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, reset_type);
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 /* if mac_type is not generic, sideband is not supported
1653 * and there's nothing to do here
1654 */
1655 if (!ice_is_generic_mac(hw)) {
1656 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1657 return;
1658 }
1659
1660 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1661 return;
1662
1663 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1664 return;
1665
1666 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1667
1668 if (ice_ctrlq_pending(hw, &hw->sbq))
1669 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1670
1671 ice_flush(hw);
1672}
1673
1674/**
1675 * ice_service_task_schedule - schedule the service task to wake up
1676 * @pf: board private structure
1677 *
1678 * If not already scheduled, this puts the task into the work queue.
1679 */
1680void ice_service_task_schedule(struct ice_pf *pf)
1681{
1682 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1683 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1684 !test_bit(ICE_NEEDS_RESTART, pf->state))
1685 queue_work(ice_wq, &pf->serv_task);
1686}
1687
1688/**
1689 * ice_service_task_complete - finish up the service task
1690 * @pf: board private structure
1691 */
1692static void ice_service_task_complete(struct ice_pf *pf)
1693{
1694 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1695
1696 /* force memory (pf->state) to sync before next service task */
1697 smp_mb__before_atomic();
1698 clear_bit(ICE_SERVICE_SCHED, pf->state);
1699}
1700
1701/**
1702 * ice_service_task_stop - stop service task and cancel works
1703 * @pf: board private structure
1704 *
1705 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1706 * 1 otherwise.
1707 */
1708static int ice_service_task_stop(struct ice_pf *pf)
1709{
1710 int ret;
1711
1712 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1713
1714 if (pf->serv_tmr.function)
1715 del_timer_sync(&pf->serv_tmr);
1716 if (pf->serv_task.func)
1717 cancel_work_sync(&pf->serv_task);
1718
1719 clear_bit(ICE_SERVICE_SCHED, pf->state);
1720 return ret;
1721}
1722
1723/**
1724 * ice_service_task_restart - restart service task and schedule works
1725 * @pf: board private structure
1726 *
1727 * This function is needed for suspend and resume works (e.g WoL scenario)
1728 */
1729static void ice_service_task_restart(struct ice_pf *pf)
1730{
1731 clear_bit(ICE_SERVICE_DIS, pf->state);
1732 ice_service_task_schedule(pf);
1733}
1734
1735/**
1736 * ice_service_timer - timer callback to schedule service task
1737 * @t: pointer to timer_list
1738 */
1739static void ice_service_timer(struct timer_list *t)
1740{
1741 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1742
1743 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1744 ice_service_task_schedule(pf);
1745}
1746
1747/**
1748 * ice_handle_mdd_event - handle malicious driver detect event
1749 * @pf: pointer to the PF structure
1750 *
1751 * Called from service task. OICR interrupt handler indicates MDD event.
1752 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1753 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1754 * disable the queue, the PF can be configured to reset the VF using ethtool
1755 * private flag mdd-auto-reset-vf.
1756 */
1757static void ice_handle_mdd_event(struct ice_pf *pf)
1758{
1759 struct device *dev = ice_pf_to_dev(pf);
1760 struct ice_hw *hw = &pf->hw;
1761 struct ice_vf *vf;
1762 unsigned int bkt;
1763 u32 reg;
1764
1765 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1766 /* Since the VF MDD event logging is rate limited, check if
1767 * there are pending MDD events.
1768 */
1769 ice_print_vfs_mdd_events(pf);
1770 return;
1771 }
1772
1773 /* find what triggered an MDD event */
1774 reg = rd32(hw, GL_MDET_TX_PQM);
1775 if (reg & GL_MDET_TX_PQM_VALID_M) {
1776 u8 pf_num = FIELD_GET(GL_MDET_TX_PQM_PF_NUM_M, reg);
1777 u16 vf_num = FIELD_GET(GL_MDET_TX_PQM_VF_NUM_M, reg);
1778 u8 event = FIELD_GET(GL_MDET_TX_PQM_MAL_TYPE_M, reg);
1779 u16 queue = FIELD_GET(GL_MDET_TX_PQM_QNUM_M, reg);
1780
1781 if (netif_msg_tx_err(pf))
1782 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1783 event, queue, pf_num, vf_num);
1784 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1785 }
1786
1787 reg = rd32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw));
1788 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1789 u8 pf_num = FIELD_GET(GL_MDET_TX_TCLAN_PF_NUM_M, reg);
1790 u16 vf_num = FIELD_GET(GL_MDET_TX_TCLAN_VF_NUM_M, reg);
1791 u8 event = FIELD_GET(GL_MDET_TX_TCLAN_MAL_TYPE_M, reg);
1792 u16 queue = FIELD_GET(GL_MDET_TX_TCLAN_QNUM_M, reg);
1793
1794 if (netif_msg_tx_err(pf))
1795 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1796 event, queue, pf_num, vf_num);
1797 wr32(hw, GL_MDET_TX_TCLAN_BY_MAC(hw), U32_MAX);
1798 }
1799
1800 reg = rd32(hw, GL_MDET_RX);
1801 if (reg & GL_MDET_RX_VALID_M) {
1802 u8 pf_num = FIELD_GET(GL_MDET_RX_PF_NUM_M, reg);
1803 u16 vf_num = FIELD_GET(GL_MDET_RX_VF_NUM_M, reg);
1804 u8 event = FIELD_GET(GL_MDET_RX_MAL_TYPE_M, reg);
1805 u16 queue = FIELD_GET(GL_MDET_RX_QNUM_M, reg);
1806
1807 if (netif_msg_rx_err(pf))
1808 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1809 event, queue, pf_num, vf_num);
1810 wr32(hw, GL_MDET_RX, 0xffffffff);
1811 }
1812
1813 /* check to see if this PF caused an MDD event */
1814 reg = rd32(hw, PF_MDET_TX_PQM);
1815 if (reg & PF_MDET_TX_PQM_VALID_M) {
1816 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1817 if (netif_msg_tx_err(pf))
1818 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1819 }
1820
1821 reg = rd32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw));
1822 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1823 wr32(hw, PF_MDET_TX_TCLAN_BY_MAC(hw), 0xffff);
1824 if (netif_msg_tx_err(pf))
1825 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1826 }
1827
1828 reg = rd32(hw, PF_MDET_RX);
1829 if (reg & PF_MDET_RX_VALID_M) {
1830 wr32(hw, PF_MDET_RX, 0xFFFF);
1831 if (netif_msg_rx_err(pf))
1832 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1833 }
1834
1835 /* Check to see if one of the VFs caused an MDD event, and then
1836 * increment counters and set print pending
1837 */
1838 mutex_lock(&pf->vfs.table_lock);
1839 ice_for_each_vf(pf, bkt, vf) {
1840 reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1841 if (reg & VP_MDET_TX_PQM_VALID_M) {
1842 wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1843 vf->mdd_tx_events.count++;
1844 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1845 if (netif_msg_tx_err(pf))
1846 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1847 vf->vf_id);
1848 }
1849
1850 reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1851 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1852 wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1853 vf->mdd_tx_events.count++;
1854 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1855 if (netif_msg_tx_err(pf))
1856 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1857 vf->vf_id);
1858 }
1859
1860 reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1861 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1862 wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1863 vf->mdd_tx_events.count++;
1864 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1865 if (netif_msg_tx_err(pf))
1866 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1867 vf->vf_id);
1868 }
1869
1870 reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1871 if (reg & VP_MDET_RX_VALID_M) {
1872 wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1873 vf->mdd_rx_events.count++;
1874 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1875 if (netif_msg_rx_err(pf))
1876 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1877 vf->vf_id);
1878
1879 /* Since the queue is disabled on VF Rx MDD events, the
1880 * PF can be configured to reset the VF through ethtool
1881 * private flag mdd-auto-reset-vf.
1882 */
1883 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1884 /* VF MDD event counters will be cleared by
1885 * reset, so print the event prior to reset.
1886 */
1887 ice_print_vf_rx_mdd_event(vf);
1888 ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1889 }
1890 }
1891 }
1892 mutex_unlock(&pf->vfs.table_lock);
1893
1894 ice_print_vfs_mdd_events(pf);
1895}
1896
1897/**
1898 * ice_force_phys_link_state - Force the physical link state
1899 * @vsi: VSI to force the physical link state to up/down
1900 * @link_up: true/false indicates to set the physical link to up/down
1901 *
1902 * Force the physical link state by getting the current PHY capabilities from
1903 * hardware and setting the PHY config based on the determined capabilities. If
1904 * link changes a link event will be triggered because both the Enable Automatic
1905 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1906 *
1907 * Returns 0 on success, negative on failure
1908 */
1909static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1910{
1911 struct ice_aqc_get_phy_caps_data *pcaps;
1912 struct ice_aqc_set_phy_cfg_data *cfg;
1913 struct ice_port_info *pi;
1914 struct device *dev;
1915 int retcode;
1916
1917 if (!vsi || !vsi->port_info || !vsi->back)
1918 return -EINVAL;
1919 if (vsi->type != ICE_VSI_PF)
1920 return 0;
1921
1922 dev = ice_pf_to_dev(vsi->back);
1923
1924 pi = vsi->port_info;
1925
1926 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1927 if (!pcaps)
1928 return -ENOMEM;
1929
1930 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1931 NULL);
1932 if (retcode) {
1933 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1934 vsi->vsi_num, retcode);
1935 retcode = -EIO;
1936 goto out;
1937 }
1938
1939 /* No change in link */
1940 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1941 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1942 goto out;
1943
1944 /* Use the current user PHY configuration. The current user PHY
1945 * configuration is initialized during probe from PHY capabilities
1946 * software mode, and updated on set PHY configuration.
1947 */
1948 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1949 if (!cfg) {
1950 retcode = -ENOMEM;
1951 goto out;
1952 }
1953
1954 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1955 if (link_up)
1956 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1957 else
1958 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1959
1960 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1961 if (retcode) {
1962 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1963 vsi->vsi_num, retcode);
1964 retcode = -EIO;
1965 }
1966
1967 kfree(cfg);
1968out:
1969 kfree(pcaps);
1970 return retcode;
1971}
1972
1973/**
1974 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1975 * @pi: port info structure
1976 *
1977 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1978 */
1979static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1980{
1981 struct ice_aqc_get_phy_caps_data *pcaps;
1982 struct ice_pf *pf = pi->hw->back;
1983 int err;
1984
1985 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1986 if (!pcaps)
1987 return -ENOMEM;
1988
1989 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1990 pcaps, NULL);
1991
1992 if (err) {
1993 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1994 goto out;
1995 }
1996
1997 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1998 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1999
2000out:
2001 kfree(pcaps);
2002 return err;
2003}
2004
2005/**
2006 * ice_init_link_dflt_override - Initialize link default override
2007 * @pi: port info structure
2008 *
2009 * Initialize link default override and PHY total port shutdown during probe
2010 */
2011static void ice_init_link_dflt_override(struct ice_port_info *pi)
2012{
2013 struct ice_link_default_override_tlv *ldo;
2014 struct ice_pf *pf = pi->hw->back;
2015
2016 ldo = &pf->link_dflt_override;
2017 if (ice_get_link_default_override(ldo, pi))
2018 return;
2019
2020 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
2021 return;
2022
2023 /* Enable Total Port Shutdown (override/replace link-down-on-close
2024 * ethtool private flag) for ports with Port Disable bit set.
2025 */
2026 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
2027 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
2028}
2029
2030/**
2031 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
2032 * @pi: port info structure
2033 *
2034 * If default override is enabled, initialize the user PHY cfg speed and FEC
2035 * settings using the default override mask from the NVM.
2036 *
2037 * The PHY should only be configured with the default override settings the
2038 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
2039 * is used to indicate that the user PHY cfg default override is initialized
2040 * and the PHY has not been configured with the default override settings. The
2041 * state is set here, and cleared in ice_configure_phy the first time the PHY is
2042 * configured.
2043 *
2044 * This function should be called only if the FW doesn't support default
2045 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
2046 */
2047static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
2048{
2049 struct ice_link_default_override_tlv *ldo;
2050 struct ice_aqc_set_phy_cfg_data *cfg;
2051 struct ice_phy_info *phy = &pi->phy;
2052 struct ice_pf *pf = pi->hw->back;
2053
2054 ldo = &pf->link_dflt_override;
2055
2056 /* If link default override is enabled, use to mask NVM PHY capabilities
2057 * for speed and FEC default configuration.
2058 */
2059 cfg = &phy->curr_user_phy_cfg;
2060
2061 if (ldo->phy_type_low || ldo->phy_type_high) {
2062 cfg->phy_type_low = pf->nvm_phy_type_lo &
2063 cpu_to_le64(ldo->phy_type_low);
2064 cfg->phy_type_high = pf->nvm_phy_type_hi &
2065 cpu_to_le64(ldo->phy_type_high);
2066 }
2067 cfg->link_fec_opt = ldo->fec_options;
2068 phy->curr_user_fec_req = ICE_FEC_AUTO;
2069
2070 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2071}
2072
2073/**
2074 * ice_init_phy_user_cfg - Initialize the PHY user configuration
2075 * @pi: port info structure
2076 *
2077 * Initialize the current user PHY configuration, speed, FEC, and FC requested
2078 * mode to default. The PHY defaults are from get PHY capabilities topology
2079 * with media so call when media is first available. An error is returned if
2080 * called when media is not available. The PHY initialization completed state is
2081 * set here.
2082 *
2083 * These configurations are used when setting PHY
2084 * configuration. The user PHY configuration is updated on set PHY
2085 * configuration. Returns 0 on success, negative on failure
2086 */
2087static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2088{
2089 struct ice_aqc_get_phy_caps_data *pcaps;
2090 struct ice_phy_info *phy = &pi->phy;
2091 struct ice_pf *pf = pi->hw->back;
2092 int err;
2093
2094 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2095 return -EIO;
2096
2097 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2098 if (!pcaps)
2099 return -ENOMEM;
2100
2101 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2102 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2103 pcaps, NULL);
2104 else
2105 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2106 pcaps, NULL);
2107 if (err) {
2108 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2109 goto err_out;
2110 }
2111
2112 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2113
2114 /* check if lenient mode is supported and enabled */
2115 if (ice_fw_supports_link_override(pi->hw) &&
2116 !(pcaps->module_compliance_enforcement &
2117 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2118 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2119
2120 /* if the FW supports default PHY configuration mode, then the driver
2121 * does not have to apply link override settings. If not,
2122 * initialize user PHY configuration with link override values
2123 */
2124 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2125 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2126 ice_init_phy_cfg_dflt_override(pi);
2127 goto out;
2128 }
2129 }
2130
2131 /* if link default override is not enabled, set user flow control and
2132 * FEC settings based on what get_phy_caps returned
2133 */
2134 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2135 pcaps->link_fec_options);
2136 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2137
2138out:
2139 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2140 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2141err_out:
2142 kfree(pcaps);
2143 return err;
2144}
2145
2146/**
2147 * ice_configure_phy - configure PHY
2148 * @vsi: VSI of PHY
2149 *
2150 * Set the PHY configuration. If the current PHY configuration is the same as
2151 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2152 * configure the based get PHY capabilities for topology with media.
2153 */
2154static int ice_configure_phy(struct ice_vsi *vsi)
2155{
2156 struct device *dev = ice_pf_to_dev(vsi->back);
2157 struct ice_port_info *pi = vsi->port_info;
2158 struct ice_aqc_get_phy_caps_data *pcaps;
2159 struct ice_aqc_set_phy_cfg_data *cfg;
2160 struct ice_phy_info *phy = &pi->phy;
2161 struct ice_pf *pf = vsi->back;
2162 int err;
2163
2164 /* Ensure we have media as we cannot configure a medialess port */
2165 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2166 return -ENOMEDIUM;
2167
2168 ice_print_topo_conflict(vsi);
2169
2170 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2171 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2172 return -EPERM;
2173
2174 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2175 return ice_force_phys_link_state(vsi, true);
2176
2177 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2178 if (!pcaps)
2179 return -ENOMEM;
2180
2181 /* Get current PHY config */
2182 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2183 NULL);
2184 if (err) {
2185 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2186 vsi->vsi_num, err);
2187 goto done;
2188 }
2189
2190 /* If PHY enable link is configured and configuration has not changed,
2191 * there's nothing to do
2192 */
2193 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2194 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2195 goto done;
2196
2197 /* Use PHY topology as baseline for configuration */
2198 memset(pcaps, 0, sizeof(*pcaps));
2199 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2200 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2201 pcaps, NULL);
2202 else
2203 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2204 pcaps, NULL);
2205 if (err) {
2206 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2207 vsi->vsi_num, err);
2208 goto done;
2209 }
2210
2211 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2212 if (!cfg) {
2213 err = -ENOMEM;
2214 goto done;
2215 }
2216
2217 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2218
2219 /* Speed - If default override pending, use curr_user_phy_cfg set in
2220 * ice_init_phy_user_cfg_ldo.
2221 */
2222 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2223 vsi->back->state)) {
2224 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2225 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2226 } else {
2227 u64 phy_low = 0, phy_high = 0;
2228
2229 ice_update_phy_type(&phy_low, &phy_high,
2230 pi->phy.curr_user_speed_req);
2231 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2232 cfg->phy_type_high = pcaps->phy_type_high &
2233 cpu_to_le64(phy_high);
2234 }
2235
2236 /* Can't provide what was requested; use PHY capabilities */
2237 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2238 cfg->phy_type_low = pcaps->phy_type_low;
2239 cfg->phy_type_high = pcaps->phy_type_high;
2240 }
2241
2242 /* FEC */
2243 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2244
2245 /* Can't provide what was requested; use PHY capabilities */
2246 if (cfg->link_fec_opt !=
2247 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2248 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2249 cfg->link_fec_opt = pcaps->link_fec_options;
2250 }
2251
2252 /* Flow Control - always supported; no need to check against
2253 * capabilities
2254 */
2255 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2256
2257 /* Enable link and link update */
2258 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2259
2260 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2261 if (err)
2262 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2263 vsi->vsi_num, err);
2264
2265 kfree(cfg);
2266done:
2267 kfree(pcaps);
2268 return err;
2269}
2270
2271/**
2272 * ice_check_media_subtask - Check for media
2273 * @pf: pointer to PF struct
2274 *
2275 * If media is available, then initialize PHY user configuration if it is not
2276 * been, and configure the PHY if the interface is up.
2277 */
2278static void ice_check_media_subtask(struct ice_pf *pf)
2279{
2280 struct ice_port_info *pi;
2281 struct ice_vsi *vsi;
2282 int err;
2283
2284 /* No need to check for media if it's already present */
2285 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2286 return;
2287
2288 vsi = ice_get_main_vsi(pf);
2289 if (!vsi)
2290 return;
2291
2292 /* Refresh link info and check if media is present */
2293 pi = vsi->port_info;
2294 err = ice_update_link_info(pi);
2295 if (err)
2296 return;
2297
2298 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2299
2300 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2301 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2302 ice_init_phy_user_cfg(pi);
2303
2304 /* PHY settings are reset on media insertion, reconfigure
2305 * PHY to preserve settings.
2306 */
2307 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2308 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2309 return;
2310
2311 err = ice_configure_phy(vsi);
2312 if (!err)
2313 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2314
2315 /* A Link Status Event will be generated; the event handler
2316 * will complete bringing the interface up
2317 */
2318 }
2319}
2320
2321/**
2322 * ice_service_task - manage and run subtasks
2323 * @work: pointer to work_struct contained by the PF struct
2324 */
2325static void ice_service_task(struct work_struct *work)
2326{
2327 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2328 unsigned long start_time = jiffies;
2329
2330 /* subtasks */
2331
2332 /* process reset requests first */
2333 ice_reset_subtask(pf);
2334
2335 /* bail if a reset/recovery cycle is pending or rebuild failed */
2336 if (ice_is_reset_in_progress(pf->state) ||
2337 test_bit(ICE_SUSPENDED, pf->state) ||
2338 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2339 ice_service_task_complete(pf);
2340 return;
2341 }
2342
2343 if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2344 struct iidc_event *event;
2345
2346 event = kzalloc(sizeof(*event), GFP_KERNEL);
2347 if (event) {
2348 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2349 /* report the entire OICR value to AUX driver */
2350 swap(event->reg, pf->oicr_err_reg);
2351 ice_send_event_to_aux(pf, event);
2352 kfree(event);
2353 }
2354 }
2355
2356 /* unplug aux dev per request, if an unplug request came in
2357 * while processing a plug request, this will handle it
2358 */
2359 if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags))
2360 ice_unplug_aux_dev(pf);
2361
2362 /* Plug aux device per request */
2363 if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2364 ice_plug_aux_dev(pf);
2365
2366 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2367 struct iidc_event *event;
2368
2369 event = kzalloc(sizeof(*event), GFP_KERNEL);
2370 if (event) {
2371 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2372 ice_send_event_to_aux(pf, event);
2373 kfree(event);
2374 }
2375 }
2376
2377 ice_clean_adminq_subtask(pf);
2378 ice_check_media_subtask(pf);
2379 ice_check_for_hang_subtask(pf);
2380 ice_sync_fltr_subtask(pf);
2381 ice_handle_mdd_event(pf);
2382 ice_watchdog_subtask(pf);
2383
2384 if (ice_is_safe_mode(pf)) {
2385 ice_service_task_complete(pf);
2386 return;
2387 }
2388
2389 ice_process_vflr_event(pf);
2390 ice_clean_mailboxq_subtask(pf);
2391 ice_clean_sbq_subtask(pf);
2392 ice_sync_arfs_fltrs(pf);
2393 ice_flush_fdir_ctx(pf);
2394
2395 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2396 ice_service_task_complete(pf);
2397
2398 /* If the tasks have taken longer than one service timer period
2399 * or there is more work to be done, reset the service timer to
2400 * schedule the service task now.
2401 */
2402 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2403 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2404 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2405 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2406 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2407 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2408 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2409 mod_timer(&pf->serv_tmr, jiffies);
2410}
2411
2412/**
2413 * ice_set_ctrlq_len - helper function to set controlq length
2414 * @hw: pointer to the HW instance
2415 */
2416static void ice_set_ctrlq_len(struct ice_hw *hw)
2417{
2418 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2419 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2420 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2421 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2422 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2423 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2424 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2425 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2426 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2427 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2428 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2429 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2430}
2431
2432/**
2433 * ice_schedule_reset - schedule a reset
2434 * @pf: board private structure
2435 * @reset: reset being requested
2436 */
2437int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2438{
2439 struct device *dev = ice_pf_to_dev(pf);
2440
2441 /* bail out if earlier reset has failed */
2442 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2443 dev_dbg(dev, "earlier reset has failed\n");
2444 return -EIO;
2445 }
2446 /* bail if reset/recovery already in progress */
2447 if (ice_is_reset_in_progress(pf->state)) {
2448 dev_dbg(dev, "Reset already in progress\n");
2449 return -EBUSY;
2450 }
2451
2452 switch (reset) {
2453 case ICE_RESET_PFR:
2454 set_bit(ICE_PFR_REQ, pf->state);
2455 break;
2456 case ICE_RESET_CORER:
2457 set_bit(ICE_CORER_REQ, pf->state);
2458 break;
2459 case ICE_RESET_GLOBR:
2460 set_bit(ICE_GLOBR_REQ, pf->state);
2461 break;
2462 default:
2463 return -EINVAL;
2464 }
2465
2466 ice_service_task_schedule(pf);
2467 return 0;
2468}
2469
2470/**
2471 * ice_irq_affinity_notify - Callback for affinity changes
2472 * @notify: context as to what irq was changed
2473 * @mask: the new affinity mask
2474 *
2475 * This is a callback function used by the irq_set_affinity_notifier function
2476 * so that we may register to receive changes to the irq affinity masks.
2477 */
2478static void
2479ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2480 const cpumask_t *mask)
2481{
2482 struct ice_q_vector *q_vector =
2483 container_of(notify, struct ice_q_vector, affinity_notify);
2484
2485 cpumask_copy(&q_vector->affinity_mask, mask);
2486}
2487
2488/**
2489 * ice_irq_affinity_release - Callback for affinity notifier release
2490 * @ref: internal core kernel usage
2491 *
2492 * This is a callback function used by the irq_set_affinity_notifier function
2493 * to inform the current notification subscriber that they will no longer
2494 * receive notifications.
2495 */
2496static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2497
2498/**
2499 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2500 * @vsi: the VSI being configured
2501 */
2502static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2503{
2504 struct ice_hw *hw = &vsi->back->hw;
2505 int i;
2506
2507 ice_for_each_q_vector(vsi, i)
2508 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2509
2510 ice_flush(hw);
2511 return 0;
2512}
2513
2514/**
2515 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2516 * @vsi: the VSI being configured
2517 * @basename: name for the vector
2518 */
2519static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2520{
2521 int q_vectors = vsi->num_q_vectors;
2522 struct ice_pf *pf = vsi->back;
2523 struct device *dev;
2524 int rx_int_idx = 0;
2525 int tx_int_idx = 0;
2526 int vector, err;
2527 int irq_num;
2528
2529 dev = ice_pf_to_dev(pf);
2530 for (vector = 0; vector < q_vectors; vector++) {
2531 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2532
2533 irq_num = q_vector->irq.virq;
2534
2535 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2536 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2537 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2538 tx_int_idx++;
2539 } else if (q_vector->rx.rx_ring) {
2540 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2541 "%s-%s-%d", basename, "rx", rx_int_idx++);
2542 } else if (q_vector->tx.tx_ring) {
2543 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2544 "%s-%s-%d", basename, "tx", tx_int_idx++);
2545 } else {
2546 /* skip this unused q_vector */
2547 continue;
2548 }
2549 if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2550 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2551 IRQF_SHARED, q_vector->name,
2552 q_vector);
2553 else
2554 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2555 0, q_vector->name, q_vector);
2556 if (err) {
2557 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2558 err);
2559 goto free_q_irqs;
2560 }
2561
2562 /* register for affinity change notifications */
2563 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2564 struct irq_affinity_notify *affinity_notify;
2565
2566 affinity_notify = &q_vector->affinity_notify;
2567 affinity_notify->notify = ice_irq_affinity_notify;
2568 affinity_notify->release = ice_irq_affinity_release;
2569 irq_set_affinity_notifier(irq_num, affinity_notify);
2570 }
2571
2572 /* assign the mask for this irq */
2573 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2574 }
2575
2576 err = ice_set_cpu_rx_rmap(vsi);
2577 if (err) {
2578 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2579 vsi->vsi_num, ERR_PTR(err));
2580 goto free_q_irqs;
2581 }
2582
2583 vsi->irqs_ready = true;
2584 return 0;
2585
2586free_q_irqs:
2587 while (vector--) {
2588 irq_num = vsi->q_vectors[vector]->irq.virq;
2589 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2590 irq_set_affinity_notifier(irq_num, NULL);
2591 irq_set_affinity_hint(irq_num, NULL);
2592 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2593 }
2594 return err;
2595}
2596
2597/**
2598 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2599 * @vsi: VSI to setup Tx rings used by XDP
2600 *
2601 * Return 0 on success and negative value on error
2602 */
2603static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2604{
2605 struct device *dev = ice_pf_to_dev(vsi->back);
2606 struct ice_tx_desc *tx_desc;
2607 int i, j;
2608
2609 ice_for_each_xdp_txq(vsi, i) {
2610 u16 xdp_q_idx = vsi->alloc_txq + i;
2611 struct ice_ring_stats *ring_stats;
2612 struct ice_tx_ring *xdp_ring;
2613
2614 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2615 if (!xdp_ring)
2616 goto free_xdp_rings;
2617
2618 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
2619 if (!ring_stats) {
2620 ice_free_tx_ring(xdp_ring);
2621 goto free_xdp_rings;
2622 }
2623
2624 xdp_ring->ring_stats = ring_stats;
2625 xdp_ring->q_index = xdp_q_idx;
2626 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2627 xdp_ring->vsi = vsi;
2628 xdp_ring->netdev = NULL;
2629 xdp_ring->dev = dev;
2630 xdp_ring->count = vsi->num_tx_desc;
2631 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2632 if (ice_setup_tx_ring(xdp_ring))
2633 goto free_xdp_rings;
2634 ice_set_ring_xdp(xdp_ring);
2635 spin_lock_init(&xdp_ring->tx_lock);
2636 for (j = 0; j < xdp_ring->count; j++) {
2637 tx_desc = ICE_TX_DESC(xdp_ring, j);
2638 tx_desc->cmd_type_offset_bsz = 0;
2639 }
2640 }
2641
2642 return 0;
2643
2644free_xdp_rings:
2645 for (; i >= 0; i--) {
2646 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc) {
2647 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2648 vsi->xdp_rings[i]->ring_stats = NULL;
2649 ice_free_tx_ring(vsi->xdp_rings[i]);
2650 }
2651 }
2652 return -ENOMEM;
2653}
2654
2655/**
2656 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2657 * @vsi: VSI to set the bpf prog on
2658 * @prog: the bpf prog pointer
2659 */
2660static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2661{
2662 struct bpf_prog *old_prog;
2663 int i;
2664
2665 old_prog = xchg(&vsi->xdp_prog, prog);
2666 ice_for_each_rxq(vsi, i)
2667 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2668
2669 if (old_prog)
2670 bpf_prog_put(old_prog);
2671}
2672
2673/**
2674 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2675 * @vsi: VSI to bring up Tx rings used by XDP
2676 * @prog: bpf program that will be assigned to VSI
2677 *
2678 * Return 0 on success and negative value on error
2679 */
2680int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2681{
2682 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2683 int xdp_rings_rem = vsi->num_xdp_txq;
2684 struct ice_pf *pf = vsi->back;
2685 struct ice_qs_cfg xdp_qs_cfg = {
2686 .qs_mutex = &pf->avail_q_mutex,
2687 .pf_map = pf->avail_txqs,
2688 .pf_map_size = pf->max_pf_txqs,
2689 .q_count = vsi->num_xdp_txq,
2690 .scatter_count = ICE_MAX_SCATTER_TXQS,
2691 .vsi_map = vsi->txq_map,
2692 .vsi_map_offset = vsi->alloc_txq,
2693 .mapping_mode = ICE_VSI_MAP_CONTIG
2694 };
2695 struct device *dev;
2696 int i, v_idx;
2697 int status;
2698
2699 dev = ice_pf_to_dev(pf);
2700 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2701 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2702 if (!vsi->xdp_rings)
2703 return -ENOMEM;
2704
2705 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2706 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2707 goto err_map_xdp;
2708
2709 if (static_key_enabled(&ice_xdp_locking_key))
2710 netdev_warn(vsi->netdev,
2711 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2712
2713 if (ice_xdp_alloc_setup_rings(vsi))
2714 goto clear_xdp_rings;
2715
2716 /* follow the logic from ice_vsi_map_rings_to_vectors */
2717 ice_for_each_q_vector(vsi, v_idx) {
2718 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2719 int xdp_rings_per_v, q_id, q_base;
2720
2721 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2722 vsi->num_q_vectors - v_idx);
2723 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2724
2725 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2726 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2727
2728 xdp_ring->q_vector = q_vector;
2729 xdp_ring->next = q_vector->tx.tx_ring;
2730 q_vector->tx.tx_ring = xdp_ring;
2731 }
2732 xdp_rings_rem -= xdp_rings_per_v;
2733 }
2734
2735 ice_for_each_rxq(vsi, i) {
2736 if (static_key_enabled(&ice_xdp_locking_key)) {
2737 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2738 } else {
2739 struct ice_q_vector *q_vector = vsi->rx_rings[i]->q_vector;
2740 struct ice_tx_ring *ring;
2741
2742 ice_for_each_tx_ring(ring, q_vector->tx) {
2743 if (ice_ring_is_xdp(ring)) {
2744 vsi->rx_rings[i]->xdp_ring = ring;
2745 break;
2746 }
2747 }
2748 }
2749 ice_tx_xsk_pool(vsi, i);
2750 }
2751
2752 /* omit the scheduler update if in reset path; XDP queues will be
2753 * taken into account at the end of ice_vsi_rebuild, where
2754 * ice_cfg_vsi_lan is being called
2755 */
2756 if (ice_is_reset_in_progress(pf->state))
2757 return 0;
2758
2759 /* tell the Tx scheduler that right now we have
2760 * additional queues
2761 */
2762 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2763 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2764
2765 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2766 max_txqs);
2767 if (status) {
2768 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2769 status);
2770 goto clear_xdp_rings;
2771 }
2772
2773 /* assign the prog only when it's not already present on VSI;
2774 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2775 * VSI rebuild that happens under ethtool -L can expose us to
2776 * the bpf_prog refcount issues as we would be swapping same
2777 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2778 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2779 * this is not harmful as dev_xdp_install bumps the refcount
2780 * before calling the op exposed by the driver;
2781 */
2782 if (!ice_is_xdp_ena_vsi(vsi))
2783 ice_vsi_assign_bpf_prog(vsi, prog);
2784
2785 return 0;
2786clear_xdp_rings:
2787 ice_for_each_xdp_txq(vsi, i)
2788 if (vsi->xdp_rings[i]) {
2789 kfree_rcu(vsi->xdp_rings[i], rcu);
2790 vsi->xdp_rings[i] = NULL;
2791 }
2792
2793err_map_xdp:
2794 mutex_lock(&pf->avail_q_mutex);
2795 ice_for_each_xdp_txq(vsi, i) {
2796 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2797 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2798 }
2799 mutex_unlock(&pf->avail_q_mutex);
2800
2801 devm_kfree(dev, vsi->xdp_rings);
2802 return -ENOMEM;
2803}
2804
2805/**
2806 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2807 * @vsi: VSI to remove XDP rings
2808 *
2809 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2810 * resources
2811 */
2812int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2813{
2814 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2815 struct ice_pf *pf = vsi->back;
2816 int i, v_idx;
2817
2818 /* q_vectors are freed in reset path so there's no point in detaching
2819 * rings; in case of rebuild being triggered not from reset bits
2820 * in pf->state won't be set, so additionally check first q_vector
2821 * against NULL
2822 */
2823 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2824 goto free_qmap;
2825
2826 ice_for_each_q_vector(vsi, v_idx) {
2827 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2828 struct ice_tx_ring *ring;
2829
2830 ice_for_each_tx_ring(ring, q_vector->tx)
2831 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2832 break;
2833
2834 /* restore the value of last node prior to XDP setup */
2835 q_vector->tx.tx_ring = ring;
2836 }
2837
2838free_qmap:
2839 mutex_lock(&pf->avail_q_mutex);
2840 ice_for_each_xdp_txq(vsi, i) {
2841 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2842 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2843 }
2844 mutex_unlock(&pf->avail_q_mutex);
2845
2846 ice_for_each_xdp_txq(vsi, i)
2847 if (vsi->xdp_rings[i]) {
2848 if (vsi->xdp_rings[i]->desc) {
2849 synchronize_rcu();
2850 ice_free_tx_ring(vsi->xdp_rings[i]);
2851 }
2852 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2853 vsi->xdp_rings[i]->ring_stats = NULL;
2854 kfree_rcu(vsi->xdp_rings[i], rcu);
2855 vsi->xdp_rings[i] = NULL;
2856 }
2857
2858 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2859 vsi->xdp_rings = NULL;
2860
2861 if (static_key_enabled(&ice_xdp_locking_key))
2862 static_branch_dec(&ice_xdp_locking_key);
2863
2864 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2865 return 0;
2866
2867 ice_vsi_assign_bpf_prog(vsi, NULL);
2868
2869 /* notify Tx scheduler that we destroyed XDP queues and bring
2870 * back the old number of child nodes
2871 */
2872 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2873 max_txqs[i] = vsi->num_txq;
2874
2875 /* change number of XDP Tx queues to 0 */
2876 vsi->num_xdp_txq = 0;
2877
2878 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2879 max_txqs);
2880}
2881
2882/**
2883 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2884 * @vsi: VSI to schedule napi on
2885 */
2886static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2887{
2888 int i;
2889
2890 ice_for_each_rxq(vsi, i) {
2891 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2892
2893 if (rx_ring->xsk_pool)
2894 napi_schedule(&rx_ring->q_vector->napi);
2895 }
2896}
2897
2898/**
2899 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2900 * @vsi: VSI to determine the count of XDP Tx qs
2901 *
2902 * returns 0 if Tx qs count is higher than at least half of CPU count,
2903 * -ENOMEM otherwise
2904 */
2905int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2906{
2907 u16 avail = ice_get_avail_txq_count(vsi->back);
2908 u16 cpus = num_possible_cpus();
2909
2910 if (avail < cpus / 2)
2911 return -ENOMEM;
2912
2913 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2914
2915 if (vsi->num_xdp_txq < cpus)
2916 static_branch_inc(&ice_xdp_locking_key);
2917
2918 return 0;
2919}
2920
2921/**
2922 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2923 * @vsi: Pointer to VSI structure
2924 */
2925static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
2926{
2927 if (test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
2928 return ICE_RXBUF_1664;
2929 else
2930 return ICE_RXBUF_3072;
2931}
2932
2933/**
2934 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2935 * @vsi: VSI to setup XDP for
2936 * @prog: XDP program
2937 * @extack: netlink extended ack
2938 */
2939static int
2940ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2941 struct netlink_ext_ack *extack)
2942{
2943 unsigned int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2944 bool if_running = netif_running(vsi->netdev);
2945 int ret = 0, xdp_ring_err = 0;
2946
2947 if (prog && !prog->aux->xdp_has_frags) {
2948 if (frame_size > ice_max_xdp_frame_size(vsi)) {
2949 NL_SET_ERR_MSG_MOD(extack,
2950 "MTU is too large for linear frames and XDP prog does not support frags");
2951 return -EOPNOTSUPP;
2952 }
2953 }
2954
2955 /* hot swap progs and avoid toggling link */
2956 if (ice_is_xdp_ena_vsi(vsi) == !!prog) {
2957 ice_vsi_assign_bpf_prog(vsi, prog);
2958 return 0;
2959 }
2960
2961 /* need to stop netdev while setting up the program for Rx rings */
2962 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2963 ret = ice_down(vsi);
2964 if (ret) {
2965 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2966 return ret;
2967 }
2968 }
2969
2970 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2971 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2972 if (xdp_ring_err) {
2973 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2974 } else {
2975 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2976 if (xdp_ring_err)
2977 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2978 }
2979 xdp_features_set_redirect_target(vsi->netdev, true);
2980 /* reallocate Rx queues that are used for zero-copy */
2981 xdp_ring_err = ice_realloc_zc_buf(vsi, true);
2982 if (xdp_ring_err)
2983 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed");
2984 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2985 xdp_features_clear_redirect_target(vsi->netdev);
2986 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2987 if (xdp_ring_err)
2988 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2989 /* reallocate Rx queues that were used for zero-copy */
2990 xdp_ring_err = ice_realloc_zc_buf(vsi, false);
2991 if (xdp_ring_err)
2992 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed");
2993 }
2994
2995 if (if_running)
2996 ret = ice_up(vsi);
2997
2998 if (!ret && prog)
2999 ice_vsi_rx_napi_schedule(vsi);
3000
3001 return (ret || xdp_ring_err) ? -ENOMEM : 0;
3002}
3003
3004/**
3005 * ice_xdp_safe_mode - XDP handler for safe mode
3006 * @dev: netdevice
3007 * @xdp: XDP command
3008 */
3009static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
3010 struct netdev_bpf *xdp)
3011{
3012 NL_SET_ERR_MSG_MOD(xdp->extack,
3013 "Please provide working DDP firmware package in order to use XDP\n"
3014 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
3015 return -EOPNOTSUPP;
3016}
3017
3018/**
3019 * ice_xdp - implements XDP handler
3020 * @dev: netdevice
3021 * @xdp: XDP command
3022 */
3023static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
3024{
3025 struct ice_netdev_priv *np = netdev_priv(dev);
3026 struct ice_vsi *vsi = np->vsi;
3027
3028 if (vsi->type != ICE_VSI_PF) {
3029 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
3030 return -EINVAL;
3031 }
3032
3033 switch (xdp->command) {
3034 case XDP_SETUP_PROG:
3035 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
3036 case XDP_SETUP_XSK_POOL:
3037 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
3038 xdp->xsk.queue_id);
3039 default:
3040 return -EINVAL;
3041 }
3042}
3043
3044/**
3045 * ice_ena_misc_vector - enable the non-queue interrupts
3046 * @pf: board private structure
3047 */
3048static void ice_ena_misc_vector(struct ice_pf *pf)
3049{
3050 struct ice_hw *hw = &pf->hw;
3051 u32 pf_intr_start_offset;
3052 u32 val;
3053
3054 /* Disable anti-spoof detection interrupt to prevent spurious event
3055 * interrupts during a function reset. Anti-spoof functionally is
3056 * still supported.
3057 */
3058 val = rd32(hw, GL_MDCK_TX_TDPU);
3059 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
3060 wr32(hw, GL_MDCK_TX_TDPU, val);
3061
3062 /* clear things first */
3063 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
3064 rd32(hw, PFINT_OICR); /* read to clear */
3065
3066 val = (PFINT_OICR_ECC_ERR_M |
3067 PFINT_OICR_MAL_DETECT_M |
3068 PFINT_OICR_GRST_M |
3069 PFINT_OICR_PCI_EXCEPTION_M |
3070 PFINT_OICR_VFLR_M |
3071 PFINT_OICR_HMC_ERR_M |
3072 PFINT_OICR_PE_PUSH_M |
3073 PFINT_OICR_PE_CRITERR_M);
3074
3075 wr32(hw, PFINT_OICR_ENA, val);
3076
3077 /* SW_ITR_IDX = 0, but don't change INTENA */
3078 wr32(hw, GLINT_DYN_CTL(pf->oicr_irq.index),
3079 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3080
3081 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3082 return;
3083 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3084 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset),
3085 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3086}
3087
3088/**
3089 * ice_ll_ts_intr - ll_ts interrupt handler
3090 * @irq: interrupt number
3091 * @data: pointer to a q_vector
3092 */
3093static irqreturn_t ice_ll_ts_intr(int __always_unused irq, void *data)
3094{
3095 struct ice_pf *pf = data;
3096 u32 pf_intr_start_offset;
3097 struct ice_ptp_tx *tx;
3098 unsigned long flags;
3099 struct ice_hw *hw;
3100 u32 val;
3101 u8 idx;
3102
3103 hw = &pf->hw;
3104 tx = &pf->ptp.port.tx;
3105 spin_lock_irqsave(&tx->lock, flags);
3106 ice_ptp_complete_tx_single_tstamp(tx);
3107
3108 idx = find_next_bit_wrap(tx->in_use, tx->len,
3109 tx->last_ll_ts_idx_read + 1);
3110 if (idx != tx->len)
3111 ice_ptp_req_tx_single_tstamp(tx, idx);
3112 spin_unlock_irqrestore(&tx->lock, flags);
3113
3114 val = GLINT_DYN_CTL_INTENA_M | GLINT_DYN_CTL_CLEARPBA_M |
3115 (ICE_ITR_NONE << GLINT_DYN_CTL_ITR_INDX_S);
3116 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3117 wr32(hw, GLINT_DYN_CTL(pf->ll_ts_irq.index + pf_intr_start_offset),
3118 val);
3119
3120 return IRQ_HANDLED;
3121}
3122
3123/**
3124 * ice_misc_intr - misc interrupt handler
3125 * @irq: interrupt number
3126 * @data: pointer to a q_vector
3127 */
3128static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
3129{
3130 struct ice_pf *pf = (struct ice_pf *)data;
3131 irqreturn_t ret = IRQ_HANDLED;
3132 struct ice_hw *hw = &pf->hw;
3133 struct device *dev;
3134 u32 oicr, ena_mask;
3135
3136 dev = ice_pf_to_dev(pf);
3137 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3138 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3139 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3140
3141 oicr = rd32(hw, PFINT_OICR);
3142 ena_mask = rd32(hw, PFINT_OICR_ENA);
3143
3144 if (oicr & PFINT_OICR_SWINT_M) {
3145 ena_mask &= ~PFINT_OICR_SWINT_M;
3146 pf->sw_int_count++;
3147 }
3148
3149 if (oicr & PFINT_OICR_MAL_DETECT_M) {
3150 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3151 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3152 }
3153 if (oicr & PFINT_OICR_VFLR_M) {
3154 /* disable any further VFLR event notifications */
3155 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3156 u32 reg = rd32(hw, PFINT_OICR_ENA);
3157
3158 reg &= ~PFINT_OICR_VFLR_M;
3159 wr32(hw, PFINT_OICR_ENA, reg);
3160 } else {
3161 ena_mask &= ~PFINT_OICR_VFLR_M;
3162 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3163 }
3164 }
3165
3166 if (oicr & PFINT_OICR_GRST_M) {
3167 u32 reset;
3168
3169 /* we have a reset warning */
3170 ena_mask &= ~PFINT_OICR_GRST_M;
3171 reset = FIELD_GET(GLGEN_RSTAT_RESET_TYPE_M,
3172 rd32(hw, GLGEN_RSTAT));
3173
3174 if (reset == ICE_RESET_CORER)
3175 pf->corer_count++;
3176 else if (reset == ICE_RESET_GLOBR)
3177 pf->globr_count++;
3178 else if (reset == ICE_RESET_EMPR)
3179 pf->empr_count++;
3180 else
3181 dev_dbg(dev, "Invalid reset type %d\n", reset);
3182
3183 /* If a reset cycle isn't already in progress, we set a bit in
3184 * pf->state so that the service task can start a reset/rebuild.
3185 */
3186 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3187 if (reset == ICE_RESET_CORER)
3188 set_bit(ICE_CORER_RECV, pf->state);
3189 else if (reset == ICE_RESET_GLOBR)
3190 set_bit(ICE_GLOBR_RECV, pf->state);
3191 else
3192 set_bit(ICE_EMPR_RECV, pf->state);
3193
3194 /* There are couple of different bits at play here.
3195 * hw->reset_ongoing indicates whether the hardware is
3196 * in reset. This is set to true when a reset interrupt
3197 * is received and set back to false after the driver
3198 * has determined that the hardware is out of reset.
3199 *
3200 * ICE_RESET_OICR_RECV in pf->state indicates
3201 * that a post reset rebuild is required before the
3202 * driver is operational again. This is set above.
3203 *
3204 * As this is the start of the reset/rebuild cycle, set
3205 * both to indicate that.
3206 */
3207 hw->reset_ongoing = true;
3208 }
3209 }
3210
3211 if (oicr & PFINT_OICR_TSYN_TX_M) {
3212 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3213 if (ice_pf_state_is_nominal(pf) &&
3214 pf->hw.dev_caps.ts_dev_info.ts_ll_int_read) {
3215 struct ice_ptp_tx *tx = &pf->ptp.port.tx;
3216 unsigned long flags;
3217 u8 idx;
3218
3219 spin_lock_irqsave(&tx->lock, flags);
3220 idx = find_next_bit_wrap(tx->in_use, tx->len,
3221 tx->last_ll_ts_idx_read + 1);
3222 if (idx != tx->len)
3223 ice_ptp_req_tx_single_tstamp(tx, idx);
3224 spin_unlock_irqrestore(&tx->lock, flags);
3225 } else if (ice_ptp_pf_handles_tx_interrupt(pf)) {
3226 set_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread);
3227 ret = IRQ_WAKE_THREAD;
3228 }
3229 }
3230
3231 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3232 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3233 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3234
3235 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3236
3237 if (ice_pf_src_tmr_owned(pf)) {
3238 /* Save EVENTs from GLTSYN register */
3239 pf->ptp.ext_ts_irq |= gltsyn_stat &
3240 (GLTSYN_STAT_EVENT0_M |
3241 GLTSYN_STAT_EVENT1_M |
3242 GLTSYN_STAT_EVENT2_M);
3243
3244 ice_ptp_extts_event(pf);
3245 }
3246 }
3247
3248#define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3249 if (oicr & ICE_AUX_CRIT_ERR) {
3250 pf->oicr_err_reg |= oicr;
3251 set_bit(ICE_AUX_ERR_PENDING, pf->state);
3252 ena_mask &= ~ICE_AUX_CRIT_ERR;
3253 }
3254
3255 /* Report any remaining unexpected interrupts */
3256 oicr &= ena_mask;
3257 if (oicr) {
3258 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3259 /* If a critical error is pending there is no choice but to
3260 * reset the device.
3261 */
3262 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3263 PFINT_OICR_ECC_ERR_M)) {
3264 set_bit(ICE_PFR_REQ, pf->state);
3265 }
3266 }
3267 ice_service_task_schedule(pf);
3268 if (ret == IRQ_HANDLED)
3269 ice_irq_dynamic_ena(hw, NULL, NULL);
3270
3271 return ret;
3272}
3273
3274/**
3275 * ice_misc_intr_thread_fn - misc interrupt thread function
3276 * @irq: interrupt number
3277 * @data: pointer to a q_vector
3278 */
3279static irqreturn_t ice_misc_intr_thread_fn(int __always_unused irq, void *data)
3280{
3281 struct ice_pf *pf = data;
3282 struct ice_hw *hw;
3283
3284 hw = &pf->hw;
3285
3286 if (ice_is_reset_in_progress(pf->state))
3287 goto skip_irq;
3288
3289 if (test_and_clear_bit(ICE_MISC_THREAD_TX_TSTAMP, pf->misc_thread)) {
3290 /* Process outstanding Tx timestamps. If there is more work,
3291 * re-arm the interrupt to trigger again.
3292 */
3293 if (ice_ptp_process_ts(pf) == ICE_TX_TSTAMP_WORK_PENDING) {
3294 wr32(hw, PFINT_OICR, PFINT_OICR_TSYN_TX_M);
3295 ice_flush(hw);
3296 }
3297 }
3298
3299skip_irq:
3300 ice_irq_dynamic_ena(hw, NULL, NULL);
3301
3302 return IRQ_HANDLED;
3303}
3304
3305/**
3306 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3307 * @hw: pointer to HW structure
3308 */
3309static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3310{
3311 /* disable Admin queue Interrupt causes */
3312 wr32(hw, PFINT_FW_CTL,
3313 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3314
3315 /* disable Mailbox queue Interrupt causes */
3316 wr32(hw, PFINT_MBX_CTL,
3317 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3318
3319 wr32(hw, PFINT_SB_CTL,
3320 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3321
3322 /* disable Control queue Interrupt causes */
3323 wr32(hw, PFINT_OICR_CTL,
3324 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3325
3326 ice_flush(hw);
3327}
3328
3329/**
3330 * ice_free_irq_msix_ll_ts- Unroll ll_ts vector setup
3331 * @pf: board private structure
3332 */
3333static void ice_free_irq_msix_ll_ts(struct ice_pf *pf)
3334{
3335 int irq_num = pf->ll_ts_irq.virq;
3336
3337 synchronize_irq(irq_num);
3338 devm_free_irq(ice_pf_to_dev(pf), irq_num, pf);
3339
3340 ice_free_irq(pf, pf->ll_ts_irq);
3341}
3342
3343/**
3344 * ice_free_irq_msix_misc - Unroll misc vector setup
3345 * @pf: board private structure
3346 */
3347static void ice_free_irq_msix_misc(struct ice_pf *pf)
3348{
3349 int misc_irq_num = pf->oicr_irq.virq;
3350 struct ice_hw *hw = &pf->hw;
3351
3352 ice_dis_ctrlq_interrupts(hw);
3353
3354 /* disable OICR interrupt */
3355 wr32(hw, PFINT_OICR_ENA, 0);
3356 ice_flush(hw);
3357
3358 synchronize_irq(misc_irq_num);
3359 devm_free_irq(ice_pf_to_dev(pf), misc_irq_num, pf);
3360
3361 ice_free_irq(pf, pf->oicr_irq);
3362 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3363 ice_free_irq_msix_ll_ts(pf);
3364}
3365
3366/**
3367 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3368 * @hw: pointer to HW structure
3369 * @reg_idx: HW vector index to associate the control queue interrupts with
3370 */
3371static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3372{
3373 u32 val;
3374
3375 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3376 PFINT_OICR_CTL_CAUSE_ENA_M);
3377 wr32(hw, PFINT_OICR_CTL, val);
3378
3379 /* enable Admin queue Interrupt causes */
3380 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3381 PFINT_FW_CTL_CAUSE_ENA_M);
3382 wr32(hw, PFINT_FW_CTL, val);
3383
3384 /* enable Mailbox queue Interrupt causes */
3385 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3386 PFINT_MBX_CTL_CAUSE_ENA_M);
3387 wr32(hw, PFINT_MBX_CTL, val);
3388
3389 if (!hw->dev_caps.ts_dev_info.ts_ll_int_read) {
3390 /* enable Sideband queue Interrupt causes */
3391 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3392 PFINT_SB_CTL_CAUSE_ENA_M);
3393 wr32(hw, PFINT_SB_CTL, val);
3394 }
3395
3396 ice_flush(hw);
3397}
3398
3399/**
3400 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3401 * @pf: board private structure
3402 *
3403 * This sets up the handler for MSIX 0, which is used to manage the
3404 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3405 * when in MSI or Legacy interrupt mode.
3406 */
3407static int ice_req_irq_msix_misc(struct ice_pf *pf)
3408{
3409 struct device *dev = ice_pf_to_dev(pf);
3410 struct ice_hw *hw = &pf->hw;
3411 u32 pf_intr_start_offset;
3412 struct msi_map irq;
3413 int err = 0;
3414
3415 if (!pf->int_name[0])
3416 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3417 dev_driver_string(dev), dev_name(dev));
3418
3419 if (!pf->int_name_ll_ts[0])
3420 snprintf(pf->int_name_ll_ts, sizeof(pf->int_name_ll_ts) - 1,
3421 "%s-%s:ll_ts", dev_driver_string(dev), dev_name(dev));
3422 /* Do not request IRQ but do enable OICR interrupt since settings are
3423 * lost during reset. Note that this function is called only during
3424 * rebuild path and not while reset is in progress.
3425 */
3426 if (ice_is_reset_in_progress(pf->state))
3427 goto skip_req_irq;
3428
3429 /* reserve one vector in irq_tracker for misc interrupts */
3430 irq = ice_alloc_irq(pf, false);
3431 if (irq.index < 0)
3432 return irq.index;
3433
3434 pf->oicr_irq = irq;
3435 err = devm_request_threaded_irq(dev, pf->oicr_irq.virq, ice_misc_intr,
3436 ice_misc_intr_thread_fn, 0,
3437 pf->int_name, pf);
3438 if (err) {
3439 dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n",
3440 pf->int_name, err);
3441 ice_free_irq(pf, pf->oicr_irq);
3442 return err;
3443 }
3444
3445 /* reserve one vector in irq_tracker for ll_ts interrupt */
3446 if (!pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3447 goto skip_req_irq;
3448
3449 irq = ice_alloc_irq(pf, false);
3450 if (irq.index < 0)
3451 return irq.index;
3452
3453 pf->ll_ts_irq = irq;
3454 err = devm_request_irq(dev, pf->ll_ts_irq.virq, ice_ll_ts_intr, 0,
3455 pf->int_name_ll_ts, pf);
3456 if (err) {
3457 dev_err(dev, "devm_request_irq for %s failed: %d\n",
3458 pf->int_name_ll_ts, err);
3459 ice_free_irq(pf, pf->ll_ts_irq);
3460 return err;
3461 }
3462
3463skip_req_irq:
3464 ice_ena_misc_vector(pf);
3465
3466 ice_ena_ctrlq_interrupts(hw, pf->oicr_irq.index);
3467 /* This enables LL TS interrupt */
3468 pf_intr_start_offset = rd32(hw, PFINT_ALLOC) & PFINT_ALLOC_FIRST;
3469 if (pf->hw.dev_caps.ts_dev_info.ts_ll_int_read)
3470 wr32(hw, PFINT_SB_CTL,
3471 ((pf->ll_ts_irq.index + pf_intr_start_offset) &
3472 PFINT_SB_CTL_MSIX_INDX_M) | PFINT_SB_CTL_CAUSE_ENA_M);
3473 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_irq.index),
3474 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3475
3476 ice_flush(hw);
3477 ice_irq_dynamic_ena(hw, NULL, NULL);
3478
3479 return 0;
3480}
3481
3482/**
3483 * ice_napi_add - register NAPI handler for the VSI
3484 * @vsi: VSI for which NAPI handler is to be registered
3485 *
3486 * This function is only called in the driver's load path. Registering the NAPI
3487 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3488 * reset/rebuild, etc.)
3489 */
3490static void ice_napi_add(struct ice_vsi *vsi)
3491{
3492 int v_idx;
3493
3494 if (!vsi->netdev)
3495 return;
3496
3497 ice_for_each_q_vector(vsi, v_idx) {
3498 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3499 ice_napi_poll);
3500 __ice_q_vector_set_napi_queues(vsi->q_vectors[v_idx], false);
3501 }
3502}
3503
3504/**
3505 * ice_set_ops - set netdev and ethtools ops for the given netdev
3506 * @vsi: the VSI associated with the new netdev
3507 */
3508static void ice_set_ops(struct ice_vsi *vsi)
3509{
3510 struct net_device *netdev = vsi->netdev;
3511 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3512
3513 if (ice_is_safe_mode(pf)) {
3514 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3515 ice_set_ethtool_safe_mode_ops(netdev);
3516 return;
3517 }
3518
3519 netdev->netdev_ops = &ice_netdev_ops;
3520 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3521 netdev->xdp_metadata_ops = &ice_xdp_md_ops;
3522 ice_set_ethtool_ops(netdev);
3523
3524 if (vsi->type != ICE_VSI_PF)
3525 return;
3526
3527 netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
3528 NETDEV_XDP_ACT_XSK_ZEROCOPY |
3529 NETDEV_XDP_ACT_RX_SG;
3530 netdev->xdp_zc_max_segs = ICE_MAX_BUF_TXD;
3531}
3532
3533/**
3534 * ice_set_netdev_features - set features for the given netdev
3535 * @netdev: netdev instance
3536 */
3537static void ice_set_netdev_features(struct net_device *netdev)
3538{
3539 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3540 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3541 netdev_features_t csumo_features;
3542 netdev_features_t vlano_features;
3543 netdev_features_t dflt_features;
3544 netdev_features_t tso_features;
3545
3546 if (ice_is_safe_mode(pf)) {
3547 /* safe mode */
3548 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3549 netdev->hw_features = netdev->features;
3550 return;
3551 }
3552
3553 dflt_features = NETIF_F_SG |
3554 NETIF_F_HIGHDMA |
3555 NETIF_F_NTUPLE |
3556 NETIF_F_RXHASH;
3557
3558 csumo_features = NETIF_F_RXCSUM |
3559 NETIF_F_IP_CSUM |
3560 NETIF_F_SCTP_CRC |
3561 NETIF_F_IPV6_CSUM;
3562
3563 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3564 NETIF_F_HW_VLAN_CTAG_TX |
3565 NETIF_F_HW_VLAN_CTAG_RX;
3566
3567 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3568 if (is_dvm_ena)
3569 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3570
3571 tso_features = NETIF_F_TSO |
3572 NETIF_F_TSO_ECN |
3573 NETIF_F_TSO6 |
3574 NETIF_F_GSO_GRE |
3575 NETIF_F_GSO_UDP_TUNNEL |
3576 NETIF_F_GSO_GRE_CSUM |
3577 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3578 NETIF_F_GSO_PARTIAL |
3579 NETIF_F_GSO_IPXIP4 |
3580 NETIF_F_GSO_IPXIP6 |
3581 NETIF_F_GSO_UDP_L4;
3582
3583 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3584 NETIF_F_GSO_GRE_CSUM;
3585 /* set features that user can change */
3586 netdev->hw_features = dflt_features | csumo_features |
3587 vlano_features | tso_features;
3588
3589 /* add support for HW_CSUM on packets with MPLS header */
3590 netdev->mpls_features = NETIF_F_HW_CSUM |
3591 NETIF_F_TSO |
3592 NETIF_F_TSO6;
3593
3594 /* enable features */
3595 netdev->features |= netdev->hw_features;
3596
3597 netdev->hw_features |= NETIF_F_HW_TC;
3598 netdev->hw_features |= NETIF_F_LOOPBACK;
3599
3600 /* encap and VLAN devices inherit default, csumo and tso features */
3601 netdev->hw_enc_features |= dflt_features | csumo_features |
3602 tso_features;
3603 netdev->vlan_features |= dflt_features | csumo_features |
3604 tso_features;
3605
3606 /* advertise support but don't enable by default since only one type of
3607 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3608 * type turns on the other has to be turned off. This is enforced by the
3609 * ice_fix_features() ndo callback.
3610 */
3611 if (is_dvm_ena)
3612 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3613 NETIF_F_HW_VLAN_STAG_TX;
3614
3615 /* Leave CRC / FCS stripping enabled by default, but allow the value to
3616 * be changed at runtime
3617 */
3618 netdev->hw_features |= NETIF_F_RXFCS;
3619
3620 netif_set_tso_max_size(netdev, ICE_MAX_TSO_SIZE);
3621}
3622
3623/**
3624 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3625 * @lut: Lookup table
3626 * @rss_table_size: Lookup table size
3627 * @rss_size: Range of queue number for hashing
3628 */
3629void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3630{
3631 u16 i;
3632
3633 for (i = 0; i < rss_table_size; i++)
3634 lut[i] = i % rss_size;
3635}
3636
3637/**
3638 * ice_pf_vsi_setup - Set up a PF VSI
3639 * @pf: board private structure
3640 * @pi: pointer to the port_info instance
3641 *
3642 * Returns pointer to the successfully allocated VSI software struct
3643 * on success, otherwise returns NULL on failure.
3644 */
3645static struct ice_vsi *
3646ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3647{
3648 struct ice_vsi_cfg_params params = {};
3649
3650 params.type = ICE_VSI_PF;
3651 params.pi = pi;
3652 params.flags = ICE_VSI_FLAG_INIT;
3653
3654 return ice_vsi_setup(pf, ¶ms);
3655}
3656
3657static struct ice_vsi *
3658ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3659 struct ice_channel *ch)
3660{
3661 struct ice_vsi_cfg_params params = {};
3662
3663 params.type = ICE_VSI_CHNL;
3664 params.pi = pi;
3665 params.ch = ch;
3666 params.flags = ICE_VSI_FLAG_INIT;
3667
3668 return ice_vsi_setup(pf, ¶ms);
3669}
3670
3671/**
3672 * ice_ctrl_vsi_setup - Set up a control VSI
3673 * @pf: board private structure
3674 * @pi: pointer to the port_info instance
3675 *
3676 * Returns pointer to the successfully allocated VSI software struct
3677 * on success, otherwise returns NULL on failure.
3678 */
3679static struct ice_vsi *
3680ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3681{
3682 struct ice_vsi_cfg_params params = {};
3683
3684 params.type = ICE_VSI_CTRL;
3685 params.pi = pi;
3686 params.flags = ICE_VSI_FLAG_INIT;
3687
3688 return ice_vsi_setup(pf, ¶ms);
3689}
3690
3691/**
3692 * ice_lb_vsi_setup - Set up a loopback VSI
3693 * @pf: board private structure
3694 * @pi: pointer to the port_info instance
3695 *
3696 * Returns pointer to the successfully allocated VSI software struct
3697 * on success, otherwise returns NULL on failure.
3698 */
3699struct ice_vsi *
3700ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3701{
3702 struct ice_vsi_cfg_params params = {};
3703
3704 params.type = ICE_VSI_LB;
3705 params.pi = pi;
3706 params.flags = ICE_VSI_FLAG_INIT;
3707
3708 return ice_vsi_setup(pf, ¶ms);
3709}
3710
3711/**
3712 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3713 * @netdev: network interface to be adjusted
3714 * @proto: VLAN TPID
3715 * @vid: VLAN ID to be added
3716 *
3717 * net_device_ops implementation for adding VLAN IDs
3718 */
3719static int
3720ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3721{
3722 struct ice_netdev_priv *np = netdev_priv(netdev);
3723 struct ice_vsi_vlan_ops *vlan_ops;
3724 struct ice_vsi *vsi = np->vsi;
3725 struct ice_vlan vlan;
3726 int ret;
3727
3728 /* VLAN 0 is added by default during load/reset */
3729 if (!vid)
3730 return 0;
3731
3732 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3733 usleep_range(1000, 2000);
3734
3735 /* Add multicast promisc rule for the VLAN ID to be added if
3736 * all-multicast is currently enabled.
3737 */
3738 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3739 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3740 ICE_MCAST_VLAN_PROMISC_BITS,
3741 vid);
3742 if (ret)
3743 goto finish;
3744 }
3745
3746 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3747
3748 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3749 * packets aren't pruned by the device's internal switch on Rx
3750 */
3751 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3752 ret = vlan_ops->add_vlan(vsi, &vlan);
3753 if (ret)
3754 goto finish;
3755
3756 /* If all-multicast is currently enabled and this VLAN ID is only one
3757 * besides VLAN-0 we have to update look-up type of multicast promisc
3758 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3759 */
3760 if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3761 ice_vsi_num_non_zero_vlans(vsi) == 1) {
3762 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3763 ICE_MCAST_PROMISC_BITS, 0);
3764 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3765 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3766 }
3767
3768finish:
3769 clear_bit(ICE_CFG_BUSY, vsi->state);
3770
3771 return ret;
3772}
3773
3774/**
3775 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3776 * @netdev: network interface to be adjusted
3777 * @proto: VLAN TPID
3778 * @vid: VLAN ID to be removed
3779 *
3780 * net_device_ops implementation for removing VLAN IDs
3781 */
3782static int
3783ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3784{
3785 struct ice_netdev_priv *np = netdev_priv(netdev);
3786 struct ice_vsi_vlan_ops *vlan_ops;
3787 struct ice_vsi *vsi = np->vsi;
3788 struct ice_vlan vlan;
3789 int ret;
3790
3791 /* don't allow removal of VLAN 0 */
3792 if (!vid)
3793 return 0;
3794
3795 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3796 usleep_range(1000, 2000);
3797
3798 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3799 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3800 if (ret) {
3801 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3802 vsi->vsi_num);
3803 vsi->current_netdev_flags |= IFF_ALLMULTI;
3804 }
3805
3806 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3807
3808 /* Make sure VLAN delete is successful before updating VLAN
3809 * information
3810 */
3811 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3812 ret = vlan_ops->del_vlan(vsi, &vlan);
3813 if (ret)
3814 goto finish;
3815
3816 /* Remove multicast promisc rule for the removed VLAN ID if
3817 * all-multicast is enabled.
3818 */
3819 if (vsi->current_netdev_flags & IFF_ALLMULTI)
3820 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3821 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3822
3823 if (!ice_vsi_has_non_zero_vlans(vsi)) {
3824 /* Update look-up type of multicast promisc rule for VLAN 0
3825 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3826 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3827 */
3828 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3829 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3830 ICE_MCAST_VLAN_PROMISC_BITS,
3831 0);
3832 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3833 ICE_MCAST_PROMISC_BITS, 0);
3834 }
3835 }
3836
3837finish:
3838 clear_bit(ICE_CFG_BUSY, vsi->state);
3839
3840 return ret;
3841}
3842
3843/**
3844 * ice_rep_indr_tc_block_unbind
3845 * @cb_priv: indirection block private data
3846 */
3847static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3848{
3849 struct ice_indr_block_priv *indr_priv = cb_priv;
3850
3851 list_del(&indr_priv->list);
3852 kfree(indr_priv);
3853}
3854
3855/**
3856 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3857 * @vsi: VSI struct which has the netdev
3858 */
3859static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3860{
3861 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3862
3863 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3864 ice_rep_indr_tc_block_unbind);
3865}
3866
3867/**
3868 * ice_tc_indir_block_register - Register TC indirect block notifications
3869 * @vsi: VSI struct which has the netdev
3870 *
3871 * Returns 0 on success, negative value on failure
3872 */
3873static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3874{
3875 struct ice_netdev_priv *np;
3876
3877 if (!vsi || !vsi->netdev)
3878 return -EINVAL;
3879
3880 np = netdev_priv(vsi->netdev);
3881
3882 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3883 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3884}
3885
3886/**
3887 * ice_get_avail_q_count - Get count of queues in use
3888 * @pf_qmap: bitmap to get queue use count from
3889 * @lock: pointer to a mutex that protects access to pf_qmap
3890 * @size: size of the bitmap
3891 */
3892static u16
3893ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3894{
3895 unsigned long bit;
3896 u16 count = 0;
3897
3898 mutex_lock(lock);
3899 for_each_clear_bit(bit, pf_qmap, size)
3900 count++;
3901 mutex_unlock(lock);
3902
3903 return count;
3904}
3905
3906/**
3907 * ice_get_avail_txq_count - Get count of Tx queues in use
3908 * @pf: pointer to an ice_pf instance
3909 */
3910u16 ice_get_avail_txq_count(struct ice_pf *pf)
3911{
3912 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3913 pf->max_pf_txqs);
3914}
3915
3916/**
3917 * ice_get_avail_rxq_count - Get count of Rx queues in use
3918 * @pf: pointer to an ice_pf instance
3919 */
3920u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3921{
3922 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3923 pf->max_pf_rxqs);
3924}
3925
3926/**
3927 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3928 * @pf: board private structure to initialize
3929 */
3930static void ice_deinit_pf(struct ice_pf *pf)
3931{
3932 ice_service_task_stop(pf);
3933 mutex_destroy(&pf->lag_mutex);
3934 mutex_destroy(&pf->adev_mutex);
3935 mutex_destroy(&pf->sw_mutex);
3936 mutex_destroy(&pf->tc_mutex);
3937 mutex_destroy(&pf->avail_q_mutex);
3938 mutex_destroy(&pf->vfs.table_lock);
3939
3940 if (pf->avail_txqs) {
3941 bitmap_free(pf->avail_txqs);
3942 pf->avail_txqs = NULL;
3943 }
3944
3945 if (pf->avail_rxqs) {
3946 bitmap_free(pf->avail_rxqs);
3947 pf->avail_rxqs = NULL;
3948 }
3949
3950 if (pf->ptp.clock)
3951 ptp_clock_unregister(pf->ptp.clock);
3952}
3953
3954/**
3955 * ice_set_pf_caps - set PFs capability flags
3956 * @pf: pointer to the PF instance
3957 */
3958static void ice_set_pf_caps(struct ice_pf *pf)
3959{
3960 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3961
3962 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3963 if (func_caps->common_cap.rdma)
3964 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3965 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3966 if (func_caps->common_cap.dcb)
3967 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3968 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3969 if (func_caps->common_cap.sr_iov_1_1) {
3970 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3971 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3972 ICE_MAX_SRIOV_VFS);
3973 }
3974 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3975 if (func_caps->common_cap.rss_table_size)
3976 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3977
3978 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3979 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3980 u16 unused;
3981
3982 /* ctrl_vsi_idx will be set to a valid value when flow director
3983 * is setup by ice_init_fdir
3984 */
3985 pf->ctrl_vsi_idx = ICE_NO_VSI;
3986 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3987 /* force guaranteed filter pool for PF */
3988 ice_alloc_fd_guar_item(&pf->hw, &unused,
3989 func_caps->fd_fltr_guar);
3990 /* force shared filter pool for PF */
3991 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3992 func_caps->fd_fltr_best_effort);
3993 }
3994
3995 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3996 if (func_caps->common_cap.ieee_1588 &&
3997 !(pf->hw.mac_type == ICE_MAC_E830))
3998 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3999
4000 pf->max_pf_txqs = func_caps->common_cap.num_txq;
4001 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
4002}
4003
4004/**
4005 * ice_init_pf - Initialize general software structures (struct ice_pf)
4006 * @pf: board private structure to initialize
4007 */
4008static int ice_init_pf(struct ice_pf *pf)
4009{
4010 ice_set_pf_caps(pf);
4011
4012 mutex_init(&pf->sw_mutex);
4013 mutex_init(&pf->tc_mutex);
4014 mutex_init(&pf->adev_mutex);
4015 mutex_init(&pf->lag_mutex);
4016
4017 INIT_HLIST_HEAD(&pf->aq_wait_list);
4018 spin_lock_init(&pf->aq_wait_lock);
4019 init_waitqueue_head(&pf->aq_wait_queue);
4020
4021 init_waitqueue_head(&pf->reset_wait_queue);
4022
4023 /* setup service timer and periodic service task */
4024 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
4025 pf->serv_tmr_period = HZ;
4026 INIT_WORK(&pf->serv_task, ice_service_task);
4027 clear_bit(ICE_SERVICE_SCHED, pf->state);
4028
4029 mutex_init(&pf->avail_q_mutex);
4030 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
4031 if (!pf->avail_txqs)
4032 return -ENOMEM;
4033
4034 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
4035 if (!pf->avail_rxqs) {
4036 bitmap_free(pf->avail_txqs);
4037 pf->avail_txqs = NULL;
4038 return -ENOMEM;
4039 }
4040
4041 mutex_init(&pf->vfs.table_lock);
4042 hash_init(pf->vfs.table);
4043 ice_mbx_init_snapshot(&pf->hw);
4044
4045 return 0;
4046}
4047
4048/**
4049 * ice_is_wol_supported - check if WoL is supported
4050 * @hw: pointer to hardware info
4051 *
4052 * Check if WoL is supported based on the HW configuration.
4053 * Returns true if NVM supports and enables WoL for this port, false otherwise
4054 */
4055bool ice_is_wol_supported(struct ice_hw *hw)
4056{
4057 u16 wol_ctrl;
4058
4059 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4060 * word) indicates WoL is not supported on the corresponding PF ID.
4061 */
4062 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4063 return false;
4064
4065 return !(BIT(hw->port_info->lport) & wol_ctrl);
4066}
4067
4068/**
4069 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4070 * @vsi: VSI being changed
4071 * @new_rx: new number of Rx queues
4072 * @new_tx: new number of Tx queues
4073 * @locked: is adev device_lock held
4074 *
4075 * Only change the number of queues if new_tx, or new_rx is non-0.
4076 *
4077 * Returns 0 on success.
4078 */
4079int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx, bool locked)
4080{
4081 struct ice_pf *pf = vsi->back;
4082 int err = 0, timeout = 50;
4083
4084 if (!new_rx && !new_tx)
4085 return -EINVAL;
4086
4087 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4088 timeout--;
4089 if (!timeout)
4090 return -EBUSY;
4091 usleep_range(1000, 2000);
4092 }
4093
4094 if (new_tx)
4095 vsi->req_txq = (u16)new_tx;
4096 if (new_rx)
4097 vsi->req_rxq = (u16)new_rx;
4098
4099 /* set for the next time the netdev is started */
4100 if (!netif_running(vsi->netdev)) {
4101 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4102 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4103 goto done;
4104 }
4105
4106 ice_vsi_close(vsi);
4107 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4108 ice_pf_dcb_recfg(pf, locked);
4109 ice_vsi_open(vsi);
4110done:
4111 clear_bit(ICE_CFG_BUSY, pf->state);
4112 return err;
4113}
4114
4115/**
4116 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4117 * @pf: PF to configure
4118 *
4119 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4120 * VSI can still Tx/Rx VLAN tagged packets.
4121 */
4122static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4123{
4124 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4125 struct ice_vsi_ctx *ctxt;
4126 struct ice_hw *hw;
4127 int status;
4128
4129 if (!vsi)
4130 return;
4131
4132 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4133 if (!ctxt)
4134 return;
4135
4136 hw = &pf->hw;
4137 ctxt->info = vsi->info;
4138
4139 ctxt->info.valid_sections =
4140 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4141 ICE_AQ_VSI_PROP_SECURITY_VALID |
4142 ICE_AQ_VSI_PROP_SW_VALID);
4143
4144 /* disable VLAN anti-spoof */
4145 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4146 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4147
4148 /* disable VLAN pruning and keep all other settings */
4149 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4150
4151 /* allow all VLANs on Tx and don't strip on Rx */
4152 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4153 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4154
4155 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4156 if (status) {
4157 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4158 status, ice_aq_str(hw->adminq.sq_last_status));
4159 } else {
4160 vsi->info.sec_flags = ctxt->info.sec_flags;
4161 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4162 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4163 }
4164
4165 kfree(ctxt);
4166}
4167
4168/**
4169 * ice_log_pkg_init - log result of DDP package load
4170 * @hw: pointer to hardware info
4171 * @state: state of package load
4172 */
4173static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4174{
4175 struct ice_pf *pf = hw->back;
4176 struct device *dev;
4177
4178 dev = ice_pf_to_dev(pf);
4179
4180 switch (state) {
4181 case ICE_DDP_PKG_SUCCESS:
4182 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4183 hw->active_pkg_name,
4184 hw->active_pkg_ver.major,
4185 hw->active_pkg_ver.minor,
4186 hw->active_pkg_ver.update,
4187 hw->active_pkg_ver.draft);
4188 break;
4189 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4190 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4191 hw->active_pkg_name,
4192 hw->active_pkg_ver.major,
4193 hw->active_pkg_ver.minor,
4194 hw->active_pkg_ver.update,
4195 hw->active_pkg_ver.draft);
4196 break;
4197 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4198 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",
4199 hw->active_pkg_name,
4200 hw->active_pkg_ver.major,
4201 hw->active_pkg_ver.minor,
4202 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4203 break;
4204 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4205 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",
4206 hw->active_pkg_name,
4207 hw->active_pkg_ver.major,
4208 hw->active_pkg_ver.minor,
4209 hw->active_pkg_ver.update,
4210 hw->active_pkg_ver.draft,
4211 hw->pkg_name,
4212 hw->pkg_ver.major,
4213 hw->pkg_ver.minor,
4214 hw->pkg_ver.update,
4215 hw->pkg_ver.draft);
4216 break;
4217 case ICE_DDP_PKG_FW_MISMATCH:
4218 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");
4219 break;
4220 case ICE_DDP_PKG_INVALID_FILE:
4221 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4222 break;
4223 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4224 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4225 break;
4226 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4227 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",
4228 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4229 break;
4230 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4231 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");
4232 break;
4233 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4234 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");
4235 break;
4236 case ICE_DDP_PKG_LOAD_ERROR:
4237 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4238 /* poll for reset to complete */
4239 if (ice_check_reset(hw))
4240 dev_err(dev, "Error resetting device. Please reload the driver\n");
4241 break;
4242 case ICE_DDP_PKG_ERR:
4243 default:
4244 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4245 break;
4246 }
4247}
4248
4249/**
4250 * ice_load_pkg - load/reload the DDP Package file
4251 * @firmware: firmware structure when firmware requested or NULL for reload
4252 * @pf: pointer to the PF instance
4253 *
4254 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4255 * initialize HW tables.
4256 */
4257static void
4258ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4259{
4260 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4261 struct device *dev = ice_pf_to_dev(pf);
4262 struct ice_hw *hw = &pf->hw;
4263
4264 /* Load DDP Package */
4265 if (firmware && !hw->pkg_copy) {
4266 state = ice_copy_and_init_pkg(hw, firmware->data,
4267 firmware->size);
4268 ice_log_pkg_init(hw, state);
4269 } else if (!firmware && hw->pkg_copy) {
4270 /* Reload package during rebuild after CORER/GLOBR reset */
4271 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4272 ice_log_pkg_init(hw, state);
4273 } else {
4274 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4275 }
4276
4277 if (!ice_is_init_pkg_successful(state)) {
4278 /* Safe Mode */
4279 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4280 return;
4281 }
4282
4283 /* Successful download package is the precondition for advanced
4284 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4285 */
4286 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4287}
4288
4289/**
4290 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4291 * @pf: pointer to the PF structure
4292 *
4293 * There is no error returned here because the driver should be able to handle
4294 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4295 * specifically with Tx.
4296 */
4297static void ice_verify_cacheline_size(struct ice_pf *pf)
4298{
4299 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4300 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4301 ICE_CACHE_LINE_BYTES);
4302}
4303
4304/**
4305 * ice_send_version - update firmware with driver version
4306 * @pf: PF struct
4307 *
4308 * Returns 0 on success, else error code
4309 */
4310static int ice_send_version(struct ice_pf *pf)
4311{
4312 struct ice_driver_ver dv;
4313
4314 dv.major_ver = 0xff;
4315 dv.minor_ver = 0xff;
4316 dv.build_ver = 0xff;
4317 dv.subbuild_ver = 0;
4318 strscpy((char *)dv.driver_string, UTS_RELEASE,
4319 sizeof(dv.driver_string));
4320 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4321}
4322
4323/**
4324 * ice_init_fdir - Initialize flow director VSI and configuration
4325 * @pf: pointer to the PF instance
4326 *
4327 * returns 0 on success, negative on error
4328 */
4329static int ice_init_fdir(struct ice_pf *pf)
4330{
4331 struct device *dev = ice_pf_to_dev(pf);
4332 struct ice_vsi *ctrl_vsi;
4333 int err;
4334
4335 /* Side Band Flow Director needs to have a control VSI.
4336 * Allocate it and store it in the PF.
4337 */
4338 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4339 if (!ctrl_vsi) {
4340 dev_dbg(dev, "could not create control VSI\n");
4341 return -ENOMEM;
4342 }
4343
4344 err = ice_vsi_open_ctrl(ctrl_vsi);
4345 if (err) {
4346 dev_dbg(dev, "could not open control VSI\n");
4347 goto err_vsi_open;
4348 }
4349
4350 mutex_init(&pf->hw.fdir_fltr_lock);
4351
4352 err = ice_fdir_create_dflt_rules(pf);
4353 if (err)
4354 goto err_fdir_rule;
4355
4356 return 0;
4357
4358err_fdir_rule:
4359 ice_fdir_release_flows(&pf->hw);
4360 ice_vsi_close(ctrl_vsi);
4361err_vsi_open:
4362 ice_vsi_release(ctrl_vsi);
4363 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4364 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4365 pf->ctrl_vsi_idx = ICE_NO_VSI;
4366 }
4367 return err;
4368}
4369
4370static void ice_deinit_fdir(struct ice_pf *pf)
4371{
4372 struct ice_vsi *vsi = ice_get_ctrl_vsi(pf);
4373
4374 if (!vsi)
4375 return;
4376
4377 ice_vsi_manage_fdir(vsi, false);
4378 ice_vsi_release(vsi);
4379 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4380 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4381 pf->ctrl_vsi_idx = ICE_NO_VSI;
4382 }
4383
4384 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4385}
4386
4387/**
4388 * ice_get_opt_fw_name - return optional firmware file name or NULL
4389 * @pf: pointer to the PF instance
4390 */
4391static char *ice_get_opt_fw_name(struct ice_pf *pf)
4392{
4393 /* Optional firmware name same as default with additional dash
4394 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4395 */
4396 struct pci_dev *pdev = pf->pdev;
4397 char *opt_fw_filename;
4398 u64 dsn;
4399
4400 /* Determine the name of the optional file using the DSN (two
4401 * dwords following the start of the DSN Capability).
4402 */
4403 dsn = pci_get_dsn(pdev);
4404 if (!dsn)
4405 return NULL;
4406
4407 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4408 if (!opt_fw_filename)
4409 return NULL;
4410
4411 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4412 ICE_DDP_PKG_PATH, dsn);
4413
4414 return opt_fw_filename;
4415}
4416
4417/**
4418 * ice_request_fw - Device initialization routine
4419 * @pf: pointer to the PF instance
4420 */
4421static void ice_request_fw(struct ice_pf *pf)
4422{
4423 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4424 const struct firmware *firmware = NULL;
4425 struct device *dev = ice_pf_to_dev(pf);
4426 int err = 0;
4427
4428 /* optional device-specific DDP (if present) overrides the default DDP
4429 * package file. kernel logs a debug message if the file doesn't exist,
4430 * and warning messages for other errors.
4431 */
4432 if (opt_fw_filename) {
4433 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4434 if (err) {
4435 kfree(opt_fw_filename);
4436 goto dflt_pkg_load;
4437 }
4438
4439 /* request for firmware was successful. Download to device */
4440 ice_load_pkg(firmware, pf);
4441 kfree(opt_fw_filename);
4442 release_firmware(firmware);
4443 return;
4444 }
4445
4446dflt_pkg_load:
4447 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4448 if (err) {
4449 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4450 return;
4451 }
4452
4453 /* request for firmware was successful. Download to device */
4454 ice_load_pkg(firmware, pf);
4455 release_firmware(firmware);
4456}
4457
4458/**
4459 * ice_print_wake_reason - show the wake up cause in the log
4460 * @pf: pointer to the PF struct
4461 */
4462static void ice_print_wake_reason(struct ice_pf *pf)
4463{
4464 u32 wus = pf->wakeup_reason;
4465 const char *wake_str;
4466
4467 /* if no wake event, nothing to print */
4468 if (!wus)
4469 return;
4470
4471 if (wus & PFPM_WUS_LNKC_M)
4472 wake_str = "Link\n";
4473 else if (wus & PFPM_WUS_MAG_M)
4474 wake_str = "Magic Packet\n";
4475 else if (wus & PFPM_WUS_MNG_M)
4476 wake_str = "Management\n";
4477 else if (wus & PFPM_WUS_FW_RST_WK_M)
4478 wake_str = "Firmware Reset\n";
4479 else
4480 wake_str = "Unknown\n";
4481
4482 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4483}
4484
4485/**
4486 * ice_pf_fwlog_update_module - update 1 module
4487 * @pf: pointer to the PF struct
4488 * @log_level: log_level to use for the @module
4489 * @module: module to update
4490 */
4491void ice_pf_fwlog_update_module(struct ice_pf *pf, int log_level, int module)
4492{
4493 struct ice_hw *hw = &pf->hw;
4494
4495 hw->fwlog_cfg.module_entries[module].log_level = log_level;
4496}
4497
4498/**
4499 * ice_register_netdev - register netdev
4500 * @vsi: pointer to the VSI struct
4501 */
4502static int ice_register_netdev(struct ice_vsi *vsi)
4503{
4504 int err;
4505
4506 if (!vsi || !vsi->netdev)
4507 return -EIO;
4508
4509 err = register_netdev(vsi->netdev);
4510 if (err)
4511 return err;
4512
4513 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4514 netif_carrier_off(vsi->netdev);
4515 netif_tx_stop_all_queues(vsi->netdev);
4516
4517 return 0;
4518}
4519
4520static void ice_unregister_netdev(struct ice_vsi *vsi)
4521{
4522 if (!vsi || !vsi->netdev)
4523 return;
4524
4525 unregister_netdev(vsi->netdev);
4526 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4527}
4528
4529/**
4530 * ice_cfg_netdev - Allocate, configure and register a netdev
4531 * @vsi: the VSI associated with the new netdev
4532 *
4533 * Returns 0 on success, negative value on failure
4534 */
4535static int ice_cfg_netdev(struct ice_vsi *vsi)
4536{
4537 struct ice_netdev_priv *np;
4538 struct net_device *netdev;
4539 u8 mac_addr[ETH_ALEN];
4540
4541 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
4542 vsi->alloc_rxq);
4543 if (!netdev)
4544 return -ENOMEM;
4545
4546 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4547 vsi->netdev = netdev;
4548 np = netdev_priv(netdev);
4549 np->vsi = vsi;
4550
4551 ice_set_netdev_features(netdev);
4552 ice_set_ops(vsi);
4553
4554 if (vsi->type == ICE_VSI_PF) {
4555 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
4556 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4557 eth_hw_addr_set(netdev, mac_addr);
4558 }
4559
4560 netdev->priv_flags |= IFF_UNICAST_FLT;
4561
4562 /* Setup netdev TC information */
4563 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
4564
4565 netdev->max_mtu = ICE_MAX_MTU;
4566
4567 return 0;
4568}
4569
4570static void ice_decfg_netdev(struct ice_vsi *vsi)
4571{
4572 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4573 free_netdev(vsi->netdev);
4574 vsi->netdev = NULL;
4575}
4576
4577/**
4578 * ice_wait_for_fw - wait for full FW readiness
4579 * @hw: pointer to the hardware structure
4580 * @timeout: milliseconds that can elapse before timing out
4581 */
4582static int ice_wait_for_fw(struct ice_hw *hw, u32 timeout)
4583{
4584 int fw_loading;
4585 u32 elapsed = 0;
4586
4587 while (elapsed <= timeout) {
4588 fw_loading = rd32(hw, GL_MNG_FWSM) & GL_MNG_FWSM_FW_LOADING_M;
4589
4590 /* firmware was not yet loaded, we have to wait more */
4591 if (fw_loading) {
4592 elapsed += 100;
4593 msleep(100);
4594 continue;
4595 }
4596 return 0;
4597 }
4598
4599 return -ETIMEDOUT;
4600}
4601
4602int ice_init_dev(struct ice_pf *pf)
4603{
4604 struct device *dev = ice_pf_to_dev(pf);
4605 struct ice_hw *hw = &pf->hw;
4606 int err;
4607
4608 err = ice_init_hw(hw);
4609 if (err) {
4610 dev_err(dev, "ice_init_hw failed: %d\n", err);
4611 return err;
4612 }
4613
4614 /* Some cards require longer initialization times
4615 * due to necessity of loading FW from an external source.
4616 * This can take even half a minute.
4617 */
4618 if (ice_is_pf_c827(hw)) {
4619 err = ice_wait_for_fw(hw, 30000);
4620 if (err) {
4621 dev_err(dev, "ice_wait_for_fw timed out");
4622 return err;
4623 }
4624 }
4625
4626 ice_init_feature_support(pf);
4627
4628 ice_request_fw(pf);
4629
4630 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4631 * set in pf->state, which will cause ice_is_safe_mode to return
4632 * true
4633 */
4634 if (ice_is_safe_mode(pf)) {
4635 /* we already got function/device capabilities but these don't
4636 * reflect what the driver needs to do in safe mode. Instead of
4637 * adding conditional logic everywhere to ignore these
4638 * device/function capabilities, override them.
4639 */
4640 ice_set_safe_mode_caps(hw);
4641 }
4642
4643 err = ice_init_pf(pf);
4644 if (err) {
4645 dev_err(dev, "ice_init_pf failed: %d\n", err);
4646 goto err_init_pf;
4647 }
4648
4649 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4650 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4651 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4652 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4653 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4654 pf->hw.udp_tunnel_nic.tables[0].n_entries =
4655 pf->hw.tnl.valid_count[TNL_VXLAN];
4656 pf->hw.udp_tunnel_nic.tables[0].tunnel_types =
4657 UDP_TUNNEL_TYPE_VXLAN;
4658 }
4659 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4660 pf->hw.udp_tunnel_nic.tables[1].n_entries =
4661 pf->hw.tnl.valid_count[TNL_GENEVE];
4662 pf->hw.udp_tunnel_nic.tables[1].tunnel_types =
4663 UDP_TUNNEL_TYPE_GENEVE;
4664 }
4665
4666 err = ice_init_interrupt_scheme(pf);
4667 if (err) {
4668 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4669 err = -EIO;
4670 goto err_init_interrupt_scheme;
4671 }
4672
4673 /* In case of MSIX we are going to setup the misc vector right here
4674 * to handle admin queue events etc. In case of legacy and MSI
4675 * the misc functionality and queue processing is combined in
4676 * the same vector and that gets setup at open.
4677 */
4678 err = ice_req_irq_msix_misc(pf);
4679 if (err) {
4680 dev_err(dev, "setup of misc vector failed: %d\n", err);
4681 goto err_req_irq_msix_misc;
4682 }
4683
4684 return 0;
4685
4686err_req_irq_msix_misc:
4687 ice_clear_interrupt_scheme(pf);
4688err_init_interrupt_scheme:
4689 ice_deinit_pf(pf);
4690err_init_pf:
4691 ice_deinit_hw(hw);
4692 return err;
4693}
4694
4695void ice_deinit_dev(struct ice_pf *pf)
4696{
4697 ice_free_irq_msix_misc(pf);
4698 ice_deinit_pf(pf);
4699 ice_deinit_hw(&pf->hw);
4700
4701 /* Service task is already stopped, so call reset directly. */
4702 ice_reset(&pf->hw, ICE_RESET_PFR);
4703 pci_wait_for_pending_transaction(pf->pdev);
4704 ice_clear_interrupt_scheme(pf);
4705}
4706
4707static void ice_init_features(struct ice_pf *pf)
4708{
4709 struct device *dev = ice_pf_to_dev(pf);
4710
4711 if (ice_is_safe_mode(pf))
4712 return;
4713
4714 /* initialize DDP driven features */
4715 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4716 ice_ptp_init(pf);
4717
4718 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4719 ice_gnss_init(pf);
4720
4721 if (ice_is_feature_supported(pf, ICE_F_CGU) ||
4722 ice_is_feature_supported(pf, ICE_F_PHY_RCLK))
4723 ice_dpll_init(pf);
4724
4725 /* Note: Flow director init failure is non-fatal to load */
4726 if (ice_init_fdir(pf))
4727 dev_err(dev, "could not initialize flow director\n");
4728
4729 /* Note: DCB init failure is non-fatal to load */
4730 if (ice_init_pf_dcb(pf, false)) {
4731 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4732 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4733 } else {
4734 ice_cfg_lldp_mib_change(&pf->hw, true);
4735 }
4736
4737 if (ice_init_lag(pf))
4738 dev_warn(dev, "Failed to init link aggregation support\n");
4739
4740 ice_hwmon_init(pf);
4741}
4742
4743static void ice_deinit_features(struct ice_pf *pf)
4744{
4745 if (ice_is_safe_mode(pf))
4746 return;
4747
4748 ice_deinit_lag(pf);
4749 if (test_bit(ICE_FLAG_DCB_CAPABLE, pf->flags))
4750 ice_cfg_lldp_mib_change(&pf->hw, false);
4751 ice_deinit_fdir(pf);
4752 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4753 ice_gnss_exit(pf);
4754 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4755 ice_ptp_release(pf);
4756 if (test_bit(ICE_FLAG_DPLL, pf->flags))
4757 ice_dpll_deinit(pf);
4758 if (pf->eswitch_mode == DEVLINK_ESWITCH_MODE_SWITCHDEV)
4759 xa_destroy(&pf->eswitch.reprs);
4760}
4761
4762static void ice_init_wakeup(struct ice_pf *pf)
4763{
4764 /* Save wakeup reason register for later use */
4765 pf->wakeup_reason = rd32(&pf->hw, PFPM_WUS);
4766
4767 /* check for a power management event */
4768 ice_print_wake_reason(pf);
4769
4770 /* clear wake status, all bits */
4771 wr32(&pf->hw, PFPM_WUS, U32_MAX);
4772
4773 /* Disable WoL at init, wait for user to enable */
4774 device_set_wakeup_enable(ice_pf_to_dev(pf), false);
4775}
4776
4777static int ice_init_link(struct ice_pf *pf)
4778{
4779 struct device *dev = ice_pf_to_dev(pf);
4780 int err;
4781
4782 err = ice_init_link_events(pf->hw.port_info);
4783 if (err) {
4784 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4785 return err;
4786 }
4787
4788 /* not a fatal error if this fails */
4789 err = ice_init_nvm_phy_type(pf->hw.port_info);
4790 if (err)
4791 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4792
4793 /* not a fatal error if this fails */
4794 err = ice_update_link_info(pf->hw.port_info);
4795 if (err)
4796 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4797
4798 ice_init_link_dflt_override(pf->hw.port_info);
4799
4800 ice_check_link_cfg_err(pf,
4801 pf->hw.port_info->phy.link_info.link_cfg_err);
4802
4803 /* if media available, initialize PHY settings */
4804 if (pf->hw.port_info->phy.link_info.link_info &
4805 ICE_AQ_MEDIA_AVAILABLE) {
4806 /* not a fatal error if this fails */
4807 err = ice_init_phy_user_cfg(pf->hw.port_info);
4808 if (err)
4809 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4810
4811 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4812 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4813
4814 if (vsi)
4815 ice_configure_phy(vsi);
4816 }
4817 } else {
4818 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4819 }
4820
4821 return err;
4822}
4823
4824static int ice_init_pf_sw(struct ice_pf *pf)
4825{
4826 bool dvm = ice_is_dvm_ena(&pf->hw);
4827 struct ice_vsi *vsi;
4828 int err;
4829
4830 /* create switch struct for the switch element created by FW on boot */
4831 pf->first_sw = kzalloc(sizeof(*pf->first_sw), GFP_KERNEL);
4832 if (!pf->first_sw)
4833 return -ENOMEM;
4834
4835 if (pf->hw.evb_veb)
4836 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4837 else
4838 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4839
4840 pf->first_sw->pf = pf;
4841
4842 /* record the sw_id available for later use */
4843 pf->first_sw->sw_id = pf->hw.port_info->sw_id;
4844
4845 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
4846 if (err)
4847 goto err_aq_set_port_params;
4848
4849 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
4850 if (!vsi) {
4851 err = -ENOMEM;
4852 goto err_pf_vsi_setup;
4853 }
4854
4855 return 0;
4856
4857err_pf_vsi_setup:
4858err_aq_set_port_params:
4859 kfree(pf->first_sw);
4860 return err;
4861}
4862
4863static void ice_deinit_pf_sw(struct ice_pf *pf)
4864{
4865 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4866
4867 if (!vsi)
4868 return;
4869
4870 ice_vsi_release(vsi);
4871 kfree(pf->first_sw);
4872}
4873
4874static int ice_alloc_vsis(struct ice_pf *pf)
4875{
4876 struct device *dev = ice_pf_to_dev(pf);
4877
4878 pf->num_alloc_vsi = pf->hw.func_caps.guar_num_vsi;
4879 if (!pf->num_alloc_vsi)
4880 return -EIO;
4881
4882 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4883 dev_warn(dev,
4884 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4885 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4886 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4887 }
4888
4889 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4890 GFP_KERNEL);
4891 if (!pf->vsi)
4892 return -ENOMEM;
4893
4894 pf->vsi_stats = devm_kcalloc(dev, pf->num_alloc_vsi,
4895 sizeof(*pf->vsi_stats), GFP_KERNEL);
4896 if (!pf->vsi_stats) {
4897 devm_kfree(dev, pf->vsi);
4898 return -ENOMEM;
4899 }
4900
4901 return 0;
4902}
4903
4904static void ice_dealloc_vsis(struct ice_pf *pf)
4905{
4906 devm_kfree(ice_pf_to_dev(pf), pf->vsi_stats);
4907 pf->vsi_stats = NULL;
4908
4909 pf->num_alloc_vsi = 0;
4910 devm_kfree(ice_pf_to_dev(pf), pf->vsi);
4911 pf->vsi = NULL;
4912}
4913
4914static int ice_init_devlink(struct ice_pf *pf)
4915{
4916 int err;
4917
4918 err = ice_devlink_register_params(pf);
4919 if (err)
4920 return err;
4921
4922 ice_devlink_init_regions(pf);
4923 ice_devlink_register(pf);
4924
4925 return 0;
4926}
4927
4928static void ice_deinit_devlink(struct ice_pf *pf)
4929{
4930 ice_devlink_unregister(pf);
4931 ice_devlink_destroy_regions(pf);
4932 ice_devlink_unregister_params(pf);
4933}
4934
4935static int ice_init(struct ice_pf *pf)
4936{
4937 int err;
4938
4939 err = ice_init_dev(pf);
4940 if (err)
4941 return err;
4942
4943 err = ice_alloc_vsis(pf);
4944 if (err)
4945 goto err_alloc_vsis;
4946
4947 err = ice_init_pf_sw(pf);
4948 if (err)
4949 goto err_init_pf_sw;
4950
4951 ice_init_wakeup(pf);
4952
4953 err = ice_init_link(pf);
4954 if (err)
4955 goto err_init_link;
4956
4957 err = ice_send_version(pf);
4958 if (err)
4959 goto err_init_link;
4960
4961 ice_verify_cacheline_size(pf);
4962
4963 if (ice_is_safe_mode(pf))
4964 ice_set_safe_mode_vlan_cfg(pf);
4965 else
4966 /* print PCI link speed and width */
4967 pcie_print_link_status(pf->pdev);
4968
4969 /* ready to go, so clear down state bit */
4970 clear_bit(ICE_DOWN, pf->state);
4971 clear_bit(ICE_SERVICE_DIS, pf->state);
4972
4973 /* since everything is good, start the service timer */
4974 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4975
4976 return 0;
4977
4978err_init_link:
4979 ice_deinit_pf_sw(pf);
4980err_init_pf_sw:
4981 ice_dealloc_vsis(pf);
4982err_alloc_vsis:
4983 ice_deinit_dev(pf);
4984 return err;
4985}
4986
4987static void ice_deinit(struct ice_pf *pf)
4988{
4989 set_bit(ICE_SERVICE_DIS, pf->state);
4990 set_bit(ICE_DOWN, pf->state);
4991
4992 ice_deinit_pf_sw(pf);
4993 ice_dealloc_vsis(pf);
4994 ice_deinit_dev(pf);
4995}
4996
4997/**
4998 * ice_load - load pf by init hw and starting VSI
4999 * @pf: pointer to the pf instance
5000 *
5001 * This function has to be called under devl_lock.
5002 */
5003int ice_load(struct ice_pf *pf)
5004{
5005 struct ice_vsi *vsi;
5006 int err;
5007
5008 devl_assert_locked(priv_to_devlink(pf));
5009
5010 vsi = ice_get_main_vsi(pf);
5011
5012 /* init channel list */
5013 INIT_LIST_HEAD(&vsi->ch_list);
5014
5015 err = ice_cfg_netdev(vsi);
5016 if (err)
5017 return err;
5018
5019 /* Setup DCB netlink interface */
5020 ice_dcbnl_setup(vsi);
5021
5022 err = ice_init_mac_fltr(pf);
5023 if (err)
5024 goto err_init_mac_fltr;
5025
5026 err = ice_devlink_create_pf_port(pf);
5027 if (err)
5028 goto err_devlink_create_pf_port;
5029
5030 SET_NETDEV_DEVLINK_PORT(vsi->netdev, &pf->devlink_port);
5031
5032 err = ice_register_netdev(vsi);
5033 if (err)
5034 goto err_register_netdev;
5035
5036 err = ice_tc_indir_block_register(vsi);
5037 if (err)
5038 goto err_tc_indir_block_register;
5039
5040 ice_napi_add(vsi);
5041
5042 err = ice_init_rdma(pf);
5043 if (err)
5044 goto err_init_rdma;
5045
5046 ice_init_features(pf);
5047 ice_service_task_restart(pf);
5048
5049 clear_bit(ICE_DOWN, pf->state);
5050
5051 return 0;
5052
5053err_init_rdma:
5054 ice_tc_indir_block_unregister(vsi);
5055err_tc_indir_block_register:
5056 ice_unregister_netdev(vsi);
5057err_register_netdev:
5058 ice_devlink_destroy_pf_port(pf);
5059err_devlink_create_pf_port:
5060err_init_mac_fltr:
5061 ice_decfg_netdev(vsi);
5062 return err;
5063}
5064
5065/**
5066 * ice_unload - unload pf by stopping VSI and deinit hw
5067 * @pf: pointer to the pf instance
5068 *
5069 * This function has to be called under devl_lock.
5070 */
5071void ice_unload(struct ice_pf *pf)
5072{
5073 struct ice_vsi *vsi = ice_get_main_vsi(pf);
5074
5075 devl_assert_locked(priv_to_devlink(pf));
5076
5077 ice_deinit_features(pf);
5078 ice_deinit_rdma(pf);
5079 ice_tc_indir_block_unregister(vsi);
5080 ice_unregister_netdev(vsi);
5081 ice_devlink_destroy_pf_port(pf);
5082 ice_decfg_netdev(vsi);
5083}
5084
5085/**
5086 * ice_probe - Device initialization routine
5087 * @pdev: PCI device information struct
5088 * @ent: entry in ice_pci_tbl
5089 *
5090 * Returns 0 on success, negative on failure
5091 */
5092static int
5093ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
5094{
5095 struct device *dev = &pdev->dev;
5096 struct ice_pf *pf;
5097 struct ice_hw *hw;
5098 int err;
5099
5100 if (pdev->is_virtfn) {
5101 dev_err(dev, "can't probe a virtual function\n");
5102 return -EINVAL;
5103 }
5104
5105 /* when under a kdump kernel initiate a reset before enabling the
5106 * device in order to clear out any pending DMA transactions. These
5107 * transactions can cause some systems to machine check when doing
5108 * the pcim_enable_device() below.
5109 */
5110 if (is_kdump_kernel()) {
5111 pci_save_state(pdev);
5112 pci_clear_master(pdev);
5113 err = pcie_flr(pdev);
5114 if (err)
5115 return err;
5116 pci_restore_state(pdev);
5117 }
5118
5119 /* this driver uses devres, see
5120 * Documentation/driver-api/driver-model/devres.rst
5121 */
5122 err = pcim_enable_device(pdev);
5123 if (err)
5124 return err;
5125
5126 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
5127 if (err) {
5128 dev_err(dev, "BAR0 I/O map error %d\n", err);
5129 return err;
5130 }
5131
5132 pf = ice_allocate_pf(dev);
5133 if (!pf)
5134 return -ENOMEM;
5135
5136 /* initialize Auxiliary index to invalid value */
5137 pf->aux_idx = -1;
5138
5139 /* set up for high or low DMA */
5140 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
5141 if (err) {
5142 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
5143 return err;
5144 }
5145
5146 pci_set_master(pdev);
5147
5148 pf->pdev = pdev;
5149 pci_set_drvdata(pdev, pf);
5150 set_bit(ICE_DOWN, pf->state);
5151 /* Disable service task until DOWN bit is cleared */
5152 set_bit(ICE_SERVICE_DIS, pf->state);
5153
5154 hw = &pf->hw;
5155 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
5156 pci_save_state(pdev);
5157
5158 hw->back = pf;
5159 hw->port_info = NULL;
5160 hw->vendor_id = pdev->vendor;
5161 hw->device_id = pdev->device;
5162 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
5163 hw->subsystem_vendor_id = pdev->subsystem_vendor;
5164 hw->subsystem_device_id = pdev->subsystem_device;
5165 hw->bus.device = PCI_SLOT(pdev->devfn);
5166 hw->bus.func = PCI_FUNC(pdev->devfn);
5167 ice_set_ctrlq_len(hw);
5168
5169 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
5170
5171#ifndef CONFIG_DYNAMIC_DEBUG
5172 if (debug < -1)
5173 hw->debug_mask = debug;
5174#endif
5175
5176 err = ice_init(pf);
5177 if (err)
5178 goto err_init;
5179
5180 devl_lock(priv_to_devlink(pf));
5181 err = ice_load(pf);
5182 devl_unlock(priv_to_devlink(pf));
5183 if (err)
5184 goto err_load;
5185
5186 err = ice_init_devlink(pf);
5187 if (err)
5188 goto err_init_devlink;
5189
5190 return 0;
5191
5192err_init_devlink:
5193 devl_lock(priv_to_devlink(pf));
5194 ice_unload(pf);
5195 devl_unlock(priv_to_devlink(pf));
5196err_load:
5197 ice_deinit(pf);
5198err_init:
5199 pci_disable_device(pdev);
5200 return err;
5201}
5202
5203/**
5204 * ice_set_wake - enable or disable Wake on LAN
5205 * @pf: pointer to the PF struct
5206 *
5207 * Simple helper for WoL control
5208 */
5209static void ice_set_wake(struct ice_pf *pf)
5210{
5211 struct ice_hw *hw = &pf->hw;
5212 bool wol = pf->wol_ena;
5213
5214 /* clear wake state, otherwise new wake events won't fire */
5215 wr32(hw, PFPM_WUS, U32_MAX);
5216
5217 /* enable / disable APM wake up, no RMW needed */
5218 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
5219
5220 /* set magic packet filter enabled */
5221 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
5222}
5223
5224/**
5225 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
5226 * @pf: pointer to the PF struct
5227 *
5228 * Issue firmware command to enable multicast magic wake, making
5229 * sure that any locally administered address (LAA) is used for
5230 * wake, and that PF reset doesn't undo the LAA.
5231 */
5232static void ice_setup_mc_magic_wake(struct ice_pf *pf)
5233{
5234 struct device *dev = ice_pf_to_dev(pf);
5235 struct ice_hw *hw = &pf->hw;
5236 u8 mac_addr[ETH_ALEN];
5237 struct ice_vsi *vsi;
5238 int status;
5239 u8 flags;
5240
5241 if (!pf->wol_ena)
5242 return;
5243
5244 vsi = ice_get_main_vsi(pf);
5245 if (!vsi)
5246 return;
5247
5248 /* Get current MAC address in case it's an LAA */
5249 if (vsi->netdev)
5250 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
5251 else
5252 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
5253
5254 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
5255 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
5256 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
5257
5258 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
5259 if (status)
5260 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
5261 status, ice_aq_str(hw->adminq.sq_last_status));
5262}
5263
5264/**
5265 * ice_remove - Device removal routine
5266 * @pdev: PCI device information struct
5267 */
5268static void ice_remove(struct pci_dev *pdev)
5269{
5270 struct ice_pf *pf = pci_get_drvdata(pdev);
5271 int i;
5272
5273 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
5274 if (!ice_is_reset_in_progress(pf->state))
5275 break;
5276 msleep(100);
5277 }
5278
5279 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
5280 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5281 ice_free_vfs(pf);
5282 }
5283
5284 ice_hwmon_exit(pf);
5285
5286 ice_service_task_stop(pf);
5287 ice_aq_cancel_waiting_tasks(pf);
5288 set_bit(ICE_DOWN, pf->state);
5289
5290 if (!ice_is_safe_mode(pf))
5291 ice_remove_arfs(pf);
5292
5293 ice_deinit_devlink(pf);
5294
5295 devl_lock(priv_to_devlink(pf));
5296 ice_unload(pf);
5297 devl_unlock(priv_to_devlink(pf));
5298
5299 ice_deinit(pf);
5300 ice_vsi_release_all(pf);
5301
5302 ice_setup_mc_magic_wake(pf);
5303 ice_set_wake(pf);
5304
5305 pci_disable_device(pdev);
5306}
5307
5308/**
5309 * ice_shutdown - PCI callback for shutting down device
5310 * @pdev: PCI device information struct
5311 */
5312static void ice_shutdown(struct pci_dev *pdev)
5313{
5314 struct ice_pf *pf = pci_get_drvdata(pdev);
5315
5316 ice_remove(pdev);
5317
5318 if (system_state == SYSTEM_POWER_OFF) {
5319 pci_wake_from_d3(pdev, pf->wol_ena);
5320 pci_set_power_state(pdev, PCI_D3hot);
5321 }
5322}
5323
5324#ifdef CONFIG_PM
5325/**
5326 * ice_prepare_for_shutdown - prep for PCI shutdown
5327 * @pf: board private structure
5328 *
5329 * Inform or close all dependent features in prep for PCI device shutdown
5330 */
5331static void ice_prepare_for_shutdown(struct ice_pf *pf)
5332{
5333 struct ice_hw *hw = &pf->hw;
5334 u32 v;
5335
5336 /* Notify VFs of impending reset */
5337 if (ice_check_sq_alive(hw, &hw->mailboxq))
5338 ice_vc_notify_reset(pf);
5339
5340 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5341
5342 /* disable the VSIs and their queues that are not already DOWN */
5343 ice_pf_dis_all_vsi(pf, false);
5344
5345 ice_for_each_vsi(pf, v)
5346 if (pf->vsi[v])
5347 pf->vsi[v]->vsi_num = 0;
5348
5349 ice_shutdown_all_ctrlq(hw);
5350}
5351
5352/**
5353 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5354 * @pf: board private structure to reinitialize
5355 *
5356 * This routine reinitialize interrupt scheme that was cleared during
5357 * power management suspend callback.
5358 *
5359 * This should be called during resume routine to re-allocate the q_vectors
5360 * and reacquire interrupts.
5361 */
5362static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5363{
5364 struct device *dev = ice_pf_to_dev(pf);
5365 int ret, v;
5366
5367 /* Since we clear MSIX flag during suspend, we need to
5368 * set it back during resume...
5369 */
5370
5371 ret = ice_init_interrupt_scheme(pf);
5372 if (ret) {
5373 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5374 return ret;
5375 }
5376
5377 /* Remap vectors and rings, after successful re-init interrupts */
5378 ice_for_each_vsi(pf, v) {
5379 if (!pf->vsi[v])
5380 continue;
5381
5382 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5383 if (ret)
5384 goto err_reinit;
5385 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5386 ice_vsi_set_napi_queues(pf->vsi[v]);
5387 }
5388
5389 ret = ice_req_irq_msix_misc(pf);
5390 if (ret) {
5391 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5392 ret);
5393 goto err_reinit;
5394 }
5395
5396 return 0;
5397
5398err_reinit:
5399 while (v--)
5400 if (pf->vsi[v])
5401 ice_vsi_free_q_vectors(pf->vsi[v]);
5402
5403 return ret;
5404}
5405
5406/**
5407 * ice_suspend
5408 * @dev: generic device information structure
5409 *
5410 * Power Management callback to quiesce the device and prepare
5411 * for D3 transition.
5412 */
5413static int __maybe_unused ice_suspend(struct device *dev)
5414{
5415 struct pci_dev *pdev = to_pci_dev(dev);
5416 struct ice_pf *pf;
5417 int disabled, v;
5418
5419 pf = pci_get_drvdata(pdev);
5420
5421 if (!ice_pf_state_is_nominal(pf)) {
5422 dev_err(dev, "Device is not ready, no need to suspend it\n");
5423 return -EBUSY;
5424 }
5425
5426 /* Stop watchdog tasks until resume completion.
5427 * Even though it is most likely that the service task is
5428 * disabled if the device is suspended or down, the service task's
5429 * state is controlled by a different state bit, and we should
5430 * store and honor whatever state that bit is in at this point.
5431 */
5432 disabled = ice_service_task_stop(pf);
5433
5434 ice_unplug_aux_dev(pf);
5435
5436 /* Already suspended?, then there is nothing to do */
5437 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5438 if (!disabled)
5439 ice_service_task_restart(pf);
5440 return 0;
5441 }
5442
5443 if (test_bit(ICE_DOWN, pf->state) ||
5444 ice_is_reset_in_progress(pf->state)) {
5445 dev_err(dev, "can't suspend device in reset or already down\n");
5446 if (!disabled)
5447 ice_service_task_restart(pf);
5448 return 0;
5449 }
5450
5451 ice_setup_mc_magic_wake(pf);
5452
5453 ice_prepare_for_shutdown(pf);
5454
5455 ice_set_wake(pf);
5456
5457 /* Free vectors, clear the interrupt scheme and release IRQs
5458 * for proper hibernation, especially with large number of CPUs.
5459 * Otherwise hibernation might fail when mapping all the vectors back
5460 * to CPU0.
5461 */
5462 ice_free_irq_msix_misc(pf);
5463 ice_for_each_vsi(pf, v) {
5464 if (!pf->vsi[v])
5465 continue;
5466 ice_vsi_free_q_vectors(pf->vsi[v]);
5467 }
5468 ice_clear_interrupt_scheme(pf);
5469
5470 pci_save_state(pdev);
5471 pci_wake_from_d3(pdev, pf->wol_ena);
5472 pci_set_power_state(pdev, PCI_D3hot);
5473 return 0;
5474}
5475
5476/**
5477 * ice_resume - PM callback for waking up from D3
5478 * @dev: generic device information structure
5479 */
5480static int __maybe_unused ice_resume(struct device *dev)
5481{
5482 struct pci_dev *pdev = to_pci_dev(dev);
5483 enum ice_reset_req reset_type;
5484 struct ice_pf *pf;
5485 struct ice_hw *hw;
5486 int ret;
5487
5488 pci_set_power_state(pdev, PCI_D0);
5489 pci_restore_state(pdev);
5490 pci_save_state(pdev);
5491
5492 if (!pci_device_is_present(pdev))
5493 return -ENODEV;
5494
5495 ret = pci_enable_device_mem(pdev);
5496 if (ret) {
5497 dev_err(dev, "Cannot enable device after suspend\n");
5498 return ret;
5499 }
5500
5501 pf = pci_get_drvdata(pdev);
5502 hw = &pf->hw;
5503
5504 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5505 ice_print_wake_reason(pf);
5506
5507 /* We cleared the interrupt scheme when we suspended, so we need to
5508 * restore it now to resume device functionality.
5509 */
5510 ret = ice_reinit_interrupt_scheme(pf);
5511 if (ret)
5512 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5513
5514 clear_bit(ICE_DOWN, pf->state);
5515 /* Now perform PF reset and rebuild */
5516 reset_type = ICE_RESET_PFR;
5517 /* re-enable service task for reset, but allow reset to schedule it */
5518 clear_bit(ICE_SERVICE_DIS, pf->state);
5519
5520 if (ice_schedule_reset(pf, reset_type))
5521 dev_err(dev, "Reset during resume failed.\n");
5522
5523 clear_bit(ICE_SUSPENDED, pf->state);
5524 ice_service_task_restart(pf);
5525
5526 /* Restart the service task */
5527 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5528
5529 return 0;
5530}
5531#endif /* CONFIG_PM */
5532
5533/**
5534 * ice_pci_err_detected - warning that PCI error has been detected
5535 * @pdev: PCI device information struct
5536 * @err: the type of PCI error
5537 *
5538 * Called to warn that something happened on the PCI bus and the error handling
5539 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5540 */
5541static pci_ers_result_t
5542ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5543{
5544 struct ice_pf *pf = pci_get_drvdata(pdev);
5545
5546 if (!pf) {
5547 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5548 __func__, err);
5549 return PCI_ERS_RESULT_DISCONNECT;
5550 }
5551
5552 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5553 ice_service_task_stop(pf);
5554
5555 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5556 set_bit(ICE_PFR_REQ, pf->state);
5557 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5558 }
5559 }
5560
5561 return PCI_ERS_RESULT_NEED_RESET;
5562}
5563
5564/**
5565 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5566 * @pdev: PCI device information struct
5567 *
5568 * Called to determine if the driver can recover from the PCI slot reset by
5569 * using a register read to determine if the device is recoverable.
5570 */
5571static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5572{
5573 struct ice_pf *pf = pci_get_drvdata(pdev);
5574 pci_ers_result_t result;
5575 int err;
5576 u32 reg;
5577
5578 err = pci_enable_device_mem(pdev);
5579 if (err) {
5580 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5581 err);
5582 result = PCI_ERS_RESULT_DISCONNECT;
5583 } else {
5584 pci_set_master(pdev);
5585 pci_restore_state(pdev);
5586 pci_save_state(pdev);
5587 pci_wake_from_d3(pdev, false);
5588
5589 /* Check for life */
5590 reg = rd32(&pf->hw, GLGEN_RTRIG);
5591 if (!reg)
5592 result = PCI_ERS_RESULT_RECOVERED;
5593 else
5594 result = PCI_ERS_RESULT_DISCONNECT;
5595 }
5596
5597 return result;
5598}
5599
5600/**
5601 * ice_pci_err_resume - restart operations after PCI error recovery
5602 * @pdev: PCI device information struct
5603 *
5604 * Called to allow the driver to bring things back up after PCI error and/or
5605 * reset recovery have finished
5606 */
5607static void ice_pci_err_resume(struct pci_dev *pdev)
5608{
5609 struct ice_pf *pf = pci_get_drvdata(pdev);
5610
5611 if (!pf) {
5612 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5613 __func__);
5614 return;
5615 }
5616
5617 if (test_bit(ICE_SUSPENDED, pf->state)) {
5618 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5619 __func__);
5620 return;
5621 }
5622
5623 ice_restore_all_vfs_msi_state(pf);
5624
5625 ice_do_reset(pf, ICE_RESET_PFR);
5626 ice_service_task_restart(pf);
5627 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5628}
5629
5630/**
5631 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5632 * @pdev: PCI device information struct
5633 */
5634static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5635{
5636 struct ice_pf *pf = pci_get_drvdata(pdev);
5637
5638 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5639 ice_service_task_stop(pf);
5640
5641 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5642 set_bit(ICE_PFR_REQ, pf->state);
5643 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5644 }
5645 }
5646}
5647
5648/**
5649 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5650 * @pdev: PCI device information struct
5651 */
5652static void ice_pci_err_reset_done(struct pci_dev *pdev)
5653{
5654 ice_pci_err_resume(pdev);
5655}
5656
5657/* ice_pci_tbl - PCI Device ID Table
5658 *
5659 * Wildcard entries (PCI_ANY_ID) should come last
5660 * Last entry must be all 0s
5661 *
5662 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5663 * Class, Class Mask, private data (not used) }
5664 */
5665static const struct pci_device_id ice_pci_tbl[] = {
5666 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE) },
5667 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP) },
5668 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP) },
5669 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE) },
5670 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP) },
5671 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP) },
5672 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE) },
5673 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP) },
5674 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP) },
5675 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T) },
5676 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII) },
5677 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE) },
5678 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP) },
5679 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP) },
5680 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T) },
5681 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII) },
5682 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE) },
5683 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP) },
5684 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T) },
5685 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII) },
5686 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE) },
5687 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP) },
5688 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T) },
5689 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE) },
5690 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP) },
5691 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT) },
5692 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_BACKPLANE), },
5693 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_QSFP), },
5694 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_SFP), },
5695 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E825C_SGMII), },
5696 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_BACKPLANE) },
5697 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_QSFP56) },
5698 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP) },
5699 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E830_SFP_DD) },
5700 /* required last entry */
5701 {}
5702};
5703MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5704
5705static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5706
5707static const struct pci_error_handlers ice_pci_err_handler = {
5708 .error_detected = ice_pci_err_detected,
5709 .slot_reset = ice_pci_err_slot_reset,
5710 .reset_prepare = ice_pci_err_reset_prepare,
5711 .reset_done = ice_pci_err_reset_done,
5712 .resume = ice_pci_err_resume
5713};
5714
5715static struct pci_driver ice_driver = {
5716 .name = KBUILD_MODNAME,
5717 .id_table = ice_pci_tbl,
5718 .probe = ice_probe,
5719 .remove = ice_remove,
5720#ifdef CONFIG_PM
5721 .driver.pm = &ice_pm_ops,
5722#endif /* CONFIG_PM */
5723 .shutdown = ice_shutdown,
5724 .sriov_configure = ice_sriov_configure,
5725 .sriov_get_vf_total_msix = ice_sriov_get_vf_total_msix,
5726 .sriov_set_msix_vec_count = ice_sriov_set_msix_vec_count,
5727 .err_handler = &ice_pci_err_handler
5728};
5729
5730/**
5731 * ice_module_init - Driver registration routine
5732 *
5733 * ice_module_init is the first routine called when the driver is
5734 * loaded. All it does is register with the PCI subsystem.
5735 */
5736static int __init ice_module_init(void)
5737{
5738 int status = -ENOMEM;
5739
5740 pr_info("%s\n", ice_driver_string);
5741 pr_info("%s\n", ice_copyright);
5742
5743 ice_adv_lnk_speed_maps_init();
5744
5745 ice_wq = alloc_workqueue("%s", 0, 0, KBUILD_MODNAME);
5746 if (!ice_wq) {
5747 pr_err("Failed to create workqueue\n");
5748 return status;
5749 }
5750
5751 ice_lag_wq = alloc_ordered_workqueue("ice_lag_wq", 0);
5752 if (!ice_lag_wq) {
5753 pr_err("Failed to create LAG workqueue\n");
5754 goto err_dest_wq;
5755 }
5756
5757 ice_debugfs_init();
5758
5759 status = pci_register_driver(&ice_driver);
5760 if (status) {
5761 pr_err("failed to register PCI driver, err %d\n", status);
5762 goto err_dest_lag_wq;
5763 }
5764
5765 return 0;
5766
5767err_dest_lag_wq:
5768 destroy_workqueue(ice_lag_wq);
5769 ice_debugfs_exit();
5770err_dest_wq:
5771 destroy_workqueue(ice_wq);
5772 return status;
5773}
5774module_init(ice_module_init);
5775
5776/**
5777 * ice_module_exit - Driver exit cleanup routine
5778 *
5779 * ice_module_exit is called just before the driver is removed
5780 * from memory.
5781 */
5782static void __exit ice_module_exit(void)
5783{
5784 pci_unregister_driver(&ice_driver);
5785 ice_debugfs_exit();
5786 destroy_workqueue(ice_wq);
5787 destroy_workqueue(ice_lag_wq);
5788 pr_info("module unloaded\n");
5789}
5790module_exit(ice_module_exit);
5791
5792/**
5793 * ice_set_mac_address - NDO callback to set MAC address
5794 * @netdev: network interface device structure
5795 * @pi: pointer to an address structure
5796 *
5797 * Returns 0 on success, negative on failure
5798 */
5799static int ice_set_mac_address(struct net_device *netdev, void *pi)
5800{
5801 struct ice_netdev_priv *np = netdev_priv(netdev);
5802 struct ice_vsi *vsi = np->vsi;
5803 struct ice_pf *pf = vsi->back;
5804 struct ice_hw *hw = &pf->hw;
5805 struct sockaddr *addr = pi;
5806 u8 old_mac[ETH_ALEN];
5807 u8 flags = 0;
5808 u8 *mac;
5809 int err;
5810
5811 mac = (u8 *)addr->sa_data;
5812
5813 if (!is_valid_ether_addr(mac))
5814 return -EADDRNOTAVAIL;
5815
5816 if (test_bit(ICE_DOWN, pf->state) ||
5817 ice_is_reset_in_progress(pf->state)) {
5818 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5819 mac);
5820 return -EBUSY;
5821 }
5822
5823 if (ice_chnl_dmac_fltr_cnt(pf)) {
5824 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5825 mac);
5826 return -EAGAIN;
5827 }
5828
5829 netif_addr_lock_bh(netdev);
5830 ether_addr_copy(old_mac, netdev->dev_addr);
5831 /* change the netdev's MAC address */
5832 eth_hw_addr_set(netdev, mac);
5833 netif_addr_unlock_bh(netdev);
5834
5835 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5836 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5837 if (err && err != -ENOENT) {
5838 err = -EADDRNOTAVAIL;
5839 goto err_update_filters;
5840 }
5841
5842 /* Add filter for new MAC. If filter exists, return success */
5843 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5844 if (err == -EEXIST) {
5845 /* Although this MAC filter is already present in hardware it's
5846 * possible in some cases (e.g. bonding) that dev_addr was
5847 * modified outside of the driver and needs to be restored back
5848 * to this value.
5849 */
5850 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5851
5852 return 0;
5853 } else if (err) {
5854 /* error if the new filter addition failed */
5855 err = -EADDRNOTAVAIL;
5856 }
5857
5858err_update_filters:
5859 if (err) {
5860 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5861 mac);
5862 netif_addr_lock_bh(netdev);
5863 eth_hw_addr_set(netdev, old_mac);
5864 netif_addr_unlock_bh(netdev);
5865 return err;
5866 }
5867
5868 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5869 netdev->dev_addr);
5870
5871 /* write new MAC address to the firmware */
5872 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5873 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5874 if (err) {
5875 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5876 mac, err);
5877 }
5878 return 0;
5879}
5880
5881/**
5882 * ice_set_rx_mode - NDO callback to set the netdev filters
5883 * @netdev: network interface device structure
5884 */
5885static void ice_set_rx_mode(struct net_device *netdev)
5886{
5887 struct ice_netdev_priv *np = netdev_priv(netdev);
5888 struct ice_vsi *vsi = np->vsi;
5889
5890 if (!vsi || ice_is_switchdev_running(vsi->back))
5891 return;
5892
5893 /* Set the flags to synchronize filters
5894 * ndo_set_rx_mode may be triggered even without a change in netdev
5895 * flags
5896 */
5897 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5898 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5899 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5900
5901 /* schedule our worker thread which will take care of
5902 * applying the new filter changes
5903 */
5904 ice_service_task_schedule(vsi->back);
5905}
5906
5907/**
5908 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5909 * @netdev: network interface device structure
5910 * @queue_index: Queue ID
5911 * @maxrate: maximum bandwidth in Mbps
5912 */
5913static int
5914ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5915{
5916 struct ice_netdev_priv *np = netdev_priv(netdev);
5917 struct ice_vsi *vsi = np->vsi;
5918 u16 q_handle;
5919 int status;
5920 u8 tc;
5921
5922 /* Validate maxrate requested is within permitted range */
5923 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5924 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5925 maxrate, queue_index);
5926 return -EINVAL;
5927 }
5928
5929 q_handle = vsi->tx_rings[queue_index]->q_handle;
5930 tc = ice_dcb_get_tc(vsi, queue_index);
5931
5932 vsi = ice_locate_vsi_using_queue(vsi, queue_index);
5933 if (!vsi) {
5934 netdev_err(netdev, "Invalid VSI for given queue %d\n",
5935 queue_index);
5936 return -EINVAL;
5937 }
5938
5939 /* Set BW back to default, when user set maxrate to 0 */
5940 if (!maxrate)
5941 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5942 q_handle, ICE_MAX_BW);
5943 else
5944 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5945 q_handle, ICE_MAX_BW, maxrate * 1000);
5946 if (status)
5947 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5948 status);
5949
5950 return status;
5951}
5952
5953/**
5954 * ice_fdb_add - add an entry to the hardware database
5955 * @ndm: the input from the stack
5956 * @tb: pointer to array of nladdr (unused)
5957 * @dev: the net device pointer
5958 * @addr: the MAC address entry being added
5959 * @vid: VLAN ID
5960 * @flags: instructions from stack about fdb operation
5961 * @extack: netlink extended ack
5962 */
5963static int
5964ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5965 struct net_device *dev, const unsigned char *addr, u16 vid,
5966 u16 flags, struct netlink_ext_ack __always_unused *extack)
5967{
5968 int err;
5969
5970 if (vid) {
5971 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5972 return -EINVAL;
5973 }
5974 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5975 netdev_err(dev, "FDB only supports static addresses\n");
5976 return -EINVAL;
5977 }
5978
5979 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5980 err = dev_uc_add_excl(dev, addr);
5981 else if (is_multicast_ether_addr(addr))
5982 err = dev_mc_add_excl(dev, addr);
5983 else
5984 err = -EINVAL;
5985
5986 /* Only return duplicate errors if NLM_F_EXCL is set */
5987 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5988 err = 0;
5989
5990 return err;
5991}
5992
5993/**
5994 * ice_fdb_del - delete an entry from the hardware database
5995 * @ndm: the input from the stack
5996 * @tb: pointer to array of nladdr (unused)
5997 * @dev: the net device pointer
5998 * @addr: the MAC address entry being added
5999 * @vid: VLAN ID
6000 * @extack: netlink extended ack
6001 */
6002static int
6003ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
6004 struct net_device *dev, const unsigned char *addr,
6005 __always_unused u16 vid, struct netlink_ext_ack *extack)
6006{
6007 int err;
6008
6009 if (ndm->ndm_state & NUD_PERMANENT) {
6010 netdev_err(dev, "FDB only supports static addresses\n");
6011 return -EINVAL;
6012 }
6013
6014 if (is_unicast_ether_addr(addr))
6015 err = dev_uc_del(dev, addr);
6016 else if (is_multicast_ether_addr(addr))
6017 err = dev_mc_del(dev, addr);
6018 else
6019 err = -EINVAL;
6020
6021 return err;
6022}
6023
6024#define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6025 NETIF_F_HW_VLAN_CTAG_TX | \
6026 NETIF_F_HW_VLAN_STAG_RX | \
6027 NETIF_F_HW_VLAN_STAG_TX)
6028
6029#define NETIF_VLAN_STRIPPING_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6030 NETIF_F_HW_VLAN_STAG_RX)
6031
6032#define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
6033 NETIF_F_HW_VLAN_STAG_FILTER)
6034
6035/**
6036 * ice_fix_features - fix the netdev features flags based on device limitations
6037 * @netdev: ptr to the netdev that flags are being fixed on
6038 * @features: features that need to be checked and possibly fixed
6039 *
6040 * Make sure any fixups are made to features in this callback. This enables the
6041 * driver to not have to check unsupported configurations throughout the driver
6042 * because that's the responsiblity of this callback.
6043 *
6044 * Single VLAN Mode (SVM) Supported Features:
6045 * NETIF_F_HW_VLAN_CTAG_FILTER
6046 * NETIF_F_HW_VLAN_CTAG_RX
6047 * NETIF_F_HW_VLAN_CTAG_TX
6048 *
6049 * Double VLAN Mode (DVM) Supported Features:
6050 * NETIF_F_HW_VLAN_CTAG_FILTER
6051 * NETIF_F_HW_VLAN_CTAG_RX
6052 * NETIF_F_HW_VLAN_CTAG_TX
6053 *
6054 * NETIF_F_HW_VLAN_STAG_FILTER
6055 * NETIF_HW_VLAN_STAG_RX
6056 * NETIF_HW_VLAN_STAG_TX
6057 *
6058 * Features that need fixing:
6059 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
6060 * These are mutually exlusive as the VSI context cannot support multiple
6061 * VLAN ethertypes simultaneously for stripping and/or insertion. If this
6062 * is not done, then default to clearing the requested STAG offload
6063 * settings.
6064 *
6065 * All supported filtering has to be enabled or disabled together. For
6066 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled
6067 * together. If this is not done, then default to VLAN filtering disabled.
6068 * These are mutually exclusive as there is currently no way to
6069 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN
6070 * prune rules.
6071 */
6072static netdev_features_t
6073ice_fix_features(struct net_device *netdev, netdev_features_t features)
6074{
6075 struct ice_netdev_priv *np = netdev_priv(netdev);
6076 netdev_features_t req_vlan_fltr, cur_vlan_fltr;
6077 bool cur_ctag, cur_stag, req_ctag, req_stag;
6078
6079 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
6080 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6081 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6082
6083 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
6084 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6085 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6086
6087 if (req_vlan_fltr != cur_vlan_fltr) {
6088 if (ice_is_dvm_ena(&np->vsi->back->hw)) {
6089 if (req_ctag && req_stag) {
6090 features |= NETIF_VLAN_FILTERING_FEATURES;
6091 } else if (!req_ctag && !req_stag) {
6092 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6093 } else if ((!cur_ctag && req_ctag && !cur_stag) ||
6094 (!cur_stag && req_stag && !cur_ctag)) {
6095 features |= NETIF_VLAN_FILTERING_FEATURES;
6096 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");
6097 } else if ((cur_ctag && !req_ctag && cur_stag) ||
6098 (cur_stag && !req_stag && cur_ctag)) {
6099 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6100 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");
6101 }
6102 } else {
6103 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
6104 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
6105
6106 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
6107 features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6108 }
6109 }
6110
6111 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
6112 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
6113 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");
6114 features &= ~(NETIF_F_HW_VLAN_STAG_RX |
6115 NETIF_F_HW_VLAN_STAG_TX);
6116 }
6117
6118 if (!(netdev->features & NETIF_F_RXFCS) &&
6119 (features & NETIF_F_RXFCS) &&
6120 (features & NETIF_VLAN_STRIPPING_FEATURES) &&
6121 !ice_vsi_has_non_zero_vlans(np->vsi)) {
6122 netdev_warn(netdev, "Disabling VLAN stripping as FCS/CRC stripping is also disabled and there is no VLAN configured\n");
6123 features &= ~NETIF_VLAN_STRIPPING_FEATURES;
6124 }
6125
6126 return features;
6127}
6128
6129/**
6130 * ice_set_rx_rings_vlan_proto - update rings with new stripped VLAN proto
6131 * @vsi: PF's VSI
6132 * @vlan_ethertype: VLAN ethertype (802.1Q or 802.1ad) in network byte order
6133 *
6134 * Store current stripped VLAN proto in ring packet context,
6135 * so it can be accessed more efficiently by packet processing code.
6136 */
6137static void
6138ice_set_rx_rings_vlan_proto(struct ice_vsi *vsi, __be16 vlan_ethertype)
6139{
6140 u16 i;
6141
6142 ice_for_each_alloc_rxq(vsi, i)
6143 vsi->rx_rings[i]->pkt_ctx.vlan_proto = vlan_ethertype;
6144}
6145
6146/**
6147 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
6148 * @vsi: PF's VSI
6149 * @features: features used to determine VLAN offload settings
6150 *
6151 * First, determine the vlan_ethertype based on the VLAN offload bits in
6152 * features. Then determine if stripping and insertion should be enabled or
6153 * disabled. Finally enable or disable VLAN stripping and insertion.
6154 */
6155static int
6156ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
6157{
6158 bool enable_stripping = true, enable_insertion = true;
6159 struct ice_vsi_vlan_ops *vlan_ops;
6160 int strip_err = 0, insert_err = 0;
6161 u16 vlan_ethertype = 0;
6162
6163 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6164
6165 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
6166 vlan_ethertype = ETH_P_8021AD;
6167 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
6168 vlan_ethertype = ETH_P_8021Q;
6169
6170 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
6171 enable_stripping = false;
6172 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
6173 enable_insertion = false;
6174
6175 if (enable_stripping)
6176 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
6177 else
6178 strip_err = vlan_ops->dis_stripping(vsi);
6179
6180 if (enable_insertion)
6181 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
6182 else
6183 insert_err = vlan_ops->dis_insertion(vsi);
6184
6185 if (strip_err || insert_err)
6186 return -EIO;
6187
6188 ice_set_rx_rings_vlan_proto(vsi, enable_stripping ?
6189 htons(vlan_ethertype) : 0);
6190
6191 return 0;
6192}
6193
6194/**
6195 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
6196 * @vsi: PF's VSI
6197 * @features: features used to determine VLAN filtering settings
6198 *
6199 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
6200 * features.
6201 */
6202static int
6203ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
6204{
6205 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6206 int err = 0;
6207
6208 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
6209 * if either bit is set
6210 */
6211 if (features &
6212 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
6213 err = vlan_ops->ena_rx_filtering(vsi);
6214 else
6215 err = vlan_ops->dis_rx_filtering(vsi);
6216
6217 return err;
6218}
6219
6220/**
6221 * ice_set_vlan_features - set VLAN settings based on suggested feature set
6222 * @netdev: ptr to the netdev being adjusted
6223 * @features: the feature set that the stack is suggesting
6224 *
6225 * Only update VLAN settings if the requested_vlan_features are different than
6226 * the current_vlan_features.
6227 */
6228static int
6229ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
6230{
6231 netdev_features_t current_vlan_features, requested_vlan_features;
6232 struct ice_netdev_priv *np = netdev_priv(netdev);
6233 struct ice_vsi *vsi = np->vsi;
6234 int err;
6235
6236 current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
6237 requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
6238 if (current_vlan_features ^ requested_vlan_features) {
6239 if ((features & NETIF_F_RXFCS) &&
6240 (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6241 dev_err(ice_pf_to_dev(vsi->back),
6242 "To enable VLAN stripping, you must first enable FCS/CRC stripping\n");
6243 return -EIO;
6244 }
6245
6246 err = ice_set_vlan_offload_features(vsi, features);
6247 if (err)
6248 return err;
6249 }
6250
6251 current_vlan_features = netdev->features &
6252 NETIF_VLAN_FILTERING_FEATURES;
6253 requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
6254 if (current_vlan_features ^ requested_vlan_features) {
6255 err = ice_set_vlan_filtering_features(vsi, features);
6256 if (err)
6257 return err;
6258 }
6259
6260 return 0;
6261}
6262
6263/**
6264 * ice_set_loopback - turn on/off loopback mode on underlying PF
6265 * @vsi: ptr to VSI
6266 * @ena: flag to indicate the on/off setting
6267 */
6268static int ice_set_loopback(struct ice_vsi *vsi, bool ena)
6269{
6270 bool if_running = netif_running(vsi->netdev);
6271 int ret;
6272
6273 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6274 ret = ice_down(vsi);
6275 if (ret) {
6276 netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
6277 return ret;
6278 }
6279 }
6280 ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
6281 if (ret)
6282 netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
6283 if (if_running)
6284 ret = ice_up(vsi);
6285
6286 return ret;
6287}
6288
6289/**
6290 * ice_set_features - set the netdev feature flags
6291 * @netdev: ptr to the netdev being adjusted
6292 * @features: the feature set that the stack is suggesting
6293 */
6294static int
6295ice_set_features(struct net_device *netdev, netdev_features_t features)
6296{
6297 netdev_features_t changed = netdev->features ^ features;
6298 struct ice_netdev_priv *np = netdev_priv(netdev);
6299 struct ice_vsi *vsi = np->vsi;
6300 struct ice_pf *pf = vsi->back;
6301 int ret = 0;
6302
6303 /* Don't set any netdev advanced features with device in Safe Mode */
6304 if (ice_is_safe_mode(pf)) {
6305 dev_err(ice_pf_to_dev(pf),
6306 "Device is in Safe Mode - not enabling advanced netdev features\n");
6307 return ret;
6308 }
6309
6310 /* Do not change setting during reset */
6311 if (ice_is_reset_in_progress(pf->state)) {
6312 dev_err(ice_pf_to_dev(pf),
6313 "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
6314 return -EBUSY;
6315 }
6316
6317 /* Multiple features can be changed in one call so keep features in
6318 * separate if/else statements to guarantee each feature is checked
6319 */
6320 if (changed & NETIF_F_RXHASH)
6321 ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH));
6322
6323 ret = ice_set_vlan_features(netdev, features);
6324 if (ret)
6325 return ret;
6326
6327 /* Turn on receive of FCS aka CRC, and after setting this
6328 * flag the packet data will have the 4 byte CRC appended
6329 */
6330 if (changed & NETIF_F_RXFCS) {
6331 if ((features & NETIF_F_RXFCS) &&
6332 (features & NETIF_VLAN_STRIPPING_FEATURES)) {
6333 dev_err(ice_pf_to_dev(vsi->back),
6334 "To disable FCS/CRC stripping, you must first disable VLAN stripping\n");
6335 return -EIO;
6336 }
6337
6338 ice_vsi_cfg_crc_strip(vsi, !!(features & NETIF_F_RXFCS));
6339 ret = ice_down_up(vsi);
6340 if (ret)
6341 return ret;
6342 }
6343
6344 if (changed & NETIF_F_NTUPLE) {
6345 bool ena = !!(features & NETIF_F_NTUPLE);
6346
6347 ice_vsi_manage_fdir(vsi, ena);
6348 ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi);
6349 }
6350
6351 /* don't turn off hw_tc_offload when ADQ is already enabled */
6352 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
6353 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
6354 return -EACCES;
6355 }
6356
6357 if (changed & NETIF_F_HW_TC) {
6358 bool ena = !!(features & NETIF_F_HW_TC);
6359
6360 ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) :
6361 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6362 }
6363
6364 if (changed & NETIF_F_LOOPBACK)
6365 ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
6366
6367 return ret;
6368}
6369
6370/**
6371 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6372 * @vsi: VSI to setup VLAN properties for
6373 */
6374static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6375{
6376 int err;
6377
6378 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6379 if (err)
6380 return err;
6381
6382 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6383 if (err)
6384 return err;
6385
6386 return ice_vsi_add_vlan_zero(vsi);
6387}
6388
6389/**
6390 * ice_vsi_cfg_lan - Setup the VSI lan related config
6391 * @vsi: the VSI being configured
6392 *
6393 * Return 0 on success and negative value on error
6394 */
6395int ice_vsi_cfg_lan(struct ice_vsi *vsi)
6396{
6397 int err;
6398
6399 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6400 ice_set_rx_mode(vsi->netdev);
6401
6402 err = ice_vsi_vlan_setup(vsi);
6403 if (err)
6404 return err;
6405 }
6406 ice_vsi_cfg_dcb_rings(vsi);
6407
6408 err = ice_vsi_cfg_lan_txqs(vsi);
6409 if (!err && ice_is_xdp_ena_vsi(vsi))
6410 err = ice_vsi_cfg_xdp_txqs(vsi);
6411 if (!err)
6412 err = ice_vsi_cfg_rxqs(vsi);
6413
6414 return err;
6415}
6416
6417/* THEORY OF MODERATION:
6418 * The ice driver hardware works differently than the hardware that DIMLIB was
6419 * originally made for. ice hardware doesn't have packet count limits that
6420 * can trigger an interrupt, but it *does* have interrupt rate limit support,
6421 * which is hard-coded to a limit of 250,000 ints/second.
6422 * If not using dynamic moderation, the INTRL value can be modified
6423 * by ethtool rx-usecs-high.
6424 */
6425struct ice_dim {
6426 /* the throttle rate for interrupts, basically worst case delay before
6427 * an initial interrupt fires, value is stored in microseconds.
6428 */
6429 u16 itr;
6430};
6431
6432/* Make a different profile for Rx that doesn't allow quite so aggressive
6433 * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6434 * second.
6435 */
6436static const struct ice_dim rx_profile[] = {
6437 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6438 {8}, /* 125,000 ints/s */
6439 {16}, /* 62,500 ints/s */
6440 {62}, /* 16,129 ints/s */
6441 {126} /* 7,936 ints/s */
6442};
6443
6444/* The transmit profile, which has the same sorts of values
6445 * as the previous struct
6446 */
6447static const struct ice_dim tx_profile[] = {
6448 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6449 {8}, /* 125,000 ints/s */
6450 {40}, /* 16,125 ints/s */
6451 {128}, /* 7,812 ints/s */
6452 {256} /* 3,906 ints/s */
6453};
6454
6455static void ice_tx_dim_work(struct work_struct *work)
6456{
6457 struct ice_ring_container *rc;
6458 struct dim *dim;
6459 u16 itr;
6460
6461 dim = container_of(work, struct dim, work);
6462 rc = dim->priv;
6463
6464 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6465
6466 /* look up the values in our local table */
6467 itr = tx_profile[dim->profile_ix].itr;
6468
6469 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6470 ice_write_itr(rc, itr);
6471
6472 dim->state = DIM_START_MEASURE;
6473}
6474
6475static void ice_rx_dim_work(struct work_struct *work)
6476{
6477 struct ice_ring_container *rc;
6478 struct dim *dim;
6479 u16 itr;
6480
6481 dim = container_of(work, struct dim, work);
6482 rc = dim->priv;
6483
6484 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6485
6486 /* look up the values in our local table */
6487 itr = rx_profile[dim->profile_ix].itr;
6488
6489 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6490 ice_write_itr(rc, itr);
6491
6492 dim->state = DIM_START_MEASURE;
6493}
6494
6495#define ICE_DIM_DEFAULT_PROFILE_IX 1
6496
6497/**
6498 * ice_init_moderation - set up interrupt moderation
6499 * @q_vector: the vector containing rings to be configured
6500 *
6501 * Set up interrupt moderation registers, with the intent to do the right thing
6502 * when called from reset or from probe, and whether or not dynamic moderation
6503 * is enabled or not. Take special care to write all the registers in both
6504 * dynamic moderation mode or not in order to make sure hardware is in a known
6505 * state.
6506 */
6507static void ice_init_moderation(struct ice_q_vector *q_vector)
6508{
6509 struct ice_ring_container *rc;
6510 bool tx_dynamic, rx_dynamic;
6511
6512 rc = &q_vector->tx;
6513 INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6514 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6515 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6516 rc->dim.priv = rc;
6517 tx_dynamic = ITR_IS_DYNAMIC(rc);
6518
6519 /* set the initial TX ITR to match the above */
6520 ice_write_itr(rc, tx_dynamic ?
6521 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6522
6523 rc = &q_vector->rx;
6524 INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6525 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6526 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6527 rc->dim.priv = rc;
6528 rx_dynamic = ITR_IS_DYNAMIC(rc);
6529
6530 /* set the initial RX ITR to match the above */
6531 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6532 rc->itr_setting);
6533
6534 ice_set_q_vector_intrl(q_vector);
6535}
6536
6537/**
6538 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6539 * @vsi: the VSI being configured
6540 */
6541static void ice_napi_enable_all(struct ice_vsi *vsi)
6542{
6543 int q_idx;
6544
6545 if (!vsi->netdev)
6546 return;
6547
6548 ice_for_each_q_vector(vsi, q_idx) {
6549 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6550
6551 ice_init_moderation(q_vector);
6552
6553 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6554 napi_enable(&q_vector->napi);
6555 }
6556}
6557
6558/**
6559 * ice_up_complete - Finish the last steps of bringing up a connection
6560 * @vsi: The VSI being configured
6561 *
6562 * Return 0 on success and negative value on error
6563 */
6564static int ice_up_complete(struct ice_vsi *vsi)
6565{
6566 struct ice_pf *pf = vsi->back;
6567 int err;
6568
6569 ice_vsi_cfg_msix(vsi);
6570
6571 /* Enable only Rx rings, Tx rings were enabled by the FW when the
6572 * Tx queue group list was configured and the context bits were
6573 * programmed using ice_vsi_cfg_txqs
6574 */
6575 err = ice_vsi_start_all_rx_rings(vsi);
6576 if (err)
6577 return err;
6578
6579 clear_bit(ICE_VSI_DOWN, vsi->state);
6580 ice_napi_enable_all(vsi);
6581 ice_vsi_ena_irq(vsi);
6582
6583 if (vsi->port_info &&
6584 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6585 vsi->netdev && vsi->type == ICE_VSI_PF) {
6586 ice_print_link_msg(vsi, true);
6587 netif_tx_start_all_queues(vsi->netdev);
6588 netif_carrier_on(vsi->netdev);
6589 ice_ptp_link_change(pf, pf->hw.pf_id, true);
6590 }
6591
6592 /* Perform an initial read of the statistics registers now to
6593 * set the baseline so counters are ready when interface is up
6594 */
6595 ice_update_eth_stats(vsi);
6596
6597 if (vsi->type == ICE_VSI_PF)
6598 ice_service_task_schedule(pf);
6599
6600 return 0;
6601}
6602
6603/**
6604 * ice_up - Bring the connection back up after being down
6605 * @vsi: VSI being configured
6606 */
6607int ice_up(struct ice_vsi *vsi)
6608{
6609 int err;
6610
6611 err = ice_vsi_cfg_lan(vsi);
6612 if (!err)
6613 err = ice_up_complete(vsi);
6614
6615 return err;
6616}
6617
6618/**
6619 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6620 * @syncp: pointer to u64_stats_sync
6621 * @stats: stats that pkts and bytes count will be taken from
6622 * @pkts: packets stats counter
6623 * @bytes: bytes stats counter
6624 *
6625 * This function fetches stats from the ring considering the atomic operations
6626 * that needs to be performed to read u64 values in 32 bit machine.
6627 */
6628void
6629ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6630 struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6631{
6632 unsigned int start;
6633
6634 do {
6635 start = u64_stats_fetch_begin(syncp);
6636 *pkts = stats.pkts;
6637 *bytes = stats.bytes;
6638 } while (u64_stats_fetch_retry(syncp, start));
6639}
6640
6641/**
6642 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6643 * @vsi: the VSI to be updated
6644 * @vsi_stats: the stats struct to be updated
6645 * @rings: rings to work on
6646 * @count: number of rings
6647 */
6648static void
6649ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6650 struct rtnl_link_stats64 *vsi_stats,
6651 struct ice_tx_ring **rings, u16 count)
6652{
6653 u16 i;
6654
6655 for (i = 0; i < count; i++) {
6656 struct ice_tx_ring *ring;
6657 u64 pkts = 0, bytes = 0;
6658
6659 ring = READ_ONCE(rings[i]);
6660 if (!ring || !ring->ring_stats)
6661 continue;
6662 ice_fetch_u64_stats_per_ring(&ring->ring_stats->syncp,
6663 ring->ring_stats->stats, &pkts,
6664 &bytes);
6665 vsi_stats->tx_packets += pkts;
6666 vsi_stats->tx_bytes += bytes;
6667 vsi->tx_restart += ring->ring_stats->tx_stats.restart_q;
6668 vsi->tx_busy += ring->ring_stats->tx_stats.tx_busy;
6669 vsi->tx_linearize += ring->ring_stats->tx_stats.tx_linearize;
6670 }
6671}
6672
6673/**
6674 * ice_update_vsi_ring_stats - Update VSI stats counters
6675 * @vsi: the VSI to be updated
6676 */
6677static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6678{
6679 struct rtnl_link_stats64 *net_stats, *stats_prev;
6680 struct rtnl_link_stats64 *vsi_stats;
6681 struct ice_pf *pf = vsi->back;
6682 u64 pkts, bytes;
6683 int i;
6684
6685 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6686 if (!vsi_stats)
6687 return;
6688
6689 /* reset non-netdev (extended) stats */
6690 vsi->tx_restart = 0;
6691 vsi->tx_busy = 0;
6692 vsi->tx_linearize = 0;
6693 vsi->rx_buf_failed = 0;
6694 vsi->rx_page_failed = 0;
6695
6696 rcu_read_lock();
6697
6698 /* update Tx rings counters */
6699 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6700 vsi->num_txq);
6701
6702 /* update Rx rings counters */
6703 ice_for_each_rxq(vsi, i) {
6704 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6705 struct ice_ring_stats *ring_stats;
6706
6707 ring_stats = ring->ring_stats;
6708 ice_fetch_u64_stats_per_ring(&ring_stats->syncp,
6709 ring_stats->stats, &pkts,
6710 &bytes);
6711 vsi_stats->rx_packets += pkts;
6712 vsi_stats->rx_bytes += bytes;
6713 vsi->rx_buf_failed += ring_stats->rx_stats.alloc_buf_failed;
6714 vsi->rx_page_failed += ring_stats->rx_stats.alloc_page_failed;
6715 }
6716
6717 /* update XDP Tx rings counters */
6718 if (ice_is_xdp_ena_vsi(vsi))
6719 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6720 vsi->num_xdp_txq);
6721
6722 rcu_read_unlock();
6723
6724 net_stats = &vsi->net_stats;
6725 stats_prev = &vsi->net_stats_prev;
6726
6727 /* Update netdev counters, but keep in mind that values could start at
6728 * random value after PF reset. And as we increase the reported stat by
6729 * diff of Prev-Cur, we need to be sure that Prev is valid. If it's not,
6730 * let's skip this round.
6731 */
6732 if (likely(pf->stat_prev_loaded)) {
6733 net_stats->tx_packets += vsi_stats->tx_packets - stats_prev->tx_packets;
6734 net_stats->tx_bytes += vsi_stats->tx_bytes - stats_prev->tx_bytes;
6735 net_stats->rx_packets += vsi_stats->rx_packets - stats_prev->rx_packets;
6736 net_stats->rx_bytes += vsi_stats->rx_bytes - stats_prev->rx_bytes;
6737 }
6738
6739 stats_prev->tx_packets = vsi_stats->tx_packets;
6740 stats_prev->tx_bytes = vsi_stats->tx_bytes;
6741 stats_prev->rx_packets = vsi_stats->rx_packets;
6742 stats_prev->rx_bytes = vsi_stats->rx_bytes;
6743
6744 kfree(vsi_stats);
6745}
6746
6747/**
6748 * ice_update_vsi_stats - Update VSI stats counters
6749 * @vsi: the VSI to be updated
6750 */
6751void ice_update_vsi_stats(struct ice_vsi *vsi)
6752{
6753 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6754 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6755 struct ice_pf *pf = vsi->back;
6756
6757 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6758 test_bit(ICE_CFG_BUSY, pf->state))
6759 return;
6760
6761 /* get stats as recorded by Tx/Rx rings */
6762 ice_update_vsi_ring_stats(vsi);
6763
6764 /* get VSI stats as recorded by the hardware */
6765 ice_update_eth_stats(vsi);
6766
6767 cur_ns->tx_errors = cur_es->tx_errors;
6768 cur_ns->rx_dropped = cur_es->rx_discards;
6769 cur_ns->tx_dropped = cur_es->tx_discards;
6770 cur_ns->multicast = cur_es->rx_multicast;
6771
6772 /* update some more netdev stats if this is main VSI */
6773 if (vsi->type == ICE_VSI_PF) {
6774 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6775 cur_ns->rx_errors = pf->stats.crc_errors +
6776 pf->stats.illegal_bytes +
6777 pf->stats.rx_undersize +
6778 pf->hw_csum_rx_error +
6779 pf->stats.rx_jabber +
6780 pf->stats.rx_fragments +
6781 pf->stats.rx_oversize;
6782 /* record drops from the port level */
6783 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6784 }
6785}
6786
6787/**
6788 * ice_update_pf_stats - Update PF port stats counters
6789 * @pf: PF whose stats needs to be updated
6790 */
6791void ice_update_pf_stats(struct ice_pf *pf)
6792{
6793 struct ice_hw_port_stats *prev_ps, *cur_ps;
6794 struct ice_hw *hw = &pf->hw;
6795 u16 fd_ctr_base;
6796 u8 port;
6797
6798 port = hw->port_info->lport;
6799 prev_ps = &pf->stats_prev;
6800 cur_ps = &pf->stats;
6801
6802 if (ice_is_reset_in_progress(pf->state))
6803 pf->stat_prev_loaded = false;
6804
6805 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6806 &prev_ps->eth.rx_bytes,
6807 &cur_ps->eth.rx_bytes);
6808
6809 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6810 &prev_ps->eth.rx_unicast,
6811 &cur_ps->eth.rx_unicast);
6812
6813 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6814 &prev_ps->eth.rx_multicast,
6815 &cur_ps->eth.rx_multicast);
6816
6817 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6818 &prev_ps->eth.rx_broadcast,
6819 &cur_ps->eth.rx_broadcast);
6820
6821 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6822 &prev_ps->eth.rx_discards,
6823 &cur_ps->eth.rx_discards);
6824
6825 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6826 &prev_ps->eth.tx_bytes,
6827 &cur_ps->eth.tx_bytes);
6828
6829 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6830 &prev_ps->eth.tx_unicast,
6831 &cur_ps->eth.tx_unicast);
6832
6833 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6834 &prev_ps->eth.tx_multicast,
6835 &cur_ps->eth.tx_multicast);
6836
6837 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6838 &prev_ps->eth.tx_broadcast,
6839 &cur_ps->eth.tx_broadcast);
6840
6841 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6842 &prev_ps->tx_dropped_link_down,
6843 &cur_ps->tx_dropped_link_down);
6844
6845 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6846 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6847
6848 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6849 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6850
6851 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6852 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6853
6854 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6855 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6856
6857 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6858 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6859
6860 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6861 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6862
6863 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6864 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6865
6866 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6867 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6868
6869 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6870 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6871
6872 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6873 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6874
6875 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6876 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6877
6878 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6879 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6880
6881 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6882 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6883
6884 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6885 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6886
6887 fd_ctr_base = hw->fd_ctr_base;
6888
6889 ice_stat_update40(hw,
6890 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6891 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6892 &cur_ps->fd_sb_match);
6893 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6894 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6895
6896 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6897 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6898
6899 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6900 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6901
6902 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6903 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6904
6905 ice_update_dcb_stats(pf);
6906
6907 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6908 &prev_ps->crc_errors, &cur_ps->crc_errors);
6909
6910 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6911 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6912
6913 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6914 &prev_ps->mac_local_faults,
6915 &cur_ps->mac_local_faults);
6916
6917 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6918 &prev_ps->mac_remote_faults,
6919 &cur_ps->mac_remote_faults);
6920
6921 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6922 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6923
6924 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6925 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6926
6927 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6928 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6929
6930 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6931 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6932
6933 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6934
6935 pf->stat_prev_loaded = true;
6936}
6937
6938/**
6939 * ice_get_stats64 - get statistics for network device structure
6940 * @netdev: network interface device structure
6941 * @stats: main device statistics structure
6942 */
6943static
6944void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6945{
6946 struct ice_netdev_priv *np = netdev_priv(netdev);
6947 struct rtnl_link_stats64 *vsi_stats;
6948 struct ice_vsi *vsi = np->vsi;
6949
6950 vsi_stats = &vsi->net_stats;
6951
6952 if (!vsi->num_txq || !vsi->num_rxq)
6953 return;
6954
6955 /* netdev packet/byte stats come from ring counter. These are obtained
6956 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6957 * But, only call the update routine and read the registers if VSI is
6958 * not down.
6959 */
6960 if (!test_bit(ICE_VSI_DOWN, vsi->state))
6961 ice_update_vsi_ring_stats(vsi);
6962 stats->tx_packets = vsi_stats->tx_packets;
6963 stats->tx_bytes = vsi_stats->tx_bytes;
6964 stats->rx_packets = vsi_stats->rx_packets;
6965 stats->rx_bytes = vsi_stats->rx_bytes;
6966
6967 /* The rest of the stats can be read from the hardware but instead we
6968 * just return values that the watchdog task has already obtained from
6969 * the hardware.
6970 */
6971 stats->multicast = vsi_stats->multicast;
6972 stats->tx_errors = vsi_stats->tx_errors;
6973 stats->tx_dropped = vsi_stats->tx_dropped;
6974 stats->rx_errors = vsi_stats->rx_errors;
6975 stats->rx_dropped = vsi_stats->rx_dropped;
6976 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6977 stats->rx_length_errors = vsi_stats->rx_length_errors;
6978}
6979
6980/**
6981 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6982 * @vsi: VSI having NAPI disabled
6983 */
6984static void ice_napi_disable_all(struct ice_vsi *vsi)
6985{
6986 int q_idx;
6987
6988 if (!vsi->netdev)
6989 return;
6990
6991 ice_for_each_q_vector(vsi, q_idx) {
6992 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6993
6994 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6995 napi_disable(&q_vector->napi);
6996
6997 cancel_work_sync(&q_vector->tx.dim.work);
6998 cancel_work_sync(&q_vector->rx.dim.work);
6999 }
7000}
7001
7002/**
7003 * ice_vsi_dis_irq - Mask off queue interrupt generation on the VSI
7004 * @vsi: the VSI being un-configured
7005 */
7006static void ice_vsi_dis_irq(struct ice_vsi *vsi)
7007{
7008 struct ice_pf *pf = vsi->back;
7009 struct ice_hw *hw = &pf->hw;
7010 u32 val;
7011 int i;
7012
7013 /* disable interrupt causation from each Rx queue; Tx queues are
7014 * handled in ice_vsi_stop_tx_ring()
7015 */
7016 if (vsi->rx_rings) {
7017 ice_for_each_rxq(vsi, i) {
7018 if (vsi->rx_rings[i]) {
7019 u16 reg;
7020
7021 reg = vsi->rx_rings[i]->reg_idx;
7022 val = rd32(hw, QINT_RQCTL(reg));
7023 val &= ~QINT_RQCTL_CAUSE_ENA_M;
7024 wr32(hw, QINT_RQCTL(reg), val);
7025 }
7026 }
7027 }
7028
7029 /* disable each interrupt */
7030 ice_for_each_q_vector(vsi, i) {
7031 if (!vsi->q_vectors[i])
7032 continue;
7033 wr32(hw, GLINT_DYN_CTL(vsi->q_vectors[i]->reg_idx), 0);
7034 }
7035
7036 ice_flush(hw);
7037
7038 /* don't call synchronize_irq() for VF's from the host */
7039 if (vsi->type == ICE_VSI_VF)
7040 return;
7041
7042 ice_for_each_q_vector(vsi, i)
7043 synchronize_irq(vsi->q_vectors[i]->irq.virq);
7044}
7045
7046/**
7047 * ice_down - Shutdown the connection
7048 * @vsi: The VSI being stopped
7049 *
7050 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
7051 */
7052int ice_down(struct ice_vsi *vsi)
7053{
7054 int i, tx_err, rx_err, vlan_err = 0;
7055
7056 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
7057
7058 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
7059 vlan_err = ice_vsi_del_vlan_zero(vsi);
7060 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
7061 netif_carrier_off(vsi->netdev);
7062 netif_tx_disable(vsi->netdev);
7063 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
7064 ice_eswitch_stop_all_tx_queues(vsi->back);
7065 }
7066
7067 ice_vsi_dis_irq(vsi);
7068
7069 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
7070 if (tx_err)
7071 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
7072 vsi->vsi_num, tx_err);
7073 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
7074 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
7075 if (tx_err)
7076 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
7077 vsi->vsi_num, tx_err);
7078 }
7079
7080 rx_err = ice_vsi_stop_all_rx_rings(vsi);
7081 if (rx_err)
7082 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
7083 vsi->vsi_num, rx_err);
7084
7085 ice_napi_disable_all(vsi);
7086
7087 ice_for_each_txq(vsi, i)
7088 ice_clean_tx_ring(vsi->tx_rings[i]);
7089
7090 if (ice_is_xdp_ena_vsi(vsi))
7091 ice_for_each_xdp_txq(vsi, i)
7092 ice_clean_tx_ring(vsi->xdp_rings[i]);
7093
7094 ice_for_each_rxq(vsi, i)
7095 ice_clean_rx_ring(vsi->rx_rings[i]);
7096
7097 if (tx_err || rx_err || vlan_err) {
7098 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
7099 vsi->vsi_num, vsi->vsw->sw_id);
7100 return -EIO;
7101 }
7102
7103 return 0;
7104}
7105
7106/**
7107 * ice_down_up - shutdown the VSI connection and bring it up
7108 * @vsi: the VSI to be reconnected
7109 */
7110int ice_down_up(struct ice_vsi *vsi)
7111{
7112 int ret;
7113
7114 /* if DOWN already set, nothing to do */
7115 if (test_and_set_bit(ICE_VSI_DOWN, vsi->state))
7116 return 0;
7117
7118 ret = ice_down(vsi);
7119 if (ret)
7120 return ret;
7121
7122 ret = ice_up(vsi);
7123 if (ret) {
7124 netdev_err(vsi->netdev, "reallocating resources failed during netdev features change, may need to reload driver\n");
7125 return ret;
7126 }
7127
7128 return 0;
7129}
7130
7131/**
7132 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
7133 * @vsi: VSI having resources allocated
7134 *
7135 * Return 0 on success, negative on failure
7136 */
7137int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
7138{
7139 int i, err = 0;
7140
7141 if (!vsi->num_txq) {
7142 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
7143 vsi->vsi_num);
7144 return -EINVAL;
7145 }
7146
7147 ice_for_each_txq(vsi, i) {
7148 struct ice_tx_ring *ring = vsi->tx_rings[i];
7149
7150 if (!ring)
7151 return -EINVAL;
7152
7153 if (vsi->netdev)
7154 ring->netdev = vsi->netdev;
7155 err = ice_setup_tx_ring(ring);
7156 if (err)
7157 break;
7158 }
7159
7160 return err;
7161}
7162
7163/**
7164 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
7165 * @vsi: VSI having resources allocated
7166 *
7167 * Return 0 on success, negative on failure
7168 */
7169int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
7170{
7171 int i, err = 0;
7172
7173 if (!vsi->num_rxq) {
7174 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
7175 vsi->vsi_num);
7176 return -EINVAL;
7177 }
7178
7179 ice_for_each_rxq(vsi, i) {
7180 struct ice_rx_ring *ring = vsi->rx_rings[i];
7181
7182 if (!ring)
7183 return -EINVAL;
7184
7185 if (vsi->netdev)
7186 ring->netdev = vsi->netdev;
7187 err = ice_setup_rx_ring(ring);
7188 if (err)
7189 break;
7190 }
7191
7192 return err;
7193}
7194
7195/**
7196 * ice_vsi_open_ctrl - open control VSI for use
7197 * @vsi: the VSI to open
7198 *
7199 * Initialization of the Control VSI
7200 *
7201 * Returns 0 on success, negative value on error
7202 */
7203int ice_vsi_open_ctrl(struct ice_vsi *vsi)
7204{
7205 char int_name[ICE_INT_NAME_STR_LEN];
7206 struct ice_pf *pf = vsi->back;
7207 struct device *dev;
7208 int err;
7209
7210 dev = ice_pf_to_dev(pf);
7211 /* allocate descriptors */
7212 err = ice_vsi_setup_tx_rings(vsi);
7213 if (err)
7214 goto err_setup_tx;
7215
7216 err = ice_vsi_setup_rx_rings(vsi);
7217 if (err)
7218 goto err_setup_rx;
7219
7220 err = ice_vsi_cfg_lan(vsi);
7221 if (err)
7222 goto err_setup_rx;
7223
7224 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
7225 dev_driver_string(dev), dev_name(dev));
7226 err = ice_vsi_req_irq_msix(vsi, int_name);
7227 if (err)
7228 goto err_setup_rx;
7229
7230 ice_vsi_cfg_msix(vsi);
7231
7232 err = ice_vsi_start_all_rx_rings(vsi);
7233 if (err)
7234 goto err_up_complete;
7235
7236 clear_bit(ICE_VSI_DOWN, vsi->state);
7237 ice_vsi_ena_irq(vsi);
7238
7239 return 0;
7240
7241err_up_complete:
7242 ice_down(vsi);
7243err_setup_rx:
7244 ice_vsi_free_rx_rings(vsi);
7245err_setup_tx:
7246 ice_vsi_free_tx_rings(vsi);
7247
7248 return err;
7249}
7250
7251/**
7252 * ice_vsi_open - Called when a network interface is made active
7253 * @vsi: the VSI to open
7254 *
7255 * Initialization of the VSI
7256 *
7257 * Returns 0 on success, negative value on error
7258 */
7259int ice_vsi_open(struct ice_vsi *vsi)
7260{
7261 char int_name[ICE_INT_NAME_STR_LEN];
7262 struct ice_pf *pf = vsi->back;
7263 int err;
7264
7265 /* allocate descriptors */
7266 err = ice_vsi_setup_tx_rings(vsi);
7267 if (err)
7268 goto err_setup_tx;
7269
7270 err = ice_vsi_setup_rx_rings(vsi);
7271 if (err)
7272 goto err_setup_rx;
7273
7274 err = ice_vsi_cfg_lan(vsi);
7275 if (err)
7276 goto err_setup_rx;
7277
7278 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
7279 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
7280 err = ice_vsi_req_irq_msix(vsi, int_name);
7281 if (err)
7282 goto err_setup_rx;
7283
7284 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
7285
7286 if (vsi->type == ICE_VSI_PF) {
7287 /* Notify the stack of the actual queue counts. */
7288 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
7289 if (err)
7290 goto err_set_qs;
7291
7292 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
7293 if (err)
7294 goto err_set_qs;
7295 }
7296
7297 err = ice_up_complete(vsi);
7298 if (err)
7299 goto err_up_complete;
7300
7301 return 0;
7302
7303err_up_complete:
7304 ice_down(vsi);
7305err_set_qs:
7306 ice_vsi_free_irq(vsi);
7307err_setup_rx:
7308 ice_vsi_free_rx_rings(vsi);
7309err_setup_tx:
7310 ice_vsi_free_tx_rings(vsi);
7311
7312 return err;
7313}
7314
7315/**
7316 * ice_vsi_release_all - Delete all VSIs
7317 * @pf: PF from which all VSIs are being removed
7318 */
7319static void ice_vsi_release_all(struct ice_pf *pf)
7320{
7321 int err, i;
7322
7323 if (!pf->vsi)
7324 return;
7325
7326 ice_for_each_vsi(pf, i) {
7327 if (!pf->vsi[i])
7328 continue;
7329
7330 if (pf->vsi[i]->type == ICE_VSI_CHNL)
7331 continue;
7332
7333 err = ice_vsi_release(pf->vsi[i]);
7334 if (err)
7335 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
7336 i, err, pf->vsi[i]->vsi_num);
7337 }
7338}
7339
7340/**
7341 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
7342 * @pf: pointer to the PF instance
7343 * @type: VSI type to rebuild
7344 *
7345 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
7346 */
7347static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
7348{
7349 struct device *dev = ice_pf_to_dev(pf);
7350 int i, err;
7351
7352 ice_for_each_vsi(pf, i) {
7353 struct ice_vsi *vsi = pf->vsi[i];
7354
7355 if (!vsi || vsi->type != type)
7356 continue;
7357
7358 /* rebuild the VSI */
7359 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
7360 if (err) {
7361 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
7362 err, vsi->idx, ice_vsi_type_str(type));
7363 return err;
7364 }
7365
7366 /* replay filters for the VSI */
7367 err = ice_replay_vsi(&pf->hw, vsi->idx);
7368 if (err) {
7369 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
7370 err, vsi->idx, ice_vsi_type_str(type));
7371 return err;
7372 }
7373
7374 /* Re-map HW VSI number, using VSI handle that has been
7375 * previously validated in ice_replay_vsi() call above
7376 */
7377 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
7378
7379 /* enable the VSI */
7380 err = ice_ena_vsi(vsi, false);
7381 if (err) {
7382 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
7383 err, vsi->idx, ice_vsi_type_str(type));
7384 return err;
7385 }
7386
7387 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
7388 ice_vsi_type_str(type));
7389 }
7390
7391 return 0;
7392}
7393
7394/**
7395 * ice_update_pf_netdev_link - Update PF netdev link status
7396 * @pf: pointer to the PF instance
7397 */
7398static void ice_update_pf_netdev_link(struct ice_pf *pf)
7399{
7400 bool link_up;
7401 int i;
7402
7403 ice_for_each_vsi(pf, i) {
7404 struct ice_vsi *vsi = pf->vsi[i];
7405
7406 if (!vsi || vsi->type != ICE_VSI_PF)
7407 return;
7408
7409 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
7410 if (link_up) {
7411 netif_carrier_on(pf->vsi[i]->netdev);
7412 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
7413 } else {
7414 netif_carrier_off(pf->vsi[i]->netdev);
7415 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
7416 }
7417 }
7418}
7419
7420/**
7421 * ice_rebuild - rebuild after reset
7422 * @pf: PF to rebuild
7423 * @reset_type: type of reset
7424 *
7425 * Do not rebuild VF VSI in this flow because that is already handled via
7426 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
7427 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
7428 * to reset/rebuild all the VF VSI twice.
7429 */
7430static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
7431{
7432 struct device *dev = ice_pf_to_dev(pf);
7433 struct ice_hw *hw = &pf->hw;
7434 bool dvm;
7435 int err;
7436
7437 if (test_bit(ICE_DOWN, pf->state))
7438 goto clear_recovery;
7439
7440 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
7441
7442#define ICE_EMP_RESET_SLEEP_MS 5000
7443 if (reset_type == ICE_RESET_EMPR) {
7444 /* If an EMP reset has occurred, any previously pending flash
7445 * update will have completed. We no longer know whether or
7446 * not the NVM update EMP reset is restricted.
7447 */
7448 pf->fw_emp_reset_disabled = false;
7449
7450 msleep(ICE_EMP_RESET_SLEEP_MS);
7451 }
7452
7453 err = ice_init_all_ctrlq(hw);
7454 if (err) {
7455 dev_err(dev, "control queues init failed %d\n", err);
7456 goto err_init_ctrlq;
7457 }
7458
7459 /* if DDP was previously loaded successfully */
7460 if (!ice_is_safe_mode(pf)) {
7461 /* reload the SW DB of filter tables */
7462 if (reset_type == ICE_RESET_PFR)
7463 ice_fill_blk_tbls(hw);
7464 else
7465 /* Reload DDP Package after CORER/GLOBR reset */
7466 ice_load_pkg(NULL, pf);
7467 }
7468
7469 err = ice_clear_pf_cfg(hw);
7470 if (err) {
7471 dev_err(dev, "clear PF configuration failed %d\n", err);
7472 goto err_init_ctrlq;
7473 }
7474
7475 ice_clear_pxe_mode(hw);
7476
7477 err = ice_init_nvm(hw);
7478 if (err) {
7479 dev_err(dev, "ice_init_nvm failed %d\n", err);
7480 goto err_init_ctrlq;
7481 }
7482
7483 err = ice_get_caps(hw);
7484 if (err) {
7485 dev_err(dev, "ice_get_caps failed %d\n", err);
7486 goto err_init_ctrlq;
7487 }
7488
7489 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7490 if (err) {
7491 dev_err(dev, "set_mac_cfg failed %d\n", err);
7492 goto err_init_ctrlq;
7493 }
7494
7495 dvm = ice_is_dvm_ena(hw);
7496
7497 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7498 if (err)
7499 goto err_init_ctrlq;
7500
7501 err = ice_sched_init_port(hw->port_info);
7502 if (err)
7503 goto err_sched_init_port;
7504
7505 /* start misc vector */
7506 err = ice_req_irq_msix_misc(pf);
7507 if (err) {
7508 dev_err(dev, "misc vector setup failed: %d\n", err);
7509 goto err_sched_init_port;
7510 }
7511
7512 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7513 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7514 if (!rd32(hw, PFQF_FD_SIZE)) {
7515 u16 unused, guar, b_effort;
7516
7517 guar = hw->func_caps.fd_fltr_guar;
7518 b_effort = hw->func_caps.fd_fltr_best_effort;
7519
7520 /* force guaranteed filter pool for PF */
7521 ice_alloc_fd_guar_item(hw, &unused, guar);
7522 /* force shared filter pool for PF */
7523 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7524 }
7525 }
7526
7527 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7528 ice_dcb_rebuild(pf);
7529
7530 /* If the PF previously had enabled PTP, PTP init needs to happen before
7531 * the VSI rebuild. If not, this causes the PTP link status events to
7532 * fail.
7533 */
7534 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7535 ice_ptp_rebuild(pf, reset_type);
7536
7537 if (ice_is_feature_supported(pf, ICE_F_GNSS))
7538 ice_gnss_init(pf);
7539
7540 /* rebuild PF VSI */
7541 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7542 if (err) {
7543 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7544 goto err_vsi_rebuild;
7545 }
7546
7547 err = ice_eswitch_rebuild(pf);
7548 if (err) {
7549 dev_err(dev, "Switchdev rebuild failed: %d\n", err);
7550 goto err_vsi_rebuild;
7551 }
7552
7553 if (reset_type == ICE_RESET_PFR) {
7554 err = ice_rebuild_channels(pf);
7555 if (err) {
7556 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7557 err);
7558 goto err_vsi_rebuild;
7559 }
7560 }
7561
7562 /* If Flow Director is active */
7563 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7564 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7565 if (err) {
7566 dev_err(dev, "control VSI rebuild failed: %d\n", err);
7567 goto err_vsi_rebuild;
7568 }
7569
7570 /* replay HW Flow Director recipes */
7571 if (hw->fdir_prof)
7572 ice_fdir_replay_flows(hw);
7573
7574 /* replay Flow Director filters */
7575 ice_fdir_replay_fltrs(pf);
7576
7577 ice_rebuild_arfs(pf);
7578 }
7579
7580 ice_update_pf_netdev_link(pf);
7581
7582 /* tell the firmware we are up */
7583 err = ice_send_version(pf);
7584 if (err) {
7585 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7586 err);
7587 goto err_vsi_rebuild;
7588 }
7589
7590 ice_replay_post(hw);
7591
7592 /* if we get here, reset flow is successful */
7593 clear_bit(ICE_RESET_FAILED, pf->state);
7594
7595 ice_plug_aux_dev(pf);
7596 if (ice_is_feature_supported(pf, ICE_F_SRIOV_LAG))
7597 ice_lag_rebuild(pf);
7598
7599 /* Restore timestamp mode settings after VSI rebuild */
7600 ice_ptp_restore_timestamp_mode(pf);
7601 return;
7602
7603err_vsi_rebuild:
7604err_sched_init_port:
7605 ice_sched_cleanup_all(hw);
7606err_init_ctrlq:
7607 ice_shutdown_all_ctrlq(hw);
7608 set_bit(ICE_RESET_FAILED, pf->state);
7609clear_recovery:
7610 /* set this bit in PF state to control service task scheduling */
7611 set_bit(ICE_NEEDS_RESTART, pf->state);
7612 dev_err(dev, "Rebuild failed, unload and reload driver\n");
7613}
7614
7615/**
7616 * ice_change_mtu - NDO callback to change the MTU
7617 * @netdev: network interface device structure
7618 * @new_mtu: new value for maximum frame size
7619 *
7620 * Returns 0 on success, negative on failure
7621 */
7622static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7623{
7624 struct ice_netdev_priv *np = netdev_priv(netdev);
7625 struct ice_vsi *vsi = np->vsi;
7626 struct ice_pf *pf = vsi->back;
7627 struct bpf_prog *prog;
7628 u8 count = 0;
7629 int err = 0;
7630
7631 if (new_mtu == (int)netdev->mtu) {
7632 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7633 return 0;
7634 }
7635
7636 prog = vsi->xdp_prog;
7637 if (prog && !prog->aux->xdp_has_frags) {
7638 int frame_size = ice_max_xdp_frame_size(vsi);
7639
7640 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7641 netdev_err(netdev, "max MTU for XDP usage is %d\n",
7642 frame_size - ICE_ETH_PKT_HDR_PAD);
7643 return -EINVAL;
7644 }
7645 } else if (test_bit(ICE_FLAG_LEGACY_RX, pf->flags)) {
7646 if (new_mtu + ICE_ETH_PKT_HDR_PAD > ICE_MAX_FRAME_LEGACY_RX) {
7647 netdev_err(netdev, "Too big MTU for legacy-rx; Max is %d\n",
7648 ICE_MAX_FRAME_LEGACY_RX - ICE_ETH_PKT_HDR_PAD);
7649 return -EINVAL;
7650 }
7651 }
7652
7653 /* if a reset is in progress, wait for some time for it to complete */
7654 do {
7655 if (ice_is_reset_in_progress(pf->state)) {
7656 count++;
7657 usleep_range(1000, 2000);
7658 } else {
7659 break;
7660 }
7661
7662 } while (count < 100);
7663
7664 if (count == 100) {
7665 netdev_err(netdev, "can't change MTU. Device is busy\n");
7666 return -EBUSY;
7667 }
7668
7669 netdev->mtu = (unsigned int)new_mtu;
7670 err = ice_down_up(vsi);
7671 if (err)
7672 return err;
7673
7674 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7675 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7676
7677 return err;
7678}
7679
7680/**
7681 * ice_eth_ioctl - Access the hwtstamp interface
7682 * @netdev: network interface device structure
7683 * @ifr: interface request data
7684 * @cmd: ioctl command
7685 */
7686static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7687{
7688 struct ice_netdev_priv *np = netdev_priv(netdev);
7689 struct ice_pf *pf = np->vsi->back;
7690
7691 switch (cmd) {
7692 case SIOCGHWTSTAMP:
7693 return ice_ptp_get_ts_config(pf, ifr);
7694 case SIOCSHWTSTAMP:
7695 return ice_ptp_set_ts_config(pf, ifr);
7696 default:
7697 return -EOPNOTSUPP;
7698 }
7699}
7700
7701/**
7702 * ice_aq_str - convert AQ err code to a string
7703 * @aq_err: the AQ error code to convert
7704 */
7705const char *ice_aq_str(enum ice_aq_err aq_err)
7706{
7707 switch (aq_err) {
7708 case ICE_AQ_RC_OK:
7709 return "OK";
7710 case ICE_AQ_RC_EPERM:
7711 return "ICE_AQ_RC_EPERM";
7712 case ICE_AQ_RC_ENOENT:
7713 return "ICE_AQ_RC_ENOENT";
7714 case ICE_AQ_RC_ENOMEM:
7715 return "ICE_AQ_RC_ENOMEM";
7716 case ICE_AQ_RC_EBUSY:
7717 return "ICE_AQ_RC_EBUSY";
7718 case ICE_AQ_RC_EEXIST:
7719 return "ICE_AQ_RC_EEXIST";
7720 case ICE_AQ_RC_EINVAL:
7721 return "ICE_AQ_RC_EINVAL";
7722 case ICE_AQ_RC_ENOSPC:
7723 return "ICE_AQ_RC_ENOSPC";
7724 case ICE_AQ_RC_ENOSYS:
7725 return "ICE_AQ_RC_ENOSYS";
7726 case ICE_AQ_RC_EMODE:
7727 return "ICE_AQ_RC_EMODE";
7728 case ICE_AQ_RC_ENOSEC:
7729 return "ICE_AQ_RC_ENOSEC";
7730 case ICE_AQ_RC_EBADSIG:
7731 return "ICE_AQ_RC_EBADSIG";
7732 case ICE_AQ_RC_ESVN:
7733 return "ICE_AQ_RC_ESVN";
7734 case ICE_AQ_RC_EBADMAN:
7735 return "ICE_AQ_RC_EBADMAN";
7736 case ICE_AQ_RC_EBADBUF:
7737 return "ICE_AQ_RC_EBADBUF";
7738 }
7739
7740 return "ICE_AQ_RC_UNKNOWN";
7741}
7742
7743/**
7744 * ice_set_rss_lut - Set RSS LUT
7745 * @vsi: Pointer to VSI structure
7746 * @lut: Lookup table
7747 * @lut_size: Lookup table size
7748 *
7749 * Returns 0 on success, negative on failure
7750 */
7751int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7752{
7753 struct ice_aq_get_set_rss_lut_params params = {};
7754 struct ice_hw *hw = &vsi->back->hw;
7755 int status;
7756
7757 if (!lut)
7758 return -EINVAL;
7759
7760 params.vsi_handle = vsi->idx;
7761 params.lut_size = lut_size;
7762 params.lut_type = vsi->rss_lut_type;
7763 params.lut = lut;
7764
7765 status = ice_aq_set_rss_lut(hw, ¶ms);
7766 if (status)
7767 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7768 status, ice_aq_str(hw->adminq.sq_last_status));
7769
7770 return status;
7771}
7772
7773/**
7774 * ice_set_rss_key - Set RSS key
7775 * @vsi: Pointer to the VSI structure
7776 * @seed: RSS hash seed
7777 *
7778 * Returns 0 on success, negative on failure
7779 */
7780int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7781{
7782 struct ice_hw *hw = &vsi->back->hw;
7783 int status;
7784
7785 if (!seed)
7786 return -EINVAL;
7787
7788 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7789 if (status)
7790 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7791 status, ice_aq_str(hw->adminq.sq_last_status));
7792
7793 return status;
7794}
7795
7796/**
7797 * ice_get_rss_lut - Get RSS LUT
7798 * @vsi: Pointer to VSI structure
7799 * @lut: Buffer to store the lookup table entries
7800 * @lut_size: Size of buffer to store the lookup table entries
7801 *
7802 * Returns 0 on success, negative on failure
7803 */
7804int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7805{
7806 struct ice_aq_get_set_rss_lut_params params = {};
7807 struct ice_hw *hw = &vsi->back->hw;
7808 int status;
7809
7810 if (!lut)
7811 return -EINVAL;
7812
7813 params.vsi_handle = vsi->idx;
7814 params.lut_size = lut_size;
7815 params.lut_type = vsi->rss_lut_type;
7816 params.lut = lut;
7817
7818 status = ice_aq_get_rss_lut(hw, ¶ms);
7819 if (status)
7820 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7821 status, ice_aq_str(hw->adminq.sq_last_status));
7822
7823 return status;
7824}
7825
7826/**
7827 * ice_get_rss_key - Get RSS key
7828 * @vsi: Pointer to VSI structure
7829 * @seed: Buffer to store the key in
7830 *
7831 * Returns 0 on success, negative on failure
7832 */
7833int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7834{
7835 struct ice_hw *hw = &vsi->back->hw;
7836 int status;
7837
7838 if (!seed)
7839 return -EINVAL;
7840
7841 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7842 if (status)
7843 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7844 status, ice_aq_str(hw->adminq.sq_last_status));
7845
7846 return status;
7847}
7848
7849/**
7850 * ice_set_rss_hfunc - Set RSS HASH function
7851 * @vsi: Pointer to VSI structure
7852 * @hfunc: hash function (ICE_AQ_VSI_Q_OPT_RSS_*)
7853 *
7854 * Returns 0 on success, negative on failure
7855 */
7856int ice_set_rss_hfunc(struct ice_vsi *vsi, u8 hfunc)
7857{
7858 struct ice_hw *hw = &vsi->back->hw;
7859 struct ice_vsi_ctx *ctx;
7860 bool symm;
7861 int err;
7862
7863 if (hfunc == vsi->rss_hfunc)
7864 return 0;
7865
7866 if (hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_TPLZ &&
7867 hfunc != ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ)
7868 return -EOPNOTSUPP;
7869
7870 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
7871 if (!ctx)
7872 return -ENOMEM;
7873
7874 ctx->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_Q_OPT_VALID);
7875 ctx->info.q_opt_rss = vsi->info.q_opt_rss;
7876 ctx->info.q_opt_rss &= ~ICE_AQ_VSI_Q_OPT_RSS_HASH_M;
7877 ctx->info.q_opt_rss |=
7878 FIELD_PREP(ICE_AQ_VSI_Q_OPT_RSS_HASH_M, hfunc);
7879 ctx->info.q_opt_tc = vsi->info.q_opt_tc;
7880 ctx->info.q_opt_flags = vsi->info.q_opt_rss;
7881
7882 err = ice_update_vsi(hw, vsi->idx, ctx, NULL);
7883 if (err) {
7884 dev_err(ice_pf_to_dev(vsi->back), "Failed to configure RSS hash for VSI %d, error %d\n",
7885 vsi->vsi_num, err);
7886 } else {
7887 vsi->info.q_opt_rss = ctx->info.q_opt_rss;
7888 vsi->rss_hfunc = hfunc;
7889 netdev_info(vsi->netdev, "Hash function set to: %sToeplitz\n",
7890 hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ ?
7891 "Symmetric " : "");
7892 }
7893 kfree(ctx);
7894 if (err)
7895 return err;
7896
7897 /* Fix the symmetry setting for all existing RSS configurations */
7898 symm = !!(hfunc == ICE_AQ_VSI_Q_OPT_RSS_HASH_SYM_TPLZ);
7899 return ice_set_rss_cfg_symm(hw, vsi, symm);
7900}
7901
7902/**
7903 * ice_bridge_getlink - Get the hardware bridge mode
7904 * @skb: skb buff
7905 * @pid: process ID
7906 * @seq: RTNL message seq
7907 * @dev: the netdev being configured
7908 * @filter_mask: filter mask passed in
7909 * @nlflags: netlink flags passed in
7910 *
7911 * Return the bridge mode (VEB/VEPA)
7912 */
7913static int
7914ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7915 struct net_device *dev, u32 filter_mask, int nlflags)
7916{
7917 struct ice_netdev_priv *np = netdev_priv(dev);
7918 struct ice_vsi *vsi = np->vsi;
7919 struct ice_pf *pf = vsi->back;
7920 u16 bmode;
7921
7922 bmode = pf->first_sw->bridge_mode;
7923
7924 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7925 filter_mask, NULL);
7926}
7927
7928/**
7929 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7930 * @vsi: Pointer to VSI structure
7931 * @bmode: Hardware bridge mode (VEB/VEPA)
7932 *
7933 * Returns 0 on success, negative on failure
7934 */
7935static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7936{
7937 struct ice_aqc_vsi_props *vsi_props;
7938 struct ice_hw *hw = &vsi->back->hw;
7939 struct ice_vsi_ctx *ctxt;
7940 int ret;
7941
7942 vsi_props = &vsi->info;
7943
7944 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7945 if (!ctxt)
7946 return -ENOMEM;
7947
7948 ctxt->info = vsi->info;
7949
7950 if (bmode == BRIDGE_MODE_VEB)
7951 /* change from VEPA to VEB mode */
7952 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7953 else
7954 /* change from VEB to VEPA mode */
7955 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7956 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7957
7958 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7959 if (ret) {
7960 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7961 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7962 goto out;
7963 }
7964 /* Update sw flags for book keeping */
7965 vsi_props->sw_flags = ctxt->info.sw_flags;
7966
7967out:
7968 kfree(ctxt);
7969 return ret;
7970}
7971
7972/**
7973 * ice_bridge_setlink - Set the hardware bridge mode
7974 * @dev: the netdev being configured
7975 * @nlh: RTNL message
7976 * @flags: bridge setlink flags
7977 * @extack: netlink extended ack
7978 *
7979 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7980 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7981 * not already set for all VSIs connected to this switch. And also update the
7982 * unicast switch filter rules for the corresponding switch of the netdev.
7983 */
7984static int
7985ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7986 u16 __always_unused flags,
7987 struct netlink_ext_ack __always_unused *extack)
7988{
7989 struct ice_netdev_priv *np = netdev_priv(dev);
7990 struct ice_pf *pf = np->vsi->back;
7991 struct nlattr *attr, *br_spec;
7992 struct ice_hw *hw = &pf->hw;
7993 struct ice_sw *pf_sw;
7994 int rem, v, err = 0;
7995
7996 pf_sw = pf->first_sw;
7997 /* find the attribute in the netlink message */
7998 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7999 if (!br_spec)
8000 return -EINVAL;
8001
8002 nla_for_each_nested(attr, br_spec, rem) {
8003 __u16 mode;
8004
8005 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8006 continue;
8007 mode = nla_get_u16(attr);
8008 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
8009 return -EINVAL;
8010 /* Continue if bridge mode is not being flipped */
8011 if (mode == pf_sw->bridge_mode)
8012 continue;
8013 /* Iterates through the PF VSI list and update the loopback
8014 * mode of the VSI
8015 */
8016 ice_for_each_vsi(pf, v) {
8017 if (!pf->vsi[v])
8018 continue;
8019 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
8020 if (err)
8021 return err;
8022 }
8023
8024 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
8025 /* Update the unicast switch filter rules for the corresponding
8026 * switch of the netdev
8027 */
8028 err = ice_update_sw_rule_bridge_mode(hw);
8029 if (err) {
8030 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
8031 mode, err,
8032 ice_aq_str(hw->adminq.sq_last_status));
8033 /* revert hw->evb_veb */
8034 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
8035 return err;
8036 }
8037
8038 pf_sw->bridge_mode = mode;
8039 }
8040
8041 return 0;
8042}
8043
8044/**
8045 * ice_tx_timeout - Respond to a Tx Hang
8046 * @netdev: network interface device structure
8047 * @txqueue: Tx queue
8048 */
8049static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
8050{
8051 struct ice_netdev_priv *np = netdev_priv(netdev);
8052 struct ice_tx_ring *tx_ring = NULL;
8053 struct ice_vsi *vsi = np->vsi;
8054 struct ice_pf *pf = vsi->back;
8055 u32 i;
8056
8057 pf->tx_timeout_count++;
8058
8059 /* Check if PFC is enabled for the TC to which the queue belongs
8060 * to. If yes then Tx timeout is not caused by a hung queue, no
8061 * need to reset and rebuild
8062 */
8063 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
8064 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
8065 txqueue);
8066 return;
8067 }
8068
8069 /* now that we have an index, find the tx_ring struct */
8070 ice_for_each_txq(vsi, i)
8071 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
8072 if (txqueue == vsi->tx_rings[i]->q_index) {
8073 tx_ring = vsi->tx_rings[i];
8074 break;
8075 }
8076
8077 /* Reset recovery level if enough time has elapsed after last timeout.
8078 * Also ensure no new reset action happens before next timeout period.
8079 */
8080 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
8081 pf->tx_timeout_recovery_level = 1;
8082 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
8083 netdev->watchdog_timeo)))
8084 return;
8085
8086 if (tx_ring) {
8087 struct ice_hw *hw = &pf->hw;
8088 u32 head, val = 0;
8089
8090 head = FIELD_GET(QTX_COMM_HEAD_HEAD_M,
8091 rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])));
8092 /* Read interrupt register */
8093 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
8094
8095 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
8096 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
8097 head, tx_ring->next_to_use, val);
8098 }
8099
8100 pf->tx_timeout_last_recovery = jiffies;
8101 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
8102 pf->tx_timeout_recovery_level, txqueue);
8103
8104 switch (pf->tx_timeout_recovery_level) {
8105 case 1:
8106 set_bit(ICE_PFR_REQ, pf->state);
8107 break;
8108 case 2:
8109 set_bit(ICE_CORER_REQ, pf->state);
8110 break;
8111 case 3:
8112 set_bit(ICE_GLOBR_REQ, pf->state);
8113 break;
8114 default:
8115 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
8116 set_bit(ICE_DOWN, pf->state);
8117 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
8118 set_bit(ICE_SERVICE_DIS, pf->state);
8119 break;
8120 }
8121
8122 ice_service_task_schedule(pf);
8123 pf->tx_timeout_recovery_level++;
8124}
8125
8126/**
8127 * ice_setup_tc_cls_flower - flower classifier offloads
8128 * @np: net device to configure
8129 * @filter_dev: device on which filter is added
8130 * @cls_flower: offload data
8131 */
8132static int
8133ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
8134 struct net_device *filter_dev,
8135 struct flow_cls_offload *cls_flower)
8136{
8137 struct ice_vsi *vsi = np->vsi;
8138
8139 if (cls_flower->common.chain_index)
8140 return -EOPNOTSUPP;
8141
8142 switch (cls_flower->command) {
8143 case FLOW_CLS_REPLACE:
8144 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
8145 case FLOW_CLS_DESTROY:
8146 return ice_del_cls_flower(vsi, cls_flower);
8147 default:
8148 return -EINVAL;
8149 }
8150}
8151
8152/**
8153 * ice_setup_tc_block_cb - callback handler registered for TC block
8154 * @type: TC SETUP type
8155 * @type_data: TC flower offload data that contains user input
8156 * @cb_priv: netdev private data
8157 */
8158static int
8159ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
8160{
8161 struct ice_netdev_priv *np = cb_priv;
8162
8163 switch (type) {
8164 case TC_SETUP_CLSFLOWER:
8165 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
8166 type_data);
8167 default:
8168 return -EOPNOTSUPP;
8169 }
8170}
8171
8172/**
8173 * ice_validate_mqprio_qopt - Validate TCF input parameters
8174 * @vsi: Pointer to VSI
8175 * @mqprio_qopt: input parameters for mqprio queue configuration
8176 *
8177 * This function validates MQPRIO params, such as qcount (power of 2 wherever
8178 * needed), and make sure user doesn't specify qcount and BW rate limit
8179 * for TCs, which are more than "num_tc"
8180 */
8181static int
8182ice_validate_mqprio_qopt(struct ice_vsi *vsi,
8183 struct tc_mqprio_qopt_offload *mqprio_qopt)
8184{
8185 int non_power_of_2_qcount = 0;
8186 struct ice_pf *pf = vsi->back;
8187 int max_rss_q_cnt = 0;
8188 u64 sum_min_rate = 0;
8189 struct device *dev;
8190 int i, speed;
8191 u8 num_tc;
8192
8193 if (vsi->type != ICE_VSI_PF)
8194 return -EINVAL;
8195
8196 if (mqprio_qopt->qopt.offset[0] != 0 ||
8197 mqprio_qopt->qopt.num_tc < 1 ||
8198 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
8199 return -EINVAL;
8200
8201 dev = ice_pf_to_dev(pf);
8202 vsi->ch_rss_size = 0;
8203 num_tc = mqprio_qopt->qopt.num_tc;
8204 speed = ice_get_link_speed_kbps(vsi);
8205
8206 for (i = 0; num_tc; i++) {
8207 int qcount = mqprio_qopt->qopt.count[i];
8208 u64 max_rate, min_rate, rem;
8209
8210 if (!qcount)
8211 return -EINVAL;
8212
8213 if (is_power_of_2(qcount)) {
8214 if (non_power_of_2_qcount &&
8215 qcount > non_power_of_2_qcount) {
8216 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
8217 qcount, non_power_of_2_qcount);
8218 return -EINVAL;
8219 }
8220 if (qcount > max_rss_q_cnt)
8221 max_rss_q_cnt = qcount;
8222 } else {
8223 if (non_power_of_2_qcount &&
8224 qcount != non_power_of_2_qcount) {
8225 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
8226 qcount, non_power_of_2_qcount);
8227 return -EINVAL;
8228 }
8229 if (qcount < max_rss_q_cnt) {
8230 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
8231 qcount, max_rss_q_cnt);
8232 return -EINVAL;
8233 }
8234 max_rss_q_cnt = qcount;
8235 non_power_of_2_qcount = qcount;
8236 }
8237
8238 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
8239 * converts the bandwidth rate limit into Bytes/s when
8240 * passing it down to the driver. So convert input bandwidth
8241 * from Bytes/s to Kbps
8242 */
8243 max_rate = mqprio_qopt->max_rate[i];
8244 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
8245
8246 /* min_rate is minimum guaranteed rate and it can't be zero */
8247 min_rate = mqprio_qopt->min_rate[i];
8248 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
8249 sum_min_rate += min_rate;
8250
8251 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
8252 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
8253 min_rate, ICE_MIN_BW_LIMIT);
8254 return -EINVAL;
8255 }
8256
8257 if (max_rate && max_rate > speed) {
8258 dev_err(dev, "TC%d: max_rate(%llu Kbps) > link speed of %u Kbps\n",
8259 i, max_rate, speed);
8260 return -EINVAL;
8261 }
8262
8263 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
8264 if (rem) {
8265 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
8266 i, ICE_MIN_BW_LIMIT);
8267 return -EINVAL;
8268 }
8269
8270 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
8271 if (rem) {
8272 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
8273 i, ICE_MIN_BW_LIMIT);
8274 return -EINVAL;
8275 }
8276
8277 /* min_rate can't be more than max_rate, except when max_rate
8278 * is zero (implies max_rate sought is max line rate). In such
8279 * a case min_rate can be more than max.
8280 */
8281 if (max_rate && min_rate > max_rate) {
8282 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
8283 min_rate, max_rate);
8284 return -EINVAL;
8285 }
8286
8287 if (i >= mqprio_qopt->qopt.num_tc - 1)
8288 break;
8289 if (mqprio_qopt->qopt.offset[i + 1] !=
8290 (mqprio_qopt->qopt.offset[i] + qcount))
8291 return -EINVAL;
8292 }
8293 if (vsi->num_rxq <
8294 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8295 return -EINVAL;
8296 if (vsi->num_txq <
8297 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8298 return -EINVAL;
8299
8300 if (sum_min_rate && sum_min_rate > (u64)speed) {
8301 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
8302 sum_min_rate, speed);
8303 return -EINVAL;
8304 }
8305
8306 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
8307 vsi->ch_rss_size = max_rss_q_cnt;
8308
8309 return 0;
8310}
8311
8312/**
8313 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
8314 * @pf: ptr to PF device
8315 * @vsi: ptr to VSI
8316 */
8317static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
8318{
8319 struct device *dev = ice_pf_to_dev(pf);
8320 bool added = false;
8321 struct ice_hw *hw;
8322 int flow;
8323
8324 if (!(vsi->num_gfltr || vsi->num_bfltr))
8325 return -EINVAL;
8326
8327 hw = &pf->hw;
8328 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
8329 struct ice_fd_hw_prof *prof;
8330 int tun, status;
8331 u64 entry_h;
8332
8333 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
8334 hw->fdir_prof[flow]->cnt))
8335 continue;
8336
8337 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
8338 enum ice_flow_priority prio;
8339
8340 /* add this VSI to FDir profile for this flow */
8341 prio = ICE_FLOW_PRIO_NORMAL;
8342 prof = hw->fdir_prof[flow];
8343 status = ice_flow_add_entry(hw, ICE_BLK_FD,
8344 prof->prof_id[tun],
8345 prof->vsi_h[0], vsi->idx,
8346 prio, prof->fdir_seg[tun],
8347 &entry_h);
8348 if (status) {
8349 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
8350 vsi->idx, flow);
8351 continue;
8352 }
8353
8354 prof->entry_h[prof->cnt][tun] = entry_h;
8355 }
8356
8357 /* store VSI for filter replay and delete */
8358 prof->vsi_h[prof->cnt] = vsi->idx;
8359 prof->cnt++;
8360
8361 added = true;
8362 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
8363 flow);
8364 }
8365
8366 if (!added)
8367 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
8368
8369 return 0;
8370}
8371
8372/**
8373 * ice_add_channel - add a channel by adding VSI
8374 * @pf: ptr to PF device
8375 * @sw_id: underlying HW switching element ID
8376 * @ch: ptr to channel structure
8377 *
8378 * Add a channel (VSI) using add_vsi and queue_map
8379 */
8380static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
8381{
8382 struct device *dev = ice_pf_to_dev(pf);
8383 struct ice_vsi *vsi;
8384
8385 if (ch->type != ICE_VSI_CHNL) {
8386 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
8387 return -EINVAL;
8388 }
8389
8390 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
8391 if (!vsi || vsi->type != ICE_VSI_CHNL) {
8392 dev_err(dev, "create chnl VSI failure\n");
8393 return -EINVAL;
8394 }
8395
8396 ice_add_vsi_to_fdir(pf, vsi);
8397
8398 ch->sw_id = sw_id;
8399 ch->vsi_num = vsi->vsi_num;
8400 ch->info.mapping_flags = vsi->info.mapping_flags;
8401 ch->ch_vsi = vsi;
8402 /* set the back pointer of channel for newly created VSI */
8403 vsi->ch = ch;
8404
8405 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
8406 sizeof(vsi->info.q_mapping));
8407 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
8408 sizeof(vsi->info.tc_mapping));
8409
8410 return 0;
8411}
8412
8413/**
8414 * ice_chnl_cfg_res
8415 * @vsi: the VSI being setup
8416 * @ch: ptr to channel structure
8417 *
8418 * Configure channel specific resources such as rings, vector.
8419 */
8420static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
8421{
8422 int i;
8423
8424 for (i = 0; i < ch->num_txq; i++) {
8425 struct ice_q_vector *tx_q_vector, *rx_q_vector;
8426 struct ice_ring_container *rc;
8427 struct ice_tx_ring *tx_ring;
8428 struct ice_rx_ring *rx_ring;
8429
8430 tx_ring = vsi->tx_rings[ch->base_q + i];
8431 rx_ring = vsi->rx_rings[ch->base_q + i];
8432 if (!tx_ring || !rx_ring)
8433 continue;
8434
8435 /* setup ring being channel enabled */
8436 tx_ring->ch = ch;
8437 rx_ring->ch = ch;
8438
8439 /* following code block sets up vector specific attributes */
8440 tx_q_vector = tx_ring->q_vector;
8441 rx_q_vector = rx_ring->q_vector;
8442 if (!tx_q_vector && !rx_q_vector)
8443 continue;
8444
8445 if (tx_q_vector) {
8446 tx_q_vector->ch = ch;
8447 /* setup Tx and Rx ITR setting if DIM is off */
8448 rc = &tx_q_vector->tx;
8449 if (!ITR_IS_DYNAMIC(rc))
8450 ice_write_itr(rc, rc->itr_setting);
8451 }
8452 if (rx_q_vector) {
8453 rx_q_vector->ch = ch;
8454 /* setup Tx and Rx ITR setting if DIM is off */
8455 rc = &rx_q_vector->rx;
8456 if (!ITR_IS_DYNAMIC(rc))
8457 ice_write_itr(rc, rc->itr_setting);
8458 }
8459 }
8460
8461 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
8462 * GLINT_ITR register would have written to perform in-context
8463 * update, hence perform flush
8464 */
8465 if (ch->num_txq || ch->num_rxq)
8466 ice_flush(&vsi->back->hw);
8467}
8468
8469/**
8470 * ice_cfg_chnl_all_res - configure channel resources
8471 * @vsi: pte to main_vsi
8472 * @ch: ptr to channel structure
8473 *
8474 * This function configures channel specific resources such as flow-director
8475 * counter index, and other resources such as queues, vectors, ITR settings
8476 */
8477static void
8478ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8479{
8480 /* configure channel (aka ADQ) resources such as queues, vectors,
8481 * ITR settings for channel specific vectors and anything else
8482 */
8483 ice_chnl_cfg_res(vsi, ch);
8484}
8485
8486/**
8487 * ice_setup_hw_channel - setup new channel
8488 * @pf: ptr to PF device
8489 * @vsi: the VSI being setup
8490 * @ch: ptr to channel structure
8491 * @sw_id: underlying HW switching element ID
8492 * @type: type of channel to be created (VMDq2/VF)
8493 *
8494 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8495 * and configures Tx rings accordingly
8496 */
8497static int
8498ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8499 struct ice_channel *ch, u16 sw_id, u8 type)
8500{
8501 struct device *dev = ice_pf_to_dev(pf);
8502 int ret;
8503
8504 ch->base_q = vsi->next_base_q;
8505 ch->type = type;
8506
8507 ret = ice_add_channel(pf, sw_id, ch);
8508 if (ret) {
8509 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8510 return ret;
8511 }
8512
8513 /* configure/setup ADQ specific resources */
8514 ice_cfg_chnl_all_res(vsi, ch);
8515
8516 /* make sure to update the next_base_q so that subsequent channel's
8517 * (aka ADQ) VSI queue map is correct
8518 */
8519 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8520 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8521 ch->num_rxq);
8522
8523 return 0;
8524}
8525
8526/**
8527 * ice_setup_channel - setup new channel using uplink element
8528 * @pf: ptr to PF device
8529 * @vsi: the VSI being setup
8530 * @ch: ptr to channel structure
8531 *
8532 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8533 * and uplink switching element
8534 */
8535static bool
8536ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8537 struct ice_channel *ch)
8538{
8539 struct device *dev = ice_pf_to_dev(pf);
8540 u16 sw_id;
8541 int ret;
8542
8543 if (vsi->type != ICE_VSI_PF) {
8544 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8545 return false;
8546 }
8547
8548 sw_id = pf->first_sw->sw_id;
8549
8550 /* create channel (VSI) */
8551 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8552 if (ret) {
8553 dev_err(dev, "failed to setup hw_channel\n");
8554 return false;
8555 }
8556 dev_dbg(dev, "successfully created channel()\n");
8557
8558 return ch->ch_vsi ? true : false;
8559}
8560
8561/**
8562 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8563 * @vsi: VSI to be configured
8564 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8565 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8566 */
8567static int
8568ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8569{
8570 int err;
8571
8572 err = ice_set_min_bw_limit(vsi, min_tx_rate);
8573 if (err)
8574 return err;
8575
8576 return ice_set_max_bw_limit(vsi, max_tx_rate);
8577}
8578
8579/**
8580 * ice_create_q_channel - function to create channel
8581 * @vsi: VSI to be configured
8582 * @ch: ptr to channel (it contains channel specific params)
8583 *
8584 * This function creates channel (VSI) using num_queues specified by user,
8585 * reconfigs RSS if needed.
8586 */
8587static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8588{
8589 struct ice_pf *pf = vsi->back;
8590 struct device *dev;
8591
8592 if (!ch)
8593 return -EINVAL;
8594
8595 dev = ice_pf_to_dev(pf);
8596 if (!ch->num_txq || !ch->num_rxq) {
8597 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8598 return -EINVAL;
8599 }
8600
8601 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8602 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8603 vsi->cnt_q_avail, ch->num_txq);
8604 return -EINVAL;
8605 }
8606
8607 if (!ice_setup_channel(pf, vsi, ch)) {
8608 dev_info(dev, "Failed to setup channel\n");
8609 return -EINVAL;
8610 }
8611 /* configure BW rate limit */
8612 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8613 int ret;
8614
8615 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8616 ch->min_tx_rate);
8617 if (ret)
8618 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8619 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8620 else
8621 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8622 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8623 }
8624
8625 vsi->cnt_q_avail -= ch->num_txq;
8626
8627 return 0;
8628}
8629
8630/**
8631 * ice_rem_all_chnl_fltrs - removes all channel filters
8632 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8633 *
8634 * Remove all advanced switch filters only if they are channel specific
8635 * tc-flower based filter
8636 */
8637static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8638{
8639 struct ice_tc_flower_fltr *fltr;
8640 struct hlist_node *node;
8641
8642 /* to remove all channel filters, iterate an ordered list of filters */
8643 hlist_for_each_entry_safe(fltr, node,
8644 &pf->tc_flower_fltr_list,
8645 tc_flower_node) {
8646 struct ice_rule_query_data rule;
8647 int status;
8648
8649 /* for now process only channel specific filters */
8650 if (!ice_is_chnl_fltr(fltr))
8651 continue;
8652
8653 rule.rid = fltr->rid;
8654 rule.rule_id = fltr->rule_id;
8655 rule.vsi_handle = fltr->dest_vsi_handle;
8656 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8657 if (status) {
8658 if (status == -ENOENT)
8659 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8660 rule.rule_id);
8661 else
8662 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8663 status);
8664 } else if (fltr->dest_vsi) {
8665 /* update advanced switch filter count */
8666 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8667 u32 flags = fltr->flags;
8668
8669 fltr->dest_vsi->num_chnl_fltr--;
8670 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8671 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8672 pf->num_dmac_chnl_fltrs--;
8673 }
8674 }
8675
8676 hlist_del(&fltr->tc_flower_node);
8677 kfree(fltr);
8678 }
8679}
8680
8681/**
8682 * ice_remove_q_channels - Remove queue channels for the TCs
8683 * @vsi: VSI to be configured
8684 * @rem_fltr: delete advanced switch filter or not
8685 *
8686 * Remove queue channels for the TCs
8687 */
8688static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8689{
8690 struct ice_channel *ch, *ch_tmp;
8691 struct ice_pf *pf = vsi->back;
8692 int i;
8693
8694 /* remove all tc-flower based filter if they are channel filters only */
8695 if (rem_fltr)
8696 ice_rem_all_chnl_fltrs(pf);
8697
8698 /* remove ntuple filters since queue configuration is being changed */
8699 if (vsi->netdev->features & NETIF_F_NTUPLE) {
8700 struct ice_hw *hw = &pf->hw;
8701
8702 mutex_lock(&hw->fdir_fltr_lock);
8703 ice_fdir_del_all_fltrs(vsi);
8704 mutex_unlock(&hw->fdir_fltr_lock);
8705 }
8706
8707 /* perform cleanup for channels if they exist */
8708 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8709 struct ice_vsi *ch_vsi;
8710
8711 list_del(&ch->list);
8712 ch_vsi = ch->ch_vsi;
8713 if (!ch_vsi) {
8714 kfree(ch);
8715 continue;
8716 }
8717
8718 /* Reset queue contexts */
8719 for (i = 0; i < ch->num_rxq; i++) {
8720 struct ice_tx_ring *tx_ring;
8721 struct ice_rx_ring *rx_ring;
8722
8723 tx_ring = vsi->tx_rings[ch->base_q + i];
8724 rx_ring = vsi->rx_rings[ch->base_q + i];
8725 if (tx_ring) {
8726 tx_ring->ch = NULL;
8727 if (tx_ring->q_vector)
8728 tx_ring->q_vector->ch = NULL;
8729 }
8730 if (rx_ring) {
8731 rx_ring->ch = NULL;
8732 if (rx_ring->q_vector)
8733 rx_ring->q_vector->ch = NULL;
8734 }
8735 }
8736
8737 /* Release FD resources for the channel VSI */
8738 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8739
8740 /* clear the VSI from scheduler tree */
8741 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8742
8743 /* Delete VSI from FW, PF and HW VSI arrays */
8744 ice_vsi_delete(ch->ch_vsi);
8745
8746 /* free the channel */
8747 kfree(ch);
8748 }
8749
8750 /* clear the channel VSI map which is stored in main VSI */
8751 ice_for_each_chnl_tc(i)
8752 vsi->tc_map_vsi[i] = NULL;
8753
8754 /* reset main VSI's all TC information */
8755 vsi->all_enatc = 0;
8756 vsi->all_numtc = 0;
8757}
8758
8759/**
8760 * ice_rebuild_channels - rebuild channel
8761 * @pf: ptr to PF
8762 *
8763 * Recreate channel VSIs and replay filters
8764 */
8765static int ice_rebuild_channels(struct ice_pf *pf)
8766{
8767 struct device *dev = ice_pf_to_dev(pf);
8768 struct ice_vsi *main_vsi;
8769 bool rem_adv_fltr = true;
8770 struct ice_channel *ch;
8771 struct ice_vsi *vsi;
8772 int tc_idx = 1;
8773 int i, err;
8774
8775 main_vsi = ice_get_main_vsi(pf);
8776 if (!main_vsi)
8777 return 0;
8778
8779 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8780 main_vsi->old_numtc == 1)
8781 return 0; /* nothing to be done */
8782
8783 /* reconfigure main VSI based on old value of TC and cached values
8784 * for MQPRIO opts
8785 */
8786 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8787 if (err) {
8788 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8789 main_vsi->old_ena_tc, main_vsi->vsi_num);
8790 return err;
8791 }
8792
8793 /* rebuild ADQ VSIs */
8794 ice_for_each_vsi(pf, i) {
8795 enum ice_vsi_type type;
8796
8797 vsi = pf->vsi[i];
8798 if (!vsi || vsi->type != ICE_VSI_CHNL)
8799 continue;
8800
8801 type = vsi->type;
8802
8803 /* rebuild ADQ VSI */
8804 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
8805 if (err) {
8806 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8807 ice_vsi_type_str(type), vsi->idx, err);
8808 goto cleanup;
8809 }
8810
8811 /* Re-map HW VSI number, using VSI handle that has been
8812 * previously validated in ice_replay_vsi() call above
8813 */
8814 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8815
8816 /* replay filters for the VSI */
8817 err = ice_replay_vsi(&pf->hw, vsi->idx);
8818 if (err) {
8819 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8820 ice_vsi_type_str(type), err, vsi->idx);
8821 rem_adv_fltr = false;
8822 goto cleanup;
8823 }
8824 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8825 ice_vsi_type_str(type), vsi->idx);
8826
8827 /* store ADQ VSI at correct TC index in main VSI's
8828 * map of TC to VSI
8829 */
8830 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8831 }
8832
8833 /* ADQ VSI(s) has been rebuilt successfully, so setup
8834 * channel for main VSI's Tx and Rx rings
8835 */
8836 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8837 struct ice_vsi *ch_vsi;
8838
8839 ch_vsi = ch->ch_vsi;
8840 if (!ch_vsi)
8841 continue;
8842
8843 /* reconfig channel resources */
8844 ice_cfg_chnl_all_res(main_vsi, ch);
8845
8846 /* replay BW rate limit if it is non-zero */
8847 if (!ch->max_tx_rate && !ch->min_tx_rate)
8848 continue;
8849
8850 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8851 ch->min_tx_rate);
8852 if (err)
8853 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",
8854 err, ch->max_tx_rate, ch->min_tx_rate,
8855 ch_vsi->vsi_num);
8856 else
8857 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8858 ch->max_tx_rate, ch->min_tx_rate,
8859 ch_vsi->vsi_num);
8860 }
8861
8862 /* reconfig RSS for main VSI */
8863 if (main_vsi->ch_rss_size)
8864 ice_vsi_cfg_rss_lut_key(main_vsi);
8865
8866 return 0;
8867
8868cleanup:
8869 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8870 return err;
8871}
8872
8873/**
8874 * ice_create_q_channels - Add queue channel for the given TCs
8875 * @vsi: VSI to be configured
8876 *
8877 * Configures queue channel mapping to the given TCs
8878 */
8879static int ice_create_q_channels(struct ice_vsi *vsi)
8880{
8881 struct ice_pf *pf = vsi->back;
8882 struct ice_channel *ch;
8883 int ret = 0, i;
8884
8885 ice_for_each_chnl_tc(i) {
8886 if (!(vsi->all_enatc & BIT(i)))
8887 continue;
8888
8889 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8890 if (!ch) {
8891 ret = -ENOMEM;
8892 goto err_free;
8893 }
8894 INIT_LIST_HEAD(&ch->list);
8895 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8896 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8897 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8898 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8899 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8900
8901 /* convert to Kbits/s */
8902 if (ch->max_tx_rate)
8903 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8904 ICE_BW_KBPS_DIVISOR);
8905 if (ch->min_tx_rate)
8906 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8907 ICE_BW_KBPS_DIVISOR);
8908
8909 ret = ice_create_q_channel(vsi, ch);
8910 if (ret) {
8911 dev_err(ice_pf_to_dev(pf),
8912 "failed creating channel TC:%d\n", i);
8913 kfree(ch);
8914 goto err_free;
8915 }
8916 list_add_tail(&ch->list, &vsi->ch_list);
8917 vsi->tc_map_vsi[i] = ch->ch_vsi;
8918 dev_dbg(ice_pf_to_dev(pf),
8919 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8920 }
8921 return 0;
8922
8923err_free:
8924 ice_remove_q_channels(vsi, false);
8925
8926 return ret;
8927}
8928
8929/**
8930 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8931 * @netdev: net device to configure
8932 * @type_data: TC offload data
8933 */
8934static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8935{
8936 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8937 struct ice_netdev_priv *np = netdev_priv(netdev);
8938 struct ice_vsi *vsi = np->vsi;
8939 struct ice_pf *pf = vsi->back;
8940 u16 mode, ena_tc_qdisc = 0;
8941 int cur_txq, cur_rxq;
8942 u8 hw = 0, num_tcf;
8943 struct device *dev;
8944 int ret, i;
8945
8946 dev = ice_pf_to_dev(pf);
8947 num_tcf = mqprio_qopt->qopt.num_tc;
8948 hw = mqprio_qopt->qopt.hw;
8949 mode = mqprio_qopt->mode;
8950 if (!hw) {
8951 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8952 vsi->ch_rss_size = 0;
8953 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8954 goto config_tcf;
8955 }
8956
8957 /* Generate queue region map for number of TCF requested */
8958 for (i = 0; i < num_tcf; i++)
8959 ena_tc_qdisc |= BIT(i);
8960
8961 switch (mode) {
8962 case TC_MQPRIO_MODE_CHANNEL:
8963
8964 if (pf->hw.port_info->is_custom_tx_enabled) {
8965 dev_err(dev, "Custom Tx scheduler feature enabled, can't configure ADQ\n");
8966 return -EBUSY;
8967 }
8968 ice_tear_down_devlink_rate_tree(pf);
8969
8970 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8971 if (ret) {
8972 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8973 ret);
8974 return ret;
8975 }
8976 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8977 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8978 /* don't assume state of hw_tc_offload during driver load
8979 * and set the flag for TC flower filter if hw_tc_offload
8980 * already ON
8981 */
8982 if (vsi->netdev->features & NETIF_F_HW_TC)
8983 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8984 break;
8985 default:
8986 return -EINVAL;
8987 }
8988
8989config_tcf:
8990
8991 /* Requesting same TCF configuration as already enabled */
8992 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8993 mode != TC_MQPRIO_MODE_CHANNEL)
8994 return 0;
8995
8996 /* Pause VSI queues */
8997 ice_dis_vsi(vsi, true);
8998
8999 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
9000 ice_remove_q_channels(vsi, true);
9001
9002 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
9003 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
9004 num_online_cpus());
9005 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
9006 num_online_cpus());
9007 } else {
9008 /* logic to rebuild VSI, same like ethtool -L */
9009 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
9010
9011 for (i = 0; i < num_tcf; i++) {
9012 if (!(ena_tc_qdisc & BIT(i)))
9013 continue;
9014
9015 offset = vsi->mqprio_qopt.qopt.offset[i];
9016 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
9017 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
9018 }
9019 vsi->req_txq = offset + qcount_tx;
9020 vsi->req_rxq = offset + qcount_rx;
9021
9022 /* store away original rss_size info, so that it gets reused
9023 * form ice_vsi_rebuild during tc-qdisc delete stage - to
9024 * determine, what should be the rss_sizefor main VSI
9025 */
9026 vsi->orig_rss_size = vsi->rss_size;
9027 }
9028
9029 /* save current values of Tx and Rx queues before calling VSI rebuild
9030 * for fallback option
9031 */
9032 cur_txq = vsi->num_txq;
9033 cur_rxq = vsi->num_rxq;
9034
9035 /* proceed with rebuild main VSI using correct number of queues */
9036 ret = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
9037 if (ret) {
9038 /* fallback to current number of queues */
9039 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
9040 vsi->req_txq = cur_txq;
9041 vsi->req_rxq = cur_rxq;
9042 clear_bit(ICE_RESET_FAILED, pf->state);
9043 if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) {
9044 dev_err(dev, "Rebuild of main VSI failed again\n");
9045 return ret;
9046 }
9047 }
9048
9049 vsi->all_numtc = num_tcf;
9050 vsi->all_enatc = ena_tc_qdisc;
9051 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
9052 if (ret) {
9053 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
9054 vsi->vsi_num);
9055 goto exit;
9056 }
9057
9058 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
9059 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
9060 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
9061
9062 /* set TC0 rate limit if specified */
9063 if (max_tx_rate || min_tx_rate) {
9064 /* convert to Kbits/s */
9065 if (max_tx_rate)
9066 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
9067 if (min_tx_rate)
9068 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
9069
9070 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
9071 if (!ret) {
9072 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
9073 max_tx_rate, min_tx_rate, vsi->vsi_num);
9074 } else {
9075 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
9076 max_tx_rate, min_tx_rate, vsi->vsi_num);
9077 goto exit;
9078 }
9079 }
9080 ret = ice_create_q_channels(vsi);
9081 if (ret) {
9082 netdev_err(netdev, "failed configuring queue channels\n");
9083 goto exit;
9084 } else {
9085 netdev_dbg(netdev, "successfully configured channels\n");
9086 }
9087 }
9088
9089 if (vsi->ch_rss_size)
9090 ice_vsi_cfg_rss_lut_key(vsi);
9091
9092exit:
9093 /* if error, reset the all_numtc and all_enatc */
9094 if (ret) {
9095 vsi->all_numtc = 0;
9096 vsi->all_enatc = 0;
9097 }
9098 /* resume VSI */
9099 ice_ena_vsi(vsi, true);
9100
9101 return ret;
9102}
9103
9104static LIST_HEAD(ice_block_cb_list);
9105
9106static int
9107ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
9108 void *type_data)
9109{
9110 struct ice_netdev_priv *np = netdev_priv(netdev);
9111 struct ice_pf *pf = np->vsi->back;
9112 bool locked = false;
9113 int err;
9114
9115 switch (type) {
9116 case TC_SETUP_BLOCK:
9117 return flow_block_cb_setup_simple(type_data,
9118 &ice_block_cb_list,
9119 ice_setup_tc_block_cb,
9120 np, np, true);
9121 case TC_SETUP_QDISC_MQPRIO:
9122 if (ice_is_eswitch_mode_switchdev(pf)) {
9123 netdev_err(netdev, "TC MQPRIO offload not supported, switchdev is enabled\n");
9124 return -EOPNOTSUPP;
9125 }
9126
9127 if (pf->adev) {
9128 mutex_lock(&pf->adev_mutex);
9129 device_lock(&pf->adev->dev);
9130 locked = true;
9131 if (pf->adev->dev.driver) {
9132 netdev_err(netdev, "Cannot change qdisc when RDMA is active\n");
9133 err = -EBUSY;
9134 goto adev_unlock;
9135 }
9136 }
9137
9138 /* setup traffic classifier for receive side */
9139 mutex_lock(&pf->tc_mutex);
9140 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
9141 mutex_unlock(&pf->tc_mutex);
9142
9143adev_unlock:
9144 if (locked) {
9145 device_unlock(&pf->adev->dev);
9146 mutex_unlock(&pf->adev_mutex);
9147 }
9148 return err;
9149 default:
9150 return -EOPNOTSUPP;
9151 }
9152 return -EOPNOTSUPP;
9153}
9154
9155static struct ice_indr_block_priv *
9156ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
9157 struct net_device *netdev)
9158{
9159 struct ice_indr_block_priv *cb_priv;
9160
9161 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
9162 if (!cb_priv->netdev)
9163 return NULL;
9164 if (cb_priv->netdev == netdev)
9165 return cb_priv;
9166 }
9167 return NULL;
9168}
9169
9170static int
9171ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
9172 void *indr_priv)
9173{
9174 struct ice_indr_block_priv *priv = indr_priv;
9175 struct ice_netdev_priv *np = priv->np;
9176
9177 switch (type) {
9178 case TC_SETUP_CLSFLOWER:
9179 return ice_setup_tc_cls_flower(np, priv->netdev,
9180 (struct flow_cls_offload *)
9181 type_data);
9182 default:
9183 return -EOPNOTSUPP;
9184 }
9185}
9186
9187static int
9188ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
9189 struct ice_netdev_priv *np,
9190 struct flow_block_offload *f, void *data,
9191 void (*cleanup)(struct flow_block_cb *block_cb))
9192{
9193 struct ice_indr_block_priv *indr_priv;
9194 struct flow_block_cb *block_cb;
9195
9196 if (!ice_is_tunnel_supported(netdev) &&
9197 !(is_vlan_dev(netdev) &&
9198 vlan_dev_real_dev(netdev) == np->vsi->netdev))
9199 return -EOPNOTSUPP;
9200
9201 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9202 return -EOPNOTSUPP;
9203
9204 switch (f->command) {
9205 case FLOW_BLOCK_BIND:
9206 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9207 if (indr_priv)
9208 return -EEXIST;
9209
9210 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
9211 if (!indr_priv)
9212 return -ENOMEM;
9213
9214 indr_priv->netdev = netdev;
9215 indr_priv->np = np;
9216 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
9217
9218 block_cb =
9219 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
9220 indr_priv, indr_priv,
9221 ice_rep_indr_tc_block_unbind,
9222 f, netdev, sch, data, np,
9223 cleanup);
9224
9225 if (IS_ERR(block_cb)) {
9226 list_del(&indr_priv->list);
9227 kfree(indr_priv);
9228 return PTR_ERR(block_cb);
9229 }
9230 flow_block_cb_add(block_cb, f);
9231 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
9232 break;
9233 case FLOW_BLOCK_UNBIND:
9234 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9235 if (!indr_priv)
9236 return -ENOENT;
9237
9238 block_cb = flow_block_cb_lookup(f->block,
9239 ice_indr_setup_block_cb,
9240 indr_priv);
9241 if (!block_cb)
9242 return -ENOENT;
9243
9244 flow_indr_block_cb_remove(block_cb, f);
9245
9246 list_del(&block_cb->driver_list);
9247 break;
9248 default:
9249 return -EOPNOTSUPP;
9250 }
9251 return 0;
9252}
9253
9254static int
9255ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
9256 void *cb_priv, enum tc_setup_type type, void *type_data,
9257 void *data,
9258 void (*cleanup)(struct flow_block_cb *block_cb))
9259{
9260 switch (type) {
9261 case TC_SETUP_BLOCK:
9262 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
9263 data, cleanup);
9264
9265 default:
9266 return -EOPNOTSUPP;
9267 }
9268}
9269
9270/**
9271 * ice_open - Called when a network interface becomes active
9272 * @netdev: network interface device structure
9273 *
9274 * The open entry point is called when a network interface is made
9275 * active by the system (IFF_UP). At this point all resources needed
9276 * for transmit and receive operations are allocated, the interrupt
9277 * handler is registered with the OS, the netdev watchdog is enabled,
9278 * and the stack is notified that the interface is ready.
9279 *
9280 * Returns 0 on success, negative value on failure
9281 */
9282int ice_open(struct net_device *netdev)
9283{
9284 struct ice_netdev_priv *np = netdev_priv(netdev);
9285 struct ice_pf *pf = np->vsi->back;
9286
9287 if (ice_is_reset_in_progress(pf->state)) {
9288 netdev_err(netdev, "can't open net device while reset is in progress");
9289 return -EBUSY;
9290 }
9291
9292 return ice_open_internal(netdev);
9293}
9294
9295/**
9296 * ice_open_internal - Called when a network interface becomes active
9297 * @netdev: network interface device structure
9298 *
9299 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
9300 * handling routine
9301 *
9302 * Returns 0 on success, negative value on failure
9303 */
9304int ice_open_internal(struct net_device *netdev)
9305{
9306 struct ice_netdev_priv *np = netdev_priv(netdev);
9307 struct ice_vsi *vsi = np->vsi;
9308 struct ice_pf *pf = vsi->back;
9309 struct ice_port_info *pi;
9310 int err;
9311
9312 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
9313 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
9314 return -EIO;
9315 }
9316
9317 netif_carrier_off(netdev);
9318
9319 pi = vsi->port_info;
9320 err = ice_update_link_info(pi);
9321 if (err) {
9322 netdev_err(netdev, "Failed to get link info, error %d\n", err);
9323 return err;
9324 }
9325
9326 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
9327
9328 /* Set PHY if there is media, otherwise, turn off PHY */
9329 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
9330 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9331 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
9332 err = ice_init_phy_user_cfg(pi);
9333 if (err) {
9334 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
9335 err);
9336 return err;
9337 }
9338 }
9339
9340 err = ice_configure_phy(vsi);
9341 if (err) {
9342 netdev_err(netdev, "Failed to set physical link up, error %d\n",
9343 err);
9344 return err;
9345 }
9346 } else {
9347 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9348 ice_set_link(vsi, false);
9349 }
9350
9351 err = ice_vsi_open(vsi);
9352 if (err)
9353 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
9354 vsi->vsi_num, vsi->vsw->sw_id);
9355
9356 /* Update existing tunnels information */
9357 udp_tunnel_get_rx_info(netdev);
9358
9359 return err;
9360}
9361
9362/**
9363 * ice_stop - Disables a network interface
9364 * @netdev: network interface device structure
9365 *
9366 * The stop entry point is called when an interface is de-activated by the OS,
9367 * and the netdevice enters the DOWN state. The hardware is still under the
9368 * driver's control, but the netdev interface is disabled.
9369 *
9370 * Returns success only - not allowed to fail
9371 */
9372int ice_stop(struct net_device *netdev)
9373{
9374 struct ice_netdev_priv *np = netdev_priv(netdev);
9375 struct ice_vsi *vsi = np->vsi;
9376 struct ice_pf *pf = vsi->back;
9377
9378 if (ice_is_reset_in_progress(pf->state)) {
9379 netdev_err(netdev, "can't stop net device while reset is in progress");
9380 return -EBUSY;
9381 }
9382
9383 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
9384 int link_err = ice_force_phys_link_state(vsi, false);
9385
9386 if (link_err) {
9387 if (link_err == -ENOMEDIUM)
9388 netdev_info(vsi->netdev, "Skipping link reconfig - no media attached, VSI %d\n",
9389 vsi->vsi_num);
9390 else
9391 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
9392 vsi->vsi_num, link_err);
9393
9394 ice_vsi_close(vsi);
9395 return -EIO;
9396 }
9397 }
9398
9399 ice_vsi_close(vsi);
9400
9401 return 0;
9402}
9403
9404/**
9405 * ice_features_check - Validate encapsulated packet conforms to limits
9406 * @skb: skb buffer
9407 * @netdev: This port's netdev
9408 * @features: Offload features that the stack believes apply
9409 */
9410static netdev_features_t
9411ice_features_check(struct sk_buff *skb,
9412 struct net_device __always_unused *netdev,
9413 netdev_features_t features)
9414{
9415 bool gso = skb_is_gso(skb);
9416 size_t len;
9417
9418 /* No point in doing any of this if neither checksum nor GSO are
9419 * being requested for this frame. We can rule out both by just
9420 * checking for CHECKSUM_PARTIAL
9421 */
9422 if (skb->ip_summed != CHECKSUM_PARTIAL)
9423 return features;
9424
9425 /* We cannot support GSO if the MSS is going to be less than
9426 * 64 bytes. If it is then we need to drop support for GSO.
9427 */
9428 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
9429 features &= ~NETIF_F_GSO_MASK;
9430
9431 len = skb_network_offset(skb);
9432 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
9433 goto out_rm_features;
9434
9435 len = skb_network_header_len(skb);
9436 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9437 goto out_rm_features;
9438
9439 if (skb->encapsulation) {
9440 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
9441 * the case of IPIP frames, the transport header pointer is
9442 * after the inner header! So check to make sure that this
9443 * is a GRE or UDP_TUNNEL frame before doing that math.
9444 */
9445 if (gso && (skb_shinfo(skb)->gso_type &
9446 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
9447 len = skb_inner_network_header(skb) -
9448 skb_transport_header(skb);
9449 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
9450 goto out_rm_features;
9451 }
9452
9453 len = skb_inner_network_header_len(skb);
9454 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9455 goto out_rm_features;
9456 }
9457
9458 return features;
9459out_rm_features:
9460 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9461}
9462
9463static const struct net_device_ops ice_netdev_safe_mode_ops = {
9464 .ndo_open = ice_open,
9465 .ndo_stop = ice_stop,
9466 .ndo_start_xmit = ice_start_xmit,
9467 .ndo_set_mac_address = ice_set_mac_address,
9468 .ndo_validate_addr = eth_validate_addr,
9469 .ndo_change_mtu = ice_change_mtu,
9470 .ndo_get_stats64 = ice_get_stats64,
9471 .ndo_tx_timeout = ice_tx_timeout,
9472 .ndo_bpf = ice_xdp_safe_mode,
9473};
9474
9475static const struct net_device_ops ice_netdev_ops = {
9476 .ndo_open = ice_open,
9477 .ndo_stop = ice_stop,
9478 .ndo_start_xmit = ice_start_xmit,
9479 .ndo_select_queue = ice_select_queue,
9480 .ndo_features_check = ice_features_check,
9481 .ndo_fix_features = ice_fix_features,
9482 .ndo_set_rx_mode = ice_set_rx_mode,
9483 .ndo_set_mac_address = ice_set_mac_address,
9484 .ndo_validate_addr = eth_validate_addr,
9485 .ndo_change_mtu = ice_change_mtu,
9486 .ndo_get_stats64 = ice_get_stats64,
9487 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
9488 .ndo_eth_ioctl = ice_eth_ioctl,
9489 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
9490 .ndo_set_vf_mac = ice_set_vf_mac,
9491 .ndo_get_vf_config = ice_get_vf_cfg,
9492 .ndo_set_vf_trust = ice_set_vf_trust,
9493 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
9494 .ndo_set_vf_link_state = ice_set_vf_link_state,
9495 .ndo_get_vf_stats = ice_get_vf_stats,
9496 .ndo_set_vf_rate = ice_set_vf_bw,
9497 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
9498 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
9499 .ndo_setup_tc = ice_setup_tc,
9500 .ndo_set_features = ice_set_features,
9501 .ndo_bridge_getlink = ice_bridge_getlink,
9502 .ndo_bridge_setlink = ice_bridge_setlink,
9503 .ndo_fdb_add = ice_fdb_add,
9504 .ndo_fdb_del = ice_fdb_del,
9505#ifdef CONFIG_RFS_ACCEL
9506 .ndo_rx_flow_steer = ice_rx_flow_steer,
9507#endif
9508 .ndo_tx_timeout = ice_tx_timeout,
9509 .ndo_bpf = ice_xdp,
9510 .ndo_xdp_xmit = ice_xdp_xmit,
9511 .ndo_xsk_wakeup = ice_xsk_wakeup,
9512};