9221 switch (f->command) {
9222 case FLOW_BLOCK_BIND:
9223 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9224 if (indr_priv)
9225 return -EEXIST;
9226
9227 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
9228 if (!indr_priv)
9229 return -ENOMEM;
9230
9231 indr_priv->netdev = netdev;
9232 indr_priv->np = np;
9233 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
9234
9235 block_cb =
9236 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
9237 indr_priv, indr_priv,
9238 ice_rep_indr_tc_block_unbind,
9239 f, netdev, sch, data, np,
9240 cleanup);
9241
9242 if (IS_ERR(block_cb)) {
9243 list_del(&indr_priv->list);
9244 kfree(indr_priv);
9245 return PTR_ERR(block_cb);
9246 }
9247 flow_block_cb_add(block_cb, f);
9248 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
9249 break;
9250 case FLOW_BLOCK_UNBIND:
9251 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9252 if (!indr_priv)
9253 return -ENOENT;
9254
9255 block_cb = flow_block_cb_lookup(f->block,
9256 ice_indr_setup_block_cb,
9257 indr_priv);
9258 if (!block_cb)
9259 return -ENOENT;
9260
9261 flow_indr_block_cb_remove(block_cb, f);
9262
9263 list_del(&block_cb->driver_list);
9264 break;
9265 default:
9266 return -EOPNOTSUPP;
9267 }
9268 return 0;
9269}
9270
9271static int
9272ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
9273 void *cb_priv, enum tc_setup_type type, void *type_data,
9274 void *data,
9275 void (*cleanup)(struct flow_block_cb *block_cb))
9276{
9277 switch (type) {
9278 case TC_SETUP_BLOCK:
9279 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
9280 data, cleanup);
9281
9282 default:
9283 return -EOPNOTSUPP;
9284 }
9285}
9286
9287/**
9288 * ice_open - Called when a network interface becomes active
9289 * @netdev: network interface device structure
9290 *
9291 * The open entry point is called when a network interface is made
9292 * active by the system (IFF_UP). At this point all resources needed
9293 * for transmit and receive operations are allocated, the interrupt
9294 * handler is registered with the OS, the netdev watchdog is enabled,
9295 * and the stack is notified that the interface is ready.
9296 *
9297 * Returns 0 on success, negative value on failure
9298 */
9299int ice_open(struct net_device *netdev)
9300{
9301 struct ice_netdev_priv *np = netdev_priv(netdev);
9302 struct ice_pf *pf = np->vsi->back;
9303
9304 if (ice_is_reset_in_progress(pf->state)) {
9305 netdev_err(netdev, "can't open net device while reset is in progress");
9306 return -EBUSY;
9307 }
9308
9309 return ice_open_internal(netdev);
9310}
9311
9312/**
9313 * ice_open_internal - Called when a network interface becomes active
9314 * @netdev: network interface device structure
9315 *
9316 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
9317 * handling routine
9318 *
9319 * Returns 0 on success, negative value on failure
9320 */
9321int ice_open_internal(struct net_device *netdev)
9322{
9323 struct ice_netdev_priv *np = netdev_priv(netdev);
9324 struct ice_vsi *vsi = np->vsi;
9325 struct ice_pf *pf = vsi->back;
9326 struct ice_port_info *pi;
9327 int err;
9328
9329 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
9330 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
9331 return -EIO;
9332 }
9333
9334 netif_carrier_off(netdev);
9335
9336 pi = vsi->port_info;
9337 err = ice_update_link_info(pi);
9338 if (err) {
9339 netdev_err(netdev, "Failed to get link info, error %d\n", err);
9340 return err;
9341 }
9342
9343 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
9344
9345 /* Set PHY if there is media, otherwise, turn off PHY */
9346 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
9347 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9348 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
9349 err = ice_init_phy_user_cfg(pi);
9350 if (err) {
9351 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
9352 err);
9353 return err;
9354 }
9355 }
9356
9357 err = ice_configure_phy(vsi);
9358 if (err) {
9359 netdev_err(netdev, "Failed to set physical link up, error %d\n",
9360 err);
9361 return err;
9362 }
9363 } else {
9364 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9365 ice_set_link(vsi, false);
9366 }
9367
9368 err = ice_vsi_open(vsi);
9369 if (err)
9370 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
9371 vsi->vsi_num, vsi->vsw->sw_id);
9372
9373 /* Update existing tunnels information */
9374 udp_tunnel_get_rx_info(netdev);
9375
9376 return err;
9377}
9378
9379/**
9380 * ice_stop - Disables a network interface
9381 * @netdev: network interface device structure
9382 *
9383 * The stop entry point is called when an interface is de-activated by the OS,
9384 * and the netdevice enters the DOWN state. The hardware is still under the
9385 * driver's control, but the netdev interface is disabled.
9386 *
9387 * Returns success only - not allowed to fail
9388 */
9389int ice_stop(struct net_device *netdev)
9390{
9391 struct ice_netdev_priv *np = netdev_priv(netdev);
9392 struct ice_vsi *vsi = np->vsi;
9393 struct ice_pf *pf = vsi->back;
9394
9395 if (ice_is_reset_in_progress(pf->state)) {
9396 netdev_err(netdev, "can't stop net device while reset is in progress");
9397 return -EBUSY;
9398 }
9399
9400 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
9401 int link_err = ice_force_phys_link_state(vsi, false);
9402
9403 if (link_err) {
9404 if (link_err == -ENOMEDIUM)
9405 netdev_info(vsi->netdev, "Skipping link reconfig - no media attached, VSI %d\n",
9406 vsi->vsi_num);
9407 else
9408 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
9409 vsi->vsi_num, link_err);
9410
9411 ice_vsi_close(vsi);
9412 return -EIO;
9413 }
9414 }
9415
9416 ice_vsi_close(vsi);
9417
9418 return 0;
9419}
9420
9421/**
9422 * ice_features_check - Validate encapsulated packet conforms to limits
9423 * @skb: skb buffer
9424 * @netdev: This port's netdev
9425 * @features: Offload features that the stack believes apply
9426 */
9427static netdev_features_t
9428ice_features_check(struct sk_buff *skb,
9429 struct net_device __always_unused *netdev,
9430 netdev_features_t features)
9431{
9432 bool gso = skb_is_gso(skb);
9433 size_t len;
9434
9435 /* No point in doing any of this if neither checksum nor GSO are
9436 * being requested for this frame. We can rule out both by just
9437 * checking for CHECKSUM_PARTIAL
9438 */
9439 if (skb->ip_summed != CHECKSUM_PARTIAL)
9440 return features;
9441
9442 /* We cannot support GSO if the MSS is going to be less than
9443 * 64 bytes. If it is then we need to drop support for GSO.
9444 */
9445 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
9446 features &= ~NETIF_F_GSO_MASK;
9447
9448 len = skb_network_offset(skb);
9449 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
9450 goto out_rm_features;
9451
9452 len = skb_network_header_len(skb);
9453 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9454 goto out_rm_features;
9455
9456 if (skb->encapsulation) {
9457 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
9458 * the case of IPIP frames, the transport header pointer is
9459 * after the inner header! So check to make sure that this
9460 * is a GRE or UDP_TUNNEL frame before doing that math.
9461 */
9462 if (gso && (skb_shinfo(skb)->gso_type &
9463 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
9464 len = skb_inner_network_header(skb) -
9465 skb_transport_header(skb);
9466 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
9467 goto out_rm_features;
9468 }
9469
9470 len = skb_inner_network_header_len(skb);
9471 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9472 goto out_rm_features;
9473 }
9474
9475 return features;
9476out_rm_features:
9477 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9478}
9479
9480static const struct net_device_ops ice_netdev_safe_mode_ops = {
9481 .ndo_open = ice_open,
9482 .ndo_stop = ice_stop,
9483 .ndo_start_xmit = ice_start_xmit,
9484 .ndo_set_mac_address = ice_set_mac_address,
9485 .ndo_validate_addr = eth_validate_addr,
9486 .ndo_change_mtu = ice_change_mtu,
9487 .ndo_get_stats64 = ice_get_stats64,
9488 .ndo_tx_timeout = ice_tx_timeout,
9489 .ndo_bpf = ice_xdp_safe_mode,
9490};
9491
9492static const struct net_device_ops ice_netdev_ops = {
9493 .ndo_open = ice_open,
9494 .ndo_stop = ice_stop,
9495 .ndo_start_xmit = ice_start_xmit,
9496 .ndo_select_queue = ice_select_queue,
9497 .ndo_features_check = ice_features_check,
9498 .ndo_fix_features = ice_fix_features,
9499 .ndo_set_rx_mode = ice_set_rx_mode,
9500 .ndo_set_mac_address = ice_set_mac_address,
9501 .ndo_validate_addr = eth_validate_addr,
9502 .ndo_change_mtu = ice_change_mtu,
9503 .ndo_get_stats64 = ice_get_stats64,
9504 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
9505 .ndo_eth_ioctl = ice_eth_ioctl,
9506 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
9507 .ndo_set_vf_mac = ice_set_vf_mac,
9508 .ndo_get_vf_config = ice_get_vf_cfg,
9509 .ndo_set_vf_trust = ice_set_vf_trust,
9510 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
9511 .ndo_set_vf_link_state = ice_set_vf_link_state,
9512 .ndo_get_vf_stats = ice_get_vf_stats,
9513 .ndo_set_vf_rate = ice_set_vf_bw,
9514 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
9515 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
9516 .ndo_setup_tc = ice_setup_tc,
9517 .ndo_set_features = ice_set_features,
9518 .ndo_bridge_getlink = ice_bridge_getlink,
9519 .ndo_bridge_setlink = ice_bridge_setlink,
9520 .ndo_fdb_add = ice_fdb_add,
9521 .ndo_fdb_del = ice_fdb_del,
9522#ifdef CONFIG_RFS_ACCEL
9523 .ndo_rx_flow_steer = ice_rx_flow_steer,
9524#endif
9525 .ndo_tx_timeout = ice_tx_timeout,
9526 .ndo_bpf = ice_xdp,
9527 .ndo_xdp_xmit = ice_xdp_xmit,
9528 .ndo_xsk_wakeup = ice_xsk_wakeup,
9529};