1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright (c) 2018, Intel Corporation. */
4 /* Intel(R) Ethernet Connection E800 Series Linux Driver */
6 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
8 #include <generated/utsrelease.h>
13 #include "ice_dcb_lib.h"
14 #include "ice_dcb_nl.h"
15 #include "ice_devlink.h"
16 /* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
17 * ice tracepoint functions. This must be done exactly once across the
20 #define CREATE_TRACE_POINTS
21 #include "ice_trace.h"
22 #include "ice_eswitch.h"
23 #include "ice_tc_lib.h"
24 #include "ice_vsi_vlan_ops.h"
25 #include <net/xdp_sock_drv.h>
27 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
28 static const char ice_driver_string[] = DRV_SUMMARY;
29 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
31 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
32 #define ICE_DDP_PKG_PATH "intel/ice/ddp/"
33 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
35 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
36 MODULE_DESCRIPTION(DRV_SUMMARY);
37 MODULE_LICENSE("GPL v2");
38 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
40 static int debug = -1;
41 module_param(debug, int, 0644);
42 #ifndef CONFIG_DYNAMIC_DEBUG
43 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
45 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
46 #endif /* !CONFIG_DYNAMIC_DEBUG */
48 DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key);
49 EXPORT_SYMBOL(ice_xdp_locking_key);
52 * ice_hw_to_dev - Get device pointer from the hardware structure
53 * @hw: pointer to the device HW structure
55 * Used to access the device pointer from compilation units which can't easily
56 * include the definition of struct ice_pf without leading to circular header
59 struct device *ice_hw_to_dev(struct ice_hw *hw)
61 struct ice_pf *pf = container_of(hw, struct ice_pf, hw);
63 return &pf->pdev->dev;
66 static struct workqueue_struct *ice_wq;
67 static const struct net_device_ops ice_netdev_safe_mode_ops;
68 static const struct net_device_ops ice_netdev_ops;
70 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
72 static void ice_vsi_release_all(struct ice_pf *pf);
74 static int ice_rebuild_channels(struct ice_pf *pf);
75 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr);
78 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
79 void *cb_priv, enum tc_setup_type type, void *type_data,
81 void (*cleanup)(struct flow_block_cb *block_cb));
83 bool netif_is_ice(struct net_device *dev)
85 return dev && (dev->netdev_ops == &ice_netdev_ops);
89 * ice_get_tx_pending - returns number of Tx descriptors not processed
90 * @ring: the ring of descriptors
92 static u16 ice_get_tx_pending(struct ice_tx_ring *ring)
96 head = ring->next_to_clean;
97 tail = ring->next_to_use;
100 return (head < tail) ?
101 tail - head : (tail + ring->count - head);
106 * ice_check_for_hang_subtask - check for and recover hung queues
107 * @pf: pointer to PF struct
109 static void ice_check_for_hang_subtask(struct ice_pf *pf)
111 struct ice_vsi *vsi = NULL;
117 ice_for_each_vsi(pf, v)
118 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
123 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
126 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
131 ice_for_each_txq(vsi, i) {
132 struct ice_tx_ring *tx_ring = vsi->tx_rings[i];
133 struct ice_ring_stats *ring_stats;
137 if (ice_ring_ch_enabled(tx_ring))
140 ring_stats = tx_ring->ring_stats;
145 /* If packet counter has not changed the queue is
146 * likely stalled, so force an interrupt for this
149 * prev_pkt would be negative if there was no
152 packets = ring_stats->stats.pkts & INT_MAX;
153 if (ring_stats->tx_stats.prev_pkt == packets) {
154 /* Trigger sw interrupt to revive the queue */
155 ice_trigger_sw_intr(hw, tx_ring->q_vector);
159 /* Memory barrier between read of packet count and call
160 * to ice_get_tx_pending()
163 ring_stats->tx_stats.prev_pkt =
164 ice_get_tx_pending(tx_ring) ? packets : -1;
170 * ice_init_mac_fltr - Set initial MAC filters
171 * @pf: board private structure
173 * Set initial set of MAC filters for PF VSI; configure filters for permanent
174 * address and broadcast address. If an error is encountered, netdevice will be
177 static int ice_init_mac_fltr(struct ice_pf *pf)
182 vsi = ice_get_main_vsi(pf);
186 perm_addr = vsi->port_info->mac.perm_addr;
187 return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
191 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
192 * @netdev: the net device on which the sync is happening
193 * @addr: MAC address to sync
195 * This is a callback function which is called by the in kernel device sync
196 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
197 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
198 * MAC filters from the hardware.
200 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
202 struct ice_netdev_priv *np = netdev_priv(netdev);
203 struct ice_vsi *vsi = np->vsi;
205 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
213 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
214 * @netdev: the net device on which the unsync is happening
215 * @addr: MAC address to unsync
217 * This is a callback function which is called by the in kernel device unsync
218 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
219 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
220 * delete the MAC filters from the hardware.
222 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
224 struct ice_netdev_priv *np = netdev_priv(netdev);
225 struct ice_vsi *vsi = np->vsi;
227 /* Under some circumstances, we might receive a request to delete our
228 * own device address from our uc list. Because we store the device
229 * address in the VSI's MAC filter list, we need to ignore such
230 * requests and not delete our device address from this list.
232 if (ether_addr_equal(addr, netdev->dev_addr))
235 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
243 * ice_vsi_fltr_changed - check if filter state changed
244 * @vsi: VSI to be checked
246 * returns true if filter state has changed, false otherwise.
248 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
250 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
251 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
255 * ice_set_promisc - Enable promiscuous mode for a given PF
256 * @vsi: the VSI being configured
257 * @promisc_m: mask of promiscuous config bits
260 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
264 if (vsi->type != ICE_VSI_PF)
267 if (ice_vsi_has_non_zero_vlans(vsi)) {
268 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
269 status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi,
272 status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
275 if (status && status != -EEXIST)
278 netdev_dbg(vsi->netdev, "set promisc filter bits for VSI %i: 0x%x\n",
279 vsi->vsi_num, promisc_m);
284 * ice_clear_promisc - Disable promiscuous mode for a given PF
285 * @vsi: the VSI being configured
286 * @promisc_m: mask of promiscuous config bits
289 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
293 if (vsi->type != ICE_VSI_PF)
296 if (ice_vsi_has_non_zero_vlans(vsi)) {
297 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
298 status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi,
301 status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
305 netdev_dbg(vsi->netdev, "clear promisc filter bits for VSI %i: 0x%x\n",
306 vsi->vsi_num, promisc_m);
311 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
312 * @vsi: ptr to the VSI
314 * Push any outstanding VSI filter changes through the AdminQ.
316 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
318 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
319 struct device *dev = ice_pf_to_dev(vsi->back);
320 struct net_device *netdev = vsi->netdev;
321 bool promisc_forced_on = false;
322 struct ice_pf *pf = vsi->back;
323 struct ice_hw *hw = &pf->hw;
324 u32 changed_flags = 0;
330 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
331 usleep_range(1000, 2000);
333 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
334 vsi->current_netdev_flags = vsi->netdev->flags;
336 INIT_LIST_HEAD(&vsi->tmp_sync_list);
337 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
339 if (ice_vsi_fltr_changed(vsi)) {
340 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
341 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
343 /* grab the netdev's addr_list_lock */
344 netif_addr_lock_bh(netdev);
345 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
346 ice_add_mac_to_unsync_list);
347 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
348 ice_add_mac_to_unsync_list);
349 /* our temp lists are populated. release lock */
350 netif_addr_unlock_bh(netdev);
353 /* Remove MAC addresses in the unsync list */
354 err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
355 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
357 netdev_err(netdev, "Failed to delete MAC filters\n");
358 /* if we failed because of alloc failures, just bail */
363 /* Add MAC addresses in the sync list */
364 err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
365 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
366 /* If filter is added successfully or already exists, do not go into
367 * 'if' condition and report it as error. Instead continue processing
368 * rest of the function.
370 if (err && err != -EEXIST) {
371 netdev_err(netdev, "Failed to add MAC filters\n");
372 /* If there is no more space for new umac filters, VSI
373 * should go into promiscuous mode. There should be some
374 * space reserved for promiscuous filters.
376 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
377 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
379 promisc_forced_on = true;
380 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
387 /* check for changes in promiscuous modes */
388 if (changed_flags & IFF_ALLMULTI) {
389 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
390 err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS);
392 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
396 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
397 err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS);
399 vsi->current_netdev_flags |= IFF_ALLMULTI;
405 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
406 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
407 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
408 if (vsi->current_netdev_flags & IFF_PROMISC) {
409 /* Apply Rx filter rule to get traffic from wire */
410 if (!ice_is_dflt_vsi_in_use(vsi->port_info)) {
411 err = ice_set_dflt_vsi(vsi);
412 if (err && err != -EEXIST) {
413 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
415 vsi->current_netdev_flags &=
420 vlan_ops->dis_rx_filtering(vsi);
422 /* promiscuous mode implies allmulticast so
423 * that VSIs that are in promiscuous mode are
424 * subscribed to multicast packets coming to
427 err = ice_set_promisc(vsi,
428 ICE_MCAST_PROMISC_BITS);
433 /* Clear Rx filter to remove traffic from wire */
434 if (ice_is_vsi_dflt_vsi(vsi)) {
435 err = ice_clear_dflt_vsi(vsi);
437 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
439 vsi->current_netdev_flags |=
443 if (vsi->netdev->features &
444 NETIF_F_HW_VLAN_CTAG_FILTER)
445 vlan_ops->ena_rx_filtering(vsi);
448 /* disable allmulti here, but only if allmulti is not
449 * still enabled for the netdev
451 if (!(vsi->current_netdev_flags & IFF_ALLMULTI)) {
452 err = ice_clear_promisc(vsi,
453 ICE_MCAST_PROMISC_BITS);
455 netdev_err(netdev, "Error %d clearing multicast promiscuous on VSI %i\n",
464 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
467 /* if something went wrong then set the changed flag so we try again */
468 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
469 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
471 clear_bit(ICE_CFG_BUSY, vsi->state);
476 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
477 * @pf: board private structure
479 static void ice_sync_fltr_subtask(struct ice_pf *pf)
483 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
486 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
488 ice_for_each_vsi(pf, v)
489 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
490 ice_vsi_sync_fltr(pf->vsi[v])) {
491 /* come back and try again later */
492 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
498 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
500 * @locked: is the rtnl_lock already held
502 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
507 ice_for_each_vsi(pf, v)
509 ice_dis_vsi(pf->vsi[v], locked);
511 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
512 pf->pf_agg_node[node].num_vsis = 0;
514 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
515 pf->vf_agg_node[node].num_vsis = 0;
519 * ice_clear_sw_switch_recipes - clear switch recipes
520 * @pf: board private structure
522 * Mark switch recipes as not created in sw structures. There are cases where
523 * rules (especially advanced rules) need to be restored, either re-read from
524 * hardware or added again. For example after the reset. 'recp_created' flag
525 * prevents from doing that and need to be cleared upfront.
527 static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
529 struct ice_sw_recipe *recp;
532 recp = pf->hw.switch_info->recp_list;
533 for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
534 recp[i].recp_created = false;
538 * ice_prepare_for_reset - prep for reset
539 * @pf: board private structure
540 * @reset_type: reset type requested
542 * Inform or close all dependent features in prep for reset.
545 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
547 struct ice_hw *hw = &pf->hw;
552 dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
554 /* already prepared for reset */
555 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
558 ice_unplug_aux_dev(pf);
560 /* Notify VFs of impending reset */
561 if (ice_check_sq_alive(hw, &hw->mailboxq))
562 ice_vc_notify_reset(pf);
564 /* Disable VFs until reset is completed */
565 mutex_lock(&pf->vfs.table_lock);
566 ice_for_each_vf(pf, bkt, vf)
567 ice_set_vf_state_dis(vf);
568 mutex_unlock(&pf->vfs.table_lock);
570 if (ice_is_eswitch_mode_switchdev(pf)) {
571 if (reset_type != ICE_RESET_PFR)
572 ice_clear_sw_switch_recipes(pf);
575 /* release ADQ specific HW and SW resources */
576 vsi = ice_get_main_vsi(pf);
580 /* to be on safe side, reset orig_rss_size so that normal flow
581 * of deciding rss_size can take precedence
583 vsi->orig_rss_size = 0;
585 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
586 if (reset_type == ICE_RESET_PFR) {
587 vsi->old_ena_tc = vsi->all_enatc;
588 vsi->old_numtc = vsi->all_numtc;
590 ice_remove_q_channels(vsi, true);
592 /* for other reset type, do not support channel rebuild
593 * hence reset needed info
601 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
602 memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
607 /* clear SW filtering DB */
608 ice_clear_hw_tbls(hw);
609 /* disable the VSIs and their queues that are not already DOWN */
610 ice_pf_dis_all_vsi(pf, false);
612 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
613 ice_ptp_prepare_for_reset(pf);
615 if (ice_is_feature_supported(pf, ICE_F_GNSS))
619 ice_sched_clear_port(hw->port_info);
621 ice_shutdown_all_ctrlq(hw);
623 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
627 * ice_do_reset - Initiate one of many types of resets
628 * @pf: board private structure
629 * @reset_type: reset type requested before this function was called.
631 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
633 struct device *dev = ice_pf_to_dev(pf);
634 struct ice_hw *hw = &pf->hw;
636 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
638 ice_prepare_for_reset(pf, reset_type);
640 /* trigger the reset */
641 if (ice_reset(hw, reset_type)) {
642 dev_err(dev, "reset %d failed\n", reset_type);
643 set_bit(ICE_RESET_FAILED, pf->state);
644 clear_bit(ICE_RESET_OICR_RECV, pf->state);
645 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
646 clear_bit(ICE_PFR_REQ, pf->state);
647 clear_bit(ICE_CORER_REQ, pf->state);
648 clear_bit(ICE_GLOBR_REQ, pf->state);
649 wake_up(&pf->reset_wait_queue);
653 /* PFR is a bit of a special case because it doesn't result in an OICR
654 * interrupt. So for PFR, rebuild after the reset and clear the reset-
655 * associated state bits.
657 if (reset_type == ICE_RESET_PFR) {
659 ice_rebuild(pf, reset_type);
660 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
661 clear_bit(ICE_PFR_REQ, pf->state);
662 wake_up(&pf->reset_wait_queue);
663 ice_reset_all_vfs(pf);
668 * ice_reset_subtask - Set up for resetting the device and driver
669 * @pf: board private structure
671 static void ice_reset_subtask(struct ice_pf *pf)
673 enum ice_reset_req reset_type = ICE_RESET_INVAL;
675 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
676 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
677 * of reset is pending and sets bits in pf->state indicating the reset
678 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
679 * prepare for pending reset if not already (for PF software-initiated
680 * global resets the software should already be prepared for it as
681 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
682 * by firmware or software on other PFs, that bit is not set so prepare
683 * for the reset now), poll for reset done, rebuild and return.
685 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
686 /* Perform the largest reset requested */
687 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
688 reset_type = ICE_RESET_CORER;
689 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
690 reset_type = ICE_RESET_GLOBR;
691 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
692 reset_type = ICE_RESET_EMPR;
693 /* return if no valid reset type requested */
694 if (reset_type == ICE_RESET_INVAL)
696 ice_prepare_for_reset(pf, reset_type);
698 /* make sure we are ready to rebuild */
699 if (ice_check_reset(&pf->hw)) {
700 set_bit(ICE_RESET_FAILED, pf->state);
702 /* done with reset. start rebuild */
703 pf->hw.reset_ongoing = false;
704 ice_rebuild(pf, reset_type);
705 /* clear bit to resume normal operations, but
706 * ICE_NEEDS_RESTART bit is set in case rebuild failed
708 clear_bit(ICE_RESET_OICR_RECV, pf->state);
709 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
710 clear_bit(ICE_PFR_REQ, pf->state);
711 clear_bit(ICE_CORER_REQ, pf->state);
712 clear_bit(ICE_GLOBR_REQ, pf->state);
713 wake_up(&pf->reset_wait_queue);
714 ice_reset_all_vfs(pf);
720 /* No pending resets to finish processing. Check for new resets */
721 if (test_bit(ICE_PFR_REQ, pf->state))
722 reset_type = ICE_RESET_PFR;
723 if (test_bit(ICE_CORER_REQ, pf->state))
724 reset_type = ICE_RESET_CORER;
725 if (test_bit(ICE_GLOBR_REQ, pf->state))
726 reset_type = ICE_RESET_GLOBR;
727 /* If no valid reset type requested just return */
728 if (reset_type == ICE_RESET_INVAL)
731 /* reset if not already down or busy */
732 if (!test_bit(ICE_DOWN, pf->state) &&
733 !test_bit(ICE_CFG_BUSY, pf->state)) {
734 ice_do_reset(pf, reset_type);
739 * ice_print_topo_conflict - print topology conflict message
740 * @vsi: the VSI whose topology status is being checked
742 static void ice_print_topo_conflict(struct ice_vsi *vsi)
744 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
745 case ICE_AQ_LINK_TOPO_CONFLICT:
746 case ICE_AQ_LINK_MEDIA_CONFLICT:
747 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
748 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
749 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
750 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");
752 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
753 if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
754 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");
756 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");
764 * ice_print_link_msg - print link up or down message
765 * @vsi: the VSI whose link status is being queried
766 * @isup: boolean for if the link is now up or down
768 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
770 struct ice_aqc_get_phy_caps_data *caps;
771 const char *an_advertised;
782 if (vsi->current_isup == isup)
785 vsi->current_isup = isup;
788 netdev_info(vsi->netdev, "NIC Link is Down\n");
792 switch (vsi->port_info->phy.link_info.link_speed) {
793 case ICE_AQ_LINK_SPEED_100GB:
796 case ICE_AQ_LINK_SPEED_50GB:
799 case ICE_AQ_LINK_SPEED_40GB:
802 case ICE_AQ_LINK_SPEED_25GB:
805 case ICE_AQ_LINK_SPEED_20GB:
808 case ICE_AQ_LINK_SPEED_10GB:
811 case ICE_AQ_LINK_SPEED_5GB:
814 case ICE_AQ_LINK_SPEED_2500MB:
817 case ICE_AQ_LINK_SPEED_1000MB:
820 case ICE_AQ_LINK_SPEED_100MB:
828 switch (vsi->port_info->fc.current_mode) {
832 case ICE_FC_TX_PAUSE:
835 case ICE_FC_RX_PAUSE:
846 /* Get FEC mode based on negotiated link info */
847 switch (vsi->port_info->phy.link_info.fec_info) {
848 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
849 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
852 case ICE_AQ_LINK_25G_KR_FEC_EN:
853 fec = "FC-FEC/BASE-R";
860 /* check if autoneg completed, might be false due to not supported */
861 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
866 /* Get FEC mode requested based on PHY caps last SW configuration */
867 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
870 an_advertised = "Unknown";
874 status = ice_aq_get_phy_caps(vsi->port_info, false,
875 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
877 netdev_info(vsi->netdev, "Get phy capability failed.\n");
879 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
881 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
882 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
884 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
885 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
886 fec_req = "FC-FEC/BASE-R";
893 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",
894 speed, fec_req, fec, an_advertised, an, fc);
895 ice_print_topo_conflict(vsi);
899 * ice_vsi_link_event - update the VSI's netdev
900 * @vsi: the VSI on which the link event occurred
901 * @link_up: whether or not the VSI needs to be set up or down
903 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
908 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
911 if (vsi->type == ICE_VSI_PF) {
912 if (link_up == netif_carrier_ok(vsi->netdev))
916 netif_carrier_on(vsi->netdev);
917 netif_tx_wake_all_queues(vsi->netdev);
919 netif_carrier_off(vsi->netdev);
920 netif_tx_stop_all_queues(vsi->netdev);
926 * ice_set_dflt_mib - send a default config MIB to the FW
927 * @pf: private PF struct
929 * This function sends a default configuration MIB to the FW.
931 * If this function errors out at any point, the driver is still able to
932 * function. The main impact is that LFC may not operate as expected.
933 * Therefore an error state in this function should be treated with a DBG
934 * message and continue on with driver rebuild/reenable.
936 static void ice_set_dflt_mib(struct ice_pf *pf)
938 struct device *dev = ice_pf_to_dev(pf);
939 u8 mib_type, *buf, *lldpmib = NULL;
940 u16 len, typelen, offset = 0;
941 struct ice_lldp_org_tlv *tlv;
942 struct ice_hw *hw = &pf->hw;
945 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
946 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
948 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
953 /* Add ETS CFG TLV */
954 tlv = (struct ice_lldp_org_tlv *)lldpmib;
955 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
956 ICE_IEEE_ETS_TLV_LEN);
957 tlv->typelen = htons(typelen);
958 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
959 ICE_IEEE_SUBTYPE_ETS_CFG);
960 tlv->ouisubtype = htonl(ouisubtype);
965 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
966 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
967 * Octets 13 - 20 are TSA values - leave as zeros
970 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
972 tlv = (struct ice_lldp_org_tlv *)
973 ((char *)tlv + sizeof(tlv->typelen) + len);
975 /* Add ETS REC TLV */
977 tlv->typelen = htons(typelen);
979 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
980 ICE_IEEE_SUBTYPE_ETS_REC);
981 tlv->ouisubtype = htonl(ouisubtype);
983 /* First octet of buf is reserved
984 * Octets 1 - 4 map UP to TC - all UPs map to zero
985 * Octets 5 - 12 are BW values - set TC 0 to 100%.
986 * Octets 13 - 20 are TSA value - leave as zeros
990 tlv = (struct ice_lldp_org_tlv *)
991 ((char *)tlv + sizeof(tlv->typelen) + len);
993 /* Add PFC CFG TLV */
994 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
995 ICE_IEEE_PFC_TLV_LEN);
996 tlv->typelen = htons(typelen);
998 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
999 ICE_IEEE_SUBTYPE_PFC_CFG);
1000 tlv->ouisubtype = htonl(ouisubtype);
1002 /* Octet 1 left as all zeros - PFC disabled */
1004 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
1007 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
1008 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
1014 * ice_check_phy_fw_load - check if PHY FW load failed
1015 * @pf: pointer to PF struct
1016 * @link_cfg_err: bitmap from the link info structure
1018 * check if external PHY FW load failed and print an error message if it did
1020 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
1022 if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
1023 clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1027 if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
1030 if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
1031 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");
1032 set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1037 * ice_check_module_power
1038 * @pf: pointer to PF struct
1039 * @link_cfg_err: bitmap from the link info structure
1041 * check module power level returned by a previous call to aq_get_link_info
1042 * and print error messages if module power level is not supported
1044 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
1046 /* if module power level is supported, clear the flag */
1047 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
1048 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
1049 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1053 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1054 * above block didn't clear this bit, there's nothing to do
1056 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1059 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1060 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1061 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1062 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1063 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1064 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1069 * ice_check_link_cfg_err - check if link configuration failed
1070 * @pf: pointer to the PF struct
1071 * @link_cfg_err: bitmap from the link info structure
1073 * print if any link configuration failure happens due to the value in the
1074 * link_cfg_err parameter in the link info structure
1076 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1078 ice_check_module_power(pf, link_cfg_err);
1079 ice_check_phy_fw_load(pf, link_cfg_err);
1083 * ice_link_event - process the link event
1084 * @pf: PF that the link event is associated with
1085 * @pi: port_info for the port that the link event is associated with
1086 * @link_up: true if the physical link is up and false if it is down
1087 * @link_speed: current link speed received from the link event
1089 * Returns 0 on success and negative on failure
1092 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1095 struct device *dev = ice_pf_to_dev(pf);
1096 struct ice_phy_info *phy_info;
1097 struct ice_vsi *vsi;
1102 phy_info = &pi->phy;
1103 phy_info->link_info_old = phy_info->link_info;
1105 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1106 old_link_speed = phy_info->link_info_old.link_speed;
1108 /* update the link info structures and re-enable link events,
1109 * don't bail on failure due to other book keeping needed
1111 status = ice_update_link_info(pi);
1113 dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1115 ice_aq_str(pi->hw->adminq.sq_last_status));
1117 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1119 /* Check if the link state is up after updating link info, and treat
1120 * this event as an UP event since the link is actually UP now.
1122 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1125 vsi = ice_get_main_vsi(pf);
1126 if (!vsi || !vsi->port_info)
1129 /* turn off PHY if media was removed */
1130 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1131 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1132 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1133 ice_set_link(vsi, false);
1136 /* if the old link up/down and speed is the same as the new */
1137 if (link_up == old_link && link_speed == old_link_speed)
1140 ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1142 if (ice_is_dcb_active(pf)) {
1143 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1144 ice_dcb_rebuild(pf);
1147 ice_set_dflt_mib(pf);
1149 ice_vsi_link_event(vsi, link_up);
1150 ice_print_link_msg(vsi, link_up);
1152 ice_vc_notify_link_state(pf);
1158 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1159 * @pf: board private structure
1161 static void ice_watchdog_subtask(struct ice_pf *pf)
1165 /* if interface is down do nothing */
1166 if (test_bit(ICE_DOWN, pf->state) ||
1167 test_bit(ICE_CFG_BUSY, pf->state))
1170 /* make sure we don't do these things too often */
1171 if (time_before(jiffies,
1172 pf->serv_tmr_prev + pf->serv_tmr_period))
1175 pf->serv_tmr_prev = jiffies;
1177 /* Update the stats for active netdevs so the network stack
1178 * can look at updated numbers whenever it cares to
1180 ice_update_pf_stats(pf);
1181 ice_for_each_vsi(pf, i)
1182 if (pf->vsi[i] && pf->vsi[i]->netdev)
1183 ice_update_vsi_stats(pf->vsi[i]);
1187 * ice_init_link_events - enable/initialize link events
1188 * @pi: pointer to the port_info instance
1190 * Returns -EIO on failure, 0 on success
1192 static int ice_init_link_events(struct ice_port_info *pi)
1196 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1197 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1198 ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1200 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1201 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1206 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1207 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1216 * ice_handle_link_event - handle link event via ARQ
1217 * @pf: PF that the link event is associated with
1218 * @event: event structure containing link status info
1221 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1223 struct ice_aqc_get_link_status_data *link_data;
1224 struct ice_port_info *port_info;
1227 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1228 port_info = pf->hw.port_info;
1232 status = ice_link_event(pf, port_info,
1233 !!(link_data->link_info & ICE_AQ_LINK_UP),
1234 le16_to_cpu(link_data->link_speed));
1236 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1242 enum ice_aq_task_state {
1243 ICE_AQ_TASK_WAITING = 0,
1244 ICE_AQ_TASK_COMPLETE,
1245 ICE_AQ_TASK_CANCELED,
1248 struct ice_aq_task {
1249 struct hlist_node entry;
1252 struct ice_rq_event_info *event;
1253 enum ice_aq_task_state state;
1257 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1258 * @pf: pointer to the PF private structure
1259 * @opcode: the opcode to wait for
1260 * @timeout: how long to wait, in jiffies
1261 * @event: storage for the event info
1263 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1264 * current thread will be put to sleep until the specified event occurs or
1265 * until the given timeout is reached.
1267 * To obtain only the descriptor contents, pass an event without an allocated
1268 * msg_buf. If the complete data buffer is desired, allocate the
1269 * event->msg_buf with enough space ahead of time.
1271 * Returns: zero on success, or a negative error code on failure.
1273 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1274 struct ice_rq_event_info *event)
1276 struct device *dev = ice_pf_to_dev(pf);
1277 struct ice_aq_task *task;
1278 unsigned long start;
1282 task = kzalloc(sizeof(*task), GFP_KERNEL);
1286 INIT_HLIST_NODE(&task->entry);
1287 task->opcode = opcode;
1288 task->event = event;
1289 task->state = ICE_AQ_TASK_WAITING;
1291 spin_lock_bh(&pf->aq_wait_lock);
1292 hlist_add_head(&task->entry, &pf->aq_wait_list);
1293 spin_unlock_bh(&pf->aq_wait_lock);
1297 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1299 switch (task->state) {
1300 case ICE_AQ_TASK_WAITING:
1301 err = ret < 0 ? ret : -ETIMEDOUT;
1303 case ICE_AQ_TASK_CANCELED:
1304 err = ret < 0 ? ret : -ECANCELED;
1306 case ICE_AQ_TASK_COMPLETE:
1307 err = ret < 0 ? ret : 0;
1310 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1315 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1316 jiffies_to_msecs(jiffies - start),
1317 jiffies_to_msecs(timeout),
1320 spin_lock_bh(&pf->aq_wait_lock);
1321 hlist_del(&task->entry);
1322 spin_unlock_bh(&pf->aq_wait_lock);
1329 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1330 * @pf: pointer to the PF private structure
1331 * @opcode: the opcode of the event
1332 * @event: the event to check
1334 * Loops over the current list of pending threads waiting for an AdminQ event.
1335 * For each matching task, copy the contents of the event into the task
1336 * structure and wake up the thread.
1338 * If multiple threads wait for the same opcode, they will all be woken up.
1340 * Note that event->msg_buf will only be duplicated if the event has a buffer
1341 * with enough space already allocated. Otherwise, only the descriptor and
1342 * message length will be copied.
1344 * Returns: true if an event was found, false otherwise
1346 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1347 struct ice_rq_event_info *event)
1349 struct ice_aq_task *task;
1352 spin_lock_bh(&pf->aq_wait_lock);
1353 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1354 if (task->state || task->opcode != opcode)
1357 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1358 task->event->msg_len = event->msg_len;
1360 /* Only copy the data buffer if a destination was set */
1361 if (task->event->msg_buf &&
1362 task->event->buf_len > event->buf_len) {
1363 memcpy(task->event->msg_buf, event->msg_buf,
1365 task->event->buf_len = event->buf_len;
1368 task->state = ICE_AQ_TASK_COMPLETE;
1371 spin_unlock_bh(&pf->aq_wait_lock);
1374 wake_up(&pf->aq_wait_queue);
1378 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1379 * @pf: the PF private structure
1381 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1382 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1384 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1386 struct ice_aq_task *task;
1388 spin_lock_bh(&pf->aq_wait_lock);
1389 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1390 task->state = ICE_AQ_TASK_CANCELED;
1391 spin_unlock_bh(&pf->aq_wait_lock);
1393 wake_up(&pf->aq_wait_queue);
1396 #define ICE_MBX_OVERFLOW_WATERMARK 64
1399 * __ice_clean_ctrlq - helper function to clean controlq rings
1400 * @pf: ptr to struct ice_pf
1401 * @q_type: specific Control queue type
1403 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1405 struct device *dev = ice_pf_to_dev(pf);
1406 struct ice_rq_event_info event;
1407 struct ice_hw *hw = &pf->hw;
1408 struct ice_ctl_q_info *cq;
1413 /* Do not clean control queue if/when PF reset fails */
1414 if (test_bit(ICE_RESET_FAILED, pf->state))
1418 case ICE_CTL_Q_ADMIN:
1426 case ICE_CTL_Q_MAILBOX:
1429 /* we are going to try to detect a malicious VF, so set the
1430 * state to begin detection
1432 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1435 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1439 /* check for error indications - PF_xx_AxQLEN register layout for
1440 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1442 val = rd32(hw, cq->rq.len);
1443 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1444 PF_FW_ARQLEN_ARQCRIT_M)) {
1446 if (val & PF_FW_ARQLEN_ARQVFE_M)
1447 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1449 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1450 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1453 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1454 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1456 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1457 PF_FW_ARQLEN_ARQCRIT_M);
1459 wr32(hw, cq->rq.len, val);
1462 val = rd32(hw, cq->sq.len);
1463 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1464 PF_FW_ATQLEN_ATQCRIT_M)) {
1466 if (val & PF_FW_ATQLEN_ATQVFE_M)
1467 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1469 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1470 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1473 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1474 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1476 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1477 PF_FW_ATQLEN_ATQCRIT_M);
1479 wr32(hw, cq->sq.len, val);
1482 event.buf_len = cq->rq_buf_size;
1483 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1488 struct ice_mbx_data data = {};
1492 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1493 if (ret == -EALREADY)
1496 dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1501 opcode = le16_to_cpu(event.desc.opcode);
1503 /* Notify any thread that might be waiting for this event */
1504 ice_aq_check_events(pf, opcode, &event);
1507 case ice_aqc_opc_get_link_status:
1508 if (ice_handle_link_event(pf, &event))
1509 dev_err(dev, "Could not handle link event\n");
1511 case ice_aqc_opc_event_lan_overflow:
1512 ice_vf_lan_overflow_event(pf, &event);
1514 case ice_mbx_opc_send_msg_to_pf:
1515 data.num_msg_proc = i;
1516 data.num_pending_arq = pending;
1517 data.max_num_msgs_mbx = hw->mailboxq.num_rq_entries;
1518 data.async_watermark_val = ICE_MBX_OVERFLOW_WATERMARK;
1520 ice_vc_process_vf_msg(pf, &event, &data);
1522 case ice_aqc_opc_fw_logging:
1523 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1525 case ice_aqc_opc_lldp_set_mib_change:
1526 ice_dcb_process_lldp_set_mib_change(pf, &event);
1529 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1533 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1535 kfree(event.msg_buf);
1537 return pending && (i == ICE_DFLT_IRQ_WORK);
1541 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1542 * @hw: pointer to hardware info
1543 * @cq: control queue information
1545 * returns true if there are pending messages in a queue, false if there aren't
1547 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1551 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1552 return cq->rq.next_to_clean != ntu;
1556 * ice_clean_adminq_subtask - clean the AdminQ rings
1557 * @pf: board private structure
1559 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1561 struct ice_hw *hw = &pf->hw;
1563 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1566 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1569 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1571 /* There might be a situation where new messages arrive to a control
1572 * queue between processing the last message and clearing the
1573 * EVENT_PENDING bit. So before exiting, check queue head again (using
1574 * ice_ctrlq_pending) and process new messages if any.
1576 if (ice_ctrlq_pending(hw, &hw->adminq))
1577 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1583 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1584 * @pf: board private structure
1586 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1588 struct ice_hw *hw = &pf->hw;
1590 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1593 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1596 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1598 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1599 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1605 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1606 * @pf: board private structure
1608 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1610 struct ice_hw *hw = &pf->hw;
1612 /* Nothing to do here if sideband queue is not supported */
1613 if (!ice_is_sbq_supported(hw)) {
1614 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1618 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1621 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1624 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1626 if (ice_ctrlq_pending(hw, &hw->sbq))
1627 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1633 * ice_service_task_schedule - schedule the service task to wake up
1634 * @pf: board private structure
1636 * If not already scheduled, this puts the task into the work queue.
1638 void ice_service_task_schedule(struct ice_pf *pf)
1640 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1641 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1642 !test_bit(ICE_NEEDS_RESTART, pf->state))
1643 queue_work(ice_wq, &pf->serv_task);
1647 * ice_service_task_complete - finish up the service task
1648 * @pf: board private structure
1650 static void ice_service_task_complete(struct ice_pf *pf)
1652 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1654 /* force memory (pf->state) to sync before next service task */
1655 smp_mb__before_atomic();
1656 clear_bit(ICE_SERVICE_SCHED, pf->state);
1660 * ice_service_task_stop - stop service task and cancel works
1661 * @pf: board private structure
1663 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1666 static int ice_service_task_stop(struct ice_pf *pf)
1670 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1672 if (pf->serv_tmr.function)
1673 del_timer_sync(&pf->serv_tmr);
1674 if (pf->serv_task.func)
1675 cancel_work_sync(&pf->serv_task);
1677 clear_bit(ICE_SERVICE_SCHED, pf->state);
1682 * ice_service_task_restart - restart service task and schedule works
1683 * @pf: board private structure
1685 * This function is needed for suspend and resume works (e.g WoL scenario)
1687 static void ice_service_task_restart(struct ice_pf *pf)
1689 clear_bit(ICE_SERVICE_DIS, pf->state);
1690 ice_service_task_schedule(pf);
1694 * ice_service_timer - timer callback to schedule service task
1695 * @t: pointer to timer_list
1697 static void ice_service_timer(struct timer_list *t)
1699 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1701 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1702 ice_service_task_schedule(pf);
1706 * ice_handle_mdd_event - handle malicious driver detect event
1707 * @pf: pointer to the PF structure
1709 * Called from service task. OICR interrupt handler indicates MDD event.
1710 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1711 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1712 * disable the queue, the PF can be configured to reset the VF using ethtool
1713 * private flag mdd-auto-reset-vf.
1715 static void ice_handle_mdd_event(struct ice_pf *pf)
1717 struct device *dev = ice_pf_to_dev(pf);
1718 struct ice_hw *hw = &pf->hw;
1723 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1724 /* Since the VF MDD event logging is rate limited, check if
1725 * there are pending MDD events.
1727 ice_print_vfs_mdd_events(pf);
1731 /* find what triggered an MDD event */
1732 reg = rd32(hw, GL_MDET_TX_PQM);
1733 if (reg & GL_MDET_TX_PQM_VALID_M) {
1734 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1735 GL_MDET_TX_PQM_PF_NUM_S;
1736 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1737 GL_MDET_TX_PQM_VF_NUM_S;
1738 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1739 GL_MDET_TX_PQM_MAL_TYPE_S;
1740 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1741 GL_MDET_TX_PQM_QNUM_S);
1743 if (netif_msg_tx_err(pf))
1744 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1745 event, queue, pf_num, vf_num);
1746 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1749 reg = rd32(hw, GL_MDET_TX_TCLAN);
1750 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1751 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1752 GL_MDET_TX_TCLAN_PF_NUM_S;
1753 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1754 GL_MDET_TX_TCLAN_VF_NUM_S;
1755 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1756 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1757 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1758 GL_MDET_TX_TCLAN_QNUM_S);
1760 if (netif_msg_tx_err(pf))
1761 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1762 event, queue, pf_num, vf_num);
1763 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1766 reg = rd32(hw, GL_MDET_RX);
1767 if (reg & GL_MDET_RX_VALID_M) {
1768 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1769 GL_MDET_RX_PF_NUM_S;
1770 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1771 GL_MDET_RX_VF_NUM_S;
1772 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1773 GL_MDET_RX_MAL_TYPE_S;
1774 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1777 if (netif_msg_rx_err(pf))
1778 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1779 event, queue, pf_num, vf_num);
1780 wr32(hw, GL_MDET_RX, 0xffffffff);
1783 /* check to see if this PF caused an MDD event */
1784 reg = rd32(hw, PF_MDET_TX_PQM);
1785 if (reg & PF_MDET_TX_PQM_VALID_M) {
1786 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1787 if (netif_msg_tx_err(pf))
1788 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1791 reg = rd32(hw, PF_MDET_TX_TCLAN);
1792 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1793 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1794 if (netif_msg_tx_err(pf))
1795 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1798 reg = rd32(hw, PF_MDET_RX);
1799 if (reg & PF_MDET_RX_VALID_M) {
1800 wr32(hw, PF_MDET_RX, 0xFFFF);
1801 if (netif_msg_rx_err(pf))
1802 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1805 /* Check to see if one of the VFs caused an MDD event, and then
1806 * increment counters and set print pending
1808 mutex_lock(&pf->vfs.table_lock);
1809 ice_for_each_vf(pf, bkt, vf) {
1810 reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1811 if (reg & VP_MDET_TX_PQM_VALID_M) {
1812 wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1813 vf->mdd_tx_events.count++;
1814 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1815 if (netif_msg_tx_err(pf))
1816 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1820 reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1821 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1822 wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1823 vf->mdd_tx_events.count++;
1824 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1825 if (netif_msg_tx_err(pf))
1826 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1830 reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1831 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1832 wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1833 vf->mdd_tx_events.count++;
1834 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1835 if (netif_msg_tx_err(pf))
1836 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1840 reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1841 if (reg & VP_MDET_RX_VALID_M) {
1842 wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1843 vf->mdd_rx_events.count++;
1844 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1845 if (netif_msg_rx_err(pf))
1846 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1849 /* Since the queue is disabled on VF Rx MDD events, the
1850 * PF can be configured to reset the VF through ethtool
1851 * private flag mdd-auto-reset-vf.
1853 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1854 /* VF MDD event counters will be cleared by
1855 * reset, so print the event prior to reset.
1857 ice_print_vf_rx_mdd_event(vf);
1858 ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1862 mutex_unlock(&pf->vfs.table_lock);
1864 ice_print_vfs_mdd_events(pf);
1868 * ice_force_phys_link_state - Force the physical link state
1869 * @vsi: VSI to force the physical link state to up/down
1870 * @link_up: true/false indicates to set the physical link to up/down
1872 * Force the physical link state by getting the current PHY capabilities from
1873 * hardware and setting the PHY config based on the determined capabilities. If
1874 * link changes a link event will be triggered because both the Enable Automatic
1875 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1877 * Returns 0 on success, negative on failure
1879 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1881 struct ice_aqc_get_phy_caps_data *pcaps;
1882 struct ice_aqc_set_phy_cfg_data *cfg;
1883 struct ice_port_info *pi;
1887 if (!vsi || !vsi->port_info || !vsi->back)
1889 if (vsi->type != ICE_VSI_PF)
1892 dev = ice_pf_to_dev(vsi->back);
1894 pi = vsi->port_info;
1896 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1900 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1903 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1904 vsi->vsi_num, retcode);
1909 /* No change in link */
1910 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1911 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1914 /* Use the current user PHY configuration. The current user PHY
1915 * configuration is initialized during probe from PHY capabilities
1916 * software mode, and updated on set PHY configuration.
1918 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1924 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1926 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1928 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1930 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1932 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1933 vsi->vsi_num, retcode);
1944 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1945 * @pi: port info structure
1947 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1949 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1951 struct ice_aqc_get_phy_caps_data *pcaps;
1952 struct ice_pf *pf = pi->hw->back;
1955 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1959 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1963 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1967 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1968 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1976 * ice_init_link_dflt_override - Initialize link default override
1977 * @pi: port info structure
1979 * Initialize link default override and PHY total port shutdown during probe
1981 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1983 struct ice_link_default_override_tlv *ldo;
1984 struct ice_pf *pf = pi->hw->back;
1986 ldo = &pf->link_dflt_override;
1987 if (ice_get_link_default_override(ldo, pi))
1990 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1993 /* Enable Total Port Shutdown (override/replace link-down-on-close
1994 * ethtool private flag) for ports with Port Disable bit set.
1996 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1997 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
2001 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
2002 * @pi: port info structure
2004 * If default override is enabled, initialize the user PHY cfg speed and FEC
2005 * settings using the default override mask from the NVM.
2007 * The PHY should only be configured with the default override settings the
2008 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
2009 * is used to indicate that the user PHY cfg default override is initialized
2010 * and the PHY has not been configured with the default override settings. The
2011 * state is set here, and cleared in ice_configure_phy the first time the PHY is
2014 * This function should be called only if the FW doesn't support default
2015 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
2017 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
2019 struct ice_link_default_override_tlv *ldo;
2020 struct ice_aqc_set_phy_cfg_data *cfg;
2021 struct ice_phy_info *phy = &pi->phy;
2022 struct ice_pf *pf = pi->hw->back;
2024 ldo = &pf->link_dflt_override;
2026 /* If link default override is enabled, use to mask NVM PHY capabilities
2027 * for speed and FEC default configuration.
2029 cfg = &phy->curr_user_phy_cfg;
2031 if (ldo->phy_type_low || ldo->phy_type_high) {
2032 cfg->phy_type_low = pf->nvm_phy_type_lo &
2033 cpu_to_le64(ldo->phy_type_low);
2034 cfg->phy_type_high = pf->nvm_phy_type_hi &
2035 cpu_to_le64(ldo->phy_type_high);
2037 cfg->link_fec_opt = ldo->fec_options;
2038 phy->curr_user_fec_req = ICE_FEC_AUTO;
2040 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2044 * ice_init_phy_user_cfg - Initialize the PHY user configuration
2045 * @pi: port info structure
2047 * Initialize the current user PHY configuration, speed, FEC, and FC requested
2048 * mode to default. The PHY defaults are from get PHY capabilities topology
2049 * with media so call when media is first available. An error is returned if
2050 * called when media is not available. The PHY initialization completed state is
2053 * These configurations are used when setting PHY
2054 * configuration. The user PHY configuration is updated on set PHY
2055 * configuration. Returns 0 on success, negative on failure
2057 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2059 struct ice_aqc_get_phy_caps_data *pcaps;
2060 struct ice_phy_info *phy = &pi->phy;
2061 struct ice_pf *pf = pi->hw->back;
2064 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2067 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2071 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2072 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2075 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2078 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2082 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2084 /* check if lenient mode is supported and enabled */
2085 if (ice_fw_supports_link_override(pi->hw) &&
2086 !(pcaps->module_compliance_enforcement &
2087 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2088 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2090 /* if the FW supports default PHY configuration mode, then the driver
2091 * does not have to apply link override settings. If not,
2092 * initialize user PHY configuration with link override values
2094 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2095 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2096 ice_init_phy_cfg_dflt_override(pi);
2101 /* if link default override is not enabled, set user flow control and
2102 * FEC settings based on what get_phy_caps returned
2104 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2105 pcaps->link_fec_options);
2106 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2109 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2110 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2117 * ice_configure_phy - configure PHY
2120 * Set the PHY configuration. If the current PHY configuration is the same as
2121 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2122 * configure the based get PHY capabilities for topology with media.
2124 static int ice_configure_phy(struct ice_vsi *vsi)
2126 struct device *dev = ice_pf_to_dev(vsi->back);
2127 struct ice_port_info *pi = vsi->port_info;
2128 struct ice_aqc_get_phy_caps_data *pcaps;
2129 struct ice_aqc_set_phy_cfg_data *cfg;
2130 struct ice_phy_info *phy = &pi->phy;
2131 struct ice_pf *pf = vsi->back;
2134 /* Ensure we have media as we cannot configure a medialess port */
2135 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2138 ice_print_topo_conflict(vsi);
2140 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2141 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2144 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2145 return ice_force_phys_link_state(vsi, true);
2147 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2151 /* Get current PHY config */
2152 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2155 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2160 /* If PHY enable link is configured and configuration has not changed,
2161 * there's nothing to do
2163 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2164 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2167 /* Use PHY topology as baseline for configuration */
2168 memset(pcaps, 0, sizeof(*pcaps));
2169 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2170 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2173 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2176 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2181 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2187 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2189 /* Speed - If default override pending, use curr_user_phy_cfg set in
2190 * ice_init_phy_user_cfg_ldo.
2192 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2193 vsi->back->state)) {
2194 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2195 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2197 u64 phy_low = 0, phy_high = 0;
2199 ice_update_phy_type(&phy_low, &phy_high,
2200 pi->phy.curr_user_speed_req);
2201 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2202 cfg->phy_type_high = pcaps->phy_type_high &
2203 cpu_to_le64(phy_high);
2206 /* Can't provide what was requested; use PHY capabilities */
2207 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2208 cfg->phy_type_low = pcaps->phy_type_low;
2209 cfg->phy_type_high = pcaps->phy_type_high;
2213 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2215 /* Can't provide what was requested; use PHY capabilities */
2216 if (cfg->link_fec_opt !=
2217 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2218 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2219 cfg->link_fec_opt = pcaps->link_fec_options;
2222 /* Flow Control - always supported; no need to check against
2225 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2227 /* Enable link and link update */
2228 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2230 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2232 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2242 * ice_check_media_subtask - Check for media
2243 * @pf: pointer to PF struct
2245 * If media is available, then initialize PHY user configuration if it is not
2246 * been, and configure the PHY if the interface is up.
2248 static void ice_check_media_subtask(struct ice_pf *pf)
2250 struct ice_port_info *pi;
2251 struct ice_vsi *vsi;
2254 /* No need to check for media if it's already present */
2255 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2258 vsi = ice_get_main_vsi(pf);
2262 /* Refresh link info and check if media is present */
2263 pi = vsi->port_info;
2264 err = ice_update_link_info(pi);
2268 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2270 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2271 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2272 ice_init_phy_user_cfg(pi);
2274 /* PHY settings are reset on media insertion, reconfigure
2275 * PHY to preserve settings.
2277 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2278 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2281 err = ice_configure_phy(vsi);
2283 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2285 /* A Link Status Event will be generated; the event handler
2286 * will complete bringing the interface up
2292 * ice_service_task - manage and run subtasks
2293 * @work: pointer to work_struct contained by the PF struct
2295 static void ice_service_task(struct work_struct *work)
2297 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2298 unsigned long start_time = jiffies;
2302 /* process reset requests first */
2303 ice_reset_subtask(pf);
2305 /* bail if a reset/recovery cycle is pending or rebuild failed */
2306 if (ice_is_reset_in_progress(pf->state) ||
2307 test_bit(ICE_SUSPENDED, pf->state) ||
2308 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2309 ice_service_task_complete(pf);
2313 if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2314 struct iidc_event *event;
2316 event = kzalloc(sizeof(*event), GFP_KERNEL);
2318 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2319 /* report the entire OICR value to AUX driver */
2320 swap(event->reg, pf->oicr_err_reg);
2321 ice_send_event_to_aux(pf, event);
2326 /* unplug aux dev per request, if an unplug request came in
2327 * while processing a plug request, this will handle it
2329 if (test_and_clear_bit(ICE_FLAG_UNPLUG_AUX_DEV, pf->flags))
2330 ice_unplug_aux_dev(pf);
2332 /* Plug aux device per request */
2333 if (test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2334 ice_plug_aux_dev(pf);
2336 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2337 struct iidc_event *event;
2339 event = kzalloc(sizeof(*event), GFP_KERNEL);
2341 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2342 ice_send_event_to_aux(pf, event);
2347 ice_clean_adminq_subtask(pf);
2348 ice_check_media_subtask(pf);
2349 ice_check_for_hang_subtask(pf);
2350 ice_sync_fltr_subtask(pf);
2351 ice_handle_mdd_event(pf);
2352 ice_watchdog_subtask(pf);
2354 if (ice_is_safe_mode(pf)) {
2355 ice_service_task_complete(pf);
2359 ice_process_vflr_event(pf);
2360 ice_clean_mailboxq_subtask(pf);
2361 ice_clean_sbq_subtask(pf);
2362 ice_sync_arfs_fltrs(pf);
2363 ice_flush_fdir_ctx(pf);
2365 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2366 ice_service_task_complete(pf);
2368 /* If the tasks have taken longer than one service timer period
2369 * or there is more work to be done, reset the service timer to
2370 * schedule the service task now.
2372 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2373 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2374 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2375 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2376 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2377 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2378 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2379 mod_timer(&pf->serv_tmr, jiffies);
2383 * ice_set_ctrlq_len - helper function to set controlq length
2384 * @hw: pointer to the HW instance
2386 static void ice_set_ctrlq_len(struct ice_hw *hw)
2388 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2389 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2390 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2391 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2392 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2393 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2394 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2395 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2396 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2397 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2398 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2399 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2403 * ice_schedule_reset - schedule a reset
2404 * @pf: board private structure
2405 * @reset: reset being requested
2407 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2409 struct device *dev = ice_pf_to_dev(pf);
2411 /* bail out if earlier reset has failed */
2412 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2413 dev_dbg(dev, "earlier reset has failed\n");
2416 /* bail if reset/recovery already in progress */
2417 if (ice_is_reset_in_progress(pf->state)) {
2418 dev_dbg(dev, "Reset already in progress\n");
2424 set_bit(ICE_PFR_REQ, pf->state);
2426 case ICE_RESET_CORER:
2427 set_bit(ICE_CORER_REQ, pf->state);
2429 case ICE_RESET_GLOBR:
2430 set_bit(ICE_GLOBR_REQ, pf->state);
2436 ice_service_task_schedule(pf);
2441 * ice_irq_affinity_notify - Callback for affinity changes
2442 * @notify: context as to what irq was changed
2443 * @mask: the new affinity mask
2445 * This is a callback function used by the irq_set_affinity_notifier function
2446 * so that we may register to receive changes to the irq affinity masks.
2449 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2450 const cpumask_t *mask)
2452 struct ice_q_vector *q_vector =
2453 container_of(notify, struct ice_q_vector, affinity_notify);
2455 cpumask_copy(&q_vector->affinity_mask, mask);
2459 * ice_irq_affinity_release - Callback for affinity notifier release
2460 * @ref: internal core kernel usage
2462 * This is a callback function used by the irq_set_affinity_notifier function
2463 * to inform the current notification subscriber that they will no longer
2464 * receive notifications.
2466 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2469 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2470 * @vsi: the VSI being configured
2472 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2474 struct ice_hw *hw = &vsi->back->hw;
2477 ice_for_each_q_vector(vsi, i)
2478 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2485 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2486 * @vsi: the VSI being configured
2487 * @basename: name for the vector
2489 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2491 int q_vectors = vsi->num_q_vectors;
2492 struct ice_pf *pf = vsi->back;
2493 int base = vsi->base_vector;
2500 dev = ice_pf_to_dev(pf);
2501 for (vector = 0; vector < q_vectors; vector++) {
2502 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2504 irq_num = pf->msix_entries[base + vector].vector;
2506 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2507 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2508 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2510 } else if (q_vector->rx.rx_ring) {
2511 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2512 "%s-%s-%d", basename, "rx", rx_int_idx++);
2513 } else if (q_vector->tx.tx_ring) {
2514 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2515 "%s-%s-%d", basename, "tx", tx_int_idx++);
2517 /* skip this unused q_vector */
2520 if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2521 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2522 IRQF_SHARED, q_vector->name,
2525 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2526 0, q_vector->name, q_vector);
2528 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2533 /* register for affinity change notifications */
2534 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2535 struct irq_affinity_notify *affinity_notify;
2537 affinity_notify = &q_vector->affinity_notify;
2538 affinity_notify->notify = ice_irq_affinity_notify;
2539 affinity_notify->release = ice_irq_affinity_release;
2540 irq_set_affinity_notifier(irq_num, affinity_notify);
2543 /* assign the mask for this irq */
2544 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2547 err = ice_set_cpu_rx_rmap(vsi);
2549 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2550 vsi->vsi_num, ERR_PTR(err));
2554 vsi->irqs_ready = true;
2560 irq_num = pf->msix_entries[base + vector].vector;
2561 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2562 irq_set_affinity_notifier(irq_num, NULL);
2563 irq_set_affinity_hint(irq_num, NULL);
2564 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2570 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2571 * @vsi: VSI to setup Tx rings used by XDP
2573 * Return 0 on success and negative value on error
2575 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2577 struct device *dev = ice_pf_to_dev(vsi->back);
2578 struct ice_tx_desc *tx_desc;
2581 ice_for_each_xdp_txq(vsi, i) {
2582 u16 xdp_q_idx = vsi->alloc_txq + i;
2583 struct ice_ring_stats *ring_stats;
2584 struct ice_tx_ring *xdp_ring;
2586 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2588 goto free_xdp_rings;
2590 ring_stats = kzalloc(sizeof(*ring_stats), GFP_KERNEL);
2592 ice_free_tx_ring(xdp_ring);
2593 goto free_xdp_rings;
2596 xdp_ring->ring_stats = ring_stats;
2597 xdp_ring->q_index = xdp_q_idx;
2598 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2599 xdp_ring->vsi = vsi;
2600 xdp_ring->netdev = NULL;
2601 xdp_ring->dev = dev;
2602 xdp_ring->count = vsi->num_tx_desc;
2603 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2604 if (ice_setup_tx_ring(xdp_ring))
2605 goto free_xdp_rings;
2606 ice_set_ring_xdp(xdp_ring);
2607 spin_lock_init(&xdp_ring->tx_lock);
2608 for (j = 0; j < xdp_ring->count; j++) {
2609 tx_desc = ICE_TX_DESC(xdp_ring, j);
2610 tx_desc->cmd_type_offset_bsz = 0;
2617 for (; i >= 0; i--) {
2618 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc) {
2619 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2620 vsi->xdp_rings[i]->ring_stats = NULL;
2621 ice_free_tx_ring(vsi->xdp_rings[i]);
2628 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2629 * @vsi: VSI to set the bpf prog on
2630 * @prog: the bpf prog pointer
2632 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2634 struct bpf_prog *old_prog;
2637 old_prog = xchg(&vsi->xdp_prog, prog);
2639 bpf_prog_put(old_prog);
2641 ice_for_each_rxq(vsi, i)
2642 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2646 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2647 * @vsi: VSI to bring up Tx rings used by XDP
2648 * @prog: bpf program that will be assigned to VSI
2650 * Return 0 on success and negative value on error
2652 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2654 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2655 int xdp_rings_rem = vsi->num_xdp_txq;
2656 struct ice_pf *pf = vsi->back;
2657 struct ice_qs_cfg xdp_qs_cfg = {
2658 .qs_mutex = &pf->avail_q_mutex,
2659 .pf_map = pf->avail_txqs,
2660 .pf_map_size = pf->max_pf_txqs,
2661 .q_count = vsi->num_xdp_txq,
2662 .scatter_count = ICE_MAX_SCATTER_TXQS,
2663 .vsi_map = vsi->txq_map,
2664 .vsi_map_offset = vsi->alloc_txq,
2665 .mapping_mode = ICE_VSI_MAP_CONTIG
2671 dev = ice_pf_to_dev(pf);
2672 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2673 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2674 if (!vsi->xdp_rings)
2677 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2678 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2681 if (static_key_enabled(&ice_xdp_locking_key))
2682 netdev_warn(vsi->netdev,
2683 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2685 if (ice_xdp_alloc_setup_rings(vsi))
2686 goto clear_xdp_rings;
2688 /* follow the logic from ice_vsi_map_rings_to_vectors */
2689 ice_for_each_q_vector(vsi, v_idx) {
2690 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2691 int xdp_rings_per_v, q_id, q_base;
2693 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2694 vsi->num_q_vectors - v_idx);
2695 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2697 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2698 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2700 xdp_ring->q_vector = q_vector;
2701 xdp_ring->next = q_vector->tx.tx_ring;
2702 q_vector->tx.tx_ring = xdp_ring;
2704 xdp_rings_rem -= xdp_rings_per_v;
2707 ice_for_each_rxq(vsi, i) {
2708 if (static_key_enabled(&ice_xdp_locking_key)) {
2709 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2711 struct ice_q_vector *q_vector = vsi->rx_rings[i]->q_vector;
2712 struct ice_tx_ring *ring;
2714 ice_for_each_tx_ring(ring, q_vector->tx) {
2715 if (ice_ring_is_xdp(ring)) {
2716 vsi->rx_rings[i]->xdp_ring = ring;
2721 ice_tx_xsk_pool(vsi, i);
2724 /* omit the scheduler update if in reset path; XDP queues will be
2725 * taken into account at the end of ice_vsi_rebuild, where
2726 * ice_cfg_vsi_lan is being called
2728 if (ice_is_reset_in_progress(pf->state))
2731 /* tell the Tx scheduler that right now we have
2734 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2735 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2737 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2740 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2742 goto clear_xdp_rings;
2745 /* assign the prog only when it's not already present on VSI;
2746 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2747 * VSI rebuild that happens under ethtool -L can expose us to
2748 * the bpf_prog refcount issues as we would be swapping same
2749 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2750 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2751 * this is not harmful as dev_xdp_install bumps the refcount
2752 * before calling the op exposed by the driver;
2754 if (!ice_is_xdp_ena_vsi(vsi))
2755 ice_vsi_assign_bpf_prog(vsi, prog);
2759 ice_for_each_xdp_txq(vsi, i)
2760 if (vsi->xdp_rings[i]) {
2761 kfree_rcu(vsi->xdp_rings[i], rcu);
2762 vsi->xdp_rings[i] = NULL;
2766 mutex_lock(&pf->avail_q_mutex);
2767 ice_for_each_xdp_txq(vsi, i) {
2768 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2769 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2771 mutex_unlock(&pf->avail_q_mutex);
2773 devm_kfree(dev, vsi->xdp_rings);
2778 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2779 * @vsi: VSI to remove XDP rings
2781 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2784 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2786 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2787 struct ice_pf *pf = vsi->back;
2790 /* q_vectors are freed in reset path so there's no point in detaching
2791 * rings; in case of rebuild being triggered not from reset bits
2792 * in pf->state won't be set, so additionally check first q_vector
2795 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2798 ice_for_each_q_vector(vsi, v_idx) {
2799 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2800 struct ice_tx_ring *ring;
2802 ice_for_each_tx_ring(ring, q_vector->tx)
2803 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2806 /* restore the value of last node prior to XDP setup */
2807 q_vector->tx.tx_ring = ring;
2811 mutex_lock(&pf->avail_q_mutex);
2812 ice_for_each_xdp_txq(vsi, i) {
2813 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2814 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2816 mutex_unlock(&pf->avail_q_mutex);
2818 ice_for_each_xdp_txq(vsi, i)
2819 if (vsi->xdp_rings[i]) {
2820 if (vsi->xdp_rings[i]->desc) {
2822 ice_free_tx_ring(vsi->xdp_rings[i]);
2824 kfree_rcu(vsi->xdp_rings[i]->ring_stats, rcu);
2825 vsi->xdp_rings[i]->ring_stats = NULL;
2826 kfree_rcu(vsi->xdp_rings[i], rcu);
2827 vsi->xdp_rings[i] = NULL;
2830 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2831 vsi->xdp_rings = NULL;
2833 if (static_key_enabled(&ice_xdp_locking_key))
2834 static_branch_dec(&ice_xdp_locking_key);
2836 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2839 ice_vsi_assign_bpf_prog(vsi, NULL);
2841 /* notify Tx scheduler that we destroyed XDP queues and bring
2842 * back the old number of child nodes
2844 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2845 max_txqs[i] = vsi->num_txq;
2847 /* change number of XDP Tx queues to 0 */
2848 vsi->num_xdp_txq = 0;
2850 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2855 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2856 * @vsi: VSI to schedule napi on
2858 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2862 ice_for_each_rxq(vsi, i) {
2863 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2865 if (rx_ring->xsk_pool)
2866 napi_schedule(&rx_ring->q_vector->napi);
2871 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2872 * @vsi: VSI to determine the count of XDP Tx qs
2874 * returns 0 if Tx qs count is higher than at least half of CPU count,
2877 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2879 u16 avail = ice_get_avail_txq_count(vsi->back);
2880 u16 cpus = num_possible_cpus();
2882 if (avail < cpus / 2)
2885 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2887 if (vsi->num_xdp_txq < cpus)
2888 static_branch_inc(&ice_xdp_locking_key);
2894 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
2895 * @vsi: Pointer to VSI structure
2897 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
2899 if (test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
2900 return ICE_RXBUF_1664;
2902 return ICE_RXBUF_3072;
2906 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2907 * @vsi: VSI to setup XDP for
2908 * @prog: XDP program
2909 * @extack: netlink extended ack
2912 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2913 struct netlink_ext_ack *extack)
2915 unsigned int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2916 bool if_running = netif_running(vsi->netdev);
2917 int ret = 0, xdp_ring_err = 0;
2919 if (prog && !prog->aux->xdp_has_frags) {
2920 if (frame_size > ice_max_xdp_frame_size(vsi)) {
2921 NL_SET_ERR_MSG_MOD(extack,
2922 "MTU is too large for linear frames and XDP prog does not support frags");
2927 /* need to stop netdev while setting up the program for Rx rings */
2928 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2929 ret = ice_down(vsi);
2931 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2936 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2937 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2939 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2941 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2943 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2945 xdp_features_set_redirect_target(vsi->netdev, true);
2946 /* reallocate Rx queues that are used for zero-copy */
2947 xdp_ring_err = ice_realloc_zc_buf(vsi, true);
2949 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed");
2950 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2951 xdp_features_clear_redirect_target(vsi->netdev);
2952 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2954 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2955 /* reallocate Rx queues that were used for zero-copy */
2956 xdp_ring_err = ice_realloc_zc_buf(vsi, false);
2958 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed");
2960 /* safe to call even when prog == vsi->xdp_prog as
2961 * dev_xdp_install in net/core/dev.c incremented prog's
2962 * refcount so corresponding bpf_prog_put won't cause
2965 ice_vsi_assign_bpf_prog(vsi, prog);
2972 ice_vsi_rx_napi_schedule(vsi);
2974 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2978 * ice_xdp_safe_mode - XDP handler for safe mode
2982 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2983 struct netdev_bpf *xdp)
2985 NL_SET_ERR_MSG_MOD(xdp->extack,
2986 "Please provide working DDP firmware package in order to use XDP\n"
2987 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2992 * ice_xdp - implements XDP handler
2996 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2998 struct ice_netdev_priv *np = netdev_priv(dev);
2999 struct ice_vsi *vsi = np->vsi;
3001 if (vsi->type != ICE_VSI_PF) {
3002 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
3006 switch (xdp->command) {
3007 case XDP_SETUP_PROG:
3008 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
3009 case XDP_SETUP_XSK_POOL:
3010 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
3018 * ice_ena_misc_vector - enable the non-queue interrupts
3019 * @pf: board private structure
3021 static void ice_ena_misc_vector(struct ice_pf *pf)
3023 struct ice_hw *hw = &pf->hw;
3026 /* Disable anti-spoof detection interrupt to prevent spurious event
3027 * interrupts during a function reset. Anti-spoof functionally is
3030 val = rd32(hw, GL_MDCK_TX_TDPU);
3031 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
3032 wr32(hw, GL_MDCK_TX_TDPU, val);
3034 /* clear things first */
3035 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
3036 rd32(hw, PFINT_OICR); /* read to clear */
3038 val = (PFINT_OICR_ECC_ERR_M |
3039 PFINT_OICR_MAL_DETECT_M |
3041 PFINT_OICR_PCI_EXCEPTION_M |
3043 PFINT_OICR_HMC_ERR_M |
3044 PFINT_OICR_PE_PUSH_M |
3045 PFINT_OICR_PE_CRITERR_M);
3047 wr32(hw, PFINT_OICR_ENA, val);
3049 /* SW_ITR_IDX = 0, but don't change INTENA */
3050 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
3051 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3055 * ice_misc_intr - misc interrupt handler
3056 * @irq: interrupt number
3057 * @data: pointer to a q_vector
3059 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
3061 struct ice_pf *pf = (struct ice_pf *)data;
3062 struct ice_hw *hw = &pf->hw;
3063 irqreturn_t ret = IRQ_NONE;
3067 dev = ice_pf_to_dev(pf);
3068 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3069 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3070 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3072 oicr = rd32(hw, PFINT_OICR);
3073 ena_mask = rd32(hw, PFINT_OICR_ENA);
3075 if (oicr & PFINT_OICR_SWINT_M) {
3076 ena_mask &= ~PFINT_OICR_SWINT_M;
3080 if (oicr & PFINT_OICR_MAL_DETECT_M) {
3081 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3082 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3084 if (oicr & PFINT_OICR_VFLR_M) {
3085 /* disable any further VFLR event notifications */
3086 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3087 u32 reg = rd32(hw, PFINT_OICR_ENA);
3089 reg &= ~PFINT_OICR_VFLR_M;
3090 wr32(hw, PFINT_OICR_ENA, reg);
3092 ena_mask &= ~PFINT_OICR_VFLR_M;
3093 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3097 if (oicr & PFINT_OICR_GRST_M) {
3100 /* we have a reset warning */
3101 ena_mask &= ~PFINT_OICR_GRST_M;
3102 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
3103 GLGEN_RSTAT_RESET_TYPE_S;
3105 if (reset == ICE_RESET_CORER)
3107 else if (reset == ICE_RESET_GLOBR)
3109 else if (reset == ICE_RESET_EMPR)
3112 dev_dbg(dev, "Invalid reset type %d\n", reset);
3114 /* If a reset cycle isn't already in progress, we set a bit in
3115 * pf->state so that the service task can start a reset/rebuild.
3117 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3118 if (reset == ICE_RESET_CORER)
3119 set_bit(ICE_CORER_RECV, pf->state);
3120 else if (reset == ICE_RESET_GLOBR)
3121 set_bit(ICE_GLOBR_RECV, pf->state);
3123 set_bit(ICE_EMPR_RECV, pf->state);
3125 /* There are couple of different bits at play here.
3126 * hw->reset_ongoing indicates whether the hardware is
3127 * in reset. This is set to true when a reset interrupt
3128 * is received and set back to false after the driver
3129 * has determined that the hardware is out of reset.
3131 * ICE_RESET_OICR_RECV in pf->state indicates
3132 * that a post reset rebuild is required before the
3133 * driver is operational again. This is set above.
3135 * As this is the start of the reset/rebuild cycle, set
3136 * both to indicate that.
3138 hw->reset_ongoing = true;
3142 if (oicr & PFINT_OICR_TSYN_TX_M) {
3143 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3144 if (!hw->reset_ongoing)
3145 ret = IRQ_WAKE_THREAD;
3148 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3149 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3150 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3152 /* Save EVENTs from GTSYN register */
3153 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
3154 GLTSYN_STAT_EVENT1_M |
3155 GLTSYN_STAT_EVENT2_M);
3156 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3157 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
3160 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3161 if (oicr & ICE_AUX_CRIT_ERR) {
3162 pf->oicr_err_reg |= oicr;
3163 set_bit(ICE_AUX_ERR_PENDING, pf->state);
3164 ena_mask &= ~ICE_AUX_CRIT_ERR;
3167 /* Report any remaining unexpected interrupts */
3170 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3171 /* If a critical error is pending there is no choice but to
3174 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3175 PFINT_OICR_ECC_ERR_M)) {
3176 set_bit(ICE_PFR_REQ, pf->state);
3177 ice_service_task_schedule(pf);
3183 ice_service_task_schedule(pf);
3184 ice_irq_dynamic_ena(hw, NULL, NULL);
3190 * ice_misc_intr_thread_fn - misc interrupt thread function
3191 * @irq: interrupt number
3192 * @data: pointer to a q_vector
3194 static irqreturn_t ice_misc_intr_thread_fn(int __always_unused irq, void *data)
3196 struct ice_pf *pf = data;
3198 if (ice_is_reset_in_progress(pf->state))
3201 while (!ice_ptp_process_ts(pf))
3202 usleep_range(50, 100);
3208 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3209 * @hw: pointer to HW structure
3211 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3213 /* disable Admin queue Interrupt causes */
3214 wr32(hw, PFINT_FW_CTL,
3215 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3217 /* disable Mailbox queue Interrupt causes */
3218 wr32(hw, PFINT_MBX_CTL,
3219 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3221 wr32(hw, PFINT_SB_CTL,
3222 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3224 /* disable Control queue Interrupt causes */
3225 wr32(hw, PFINT_OICR_CTL,
3226 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3232 * ice_free_irq_msix_misc - Unroll misc vector setup
3233 * @pf: board private structure
3235 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3237 struct ice_hw *hw = &pf->hw;
3239 ice_dis_ctrlq_interrupts(hw);
3241 /* disable OICR interrupt */
3242 wr32(hw, PFINT_OICR_ENA, 0);
3245 if (pf->msix_entries) {
3246 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
3247 devm_free_irq(ice_pf_to_dev(pf),
3248 pf->msix_entries[pf->oicr_idx].vector, pf);
3251 pf->num_avail_sw_msix += 1;
3252 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
3256 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3257 * @hw: pointer to HW structure
3258 * @reg_idx: HW vector index to associate the control queue interrupts with
3260 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3264 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3265 PFINT_OICR_CTL_CAUSE_ENA_M);
3266 wr32(hw, PFINT_OICR_CTL, val);
3268 /* enable Admin queue Interrupt causes */
3269 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3270 PFINT_FW_CTL_CAUSE_ENA_M);
3271 wr32(hw, PFINT_FW_CTL, val);
3273 /* enable Mailbox queue Interrupt causes */
3274 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3275 PFINT_MBX_CTL_CAUSE_ENA_M);
3276 wr32(hw, PFINT_MBX_CTL, val);
3278 /* This enables Sideband queue Interrupt causes */
3279 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3280 PFINT_SB_CTL_CAUSE_ENA_M);
3281 wr32(hw, PFINT_SB_CTL, val);
3287 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3288 * @pf: board private structure
3290 * This sets up the handler for MSIX 0, which is used to manage the
3291 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3292 * when in MSI or Legacy interrupt mode.
3294 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3296 struct device *dev = ice_pf_to_dev(pf);
3297 struct ice_hw *hw = &pf->hw;
3298 int oicr_idx, err = 0;
3300 if (!pf->int_name[0])
3301 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3302 dev_driver_string(dev), dev_name(dev));
3304 /* Do not request IRQ but do enable OICR interrupt since settings are
3305 * lost during reset. Note that this function is called only during
3306 * rebuild path and not while reset is in progress.
3308 if (ice_is_reset_in_progress(pf->state))
3311 /* reserve one vector in irq_tracker for misc interrupts */
3312 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3316 pf->num_avail_sw_msix -= 1;
3317 pf->oicr_idx = (u16)oicr_idx;
3319 err = devm_request_threaded_irq(dev,
3320 pf->msix_entries[pf->oicr_idx].vector,
3321 ice_misc_intr, ice_misc_intr_thread_fn,
3322 0, pf->int_name, pf);
3324 dev_err(dev, "devm_request_threaded_irq for %s failed: %d\n",
3326 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3327 pf->num_avail_sw_msix += 1;
3332 ice_ena_misc_vector(pf);
3334 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3335 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3336 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3339 ice_irq_dynamic_ena(hw, NULL, NULL);
3345 * ice_napi_add - register NAPI handler for the VSI
3346 * @vsi: VSI for which NAPI handler is to be registered
3348 * This function is only called in the driver's load path. Registering the NAPI
3349 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3350 * reset/rebuild, etc.)
3352 static void ice_napi_add(struct ice_vsi *vsi)
3359 ice_for_each_q_vector(vsi, v_idx)
3360 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3365 * ice_set_ops - set netdev and ethtools ops for the given netdev
3366 * @vsi: the VSI associated with the new netdev
3368 static void ice_set_ops(struct ice_vsi *vsi)
3370 struct net_device *netdev = vsi->netdev;
3371 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3373 if (ice_is_safe_mode(pf)) {
3374 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3375 ice_set_ethtool_safe_mode_ops(netdev);
3379 netdev->netdev_ops = &ice_netdev_ops;
3380 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3381 ice_set_ethtool_ops(netdev);
3383 if (vsi->type != ICE_VSI_PF)
3386 netdev->xdp_features = NETDEV_XDP_ACT_BASIC | NETDEV_XDP_ACT_REDIRECT |
3387 NETDEV_XDP_ACT_XSK_ZEROCOPY |
3388 NETDEV_XDP_ACT_RX_SG;
3392 * ice_set_netdev_features - set features for the given netdev
3393 * @netdev: netdev instance
3395 static void ice_set_netdev_features(struct net_device *netdev)
3397 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3398 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3399 netdev_features_t csumo_features;
3400 netdev_features_t vlano_features;
3401 netdev_features_t dflt_features;
3402 netdev_features_t tso_features;
3404 if (ice_is_safe_mode(pf)) {
3406 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3407 netdev->hw_features = netdev->features;
3411 dflt_features = NETIF_F_SG |
3416 csumo_features = NETIF_F_RXCSUM |
3421 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3422 NETIF_F_HW_VLAN_CTAG_TX |
3423 NETIF_F_HW_VLAN_CTAG_RX;
3425 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3427 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3429 tso_features = NETIF_F_TSO |
3433 NETIF_F_GSO_UDP_TUNNEL |
3434 NETIF_F_GSO_GRE_CSUM |
3435 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3436 NETIF_F_GSO_PARTIAL |
3437 NETIF_F_GSO_IPXIP4 |
3438 NETIF_F_GSO_IPXIP6 |
3441 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3442 NETIF_F_GSO_GRE_CSUM;
3443 /* set features that user can change */
3444 netdev->hw_features = dflt_features | csumo_features |
3445 vlano_features | tso_features;
3447 /* add support for HW_CSUM on packets with MPLS header */
3448 netdev->mpls_features = NETIF_F_HW_CSUM |
3452 /* enable features */
3453 netdev->features |= netdev->hw_features;
3455 netdev->hw_features |= NETIF_F_HW_TC;
3456 netdev->hw_features |= NETIF_F_LOOPBACK;
3458 /* encap and VLAN devices inherit default, csumo and tso features */
3459 netdev->hw_enc_features |= dflt_features | csumo_features |
3461 netdev->vlan_features |= dflt_features | csumo_features |
3464 /* advertise support but don't enable by default since only one type of
3465 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3466 * type turns on the other has to be turned off. This is enforced by the
3467 * ice_fix_features() ndo callback.
3470 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3471 NETIF_F_HW_VLAN_STAG_TX;
3473 /* Leave CRC / FCS stripping enabled by default, but allow the value to
3474 * be changed at runtime
3476 netdev->hw_features |= NETIF_F_RXFCS;
3478 netif_set_tso_max_size(netdev, ICE_MAX_TSO_SIZE);
3482 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3483 * @lut: Lookup table
3484 * @rss_table_size: Lookup table size
3485 * @rss_size: Range of queue number for hashing
3487 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3491 for (i = 0; i < rss_table_size; i++)
3492 lut[i] = i % rss_size;
3496 * ice_pf_vsi_setup - Set up a PF VSI
3497 * @pf: board private structure
3498 * @pi: pointer to the port_info instance
3500 * Returns pointer to the successfully allocated VSI software struct
3501 * on success, otherwise returns NULL on failure.
3503 static struct ice_vsi *
3504 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3506 struct ice_vsi_cfg_params params = {};
3508 params.type = ICE_VSI_PF;
3510 params.flags = ICE_VSI_FLAG_INIT;
3512 return ice_vsi_setup(pf, ¶ms);
3515 static struct ice_vsi *
3516 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3517 struct ice_channel *ch)
3519 struct ice_vsi_cfg_params params = {};
3521 params.type = ICE_VSI_CHNL;
3524 params.flags = ICE_VSI_FLAG_INIT;
3526 return ice_vsi_setup(pf, ¶ms);
3530 * ice_ctrl_vsi_setup - Set up a control VSI
3531 * @pf: board private structure
3532 * @pi: pointer to the port_info instance
3534 * Returns pointer to the successfully allocated VSI software struct
3535 * on success, otherwise returns NULL on failure.
3537 static struct ice_vsi *
3538 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3540 struct ice_vsi_cfg_params params = {};
3542 params.type = ICE_VSI_CTRL;
3544 params.flags = ICE_VSI_FLAG_INIT;
3546 return ice_vsi_setup(pf, ¶ms);
3550 * ice_lb_vsi_setup - Set up a loopback VSI
3551 * @pf: board private structure
3552 * @pi: pointer to the port_info instance
3554 * Returns pointer to the successfully allocated VSI software struct
3555 * on success, otherwise returns NULL on failure.
3558 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3560 struct ice_vsi_cfg_params params = {};
3562 params.type = ICE_VSI_LB;
3564 params.flags = ICE_VSI_FLAG_INIT;
3566 return ice_vsi_setup(pf, ¶ms);
3570 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3571 * @netdev: network interface to be adjusted
3573 * @vid: VLAN ID to be added
3575 * net_device_ops implementation for adding VLAN IDs
3578 ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3580 struct ice_netdev_priv *np = netdev_priv(netdev);
3581 struct ice_vsi_vlan_ops *vlan_ops;
3582 struct ice_vsi *vsi = np->vsi;
3583 struct ice_vlan vlan;
3586 /* VLAN 0 is added by default during load/reset */
3590 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3591 usleep_range(1000, 2000);
3593 /* Add multicast promisc rule for the VLAN ID to be added if
3594 * all-multicast is currently enabled.
3596 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3597 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3598 ICE_MCAST_VLAN_PROMISC_BITS,
3604 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3606 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3607 * packets aren't pruned by the device's internal switch on Rx
3609 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3610 ret = vlan_ops->add_vlan(vsi, &vlan);
3614 /* If all-multicast is currently enabled and this VLAN ID is only one
3615 * besides VLAN-0 we have to update look-up type of multicast promisc
3616 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3618 if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3619 ice_vsi_num_non_zero_vlans(vsi) == 1) {
3620 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3621 ICE_MCAST_PROMISC_BITS, 0);
3622 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3623 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3627 clear_bit(ICE_CFG_BUSY, vsi->state);
3633 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3634 * @netdev: network interface to be adjusted
3636 * @vid: VLAN ID to be removed
3638 * net_device_ops implementation for removing VLAN IDs
3641 ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3643 struct ice_netdev_priv *np = netdev_priv(netdev);
3644 struct ice_vsi_vlan_ops *vlan_ops;
3645 struct ice_vsi *vsi = np->vsi;
3646 struct ice_vlan vlan;
3649 /* don't allow removal of VLAN 0 */
3653 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3654 usleep_range(1000, 2000);
3656 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3657 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3659 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3661 vsi->current_netdev_flags |= IFF_ALLMULTI;
3664 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3666 /* Make sure VLAN delete is successful before updating VLAN
3669 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3670 ret = vlan_ops->del_vlan(vsi, &vlan);
3674 /* Remove multicast promisc rule for the removed VLAN ID if
3675 * all-multicast is enabled.
3677 if (vsi->current_netdev_flags & IFF_ALLMULTI)
3678 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3679 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3681 if (!ice_vsi_has_non_zero_vlans(vsi)) {
3682 /* Update look-up type of multicast promisc rule for VLAN 0
3683 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3684 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3686 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3687 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3688 ICE_MCAST_VLAN_PROMISC_BITS,
3690 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3691 ICE_MCAST_PROMISC_BITS, 0);
3696 clear_bit(ICE_CFG_BUSY, vsi->state);
3702 * ice_rep_indr_tc_block_unbind
3703 * @cb_priv: indirection block private data
3705 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3707 struct ice_indr_block_priv *indr_priv = cb_priv;
3709 list_del(&indr_priv->list);
3714 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3715 * @vsi: VSI struct which has the netdev
3717 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3719 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3721 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3722 ice_rep_indr_tc_block_unbind);
3726 * ice_tc_indir_block_register - Register TC indirect block notifications
3727 * @vsi: VSI struct which has the netdev
3729 * Returns 0 on success, negative value on failure
3731 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3733 struct ice_netdev_priv *np;
3735 if (!vsi || !vsi->netdev)
3738 np = netdev_priv(vsi->netdev);
3740 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3741 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3745 * ice_get_avail_q_count - Get count of queues in use
3746 * @pf_qmap: bitmap to get queue use count from
3747 * @lock: pointer to a mutex that protects access to pf_qmap
3748 * @size: size of the bitmap
3751 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3757 for_each_clear_bit(bit, pf_qmap, size)
3765 * ice_get_avail_txq_count - Get count of Tx queues in use
3766 * @pf: pointer to an ice_pf instance
3768 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3770 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3775 * ice_get_avail_rxq_count - Get count of Rx queues in use
3776 * @pf: pointer to an ice_pf instance
3778 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3780 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3785 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3786 * @pf: board private structure to initialize
3788 static void ice_deinit_pf(struct ice_pf *pf)
3790 ice_service_task_stop(pf);
3791 mutex_destroy(&pf->adev_mutex);
3792 mutex_destroy(&pf->sw_mutex);
3793 mutex_destroy(&pf->tc_mutex);
3794 mutex_destroy(&pf->avail_q_mutex);
3795 mutex_destroy(&pf->vfs.table_lock);
3797 if (pf->avail_txqs) {
3798 bitmap_free(pf->avail_txqs);
3799 pf->avail_txqs = NULL;
3802 if (pf->avail_rxqs) {
3803 bitmap_free(pf->avail_rxqs);
3804 pf->avail_rxqs = NULL;
3808 ptp_clock_unregister(pf->ptp.clock);
3812 * ice_set_pf_caps - set PFs capability flags
3813 * @pf: pointer to the PF instance
3815 static void ice_set_pf_caps(struct ice_pf *pf)
3817 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3819 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3820 if (func_caps->common_cap.rdma)
3821 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3822 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3823 if (func_caps->common_cap.dcb)
3824 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3825 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3826 if (func_caps->common_cap.sr_iov_1_1) {
3827 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3828 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3831 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3832 if (func_caps->common_cap.rss_table_size)
3833 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3835 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3836 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3839 /* ctrl_vsi_idx will be set to a valid value when flow director
3840 * is setup by ice_init_fdir
3842 pf->ctrl_vsi_idx = ICE_NO_VSI;
3843 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3844 /* force guaranteed filter pool for PF */
3845 ice_alloc_fd_guar_item(&pf->hw, &unused,
3846 func_caps->fd_fltr_guar);
3847 /* force shared filter pool for PF */
3848 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3849 func_caps->fd_fltr_best_effort);
3852 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3853 if (func_caps->common_cap.ieee_1588)
3854 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3856 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3857 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3861 * ice_init_pf - Initialize general software structures (struct ice_pf)
3862 * @pf: board private structure to initialize
3864 static int ice_init_pf(struct ice_pf *pf)
3866 ice_set_pf_caps(pf);
3868 mutex_init(&pf->sw_mutex);
3869 mutex_init(&pf->tc_mutex);
3870 mutex_init(&pf->adev_mutex);
3872 INIT_HLIST_HEAD(&pf->aq_wait_list);
3873 spin_lock_init(&pf->aq_wait_lock);
3874 init_waitqueue_head(&pf->aq_wait_queue);
3876 init_waitqueue_head(&pf->reset_wait_queue);
3878 /* setup service timer and periodic service task */
3879 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3880 pf->serv_tmr_period = HZ;
3881 INIT_WORK(&pf->serv_task, ice_service_task);
3882 clear_bit(ICE_SERVICE_SCHED, pf->state);
3884 mutex_init(&pf->avail_q_mutex);
3885 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3886 if (!pf->avail_txqs)
3889 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3890 if (!pf->avail_rxqs) {
3891 bitmap_free(pf->avail_txqs);
3892 pf->avail_txqs = NULL;
3896 mutex_init(&pf->vfs.table_lock);
3897 hash_init(pf->vfs.table);
3898 ice_mbx_init_snapshot(&pf->hw);
3904 * ice_reduce_msix_usage - Reduce usage of MSI-X vectors
3905 * @pf: board private structure
3906 * @v_remain: number of remaining MSI-X vectors to be distributed
3908 * Reduce the usage of MSI-X vectors when entire request cannot be fulfilled.
3909 * pf->num_lan_msix and pf->num_rdma_msix values are set based on number of
3910 * remaining vectors.
3912 static void ice_reduce_msix_usage(struct ice_pf *pf, int v_remain)
3916 if (!ice_is_rdma_ena(pf)) {
3917 pf->num_lan_msix = v_remain;
3921 /* RDMA needs at least 1 interrupt in addition to AEQ MSIX */
3922 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3924 if (v_remain < ICE_MIN_LAN_TXRX_MSIX + ICE_MIN_RDMA_MSIX) {
3925 dev_warn(ice_pf_to_dev(pf), "Not enough MSI-X vectors to support RDMA.\n");
3926 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3928 pf->num_rdma_msix = 0;
3929 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3930 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3931 (v_remain - v_rdma < v_rdma)) {
3932 /* Support minimum RDMA and give remaining vectors to LAN MSIX */
3933 pf->num_rdma_msix = ICE_MIN_RDMA_MSIX;
3934 pf->num_lan_msix = v_remain - ICE_MIN_RDMA_MSIX;
3936 /* Split remaining MSIX with RDMA after accounting for AEQ MSIX
3938 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3939 ICE_RDMA_NUM_AEQ_MSIX;
3940 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3945 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3946 * @pf: board private structure
3948 * Compute the number of MSIX vectors wanted and request from the OS. Adjust
3949 * device usage if there are not enough vectors. Return the number of vectors
3950 * reserved or negative on failure.
3952 static int ice_ena_msix_range(struct ice_pf *pf)
3954 int num_cpus, hw_num_msix, v_other, v_wanted, v_actual;
3955 struct device *dev = ice_pf_to_dev(pf);
3958 hw_num_msix = pf->hw.func_caps.common_cap.num_msix_vectors;
3959 num_cpus = num_online_cpus();
3961 /* LAN miscellaneous handler */
3962 v_other = ICE_MIN_LAN_OICR_MSIX;
3965 if (test_bit(ICE_FLAG_FD_ENA, pf->flags))
3966 v_other += ICE_FDIR_MSIX;
3969 v_other += ICE_ESWITCH_MSIX;
3974 pf->num_lan_msix = num_cpus;
3975 v_wanted += pf->num_lan_msix;
3977 /* RDMA auxiliary driver */
3978 if (ice_is_rdma_ena(pf)) {
3979 pf->num_rdma_msix = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3980 v_wanted += pf->num_rdma_msix;
3983 if (v_wanted > hw_num_msix) {
3986 dev_warn(dev, "not enough device MSI-X vectors. wanted = %d, available = %d\n",
3987 v_wanted, hw_num_msix);
3989 if (hw_num_msix < ICE_MIN_MSIX) {
3994 v_remain = hw_num_msix - v_other;
3995 if (v_remain < ICE_MIN_LAN_TXRX_MSIX) {
3996 v_other = ICE_MIN_MSIX - ICE_MIN_LAN_TXRX_MSIX;
3997 v_remain = ICE_MIN_LAN_TXRX_MSIX;
4000 ice_reduce_msix_usage(pf, v_remain);
4001 v_wanted = pf->num_lan_msix + pf->num_rdma_msix + v_other;
4003 dev_notice(dev, "Reducing request to %d MSI-X vectors for LAN traffic.\n",
4005 if (ice_is_rdma_ena(pf))
4006 dev_notice(dev, "Reducing request to %d MSI-X vectors for RDMA.\n",
4010 pf->msix_entries = devm_kcalloc(dev, v_wanted,
4011 sizeof(*pf->msix_entries), GFP_KERNEL);
4012 if (!pf->msix_entries) {
4017 for (i = 0; i < v_wanted; i++)
4018 pf->msix_entries[i].entry = i;
4020 /* actually reserve the vectors */
4021 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
4022 ICE_MIN_MSIX, v_wanted);
4024 dev_err(dev, "unable to reserve MSI-X vectors\n");
4029 if (v_actual < v_wanted) {
4030 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
4031 v_wanted, v_actual);
4033 if (v_actual < ICE_MIN_MSIX) {
4034 /* error if we can't get minimum vectors */
4035 pci_disable_msix(pf->pdev);
4039 int v_remain = v_actual - v_other;
4041 if (v_remain < ICE_MIN_LAN_TXRX_MSIX)
4042 v_remain = ICE_MIN_LAN_TXRX_MSIX;
4044 ice_reduce_msix_usage(pf, v_remain);
4046 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
4049 if (ice_is_rdma_ena(pf))
4050 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
4058 devm_kfree(dev, pf->msix_entries);
4061 pf->num_rdma_msix = 0;
4062 pf->num_lan_msix = 0;
4067 * ice_dis_msix - Disable MSI-X interrupt setup in OS
4068 * @pf: board private structure
4070 static void ice_dis_msix(struct ice_pf *pf)
4072 pci_disable_msix(pf->pdev);
4073 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
4074 pf->msix_entries = NULL;
4078 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
4079 * @pf: board private structure
4081 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
4085 if (pf->irq_tracker) {
4086 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
4087 pf->irq_tracker = NULL;
4092 * ice_init_interrupt_scheme - Determine proper interrupt scheme
4093 * @pf: board private structure to initialize
4095 static int ice_init_interrupt_scheme(struct ice_pf *pf)
4099 vectors = ice_ena_msix_range(pf);
4104 /* set up vector assignment tracking */
4105 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
4106 struct_size(pf->irq_tracker, list, vectors),
4108 if (!pf->irq_tracker) {
4113 /* populate SW interrupts pool with number of OS granted IRQs. */
4114 pf->num_avail_sw_msix = (u16)vectors;
4115 pf->irq_tracker->num_entries = (u16)vectors;
4116 pf->irq_tracker->end = pf->irq_tracker->num_entries;
4122 * ice_is_wol_supported - check if WoL is supported
4123 * @hw: pointer to hardware info
4125 * Check if WoL is supported based on the HW configuration.
4126 * Returns true if NVM supports and enables WoL for this port, false otherwise
4128 bool ice_is_wol_supported(struct ice_hw *hw)
4132 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4133 * word) indicates WoL is not supported on the corresponding PF ID.
4135 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4138 return !(BIT(hw->port_info->lport) & wol_ctrl);
4142 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4143 * @vsi: VSI being changed
4144 * @new_rx: new number of Rx queues
4145 * @new_tx: new number of Tx queues
4146 * @locked: is adev device_lock held
4148 * Only change the number of queues if new_tx, or new_rx is non-0.
4150 * Returns 0 on success.
4152 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx, bool locked)
4154 struct ice_pf *pf = vsi->back;
4155 int err = 0, timeout = 50;
4157 if (!new_rx && !new_tx)
4160 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4164 usleep_range(1000, 2000);
4168 vsi->req_txq = (u16)new_tx;
4170 vsi->req_rxq = (u16)new_rx;
4172 /* set for the next time the netdev is started */
4173 if (!netif_running(vsi->netdev)) {
4174 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4175 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4180 ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
4181 ice_pf_dcb_recfg(pf, locked);
4184 clear_bit(ICE_CFG_BUSY, pf->state);
4189 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4190 * @pf: PF to configure
4192 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4193 * VSI can still Tx/Rx VLAN tagged packets.
4195 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4197 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4198 struct ice_vsi_ctx *ctxt;
4205 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4210 ctxt->info = vsi->info;
4212 ctxt->info.valid_sections =
4213 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4214 ICE_AQ_VSI_PROP_SECURITY_VALID |
4215 ICE_AQ_VSI_PROP_SW_VALID);
4217 /* disable VLAN anti-spoof */
4218 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4219 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4221 /* disable VLAN pruning and keep all other settings */
4222 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4224 /* allow all VLANs on Tx and don't strip on Rx */
4225 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4226 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4228 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4230 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4231 status, ice_aq_str(hw->adminq.sq_last_status));
4233 vsi->info.sec_flags = ctxt->info.sec_flags;
4234 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4235 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4242 * ice_log_pkg_init - log result of DDP package load
4243 * @hw: pointer to hardware info
4244 * @state: state of package load
4246 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4248 struct ice_pf *pf = hw->back;
4251 dev = ice_pf_to_dev(pf);
4254 case ICE_DDP_PKG_SUCCESS:
4255 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4256 hw->active_pkg_name,
4257 hw->active_pkg_ver.major,
4258 hw->active_pkg_ver.minor,
4259 hw->active_pkg_ver.update,
4260 hw->active_pkg_ver.draft);
4262 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4263 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4264 hw->active_pkg_name,
4265 hw->active_pkg_ver.major,
4266 hw->active_pkg_ver.minor,
4267 hw->active_pkg_ver.update,
4268 hw->active_pkg_ver.draft);
4270 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4271 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",
4272 hw->active_pkg_name,
4273 hw->active_pkg_ver.major,
4274 hw->active_pkg_ver.minor,
4275 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4277 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4278 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",
4279 hw->active_pkg_name,
4280 hw->active_pkg_ver.major,
4281 hw->active_pkg_ver.minor,
4282 hw->active_pkg_ver.update,
4283 hw->active_pkg_ver.draft,
4290 case ICE_DDP_PKG_FW_MISMATCH:
4291 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");
4293 case ICE_DDP_PKG_INVALID_FILE:
4294 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4296 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4297 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4299 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4300 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",
4301 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4303 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4304 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");
4306 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4307 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");
4309 case ICE_DDP_PKG_LOAD_ERROR:
4310 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4311 /* poll for reset to complete */
4312 if (ice_check_reset(hw))
4313 dev_err(dev, "Error resetting device. Please reload the driver\n");
4315 case ICE_DDP_PKG_ERR:
4317 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4323 * ice_load_pkg - load/reload the DDP Package file
4324 * @firmware: firmware structure when firmware requested or NULL for reload
4325 * @pf: pointer to the PF instance
4327 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4328 * initialize HW tables.
4331 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4333 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4334 struct device *dev = ice_pf_to_dev(pf);
4335 struct ice_hw *hw = &pf->hw;
4337 /* Load DDP Package */
4338 if (firmware && !hw->pkg_copy) {
4339 state = ice_copy_and_init_pkg(hw, firmware->data,
4341 ice_log_pkg_init(hw, state);
4342 } else if (!firmware && hw->pkg_copy) {
4343 /* Reload package during rebuild after CORER/GLOBR reset */
4344 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4345 ice_log_pkg_init(hw, state);
4347 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4350 if (!ice_is_init_pkg_successful(state)) {
4352 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4356 /* Successful download package is the precondition for advanced
4357 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4359 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4363 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4364 * @pf: pointer to the PF structure
4366 * There is no error returned here because the driver should be able to handle
4367 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4368 * specifically with Tx.
4370 static void ice_verify_cacheline_size(struct ice_pf *pf)
4372 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4373 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4374 ICE_CACHE_LINE_BYTES);
4378 * ice_send_version - update firmware with driver version
4381 * Returns 0 on success, else error code
4383 static int ice_send_version(struct ice_pf *pf)
4385 struct ice_driver_ver dv;
4387 dv.major_ver = 0xff;
4388 dv.minor_ver = 0xff;
4389 dv.build_ver = 0xff;
4390 dv.subbuild_ver = 0;
4391 strscpy((char *)dv.driver_string, UTS_RELEASE,
4392 sizeof(dv.driver_string));
4393 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4397 * ice_init_fdir - Initialize flow director VSI and configuration
4398 * @pf: pointer to the PF instance
4400 * returns 0 on success, negative on error
4402 static int ice_init_fdir(struct ice_pf *pf)
4404 struct device *dev = ice_pf_to_dev(pf);
4405 struct ice_vsi *ctrl_vsi;
4408 /* Side Band Flow Director needs to have a control VSI.
4409 * Allocate it and store it in the PF.
4411 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4413 dev_dbg(dev, "could not create control VSI\n");
4417 err = ice_vsi_open_ctrl(ctrl_vsi);
4419 dev_dbg(dev, "could not open control VSI\n");
4423 mutex_init(&pf->hw.fdir_fltr_lock);
4425 err = ice_fdir_create_dflt_rules(pf);
4432 ice_fdir_release_flows(&pf->hw);
4433 ice_vsi_close(ctrl_vsi);
4435 ice_vsi_release(ctrl_vsi);
4436 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4437 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4438 pf->ctrl_vsi_idx = ICE_NO_VSI;
4443 static void ice_deinit_fdir(struct ice_pf *pf)
4445 struct ice_vsi *vsi = ice_get_ctrl_vsi(pf);
4450 ice_vsi_manage_fdir(vsi, false);
4451 ice_vsi_release(vsi);
4452 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4453 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4454 pf->ctrl_vsi_idx = ICE_NO_VSI;
4457 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4461 * ice_get_opt_fw_name - return optional firmware file name or NULL
4462 * @pf: pointer to the PF instance
4464 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4466 /* Optional firmware name same as default with additional dash
4467 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4469 struct pci_dev *pdev = pf->pdev;
4470 char *opt_fw_filename;
4473 /* Determine the name of the optional file using the DSN (two
4474 * dwords following the start of the DSN Capability).
4476 dsn = pci_get_dsn(pdev);
4480 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4481 if (!opt_fw_filename)
4484 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4485 ICE_DDP_PKG_PATH, dsn);
4487 return opt_fw_filename;
4491 * ice_request_fw - Device initialization routine
4492 * @pf: pointer to the PF instance
4494 static void ice_request_fw(struct ice_pf *pf)
4496 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4497 const struct firmware *firmware = NULL;
4498 struct device *dev = ice_pf_to_dev(pf);
4501 /* optional device-specific DDP (if present) overrides the default DDP
4502 * package file. kernel logs a debug message if the file doesn't exist,
4503 * and warning messages for other errors.
4505 if (opt_fw_filename) {
4506 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4508 kfree(opt_fw_filename);
4512 /* request for firmware was successful. Download to device */
4513 ice_load_pkg(firmware, pf);
4514 kfree(opt_fw_filename);
4515 release_firmware(firmware);
4520 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4522 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4526 /* request for firmware was successful. Download to device */
4527 ice_load_pkg(firmware, pf);
4528 release_firmware(firmware);
4532 * ice_print_wake_reason - show the wake up cause in the log
4533 * @pf: pointer to the PF struct
4535 static void ice_print_wake_reason(struct ice_pf *pf)
4537 u32 wus = pf->wakeup_reason;
4538 const char *wake_str;
4540 /* if no wake event, nothing to print */
4544 if (wus & PFPM_WUS_LNKC_M)
4545 wake_str = "Link\n";
4546 else if (wus & PFPM_WUS_MAG_M)
4547 wake_str = "Magic Packet\n";
4548 else if (wus & PFPM_WUS_MNG_M)
4549 wake_str = "Management\n";
4550 else if (wus & PFPM_WUS_FW_RST_WK_M)
4551 wake_str = "Firmware Reset\n";
4553 wake_str = "Unknown\n";
4555 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4559 * ice_register_netdev - register netdev
4560 * @vsi: pointer to the VSI struct
4562 static int ice_register_netdev(struct ice_vsi *vsi)
4566 if (!vsi || !vsi->netdev)
4569 err = register_netdev(vsi->netdev);
4573 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4574 netif_carrier_off(vsi->netdev);
4575 netif_tx_stop_all_queues(vsi->netdev);
4580 static void ice_unregister_netdev(struct ice_vsi *vsi)
4582 if (!vsi || !vsi->netdev)
4585 unregister_netdev(vsi->netdev);
4586 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4590 * ice_cfg_netdev - Allocate, configure and register a netdev
4591 * @vsi: the VSI associated with the new netdev
4593 * Returns 0 on success, negative value on failure
4595 static int ice_cfg_netdev(struct ice_vsi *vsi)
4597 struct ice_netdev_priv *np;
4598 struct net_device *netdev;
4599 u8 mac_addr[ETH_ALEN];
4601 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
4606 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4607 vsi->netdev = netdev;
4608 np = netdev_priv(netdev);
4611 ice_set_netdev_features(netdev);
4614 if (vsi->type == ICE_VSI_PF) {
4615 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
4616 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4617 eth_hw_addr_set(netdev, mac_addr);
4620 netdev->priv_flags |= IFF_UNICAST_FLT;
4622 /* Setup netdev TC information */
4623 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
4625 netdev->max_mtu = ICE_MAX_MTU;
4630 static void ice_decfg_netdev(struct ice_vsi *vsi)
4632 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4633 free_netdev(vsi->netdev);
4637 static int ice_start_eth(struct ice_vsi *vsi)
4641 err = ice_init_mac_fltr(vsi->back);
4646 err = ice_vsi_open(vsi);
4652 static void ice_stop_eth(struct ice_vsi *vsi)
4654 ice_fltr_remove_all(vsi);
4658 static int ice_init_eth(struct ice_pf *pf)
4660 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4666 /* init channel list */
4667 INIT_LIST_HEAD(&vsi->ch_list);
4669 err = ice_cfg_netdev(vsi);
4672 /* Setup DCB netlink interface */
4673 ice_dcbnl_setup(vsi);
4675 err = ice_init_mac_fltr(pf);
4677 goto err_init_mac_fltr;
4679 err = ice_devlink_create_pf_port(pf);
4681 goto err_devlink_create_pf_port;
4683 SET_NETDEV_DEVLINK_PORT(vsi->netdev, &pf->devlink_port);
4685 err = ice_register_netdev(vsi);
4687 goto err_register_netdev;
4689 err = ice_tc_indir_block_register(vsi);
4691 goto err_tc_indir_block_register;
4697 err_tc_indir_block_register:
4698 ice_unregister_netdev(vsi);
4699 err_register_netdev:
4700 ice_devlink_destroy_pf_port(pf);
4701 err_devlink_create_pf_port:
4703 ice_decfg_netdev(vsi);
4707 static void ice_deinit_eth(struct ice_pf *pf)
4709 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4715 ice_unregister_netdev(vsi);
4716 ice_devlink_destroy_pf_port(pf);
4717 ice_tc_indir_block_unregister(vsi);
4718 ice_decfg_netdev(vsi);
4721 static int ice_init_dev(struct ice_pf *pf)
4723 struct device *dev = ice_pf_to_dev(pf);
4724 struct ice_hw *hw = &pf->hw;
4727 err = ice_init_hw(hw);
4729 dev_err(dev, "ice_init_hw failed: %d\n", err);
4733 ice_init_feature_support(pf);
4737 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4738 * set in pf->state, which will cause ice_is_safe_mode to return
4741 if (ice_is_safe_mode(pf)) {
4742 /* we already got function/device capabilities but these don't
4743 * reflect what the driver needs to do in safe mode. Instead of
4744 * adding conditional logic everywhere to ignore these
4745 * device/function capabilities, override them.
4747 ice_set_safe_mode_caps(hw);
4750 err = ice_init_pf(pf);
4752 dev_err(dev, "ice_init_pf failed: %d\n", err);
4756 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4757 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4758 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4759 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4760 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4761 pf->hw.udp_tunnel_nic.tables[0].n_entries =
4762 pf->hw.tnl.valid_count[TNL_VXLAN];
4763 pf->hw.udp_tunnel_nic.tables[0].tunnel_types =
4764 UDP_TUNNEL_TYPE_VXLAN;
4766 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4767 pf->hw.udp_tunnel_nic.tables[1].n_entries =
4768 pf->hw.tnl.valid_count[TNL_GENEVE];
4769 pf->hw.udp_tunnel_nic.tables[1].tunnel_types =
4770 UDP_TUNNEL_TYPE_GENEVE;
4773 err = ice_init_interrupt_scheme(pf);
4775 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4777 goto err_init_interrupt_scheme;
4780 /* In case of MSIX we are going to setup the misc vector right here
4781 * to handle admin queue events etc. In case of legacy and MSI
4782 * the misc functionality and queue processing is combined in
4783 * the same vector and that gets setup at open.
4785 err = ice_req_irq_msix_misc(pf);
4787 dev_err(dev, "setup of misc vector failed: %d\n", err);
4788 goto err_req_irq_msix_misc;
4793 err_req_irq_msix_misc:
4794 ice_clear_interrupt_scheme(pf);
4795 err_init_interrupt_scheme:
4802 static void ice_deinit_dev(struct ice_pf *pf)
4804 ice_free_irq_msix_misc(pf);
4806 ice_deinit_hw(&pf->hw);
4808 /* Service task is already stopped, so call reset directly. */
4809 ice_reset(&pf->hw, ICE_RESET_PFR);
4810 pci_wait_for_pending_transaction(pf->pdev);
4811 ice_clear_interrupt_scheme(pf);
4814 static void ice_init_features(struct ice_pf *pf)
4816 struct device *dev = ice_pf_to_dev(pf);
4818 if (ice_is_safe_mode(pf))
4821 /* initialize DDP driven features */
4822 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4825 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4828 /* Note: Flow director init failure is non-fatal to load */
4829 if (ice_init_fdir(pf))
4830 dev_err(dev, "could not initialize flow director\n");
4832 /* Note: DCB init failure is non-fatal to load */
4833 if (ice_init_pf_dcb(pf, false)) {
4834 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4835 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4837 ice_cfg_lldp_mib_change(&pf->hw, true);
4840 if (ice_init_lag(pf))
4841 dev_warn(dev, "Failed to init link aggregation support\n");
4844 static void ice_deinit_features(struct ice_pf *pf)
4847 if (test_bit(ICE_FLAG_DCB_CAPABLE, pf->flags))
4848 ice_cfg_lldp_mib_change(&pf->hw, false);
4849 ice_deinit_fdir(pf);
4850 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4852 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4853 ice_ptp_release(pf);
4856 static void ice_init_wakeup(struct ice_pf *pf)
4858 /* Save wakeup reason register for later use */
4859 pf->wakeup_reason = rd32(&pf->hw, PFPM_WUS);
4861 /* check for a power management event */
4862 ice_print_wake_reason(pf);
4864 /* clear wake status, all bits */
4865 wr32(&pf->hw, PFPM_WUS, U32_MAX);
4867 /* Disable WoL at init, wait for user to enable */
4868 device_set_wakeup_enable(ice_pf_to_dev(pf), false);
4871 static int ice_init_link(struct ice_pf *pf)
4873 struct device *dev = ice_pf_to_dev(pf);
4876 err = ice_init_link_events(pf->hw.port_info);
4878 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4882 /* not a fatal error if this fails */
4883 err = ice_init_nvm_phy_type(pf->hw.port_info);
4885 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4887 /* not a fatal error if this fails */
4888 err = ice_update_link_info(pf->hw.port_info);
4890 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4892 ice_init_link_dflt_override(pf->hw.port_info);
4894 ice_check_link_cfg_err(pf,
4895 pf->hw.port_info->phy.link_info.link_cfg_err);
4897 /* if media available, initialize PHY settings */
4898 if (pf->hw.port_info->phy.link_info.link_info &
4899 ICE_AQ_MEDIA_AVAILABLE) {
4900 /* not a fatal error if this fails */
4901 err = ice_init_phy_user_cfg(pf->hw.port_info);
4903 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4905 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4906 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4909 ice_configure_phy(vsi);
4912 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4918 static int ice_init_pf_sw(struct ice_pf *pf)
4920 bool dvm = ice_is_dvm_ena(&pf->hw);
4921 struct ice_vsi *vsi;
4924 /* create switch struct for the switch element created by FW on boot */
4925 pf->first_sw = kzalloc(sizeof(*pf->first_sw), GFP_KERNEL);
4930 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4932 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4934 pf->first_sw->pf = pf;
4936 /* record the sw_id available for later use */
4937 pf->first_sw->sw_id = pf->hw.port_info->sw_id;
4939 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
4941 goto err_aq_set_port_params;
4943 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
4946 goto err_pf_vsi_setup;
4952 err_aq_set_port_params:
4953 kfree(pf->first_sw);
4957 static void ice_deinit_pf_sw(struct ice_pf *pf)
4959 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4964 ice_vsi_release(vsi);
4965 kfree(pf->first_sw);
4968 static int ice_alloc_vsis(struct ice_pf *pf)
4970 struct device *dev = ice_pf_to_dev(pf);
4972 pf->num_alloc_vsi = pf->hw.func_caps.guar_num_vsi;
4973 if (!pf->num_alloc_vsi)
4976 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4978 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4979 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4980 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4983 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4988 pf->vsi_stats = devm_kcalloc(dev, pf->num_alloc_vsi,
4989 sizeof(*pf->vsi_stats), GFP_KERNEL);
4990 if (!pf->vsi_stats) {
4991 devm_kfree(dev, pf->vsi);
4998 static void ice_dealloc_vsis(struct ice_pf *pf)
5000 devm_kfree(ice_pf_to_dev(pf), pf->vsi_stats);
5001 pf->vsi_stats = NULL;
5003 pf->num_alloc_vsi = 0;
5004 devm_kfree(ice_pf_to_dev(pf), pf->vsi);
5008 static int ice_init_devlink(struct ice_pf *pf)
5012 err = ice_devlink_register_params(pf);
5016 ice_devlink_init_regions(pf);
5017 ice_devlink_register(pf);
5022 static void ice_deinit_devlink(struct ice_pf *pf)
5024 ice_devlink_unregister(pf);
5025 ice_devlink_destroy_regions(pf);
5026 ice_devlink_unregister_params(pf);
5029 static int ice_init(struct ice_pf *pf)
5033 err = ice_init_dev(pf);
5037 err = ice_alloc_vsis(pf);
5039 goto err_alloc_vsis;
5041 err = ice_init_pf_sw(pf);
5043 goto err_init_pf_sw;
5045 ice_init_wakeup(pf);
5047 err = ice_init_link(pf);
5051 err = ice_send_version(pf);
5055 ice_verify_cacheline_size(pf);
5057 if (ice_is_safe_mode(pf))
5058 ice_set_safe_mode_vlan_cfg(pf);
5060 /* print PCI link speed and width */
5061 pcie_print_link_status(pf->pdev);
5063 /* ready to go, so clear down state bit */
5064 clear_bit(ICE_DOWN, pf->state);
5065 clear_bit(ICE_SERVICE_DIS, pf->state);
5067 /* since everything is good, start the service timer */
5068 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5073 ice_deinit_pf_sw(pf);
5075 ice_dealloc_vsis(pf);
5081 static void ice_deinit(struct ice_pf *pf)
5083 set_bit(ICE_SERVICE_DIS, pf->state);
5084 set_bit(ICE_DOWN, pf->state);
5086 ice_deinit_pf_sw(pf);
5087 ice_dealloc_vsis(pf);
5092 * ice_load - load pf by init hw and starting VSI
5093 * @pf: pointer to the pf instance
5095 int ice_load(struct ice_pf *pf)
5097 struct ice_vsi_cfg_params params = {};
5098 struct ice_vsi *vsi;
5101 err = ice_init_dev(pf);
5105 vsi = ice_get_main_vsi(pf);
5107 params = ice_vsi_to_params(vsi);
5108 params.flags = ICE_VSI_FLAG_INIT;
5110 err = ice_vsi_cfg(vsi, ¶ms);
5114 err = ice_start_eth(ice_get_main_vsi(pf));
5118 err = ice_init_rdma(pf);
5122 ice_init_features(pf);
5123 ice_service_task_restart(pf);
5125 clear_bit(ICE_DOWN, pf->state);
5130 ice_vsi_close(ice_get_main_vsi(pf));
5132 ice_vsi_decfg(ice_get_main_vsi(pf));
5139 * ice_unload - unload pf by stopping VSI and deinit hw
5140 * @pf: pointer to the pf instance
5142 void ice_unload(struct ice_pf *pf)
5144 ice_deinit_features(pf);
5145 ice_deinit_rdma(pf);
5146 ice_stop_eth(ice_get_main_vsi(pf));
5147 ice_vsi_decfg(ice_get_main_vsi(pf));
5152 * ice_probe - Device initialization routine
5153 * @pdev: PCI device information struct
5154 * @ent: entry in ice_pci_tbl
5156 * Returns 0 on success, negative on failure
5159 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
5161 struct device *dev = &pdev->dev;
5166 if (pdev->is_virtfn) {
5167 dev_err(dev, "can't probe a virtual function\n");
5171 /* this driver uses devres, see
5172 * Documentation/driver-api/driver-model/devres.rst
5174 err = pcim_enable_device(pdev);
5178 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
5180 dev_err(dev, "BAR0 I/O map error %d\n", err);
5184 pf = ice_allocate_pf(dev);
5188 /* initialize Auxiliary index to invalid value */
5191 /* set up for high or low DMA */
5192 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
5194 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
5198 pci_set_master(pdev);
5201 pci_set_drvdata(pdev, pf);
5202 set_bit(ICE_DOWN, pf->state);
5203 /* Disable service task until DOWN bit is cleared */
5204 set_bit(ICE_SERVICE_DIS, pf->state);
5207 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
5208 pci_save_state(pdev);
5211 hw->port_info = NULL;
5212 hw->vendor_id = pdev->vendor;
5213 hw->device_id = pdev->device;
5214 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
5215 hw->subsystem_vendor_id = pdev->subsystem_vendor;
5216 hw->subsystem_device_id = pdev->subsystem_device;
5217 hw->bus.device = PCI_SLOT(pdev->devfn);
5218 hw->bus.func = PCI_FUNC(pdev->devfn);
5219 ice_set_ctrlq_len(hw);
5221 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
5223 #ifndef CONFIG_DYNAMIC_DEBUG
5225 hw->debug_mask = debug;
5232 err = ice_init_eth(pf);
5236 err = ice_init_rdma(pf);
5240 err = ice_init_devlink(pf);
5242 goto err_init_devlink;
5244 ice_init_features(pf);
5249 ice_deinit_rdma(pf);
5255 pci_disable_device(pdev);
5260 * ice_set_wake - enable or disable Wake on LAN
5261 * @pf: pointer to the PF struct
5263 * Simple helper for WoL control
5265 static void ice_set_wake(struct ice_pf *pf)
5267 struct ice_hw *hw = &pf->hw;
5268 bool wol = pf->wol_ena;
5270 /* clear wake state, otherwise new wake events won't fire */
5271 wr32(hw, PFPM_WUS, U32_MAX);
5273 /* enable / disable APM wake up, no RMW needed */
5274 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
5276 /* set magic packet filter enabled */
5277 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
5281 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
5282 * @pf: pointer to the PF struct
5284 * Issue firmware command to enable multicast magic wake, making
5285 * sure that any locally administered address (LAA) is used for
5286 * wake, and that PF reset doesn't undo the LAA.
5288 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
5290 struct device *dev = ice_pf_to_dev(pf);
5291 struct ice_hw *hw = &pf->hw;
5292 u8 mac_addr[ETH_ALEN];
5293 struct ice_vsi *vsi;
5300 vsi = ice_get_main_vsi(pf);
5304 /* Get current MAC address in case it's an LAA */
5306 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
5308 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
5310 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
5311 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
5312 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
5314 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
5316 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
5317 status, ice_aq_str(hw->adminq.sq_last_status));
5321 * ice_remove - Device removal routine
5322 * @pdev: PCI device information struct
5324 static void ice_remove(struct pci_dev *pdev)
5326 struct ice_pf *pf = pci_get_drvdata(pdev);
5329 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
5330 if (!ice_is_reset_in_progress(pf->state))
5335 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
5336 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5340 ice_service_task_stop(pf);
5341 ice_aq_cancel_waiting_tasks(pf);
5342 set_bit(ICE_DOWN, pf->state);
5344 if (!ice_is_safe_mode(pf))
5345 ice_remove_arfs(pf);
5346 ice_deinit_features(pf);
5347 ice_deinit_devlink(pf);
5348 ice_deinit_rdma(pf);
5352 ice_vsi_release_all(pf);
5354 ice_setup_mc_magic_wake(pf);
5357 pci_disable_device(pdev);
5361 * ice_shutdown - PCI callback for shutting down device
5362 * @pdev: PCI device information struct
5364 static void ice_shutdown(struct pci_dev *pdev)
5366 struct ice_pf *pf = pci_get_drvdata(pdev);
5370 if (system_state == SYSTEM_POWER_OFF) {
5371 pci_wake_from_d3(pdev, pf->wol_ena);
5372 pci_set_power_state(pdev, PCI_D3hot);
5378 * ice_prepare_for_shutdown - prep for PCI shutdown
5379 * @pf: board private structure
5381 * Inform or close all dependent features in prep for PCI device shutdown
5383 static void ice_prepare_for_shutdown(struct ice_pf *pf)
5385 struct ice_hw *hw = &pf->hw;
5388 /* Notify VFs of impending reset */
5389 if (ice_check_sq_alive(hw, &hw->mailboxq))
5390 ice_vc_notify_reset(pf);
5392 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5394 /* disable the VSIs and their queues that are not already DOWN */
5395 ice_pf_dis_all_vsi(pf, false);
5397 ice_for_each_vsi(pf, v)
5399 pf->vsi[v]->vsi_num = 0;
5401 ice_shutdown_all_ctrlq(hw);
5405 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5406 * @pf: board private structure to reinitialize
5408 * This routine reinitialize interrupt scheme that was cleared during
5409 * power management suspend callback.
5411 * This should be called during resume routine to re-allocate the q_vectors
5412 * and reacquire interrupts.
5414 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5416 struct device *dev = ice_pf_to_dev(pf);
5419 /* Since we clear MSIX flag during suspend, we need to
5420 * set it back during resume...
5423 ret = ice_init_interrupt_scheme(pf);
5425 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5429 /* Remap vectors and rings, after successful re-init interrupts */
5430 ice_for_each_vsi(pf, v) {
5434 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5437 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5440 ret = ice_req_irq_msix_misc(pf);
5442 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5452 ice_vsi_free_q_vectors(pf->vsi[v]);
5459 * @dev: generic device information structure
5461 * Power Management callback to quiesce the device and prepare
5462 * for D3 transition.
5464 static int __maybe_unused ice_suspend(struct device *dev)
5466 struct pci_dev *pdev = to_pci_dev(dev);
5470 pf = pci_get_drvdata(pdev);
5472 if (!ice_pf_state_is_nominal(pf)) {
5473 dev_err(dev, "Device is not ready, no need to suspend it\n");
5477 /* Stop watchdog tasks until resume completion.
5478 * Even though it is most likely that the service task is
5479 * disabled if the device is suspended or down, the service task's
5480 * state is controlled by a different state bit, and we should
5481 * store and honor whatever state that bit is in at this point.
5483 disabled = ice_service_task_stop(pf);
5485 ice_unplug_aux_dev(pf);
5487 /* Already suspended?, then there is nothing to do */
5488 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5490 ice_service_task_restart(pf);
5494 if (test_bit(ICE_DOWN, pf->state) ||
5495 ice_is_reset_in_progress(pf->state)) {
5496 dev_err(dev, "can't suspend device in reset or already down\n");
5498 ice_service_task_restart(pf);
5502 ice_setup_mc_magic_wake(pf);
5504 ice_prepare_for_shutdown(pf);
5508 /* Free vectors, clear the interrupt scheme and release IRQs
5509 * for proper hibernation, especially with large number of CPUs.
5510 * Otherwise hibernation might fail when mapping all the vectors back
5513 ice_free_irq_msix_misc(pf);
5514 ice_for_each_vsi(pf, v) {
5517 ice_vsi_free_q_vectors(pf->vsi[v]);
5519 ice_clear_interrupt_scheme(pf);
5521 pci_save_state(pdev);
5522 pci_wake_from_d3(pdev, pf->wol_ena);
5523 pci_set_power_state(pdev, PCI_D3hot);
5528 * ice_resume - PM callback for waking up from D3
5529 * @dev: generic device information structure
5531 static int __maybe_unused ice_resume(struct device *dev)
5533 struct pci_dev *pdev = to_pci_dev(dev);
5534 enum ice_reset_req reset_type;
5539 pci_set_power_state(pdev, PCI_D0);
5540 pci_restore_state(pdev);
5541 pci_save_state(pdev);
5543 if (!pci_device_is_present(pdev))
5546 ret = pci_enable_device_mem(pdev);
5548 dev_err(dev, "Cannot enable device after suspend\n");
5552 pf = pci_get_drvdata(pdev);
5555 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5556 ice_print_wake_reason(pf);
5558 /* We cleared the interrupt scheme when we suspended, so we need to
5559 * restore it now to resume device functionality.
5561 ret = ice_reinit_interrupt_scheme(pf);
5563 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5565 clear_bit(ICE_DOWN, pf->state);
5566 /* Now perform PF reset and rebuild */
5567 reset_type = ICE_RESET_PFR;
5568 /* re-enable service task for reset, but allow reset to schedule it */
5569 clear_bit(ICE_SERVICE_DIS, pf->state);
5571 if (ice_schedule_reset(pf, reset_type))
5572 dev_err(dev, "Reset during resume failed.\n");
5574 clear_bit(ICE_SUSPENDED, pf->state);
5575 ice_service_task_restart(pf);
5577 /* Restart the service task */
5578 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5582 #endif /* CONFIG_PM */
5585 * ice_pci_err_detected - warning that PCI error has been detected
5586 * @pdev: PCI device information struct
5587 * @err: the type of PCI error
5589 * Called to warn that something happened on the PCI bus and the error handling
5590 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5592 static pci_ers_result_t
5593 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5595 struct ice_pf *pf = pci_get_drvdata(pdev);
5598 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5600 return PCI_ERS_RESULT_DISCONNECT;
5603 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5604 ice_service_task_stop(pf);
5606 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5607 set_bit(ICE_PFR_REQ, pf->state);
5608 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5612 return PCI_ERS_RESULT_NEED_RESET;
5616 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5617 * @pdev: PCI device information struct
5619 * Called to determine if the driver can recover from the PCI slot reset by
5620 * using a register read to determine if the device is recoverable.
5622 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5624 struct ice_pf *pf = pci_get_drvdata(pdev);
5625 pci_ers_result_t result;
5629 err = pci_enable_device_mem(pdev);
5631 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5633 result = PCI_ERS_RESULT_DISCONNECT;
5635 pci_set_master(pdev);
5636 pci_restore_state(pdev);
5637 pci_save_state(pdev);
5638 pci_wake_from_d3(pdev, false);
5640 /* Check for life */
5641 reg = rd32(&pf->hw, GLGEN_RTRIG);
5643 result = PCI_ERS_RESULT_RECOVERED;
5645 result = PCI_ERS_RESULT_DISCONNECT;
5652 * ice_pci_err_resume - restart operations after PCI error recovery
5653 * @pdev: PCI device information struct
5655 * Called to allow the driver to bring things back up after PCI error and/or
5656 * reset recovery have finished
5658 static void ice_pci_err_resume(struct pci_dev *pdev)
5660 struct ice_pf *pf = pci_get_drvdata(pdev);
5663 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5668 if (test_bit(ICE_SUSPENDED, pf->state)) {
5669 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5674 ice_restore_all_vfs_msi_state(pdev);
5676 ice_do_reset(pf, ICE_RESET_PFR);
5677 ice_service_task_restart(pf);
5678 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5682 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5683 * @pdev: PCI device information struct
5685 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5687 struct ice_pf *pf = pci_get_drvdata(pdev);
5689 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5690 ice_service_task_stop(pf);
5692 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5693 set_bit(ICE_PFR_REQ, pf->state);
5694 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5700 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5701 * @pdev: PCI device information struct
5703 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5705 ice_pci_err_resume(pdev);
5708 /* ice_pci_tbl - PCI Device ID Table
5710 * Wildcard entries (PCI_ANY_ID) should come last
5711 * Last entry must be all 0s
5713 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5714 * Class, Class Mask, private data (not used) }
5716 static const struct pci_device_id ice_pci_tbl[] = {
5717 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5718 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5719 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5720 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5721 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5722 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5723 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5724 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5725 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5726 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5727 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5728 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5729 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5730 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5731 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5732 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5733 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5734 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5735 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5736 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5737 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5738 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5739 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5740 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5741 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5742 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT), 0 },
5743 /* required last entry */
5746 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5748 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5750 static const struct pci_error_handlers ice_pci_err_handler = {
5751 .error_detected = ice_pci_err_detected,
5752 .slot_reset = ice_pci_err_slot_reset,
5753 .reset_prepare = ice_pci_err_reset_prepare,
5754 .reset_done = ice_pci_err_reset_done,
5755 .resume = ice_pci_err_resume
5758 static struct pci_driver ice_driver = {
5759 .name = KBUILD_MODNAME,
5760 .id_table = ice_pci_tbl,
5762 .remove = ice_remove,
5764 .driver.pm = &ice_pm_ops,
5765 #endif /* CONFIG_PM */
5766 .shutdown = ice_shutdown,
5767 .sriov_configure = ice_sriov_configure,
5768 .err_handler = &ice_pci_err_handler
5772 * ice_module_init - Driver registration routine
5774 * ice_module_init is the first routine called when the driver is
5775 * loaded. All it does is register with the PCI subsystem.
5777 static int __init ice_module_init(void)
5781 pr_info("%s\n", ice_driver_string);
5782 pr_info("%s\n", ice_copyright);
5784 ice_wq = alloc_workqueue("%s", 0, 0, KBUILD_MODNAME);
5786 pr_err("Failed to create workqueue\n");
5790 status = pci_register_driver(&ice_driver);
5792 pr_err("failed to register PCI driver, err %d\n", status);
5793 destroy_workqueue(ice_wq);
5798 module_init(ice_module_init);
5801 * ice_module_exit - Driver exit cleanup routine
5803 * ice_module_exit is called just before the driver is removed
5806 static void __exit ice_module_exit(void)
5808 pci_unregister_driver(&ice_driver);
5809 destroy_workqueue(ice_wq);
5810 pr_info("module unloaded\n");
5812 module_exit(ice_module_exit);
5815 * ice_set_mac_address - NDO callback to set MAC address
5816 * @netdev: network interface device structure
5817 * @pi: pointer to an address structure
5819 * Returns 0 on success, negative on failure
5821 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5823 struct ice_netdev_priv *np = netdev_priv(netdev);
5824 struct ice_vsi *vsi = np->vsi;
5825 struct ice_pf *pf = vsi->back;
5826 struct ice_hw *hw = &pf->hw;
5827 struct sockaddr *addr = pi;
5828 u8 old_mac[ETH_ALEN];
5833 mac = (u8 *)addr->sa_data;
5835 if (!is_valid_ether_addr(mac))
5836 return -EADDRNOTAVAIL;
5838 if (ether_addr_equal(netdev->dev_addr, mac)) {
5839 netdev_dbg(netdev, "already using mac %pM\n", mac);
5843 if (test_bit(ICE_DOWN, pf->state) ||
5844 ice_is_reset_in_progress(pf->state)) {
5845 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5850 if (ice_chnl_dmac_fltr_cnt(pf)) {
5851 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5856 netif_addr_lock_bh(netdev);
5857 ether_addr_copy(old_mac, netdev->dev_addr);
5858 /* change the netdev's MAC address */
5859 eth_hw_addr_set(netdev, mac);
5860 netif_addr_unlock_bh(netdev);
5862 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5863 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5864 if (err && err != -ENOENT) {
5865 err = -EADDRNOTAVAIL;
5866 goto err_update_filters;
5869 /* Add filter for new MAC. If filter exists, return success */
5870 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5871 if (err == -EEXIST) {
5872 /* Although this MAC filter is already present in hardware it's
5873 * possible in some cases (e.g. bonding) that dev_addr was
5874 * modified outside of the driver and needs to be restored back
5877 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5881 /* error if the new filter addition failed */
5882 err = -EADDRNOTAVAIL;
5887 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5889 netif_addr_lock_bh(netdev);
5890 eth_hw_addr_set(netdev, old_mac);
5891 netif_addr_unlock_bh(netdev);
5895 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5898 /* write new MAC address to the firmware */
5899 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5900 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5902 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5909 * ice_set_rx_mode - NDO callback to set the netdev filters
5910 * @netdev: network interface device structure
5912 static void ice_set_rx_mode(struct net_device *netdev)
5914 struct ice_netdev_priv *np = netdev_priv(netdev);
5915 struct ice_vsi *vsi = np->vsi;
5920 /* Set the flags to synchronize filters
5921 * ndo_set_rx_mode may be triggered even without a change in netdev
5924 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5925 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5926 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5928 /* schedule our worker thread which will take care of
5929 * applying the new filter changes
5931 ice_service_task_schedule(vsi->back);
5935 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5936 * @netdev: network interface device structure
5937 * @queue_index: Queue ID
5938 * @maxrate: maximum bandwidth in Mbps
5941 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5943 struct ice_netdev_priv *np = netdev_priv(netdev);
5944 struct ice_vsi *vsi = np->vsi;
5949 /* Validate maxrate requested is within permitted range */
5950 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5951 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5952 maxrate, queue_index);
5956 q_handle = vsi->tx_rings[queue_index]->q_handle;
5957 tc = ice_dcb_get_tc(vsi, queue_index);
5959 /* Set BW back to default, when user set maxrate to 0 */
5961 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5962 q_handle, ICE_MAX_BW);
5964 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5965 q_handle, ICE_MAX_BW, maxrate * 1000);
5967 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5974 * ice_fdb_add - add an entry to the hardware database
5975 * @ndm: the input from the stack
5976 * @tb: pointer to array of nladdr (unused)
5977 * @dev: the net device pointer
5978 * @addr: the MAC address entry being added
5980 * @flags: instructions from stack about fdb operation
5981 * @extack: netlink extended ack
5984 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5985 struct net_device *dev, const unsigned char *addr, u16 vid,
5986 u16 flags, struct netlink_ext_ack __always_unused *extack)
5991 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5994 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5995 netdev_err(dev, "FDB only supports static addresses\n");
5999 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
6000 err = dev_uc_add_excl(dev, addr);
6001 else if (is_multicast_ether_addr(addr))
6002 err = dev_mc_add_excl(dev, addr);
6006 /* Only return duplicate errors if NLM_F_EXCL is set */
6007 if (err == -EEXIST && !(flags & NLM_F_EXCL))
6014 * ice_fdb_del - delete an entry from the hardware database
6015 * @ndm: the input from the stack
6016 * @tb: pointer to array of nladdr (unused)
6017 * @dev: the net device pointer
6018 * @addr: the MAC address entry being added
6020 * @extack: netlink extended ack
6023 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
6024 struct net_device *dev, const unsigned char *addr,
6025 __always_unused u16 vid, struct netlink_ext_ack *extack)
6029 if (ndm->ndm_state & NUD_PERMANENT) {
6030 netdev_err(dev, "FDB only supports static addresses\n");
6034 if (is_unicast_ether_addr(addr))
6035 err = dev_uc_del(dev, addr);
6036 else if (is_multicast_ether_addr(addr))
6037 err = dev_mc_del(dev, addr);
6044 #define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6045 NETIF_F_HW_VLAN_CTAG_TX | \
6046 NETIF_F_HW_VLAN_STAG_RX | \
6047 NETIF_F_HW_VLAN_STAG_TX)
6049 #define NETIF_VLAN_STRIPPING_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
6050 NETIF_F_HW_VLAN_STAG_RX)
6052 #define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
6053 NETIF_F_HW_VLAN_STAG_FILTER)
6056 * ice_fix_features - fix the netdev features flags based on device limitations
6057 * @netdev: ptr to the netdev that flags are being fixed on
6058 * @features: features that need to be checked and possibly fixed
6060 * Make sure any fixups are made to features in this callback. This enables the
6061 * driver to not have to check unsupported configurations throughout the driver
6062 * because that's the responsiblity of this callback.
6064 * Single VLAN Mode (SVM) Supported Features:
6065 * NETIF_F_HW_VLAN_CTAG_FILTER
6066 * NETIF_F_HW_VLAN_CTAG_RX
6067 * NETIF_F_HW_VLAN_CTAG_TX
6069 * Double VLAN Mode (DVM) Supported Features:
6070 * NETIF_F_HW_VLAN_CTAG_FILTER
6071 * NETIF_F_HW_VLAN_CTAG_RX
6072 * NETIF_F_HW_VLAN_CTAG_TX
6074 * NETIF_F_HW_VLAN_STAG_FILTER
6075 * NETIF_HW_VLAN_STAG_RX
6076 * NETIF_HW_VLAN_STAG_TX
6078 * Features that need fixing:
6079 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
6080 * These are mutually exlusive as the VSI context cannot support multiple
6081 * VLAN ethertypes simultaneously for stripping and/or insertion. If this
6082 * is not done, then default to clearing the requested STAG offload
6085 * All supported filtering has to be enabled or disabled together. For
6086 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled
6087 * together. If this is not done, then default to VLAN filtering disabled.
6088 * These are mutually exclusive as there is currently no way to
6089 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN
6092 static netdev_features_t
6093 ice_fix_features(struct net_device *netdev, netdev_features_t features)
6095 struct ice_netdev_priv *np = netdev_priv(netdev);
6096 netdev_features_t req_vlan_fltr, cur_vlan_fltr;
6097 bool cur_ctag, cur_stag, req_ctag, req_stag;
6099 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
6100 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6101 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6103 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
6104 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
6105 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
6107 if (req_vlan_fltr != cur_vlan_fltr) {
6108 if (ice_is_dvm_ena(&np->vsi->back->hw)) {
6109 if (req_ctag && req_stag) {
6110 features |= NETIF_VLAN_FILTERING_FEATURES;
6111 } else if (!req_ctag && !req_stag) {
6112 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6113 } else if ((!cur_ctag && req_ctag && !cur_stag) ||
6114 (!cur_stag && req_stag && !cur_ctag)) {
6115 features |= NETIF_VLAN_FILTERING_FEATURES;
6116 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");
6117 } else if ((cur_ctag && !req_ctag && cur_stag) ||
6118 (cur_stag && !req_stag && cur_ctag)) {
6119 features &= ~NETIF_VLAN_FILTERING_FEATURES;
6120 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");
6123 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
6124 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
6126 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
6127 features |= NETIF_F_HW_VLAN_CTAG_FILTER;
6131 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
6132 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
6133 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");
6134 features &= ~(NETIF_F_HW_VLAN_STAG_RX |
6135 NETIF_F_HW_VLAN_STAG_TX);
6138 if (!(netdev->features & NETIF_F_RXFCS) &&
6139 (features & NETIF_F_RXFCS) &&
6140 (features & NETIF_VLAN_STRIPPING_FEATURES) &&
6141 !ice_vsi_has_non_zero_vlans(np->vsi)) {
6142 netdev_warn(netdev, "Disabling VLAN stripping as FCS/CRC stripping is also disabled and there is no VLAN configured\n");
6143 features &= ~NETIF_VLAN_STRIPPING_FEATURES;
6150 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
6152 * @features: features used to determine VLAN offload settings
6154 * First, determine the vlan_ethertype based on the VLAN offload bits in
6155 * features. Then determine if stripping and insertion should be enabled or
6156 * disabled. Finally enable or disable VLAN stripping and insertion.
6159 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
6161 bool enable_stripping = true, enable_insertion = true;
6162 struct ice_vsi_vlan_ops *vlan_ops;
6163 int strip_err = 0, insert_err = 0;
6164 u16 vlan_ethertype = 0;
6166 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6168 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
6169 vlan_ethertype = ETH_P_8021AD;
6170 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
6171 vlan_ethertype = ETH_P_8021Q;
6173 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
6174 enable_stripping = false;
6175 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
6176 enable_insertion = false;
6178 if (enable_stripping)
6179 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
6181 strip_err = vlan_ops->dis_stripping(vsi);
6183 if (enable_insertion)
6184 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
6186 insert_err = vlan_ops->dis_insertion(vsi);
6188 if (strip_err || insert_err)
6195 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
6197 * @features: features used to determine VLAN filtering settings
6199 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
6203 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
6205 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
6208 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
6209 * if either bit is set
6212 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
6213 err = vlan_ops->ena_rx_filtering(vsi);
6215 err = vlan_ops->dis_rx_filtering(vsi);
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
6225 * Only update VLAN settings if the requested_vlan_features are different than
6226 * the current_vlan_features.
6229 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
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;
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");
6246 err = ice_set_vlan_offload_features(vsi, features);
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);
6264 * ice_set_loopback - turn on/off loopback mode on underlying PF
6266 * @ena: flag to indicate the on/off setting
6268 static int ice_set_loopback(struct ice_vsi *vsi, bool ena)
6270 bool if_running = netif_running(vsi->netdev);
6273 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6274 ret = ice_down(vsi);
6276 netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
6280 ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
6282 netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
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
6295 ice_set_features(struct net_device *netdev, netdev_features_t features)
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;
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");
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");
6317 /* Multiple features can be changed in one call so keep features in
6318 * separate if/else statements to guarantee each feature is checked
6320 if (changed & NETIF_F_RXHASH)
6321 ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH));
6323 ret = ice_set_vlan_features(netdev, features);
6327 /* Turn on receive of FCS aka CRC, and after setting this
6328 * flag the packet data will have the 4 byte CRC appended
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");
6338 ice_vsi_cfg_crc_strip(vsi, !!(features & NETIF_F_RXFCS));
6339 ret = ice_down_up(vsi);
6344 if (changed & NETIF_F_NTUPLE) {
6345 bool ena = !!(features & NETIF_F_NTUPLE);
6347 ice_vsi_manage_fdir(vsi, ena);
6348 ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi);
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");
6357 if (changed & NETIF_F_HW_TC) {
6358 bool ena = !!(features & NETIF_F_HW_TC);
6360 ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) :
6361 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6364 if (changed & NETIF_F_LOOPBACK)
6365 ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
6371 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6372 * @vsi: VSI to setup VLAN properties for
6374 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6378 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6382 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6386 return ice_vsi_add_vlan_zero(vsi);
6390 * ice_vsi_cfg_lan - Setup the VSI lan related config
6391 * @vsi: the VSI being configured
6393 * Return 0 on success and negative value on error
6395 int ice_vsi_cfg_lan(struct ice_vsi *vsi)
6399 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6400 ice_set_rx_mode(vsi->netdev);
6402 err = ice_vsi_vlan_setup(vsi);
6406 ice_vsi_cfg_dcb_rings(vsi);
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);
6412 err = ice_vsi_cfg_rxqs(vsi);
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.
6426 /* the throttle rate for interrupts, basically worst case delay before
6427 * an initial interrupt fires, value is stored in microseconds.
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
6436 static 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 */
6444 /* The transmit profile, which has the same sorts of values
6445 * as the previous struct
6447 static 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 */
6455 static void ice_tx_dim_work(struct work_struct *work)
6457 struct ice_ring_container *rc;
6461 dim = container_of(work, struct dim, work);
6462 rc = (struct ice_ring_container *)dim->priv;
6464 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6466 /* look up the values in our local table */
6467 itr = tx_profile[dim->profile_ix].itr;
6469 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6470 ice_write_itr(rc, itr);
6472 dim->state = DIM_START_MEASURE;
6475 static void ice_rx_dim_work(struct work_struct *work)
6477 struct ice_ring_container *rc;
6481 dim = container_of(work, struct dim, work);
6482 rc = (struct ice_ring_container *)dim->priv;
6484 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6486 /* look up the values in our local table */
6487 itr = rx_profile[dim->profile_ix].itr;
6489 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6490 ice_write_itr(rc, itr);
6492 dim->state = DIM_START_MEASURE;
6495 #define ICE_DIM_DEFAULT_PROFILE_IX 1
6498 * ice_init_moderation - set up interrupt moderation
6499 * @q_vector: the vector containing rings to be configured
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
6507 static void ice_init_moderation(struct ice_q_vector *q_vector)
6509 struct ice_ring_container *rc;
6510 bool tx_dynamic, rx_dynamic;
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;
6517 tx_dynamic = ITR_IS_DYNAMIC(rc);
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);
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;
6528 rx_dynamic = ITR_IS_DYNAMIC(rc);
6530 /* set the initial RX ITR to match the above */
6531 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6534 ice_set_q_vector_intrl(q_vector);
6538 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6539 * @vsi: the VSI being configured
6541 static void ice_napi_enable_all(struct ice_vsi *vsi)
6548 ice_for_each_q_vector(vsi, q_idx) {
6549 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6551 ice_init_moderation(q_vector);
6553 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6554 napi_enable(&q_vector->napi);
6559 * ice_up_complete - Finish the last steps of bringing up a connection
6560 * @vsi: The VSI being configured
6562 * Return 0 on success and negative value on error
6564 static int ice_up_complete(struct ice_vsi *vsi)
6566 struct ice_pf *pf = vsi->back;
6569 ice_vsi_cfg_msix(vsi);
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
6575 err = ice_vsi_start_all_rx_rings(vsi);
6579 clear_bit(ICE_VSI_DOWN, vsi->state);
6580 ice_napi_enable_all(vsi);
6581 ice_vsi_ena_irq(vsi);
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);
6592 /* Perform an initial read of the statistics registers now to
6593 * set the baseline so counters are ready when interface is up
6595 ice_update_eth_stats(vsi);
6597 if (vsi->type == ICE_VSI_PF)
6598 ice_service_task_schedule(pf);
6604 * ice_up - Bring the connection back up after being down
6605 * @vsi: VSI being configured
6607 int ice_up(struct ice_vsi *vsi)
6611 err = ice_vsi_cfg_lan(vsi);
6613 err = ice_up_complete(vsi);
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
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.
6629 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6630 struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6635 start = u64_stats_fetch_begin(syncp);
6637 *bytes = stats.bytes;
6638 } while (u64_stats_fetch_retry(syncp, start));
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
6649 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6650 struct rtnl_link_stats64 *vsi_stats,
6651 struct ice_tx_ring **rings, u16 count)
6655 for (i = 0; i < count; i++) {
6656 struct ice_tx_ring *ring;
6657 u64 pkts = 0, bytes = 0;
6659 ring = READ_ONCE(rings[i]);
6660 if (!ring || !ring->ring_stats)
6662 ice_fetch_u64_stats_per_ring(&ring->ring_stats->syncp,
6663 ring->ring_stats->stats, &pkts,
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;
6674 * ice_update_vsi_ring_stats - Update VSI stats counters
6675 * @vsi: the VSI to be updated
6677 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6679 struct rtnl_link_stats64 *net_stats, *stats_prev;
6680 struct rtnl_link_stats64 *vsi_stats;
6684 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6688 /* reset non-netdev (extended) stats */
6689 vsi->tx_restart = 0;
6691 vsi->tx_linearize = 0;
6692 vsi->rx_buf_failed = 0;
6693 vsi->rx_page_failed = 0;
6697 /* update Tx rings counters */
6698 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6701 /* update Rx rings counters */
6702 ice_for_each_rxq(vsi, i) {
6703 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6704 struct ice_ring_stats *ring_stats;
6706 ring_stats = ring->ring_stats;
6707 ice_fetch_u64_stats_per_ring(&ring_stats->syncp,
6708 ring_stats->stats, &pkts,
6710 vsi_stats->rx_packets += pkts;
6711 vsi_stats->rx_bytes += bytes;
6712 vsi->rx_buf_failed += ring_stats->rx_stats.alloc_buf_failed;
6713 vsi->rx_page_failed += ring_stats->rx_stats.alloc_page_failed;
6716 /* update XDP Tx rings counters */
6717 if (ice_is_xdp_ena_vsi(vsi))
6718 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6723 net_stats = &vsi->net_stats;
6724 stats_prev = &vsi->net_stats_prev;
6726 /* clear prev counters after reset */
6727 if (vsi_stats->tx_packets < stats_prev->tx_packets ||
6728 vsi_stats->rx_packets < stats_prev->rx_packets) {
6729 stats_prev->tx_packets = 0;
6730 stats_prev->tx_bytes = 0;
6731 stats_prev->rx_packets = 0;
6732 stats_prev->rx_bytes = 0;
6735 /* update netdev counters */
6736 net_stats->tx_packets += vsi_stats->tx_packets - stats_prev->tx_packets;
6737 net_stats->tx_bytes += vsi_stats->tx_bytes - stats_prev->tx_bytes;
6738 net_stats->rx_packets += vsi_stats->rx_packets - stats_prev->rx_packets;
6739 net_stats->rx_bytes += vsi_stats->rx_bytes - stats_prev->rx_bytes;
6741 stats_prev->tx_packets = vsi_stats->tx_packets;
6742 stats_prev->tx_bytes = vsi_stats->tx_bytes;
6743 stats_prev->rx_packets = vsi_stats->rx_packets;
6744 stats_prev->rx_bytes = vsi_stats->rx_bytes;
6750 * ice_update_vsi_stats - Update VSI stats counters
6751 * @vsi: the VSI to be updated
6753 void ice_update_vsi_stats(struct ice_vsi *vsi)
6755 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6756 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6757 struct ice_pf *pf = vsi->back;
6759 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6760 test_bit(ICE_CFG_BUSY, pf->state))
6763 /* get stats as recorded by Tx/Rx rings */
6764 ice_update_vsi_ring_stats(vsi);
6766 /* get VSI stats as recorded by the hardware */
6767 ice_update_eth_stats(vsi);
6769 cur_ns->tx_errors = cur_es->tx_errors;
6770 cur_ns->rx_dropped = cur_es->rx_discards;
6771 cur_ns->tx_dropped = cur_es->tx_discards;
6772 cur_ns->multicast = cur_es->rx_multicast;
6774 /* update some more netdev stats if this is main VSI */
6775 if (vsi->type == ICE_VSI_PF) {
6776 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6777 cur_ns->rx_errors = pf->stats.crc_errors +
6778 pf->stats.illegal_bytes +
6779 pf->stats.rx_len_errors +
6780 pf->stats.rx_undersize +
6781 pf->hw_csum_rx_error +
6782 pf->stats.rx_jabber +
6783 pf->stats.rx_fragments +
6784 pf->stats.rx_oversize;
6785 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6786 /* record drops from the port level */
6787 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6792 * ice_update_pf_stats - Update PF port stats counters
6793 * @pf: PF whose stats needs to be updated
6795 void ice_update_pf_stats(struct ice_pf *pf)
6797 struct ice_hw_port_stats *prev_ps, *cur_ps;
6798 struct ice_hw *hw = &pf->hw;
6802 port = hw->port_info->lport;
6803 prev_ps = &pf->stats_prev;
6804 cur_ps = &pf->stats;
6806 if (ice_is_reset_in_progress(pf->state))
6807 pf->stat_prev_loaded = false;
6809 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6810 &prev_ps->eth.rx_bytes,
6811 &cur_ps->eth.rx_bytes);
6813 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6814 &prev_ps->eth.rx_unicast,
6815 &cur_ps->eth.rx_unicast);
6817 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6818 &prev_ps->eth.rx_multicast,
6819 &cur_ps->eth.rx_multicast);
6821 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6822 &prev_ps->eth.rx_broadcast,
6823 &cur_ps->eth.rx_broadcast);
6825 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6826 &prev_ps->eth.rx_discards,
6827 &cur_ps->eth.rx_discards);
6829 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6830 &prev_ps->eth.tx_bytes,
6831 &cur_ps->eth.tx_bytes);
6833 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6834 &prev_ps->eth.tx_unicast,
6835 &cur_ps->eth.tx_unicast);
6837 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6838 &prev_ps->eth.tx_multicast,
6839 &cur_ps->eth.tx_multicast);
6841 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6842 &prev_ps->eth.tx_broadcast,
6843 &cur_ps->eth.tx_broadcast);
6845 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6846 &prev_ps->tx_dropped_link_down,
6847 &cur_ps->tx_dropped_link_down);
6849 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6850 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6852 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6853 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6855 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6856 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6858 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6859 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6861 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6862 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6864 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6865 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6867 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6868 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6870 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6871 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6873 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6874 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6876 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6877 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6879 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6880 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6882 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6883 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6885 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6886 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6888 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6889 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6891 fd_ctr_base = hw->fd_ctr_base;
6893 ice_stat_update40(hw,
6894 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6895 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6896 &cur_ps->fd_sb_match);
6897 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6898 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6900 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6901 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6903 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6904 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6906 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6907 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6909 ice_update_dcb_stats(pf);
6911 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6912 &prev_ps->crc_errors, &cur_ps->crc_errors);
6914 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6915 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6917 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6918 &prev_ps->mac_local_faults,
6919 &cur_ps->mac_local_faults);
6921 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6922 &prev_ps->mac_remote_faults,
6923 &cur_ps->mac_remote_faults);
6925 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6926 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6928 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6929 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6931 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6932 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6934 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6935 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6937 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6938 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6940 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6942 pf->stat_prev_loaded = true;
6946 * ice_get_stats64 - get statistics for network device structure
6947 * @netdev: network interface device structure
6948 * @stats: main device statistics structure
6951 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6953 struct ice_netdev_priv *np = netdev_priv(netdev);
6954 struct rtnl_link_stats64 *vsi_stats;
6955 struct ice_vsi *vsi = np->vsi;
6957 vsi_stats = &vsi->net_stats;
6959 if (!vsi->num_txq || !vsi->num_rxq)
6962 /* netdev packet/byte stats come from ring counter. These are obtained
6963 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6964 * But, only call the update routine and read the registers if VSI is
6967 if (!test_bit(ICE_VSI_DOWN, vsi->state))
6968 ice_update_vsi_ring_stats(vsi);
6969 stats->tx_packets = vsi_stats->tx_packets;
6970 stats->tx_bytes = vsi_stats->tx_bytes;
6971 stats->rx_packets = vsi_stats->rx_packets;
6972 stats->rx_bytes = vsi_stats->rx_bytes;
6974 /* The rest of the stats can be read from the hardware but instead we
6975 * just return values that the watchdog task has already obtained from
6978 stats->multicast = vsi_stats->multicast;
6979 stats->tx_errors = vsi_stats->tx_errors;
6980 stats->tx_dropped = vsi_stats->tx_dropped;
6981 stats->rx_errors = vsi_stats->rx_errors;
6982 stats->rx_dropped = vsi_stats->rx_dropped;
6983 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6984 stats->rx_length_errors = vsi_stats->rx_length_errors;
6988 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6989 * @vsi: VSI having NAPI disabled
6991 static void ice_napi_disable_all(struct ice_vsi *vsi)
6998 ice_for_each_q_vector(vsi, q_idx) {
6999 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
7001 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
7002 napi_disable(&q_vector->napi);
7004 cancel_work_sync(&q_vector->tx.dim.work);
7005 cancel_work_sync(&q_vector->rx.dim.work);
7010 * ice_down - Shutdown the connection
7011 * @vsi: The VSI being stopped
7013 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
7015 int ice_down(struct ice_vsi *vsi)
7017 int i, tx_err, rx_err, vlan_err = 0;
7019 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
7021 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
7022 vlan_err = ice_vsi_del_vlan_zero(vsi);
7023 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
7024 netif_carrier_off(vsi->netdev);
7025 netif_tx_disable(vsi->netdev);
7026 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
7027 ice_eswitch_stop_all_tx_queues(vsi->back);
7030 ice_vsi_dis_irq(vsi);
7032 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
7034 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
7035 vsi->vsi_num, tx_err);
7036 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
7037 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
7039 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
7040 vsi->vsi_num, tx_err);
7043 rx_err = ice_vsi_stop_all_rx_rings(vsi);
7045 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
7046 vsi->vsi_num, rx_err);
7048 ice_napi_disable_all(vsi);
7050 ice_for_each_txq(vsi, i)
7051 ice_clean_tx_ring(vsi->tx_rings[i]);
7053 if (ice_is_xdp_ena_vsi(vsi))
7054 ice_for_each_xdp_txq(vsi, i)
7055 ice_clean_tx_ring(vsi->xdp_rings[i]);
7057 ice_for_each_rxq(vsi, i)
7058 ice_clean_rx_ring(vsi->rx_rings[i]);
7060 if (tx_err || rx_err || vlan_err) {
7061 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
7062 vsi->vsi_num, vsi->vsw->sw_id);
7070 * ice_down_up - shutdown the VSI connection and bring it up
7071 * @vsi: the VSI to be reconnected
7073 int ice_down_up(struct ice_vsi *vsi)
7077 /* if DOWN already set, nothing to do */
7078 if (test_and_set_bit(ICE_VSI_DOWN, vsi->state))
7081 ret = ice_down(vsi);
7087 netdev_err(vsi->netdev, "reallocating resources failed during netdev features change, may need to reload driver\n");
7095 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
7096 * @vsi: VSI having resources allocated
7098 * Return 0 on success, negative on failure
7100 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
7104 if (!vsi->num_txq) {
7105 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
7110 ice_for_each_txq(vsi, i) {
7111 struct ice_tx_ring *ring = vsi->tx_rings[i];
7117 ring->netdev = vsi->netdev;
7118 err = ice_setup_tx_ring(ring);
7127 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
7128 * @vsi: VSI having resources allocated
7130 * Return 0 on success, negative on failure
7132 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
7136 if (!vsi->num_rxq) {
7137 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
7142 ice_for_each_rxq(vsi, i) {
7143 struct ice_rx_ring *ring = vsi->rx_rings[i];
7149 ring->netdev = vsi->netdev;
7150 err = ice_setup_rx_ring(ring);
7159 * ice_vsi_open_ctrl - open control VSI for use
7160 * @vsi: the VSI to open
7162 * Initialization of the Control VSI
7164 * Returns 0 on success, negative value on error
7166 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
7168 char int_name[ICE_INT_NAME_STR_LEN];
7169 struct ice_pf *pf = vsi->back;
7173 dev = ice_pf_to_dev(pf);
7174 /* allocate descriptors */
7175 err = ice_vsi_setup_tx_rings(vsi);
7179 err = ice_vsi_setup_rx_rings(vsi);
7183 err = ice_vsi_cfg_lan(vsi);
7187 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
7188 dev_driver_string(dev), dev_name(dev));
7189 err = ice_vsi_req_irq_msix(vsi, int_name);
7193 ice_vsi_cfg_msix(vsi);
7195 err = ice_vsi_start_all_rx_rings(vsi);
7197 goto err_up_complete;
7199 clear_bit(ICE_VSI_DOWN, vsi->state);
7200 ice_vsi_ena_irq(vsi);
7207 ice_vsi_free_rx_rings(vsi);
7209 ice_vsi_free_tx_rings(vsi);
7215 * ice_vsi_open - Called when a network interface is made active
7216 * @vsi: the VSI to open
7218 * Initialization of the VSI
7220 * Returns 0 on success, negative value on error
7222 int ice_vsi_open(struct ice_vsi *vsi)
7224 char int_name[ICE_INT_NAME_STR_LEN];
7225 struct ice_pf *pf = vsi->back;
7228 /* allocate descriptors */
7229 err = ice_vsi_setup_tx_rings(vsi);
7233 err = ice_vsi_setup_rx_rings(vsi);
7237 err = ice_vsi_cfg_lan(vsi);
7241 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
7242 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
7243 err = ice_vsi_req_irq_msix(vsi, int_name);
7247 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
7249 if (vsi->type == ICE_VSI_PF) {
7250 /* Notify the stack of the actual queue counts. */
7251 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
7255 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
7260 err = ice_up_complete(vsi);
7262 goto err_up_complete;
7269 ice_vsi_free_irq(vsi);
7271 ice_vsi_free_rx_rings(vsi);
7273 ice_vsi_free_tx_rings(vsi);
7279 * ice_vsi_release_all - Delete all VSIs
7280 * @pf: PF from which all VSIs are being removed
7282 static void ice_vsi_release_all(struct ice_pf *pf)
7289 ice_for_each_vsi(pf, i) {
7293 if (pf->vsi[i]->type == ICE_VSI_CHNL)
7296 err = ice_vsi_release(pf->vsi[i]);
7298 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
7299 i, err, pf->vsi[i]->vsi_num);
7304 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
7305 * @pf: pointer to the PF instance
7306 * @type: VSI type to rebuild
7308 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
7310 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
7312 struct device *dev = ice_pf_to_dev(pf);
7315 ice_for_each_vsi(pf, i) {
7316 struct ice_vsi *vsi = pf->vsi[i];
7318 if (!vsi || vsi->type != type)
7321 /* rebuild the VSI */
7322 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
7324 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
7325 err, vsi->idx, ice_vsi_type_str(type));
7329 /* replay filters for the VSI */
7330 err = ice_replay_vsi(&pf->hw, vsi->idx);
7332 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
7333 err, vsi->idx, ice_vsi_type_str(type));
7337 /* Re-map HW VSI number, using VSI handle that has been
7338 * previously validated in ice_replay_vsi() call above
7340 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
7342 /* enable the VSI */
7343 err = ice_ena_vsi(vsi, false);
7345 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
7346 err, vsi->idx, ice_vsi_type_str(type));
7350 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
7351 ice_vsi_type_str(type));
7358 * ice_update_pf_netdev_link - Update PF netdev link status
7359 * @pf: pointer to the PF instance
7361 static void ice_update_pf_netdev_link(struct ice_pf *pf)
7366 ice_for_each_vsi(pf, i) {
7367 struct ice_vsi *vsi = pf->vsi[i];
7369 if (!vsi || vsi->type != ICE_VSI_PF)
7372 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
7374 netif_carrier_on(pf->vsi[i]->netdev);
7375 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
7377 netif_carrier_off(pf->vsi[i]->netdev);
7378 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
7384 * ice_rebuild - rebuild after reset
7385 * @pf: PF to rebuild
7386 * @reset_type: type of reset
7388 * Do not rebuild VF VSI in this flow because that is already handled via
7389 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
7390 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
7391 * to reset/rebuild all the VF VSI twice.
7393 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
7395 struct device *dev = ice_pf_to_dev(pf);
7396 struct ice_hw *hw = &pf->hw;
7400 if (test_bit(ICE_DOWN, pf->state))
7401 goto clear_recovery;
7403 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
7405 #define ICE_EMP_RESET_SLEEP_MS 5000
7406 if (reset_type == ICE_RESET_EMPR) {
7407 /* If an EMP reset has occurred, any previously pending flash
7408 * update will have completed. We no longer know whether or
7409 * not the NVM update EMP reset is restricted.
7411 pf->fw_emp_reset_disabled = false;
7413 msleep(ICE_EMP_RESET_SLEEP_MS);
7416 err = ice_init_all_ctrlq(hw);
7418 dev_err(dev, "control queues init failed %d\n", err);
7419 goto err_init_ctrlq;
7422 /* if DDP was previously loaded successfully */
7423 if (!ice_is_safe_mode(pf)) {
7424 /* reload the SW DB of filter tables */
7425 if (reset_type == ICE_RESET_PFR)
7426 ice_fill_blk_tbls(hw);
7428 /* Reload DDP Package after CORER/GLOBR reset */
7429 ice_load_pkg(NULL, pf);
7432 err = ice_clear_pf_cfg(hw);
7434 dev_err(dev, "clear PF configuration failed %d\n", err);
7435 goto err_init_ctrlq;
7438 ice_clear_pxe_mode(hw);
7440 err = ice_init_nvm(hw);
7442 dev_err(dev, "ice_init_nvm failed %d\n", err);
7443 goto err_init_ctrlq;
7446 err = ice_get_caps(hw);
7448 dev_err(dev, "ice_get_caps failed %d\n", err);
7449 goto err_init_ctrlq;
7452 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7454 dev_err(dev, "set_mac_cfg failed %d\n", err);
7455 goto err_init_ctrlq;
7458 dvm = ice_is_dvm_ena(hw);
7460 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7462 goto err_init_ctrlq;
7464 err = ice_sched_init_port(hw->port_info);
7466 goto err_sched_init_port;
7468 /* start misc vector */
7469 err = ice_req_irq_msix_misc(pf);
7471 dev_err(dev, "misc vector setup failed: %d\n", err);
7472 goto err_sched_init_port;
7475 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7476 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7477 if (!rd32(hw, PFQF_FD_SIZE)) {
7478 u16 unused, guar, b_effort;
7480 guar = hw->func_caps.fd_fltr_guar;
7481 b_effort = hw->func_caps.fd_fltr_best_effort;
7483 /* force guaranteed filter pool for PF */
7484 ice_alloc_fd_guar_item(hw, &unused, guar);
7485 /* force shared filter pool for PF */
7486 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7490 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7491 ice_dcb_rebuild(pf);
7493 /* If the PF previously had enabled PTP, PTP init needs to happen before
7494 * the VSI rebuild. If not, this causes the PTP link status events to
7497 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7500 if (ice_is_feature_supported(pf, ICE_F_GNSS))
7503 /* rebuild PF VSI */
7504 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7506 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7507 goto err_vsi_rebuild;
7510 /* configure PTP timestamping after VSI rebuild */
7511 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7512 ice_ptp_cfg_timestamp(pf, false);
7514 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
7516 dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
7517 goto err_vsi_rebuild;
7520 if (reset_type == ICE_RESET_PFR) {
7521 err = ice_rebuild_channels(pf);
7523 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7525 goto err_vsi_rebuild;
7529 /* If Flow Director is active */
7530 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7531 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7533 dev_err(dev, "control VSI rebuild failed: %d\n", err);
7534 goto err_vsi_rebuild;
7537 /* replay HW Flow Director recipes */
7539 ice_fdir_replay_flows(hw);
7541 /* replay Flow Director filters */
7542 ice_fdir_replay_fltrs(pf);
7544 ice_rebuild_arfs(pf);
7547 ice_update_pf_netdev_link(pf);
7549 /* tell the firmware we are up */
7550 err = ice_send_version(pf);
7552 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7554 goto err_vsi_rebuild;
7557 ice_replay_post(hw);
7559 /* if we get here, reset flow is successful */
7560 clear_bit(ICE_RESET_FAILED, pf->state);
7562 ice_plug_aux_dev(pf);
7566 err_sched_init_port:
7567 ice_sched_cleanup_all(hw);
7569 ice_shutdown_all_ctrlq(hw);
7570 set_bit(ICE_RESET_FAILED, pf->state);
7572 /* set this bit in PF state to control service task scheduling */
7573 set_bit(ICE_NEEDS_RESTART, pf->state);
7574 dev_err(dev, "Rebuild failed, unload and reload driver\n");
7578 * ice_change_mtu - NDO callback to change the MTU
7579 * @netdev: network interface device structure
7580 * @new_mtu: new value for maximum frame size
7582 * Returns 0 on success, negative on failure
7584 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7586 struct ice_netdev_priv *np = netdev_priv(netdev);
7587 struct ice_vsi *vsi = np->vsi;
7588 struct ice_pf *pf = vsi->back;
7589 struct bpf_prog *prog;
7593 if (new_mtu == (int)netdev->mtu) {
7594 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7598 prog = vsi->xdp_prog;
7599 if (prog && !prog->aux->xdp_has_frags) {
7600 int frame_size = ice_max_xdp_frame_size(vsi);
7602 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7603 netdev_err(netdev, "max MTU for XDP usage is %d\n",
7604 frame_size - ICE_ETH_PKT_HDR_PAD);
7607 } else if (test_bit(ICE_FLAG_LEGACY_RX, pf->flags)) {
7608 if (new_mtu + ICE_ETH_PKT_HDR_PAD > ICE_MAX_FRAME_LEGACY_RX) {
7609 netdev_err(netdev, "Too big MTU for legacy-rx; Max is %d\n",
7610 ICE_MAX_FRAME_LEGACY_RX - ICE_ETH_PKT_HDR_PAD);
7615 /* if a reset is in progress, wait for some time for it to complete */
7617 if (ice_is_reset_in_progress(pf->state)) {
7619 usleep_range(1000, 2000);
7624 } while (count < 100);
7627 netdev_err(netdev, "can't change MTU. Device is busy\n");
7631 netdev->mtu = (unsigned int)new_mtu;
7633 /* if VSI is up, bring it down and then back up */
7634 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
7635 err = ice_down(vsi);
7637 netdev_err(netdev, "change MTU if_down err %d\n", err);
7643 netdev_err(netdev, "change MTU if_up err %d\n", err);
7648 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7649 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7655 * ice_eth_ioctl - Access the hwtstamp interface
7656 * @netdev: network interface device structure
7657 * @ifr: interface request data
7658 * @cmd: ioctl command
7660 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7662 struct ice_netdev_priv *np = netdev_priv(netdev);
7663 struct ice_pf *pf = np->vsi->back;
7667 return ice_ptp_get_ts_config(pf, ifr);
7669 return ice_ptp_set_ts_config(pf, ifr);
7676 * ice_aq_str - convert AQ err code to a string
7677 * @aq_err: the AQ error code to convert
7679 const char *ice_aq_str(enum ice_aq_err aq_err)
7684 case ICE_AQ_RC_EPERM:
7685 return "ICE_AQ_RC_EPERM";
7686 case ICE_AQ_RC_ENOENT:
7687 return "ICE_AQ_RC_ENOENT";
7688 case ICE_AQ_RC_ENOMEM:
7689 return "ICE_AQ_RC_ENOMEM";
7690 case ICE_AQ_RC_EBUSY:
7691 return "ICE_AQ_RC_EBUSY";
7692 case ICE_AQ_RC_EEXIST:
7693 return "ICE_AQ_RC_EEXIST";
7694 case ICE_AQ_RC_EINVAL:
7695 return "ICE_AQ_RC_EINVAL";
7696 case ICE_AQ_RC_ENOSPC:
7697 return "ICE_AQ_RC_ENOSPC";
7698 case ICE_AQ_RC_ENOSYS:
7699 return "ICE_AQ_RC_ENOSYS";
7700 case ICE_AQ_RC_EMODE:
7701 return "ICE_AQ_RC_EMODE";
7702 case ICE_AQ_RC_ENOSEC:
7703 return "ICE_AQ_RC_ENOSEC";
7704 case ICE_AQ_RC_EBADSIG:
7705 return "ICE_AQ_RC_EBADSIG";
7706 case ICE_AQ_RC_ESVN:
7707 return "ICE_AQ_RC_ESVN";
7708 case ICE_AQ_RC_EBADMAN:
7709 return "ICE_AQ_RC_EBADMAN";
7710 case ICE_AQ_RC_EBADBUF:
7711 return "ICE_AQ_RC_EBADBUF";
7714 return "ICE_AQ_RC_UNKNOWN";
7718 * ice_set_rss_lut - Set RSS LUT
7719 * @vsi: Pointer to VSI structure
7720 * @lut: Lookup table
7721 * @lut_size: Lookup table size
7723 * Returns 0 on success, negative on failure
7725 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7727 struct ice_aq_get_set_rss_lut_params params = {};
7728 struct ice_hw *hw = &vsi->back->hw;
7734 params.vsi_handle = vsi->idx;
7735 params.lut_size = lut_size;
7736 params.lut_type = vsi->rss_lut_type;
7739 status = ice_aq_set_rss_lut(hw, ¶ms);
7741 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7742 status, ice_aq_str(hw->adminq.sq_last_status));
7748 * ice_set_rss_key - Set RSS key
7749 * @vsi: Pointer to the VSI structure
7750 * @seed: RSS hash seed
7752 * Returns 0 on success, negative on failure
7754 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7756 struct ice_hw *hw = &vsi->back->hw;
7762 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7764 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7765 status, ice_aq_str(hw->adminq.sq_last_status));
7771 * ice_get_rss_lut - Get RSS LUT
7772 * @vsi: Pointer to VSI structure
7773 * @lut: Buffer to store the lookup table entries
7774 * @lut_size: Size of buffer to store the lookup table entries
7776 * Returns 0 on success, negative on failure
7778 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7780 struct ice_aq_get_set_rss_lut_params params = {};
7781 struct ice_hw *hw = &vsi->back->hw;
7787 params.vsi_handle = vsi->idx;
7788 params.lut_size = lut_size;
7789 params.lut_type = vsi->rss_lut_type;
7792 status = ice_aq_get_rss_lut(hw, ¶ms);
7794 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7795 status, ice_aq_str(hw->adminq.sq_last_status));
7801 * ice_get_rss_key - Get RSS key
7802 * @vsi: Pointer to VSI structure
7803 * @seed: Buffer to store the key in
7805 * Returns 0 on success, negative on failure
7807 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7809 struct ice_hw *hw = &vsi->back->hw;
7815 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7817 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7818 status, ice_aq_str(hw->adminq.sq_last_status));
7824 * ice_bridge_getlink - Get the hardware bridge mode
7827 * @seq: RTNL message seq
7828 * @dev: the netdev being configured
7829 * @filter_mask: filter mask passed in
7830 * @nlflags: netlink flags passed in
7832 * Return the bridge mode (VEB/VEPA)
7835 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7836 struct net_device *dev, u32 filter_mask, int nlflags)
7838 struct ice_netdev_priv *np = netdev_priv(dev);
7839 struct ice_vsi *vsi = np->vsi;
7840 struct ice_pf *pf = vsi->back;
7843 bmode = pf->first_sw->bridge_mode;
7845 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7850 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7851 * @vsi: Pointer to VSI structure
7852 * @bmode: Hardware bridge mode (VEB/VEPA)
7854 * Returns 0 on success, negative on failure
7856 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7858 struct ice_aqc_vsi_props *vsi_props;
7859 struct ice_hw *hw = &vsi->back->hw;
7860 struct ice_vsi_ctx *ctxt;
7863 vsi_props = &vsi->info;
7865 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7869 ctxt->info = vsi->info;
7871 if (bmode == BRIDGE_MODE_VEB)
7872 /* change from VEPA to VEB mode */
7873 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7875 /* change from VEB to VEPA mode */
7876 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7877 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7879 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7881 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7882 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7885 /* Update sw flags for book keeping */
7886 vsi_props->sw_flags = ctxt->info.sw_flags;
7894 * ice_bridge_setlink - Set the hardware bridge mode
7895 * @dev: the netdev being configured
7896 * @nlh: RTNL message
7897 * @flags: bridge setlink flags
7898 * @extack: netlink extended ack
7900 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7901 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7902 * not already set for all VSIs connected to this switch. And also update the
7903 * unicast switch filter rules for the corresponding switch of the netdev.
7906 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7907 u16 __always_unused flags,
7908 struct netlink_ext_ack __always_unused *extack)
7910 struct ice_netdev_priv *np = netdev_priv(dev);
7911 struct ice_pf *pf = np->vsi->back;
7912 struct nlattr *attr, *br_spec;
7913 struct ice_hw *hw = &pf->hw;
7914 struct ice_sw *pf_sw;
7915 int rem, v, err = 0;
7917 pf_sw = pf->first_sw;
7918 /* find the attribute in the netlink message */
7919 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7921 nla_for_each_nested(attr, br_spec, rem) {
7924 if (nla_type(attr) != IFLA_BRIDGE_MODE)
7926 mode = nla_get_u16(attr);
7927 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7929 /* Continue if bridge mode is not being flipped */
7930 if (mode == pf_sw->bridge_mode)
7932 /* Iterates through the PF VSI list and update the loopback
7935 ice_for_each_vsi(pf, v) {
7938 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7943 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7944 /* Update the unicast switch filter rules for the corresponding
7945 * switch of the netdev
7947 err = ice_update_sw_rule_bridge_mode(hw);
7949 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7951 ice_aq_str(hw->adminq.sq_last_status));
7952 /* revert hw->evb_veb */
7953 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7957 pf_sw->bridge_mode = mode;
7964 * ice_tx_timeout - Respond to a Tx Hang
7965 * @netdev: network interface device structure
7966 * @txqueue: Tx queue
7968 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7970 struct ice_netdev_priv *np = netdev_priv(netdev);
7971 struct ice_tx_ring *tx_ring = NULL;
7972 struct ice_vsi *vsi = np->vsi;
7973 struct ice_pf *pf = vsi->back;
7976 pf->tx_timeout_count++;
7978 /* Check if PFC is enabled for the TC to which the queue belongs
7979 * to. If yes then Tx timeout is not caused by a hung queue, no
7980 * need to reset and rebuild
7982 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7983 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7988 /* now that we have an index, find the tx_ring struct */
7989 ice_for_each_txq(vsi, i)
7990 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7991 if (txqueue == vsi->tx_rings[i]->q_index) {
7992 tx_ring = vsi->tx_rings[i];
7996 /* Reset recovery level if enough time has elapsed after last timeout.
7997 * Also ensure no new reset action happens before next timeout period.
7999 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
8000 pf->tx_timeout_recovery_level = 1;
8001 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
8002 netdev->watchdog_timeo)))
8006 struct ice_hw *hw = &pf->hw;
8009 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
8010 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
8011 /* Read interrupt register */
8012 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
8014 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
8015 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
8016 head, tx_ring->next_to_use, val);
8019 pf->tx_timeout_last_recovery = jiffies;
8020 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
8021 pf->tx_timeout_recovery_level, txqueue);
8023 switch (pf->tx_timeout_recovery_level) {
8025 set_bit(ICE_PFR_REQ, pf->state);
8028 set_bit(ICE_CORER_REQ, pf->state);
8031 set_bit(ICE_GLOBR_REQ, pf->state);
8034 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
8035 set_bit(ICE_DOWN, pf->state);
8036 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
8037 set_bit(ICE_SERVICE_DIS, pf->state);
8041 ice_service_task_schedule(pf);
8042 pf->tx_timeout_recovery_level++;
8046 * ice_setup_tc_cls_flower - flower classifier offloads
8047 * @np: net device to configure
8048 * @filter_dev: device on which filter is added
8049 * @cls_flower: offload data
8052 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
8053 struct net_device *filter_dev,
8054 struct flow_cls_offload *cls_flower)
8056 struct ice_vsi *vsi = np->vsi;
8058 if (cls_flower->common.chain_index)
8061 switch (cls_flower->command) {
8062 case FLOW_CLS_REPLACE:
8063 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
8064 case FLOW_CLS_DESTROY:
8065 return ice_del_cls_flower(vsi, cls_flower);
8072 * ice_setup_tc_block_cb - callback handler registered for TC block
8073 * @type: TC SETUP type
8074 * @type_data: TC flower offload data that contains user input
8075 * @cb_priv: netdev private data
8078 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
8080 struct ice_netdev_priv *np = cb_priv;
8083 case TC_SETUP_CLSFLOWER:
8084 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
8092 * ice_validate_mqprio_qopt - Validate TCF input parameters
8093 * @vsi: Pointer to VSI
8094 * @mqprio_qopt: input parameters for mqprio queue configuration
8096 * This function validates MQPRIO params, such as qcount (power of 2 wherever
8097 * needed), and make sure user doesn't specify qcount and BW rate limit
8098 * for TCs, which are more than "num_tc"
8101 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
8102 struct tc_mqprio_qopt_offload *mqprio_qopt)
8104 u64 sum_max_rate = 0, sum_min_rate = 0;
8105 int non_power_of_2_qcount = 0;
8106 struct ice_pf *pf = vsi->back;
8107 int max_rss_q_cnt = 0;
8112 if (vsi->type != ICE_VSI_PF)
8115 if (mqprio_qopt->qopt.offset[0] != 0 ||
8116 mqprio_qopt->qopt.num_tc < 1 ||
8117 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
8120 dev = ice_pf_to_dev(pf);
8121 vsi->ch_rss_size = 0;
8122 num_tc = mqprio_qopt->qopt.num_tc;
8124 for (i = 0; num_tc; i++) {
8125 int qcount = mqprio_qopt->qopt.count[i];
8126 u64 max_rate, min_rate, rem;
8131 if (is_power_of_2(qcount)) {
8132 if (non_power_of_2_qcount &&
8133 qcount > non_power_of_2_qcount) {
8134 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
8135 qcount, non_power_of_2_qcount);
8138 if (qcount > max_rss_q_cnt)
8139 max_rss_q_cnt = qcount;
8141 if (non_power_of_2_qcount &&
8142 qcount != non_power_of_2_qcount) {
8143 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
8144 qcount, non_power_of_2_qcount);
8147 if (qcount < max_rss_q_cnt) {
8148 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
8149 qcount, max_rss_q_cnt);
8152 max_rss_q_cnt = qcount;
8153 non_power_of_2_qcount = qcount;
8156 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
8157 * converts the bandwidth rate limit into Bytes/s when
8158 * passing it down to the driver. So convert input bandwidth
8159 * from Bytes/s to Kbps
8161 max_rate = mqprio_qopt->max_rate[i];
8162 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
8163 sum_max_rate += max_rate;
8165 /* min_rate is minimum guaranteed rate and it can't be zero */
8166 min_rate = mqprio_qopt->min_rate[i];
8167 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
8168 sum_min_rate += min_rate;
8170 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
8171 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
8172 min_rate, ICE_MIN_BW_LIMIT);
8176 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
8178 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
8179 i, ICE_MIN_BW_LIMIT);
8183 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
8185 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
8186 i, ICE_MIN_BW_LIMIT);
8190 /* min_rate can't be more than max_rate, except when max_rate
8191 * is zero (implies max_rate sought is max line rate). In such
8192 * a case min_rate can be more than max.
8194 if (max_rate && min_rate > max_rate) {
8195 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
8196 min_rate, max_rate);
8200 if (i >= mqprio_qopt->qopt.num_tc - 1)
8202 if (mqprio_qopt->qopt.offset[i + 1] !=
8203 (mqprio_qopt->qopt.offset[i] + qcount))
8207 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8210 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
8213 speed = ice_get_link_speed_kbps(vsi);
8214 if (sum_max_rate && sum_max_rate > (u64)speed) {
8215 dev_err(dev, "Invalid max Tx rate(%llu) Kbps > speed(%u) Kbps specified\n",
8216 sum_max_rate, speed);
8219 if (sum_min_rate && sum_min_rate > (u64)speed) {
8220 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
8221 sum_min_rate, speed);
8225 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
8226 vsi->ch_rss_size = max_rss_q_cnt;
8232 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
8233 * @pf: ptr to PF device
8236 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
8238 struct device *dev = ice_pf_to_dev(pf);
8243 if (!(vsi->num_gfltr || vsi->num_bfltr))
8247 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
8248 struct ice_fd_hw_prof *prof;
8252 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
8253 hw->fdir_prof[flow]->cnt))
8256 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
8257 enum ice_flow_priority prio;
8260 /* add this VSI to FDir profile for this flow */
8261 prio = ICE_FLOW_PRIO_NORMAL;
8262 prof = hw->fdir_prof[flow];
8263 prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
8264 status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
8265 prof->vsi_h[0], vsi->idx,
8266 prio, prof->fdir_seg[tun],
8269 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
8274 prof->entry_h[prof->cnt][tun] = entry_h;
8277 /* store VSI for filter replay and delete */
8278 prof->vsi_h[prof->cnt] = vsi->idx;
8282 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
8287 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
8293 * ice_add_channel - add a channel by adding VSI
8294 * @pf: ptr to PF device
8295 * @sw_id: underlying HW switching element ID
8296 * @ch: ptr to channel structure
8298 * Add a channel (VSI) using add_vsi and queue_map
8300 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
8302 struct device *dev = ice_pf_to_dev(pf);
8303 struct ice_vsi *vsi;
8305 if (ch->type != ICE_VSI_CHNL) {
8306 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
8310 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
8311 if (!vsi || vsi->type != ICE_VSI_CHNL) {
8312 dev_err(dev, "create chnl VSI failure\n");
8316 ice_add_vsi_to_fdir(pf, vsi);
8319 ch->vsi_num = vsi->vsi_num;
8320 ch->info.mapping_flags = vsi->info.mapping_flags;
8322 /* set the back pointer of channel for newly created VSI */
8325 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
8326 sizeof(vsi->info.q_mapping));
8327 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
8328 sizeof(vsi->info.tc_mapping));
8335 * @vsi: the VSI being setup
8336 * @ch: ptr to channel structure
8338 * Configure channel specific resources such as rings, vector.
8340 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
8344 for (i = 0; i < ch->num_txq; i++) {
8345 struct ice_q_vector *tx_q_vector, *rx_q_vector;
8346 struct ice_ring_container *rc;
8347 struct ice_tx_ring *tx_ring;
8348 struct ice_rx_ring *rx_ring;
8350 tx_ring = vsi->tx_rings[ch->base_q + i];
8351 rx_ring = vsi->rx_rings[ch->base_q + i];
8352 if (!tx_ring || !rx_ring)
8355 /* setup ring being channel enabled */
8359 /* following code block sets up vector specific attributes */
8360 tx_q_vector = tx_ring->q_vector;
8361 rx_q_vector = rx_ring->q_vector;
8362 if (!tx_q_vector && !rx_q_vector)
8366 tx_q_vector->ch = ch;
8367 /* setup Tx and Rx ITR setting if DIM is off */
8368 rc = &tx_q_vector->tx;
8369 if (!ITR_IS_DYNAMIC(rc))
8370 ice_write_itr(rc, rc->itr_setting);
8373 rx_q_vector->ch = ch;
8374 /* setup Tx and Rx ITR setting if DIM is off */
8375 rc = &rx_q_vector->rx;
8376 if (!ITR_IS_DYNAMIC(rc))
8377 ice_write_itr(rc, rc->itr_setting);
8381 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
8382 * GLINT_ITR register would have written to perform in-context
8383 * update, hence perform flush
8385 if (ch->num_txq || ch->num_rxq)
8386 ice_flush(&vsi->back->hw);
8390 * ice_cfg_chnl_all_res - configure channel resources
8391 * @vsi: pte to main_vsi
8392 * @ch: ptr to channel structure
8394 * This function configures channel specific resources such as flow-director
8395 * counter index, and other resources such as queues, vectors, ITR settings
8398 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8400 /* configure channel (aka ADQ) resources such as queues, vectors,
8401 * ITR settings for channel specific vectors and anything else
8403 ice_chnl_cfg_res(vsi, ch);
8407 * ice_setup_hw_channel - setup new channel
8408 * @pf: ptr to PF device
8409 * @vsi: the VSI being setup
8410 * @ch: ptr to channel structure
8411 * @sw_id: underlying HW switching element ID
8412 * @type: type of channel to be created (VMDq2/VF)
8414 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8415 * and configures Tx rings accordingly
8418 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8419 struct ice_channel *ch, u16 sw_id, u8 type)
8421 struct device *dev = ice_pf_to_dev(pf);
8424 ch->base_q = vsi->next_base_q;
8427 ret = ice_add_channel(pf, sw_id, ch);
8429 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8433 /* configure/setup ADQ specific resources */
8434 ice_cfg_chnl_all_res(vsi, ch);
8436 /* make sure to update the next_base_q so that subsequent channel's
8437 * (aka ADQ) VSI queue map is correct
8439 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8440 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8447 * ice_setup_channel - setup new channel using uplink element
8448 * @pf: ptr to PF device
8449 * @vsi: the VSI being setup
8450 * @ch: ptr to channel structure
8452 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8453 * and uplink switching element
8456 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8457 struct ice_channel *ch)
8459 struct device *dev = ice_pf_to_dev(pf);
8463 if (vsi->type != ICE_VSI_PF) {
8464 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8468 sw_id = pf->first_sw->sw_id;
8470 /* create channel (VSI) */
8471 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8473 dev_err(dev, "failed to setup hw_channel\n");
8476 dev_dbg(dev, "successfully created channel()\n");
8478 return ch->ch_vsi ? true : false;
8482 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8483 * @vsi: VSI to be configured
8484 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8485 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8488 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8492 err = ice_set_min_bw_limit(vsi, min_tx_rate);
8496 return ice_set_max_bw_limit(vsi, max_tx_rate);
8500 * ice_create_q_channel - function to create channel
8501 * @vsi: VSI to be configured
8502 * @ch: ptr to channel (it contains channel specific params)
8504 * This function creates channel (VSI) using num_queues specified by user,
8505 * reconfigs RSS if needed.
8507 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8509 struct ice_pf *pf = vsi->back;
8515 dev = ice_pf_to_dev(pf);
8516 if (!ch->num_txq || !ch->num_rxq) {
8517 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8521 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8522 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8523 vsi->cnt_q_avail, ch->num_txq);
8527 if (!ice_setup_channel(pf, vsi, ch)) {
8528 dev_info(dev, "Failed to setup channel\n");
8531 /* configure BW rate limit */
8532 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8535 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8538 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8539 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8541 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8542 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8545 vsi->cnt_q_avail -= ch->num_txq;
8551 * ice_rem_all_chnl_fltrs - removes all channel filters
8552 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8554 * Remove all advanced switch filters only if they are channel specific
8555 * tc-flower based filter
8557 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8559 struct ice_tc_flower_fltr *fltr;
8560 struct hlist_node *node;
8562 /* to remove all channel filters, iterate an ordered list of filters */
8563 hlist_for_each_entry_safe(fltr, node,
8564 &pf->tc_flower_fltr_list,
8566 struct ice_rule_query_data rule;
8569 /* for now process only channel specific filters */
8570 if (!ice_is_chnl_fltr(fltr))
8573 rule.rid = fltr->rid;
8574 rule.rule_id = fltr->rule_id;
8575 rule.vsi_handle = fltr->dest_vsi_handle;
8576 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8578 if (status == -ENOENT)
8579 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8582 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8584 } else if (fltr->dest_vsi) {
8585 /* update advanced switch filter count */
8586 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8587 u32 flags = fltr->flags;
8589 fltr->dest_vsi->num_chnl_fltr--;
8590 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8591 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8592 pf->num_dmac_chnl_fltrs--;
8596 hlist_del(&fltr->tc_flower_node);
8602 * ice_remove_q_channels - Remove queue channels for the TCs
8603 * @vsi: VSI to be configured
8604 * @rem_fltr: delete advanced switch filter or not
8606 * Remove queue channels for the TCs
8608 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8610 struct ice_channel *ch, *ch_tmp;
8611 struct ice_pf *pf = vsi->back;
8614 /* remove all tc-flower based filter if they are channel filters only */
8616 ice_rem_all_chnl_fltrs(pf);
8618 /* remove ntuple filters since queue configuration is being changed */
8619 if (vsi->netdev->features & NETIF_F_NTUPLE) {
8620 struct ice_hw *hw = &pf->hw;
8622 mutex_lock(&hw->fdir_fltr_lock);
8623 ice_fdir_del_all_fltrs(vsi);
8624 mutex_unlock(&hw->fdir_fltr_lock);
8627 /* perform cleanup for channels if they exist */
8628 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8629 struct ice_vsi *ch_vsi;
8631 list_del(&ch->list);
8632 ch_vsi = ch->ch_vsi;
8638 /* Reset queue contexts */
8639 for (i = 0; i < ch->num_rxq; i++) {
8640 struct ice_tx_ring *tx_ring;
8641 struct ice_rx_ring *rx_ring;
8643 tx_ring = vsi->tx_rings[ch->base_q + i];
8644 rx_ring = vsi->rx_rings[ch->base_q + i];
8647 if (tx_ring->q_vector)
8648 tx_ring->q_vector->ch = NULL;
8652 if (rx_ring->q_vector)
8653 rx_ring->q_vector->ch = NULL;
8657 /* Release FD resources for the channel VSI */
8658 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8660 /* clear the VSI from scheduler tree */
8661 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8663 /* Delete VSI from FW, PF and HW VSI arrays */
8664 ice_vsi_delete(ch->ch_vsi);
8666 /* free the channel */
8670 /* clear the channel VSI map which is stored in main VSI */
8671 ice_for_each_chnl_tc(i)
8672 vsi->tc_map_vsi[i] = NULL;
8674 /* reset main VSI's all TC information */
8680 * ice_rebuild_channels - rebuild channel
8683 * Recreate channel VSIs and replay filters
8685 static int ice_rebuild_channels(struct ice_pf *pf)
8687 struct device *dev = ice_pf_to_dev(pf);
8688 struct ice_vsi *main_vsi;
8689 bool rem_adv_fltr = true;
8690 struct ice_channel *ch;
8691 struct ice_vsi *vsi;
8695 main_vsi = ice_get_main_vsi(pf);
8699 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8700 main_vsi->old_numtc == 1)
8701 return 0; /* nothing to be done */
8703 /* reconfigure main VSI based on old value of TC and cached values
8706 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8708 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8709 main_vsi->old_ena_tc, main_vsi->vsi_num);
8713 /* rebuild ADQ VSIs */
8714 ice_for_each_vsi(pf, i) {
8715 enum ice_vsi_type type;
8718 if (!vsi || vsi->type != ICE_VSI_CHNL)
8723 /* rebuild ADQ VSI */
8724 err = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_INIT);
8726 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8727 ice_vsi_type_str(type), vsi->idx, err);
8731 /* Re-map HW VSI number, using VSI handle that has been
8732 * previously validated in ice_replay_vsi() call above
8734 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8736 /* replay filters for the VSI */
8737 err = ice_replay_vsi(&pf->hw, vsi->idx);
8739 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8740 ice_vsi_type_str(type), err, vsi->idx);
8741 rem_adv_fltr = false;
8744 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8745 ice_vsi_type_str(type), vsi->idx);
8747 /* store ADQ VSI at correct TC index in main VSI's
8750 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8753 /* ADQ VSI(s) has been rebuilt successfully, so setup
8754 * channel for main VSI's Tx and Rx rings
8756 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8757 struct ice_vsi *ch_vsi;
8759 ch_vsi = ch->ch_vsi;
8763 /* reconfig channel resources */
8764 ice_cfg_chnl_all_res(main_vsi, ch);
8766 /* replay BW rate limit if it is non-zero */
8767 if (!ch->max_tx_rate && !ch->min_tx_rate)
8770 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8773 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",
8774 err, ch->max_tx_rate, ch->min_tx_rate,
8777 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8778 ch->max_tx_rate, ch->min_tx_rate,
8782 /* reconfig RSS for main VSI */
8783 if (main_vsi->ch_rss_size)
8784 ice_vsi_cfg_rss_lut_key(main_vsi);
8789 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8794 * ice_create_q_channels - Add queue channel for the given TCs
8795 * @vsi: VSI to be configured
8797 * Configures queue channel mapping to the given TCs
8799 static int ice_create_q_channels(struct ice_vsi *vsi)
8801 struct ice_pf *pf = vsi->back;
8802 struct ice_channel *ch;
8805 ice_for_each_chnl_tc(i) {
8806 if (!(vsi->all_enatc & BIT(i)))
8809 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8814 INIT_LIST_HEAD(&ch->list);
8815 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8816 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8817 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8818 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8819 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8821 /* convert to Kbits/s */
8822 if (ch->max_tx_rate)
8823 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8824 ICE_BW_KBPS_DIVISOR);
8825 if (ch->min_tx_rate)
8826 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8827 ICE_BW_KBPS_DIVISOR);
8829 ret = ice_create_q_channel(vsi, ch);
8831 dev_err(ice_pf_to_dev(pf),
8832 "failed creating channel TC:%d\n", i);
8836 list_add_tail(&ch->list, &vsi->ch_list);
8837 vsi->tc_map_vsi[i] = ch->ch_vsi;
8838 dev_dbg(ice_pf_to_dev(pf),
8839 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8844 ice_remove_q_channels(vsi, false);
8850 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8851 * @netdev: net device to configure
8852 * @type_data: TC offload data
8854 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8856 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8857 struct ice_netdev_priv *np = netdev_priv(netdev);
8858 struct ice_vsi *vsi = np->vsi;
8859 struct ice_pf *pf = vsi->back;
8860 u16 mode, ena_tc_qdisc = 0;
8861 int cur_txq, cur_rxq;
8866 dev = ice_pf_to_dev(pf);
8867 num_tcf = mqprio_qopt->qopt.num_tc;
8868 hw = mqprio_qopt->qopt.hw;
8869 mode = mqprio_qopt->mode;
8871 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8872 vsi->ch_rss_size = 0;
8873 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8877 /* Generate queue region map for number of TCF requested */
8878 for (i = 0; i < num_tcf; i++)
8879 ena_tc_qdisc |= BIT(i);
8882 case TC_MQPRIO_MODE_CHANNEL:
8884 if (pf->hw.port_info->is_custom_tx_enabled) {
8885 dev_err(dev, "Custom Tx scheduler feature enabled, can't configure ADQ\n");
8888 ice_tear_down_devlink_rate_tree(pf);
8890 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8892 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8896 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8897 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8898 /* don't assume state of hw_tc_offload during driver load
8899 * and set the flag for TC flower filter if hw_tc_offload
8902 if (vsi->netdev->features & NETIF_F_HW_TC)
8903 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8911 /* Requesting same TCF configuration as already enabled */
8912 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8913 mode != TC_MQPRIO_MODE_CHANNEL)
8916 /* Pause VSI queues */
8917 ice_dis_vsi(vsi, true);
8919 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8920 ice_remove_q_channels(vsi, true);
8922 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8923 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8925 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8928 /* logic to rebuild VSI, same like ethtool -L */
8929 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8931 for (i = 0; i < num_tcf; i++) {
8932 if (!(ena_tc_qdisc & BIT(i)))
8935 offset = vsi->mqprio_qopt.qopt.offset[i];
8936 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8937 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8939 vsi->req_txq = offset + qcount_tx;
8940 vsi->req_rxq = offset + qcount_rx;
8942 /* store away original rss_size info, so that it gets reused
8943 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8944 * determine, what should be the rss_sizefor main VSI
8946 vsi->orig_rss_size = vsi->rss_size;
8949 /* save current values of Tx and Rx queues before calling VSI rebuild
8950 * for fallback option
8952 cur_txq = vsi->num_txq;
8953 cur_rxq = vsi->num_rxq;
8955 /* proceed with rebuild main VSI using correct number of queues */
8956 ret = ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT);
8958 /* fallback to current number of queues */
8959 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8960 vsi->req_txq = cur_txq;
8961 vsi->req_rxq = cur_rxq;
8962 clear_bit(ICE_RESET_FAILED, pf->state);
8963 if (ice_vsi_rebuild(vsi, ICE_VSI_FLAG_NO_INIT)) {
8964 dev_err(dev, "Rebuild of main VSI failed again\n");
8969 vsi->all_numtc = num_tcf;
8970 vsi->all_enatc = ena_tc_qdisc;
8971 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8973 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8978 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8979 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8980 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8982 /* set TC0 rate limit if specified */
8983 if (max_tx_rate || min_tx_rate) {
8984 /* convert to Kbits/s */
8986 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8988 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8990 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8992 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8993 max_tx_rate, min_tx_rate, vsi->vsi_num);
8995 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8996 max_tx_rate, min_tx_rate, vsi->vsi_num);
9000 ret = ice_create_q_channels(vsi);
9002 netdev_err(netdev, "failed configuring queue channels\n");
9005 netdev_dbg(netdev, "successfully configured channels\n");
9009 if (vsi->ch_rss_size)
9010 ice_vsi_cfg_rss_lut_key(vsi);
9013 /* if error, reset the all_numtc and all_enatc */
9019 ice_ena_vsi(vsi, true);
9024 static LIST_HEAD(ice_block_cb_list);
9027 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
9030 struct ice_netdev_priv *np = netdev_priv(netdev);
9031 struct ice_pf *pf = np->vsi->back;
9035 case TC_SETUP_BLOCK:
9036 return flow_block_cb_setup_simple(type_data,
9038 ice_setup_tc_block_cb,
9040 case TC_SETUP_QDISC_MQPRIO:
9041 /* setup traffic classifier for receive side */
9042 mutex_lock(&pf->tc_mutex);
9043 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
9044 mutex_unlock(&pf->tc_mutex);
9052 static struct ice_indr_block_priv *
9053 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
9054 struct net_device *netdev)
9056 struct ice_indr_block_priv *cb_priv;
9058 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
9059 if (!cb_priv->netdev)
9061 if (cb_priv->netdev == netdev)
9068 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
9071 struct ice_indr_block_priv *priv = indr_priv;
9072 struct ice_netdev_priv *np = priv->np;
9075 case TC_SETUP_CLSFLOWER:
9076 return ice_setup_tc_cls_flower(np, priv->netdev,
9077 (struct flow_cls_offload *)
9085 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
9086 struct ice_netdev_priv *np,
9087 struct flow_block_offload *f, void *data,
9088 void (*cleanup)(struct flow_block_cb *block_cb))
9090 struct ice_indr_block_priv *indr_priv;
9091 struct flow_block_cb *block_cb;
9093 if (!ice_is_tunnel_supported(netdev) &&
9094 !(is_vlan_dev(netdev) &&
9095 vlan_dev_real_dev(netdev) == np->vsi->netdev))
9098 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
9101 switch (f->command) {
9102 case FLOW_BLOCK_BIND:
9103 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9107 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
9111 indr_priv->netdev = netdev;
9113 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
9116 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
9117 indr_priv, indr_priv,
9118 ice_rep_indr_tc_block_unbind,
9119 f, netdev, sch, data, np,
9122 if (IS_ERR(block_cb)) {
9123 list_del(&indr_priv->list);
9125 return PTR_ERR(block_cb);
9127 flow_block_cb_add(block_cb, f);
9128 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
9130 case FLOW_BLOCK_UNBIND:
9131 indr_priv = ice_indr_block_priv_lookup(np, netdev);
9135 block_cb = flow_block_cb_lookup(f->block,
9136 ice_indr_setup_block_cb,
9141 flow_indr_block_cb_remove(block_cb, f);
9143 list_del(&block_cb->driver_list);
9152 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
9153 void *cb_priv, enum tc_setup_type type, void *type_data,
9155 void (*cleanup)(struct flow_block_cb *block_cb))
9158 case TC_SETUP_BLOCK:
9159 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
9168 * ice_open - Called when a network interface becomes active
9169 * @netdev: network interface device structure
9171 * The open entry point is called when a network interface is made
9172 * active by the system (IFF_UP). At this point all resources needed
9173 * for transmit and receive operations are allocated, the interrupt
9174 * handler is registered with the OS, the netdev watchdog is enabled,
9175 * and the stack is notified that the interface is ready.
9177 * Returns 0 on success, negative value on failure
9179 int ice_open(struct net_device *netdev)
9181 struct ice_netdev_priv *np = netdev_priv(netdev);
9182 struct ice_pf *pf = np->vsi->back;
9184 if (ice_is_reset_in_progress(pf->state)) {
9185 netdev_err(netdev, "can't open net device while reset is in progress");
9189 return ice_open_internal(netdev);
9193 * ice_open_internal - Called when a network interface becomes active
9194 * @netdev: network interface device structure
9196 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
9199 * Returns 0 on success, negative value on failure
9201 int ice_open_internal(struct net_device *netdev)
9203 struct ice_netdev_priv *np = netdev_priv(netdev);
9204 struct ice_vsi *vsi = np->vsi;
9205 struct ice_pf *pf = vsi->back;
9206 struct ice_port_info *pi;
9209 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
9210 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
9214 netif_carrier_off(netdev);
9216 pi = vsi->port_info;
9217 err = ice_update_link_info(pi);
9219 netdev_err(netdev, "Failed to get link info, error %d\n", err);
9223 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
9225 /* Set PHY if there is media, otherwise, turn off PHY */
9226 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
9227 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9228 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
9229 err = ice_init_phy_user_cfg(pi);
9231 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
9237 err = ice_configure_phy(vsi);
9239 netdev_err(netdev, "Failed to set physical link up, error %d\n",
9244 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
9245 ice_set_link(vsi, false);
9248 err = ice_vsi_open(vsi);
9250 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
9251 vsi->vsi_num, vsi->vsw->sw_id);
9253 /* Update existing tunnels information */
9254 udp_tunnel_get_rx_info(netdev);
9260 * ice_stop - Disables a network interface
9261 * @netdev: network interface device structure
9263 * The stop entry point is called when an interface is de-activated by the OS,
9264 * and the netdevice enters the DOWN state. The hardware is still under the
9265 * driver's control, but the netdev interface is disabled.
9267 * Returns success only - not allowed to fail
9269 int ice_stop(struct net_device *netdev)
9271 struct ice_netdev_priv *np = netdev_priv(netdev);
9272 struct ice_vsi *vsi = np->vsi;
9273 struct ice_pf *pf = vsi->back;
9275 if (ice_is_reset_in_progress(pf->state)) {
9276 netdev_err(netdev, "can't stop net device while reset is in progress");
9280 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
9281 int link_err = ice_force_phys_link_state(vsi, false);
9284 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
9285 vsi->vsi_num, link_err);
9296 * ice_features_check - Validate encapsulated packet conforms to limits
9298 * @netdev: This port's netdev
9299 * @features: Offload features that the stack believes apply
9301 static netdev_features_t
9302 ice_features_check(struct sk_buff *skb,
9303 struct net_device __always_unused *netdev,
9304 netdev_features_t features)
9306 bool gso = skb_is_gso(skb);
9309 /* No point in doing any of this if neither checksum nor GSO are
9310 * being requested for this frame. We can rule out both by just
9311 * checking for CHECKSUM_PARTIAL
9313 if (skb->ip_summed != CHECKSUM_PARTIAL)
9316 /* We cannot support GSO if the MSS is going to be less than
9317 * 64 bytes. If it is then we need to drop support for GSO.
9319 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
9320 features &= ~NETIF_F_GSO_MASK;
9322 len = skb_network_offset(skb);
9323 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
9324 goto out_rm_features;
9326 len = skb_network_header_len(skb);
9327 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9328 goto out_rm_features;
9330 if (skb->encapsulation) {
9331 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
9332 * the case of IPIP frames, the transport header pointer is
9333 * after the inner header! So check to make sure that this
9334 * is a GRE or UDP_TUNNEL frame before doing that math.
9336 if (gso && (skb_shinfo(skb)->gso_type &
9337 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
9338 len = skb_inner_network_header(skb) -
9339 skb_transport_header(skb);
9340 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
9341 goto out_rm_features;
9344 len = skb_inner_network_header_len(skb);
9345 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
9346 goto out_rm_features;
9351 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9354 static const struct net_device_ops ice_netdev_safe_mode_ops = {
9355 .ndo_open = ice_open,
9356 .ndo_stop = ice_stop,
9357 .ndo_start_xmit = ice_start_xmit,
9358 .ndo_set_mac_address = ice_set_mac_address,
9359 .ndo_validate_addr = eth_validate_addr,
9360 .ndo_change_mtu = ice_change_mtu,
9361 .ndo_get_stats64 = ice_get_stats64,
9362 .ndo_tx_timeout = ice_tx_timeout,
9363 .ndo_bpf = ice_xdp_safe_mode,
9366 static const struct net_device_ops ice_netdev_ops = {
9367 .ndo_open = ice_open,
9368 .ndo_stop = ice_stop,
9369 .ndo_start_xmit = ice_start_xmit,
9370 .ndo_select_queue = ice_select_queue,
9371 .ndo_features_check = ice_features_check,
9372 .ndo_fix_features = ice_fix_features,
9373 .ndo_set_rx_mode = ice_set_rx_mode,
9374 .ndo_set_mac_address = ice_set_mac_address,
9375 .ndo_validate_addr = eth_validate_addr,
9376 .ndo_change_mtu = ice_change_mtu,
9377 .ndo_get_stats64 = ice_get_stats64,
9378 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
9379 .ndo_eth_ioctl = ice_eth_ioctl,
9380 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
9381 .ndo_set_vf_mac = ice_set_vf_mac,
9382 .ndo_get_vf_config = ice_get_vf_cfg,
9383 .ndo_set_vf_trust = ice_set_vf_trust,
9384 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
9385 .ndo_set_vf_link_state = ice_set_vf_link_state,
9386 .ndo_get_vf_stats = ice_get_vf_stats,
9387 .ndo_set_vf_rate = ice_set_vf_bw,
9388 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
9389 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
9390 .ndo_setup_tc = ice_setup_tc,
9391 .ndo_set_features = ice_set_features,
9392 .ndo_bridge_getlink = ice_bridge_getlink,
9393 .ndo_bridge_setlink = ice_bridge_setlink,
9394 .ndo_fdb_add = ice_fdb_add,
9395 .ndo_fdb_del = ice_fdb_del,
9396 #ifdef CONFIG_RFS_ACCEL
9397 .ndo_rx_flow_steer = ice_rx_flow_steer,
9399 .ndo_tx_timeout = ice_tx_timeout,
9401 .ndo_xdp_xmit = ice_xdp_xmit,
9402 .ndo_xsk_wakeup = ice_xsk_wakeup,