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"
26 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
27 static const char ice_driver_string[] = DRV_SUMMARY;
28 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
30 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
31 #define ICE_DDP_PKG_PATH "intel/ice/ddp/"
32 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
34 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
35 MODULE_DESCRIPTION(DRV_SUMMARY);
36 MODULE_LICENSE("GPL v2");
37 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
39 static int debug = -1;
40 module_param(debug, int, 0644);
41 #ifndef CONFIG_DYNAMIC_DEBUG
42 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
44 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
45 #endif /* !CONFIG_DYNAMIC_DEBUG */
47 static DEFINE_IDA(ice_aux_ida);
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];
136 if (ice_ring_ch_enabled(tx_ring))
140 /* If packet counter has not changed the queue is
141 * likely stalled, so force an interrupt for this
144 * prev_pkt would be negative if there was no
147 packets = tx_ring->stats.pkts & INT_MAX;
148 if (tx_ring->tx_stats.prev_pkt == packets) {
149 /* Trigger sw interrupt to revive the queue */
150 ice_trigger_sw_intr(hw, tx_ring->q_vector);
154 /* Memory barrier between read of packet count and call
155 * to ice_get_tx_pending()
158 tx_ring->tx_stats.prev_pkt =
159 ice_get_tx_pending(tx_ring) ? packets : -1;
165 * ice_init_mac_fltr - Set initial MAC filters
166 * @pf: board private structure
168 * Set initial set of MAC filters for PF VSI; configure filters for permanent
169 * address and broadcast address. If an error is encountered, netdevice will be
172 static int ice_init_mac_fltr(struct ice_pf *pf)
177 vsi = ice_get_main_vsi(pf);
181 perm_addr = vsi->port_info->mac.perm_addr;
182 return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
186 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
187 * @netdev: the net device on which the sync is happening
188 * @addr: MAC address to sync
190 * This is a callback function which is called by the in kernel device sync
191 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
192 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
193 * MAC filters from the hardware.
195 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
197 struct ice_netdev_priv *np = netdev_priv(netdev);
198 struct ice_vsi *vsi = np->vsi;
200 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
208 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
209 * @netdev: the net device on which the unsync is happening
210 * @addr: MAC address to unsync
212 * This is a callback function which is called by the in kernel device unsync
213 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
214 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
215 * delete the MAC filters from the hardware.
217 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
219 struct ice_netdev_priv *np = netdev_priv(netdev);
220 struct ice_vsi *vsi = np->vsi;
222 /* Under some circumstances, we might receive a request to delete our
223 * own device address from our uc list. Because we store the device
224 * address in the VSI's MAC filter list, we need to ignore such
225 * requests and not delete our device address from this list.
227 if (ether_addr_equal(addr, netdev->dev_addr))
230 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
238 * ice_vsi_fltr_changed - check if filter state changed
239 * @vsi: VSI to be checked
241 * returns true if filter state has changed, false otherwise.
243 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
245 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
246 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
250 * ice_set_promisc - Enable promiscuous mode for a given PF
251 * @vsi: the VSI being configured
252 * @promisc_m: mask of promiscuous config bits
255 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
259 if (vsi->type != ICE_VSI_PF)
262 if (ice_vsi_has_non_zero_vlans(vsi)) {
263 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
264 status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi,
267 status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
270 if (status && status != -EEXIST)
277 * ice_clear_promisc - Disable promiscuous mode for a given PF
278 * @vsi: the VSI being configured
279 * @promisc_m: mask of promiscuous config bits
282 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
286 if (vsi->type != ICE_VSI_PF)
289 if (ice_vsi_has_non_zero_vlans(vsi)) {
290 promisc_m |= (ICE_PROMISC_VLAN_RX | ICE_PROMISC_VLAN_TX);
291 status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi,
294 status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
302 * ice_get_devlink_port - Get devlink port from netdev
303 * @netdev: the netdevice structure
305 static struct devlink_port *ice_get_devlink_port(struct net_device *netdev)
307 struct ice_pf *pf = ice_netdev_to_pf(netdev);
309 if (!ice_is_switchdev_running(pf))
312 return &pf->devlink_port;
316 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
317 * @vsi: ptr to the VSI
319 * Push any outstanding VSI filter changes through the AdminQ.
321 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
323 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
324 struct device *dev = ice_pf_to_dev(vsi->back);
325 struct net_device *netdev = vsi->netdev;
326 bool promisc_forced_on = false;
327 struct ice_pf *pf = vsi->back;
328 struct ice_hw *hw = &pf->hw;
329 u32 changed_flags = 0;
335 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
336 usleep_range(1000, 2000);
338 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
339 vsi->current_netdev_flags = vsi->netdev->flags;
341 INIT_LIST_HEAD(&vsi->tmp_sync_list);
342 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
344 if (ice_vsi_fltr_changed(vsi)) {
345 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
346 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
348 /* grab the netdev's addr_list_lock */
349 netif_addr_lock_bh(netdev);
350 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
351 ice_add_mac_to_unsync_list);
352 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
353 ice_add_mac_to_unsync_list);
354 /* our temp lists are populated. release lock */
355 netif_addr_unlock_bh(netdev);
358 /* Remove MAC addresses in the unsync list */
359 err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
360 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
362 netdev_err(netdev, "Failed to delete MAC filters\n");
363 /* if we failed because of alloc failures, just bail */
368 /* Add MAC addresses in the sync list */
369 err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
370 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
371 /* If filter is added successfully or already exists, do not go into
372 * 'if' condition and report it as error. Instead continue processing
373 * rest of the function.
375 if (err && err != -EEXIST) {
376 netdev_err(netdev, "Failed to add MAC filters\n");
377 /* If there is no more space for new umac filters, VSI
378 * should go into promiscuous mode. There should be some
379 * space reserved for promiscuous filters.
381 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
382 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
384 promisc_forced_on = true;
385 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
392 /* check for changes in promiscuous modes */
393 if (changed_flags & IFF_ALLMULTI) {
394 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
395 err = ice_set_promisc(vsi, ICE_MCAST_PROMISC_BITS);
397 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
401 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
402 err = ice_clear_promisc(vsi, ICE_MCAST_PROMISC_BITS);
404 vsi->current_netdev_flags |= IFF_ALLMULTI;
410 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
411 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
412 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
413 if (vsi->current_netdev_flags & IFF_PROMISC) {
414 /* Apply Rx filter rule to get traffic from wire */
415 if (!ice_is_dflt_vsi_in_use(vsi->port_info)) {
416 err = ice_set_dflt_vsi(vsi);
417 if (err && err != -EEXIST) {
418 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
420 vsi->current_netdev_flags &=
425 vlan_ops->dis_rx_filtering(vsi);
428 /* Clear Rx filter to remove traffic from wire */
429 if (ice_is_vsi_dflt_vsi(vsi)) {
430 err = ice_clear_dflt_vsi(vsi);
432 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
434 vsi->current_netdev_flags |=
438 if (vsi->netdev->features &
439 NETIF_F_HW_VLAN_CTAG_FILTER)
440 vlan_ops->ena_rx_filtering(vsi);
447 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
450 /* if something went wrong then set the changed flag so we try again */
451 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
452 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
454 clear_bit(ICE_CFG_BUSY, vsi->state);
459 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
460 * @pf: board private structure
462 static void ice_sync_fltr_subtask(struct ice_pf *pf)
466 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
469 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
471 ice_for_each_vsi(pf, v)
472 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
473 ice_vsi_sync_fltr(pf->vsi[v])) {
474 /* come back and try again later */
475 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
481 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
483 * @locked: is the rtnl_lock already held
485 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
490 ice_for_each_vsi(pf, v)
492 ice_dis_vsi(pf->vsi[v], locked);
494 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
495 pf->pf_agg_node[node].num_vsis = 0;
497 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
498 pf->vf_agg_node[node].num_vsis = 0;
502 * ice_clear_sw_switch_recipes - clear switch recipes
503 * @pf: board private structure
505 * Mark switch recipes as not created in sw structures. There are cases where
506 * rules (especially advanced rules) need to be restored, either re-read from
507 * hardware or added again. For example after the reset. 'recp_created' flag
508 * prevents from doing that and need to be cleared upfront.
510 static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
512 struct ice_sw_recipe *recp;
515 recp = pf->hw.switch_info->recp_list;
516 for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
517 recp[i].recp_created = false;
521 * ice_prepare_for_reset - prep for reset
522 * @pf: board private structure
523 * @reset_type: reset type requested
525 * Inform or close all dependent features in prep for reset.
528 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
530 struct ice_hw *hw = &pf->hw;
535 dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
537 /* already prepared for reset */
538 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
541 ice_unplug_aux_dev(pf);
543 /* Notify VFs of impending reset */
544 if (ice_check_sq_alive(hw, &hw->mailboxq))
545 ice_vc_notify_reset(pf);
547 /* Disable VFs until reset is completed */
548 mutex_lock(&pf->vfs.table_lock);
549 ice_for_each_vf(pf, bkt, vf)
550 ice_set_vf_state_qs_dis(vf);
551 mutex_unlock(&pf->vfs.table_lock);
553 if (ice_is_eswitch_mode_switchdev(pf)) {
554 if (reset_type != ICE_RESET_PFR)
555 ice_clear_sw_switch_recipes(pf);
558 /* release ADQ specific HW and SW resources */
559 vsi = ice_get_main_vsi(pf);
563 /* to be on safe side, reset orig_rss_size so that normal flow
564 * of deciding rss_size can take precedence
566 vsi->orig_rss_size = 0;
568 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
569 if (reset_type == ICE_RESET_PFR) {
570 vsi->old_ena_tc = vsi->all_enatc;
571 vsi->old_numtc = vsi->all_numtc;
573 ice_remove_q_channels(vsi, true);
575 /* for other reset type, do not support channel rebuild
576 * hence reset needed info
584 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
585 memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
590 /* clear SW filtering DB */
591 ice_clear_hw_tbls(hw);
592 /* disable the VSIs and their queues that are not already DOWN */
593 ice_pf_dis_all_vsi(pf, false);
595 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
596 ice_ptp_prepare_for_reset(pf);
598 if (ice_is_feature_supported(pf, ICE_F_GNSS))
602 ice_sched_clear_port(hw->port_info);
604 ice_shutdown_all_ctrlq(hw);
606 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
610 * ice_do_reset - Initiate one of many types of resets
611 * @pf: board private structure
612 * @reset_type: reset type requested before this function was called.
614 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
616 struct device *dev = ice_pf_to_dev(pf);
617 struct ice_hw *hw = &pf->hw;
619 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
621 ice_prepare_for_reset(pf, reset_type);
623 /* trigger the reset */
624 if (ice_reset(hw, reset_type)) {
625 dev_err(dev, "reset %d failed\n", reset_type);
626 set_bit(ICE_RESET_FAILED, pf->state);
627 clear_bit(ICE_RESET_OICR_RECV, pf->state);
628 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
629 clear_bit(ICE_PFR_REQ, pf->state);
630 clear_bit(ICE_CORER_REQ, pf->state);
631 clear_bit(ICE_GLOBR_REQ, pf->state);
632 wake_up(&pf->reset_wait_queue);
636 /* PFR is a bit of a special case because it doesn't result in an OICR
637 * interrupt. So for PFR, rebuild after the reset and clear the reset-
638 * associated state bits.
640 if (reset_type == ICE_RESET_PFR) {
642 ice_rebuild(pf, reset_type);
643 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
644 clear_bit(ICE_PFR_REQ, pf->state);
645 wake_up(&pf->reset_wait_queue);
646 ice_reset_all_vfs(pf);
651 * ice_reset_subtask - Set up for resetting the device and driver
652 * @pf: board private structure
654 static void ice_reset_subtask(struct ice_pf *pf)
656 enum ice_reset_req reset_type = ICE_RESET_INVAL;
658 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
659 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
660 * of reset is pending and sets bits in pf->state indicating the reset
661 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
662 * prepare for pending reset if not already (for PF software-initiated
663 * global resets the software should already be prepared for it as
664 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
665 * by firmware or software on other PFs, that bit is not set so prepare
666 * for the reset now), poll for reset done, rebuild and return.
668 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
669 /* Perform the largest reset requested */
670 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
671 reset_type = ICE_RESET_CORER;
672 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
673 reset_type = ICE_RESET_GLOBR;
674 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
675 reset_type = ICE_RESET_EMPR;
676 /* return if no valid reset type requested */
677 if (reset_type == ICE_RESET_INVAL)
679 ice_prepare_for_reset(pf, reset_type);
681 /* make sure we are ready to rebuild */
682 if (ice_check_reset(&pf->hw)) {
683 set_bit(ICE_RESET_FAILED, pf->state);
685 /* done with reset. start rebuild */
686 pf->hw.reset_ongoing = false;
687 ice_rebuild(pf, reset_type);
688 /* clear bit to resume normal operations, but
689 * ICE_NEEDS_RESTART bit is set in case rebuild failed
691 clear_bit(ICE_RESET_OICR_RECV, pf->state);
692 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
693 clear_bit(ICE_PFR_REQ, pf->state);
694 clear_bit(ICE_CORER_REQ, pf->state);
695 clear_bit(ICE_GLOBR_REQ, pf->state);
696 wake_up(&pf->reset_wait_queue);
697 ice_reset_all_vfs(pf);
703 /* No pending resets to finish processing. Check for new resets */
704 if (test_bit(ICE_PFR_REQ, pf->state))
705 reset_type = ICE_RESET_PFR;
706 if (test_bit(ICE_CORER_REQ, pf->state))
707 reset_type = ICE_RESET_CORER;
708 if (test_bit(ICE_GLOBR_REQ, pf->state))
709 reset_type = ICE_RESET_GLOBR;
710 /* If no valid reset type requested just return */
711 if (reset_type == ICE_RESET_INVAL)
714 /* reset if not already down or busy */
715 if (!test_bit(ICE_DOWN, pf->state) &&
716 !test_bit(ICE_CFG_BUSY, pf->state)) {
717 ice_do_reset(pf, reset_type);
722 * ice_print_topo_conflict - print topology conflict message
723 * @vsi: the VSI whose topology status is being checked
725 static void ice_print_topo_conflict(struct ice_vsi *vsi)
727 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
728 case ICE_AQ_LINK_TOPO_CONFLICT:
729 case ICE_AQ_LINK_MEDIA_CONFLICT:
730 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
731 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
732 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
733 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");
735 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
736 if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
737 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");
739 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");
747 * ice_print_link_msg - print link up or down message
748 * @vsi: the VSI whose link status is being queried
749 * @isup: boolean for if the link is now up or down
751 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
753 struct ice_aqc_get_phy_caps_data *caps;
754 const char *an_advertised;
765 if (vsi->current_isup == isup)
768 vsi->current_isup = isup;
771 netdev_info(vsi->netdev, "NIC Link is Down\n");
775 switch (vsi->port_info->phy.link_info.link_speed) {
776 case ICE_AQ_LINK_SPEED_100GB:
779 case ICE_AQ_LINK_SPEED_50GB:
782 case ICE_AQ_LINK_SPEED_40GB:
785 case ICE_AQ_LINK_SPEED_25GB:
788 case ICE_AQ_LINK_SPEED_20GB:
791 case ICE_AQ_LINK_SPEED_10GB:
794 case ICE_AQ_LINK_SPEED_5GB:
797 case ICE_AQ_LINK_SPEED_2500MB:
800 case ICE_AQ_LINK_SPEED_1000MB:
803 case ICE_AQ_LINK_SPEED_100MB:
811 switch (vsi->port_info->fc.current_mode) {
815 case ICE_FC_TX_PAUSE:
818 case ICE_FC_RX_PAUSE:
829 /* Get FEC mode based on negotiated link info */
830 switch (vsi->port_info->phy.link_info.fec_info) {
831 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
832 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
835 case ICE_AQ_LINK_25G_KR_FEC_EN:
836 fec = "FC-FEC/BASE-R";
843 /* check if autoneg completed, might be false due to not supported */
844 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
849 /* Get FEC mode requested based on PHY caps last SW configuration */
850 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
853 an_advertised = "Unknown";
857 status = ice_aq_get_phy_caps(vsi->port_info, false,
858 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
860 netdev_info(vsi->netdev, "Get phy capability failed.\n");
862 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
864 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
865 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
867 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
868 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
869 fec_req = "FC-FEC/BASE-R";
876 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",
877 speed, fec_req, fec, an_advertised, an, fc);
878 ice_print_topo_conflict(vsi);
882 * ice_vsi_link_event - update the VSI's netdev
883 * @vsi: the VSI on which the link event occurred
884 * @link_up: whether or not the VSI needs to be set up or down
886 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
891 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
894 if (vsi->type == ICE_VSI_PF) {
895 if (link_up == netif_carrier_ok(vsi->netdev))
899 netif_carrier_on(vsi->netdev);
900 netif_tx_wake_all_queues(vsi->netdev);
902 netif_carrier_off(vsi->netdev);
903 netif_tx_stop_all_queues(vsi->netdev);
909 * ice_set_dflt_mib - send a default config MIB to the FW
910 * @pf: private PF struct
912 * This function sends a default configuration MIB to the FW.
914 * If this function errors out at any point, the driver is still able to
915 * function. The main impact is that LFC may not operate as expected.
916 * Therefore an error state in this function should be treated with a DBG
917 * message and continue on with driver rebuild/reenable.
919 static void ice_set_dflt_mib(struct ice_pf *pf)
921 struct device *dev = ice_pf_to_dev(pf);
922 u8 mib_type, *buf, *lldpmib = NULL;
923 u16 len, typelen, offset = 0;
924 struct ice_lldp_org_tlv *tlv;
925 struct ice_hw *hw = &pf->hw;
928 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
929 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
931 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
936 /* Add ETS CFG TLV */
937 tlv = (struct ice_lldp_org_tlv *)lldpmib;
938 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
939 ICE_IEEE_ETS_TLV_LEN);
940 tlv->typelen = htons(typelen);
941 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
942 ICE_IEEE_SUBTYPE_ETS_CFG);
943 tlv->ouisubtype = htonl(ouisubtype);
948 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
949 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
950 * Octets 13 - 20 are TSA values - leave as zeros
953 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
955 tlv = (struct ice_lldp_org_tlv *)
956 ((char *)tlv + sizeof(tlv->typelen) + len);
958 /* Add ETS REC TLV */
960 tlv->typelen = htons(typelen);
962 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
963 ICE_IEEE_SUBTYPE_ETS_REC);
964 tlv->ouisubtype = htonl(ouisubtype);
966 /* First octet of buf is reserved
967 * Octets 1 - 4 map UP to TC - all UPs map to zero
968 * Octets 5 - 12 are BW values - set TC 0 to 100%.
969 * Octets 13 - 20 are TSA value - leave as zeros
973 tlv = (struct ice_lldp_org_tlv *)
974 ((char *)tlv + sizeof(tlv->typelen) + len);
976 /* Add PFC CFG TLV */
977 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
978 ICE_IEEE_PFC_TLV_LEN);
979 tlv->typelen = htons(typelen);
981 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
982 ICE_IEEE_SUBTYPE_PFC_CFG);
983 tlv->ouisubtype = htonl(ouisubtype);
985 /* Octet 1 left as all zeros - PFC disabled */
987 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
990 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
991 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
997 * ice_check_phy_fw_load - check if PHY FW load failed
998 * @pf: pointer to PF struct
999 * @link_cfg_err: bitmap from the link info structure
1001 * check if external PHY FW load failed and print an error message if it did
1003 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
1005 if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
1006 clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1010 if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
1013 if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
1014 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");
1015 set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
1020 * ice_check_module_power
1021 * @pf: pointer to PF struct
1022 * @link_cfg_err: bitmap from the link info structure
1024 * check module power level returned by a previous call to aq_get_link_info
1025 * and print error messages if module power level is not supported
1027 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
1029 /* if module power level is supported, clear the flag */
1030 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
1031 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
1032 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1036 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1037 * above block didn't clear this bit, there's nothing to do
1039 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1042 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1043 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1044 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1045 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1046 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1047 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1052 * ice_check_link_cfg_err - check if link configuration failed
1053 * @pf: pointer to the PF struct
1054 * @link_cfg_err: bitmap from the link info structure
1056 * print if any link configuration failure happens due to the value in the
1057 * link_cfg_err parameter in the link info structure
1059 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1061 ice_check_module_power(pf, link_cfg_err);
1062 ice_check_phy_fw_load(pf, link_cfg_err);
1066 * ice_link_event - process the link event
1067 * @pf: PF that the link event is associated with
1068 * @pi: port_info for the port that the link event is associated with
1069 * @link_up: true if the physical link is up and false if it is down
1070 * @link_speed: current link speed received from the link event
1072 * Returns 0 on success and negative on failure
1075 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1078 struct device *dev = ice_pf_to_dev(pf);
1079 struct ice_phy_info *phy_info;
1080 struct ice_vsi *vsi;
1085 phy_info = &pi->phy;
1086 phy_info->link_info_old = phy_info->link_info;
1088 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1089 old_link_speed = phy_info->link_info_old.link_speed;
1091 /* update the link info structures and re-enable link events,
1092 * don't bail on failure due to other book keeping needed
1094 status = ice_update_link_info(pi);
1096 dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1098 ice_aq_str(pi->hw->adminq.sq_last_status));
1100 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1102 /* Check if the link state is up after updating link info, and treat
1103 * this event as an UP event since the link is actually UP now.
1105 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1108 vsi = ice_get_main_vsi(pf);
1109 if (!vsi || !vsi->port_info)
1112 /* turn off PHY if media was removed */
1113 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1114 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1115 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1116 ice_set_link(vsi, false);
1119 /* if the old link up/down and speed is the same as the new */
1120 if (link_up == old_link && link_speed == old_link_speed)
1123 if (!ice_is_e810(&pf->hw))
1124 ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1126 if (ice_is_dcb_active(pf)) {
1127 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1128 ice_dcb_rebuild(pf);
1131 ice_set_dflt_mib(pf);
1133 ice_vsi_link_event(vsi, link_up);
1134 ice_print_link_msg(vsi, link_up);
1136 ice_vc_notify_link_state(pf);
1142 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1143 * @pf: board private structure
1145 static void ice_watchdog_subtask(struct ice_pf *pf)
1149 /* if interface is down do nothing */
1150 if (test_bit(ICE_DOWN, pf->state) ||
1151 test_bit(ICE_CFG_BUSY, pf->state))
1154 /* make sure we don't do these things too often */
1155 if (time_before(jiffies,
1156 pf->serv_tmr_prev + pf->serv_tmr_period))
1159 pf->serv_tmr_prev = jiffies;
1161 /* Update the stats for active netdevs so the network stack
1162 * can look at updated numbers whenever it cares to
1164 ice_update_pf_stats(pf);
1165 ice_for_each_vsi(pf, i)
1166 if (pf->vsi[i] && pf->vsi[i]->netdev)
1167 ice_update_vsi_stats(pf->vsi[i]);
1171 * ice_init_link_events - enable/initialize link events
1172 * @pi: pointer to the port_info instance
1174 * Returns -EIO on failure, 0 on success
1176 static int ice_init_link_events(struct ice_port_info *pi)
1180 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1181 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1182 ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1184 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1185 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1190 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1191 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1200 * ice_handle_link_event - handle link event via ARQ
1201 * @pf: PF that the link event is associated with
1202 * @event: event structure containing link status info
1205 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1207 struct ice_aqc_get_link_status_data *link_data;
1208 struct ice_port_info *port_info;
1211 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1212 port_info = pf->hw.port_info;
1216 status = ice_link_event(pf, port_info,
1217 !!(link_data->link_info & ICE_AQ_LINK_UP),
1218 le16_to_cpu(link_data->link_speed));
1220 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1226 enum ice_aq_task_state {
1227 ICE_AQ_TASK_WAITING = 0,
1228 ICE_AQ_TASK_COMPLETE,
1229 ICE_AQ_TASK_CANCELED,
1232 struct ice_aq_task {
1233 struct hlist_node entry;
1236 struct ice_rq_event_info *event;
1237 enum ice_aq_task_state state;
1241 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1242 * @pf: pointer to the PF private structure
1243 * @opcode: the opcode to wait for
1244 * @timeout: how long to wait, in jiffies
1245 * @event: storage for the event info
1247 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1248 * current thread will be put to sleep until the specified event occurs or
1249 * until the given timeout is reached.
1251 * To obtain only the descriptor contents, pass an event without an allocated
1252 * msg_buf. If the complete data buffer is desired, allocate the
1253 * event->msg_buf with enough space ahead of time.
1255 * Returns: zero on success, or a negative error code on failure.
1257 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1258 struct ice_rq_event_info *event)
1260 struct device *dev = ice_pf_to_dev(pf);
1261 struct ice_aq_task *task;
1262 unsigned long start;
1266 task = kzalloc(sizeof(*task), GFP_KERNEL);
1270 INIT_HLIST_NODE(&task->entry);
1271 task->opcode = opcode;
1272 task->event = event;
1273 task->state = ICE_AQ_TASK_WAITING;
1275 spin_lock_bh(&pf->aq_wait_lock);
1276 hlist_add_head(&task->entry, &pf->aq_wait_list);
1277 spin_unlock_bh(&pf->aq_wait_lock);
1281 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1283 switch (task->state) {
1284 case ICE_AQ_TASK_WAITING:
1285 err = ret < 0 ? ret : -ETIMEDOUT;
1287 case ICE_AQ_TASK_CANCELED:
1288 err = ret < 0 ? ret : -ECANCELED;
1290 case ICE_AQ_TASK_COMPLETE:
1291 err = ret < 0 ? ret : 0;
1294 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1299 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1300 jiffies_to_msecs(jiffies - start),
1301 jiffies_to_msecs(timeout),
1304 spin_lock_bh(&pf->aq_wait_lock);
1305 hlist_del(&task->entry);
1306 spin_unlock_bh(&pf->aq_wait_lock);
1313 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1314 * @pf: pointer to the PF private structure
1315 * @opcode: the opcode of the event
1316 * @event: the event to check
1318 * Loops over the current list of pending threads waiting for an AdminQ event.
1319 * For each matching task, copy the contents of the event into the task
1320 * structure and wake up the thread.
1322 * If multiple threads wait for the same opcode, they will all be woken up.
1324 * Note that event->msg_buf will only be duplicated if the event has a buffer
1325 * with enough space already allocated. Otherwise, only the descriptor and
1326 * message length will be copied.
1328 * Returns: true if an event was found, false otherwise
1330 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1331 struct ice_rq_event_info *event)
1333 struct ice_aq_task *task;
1336 spin_lock_bh(&pf->aq_wait_lock);
1337 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1338 if (task->state || task->opcode != opcode)
1341 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1342 task->event->msg_len = event->msg_len;
1344 /* Only copy the data buffer if a destination was set */
1345 if (task->event->msg_buf &&
1346 task->event->buf_len > event->buf_len) {
1347 memcpy(task->event->msg_buf, event->msg_buf,
1349 task->event->buf_len = event->buf_len;
1352 task->state = ICE_AQ_TASK_COMPLETE;
1355 spin_unlock_bh(&pf->aq_wait_lock);
1358 wake_up(&pf->aq_wait_queue);
1362 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1363 * @pf: the PF private structure
1365 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1366 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1368 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1370 struct ice_aq_task *task;
1372 spin_lock_bh(&pf->aq_wait_lock);
1373 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1374 task->state = ICE_AQ_TASK_CANCELED;
1375 spin_unlock_bh(&pf->aq_wait_lock);
1377 wake_up(&pf->aq_wait_queue);
1381 * __ice_clean_ctrlq - helper function to clean controlq rings
1382 * @pf: ptr to struct ice_pf
1383 * @q_type: specific Control queue type
1385 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1387 struct device *dev = ice_pf_to_dev(pf);
1388 struct ice_rq_event_info event;
1389 struct ice_hw *hw = &pf->hw;
1390 struct ice_ctl_q_info *cq;
1395 /* Do not clean control queue if/when PF reset fails */
1396 if (test_bit(ICE_RESET_FAILED, pf->state))
1400 case ICE_CTL_Q_ADMIN:
1408 case ICE_CTL_Q_MAILBOX:
1411 /* we are going to try to detect a malicious VF, so set the
1412 * state to begin detection
1414 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1417 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1421 /* check for error indications - PF_xx_AxQLEN register layout for
1422 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1424 val = rd32(hw, cq->rq.len);
1425 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1426 PF_FW_ARQLEN_ARQCRIT_M)) {
1428 if (val & PF_FW_ARQLEN_ARQVFE_M)
1429 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1431 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1432 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1435 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1436 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1438 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1439 PF_FW_ARQLEN_ARQCRIT_M);
1441 wr32(hw, cq->rq.len, val);
1444 val = rd32(hw, cq->sq.len);
1445 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1446 PF_FW_ATQLEN_ATQCRIT_M)) {
1448 if (val & PF_FW_ATQLEN_ATQVFE_M)
1449 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1451 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1452 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1455 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1456 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1458 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1459 PF_FW_ATQLEN_ATQCRIT_M);
1461 wr32(hw, cq->sq.len, val);
1464 event.buf_len = cq->rq_buf_size;
1465 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1473 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1474 if (ret == -EALREADY)
1477 dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1482 opcode = le16_to_cpu(event.desc.opcode);
1484 /* Notify any thread that might be waiting for this event */
1485 ice_aq_check_events(pf, opcode, &event);
1488 case ice_aqc_opc_get_link_status:
1489 if (ice_handle_link_event(pf, &event))
1490 dev_err(dev, "Could not handle link event\n");
1492 case ice_aqc_opc_event_lan_overflow:
1493 ice_vf_lan_overflow_event(pf, &event);
1495 case ice_mbx_opc_send_msg_to_pf:
1496 if (!ice_is_malicious_vf(pf, &event, i, pending))
1497 ice_vc_process_vf_msg(pf, &event);
1499 case ice_aqc_opc_fw_logging:
1500 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1502 case ice_aqc_opc_lldp_set_mib_change:
1503 ice_dcb_process_lldp_set_mib_change(pf, &event);
1506 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1510 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1512 kfree(event.msg_buf);
1514 return pending && (i == ICE_DFLT_IRQ_WORK);
1518 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1519 * @hw: pointer to hardware info
1520 * @cq: control queue information
1522 * returns true if there are pending messages in a queue, false if there aren't
1524 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1528 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1529 return cq->rq.next_to_clean != ntu;
1533 * ice_clean_adminq_subtask - clean the AdminQ rings
1534 * @pf: board private structure
1536 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1538 struct ice_hw *hw = &pf->hw;
1540 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1543 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1546 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1548 /* There might be a situation where new messages arrive to a control
1549 * queue between processing the last message and clearing the
1550 * EVENT_PENDING bit. So before exiting, check queue head again (using
1551 * ice_ctrlq_pending) and process new messages if any.
1553 if (ice_ctrlq_pending(hw, &hw->adminq))
1554 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1560 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1561 * @pf: board private structure
1563 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1565 struct ice_hw *hw = &pf->hw;
1567 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1570 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1573 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1575 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1576 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1582 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1583 * @pf: board private structure
1585 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1587 struct ice_hw *hw = &pf->hw;
1589 /* Nothing to do here if sideband queue is not supported */
1590 if (!ice_is_sbq_supported(hw)) {
1591 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1595 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1598 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1601 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1603 if (ice_ctrlq_pending(hw, &hw->sbq))
1604 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1610 * ice_service_task_schedule - schedule the service task to wake up
1611 * @pf: board private structure
1613 * If not already scheduled, this puts the task into the work queue.
1615 void ice_service_task_schedule(struct ice_pf *pf)
1617 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1618 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1619 !test_bit(ICE_NEEDS_RESTART, pf->state))
1620 queue_work(ice_wq, &pf->serv_task);
1624 * ice_service_task_complete - finish up the service task
1625 * @pf: board private structure
1627 static void ice_service_task_complete(struct ice_pf *pf)
1629 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1631 /* force memory (pf->state) to sync before next service task */
1632 smp_mb__before_atomic();
1633 clear_bit(ICE_SERVICE_SCHED, pf->state);
1637 * ice_service_task_stop - stop service task and cancel works
1638 * @pf: board private structure
1640 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1643 static int ice_service_task_stop(struct ice_pf *pf)
1647 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1649 if (pf->serv_tmr.function)
1650 del_timer_sync(&pf->serv_tmr);
1651 if (pf->serv_task.func)
1652 cancel_work_sync(&pf->serv_task);
1654 clear_bit(ICE_SERVICE_SCHED, pf->state);
1659 * ice_service_task_restart - restart service task and schedule works
1660 * @pf: board private structure
1662 * This function is needed for suspend and resume works (e.g WoL scenario)
1664 static void ice_service_task_restart(struct ice_pf *pf)
1666 clear_bit(ICE_SERVICE_DIS, pf->state);
1667 ice_service_task_schedule(pf);
1671 * ice_service_timer - timer callback to schedule service task
1672 * @t: pointer to timer_list
1674 static void ice_service_timer(struct timer_list *t)
1676 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1678 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1679 ice_service_task_schedule(pf);
1683 * ice_handle_mdd_event - handle malicious driver detect event
1684 * @pf: pointer to the PF structure
1686 * Called from service task. OICR interrupt handler indicates MDD event.
1687 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1688 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1689 * disable the queue, the PF can be configured to reset the VF using ethtool
1690 * private flag mdd-auto-reset-vf.
1692 static void ice_handle_mdd_event(struct ice_pf *pf)
1694 struct device *dev = ice_pf_to_dev(pf);
1695 struct ice_hw *hw = &pf->hw;
1700 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1701 /* Since the VF MDD event logging is rate limited, check if
1702 * there are pending MDD events.
1704 ice_print_vfs_mdd_events(pf);
1708 /* find what triggered an MDD event */
1709 reg = rd32(hw, GL_MDET_TX_PQM);
1710 if (reg & GL_MDET_TX_PQM_VALID_M) {
1711 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1712 GL_MDET_TX_PQM_PF_NUM_S;
1713 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1714 GL_MDET_TX_PQM_VF_NUM_S;
1715 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1716 GL_MDET_TX_PQM_MAL_TYPE_S;
1717 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1718 GL_MDET_TX_PQM_QNUM_S);
1720 if (netif_msg_tx_err(pf))
1721 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1722 event, queue, pf_num, vf_num);
1723 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1726 reg = rd32(hw, GL_MDET_TX_TCLAN);
1727 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1728 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1729 GL_MDET_TX_TCLAN_PF_NUM_S;
1730 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1731 GL_MDET_TX_TCLAN_VF_NUM_S;
1732 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1733 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1734 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1735 GL_MDET_TX_TCLAN_QNUM_S);
1737 if (netif_msg_tx_err(pf))
1738 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1739 event, queue, pf_num, vf_num);
1740 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1743 reg = rd32(hw, GL_MDET_RX);
1744 if (reg & GL_MDET_RX_VALID_M) {
1745 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1746 GL_MDET_RX_PF_NUM_S;
1747 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1748 GL_MDET_RX_VF_NUM_S;
1749 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1750 GL_MDET_RX_MAL_TYPE_S;
1751 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1754 if (netif_msg_rx_err(pf))
1755 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1756 event, queue, pf_num, vf_num);
1757 wr32(hw, GL_MDET_RX, 0xffffffff);
1760 /* check to see if this PF caused an MDD event */
1761 reg = rd32(hw, PF_MDET_TX_PQM);
1762 if (reg & PF_MDET_TX_PQM_VALID_M) {
1763 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1764 if (netif_msg_tx_err(pf))
1765 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1768 reg = rd32(hw, PF_MDET_TX_TCLAN);
1769 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1770 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1771 if (netif_msg_tx_err(pf))
1772 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1775 reg = rd32(hw, PF_MDET_RX);
1776 if (reg & PF_MDET_RX_VALID_M) {
1777 wr32(hw, PF_MDET_RX, 0xFFFF);
1778 if (netif_msg_rx_err(pf))
1779 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1782 /* Check to see if one of the VFs caused an MDD event, and then
1783 * increment counters and set print pending
1785 mutex_lock(&pf->vfs.table_lock);
1786 ice_for_each_vf(pf, bkt, vf) {
1787 reg = rd32(hw, VP_MDET_TX_PQM(vf->vf_id));
1788 if (reg & VP_MDET_TX_PQM_VALID_M) {
1789 wr32(hw, VP_MDET_TX_PQM(vf->vf_id), 0xFFFF);
1790 vf->mdd_tx_events.count++;
1791 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1792 if (netif_msg_tx_err(pf))
1793 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1797 reg = rd32(hw, VP_MDET_TX_TCLAN(vf->vf_id));
1798 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1799 wr32(hw, VP_MDET_TX_TCLAN(vf->vf_id), 0xFFFF);
1800 vf->mdd_tx_events.count++;
1801 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1802 if (netif_msg_tx_err(pf))
1803 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1807 reg = rd32(hw, VP_MDET_TX_TDPU(vf->vf_id));
1808 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1809 wr32(hw, VP_MDET_TX_TDPU(vf->vf_id), 0xFFFF);
1810 vf->mdd_tx_events.count++;
1811 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1812 if (netif_msg_tx_err(pf))
1813 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1817 reg = rd32(hw, VP_MDET_RX(vf->vf_id));
1818 if (reg & VP_MDET_RX_VALID_M) {
1819 wr32(hw, VP_MDET_RX(vf->vf_id), 0xFFFF);
1820 vf->mdd_rx_events.count++;
1821 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1822 if (netif_msg_rx_err(pf))
1823 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1826 /* Since the queue is disabled on VF Rx MDD events, the
1827 * PF can be configured to reset the VF through ethtool
1828 * private flag mdd-auto-reset-vf.
1830 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1831 /* VF MDD event counters will be cleared by
1832 * reset, so print the event prior to reset.
1834 ice_print_vf_rx_mdd_event(vf);
1835 ice_reset_vf(vf, ICE_VF_RESET_LOCK);
1839 mutex_unlock(&pf->vfs.table_lock);
1841 ice_print_vfs_mdd_events(pf);
1845 * ice_force_phys_link_state - Force the physical link state
1846 * @vsi: VSI to force the physical link state to up/down
1847 * @link_up: true/false indicates to set the physical link to up/down
1849 * Force the physical link state by getting the current PHY capabilities from
1850 * hardware and setting the PHY config based on the determined capabilities. If
1851 * link changes a link event will be triggered because both the Enable Automatic
1852 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1854 * Returns 0 on success, negative on failure
1856 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1858 struct ice_aqc_get_phy_caps_data *pcaps;
1859 struct ice_aqc_set_phy_cfg_data *cfg;
1860 struct ice_port_info *pi;
1864 if (!vsi || !vsi->port_info || !vsi->back)
1866 if (vsi->type != ICE_VSI_PF)
1869 dev = ice_pf_to_dev(vsi->back);
1871 pi = vsi->port_info;
1873 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1877 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1880 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1881 vsi->vsi_num, retcode);
1886 /* No change in link */
1887 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1888 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1891 /* Use the current user PHY configuration. The current user PHY
1892 * configuration is initialized during probe from PHY capabilities
1893 * software mode, and updated on set PHY configuration.
1895 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1901 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1903 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1905 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1907 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1909 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1910 vsi->vsi_num, retcode);
1921 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1922 * @pi: port info structure
1924 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1926 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1928 struct ice_aqc_get_phy_caps_data *pcaps;
1929 struct ice_pf *pf = pi->hw->back;
1932 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1936 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1940 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1944 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1945 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1953 * ice_init_link_dflt_override - Initialize link default override
1954 * @pi: port info structure
1956 * Initialize link default override and PHY total port shutdown during probe
1958 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1960 struct ice_link_default_override_tlv *ldo;
1961 struct ice_pf *pf = pi->hw->back;
1963 ldo = &pf->link_dflt_override;
1964 if (ice_get_link_default_override(ldo, pi))
1967 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1970 /* Enable Total Port Shutdown (override/replace link-down-on-close
1971 * ethtool private flag) for ports with Port Disable bit set.
1973 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1974 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1978 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1979 * @pi: port info structure
1981 * If default override is enabled, initialize the user PHY cfg speed and FEC
1982 * settings using the default override mask from the NVM.
1984 * The PHY should only be configured with the default override settings the
1985 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1986 * is used to indicate that the user PHY cfg default override is initialized
1987 * and the PHY has not been configured with the default override settings. The
1988 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1991 * This function should be called only if the FW doesn't support default
1992 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1994 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1996 struct ice_link_default_override_tlv *ldo;
1997 struct ice_aqc_set_phy_cfg_data *cfg;
1998 struct ice_phy_info *phy = &pi->phy;
1999 struct ice_pf *pf = pi->hw->back;
2001 ldo = &pf->link_dflt_override;
2003 /* If link default override is enabled, use to mask NVM PHY capabilities
2004 * for speed and FEC default configuration.
2006 cfg = &phy->curr_user_phy_cfg;
2008 if (ldo->phy_type_low || ldo->phy_type_high) {
2009 cfg->phy_type_low = pf->nvm_phy_type_lo &
2010 cpu_to_le64(ldo->phy_type_low);
2011 cfg->phy_type_high = pf->nvm_phy_type_hi &
2012 cpu_to_le64(ldo->phy_type_high);
2014 cfg->link_fec_opt = ldo->fec_options;
2015 phy->curr_user_fec_req = ICE_FEC_AUTO;
2017 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
2021 * ice_init_phy_user_cfg - Initialize the PHY user configuration
2022 * @pi: port info structure
2024 * Initialize the current user PHY configuration, speed, FEC, and FC requested
2025 * mode to default. The PHY defaults are from get PHY capabilities topology
2026 * with media so call when media is first available. An error is returned if
2027 * called when media is not available. The PHY initialization completed state is
2030 * These configurations are used when setting PHY
2031 * configuration. The user PHY configuration is updated on set PHY
2032 * configuration. Returns 0 on success, negative on failure
2034 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2036 struct ice_aqc_get_phy_caps_data *pcaps;
2037 struct ice_phy_info *phy = &pi->phy;
2038 struct ice_pf *pf = pi->hw->back;
2041 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2044 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2048 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2049 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2052 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2055 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2059 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2061 /* check if lenient mode is supported and enabled */
2062 if (ice_fw_supports_link_override(pi->hw) &&
2063 !(pcaps->module_compliance_enforcement &
2064 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2065 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2067 /* if the FW supports default PHY configuration mode, then the driver
2068 * does not have to apply link override settings. If not,
2069 * initialize user PHY configuration with link override values
2071 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2072 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2073 ice_init_phy_cfg_dflt_override(pi);
2078 /* if link default override is not enabled, set user flow control and
2079 * FEC settings based on what get_phy_caps returned
2081 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2082 pcaps->link_fec_options);
2083 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2086 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2087 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2094 * ice_configure_phy - configure PHY
2097 * Set the PHY configuration. If the current PHY configuration is the same as
2098 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2099 * configure the based get PHY capabilities for topology with media.
2101 static int ice_configure_phy(struct ice_vsi *vsi)
2103 struct device *dev = ice_pf_to_dev(vsi->back);
2104 struct ice_port_info *pi = vsi->port_info;
2105 struct ice_aqc_get_phy_caps_data *pcaps;
2106 struct ice_aqc_set_phy_cfg_data *cfg;
2107 struct ice_phy_info *phy = &pi->phy;
2108 struct ice_pf *pf = vsi->back;
2111 /* Ensure we have media as we cannot configure a medialess port */
2112 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2115 ice_print_topo_conflict(vsi);
2117 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2118 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2121 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2122 return ice_force_phys_link_state(vsi, true);
2124 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2128 /* Get current PHY config */
2129 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2132 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2137 /* If PHY enable link is configured and configuration has not changed,
2138 * there's nothing to do
2140 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2141 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2144 /* Use PHY topology as baseline for configuration */
2145 memset(pcaps, 0, sizeof(*pcaps));
2146 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2147 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2150 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2153 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2158 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2164 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2166 /* Speed - If default override pending, use curr_user_phy_cfg set in
2167 * ice_init_phy_user_cfg_ldo.
2169 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2170 vsi->back->state)) {
2171 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2172 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2174 u64 phy_low = 0, phy_high = 0;
2176 ice_update_phy_type(&phy_low, &phy_high,
2177 pi->phy.curr_user_speed_req);
2178 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2179 cfg->phy_type_high = pcaps->phy_type_high &
2180 cpu_to_le64(phy_high);
2183 /* Can't provide what was requested; use PHY capabilities */
2184 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2185 cfg->phy_type_low = pcaps->phy_type_low;
2186 cfg->phy_type_high = pcaps->phy_type_high;
2190 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2192 /* Can't provide what was requested; use PHY capabilities */
2193 if (cfg->link_fec_opt !=
2194 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2195 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2196 cfg->link_fec_opt = pcaps->link_fec_options;
2199 /* Flow Control - always supported; no need to check against
2202 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2204 /* Enable link and link update */
2205 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2207 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2209 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2219 * ice_check_media_subtask - Check for media
2220 * @pf: pointer to PF struct
2222 * If media is available, then initialize PHY user configuration if it is not
2223 * been, and configure the PHY if the interface is up.
2225 static void ice_check_media_subtask(struct ice_pf *pf)
2227 struct ice_port_info *pi;
2228 struct ice_vsi *vsi;
2231 /* No need to check for media if it's already present */
2232 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2235 vsi = ice_get_main_vsi(pf);
2239 /* Refresh link info and check if media is present */
2240 pi = vsi->port_info;
2241 err = ice_update_link_info(pi);
2245 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2247 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2248 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2249 ice_init_phy_user_cfg(pi);
2251 /* PHY settings are reset on media insertion, reconfigure
2252 * PHY to preserve settings.
2254 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2255 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2258 err = ice_configure_phy(vsi);
2260 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2262 /* A Link Status Event will be generated; the event handler
2263 * will complete bringing the interface up
2269 * ice_service_task - manage and run subtasks
2270 * @work: pointer to work_struct contained by the PF struct
2272 static void ice_service_task(struct work_struct *work)
2274 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2275 unsigned long start_time = jiffies;
2279 /* process reset requests first */
2280 ice_reset_subtask(pf);
2282 /* bail if a reset/recovery cycle is pending or rebuild failed */
2283 if (ice_is_reset_in_progress(pf->state) ||
2284 test_bit(ICE_SUSPENDED, pf->state) ||
2285 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2286 ice_service_task_complete(pf);
2290 if (test_and_clear_bit(ICE_AUX_ERR_PENDING, pf->state)) {
2291 struct iidc_event *event;
2293 event = kzalloc(sizeof(*event), GFP_KERNEL);
2295 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2296 /* report the entire OICR value to AUX driver */
2297 swap(event->reg, pf->oicr_err_reg);
2298 ice_send_event_to_aux(pf, event);
2303 if (test_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) {
2304 /* Plug aux device per request */
2305 ice_plug_aux_dev(pf);
2307 /* Mark plugging as done but check whether unplug was
2308 * requested during ice_plug_aux_dev() call
2309 * (e.g. from ice_clear_rdma_cap()) and if so then
2312 if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2313 ice_unplug_aux_dev(pf);
2316 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2317 struct iidc_event *event;
2319 event = kzalloc(sizeof(*event), GFP_KERNEL);
2321 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2322 ice_send_event_to_aux(pf, event);
2327 ice_clean_adminq_subtask(pf);
2328 ice_check_media_subtask(pf);
2329 ice_check_for_hang_subtask(pf);
2330 ice_sync_fltr_subtask(pf);
2331 ice_handle_mdd_event(pf);
2332 ice_watchdog_subtask(pf);
2334 if (ice_is_safe_mode(pf)) {
2335 ice_service_task_complete(pf);
2339 ice_process_vflr_event(pf);
2340 ice_clean_mailboxq_subtask(pf);
2341 ice_clean_sbq_subtask(pf);
2342 ice_sync_arfs_fltrs(pf);
2343 ice_flush_fdir_ctx(pf);
2345 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2346 ice_service_task_complete(pf);
2348 /* If the tasks have taken longer than one service timer period
2349 * or there is more work to be done, reset the service timer to
2350 * schedule the service task now.
2352 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2353 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2354 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2355 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2356 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2357 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2358 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2359 mod_timer(&pf->serv_tmr, jiffies);
2363 * ice_set_ctrlq_len - helper function to set controlq length
2364 * @hw: pointer to the HW instance
2366 static void ice_set_ctrlq_len(struct ice_hw *hw)
2368 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2369 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2370 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2371 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2372 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2373 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2374 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2375 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2376 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2377 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2378 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2379 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2383 * ice_schedule_reset - schedule a reset
2384 * @pf: board private structure
2385 * @reset: reset being requested
2387 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2389 struct device *dev = ice_pf_to_dev(pf);
2391 /* bail out if earlier reset has failed */
2392 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2393 dev_dbg(dev, "earlier reset has failed\n");
2396 /* bail if reset/recovery already in progress */
2397 if (ice_is_reset_in_progress(pf->state)) {
2398 dev_dbg(dev, "Reset already in progress\n");
2402 ice_unplug_aux_dev(pf);
2406 set_bit(ICE_PFR_REQ, pf->state);
2408 case ICE_RESET_CORER:
2409 set_bit(ICE_CORER_REQ, pf->state);
2411 case ICE_RESET_GLOBR:
2412 set_bit(ICE_GLOBR_REQ, pf->state);
2418 ice_service_task_schedule(pf);
2423 * ice_irq_affinity_notify - Callback for affinity changes
2424 * @notify: context as to what irq was changed
2425 * @mask: the new affinity mask
2427 * This is a callback function used by the irq_set_affinity_notifier function
2428 * so that we may register to receive changes to the irq affinity masks.
2431 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2432 const cpumask_t *mask)
2434 struct ice_q_vector *q_vector =
2435 container_of(notify, struct ice_q_vector, affinity_notify);
2437 cpumask_copy(&q_vector->affinity_mask, mask);
2441 * ice_irq_affinity_release - Callback for affinity notifier release
2442 * @ref: internal core kernel usage
2444 * This is a callback function used by the irq_set_affinity_notifier function
2445 * to inform the current notification subscriber that they will no longer
2446 * receive notifications.
2448 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2451 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2452 * @vsi: the VSI being configured
2454 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2456 struct ice_hw *hw = &vsi->back->hw;
2459 ice_for_each_q_vector(vsi, i)
2460 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2467 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2468 * @vsi: the VSI being configured
2469 * @basename: name for the vector
2471 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2473 int q_vectors = vsi->num_q_vectors;
2474 struct ice_pf *pf = vsi->back;
2475 int base = vsi->base_vector;
2482 dev = ice_pf_to_dev(pf);
2483 for (vector = 0; vector < q_vectors; vector++) {
2484 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2486 irq_num = pf->msix_entries[base + vector].vector;
2488 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2489 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2490 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2492 } else if (q_vector->rx.rx_ring) {
2493 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2494 "%s-%s-%d", basename, "rx", rx_int_idx++);
2495 } else if (q_vector->tx.tx_ring) {
2496 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2497 "%s-%s-%d", basename, "tx", tx_int_idx++);
2499 /* skip this unused q_vector */
2502 if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2503 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2504 IRQF_SHARED, q_vector->name,
2507 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2508 0, q_vector->name, q_vector);
2510 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2515 /* register for affinity change notifications */
2516 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2517 struct irq_affinity_notify *affinity_notify;
2519 affinity_notify = &q_vector->affinity_notify;
2520 affinity_notify->notify = ice_irq_affinity_notify;
2521 affinity_notify->release = ice_irq_affinity_release;
2522 irq_set_affinity_notifier(irq_num, affinity_notify);
2525 /* assign the mask for this irq */
2526 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2529 err = ice_set_cpu_rx_rmap(vsi);
2531 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2532 vsi->vsi_num, ERR_PTR(err));
2536 vsi->irqs_ready = true;
2542 irq_num = pf->msix_entries[base + vector].vector;
2543 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2544 irq_set_affinity_notifier(irq_num, NULL);
2545 irq_set_affinity_hint(irq_num, NULL);
2546 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2552 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2553 * @vsi: VSI to setup Tx rings used by XDP
2555 * Return 0 on success and negative value on error
2557 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2559 struct device *dev = ice_pf_to_dev(vsi->back);
2560 struct ice_tx_desc *tx_desc;
2563 ice_for_each_xdp_txq(vsi, i) {
2564 u16 xdp_q_idx = vsi->alloc_txq + i;
2565 struct ice_tx_ring *xdp_ring;
2567 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2570 goto free_xdp_rings;
2572 xdp_ring->q_index = xdp_q_idx;
2573 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2574 xdp_ring->vsi = vsi;
2575 xdp_ring->netdev = NULL;
2576 xdp_ring->dev = dev;
2577 xdp_ring->count = vsi->num_tx_desc;
2578 xdp_ring->next_dd = ICE_RING_QUARTER(xdp_ring) - 1;
2579 xdp_ring->next_rs = ICE_RING_QUARTER(xdp_ring) - 1;
2580 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2581 if (ice_setup_tx_ring(xdp_ring))
2582 goto free_xdp_rings;
2583 ice_set_ring_xdp(xdp_ring);
2584 xdp_ring->xsk_pool = ice_tx_xsk_pool(xdp_ring);
2585 spin_lock_init(&xdp_ring->tx_lock);
2586 for (j = 0; j < xdp_ring->count; j++) {
2587 tx_desc = ICE_TX_DESC(xdp_ring, j);
2588 tx_desc->cmd_type_offset_bsz = 0;
2592 ice_for_each_rxq(vsi, i) {
2593 if (static_key_enabled(&ice_xdp_locking_key))
2594 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2596 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i];
2603 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2604 ice_free_tx_ring(vsi->xdp_rings[i]);
2609 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2610 * @vsi: VSI to set the bpf prog on
2611 * @prog: the bpf prog pointer
2613 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2615 struct bpf_prog *old_prog;
2618 old_prog = xchg(&vsi->xdp_prog, prog);
2620 bpf_prog_put(old_prog);
2622 ice_for_each_rxq(vsi, i)
2623 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2627 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2628 * @vsi: VSI to bring up Tx rings used by XDP
2629 * @prog: bpf program that will be assigned to VSI
2631 * Return 0 on success and negative value on error
2633 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2635 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2636 int xdp_rings_rem = vsi->num_xdp_txq;
2637 struct ice_pf *pf = vsi->back;
2638 struct ice_qs_cfg xdp_qs_cfg = {
2639 .qs_mutex = &pf->avail_q_mutex,
2640 .pf_map = pf->avail_txqs,
2641 .pf_map_size = pf->max_pf_txqs,
2642 .q_count = vsi->num_xdp_txq,
2643 .scatter_count = ICE_MAX_SCATTER_TXQS,
2644 .vsi_map = vsi->txq_map,
2645 .vsi_map_offset = vsi->alloc_txq,
2646 .mapping_mode = ICE_VSI_MAP_CONTIG
2652 dev = ice_pf_to_dev(pf);
2653 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2654 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2655 if (!vsi->xdp_rings)
2658 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2659 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2662 if (static_key_enabled(&ice_xdp_locking_key))
2663 netdev_warn(vsi->netdev,
2664 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2666 if (ice_xdp_alloc_setup_rings(vsi))
2667 goto clear_xdp_rings;
2669 /* follow the logic from ice_vsi_map_rings_to_vectors */
2670 ice_for_each_q_vector(vsi, v_idx) {
2671 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2672 int xdp_rings_per_v, q_id, q_base;
2674 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2675 vsi->num_q_vectors - v_idx);
2676 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2678 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2679 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2681 xdp_ring->q_vector = q_vector;
2682 xdp_ring->next = q_vector->tx.tx_ring;
2683 q_vector->tx.tx_ring = xdp_ring;
2685 xdp_rings_rem -= xdp_rings_per_v;
2688 /* omit the scheduler update if in reset path; XDP queues will be
2689 * taken into account at the end of ice_vsi_rebuild, where
2690 * ice_cfg_vsi_lan is being called
2692 if (ice_is_reset_in_progress(pf->state))
2695 /* tell the Tx scheduler that right now we have
2698 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2699 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2701 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2704 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2706 goto clear_xdp_rings;
2709 /* assign the prog only when it's not already present on VSI;
2710 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2711 * VSI rebuild that happens under ethtool -L can expose us to
2712 * the bpf_prog refcount issues as we would be swapping same
2713 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2714 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2715 * this is not harmful as dev_xdp_install bumps the refcount
2716 * before calling the op exposed by the driver;
2718 if (!ice_is_xdp_ena_vsi(vsi))
2719 ice_vsi_assign_bpf_prog(vsi, prog);
2723 ice_for_each_xdp_txq(vsi, i)
2724 if (vsi->xdp_rings[i]) {
2725 kfree_rcu(vsi->xdp_rings[i], rcu);
2726 vsi->xdp_rings[i] = NULL;
2730 mutex_lock(&pf->avail_q_mutex);
2731 ice_for_each_xdp_txq(vsi, i) {
2732 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2733 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2735 mutex_unlock(&pf->avail_q_mutex);
2737 devm_kfree(dev, vsi->xdp_rings);
2742 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2743 * @vsi: VSI to remove XDP rings
2745 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2748 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2750 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2751 struct ice_pf *pf = vsi->back;
2754 /* q_vectors are freed in reset path so there's no point in detaching
2755 * rings; in case of rebuild being triggered not from reset bits
2756 * in pf->state won't be set, so additionally check first q_vector
2759 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2762 ice_for_each_q_vector(vsi, v_idx) {
2763 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2764 struct ice_tx_ring *ring;
2766 ice_for_each_tx_ring(ring, q_vector->tx)
2767 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2770 /* restore the value of last node prior to XDP setup */
2771 q_vector->tx.tx_ring = ring;
2775 mutex_lock(&pf->avail_q_mutex);
2776 ice_for_each_xdp_txq(vsi, i) {
2777 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2778 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2780 mutex_unlock(&pf->avail_q_mutex);
2782 ice_for_each_xdp_txq(vsi, i)
2783 if (vsi->xdp_rings[i]) {
2784 if (vsi->xdp_rings[i]->desc) {
2786 ice_free_tx_ring(vsi->xdp_rings[i]);
2788 kfree_rcu(vsi->xdp_rings[i], rcu);
2789 vsi->xdp_rings[i] = NULL;
2792 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2793 vsi->xdp_rings = NULL;
2795 if (static_key_enabled(&ice_xdp_locking_key))
2796 static_branch_dec(&ice_xdp_locking_key);
2798 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2801 ice_vsi_assign_bpf_prog(vsi, NULL);
2803 /* notify Tx scheduler that we destroyed XDP queues and bring
2804 * back the old number of child nodes
2806 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2807 max_txqs[i] = vsi->num_txq;
2809 /* change number of XDP Tx queues to 0 */
2810 vsi->num_xdp_txq = 0;
2812 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2817 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2818 * @vsi: VSI to schedule napi on
2820 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2824 ice_for_each_rxq(vsi, i) {
2825 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2827 if (rx_ring->xsk_pool)
2828 napi_schedule(&rx_ring->q_vector->napi);
2833 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2834 * @vsi: VSI to determine the count of XDP Tx qs
2836 * returns 0 if Tx qs count is higher than at least half of CPU count,
2839 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2841 u16 avail = ice_get_avail_txq_count(vsi->back);
2842 u16 cpus = num_possible_cpus();
2844 if (avail < cpus / 2)
2847 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2849 if (vsi->num_xdp_txq < cpus)
2850 static_branch_inc(&ice_xdp_locking_key);
2856 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2857 * @vsi: VSI to setup XDP for
2858 * @prog: XDP program
2859 * @extack: netlink extended ack
2862 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2863 struct netlink_ext_ack *extack)
2865 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2866 bool if_running = netif_running(vsi->netdev);
2867 int ret = 0, xdp_ring_err = 0;
2869 if (frame_size > vsi->rx_buf_len) {
2870 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2874 /* need to stop netdev while setting up the program for Rx rings */
2875 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2876 ret = ice_down(vsi);
2878 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2883 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2884 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2886 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2888 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2890 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2892 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2893 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2895 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2897 /* safe to call even when prog == vsi->xdp_prog as
2898 * dev_xdp_install in net/core/dev.c incremented prog's
2899 * refcount so corresponding bpf_prog_put won't cause
2902 ice_vsi_assign_bpf_prog(vsi, prog);
2909 ice_vsi_rx_napi_schedule(vsi);
2911 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2915 * ice_xdp_safe_mode - XDP handler for safe mode
2919 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2920 struct netdev_bpf *xdp)
2922 NL_SET_ERR_MSG_MOD(xdp->extack,
2923 "Please provide working DDP firmware package in order to use XDP\n"
2924 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2929 * ice_xdp - implements XDP handler
2933 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2935 struct ice_netdev_priv *np = netdev_priv(dev);
2936 struct ice_vsi *vsi = np->vsi;
2938 if (vsi->type != ICE_VSI_PF) {
2939 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2943 switch (xdp->command) {
2944 case XDP_SETUP_PROG:
2945 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2946 case XDP_SETUP_XSK_POOL:
2947 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2955 * ice_ena_misc_vector - enable the non-queue interrupts
2956 * @pf: board private structure
2958 static void ice_ena_misc_vector(struct ice_pf *pf)
2960 struct ice_hw *hw = &pf->hw;
2963 /* Disable anti-spoof detection interrupt to prevent spurious event
2964 * interrupts during a function reset. Anti-spoof functionally is
2967 val = rd32(hw, GL_MDCK_TX_TDPU);
2968 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2969 wr32(hw, GL_MDCK_TX_TDPU, val);
2971 /* clear things first */
2972 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2973 rd32(hw, PFINT_OICR); /* read to clear */
2975 val = (PFINT_OICR_ECC_ERR_M |
2976 PFINT_OICR_MAL_DETECT_M |
2978 PFINT_OICR_PCI_EXCEPTION_M |
2980 PFINT_OICR_HMC_ERR_M |
2981 PFINT_OICR_PE_PUSH_M |
2982 PFINT_OICR_PE_CRITERR_M);
2984 wr32(hw, PFINT_OICR_ENA, val);
2986 /* SW_ITR_IDX = 0, but don't change INTENA */
2987 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2988 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2992 * ice_misc_intr - misc interrupt handler
2993 * @irq: interrupt number
2994 * @data: pointer to a q_vector
2996 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2998 struct ice_pf *pf = (struct ice_pf *)data;
2999 struct ice_hw *hw = &pf->hw;
3000 irqreturn_t ret = IRQ_NONE;
3004 dev = ice_pf_to_dev(pf);
3005 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3006 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3007 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3009 oicr = rd32(hw, PFINT_OICR);
3010 ena_mask = rd32(hw, PFINT_OICR_ENA);
3012 if (oicr & PFINT_OICR_SWINT_M) {
3013 ena_mask &= ~PFINT_OICR_SWINT_M;
3017 if (oicr & PFINT_OICR_MAL_DETECT_M) {
3018 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3019 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3021 if (oicr & PFINT_OICR_VFLR_M) {
3022 /* disable any further VFLR event notifications */
3023 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3024 u32 reg = rd32(hw, PFINT_OICR_ENA);
3026 reg &= ~PFINT_OICR_VFLR_M;
3027 wr32(hw, PFINT_OICR_ENA, reg);
3029 ena_mask &= ~PFINT_OICR_VFLR_M;
3030 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3034 if (oicr & PFINT_OICR_GRST_M) {
3037 /* we have a reset warning */
3038 ena_mask &= ~PFINT_OICR_GRST_M;
3039 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
3040 GLGEN_RSTAT_RESET_TYPE_S;
3042 if (reset == ICE_RESET_CORER)
3044 else if (reset == ICE_RESET_GLOBR)
3046 else if (reset == ICE_RESET_EMPR)
3049 dev_dbg(dev, "Invalid reset type %d\n", reset);
3051 /* If a reset cycle isn't already in progress, we set a bit in
3052 * pf->state so that the service task can start a reset/rebuild.
3054 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3055 if (reset == ICE_RESET_CORER)
3056 set_bit(ICE_CORER_RECV, pf->state);
3057 else if (reset == ICE_RESET_GLOBR)
3058 set_bit(ICE_GLOBR_RECV, pf->state);
3060 set_bit(ICE_EMPR_RECV, pf->state);
3062 /* There are couple of different bits at play here.
3063 * hw->reset_ongoing indicates whether the hardware is
3064 * in reset. This is set to true when a reset interrupt
3065 * is received and set back to false after the driver
3066 * has determined that the hardware is out of reset.
3068 * ICE_RESET_OICR_RECV in pf->state indicates
3069 * that a post reset rebuild is required before the
3070 * driver is operational again. This is set above.
3072 * As this is the start of the reset/rebuild cycle, set
3073 * both to indicate that.
3075 hw->reset_ongoing = true;
3079 if (oicr & PFINT_OICR_TSYN_TX_M) {
3080 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3081 ice_ptp_process_ts(pf);
3084 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3085 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3086 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3088 /* Save EVENTs from GTSYN register */
3089 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
3090 GLTSYN_STAT_EVENT1_M |
3091 GLTSYN_STAT_EVENT2_M);
3092 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3093 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
3096 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3097 if (oicr & ICE_AUX_CRIT_ERR) {
3098 pf->oicr_err_reg |= oicr;
3099 set_bit(ICE_AUX_ERR_PENDING, pf->state);
3100 ena_mask &= ~ICE_AUX_CRIT_ERR;
3103 /* Report any remaining unexpected interrupts */
3106 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3107 /* If a critical error is pending there is no choice but to
3110 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3111 PFINT_OICR_ECC_ERR_M)) {
3112 set_bit(ICE_PFR_REQ, pf->state);
3113 ice_service_task_schedule(pf);
3118 ice_service_task_schedule(pf);
3119 ice_irq_dynamic_ena(hw, NULL, NULL);
3125 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3126 * @hw: pointer to HW structure
3128 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3130 /* disable Admin queue Interrupt causes */
3131 wr32(hw, PFINT_FW_CTL,
3132 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3134 /* disable Mailbox queue Interrupt causes */
3135 wr32(hw, PFINT_MBX_CTL,
3136 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3138 wr32(hw, PFINT_SB_CTL,
3139 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3141 /* disable Control queue Interrupt causes */
3142 wr32(hw, PFINT_OICR_CTL,
3143 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3149 * ice_free_irq_msix_misc - Unroll misc vector setup
3150 * @pf: board private structure
3152 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3154 struct ice_hw *hw = &pf->hw;
3156 ice_dis_ctrlq_interrupts(hw);
3158 /* disable OICR interrupt */
3159 wr32(hw, PFINT_OICR_ENA, 0);
3162 if (pf->msix_entries) {
3163 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
3164 devm_free_irq(ice_pf_to_dev(pf),
3165 pf->msix_entries[pf->oicr_idx].vector, pf);
3168 pf->num_avail_sw_msix += 1;
3169 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
3173 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3174 * @hw: pointer to HW structure
3175 * @reg_idx: HW vector index to associate the control queue interrupts with
3177 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3181 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3182 PFINT_OICR_CTL_CAUSE_ENA_M);
3183 wr32(hw, PFINT_OICR_CTL, val);
3185 /* enable Admin queue Interrupt causes */
3186 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3187 PFINT_FW_CTL_CAUSE_ENA_M);
3188 wr32(hw, PFINT_FW_CTL, val);
3190 /* enable Mailbox queue Interrupt causes */
3191 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3192 PFINT_MBX_CTL_CAUSE_ENA_M);
3193 wr32(hw, PFINT_MBX_CTL, val);
3195 /* This enables Sideband queue Interrupt causes */
3196 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3197 PFINT_SB_CTL_CAUSE_ENA_M);
3198 wr32(hw, PFINT_SB_CTL, val);
3204 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3205 * @pf: board private structure
3207 * This sets up the handler for MSIX 0, which is used to manage the
3208 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3209 * when in MSI or Legacy interrupt mode.
3211 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3213 struct device *dev = ice_pf_to_dev(pf);
3214 struct ice_hw *hw = &pf->hw;
3215 int oicr_idx, err = 0;
3217 if (!pf->int_name[0])
3218 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3219 dev_driver_string(dev), dev_name(dev));
3221 /* Do not request IRQ but do enable OICR interrupt since settings are
3222 * lost during reset. Note that this function is called only during
3223 * rebuild path and not while reset is in progress.
3225 if (ice_is_reset_in_progress(pf->state))
3228 /* reserve one vector in irq_tracker for misc interrupts */
3229 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3233 pf->num_avail_sw_msix -= 1;
3234 pf->oicr_idx = (u16)oicr_idx;
3236 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
3237 ice_misc_intr, 0, pf->int_name, pf);
3239 dev_err(dev, "devm_request_irq for %s failed: %d\n",
3241 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3242 pf->num_avail_sw_msix += 1;
3247 ice_ena_misc_vector(pf);
3249 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3250 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3251 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3254 ice_irq_dynamic_ena(hw, NULL, NULL);
3260 * ice_napi_add - register NAPI handler for the VSI
3261 * @vsi: VSI for which NAPI handler is to be registered
3263 * This function is only called in the driver's load path. Registering the NAPI
3264 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3265 * reset/rebuild, etc.)
3267 static void ice_napi_add(struct ice_vsi *vsi)
3274 ice_for_each_q_vector(vsi, v_idx)
3275 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3276 ice_napi_poll, NAPI_POLL_WEIGHT);
3280 * ice_set_ops - set netdev and ethtools ops for the given netdev
3281 * @netdev: netdev instance
3283 static void ice_set_ops(struct net_device *netdev)
3285 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3287 if (ice_is_safe_mode(pf)) {
3288 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3289 ice_set_ethtool_safe_mode_ops(netdev);
3293 netdev->netdev_ops = &ice_netdev_ops;
3294 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3295 ice_set_ethtool_ops(netdev);
3299 * ice_set_netdev_features - set features for the given netdev
3300 * @netdev: netdev instance
3302 static void ice_set_netdev_features(struct net_device *netdev)
3304 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3305 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3306 netdev_features_t csumo_features;
3307 netdev_features_t vlano_features;
3308 netdev_features_t dflt_features;
3309 netdev_features_t tso_features;
3311 if (ice_is_safe_mode(pf)) {
3313 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3314 netdev->hw_features = netdev->features;
3318 dflt_features = NETIF_F_SG |
3323 csumo_features = NETIF_F_RXCSUM |
3328 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3329 NETIF_F_HW_VLAN_CTAG_TX |
3330 NETIF_F_HW_VLAN_CTAG_RX;
3332 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3334 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3336 tso_features = NETIF_F_TSO |
3340 NETIF_F_GSO_UDP_TUNNEL |
3341 NETIF_F_GSO_GRE_CSUM |
3342 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3343 NETIF_F_GSO_PARTIAL |
3344 NETIF_F_GSO_IPXIP4 |
3345 NETIF_F_GSO_IPXIP6 |
3348 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3349 NETIF_F_GSO_GRE_CSUM;
3350 /* set features that user can change */
3351 netdev->hw_features = dflt_features | csumo_features |
3352 vlano_features | tso_features;
3354 /* add support for HW_CSUM on packets with MPLS header */
3355 netdev->mpls_features = NETIF_F_HW_CSUM |
3359 /* enable features */
3360 netdev->features |= netdev->hw_features;
3362 netdev->hw_features |= NETIF_F_HW_TC;
3363 netdev->hw_features |= NETIF_F_LOOPBACK;
3365 /* encap and VLAN devices inherit default, csumo and tso features */
3366 netdev->hw_enc_features |= dflt_features | csumo_features |
3368 netdev->vlan_features |= dflt_features | csumo_features |
3371 /* advertise support but don't enable by default since only one type of
3372 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3373 * type turns on the other has to be turned off. This is enforced by the
3374 * ice_fix_features() ndo callback.
3377 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3378 NETIF_F_HW_VLAN_STAG_TX;
3382 * ice_cfg_netdev - Allocate, configure and register a netdev
3383 * @vsi: the VSI associated with the new netdev
3385 * Returns 0 on success, negative value on failure
3387 static int ice_cfg_netdev(struct ice_vsi *vsi)
3389 struct ice_netdev_priv *np;
3390 struct net_device *netdev;
3391 u8 mac_addr[ETH_ALEN];
3393 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3398 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3399 vsi->netdev = netdev;
3400 np = netdev_priv(netdev);
3403 ice_set_netdev_features(netdev);
3405 ice_set_ops(netdev);
3407 if (vsi->type == ICE_VSI_PF) {
3408 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3409 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3410 eth_hw_addr_set(netdev, mac_addr);
3411 ether_addr_copy(netdev->perm_addr, mac_addr);
3414 netdev->priv_flags |= IFF_UNICAST_FLT;
3416 /* Setup netdev TC information */
3417 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3419 /* setup watchdog timeout value to be 5 second */
3420 netdev->watchdog_timeo = 5 * HZ;
3422 netdev->min_mtu = ETH_MIN_MTU;
3423 netdev->max_mtu = ICE_MAX_MTU;
3429 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3430 * @lut: Lookup table
3431 * @rss_table_size: Lookup table size
3432 * @rss_size: Range of queue number for hashing
3434 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3438 for (i = 0; i < rss_table_size; i++)
3439 lut[i] = i % rss_size;
3443 * ice_pf_vsi_setup - Set up a PF VSI
3444 * @pf: board private structure
3445 * @pi: pointer to the port_info instance
3447 * Returns pointer to the successfully allocated VSI software struct
3448 * on success, otherwise returns NULL on failure.
3450 static struct ice_vsi *
3451 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3453 return ice_vsi_setup(pf, pi, ICE_VSI_PF, NULL, NULL);
3456 static struct ice_vsi *
3457 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3458 struct ice_channel *ch)
3460 return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, NULL, ch);
3464 * ice_ctrl_vsi_setup - Set up a control VSI
3465 * @pf: board private structure
3466 * @pi: pointer to the port_info instance
3468 * Returns pointer to the successfully allocated VSI software struct
3469 * on success, otherwise returns NULL on failure.
3471 static struct ice_vsi *
3472 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3474 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, NULL, NULL);
3478 * ice_lb_vsi_setup - Set up a loopback VSI
3479 * @pf: board private structure
3480 * @pi: pointer to the port_info instance
3482 * Returns pointer to the successfully allocated VSI software struct
3483 * on success, otherwise returns NULL on failure.
3486 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3488 return ice_vsi_setup(pf, pi, ICE_VSI_LB, NULL, NULL);
3492 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3493 * @netdev: network interface to be adjusted
3495 * @vid: VLAN ID to be added
3497 * net_device_ops implementation for adding VLAN IDs
3500 ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3502 struct ice_netdev_priv *np = netdev_priv(netdev);
3503 struct ice_vsi_vlan_ops *vlan_ops;
3504 struct ice_vsi *vsi = np->vsi;
3505 struct ice_vlan vlan;
3508 /* VLAN 0 is added by default during load/reset */
3512 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3513 usleep_range(1000, 2000);
3515 /* Add multicast promisc rule for the VLAN ID to be added if
3516 * all-multicast is currently enabled.
3518 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3519 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3520 ICE_MCAST_VLAN_PROMISC_BITS,
3526 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3528 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3529 * packets aren't pruned by the device's internal switch on Rx
3531 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3532 ret = vlan_ops->add_vlan(vsi, &vlan);
3536 /* If all-multicast is currently enabled and this VLAN ID is only one
3537 * besides VLAN-0 we have to update look-up type of multicast promisc
3538 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3540 if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3541 ice_vsi_num_non_zero_vlans(vsi) == 1) {
3542 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3543 ICE_MCAST_PROMISC_BITS, 0);
3544 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3545 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3549 clear_bit(ICE_CFG_BUSY, vsi->state);
3555 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3556 * @netdev: network interface to be adjusted
3558 * @vid: VLAN ID to be removed
3560 * net_device_ops implementation for removing VLAN IDs
3563 ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3565 struct ice_netdev_priv *np = netdev_priv(netdev);
3566 struct ice_vsi_vlan_ops *vlan_ops;
3567 struct ice_vsi *vsi = np->vsi;
3568 struct ice_vlan vlan;
3571 /* don't allow removal of VLAN 0 */
3575 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3576 usleep_range(1000, 2000);
3578 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3579 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3581 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3583 vsi->current_netdev_flags |= IFF_ALLMULTI;
3586 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3588 /* Make sure VLAN delete is successful before updating VLAN
3591 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3592 ret = vlan_ops->del_vlan(vsi, &vlan);
3596 /* Remove multicast promisc rule for the removed VLAN ID if
3597 * all-multicast is enabled.
3599 if (vsi->current_netdev_flags & IFF_ALLMULTI)
3600 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3601 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3603 if (!ice_vsi_has_non_zero_vlans(vsi)) {
3604 /* Update look-up type of multicast promisc rule for VLAN 0
3605 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3606 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3608 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3609 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3610 ICE_MCAST_VLAN_PROMISC_BITS,
3612 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3613 ICE_MCAST_PROMISC_BITS, 0);
3618 clear_bit(ICE_CFG_BUSY, vsi->state);
3624 * ice_rep_indr_tc_block_unbind
3625 * @cb_priv: indirection block private data
3627 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3629 struct ice_indr_block_priv *indr_priv = cb_priv;
3631 list_del(&indr_priv->list);
3636 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3637 * @vsi: VSI struct which has the netdev
3639 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3641 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3643 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3644 ice_rep_indr_tc_block_unbind);
3648 * ice_tc_indir_block_remove - clean indirect TC block notifications
3651 static void ice_tc_indir_block_remove(struct ice_pf *pf)
3653 struct ice_vsi *pf_vsi = ice_get_main_vsi(pf);
3658 ice_tc_indir_block_unregister(pf_vsi);
3662 * ice_tc_indir_block_register - Register TC indirect block notifications
3663 * @vsi: VSI struct which has the netdev
3665 * Returns 0 on success, negative value on failure
3667 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3669 struct ice_netdev_priv *np;
3671 if (!vsi || !vsi->netdev)
3674 np = netdev_priv(vsi->netdev);
3676 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3677 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3681 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3682 * @pf: board private structure
3684 * Returns 0 on success, negative value on failure
3686 static int ice_setup_pf_sw(struct ice_pf *pf)
3688 struct device *dev = ice_pf_to_dev(pf);
3689 bool dvm = ice_is_dvm_ena(&pf->hw);
3690 struct ice_vsi *vsi;
3693 if (ice_is_reset_in_progress(pf->state))
3696 status = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
3700 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3704 /* init channel list */
3705 INIT_LIST_HEAD(&vsi->ch_list);
3707 status = ice_cfg_netdev(vsi);
3709 goto unroll_vsi_setup;
3710 /* netdev has to be configured before setting frame size */
3711 ice_vsi_cfg_frame_size(vsi);
3713 /* init indirect block notifications */
3714 status = ice_tc_indir_block_register(vsi);
3716 dev_err(dev, "Failed to register netdev notifier\n");
3717 goto unroll_cfg_netdev;
3720 /* Setup DCB netlink interface */
3721 ice_dcbnl_setup(vsi);
3723 /* registering the NAPI handler requires both the queues and
3724 * netdev to be created, which are done in ice_pf_vsi_setup()
3725 * and ice_cfg_netdev() respectively
3729 status = ice_init_mac_fltr(pf);
3731 goto unroll_napi_add;
3736 ice_tc_indir_block_unregister(vsi);
3741 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3742 free_netdev(vsi->netdev);
3748 ice_vsi_release(vsi);
3753 * ice_get_avail_q_count - Get count of queues in use
3754 * @pf_qmap: bitmap to get queue use count from
3755 * @lock: pointer to a mutex that protects access to pf_qmap
3756 * @size: size of the bitmap
3759 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3765 for_each_clear_bit(bit, pf_qmap, size)
3773 * ice_get_avail_txq_count - Get count of Tx queues in use
3774 * @pf: pointer to an ice_pf instance
3776 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3778 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3783 * ice_get_avail_rxq_count - Get count of Rx queues in use
3784 * @pf: pointer to an ice_pf instance
3786 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3788 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3793 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3794 * @pf: board private structure to initialize
3796 static void ice_deinit_pf(struct ice_pf *pf)
3798 ice_service_task_stop(pf);
3799 mutex_destroy(&pf->adev_mutex);
3800 mutex_destroy(&pf->sw_mutex);
3801 mutex_destroy(&pf->tc_mutex);
3802 mutex_destroy(&pf->avail_q_mutex);
3803 mutex_destroy(&pf->vfs.table_lock);
3805 if (pf->avail_txqs) {
3806 bitmap_free(pf->avail_txqs);
3807 pf->avail_txqs = NULL;
3810 if (pf->avail_rxqs) {
3811 bitmap_free(pf->avail_rxqs);
3812 pf->avail_rxqs = NULL;
3816 ptp_clock_unregister(pf->ptp.clock);
3820 * ice_set_pf_caps - set PFs capability flags
3821 * @pf: pointer to the PF instance
3823 static void ice_set_pf_caps(struct ice_pf *pf)
3825 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3827 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3828 if (func_caps->common_cap.rdma)
3829 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3830 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3831 if (func_caps->common_cap.dcb)
3832 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3833 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3834 if (func_caps->common_cap.sr_iov_1_1) {
3835 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3836 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3839 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3840 if (func_caps->common_cap.rss_table_size)
3841 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3843 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3844 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3847 /* ctrl_vsi_idx will be set to a valid value when flow director
3848 * is setup by ice_init_fdir
3850 pf->ctrl_vsi_idx = ICE_NO_VSI;
3851 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3852 /* force guaranteed filter pool for PF */
3853 ice_alloc_fd_guar_item(&pf->hw, &unused,
3854 func_caps->fd_fltr_guar);
3855 /* force shared filter pool for PF */
3856 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3857 func_caps->fd_fltr_best_effort);
3860 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3861 if (func_caps->common_cap.ieee_1588)
3862 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3864 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3865 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3869 * ice_init_pf - Initialize general software structures (struct ice_pf)
3870 * @pf: board private structure to initialize
3872 static int ice_init_pf(struct ice_pf *pf)
3874 ice_set_pf_caps(pf);
3876 mutex_init(&pf->sw_mutex);
3877 mutex_init(&pf->tc_mutex);
3878 mutex_init(&pf->adev_mutex);
3880 INIT_HLIST_HEAD(&pf->aq_wait_list);
3881 spin_lock_init(&pf->aq_wait_lock);
3882 init_waitqueue_head(&pf->aq_wait_queue);
3884 init_waitqueue_head(&pf->reset_wait_queue);
3886 /* setup service timer and periodic service task */
3887 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3888 pf->serv_tmr_period = HZ;
3889 INIT_WORK(&pf->serv_task, ice_service_task);
3890 clear_bit(ICE_SERVICE_SCHED, pf->state);
3892 mutex_init(&pf->avail_q_mutex);
3893 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3894 if (!pf->avail_txqs)
3897 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3898 if (!pf->avail_rxqs) {
3899 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3900 pf->avail_txqs = NULL;
3904 mutex_init(&pf->vfs.table_lock);
3905 hash_init(pf->vfs.table);
3911 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3912 * @pf: board private structure
3914 * compute the number of MSIX vectors required (v_budget) and request from
3915 * the OS. Return the number of vectors reserved or negative on failure
3917 static int ice_ena_msix_range(struct ice_pf *pf)
3919 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3920 struct device *dev = ice_pf_to_dev(pf);
3923 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3924 num_cpus = num_online_cpus();
3926 /* reserve for LAN miscellaneous handler */
3927 needed = ICE_MIN_LAN_OICR_MSIX;
3928 if (v_left < needed)
3929 goto no_hw_vecs_left_err;
3933 /* reserve for flow director */
3934 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3935 needed = ICE_FDIR_MSIX;
3936 if (v_left < needed)
3937 goto no_hw_vecs_left_err;
3942 /* reserve for switchdev */
3943 needed = ICE_ESWITCH_MSIX;
3944 if (v_left < needed)
3945 goto no_hw_vecs_left_err;
3949 /* total used for non-traffic vectors */
3952 /* reserve vectors for LAN traffic */
3954 if (v_left < needed)
3955 goto no_hw_vecs_left_err;
3956 pf->num_lan_msix = needed;
3960 /* reserve vectors for RDMA auxiliary driver */
3961 if (ice_is_rdma_ena(pf)) {
3962 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3963 if (v_left < needed)
3964 goto no_hw_vecs_left_err;
3965 pf->num_rdma_msix = needed;
3970 pf->msix_entries = devm_kcalloc(dev, v_budget,
3971 sizeof(*pf->msix_entries), GFP_KERNEL);
3972 if (!pf->msix_entries) {
3977 for (i = 0; i < v_budget; i++)
3978 pf->msix_entries[i].entry = i;
3980 /* actually reserve the vectors */
3981 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3982 ICE_MIN_MSIX, v_budget);
3984 dev_err(dev, "unable to reserve MSI-X vectors\n");
3989 if (v_actual < v_budget) {
3990 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3991 v_budget, v_actual);
3993 if (v_actual < ICE_MIN_MSIX) {
3994 /* error if we can't get minimum vectors */
3995 pci_disable_msix(pf->pdev);
3999 int v_remain = v_actual - v_other;
4000 int v_rdma = 0, v_min_rdma = 0;
4002 if (ice_is_rdma_ena(pf)) {
4003 /* Need at least 1 interrupt in addition to
4006 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
4007 v_min_rdma = ICE_MIN_RDMA_MSIX;
4010 if (v_actual == ICE_MIN_MSIX ||
4011 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
4012 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
4013 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
4015 pf->num_rdma_msix = 0;
4016 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
4017 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
4018 (v_remain - v_rdma < v_rdma)) {
4019 /* Support minimum RDMA and give remaining
4020 * vectors to LAN MSIX
4022 pf->num_rdma_msix = v_min_rdma;
4023 pf->num_lan_msix = v_remain - v_min_rdma;
4025 /* Split remaining MSIX with RDMA after
4026 * accounting for AEQ MSIX
4028 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
4029 ICE_RDMA_NUM_AEQ_MSIX;
4030 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
4033 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
4036 if (ice_is_rdma_ena(pf))
4037 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
4045 devm_kfree(dev, pf->msix_entries);
4048 no_hw_vecs_left_err:
4049 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
4053 pf->num_rdma_msix = 0;
4054 pf->num_lan_msix = 0;
4059 * ice_dis_msix - Disable MSI-X interrupt setup in OS
4060 * @pf: board private structure
4062 static void ice_dis_msix(struct ice_pf *pf)
4064 pci_disable_msix(pf->pdev);
4065 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
4066 pf->msix_entries = NULL;
4070 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
4071 * @pf: board private structure
4073 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
4077 if (pf->irq_tracker) {
4078 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
4079 pf->irq_tracker = NULL;
4084 * ice_init_interrupt_scheme - Determine proper interrupt scheme
4085 * @pf: board private structure to initialize
4087 static int ice_init_interrupt_scheme(struct ice_pf *pf)
4091 vectors = ice_ena_msix_range(pf);
4096 /* set up vector assignment tracking */
4097 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
4098 struct_size(pf->irq_tracker, list, vectors),
4100 if (!pf->irq_tracker) {
4105 /* populate SW interrupts pool with number of OS granted IRQs. */
4106 pf->num_avail_sw_msix = (u16)vectors;
4107 pf->irq_tracker->num_entries = (u16)vectors;
4108 pf->irq_tracker->end = pf->irq_tracker->num_entries;
4114 * ice_is_wol_supported - check if WoL is supported
4115 * @hw: pointer to hardware info
4117 * Check if WoL is supported based on the HW configuration.
4118 * Returns true if NVM supports and enables WoL for this port, false otherwise
4120 bool ice_is_wol_supported(struct ice_hw *hw)
4124 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4125 * word) indicates WoL is not supported on the corresponding PF ID.
4127 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4130 return !(BIT(hw->port_info->lport) & wol_ctrl);
4134 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4135 * @vsi: VSI being changed
4136 * @new_rx: new number of Rx queues
4137 * @new_tx: new number of Tx queues
4139 * Only change the number of queues if new_tx, or new_rx is non-0.
4141 * Returns 0 on success.
4143 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
4145 struct ice_pf *pf = vsi->back;
4146 int err = 0, timeout = 50;
4148 if (!new_rx && !new_tx)
4151 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4155 usleep_range(1000, 2000);
4159 vsi->req_txq = (u16)new_tx;
4161 vsi->req_rxq = (u16)new_rx;
4163 /* set for the next time the netdev is started */
4164 if (!netif_running(vsi->netdev)) {
4165 ice_vsi_rebuild(vsi, false);
4166 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4171 ice_vsi_rebuild(vsi, false);
4172 ice_pf_dcb_recfg(pf);
4175 clear_bit(ICE_CFG_BUSY, pf->state);
4180 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4181 * @pf: PF to configure
4183 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4184 * VSI can still Tx/Rx VLAN tagged packets.
4186 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4188 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4189 struct ice_vsi_ctx *ctxt;
4196 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4201 ctxt->info = vsi->info;
4203 ctxt->info.valid_sections =
4204 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4205 ICE_AQ_VSI_PROP_SECURITY_VALID |
4206 ICE_AQ_VSI_PROP_SW_VALID);
4208 /* disable VLAN anti-spoof */
4209 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4210 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4212 /* disable VLAN pruning and keep all other settings */
4213 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4215 /* allow all VLANs on Tx and don't strip on Rx */
4216 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4217 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4219 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4221 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4222 status, ice_aq_str(hw->adminq.sq_last_status));
4224 vsi->info.sec_flags = ctxt->info.sec_flags;
4225 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4226 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4233 * ice_log_pkg_init - log result of DDP package load
4234 * @hw: pointer to hardware info
4235 * @state: state of package load
4237 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4239 struct ice_pf *pf = hw->back;
4242 dev = ice_pf_to_dev(pf);
4245 case ICE_DDP_PKG_SUCCESS:
4246 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4247 hw->active_pkg_name,
4248 hw->active_pkg_ver.major,
4249 hw->active_pkg_ver.minor,
4250 hw->active_pkg_ver.update,
4251 hw->active_pkg_ver.draft);
4253 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4254 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4255 hw->active_pkg_name,
4256 hw->active_pkg_ver.major,
4257 hw->active_pkg_ver.minor,
4258 hw->active_pkg_ver.update,
4259 hw->active_pkg_ver.draft);
4261 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4262 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",
4263 hw->active_pkg_name,
4264 hw->active_pkg_ver.major,
4265 hw->active_pkg_ver.minor,
4266 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4268 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4269 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",
4270 hw->active_pkg_name,
4271 hw->active_pkg_ver.major,
4272 hw->active_pkg_ver.minor,
4273 hw->active_pkg_ver.update,
4274 hw->active_pkg_ver.draft,
4281 case ICE_DDP_PKG_FW_MISMATCH:
4282 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");
4284 case ICE_DDP_PKG_INVALID_FILE:
4285 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4287 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4288 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4290 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4291 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",
4292 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4294 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4295 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");
4297 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4298 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");
4300 case ICE_DDP_PKG_LOAD_ERROR:
4301 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4302 /* poll for reset to complete */
4303 if (ice_check_reset(hw))
4304 dev_err(dev, "Error resetting device. Please reload the driver\n");
4306 case ICE_DDP_PKG_ERR:
4308 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4314 * ice_load_pkg - load/reload the DDP Package file
4315 * @firmware: firmware structure when firmware requested or NULL for reload
4316 * @pf: pointer to the PF instance
4318 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4319 * initialize HW tables.
4322 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4324 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4325 struct device *dev = ice_pf_to_dev(pf);
4326 struct ice_hw *hw = &pf->hw;
4328 /* Load DDP Package */
4329 if (firmware && !hw->pkg_copy) {
4330 state = ice_copy_and_init_pkg(hw, firmware->data,
4332 ice_log_pkg_init(hw, state);
4333 } else if (!firmware && hw->pkg_copy) {
4334 /* Reload package during rebuild after CORER/GLOBR reset */
4335 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4336 ice_log_pkg_init(hw, state);
4338 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4341 if (!ice_is_init_pkg_successful(state)) {
4343 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4347 /* Successful download package is the precondition for advanced
4348 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4350 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4354 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4355 * @pf: pointer to the PF structure
4357 * There is no error returned here because the driver should be able to handle
4358 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4359 * specifically with Tx.
4361 static void ice_verify_cacheline_size(struct ice_pf *pf)
4363 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4364 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4365 ICE_CACHE_LINE_BYTES);
4369 * ice_send_version - update firmware with driver version
4372 * Returns 0 on success, else error code
4374 static int ice_send_version(struct ice_pf *pf)
4376 struct ice_driver_ver dv;
4378 dv.major_ver = 0xff;
4379 dv.minor_ver = 0xff;
4380 dv.build_ver = 0xff;
4381 dv.subbuild_ver = 0;
4382 strscpy((char *)dv.driver_string, UTS_RELEASE,
4383 sizeof(dv.driver_string));
4384 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4388 * ice_init_fdir - Initialize flow director VSI and configuration
4389 * @pf: pointer to the PF instance
4391 * returns 0 on success, negative on error
4393 static int ice_init_fdir(struct ice_pf *pf)
4395 struct device *dev = ice_pf_to_dev(pf);
4396 struct ice_vsi *ctrl_vsi;
4399 /* Side Band Flow Director needs to have a control VSI.
4400 * Allocate it and store it in the PF.
4402 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4404 dev_dbg(dev, "could not create control VSI\n");
4408 err = ice_vsi_open_ctrl(ctrl_vsi);
4410 dev_dbg(dev, "could not open control VSI\n");
4414 mutex_init(&pf->hw.fdir_fltr_lock);
4416 err = ice_fdir_create_dflt_rules(pf);
4423 ice_fdir_release_flows(&pf->hw);
4424 ice_vsi_close(ctrl_vsi);
4426 ice_vsi_release(ctrl_vsi);
4427 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4428 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4429 pf->ctrl_vsi_idx = ICE_NO_VSI;
4435 * ice_get_opt_fw_name - return optional firmware file name or NULL
4436 * @pf: pointer to the PF instance
4438 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4440 /* Optional firmware name same as default with additional dash
4441 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4443 struct pci_dev *pdev = pf->pdev;
4444 char *opt_fw_filename;
4447 /* Determine the name of the optional file using the DSN (two
4448 * dwords following the start of the DSN Capability).
4450 dsn = pci_get_dsn(pdev);
4454 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4455 if (!opt_fw_filename)
4458 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4459 ICE_DDP_PKG_PATH, dsn);
4461 return opt_fw_filename;
4465 * ice_request_fw - Device initialization routine
4466 * @pf: pointer to the PF instance
4468 static void ice_request_fw(struct ice_pf *pf)
4470 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4471 const struct firmware *firmware = NULL;
4472 struct device *dev = ice_pf_to_dev(pf);
4475 /* optional device-specific DDP (if present) overrides the default DDP
4476 * package file. kernel logs a debug message if the file doesn't exist,
4477 * and warning messages for other errors.
4479 if (opt_fw_filename) {
4480 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4482 kfree(opt_fw_filename);
4486 /* request for firmware was successful. Download to device */
4487 ice_load_pkg(firmware, pf);
4488 kfree(opt_fw_filename);
4489 release_firmware(firmware);
4494 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4496 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4500 /* request for firmware was successful. Download to device */
4501 ice_load_pkg(firmware, pf);
4502 release_firmware(firmware);
4506 * ice_print_wake_reason - show the wake up cause in the log
4507 * @pf: pointer to the PF struct
4509 static void ice_print_wake_reason(struct ice_pf *pf)
4511 u32 wus = pf->wakeup_reason;
4512 const char *wake_str;
4514 /* if no wake event, nothing to print */
4518 if (wus & PFPM_WUS_LNKC_M)
4519 wake_str = "Link\n";
4520 else if (wus & PFPM_WUS_MAG_M)
4521 wake_str = "Magic Packet\n";
4522 else if (wus & PFPM_WUS_MNG_M)
4523 wake_str = "Management\n";
4524 else if (wus & PFPM_WUS_FW_RST_WK_M)
4525 wake_str = "Firmware Reset\n";
4527 wake_str = "Unknown\n";
4529 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4533 * ice_register_netdev - register netdev and devlink port
4534 * @pf: pointer to the PF struct
4536 static int ice_register_netdev(struct ice_pf *pf)
4538 struct ice_vsi *vsi;
4541 vsi = ice_get_main_vsi(pf);
4542 if (!vsi || !vsi->netdev)
4545 err = register_netdev(vsi->netdev);
4547 goto err_register_netdev;
4549 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4550 netif_carrier_off(vsi->netdev);
4551 netif_tx_stop_all_queues(vsi->netdev);
4552 err = ice_devlink_create_pf_port(pf);
4554 goto err_devlink_create;
4556 devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
4560 unregister_netdev(vsi->netdev);
4561 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4562 err_register_netdev:
4563 free_netdev(vsi->netdev);
4565 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4570 * ice_probe - Device initialization routine
4571 * @pdev: PCI device information struct
4572 * @ent: entry in ice_pci_tbl
4574 * Returns 0 on success, negative on failure
4577 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4579 struct device *dev = &pdev->dev;
4584 if (pdev->is_virtfn) {
4585 dev_err(dev, "can't probe a virtual function\n");
4589 /* this driver uses devres, see
4590 * Documentation/driver-api/driver-model/devres.rst
4592 err = pcim_enable_device(pdev);
4596 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4598 dev_err(dev, "BAR0 I/O map error %d\n", err);
4602 pf = ice_allocate_pf(dev);
4606 /* initialize Auxiliary index to invalid value */
4609 /* set up for high or low DMA */
4610 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4612 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4616 pci_enable_pcie_error_reporting(pdev);
4617 pci_set_master(pdev);
4620 pci_set_drvdata(pdev, pf);
4621 set_bit(ICE_DOWN, pf->state);
4622 /* Disable service task until DOWN bit is cleared */
4623 set_bit(ICE_SERVICE_DIS, pf->state);
4626 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4627 pci_save_state(pdev);
4630 hw->vendor_id = pdev->vendor;
4631 hw->device_id = pdev->device;
4632 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4633 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4634 hw->subsystem_device_id = pdev->subsystem_device;
4635 hw->bus.device = PCI_SLOT(pdev->devfn);
4636 hw->bus.func = PCI_FUNC(pdev->devfn);
4637 ice_set_ctrlq_len(hw);
4639 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4641 #ifndef CONFIG_DYNAMIC_DEBUG
4643 hw->debug_mask = debug;
4646 err = ice_init_hw(hw);
4648 dev_err(dev, "ice_init_hw failed: %d\n", err);
4650 goto err_exit_unroll;
4653 ice_init_feature_support(pf);
4657 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4658 * set in pf->state, which will cause ice_is_safe_mode to return
4661 if (ice_is_safe_mode(pf)) {
4662 /* we already got function/device capabilities but these don't
4663 * reflect what the driver needs to do in safe mode. Instead of
4664 * adding conditional logic everywhere to ignore these
4665 * device/function capabilities, override them.
4667 ice_set_safe_mode_caps(hw);
4670 hw->ucast_shared = true;
4672 err = ice_init_pf(pf);
4674 dev_err(dev, "ice_init_pf failed: %d\n", err);
4675 goto err_init_pf_unroll;
4678 ice_devlink_init_regions(pf);
4680 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4681 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4682 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4683 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4685 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4686 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4687 pf->hw.tnl.valid_count[TNL_VXLAN];
4688 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4689 UDP_TUNNEL_TYPE_VXLAN;
4692 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4693 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4694 pf->hw.tnl.valid_count[TNL_GENEVE];
4695 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4696 UDP_TUNNEL_TYPE_GENEVE;
4700 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4701 if (!pf->num_alloc_vsi) {
4703 goto err_init_pf_unroll;
4705 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4706 dev_warn(&pf->pdev->dev,
4707 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4708 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4709 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4712 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4716 goto err_init_pf_unroll;
4719 err = ice_init_interrupt_scheme(pf);
4721 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4723 goto err_init_vsi_unroll;
4726 /* In case of MSIX we are going to setup the misc vector right here
4727 * to handle admin queue events etc. In case of legacy and MSI
4728 * the misc functionality and queue processing is combined in
4729 * the same vector and that gets setup at open.
4731 err = ice_req_irq_msix_misc(pf);
4733 dev_err(dev, "setup of misc vector failed: %d\n", err);
4734 goto err_init_interrupt_unroll;
4737 /* create switch struct for the switch element created by FW on boot */
4738 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4739 if (!pf->first_sw) {
4741 goto err_msix_misc_unroll;
4745 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4747 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4749 pf->first_sw->pf = pf;
4751 /* record the sw_id available for later use */
4752 pf->first_sw->sw_id = hw->port_info->sw_id;
4754 err = ice_setup_pf_sw(pf);
4756 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4757 goto err_alloc_sw_unroll;
4760 clear_bit(ICE_SERVICE_DIS, pf->state);
4762 /* tell the firmware we are up */
4763 err = ice_send_version(pf);
4765 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4767 goto err_send_version_unroll;
4770 /* since everything is good, start the service timer */
4771 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4773 err = ice_init_link_events(pf->hw.port_info);
4775 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4776 goto err_send_version_unroll;
4779 /* not a fatal error if this fails */
4780 err = ice_init_nvm_phy_type(pf->hw.port_info);
4782 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4784 /* not a fatal error if this fails */
4785 err = ice_update_link_info(pf->hw.port_info);
4787 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4789 ice_init_link_dflt_override(pf->hw.port_info);
4791 ice_check_link_cfg_err(pf,
4792 pf->hw.port_info->phy.link_info.link_cfg_err);
4794 /* if media available, initialize PHY settings */
4795 if (pf->hw.port_info->phy.link_info.link_info &
4796 ICE_AQ_MEDIA_AVAILABLE) {
4797 /* not a fatal error if this fails */
4798 err = ice_init_phy_user_cfg(pf->hw.port_info);
4800 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4802 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4803 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4806 ice_configure_phy(vsi);
4809 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4812 ice_verify_cacheline_size(pf);
4814 /* Save wakeup reason register for later use */
4815 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4817 /* check for a power management event */
4818 ice_print_wake_reason(pf);
4820 /* clear wake status, all bits */
4821 wr32(hw, PFPM_WUS, U32_MAX);
4823 /* Disable WoL at init, wait for user to enable */
4824 device_set_wakeup_enable(dev, false);
4826 if (ice_is_safe_mode(pf)) {
4827 ice_set_safe_mode_vlan_cfg(pf);
4831 /* initialize DDP driven features */
4832 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4835 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4838 /* Note: Flow director init failure is non-fatal to load */
4839 if (ice_init_fdir(pf))
4840 dev_err(dev, "could not initialize flow director\n");
4842 /* Note: DCB init failure is non-fatal to load */
4843 if (ice_init_pf_dcb(pf, false)) {
4844 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4845 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4847 ice_cfg_lldp_mib_change(&pf->hw, true);
4850 if (ice_init_lag(pf))
4851 dev_warn(dev, "Failed to init link aggregation support\n");
4853 /* print PCI link speed and width */
4854 pcie_print_link_status(pf->pdev);
4857 err = ice_register_netdev(pf);
4859 goto err_netdev_reg;
4861 err = ice_devlink_register_params(pf);
4863 goto err_netdev_reg;
4865 /* ready to go, so clear down state bit */
4866 clear_bit(ICE_DOWN, pf->state);
4867 if (ice_is_rdma_ena(pf)) {
4868 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4869 if (pf->aux_idx < 0) {
4870 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4872 goto err_devlink_reg_param;
4875 err = ice_init_rdma(pf);
4877 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4879 goto err_init_aux_unroll;
4882 dev_warn(dev, "RDMA is not supported on this device\n");
4885 ice_devlink_register(pf);
4888 err_init_aux_unroll:
4890 ida_free(&ice_aux_ida, pf->aux_idx);
4891 err_devlink_reg_param:
4892 ice_devlink_unregister_params(pf);
4894 err_send_version_unroll:
4895 ice_vsi_release_all(pf);
4896 err_alloc_sw_unroll:
4897 set_bit(ICE_SERVICE_DIS, pf->state);
4898 set_bit(ICE_DOWN, pf->state);
4899 devm_kfree(dev, pf->first_sw);
4900 err_msix_misc_unroll:
4901 ice_free_irq_msix_misc(pf);
4902 err_init_interrupt_unroll:
4903 ice_clear_interrupt_scheme(pf);
4904 err_init_vsi_unroll:
4905 devm_kfree(dev, pf->vsi);
4908 ice_devlink_destroy_regions(pf);
4911 pci_disable_pcie_error_reporting(pdev);
4912 pci_disable_device(pdev);
4917 * ice_set_wake - enable or disable Wake on LAN
4918 * @pf: pointer to the PF struct
4920 * Simple helper for WoL control
4922 static void ice_set_wake(struct ice_pf *pf)
4924 struct ice_hw *hw = &pf->hw;
4925 bool wol = pf->wol_ena;
4927 /* clear wake state, otherwise new wake events won't fire */
4928 wr32(hw, PFPM_WUS, U32_MAX);
4930 /* enable / disable APM wake up, no RMW needed */
4931 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4933 /* set magic packet filter enabled */
4934 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4938 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4939 * @pf: pointer to the PF struct
4941 * Issue firmware command to enable multicast magic wake, making
4942 * sure that any locally administered address (LAA) is used for
4943 * wake, and that PF reset doesn't undo the LAA.
4945 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4947 struct device *dev = ice_pf_to_dev(pf);
4948 struct ice_hw *hw = &pf->hw;
4949 u8 mac_addr[ETH_ALEN];
4950 struct ice_vsi *vsi;
4957 vsi = ice_get_main_vsi(pf);
4961 /* Get current MAC address in case it's an LAA */
4963 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4965 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4967 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4968 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4969 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4971 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4973 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
4974 status, ice_aq_str(hw->adminq.sq_last_status));
4978 * ice_remove - Device removal routine
4979 * @pdev: PCI device information struct
4981 static void ice_remove(struct pci_dev *pdev)
4983 struct ice_pf *pf = pci_get_drvdata(pdev);
4986 ice_devlink_unregister(pf);
4987 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4988 if (!ice_is_reset_in_progress(pf->state))
4993 ice_tc_indir_block_remove(pf);
4995 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4996 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5000 ice_service_task_stop(pf);
5002 ice_aq_cancel_waiting_tasks(pf);
5003 ice_unplug_aux_dev(pf);
5004 if (pf->aux_idx >= 0)
5005 ida_free(&ice_aux_ida, pf->aux_idx);
5006 ice_devlink_unregister_params(pf);
5007 set_bit(ICE_DOWN, pf->state);
5010 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
5011 ice_ptp_release(pf);
5012 if (ice_is_feature_supported(pf, ICE_F_GNSS))
5014 if (!ice_is_safe_mode(pf))
5015 ice_remove_arfs(pf);
5016 ice_setup_mc_magic_wake(pf);
5017 ice_vsi_release_all(pf);
5018 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
5020 ice_free_irq_msix_misc(pf);
5021 ice_for_each_vsi(pf, i) {
5024 ice_vsi_free_q_vectors(pf->vsi[i]);
5027 ice_devlink_destroy_regions(pf);
5028 ice_deinit_hw(&pf->hw);
5030 /* Issue a PFR as part of the prescribed driver unload flow. Do not
5031 * do it via ice_schedule_reset() since there is no need to rebuild
5032 * and the service task is already stopped.
5034 ice_reset(&pf->hw, ICE_RESET_PFR);
5035 pci_wait_for_pending_transaction(pdev);
5036 ice_clear_interrupt_scheme(pf);
5037 pci_disable_pcie_error_reporting(pdev);
5038 pci_disable_device(pdev);
5042 * ice_shutdown - PCI callback for shutting down device
5043 * @pdev: PCI device information struct
5045 static void ice_shutdown(struct pci_dev *pdev)
5047 struct ice_pf *pf = pci_get_drvdata(pdev);
5051 if (system_state == SYSTEM_POWER_OFF) {
5052 pci_wake_from_d3(pdev, pf->wol_ena);
5053 pci_set_power_state(pdev, PCI_D3hot);
5059 * ice_prepare_for_shutdown - prep for PCI shutdown
5060 * @pf: board private structure
5062 * Inform or close all dependent features in prep for PCI device shutdown
5064 static void ice_prepare_for_shutdown(struct ice_pf *pf)
5066 struct ice_hw *hw = &pf->hw;
5069 /* Notify VFs of impending reset */
5070 if (ice_check_sq_alive(hw, &hw->mailboxq))
5071 ice_vc_notify_reset(pf);
5073 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5075 /* disable the VSIs and their queues that are not already DOWN */
5076 ice_pf_dis_all_vsi(pf, false);
5078 ice_for_each_vsi(pf, v)
5080 pf->vsi[v]->vsi_num = 0;
5082 ice_shutdown_all_ctrlq(hw);
5086 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5087 * @pf: board private structure to reinitialize
5089 * This routine reinitialize interrupt scheme that was cleared during
5090 * power management suspend callback.
5092 * This should be called during resume routine to re-allocate the q_vectors
5093 * and reacquire interrupts.
5095 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5097 struct device *dev = ice_pf_to_dev(pf);
5100 /* Since we clear MSIX flag during suspend, we need to
5101 * set it back during resume...
5104 ret = ice_init_interrupt_scheme(pf);
5106 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5110 /* Remap vectors and rings, after successful re-init interrupts */
5111 ice_for_each_vsi(pf, v) {
5115 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5118 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5121 ret = ice_req_irq_msix_misc(pf);
5123 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5133 ice_vsi_free_q_vectors(pf->vsi[v]);
5140 * @dev: generic device information structure
5142 * Power Management callback to quiesce the device and prepare
5143 * for D3 transition.
5145 static int __maybe_unused ice_suspend(struct device *dev)
5147 struct pci_dev *pdev = to_pci_dev(dev);
5151 pf = pci_get_drvdata(pdev);
5153 if (!ice_pf_state_is_nominal(pf)) {
5154 dev_err(dev, "Device is not ready, no need to suspend it\n");
5158 /* Stop watchdog tasks until resume completion.
5159 * Even though it is most likely that the service task is
5160 * disabled if the device is suspended or down, the service task's
5161 * state is controlled by a different state bit, and we should
5162 * store and honor whatever state that bit is in at this point.
5164 disabled = ice_service_task_stop(pf);
5166 ice_unplug_aux_dev(pf);
5168 /* Already suspended?, then there is nothing to do */
5169 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5171 ice_service_task_restart(pf);
5175 if (test_bit(ICE_DOWN, pf->state) ||
5176 ice_is_reset_in_progress(pf->state)) {
5177 dev_err(dev, "can't suspend device in reset or already down\n");
5179 ice_service_task_restart(pf);
5183 ice_setup_mc_magic_wake(pf);
5185 ice_prepare_for_shutdown(pf);
5189 /* Free vectors, clear the interrupt scheme and release IRQs
5190 * for proper hibernation, especially with large number of CPUs.
5191 * Otherwise hibernation might fail when mapping all the vectors back
5194 ice_free_irq_msix_misc(pf);
5195 ice_for_each_vsi(pf, v) {
5198 ice_vsi_free_q_vectors(pf->vsi[v]);
5200 ice_clear_interrupt_scheme(pf);
5202 pci_save_state(pdev);
5203 pci_wake_from_d3(pdev, pf->wol_ena);
5204 pci_set_power_state(pdev, PCI_D3hot);
5209 * ice_resume - PM callback for waking up from D3
5210 * @dev: generic device information structure
5212 static int __maybe_unused ice_resume(struct device *dev)
5214 struct pci_dev *pdev = to_pci_dev(dev);
5215 enum ice_reset_req reset_type;
5220 pci_set_power_state(pdev, PCI_D0);
5221 pci_restore_state(pdev);
5222 pci_save_state(pdev);
5224 if (!pci_device_is_present(pdev))
5227 ret = pci_enable_device_mem(pdev);
5229 dev_err(dev, "Cannot enable device after suspend\n");
5233 pf = pci_get_drvdata(pdev);
5236 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5237 ice_print_wake_reason(pf);
5239 /* We cleared the interrupt scheme when we suspended, so we need to
5240 * restore it now to resume device functionality.
5242 ret = ice_reinit_interrupt_scheme(pf);
5244 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5246 clear_bit(ICE_DOWN, pf->state);
5247 /* Now perform PF reset and rebuild */
5248 reset_type = ICE_RESET_PFR;
5249 /* re-enable service task for reset, but allow reset to schedule it */
5250 clear_bit(ICE_SERVICE_DIS, pf->state);
5252 if (ice_schedule_reset(pf, reset_type))
5253 dev_err(dev, "Reset during resume failed.\n");
5255 clear_bit(ICE_SUSPENDED, pf->state);
5256 ice_service_task_restart(pf);
5258 /* Restart the service task */
5259 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5263 #endif /* CONFIG_PM */
5266 * ice_pci_err_detected - warning that PCI error has been detected
5267 * @pdev: PCI device information struct
5268 * @err: the type of PCI error
5270 * Called to warn that something happened on the PCI bus and the error handling
5271 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5273 static pci_ers_result_t
5274 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5276 struct ice_pf *pf = pci_get_drvdata(pdev);
5279 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5281 return PCI_ERS_RESULT_DISCONNECT;
5284 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5285 ice_service_task_stop(pf);
5287 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5288 set_bit(ICE_PFR_REQ, pf->state);
5289 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5293 return PCI_ERS_RESULT_NEED_RESET;
5297 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5298 * @pdev: PCI device information struct
5300 * Called to determine if the driver can recover from the PCI slot reset by
5301 * using a register read to determine if the device is recoverable.
5303 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5305 struct ice_pf *pf = pci_get_drvdata(pdev);
5306 pci_ers_result_t result;
5310 err = pci_enable_device_mem(pdev);
5312 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5314 result = PCI_ERS_RESULT_DISCONNECT;
5316 pci_set_master(pdev);
5317 pci_restore_state(pdev);
5318 pci_save_state(pdev);
5319 pci_wake_from_d3(pdev, false);
5321 /* Check for life */
5322 reg = rd32(&pf->hw, GLGEN_RTRIG);
5324 result = PCI_ERS_RESULT_RECOVERED;
5326 result = PCI_ERS_RESULT_DISCONNECT;
5333 * ice_pci_err_resume - restart operations after PCI error recovery
5334 * @pdev: PCI device information struct
5336 * Called to allow the driver to bring things back up after PCI error and/or
5337 * reset recovery have finished
5339 static void ice_pci_err_resume(struct pci_dev *pdev)
5341 struct ice_pf *pf = pci_get_drvdata(pdev);
5344 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5349 if (test_bit(ICE_SUSPENDED, pf->state)) {
5350 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5355 ice_restore_all_vfs_msi_state(pdev);
5357 ice_do_reset(pf, ICE_RESET_PFR);
5358 ice_service_task_restart(pf);
5359 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5363 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5364 * @pdev: PCI device information struct
5366 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5368 struct ice_pf *pf = pci_get_drvdata(pdev);
5370 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5371 ice_service_task_stop(pf);
5373 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5374 set_bit(ICE_PFR_REQ, pf->state);
5375 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5381 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5382 * @pdev: PCI device information struct
5384 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5386 ice_pci_err_resume(pdev);
5389 /* ice_pci_tbl - PCI Device ID Table
5391 * Wildcard entries (PCI_ANY_ID) should come last
5392 * Last entry must be all 0s
5394 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5395 * Class, Class Mask, private data (not used) }
5397 static const struct pci_device_id ice_pci_tbl[] = {
5398 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5399 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5400 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5401 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5402 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5403 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5404 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5405 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5406 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5407 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5408 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5409 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5410 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5411 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5412 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5413 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5414 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5415 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5416 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5417 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5418 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5419 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5420 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5421 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5422 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5423 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT), 0 },
5424 /* required last entry */
5427 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5429 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5431 static const struct pci_error_handlers ice_pci_err_handler = {
5432 .error_detected = ice_pci_err_detected,
5433 .slot_reset = ice_pci_err_slot_reset,
5434 .reset_prepare = ice_pci_err_reset_prepare,
5435 .reset_done = ice_pci_err_reset_done,
5436 .resume = ice_pci_err_resume
5439 static struct pci_driver ice_driver = {
5440 .name = KBUILD_MODNAME,
5441 .id_table = ice_pci_tbl,
5443 .remove = ice_remove,
5445 .driver.pm = &ice_pm_ops,
5446 #endif /* CONFIG_PM */
5447 .shutdown = ice_shutdown,
5448 .sriov_configure = ice_sriov_configure,
5449 .err_handler = &ice_pci_err_handler
5453 * ice_module_init - Driver registration routine
5455 * ice_module_init is the first routine called when the driver is
5456 * loaded. All it does is register with the PCI subsystem.
5458 static int __init ice_module_init(void)
5462 pr_info("%s\n", ice_driver_string);
5463 pr_info("%s\n", ice_copyright);
5465 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5467 pr_err("Failed to create workqueue\n");
5471 status = pci_register_driver(&ice_driver);
5473 pr_err("failed to register PCI driver, err %d\n", status);
5474 destroy_workqueue(ice_wq);
5479 module_init(ice_module_init);
5482 * ice_module_exit - Driver exit cleanup routine
5484 * ice_module_exit is called just before the driver is removed
5487 static void __exit ice_module_exit(void)
5489 pci_unregister_driver(&ice_driver);
5490 destroy_workqueue(ice_wq);
5491 pr_info("module unloaded\n");
5493 module_exit(ice_module_exit);
5496 * ice_set_mac_address - NDO callback to set MAC address
5497 * @netdev: network interface device structure
5498 * @pi: pointer to an address structure
5500 * Returns 0 on success, negative on failure
5502 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5504 struct ice_netdev_priv *np = netdev_priv(netdev);
5505 struct ice_vsi *vsi = np->vsi;
5506 struct ice_pf *pf = vsi->back;
5507 struct ice_hw *hw = &pf->hw;
5508 struct sockaddr *addr = pi;
5509 u8 old_mac[ETH_ALEN];
5514 mac = (u8 *)addr->sa_data;
5516 if (!is_valid_ether_addr(mac))
5517 return -EADDRNOTAVAIL;
5519 if (ether_addr_equal(netdev->dev_addr, mac)) {
5520 netdev_dbg(netdev, "already using mac %pM\n", mac);
5524 if (test_bit(ICE_DOWN, pf->state) ||
5525 ice_is_reset_in_progress(pf->state)) {
5526 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5531 if (ice_chnl_dmac_fltr_cnt(pf)) {
5532 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5537 netif_addr_lock_bh(netdev);
5538 ether_addr_copy(old_mac, netdev->dev_addr);
5539 /* change the netdev's MAC address */
5540 eth_hw_addr_set(netdev, mac);
5541 netif_addr_unlock_bh(netdev);
5543 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5544 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5545 if (err && err != -ENOENT) {
5546 err = -EADDRNOTAVAIL;
5547 goto err_update_filters;
5550 /* Add filter for new MAC. If filter exists, return success */
5551 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5552 if (err == -EEXIST) {
5553 /* Although this MAC filter is already present in hardware it's
5554 * possible in some cases (e.g. bonding) that dev_addr was
5555 * modified outside of the driver and needs to be restored back
5558 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5562 /* error if the new filter addition failed */
5563 err = -EADDRNOTAVAIL;
5568 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5570 netif_addr_lock_bh(netdev);
5571 eth_hw_addr_set(netdev, old_mac);
5572 netif_addr_unlock_bh(netdev);
5576 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5579 /* write new MAC address to the firmware */
5580 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5581 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5583 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5590 * ice_set_rx_mode - NDO callback to set the netdev filters
5591 * @netdev: network interface device structure
5593 static void ice_set_rx_mode(struct net_device *netdev)
5595 struct ice_netdev_priv *np = netdev_priv(netdev);
5596 struct ice_vsi *vsi = np->vsi;
5601 /* Set the flags to synchronize filters
5602 * ndo_set_rx_mode may be triggered even without a change in netdev
5605 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5606 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5607 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5609 /* schedule our worker thread which will take care of
5610 * applying the new filter changes
5612 ice_service_task_schedule(vsi->back);
5616 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5617 * @netdev: network interface device structure
5618 * @queue_index: Queue ID
5619 * @maxrate: maximum bandwidth in Mbps
5622 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5624 struct ice_netdev_priv *np = netdev_priv(netdev);
5625 struct ice_vsi *vsi = np->vsi;
5630 /* Validate maxrate requested is within permitted range */
5631 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5632 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5633 maxrate, queue_index);
5637 q_handle = vsi->tx_rings[queue_index]->q_handle;
5638 tc = ice_dcb_get_tc(vsi, queue_index);
5640 /* Set BW back to default, when user set maxrate to 0 */
5642 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5643 q_handle, ICE_MAX_BW);
5645 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5646 q_handle, ICE_MAX_BW, maxrate * 1000);
5648 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5655 * ice_fdb_add - add an entry to the hardware database
5656 * @ndm: the input from the stack
5657 * @tb: pointer to array of nladdr (unused)
5658 * @dev: the net device pointer
5659 * @addr: the MAC address entry being added
5661 * @flags: instructions from stack about fdb operation
5662 * @extack: netlink extended ack
5665 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5666 struct net_device *dev, const unsigned char *addr, u16 vid,
5667 u16 flags, struct netlink_ext_ack __always_unused *extack)
5672 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5675 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5676 netdev_err(dev, "FDB only supports static addresses\n");
5680 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5681 err = dev_uc_add_excl(dev, addr);
5682 else if (is_multicast_ether_addr(addr))
5683 err = dev_mc_add_excl(dev, addr);
5687 /* Only return duplicate errors if NLM_F_EXCL is set */
5688 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5695 * ice_fdb_del - delete an entry from the hardware database
5696 * @ndm: the input from the stack
5697 * @tb: pointer to array of nladdr (unused)
5698 * @dev: the net device pointer
5699 * @addr: the MAC address entry being added
5701 * @extack: netlink extended ack
5704 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5705 struct net_device *dev, const unsigned char *addr,
5706 __always_unused u16 vid, struct netlink_ext_ack *extack)
5710 if (ndm->ndm_state & NUD_PERMANENT) {
5711 netdev_err(dev, "FDB only supports static addresses\n");
5715 if (is_unicast_ether_addr(addr))
5716 err = dev_uc_del(dev, addr);
5717 else if (is_multicast_ether_addr(addr))
5718 err = dev_mc_del(dev, addr);
5725 #define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
5726 NETIF_F_HW_VLAN_CTAG_TX | \
5727 NETIF_F_HW_VLAN_STAG_RX | \
5728 NETIF_F_HW_VLAN_STAG_TX)
5730 #define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
5731 NETIF_F_HW_VLAN_STAG_FILTER)
5734 * ice_fix_features - fix the netdev features flags based on device limitations
5735 * @netdev: ptr to the netdev that flags are being fixed on
5736 * @features: features that need to be checked and possibly fixed
5738 * Make sure any fixups are made to features in this callback. This enables the
5739 * driver to not have to check unsupported configurations throughout the driver
5740 * because that's the responsiblity of this callback.
5742 * Single VLAN Mode (SVM) Supported Features:
5743 * NETIF_F_HW_VLAN_CTAG_FILTER
5744 * NETIF_F_HW_VLAN_CTAG_RX
5745 * NETIF_F_HW_VLAN_CTAG_TX
5747 * Double VLAN Mode (DVM) Supported Features:
5748 * NETIF_F_HW_VLAN_CTAG_FILTER
5749 * NETIF_F_HW_VLAN_CTAG_RX
5750 * NETIF_F_HW_VLAN_CTAG_TX
5752 * NETIF_F_HW_VLAN_STAG_FILTER
5753 * NETIF_HW_VLAN_STAG_RX
5754 * NETIF_HW_VLAN_STAG_TX
5756 * Features that need fixing:
5757 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
5758 * These are mutually exlusive as the VSI context cannot support multiple
5759 * VLAN ethertypes simultaneously for stripping and/or insertion. If this
5760 * is not done, then default to clearing the requested STAG offload
5763 * All supported filtering has to be enabled or disabled together. For
5764 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled
5765 * together. If this is not done, then default to VLAN filtering disabled.
5766 * These are mutually exclusive as there is currently no way to
5767 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN
5770 static netdev_features_t
5771 ice_fix_features(struct net_device *netdev, netdev_features_t features)
5773 struct ice_netdev_priv *np = netdev_priv(netdev);
5774 netdev_features_t req_vlan_fltr, cur_vlan_fltr;
5775 bool cur_ctag, cur_stag, req_ctag, req_stag;
5777 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
5778 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5779 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
5781 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
5782 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5783 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
5785 if (req_vlan_fltr != cur_vlan_fltr) {
5786 if (ice_is_dvm_ena(&np->vsi->back->hw)) {
5787 if (req_ctag && req_stag) {
5788 features |= NETIF_VLAN_FILTERING_FEATURES;
5789 } else if (!req_ctag && !req_stag) {
5790 features &= ~NETIF_VLAN_FILTERING_FEATURES;
5791 } else if ((!cur_ctag && req_ctag && !cur_stag) ||
5792 (!cur_stag && req_stag && !cur_ctag)) {
5793 features |= NETIF_VLAN_FILTERING_FEATURES;
5794 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");
5795 } else if ((cur_ctag && !req_ctag && cur_stag) ||
5796 (cur_stag && !req_stag && cur_ctag)) {
5797 features &= ~NETIF_VLAN_FILTERING_FEATURES;
5798 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");
5801 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
5802 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
5804 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
5805 features |= NETIF_F_HW_VLAN_CTAG_FILTER;
5809 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
5810 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
5811 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");
5812 features &= ~(NETIF_F_HW_VLAN_STAG_RX |
5813 NETIF_F_HW_VLAN_STAG_TX);
5820 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
5822 * @features: features used to determine VLAN offload settings
5824 * First, determine the vlan_ethertype based on the VLAN offload bits in
5825 * features. Then determine if stripping and insertion should be enabled or
5826 * disabled. Finally enable or disable VLAN stripping and insertion.
5829 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
5831 bool enable_stripping = true, enable_insertion = true;
5832 struct ice_vsi_vlan_ops *vlan_ops;
5833 int strip_err = 0, insert_err = 0;
5834 u16 vlan_ethertype = 0;
5836 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5838 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
5839 vlan_ethertype = ETH_P_8021AD;
5840 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
5841 vlan_ethertype = ETH_P_8021Q;
5843 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
5844 enable_stripping = false;
5845 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
5846 enable_insertion = false;
5848 if (enable_stripping)
5849 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
5851 strip_err = vlan_ops->dis_stripping(vsi);
5853 if (enable_insertion)
5854 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
5856 insert_err = vlan_ops->dis_insertion(vsi);
5858 if (strip_err || insert_err)
5865 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
5867 * @features: features used to determine VLAN filtering settings
5869 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
5873 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
5875 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5878 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
5879 * if either bit is set
5882 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
5883 err = vlan_ops->ena_rx_filtering(vsi);
5885 err = vlan_ops->dis_rx_filtering(vsi);
5891 * ice_set_vlan_features - set VLAN settings based on suggested feature set
5892 * @netdev: ptr to the netdev being adjusted
5893 * @features: the feature set that the stack is suggesting
5895 * Only update VLAN settings if the requested_vlan_features are different than
5896 * the current_vlan_features.
5899 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
5901 netdev_features_t current_vlan_features, requested_vlan_features;
5902 struct ice_netdev_priv *np = netdev_priv(netdev);
5903 struct ice_vsi *vsi = np->vsi;
5906 current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
5907 requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
5908 if (current_vlan_features ^ requested_vlan_features) {
5909 err = ice_set_vlan_offload_features(vsi, features);
5914 current_vlan_features = netdev->features &
5915 NETIF_VLAN_FILTERING_FEATURES;
5916 requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
5917 if (current_vlan_features ^ requested_vlan_features) {
5918 err = ice_set_vlan_filtering_features(vsi, features);
5927 * ice_set_loopback - turn on/off loopback mode on underlying PF
5929 * @ena: flag to indicate the on/off setting
5931 static int ice_set_loopback(struct ice_vsi *vsi, bool ena)
5933 bool if_running = netif_running(vsi->netdev);
5936 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
5937 ret = ice_down(vsi);
5939 netdev_err(vsi->netdev, "Preparing device to toggle loopback failed\n");
5943 ret = ice_aq_set_mac_loopback(&vsi->back->hw, ena, NULL);
5945 netdev_err(vsi->netdev, "Failed to toggle loopback state\n");
5953 * ice_set_features - set the netdev feature flags
5954 * @netdev: ptr to the netdev being adjusted
5955 * @features: the feature set that the stack is suggesting
5958 ice_set_features(struct net_device *netdev, netdev_features_t features)
5960 netdev_features_t changed = netdev->features ^ features;
5961 struct ice_netdev_priv *np = netdev_priv(netdev);
5962 struct ice_vsi *vsi = np->vsi;
5963 struct ice_pf *pf = vsi->back;
5966 /* Don't set any netdev advanced features with device in Safe Mode */
5967 if (ice_is_safe_mode(pf)) {
5968 dev_err(ice_pf_to_dev(pf),
5969 "Device is in Safe Mode - not enabling advanced netdev features\n");
5973 /* Do not change setting during reset */
5974 if (ice_is_reset_in_progress(pf->state)) {
5975 dev_err(ice_pf_to_dev(pf),
5976 "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5980 /* Multiple features can be changed in one call so keep features in
5981 * separate if/else statements to guarantee each feature is checked
5983 if (changed & NETIF_F_RXHASH)
5984 ice_vsi_manage_rss_lut(vsi, !!(features & NETIF_F_RXHASH));
5986 ret = ice_set_vlan_features(netdev, features);
5990 if (changed & NETIF_F_NTUPLE) {
5991 bool ena = !!(features & NETIF_F_NTUPLE);
5993 ice_vsi_manage_fdir(vsi, ena);
5994 ena ? ice_init_arfs(vsi) : ice_clear_arfs(vsi);
5997 /* don't turn off hw_tc_offload when ADQ is already enabled */
5998 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
5999 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
6003 if (changed & NETIF_F_HW_TC) {
6004 bool ena = !!(features & NETIF_F_HW_TC);
6006 ena ? set_bit(ICE_FLAG_CLS_FLOWER, pf->flags) :
6007 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6010 if (changed & NETIF_F_LOOPBACK)
6011 ret = ice_set_loopback(vsi, !!(features & NETIF_F_LOOPBACK));
6017 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6018 * @vsi: VSI to setup VLAN properties for
6020 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6024 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6028 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6032 return ice_vsi_add_vlan_zero(vsi);
6036 * ice_vsi_cfg - Setup the VSI
6037 * @vsi: the VSI being configured
6039 * Return 0 on success and negative value on error
6041 int ice_vsi_cfg(struct ice_vsi *vsi)
6046 ice_set_rx_mode(vsi->netdev);
6048 if (vsi->type != ICE_VSI_LB) {
6049 err = ice_vsi_vlan_setup(vsi);
6055 ice_vsi_cfg_dcb_rings(vsi);
6057 err = ice_vsi_cfg_lan_txqs(vsi);
6058 if (!err && ice_is_xdp_ena_vsi(vsi))
6059 err = ice_vsi_cfg_xdp_txqs(vsi);
6061 err = ice_vsi_cfg_rxqs(vsi);
6066 /* THEORY OF MODERATION:
6067 * The ice driver hardware works differently than the hardware that DIMLIB was
6068 * originally made for. ice hardware doesn't have packet count limits that
6069 * can trigger an interrupt, but it *does* have interrupt rate limit support,
6070 * which is hard-coded to a limit of 250,000 ints/second.
6071 * If not using dynamic moderation, the INTRL value can be modified
6072 * by ethtool rx-usecs-high.
6075 /* the throttle rate for interrupts, basically worst case delay before
6076 * an initial interrupt fires, value is stored in microseconds.
6081 /* Make a different profile for Rx that doesn't allow quite so aggressive
6082 * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6085 static const struct ice_dim rx_profile[] = {
6086 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6087 {8}, /* 125,000 ints/s */
6088 {16}, /* 62,500 ints/s */
6089 {62}, /* 16,129 ints/s */
6090 {126} /* 7,936 ints/s */
6093 /* The transmit profile, which has the same sorts of values
6094 * as the previous struct
6096 static const struct ice_dim tx_profile[] = {
6097 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6098 {8}, /* 125,000 ints/s */
6099 {40}, /* 16,125 ints/s */
6100 {128}, /* 7,812 ints/s */
6101 {256} /* 3,906 ints/s */
6104 static void ice_tx_dim_work(struct work_struct *work)
6106 struct ice_ring_container *rc;
6110 dim = container_of(work, struct dim, work);
6111 rc = (struct ice_ring_container *)dim->priv;
6113 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6115 /* look up the values in our local table */
6116 itr = tx_profile[dim->profile_ix].itr;
6118 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6119 ice_write_itr(rc, itr);
6121 dim->state = DIM_START_MEASURE;
6124 static void ice_rx_dim_work(struct work_struct *work)
6126 struct ice_ring_container *rc;
6130 dim = container_of(work, struct dim, work);
6131 rc = (struct ice_ring_container *)dim->priv;
6133 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6135 /* look up the values in our local table */
6136 itr = rx_profile[dim->profile_ix].itr;
6138 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6139 ice_write_itr(rc, itr);
6141 dim->state = DIM_START_MEASURE;
6144 #define ICE_DIM_DEFAULT_PROFILE_IX 1
6147 * ice_init_moderation - set up interrupt moderation
6148 * @q_vector: the vector containing rings to be configured
6150 * Set up interrupt moderation registers, with the intent to do the right thing
6151 * when called from reset or from probe, and whether or not dynamic moderation
6152 * is enabled or not. Take special care to write all the registers in both
6153 * dynamic moderation mode or not in order to make sure hardware is in a known
6156 static void ice_init_moderation(struct ice_q_vector *q_vector)
6158 struct ice_ring_container *rc;
6159 bool tx_dynamic, rx_dynamic;
6162 INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6163 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6164 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6166 tx_dynamic = ITR_IS_DYNAMIC(rc);
6168 /* set the initial TX ITR to match the above */
6169 ice_write_itr(rc, tx_dynamic ?
6170 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6173 INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6174 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6175 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6177 rx_dynamic = ITR_IS_DYNAMIC(rc);
6179 /* set the initial RX ITR to match the above */
6180 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6183 ice_set_q_vector_intrl(q_vector);
6187 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6188 * @vsi: the VSI being configured
6190 static void ice_napi_enable_all(struct ice_vsi *vsi)
6197 ice_for_each_q_vector(vsi, q_idx) {
6198 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6200 ice_init_moderation(q_vector);
6202 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6203 napi_enable(&q_vector->napi);
6208 * ice_up_complete - Finish the last steps of bringing up a connection
6209 * @vsi: The VSI being configured
6211 * Return 0 on success and negative value on error
6213 static int ice_up_complete(struct ice_vsi *vsi)
6215 struct ice_pf *pf = vsi->back;
6218 ice_vsi_cfg_msix(vsi);
6220 /* Enable only Rx rings, Tx rings were enabled by the FW when the
6221 * Tx queue group list was configured and the context bits were
6222 * programmed using ice_vsi_cfg_txqs
6224 err = ice_vsi_start_all_rx_rings(vsi);
6228 clear_bit(ICE_VSI_DOWN, vsi->state);
6229 ice_napi_enable_all(vsi);
6230 ice_vsi_ena_irq(vsi);
6232 if (vsi->port_info &&
6233 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6235 ice_print_link_msg(vsi, true);
6236 netif_tx_start_all_queues(vsi->netdev);
6237 netif_carrier_on(vsi->netdev);
6238 if (!ice_is_e810(&pf->hw))
6239 ice_ptp_link_change(pf, pf->hw.pf_id, true);
6242 /* Perform an initial read of the statistics registers now to
6243 * set the baseline so counters are ready when interface is up
6245 ice_update_eth_stats(vsi);
6246 ice_service_task_schedule(pf);
6252 * ice_up - Bring the connection back up after being down
6253 * @vsi: VSI being configured
6255 int ice_up(struct ice_vsi *vsi)
6259 err = ice_vsi_cfg(vsi);
6261 err = ice_up_complete(vsi);
6267 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6268 * @syncp: pointer to u64_stats_sync
6269 * @stats: stats that pkts and bytes count will be taken from
6270 * @pkts: packets stats counter
6271 * @bytes: bytes stats counter
6273 * This function fetches stats from the ring considering the atomic operations
6274 * that needs to be performed to read u64 values in 32 bit machine.
6277 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6278 struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6283 start = u64_stats_fetch_begin_irq(syncp);
6285 *bytes = stats.bytes;
6286 } while (u64_stats_fetch_retry_irq(syncp, start));
6290 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6291 * @vsi: the VSI to be updated
6292 * @vsi_stats: the stats struct to be updated
6293 * @rings: rings to work on
6294 * @count: number of rings
6297 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6298 struct rtnl_link_stats64 *vsi_stats,
6299 struct ice_tx_ring **rings, u16 count)
6303 for (i = 0; i < count; i++) {
6304 struct ice_tx_ring *ring;
6305 u64 pkts = 0, bytes = 0;
6307 ring = READ_ONCE(rings[i]);
6310 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6311 vsi_stats->tx_packets += pkts;
6312 vsi_stats->tx_bytes += bytes;
6313 vsi->tx_restart += ring->tx_stats.restart_q;
6314 vsi->tx_busy += ring->tx_stats.tx_busy;
6315 vsi->tx_linearize += ring->tx_stats.tx_linearize;
6320 * ice_update_vsi_ring_stats - Update VSI stats counters
6321 * @vsi: the VSI to be updated
6323 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6325 struct rtnl_link_stats64 *vsi_stats;
6329 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6333 /* reset non-netdev (extended) stats */
6334 vsi->tx_restart = 0;
6336 vsi->tx_linearize = 0;
6337 vsi->rx_buf_failed = 0;
6338 vsi->rx_page_failed = 0;
6342 /* update Tx rings counters */
6343 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6346 /* update Rx rings counters */
6347 ice_for_each_rxq(vsi, i) {
6348 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6350 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6351 vsi_stats->rx_packets += pkts;
6352 vsi_stats->rx_bytes += bytes;
6353 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
6354 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
6357 /* update XDP Tx rings counters */
6358 if (ice_is_xdp_ena_vsi(vsi))
6359 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6364 vsi->net_stats.tx_packets = vsi_stats->tx_packets;
6365 vsi->net_stats.tx_bytes = vsi_stats->tx_bytes;
6366 vsi->net_stats.rx_packets = vsi_stats->rx_packets;
6367 vsi->net_stats.rx_bytes = vsi_stats->rx_bytes;
6373 * ice_update_vsi_stats - Update VSI stats counters
6374 * @vsi: the VSI to be updated
6376 void ice_update_vsi_stats(struct ice_vsi *vsi)
6378 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6379 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6380 struct ice_pf *pf = vsi->back;
6382 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6383 test_bit(ICE_CFG_BUSY, pf->state))
6386 /* get stats as recorded by Tx/Rx rings */
6387 ice_update_vsi_ring_stats(vsi);
6389 /* get VSI stats as recorded by the hardware */
6390 ice_update_eth_stats(vsi);
6392 cur_ns->tx_errors = cur_es->tx_errors;
6393 cur_ns->rx_dropped = cur_es->rx_discards;
6394 cur_ns->tx_dropped = cur_es->tx_discards;
6395 cur_ns->multicast = cur_es->rx_multicast;
6397 /* update some more netdev stats if this is main VSI */
6398 if (vsi->type == ICE_VSI_PF) {
6399 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6400 cur_ns->rx_errors = pf->stats.crc_errors +
6401 pf->stats.illegal_bytes +
6402 pf->stats.rx_len_errors +
6403 pf->stats.rx_undersize +
6404 pf->hw_csum_rx_error +
6405 pf->stats.rx_jabber +
6406 pf->stats.rx_fragments +
6407 pf->stats.rx_oversize;
6408 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6409 /* record drops from the port level */
6410 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6415 * ice_update_pf_stats - Update PF port stats counters
6416 * @pf: PF whose stats needs to be updated
6418 void ice_update_pf_stats(struct ice_pf *pf)
6420 struct ice_hw_port_stats *prev_ps, *cur_ps;
6421 struct ice_hw *hw = &pf->hw;
6425 port = hw->port_info->lport;
6426 prev_ps = &pf->stats_prev;
6427 cur_ps = &pf->stats;
6429 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6430 &prev_ps->eth.rx_bytes,
6431 &cur_ps->eth.rx_bytes);
6433 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6434 &prev_ps->eth.rx_unicast,
6435 &cur_ps->eth.rx_unicast);
6437 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6438 &prev_ps->eth.rx_multicast,
6439 &cur_ps->eth.rx_multicast);
6441 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6442 &prev_ps->eth.rx_broadcast,
6443 &cur_ps->eth.rx_broadcast);
6445 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6446 &prev_ps->eth.rx_discards,
6447 &cur_ps->eth.rx_discards);
6449 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6450 &prev_ps->eth.tx_bytes,
6451 &cur_ps->eth.tx_bytes);
6453 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6454 &prev_ps->eth.tx_unicast,
6455 &cur_ps->eth.tx_unicast);
6457 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6458 &prev_ps->eth.tx_multicast,
6459 &cur_ps->eth.tx_multicast);
6461 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6462 &prev_ps->eth.tx_broadcast,
6463 &cur_ps->eth.tx_broadcast);
6465 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6466 &prev_ps->tx_dropped_link_down,
6467 &cur_ps->tx_dropped_link_down);
6469 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6470 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6472 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6473 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6475 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6476 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6478 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6479 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6481 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6482 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6484 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6485 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6487 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6488 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6490 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6491 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6493 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6494 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6496 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6497 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6499 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6500 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6502 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6503 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6505 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6506 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6508 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6509 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6511 fd_ctr_base = hw->fd_ctr_base;
6513 ice_stat_update40(hw,
6514 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6515 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6516 &cur_ps->fd_sb_match);
6517 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6518 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6520 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6521 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6523 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6524 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6526 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6527 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6529 ice_update_dcb_stats(pf);
6531 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6532 &prev_ps->crc_errors, &cur_ps->crc_errors);
6534 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6535 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6537 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6538 &prev_ps->mac_local_faults,
6539 &cur_ps->mac_local_faults);
6541 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6542 &prev_ps->mac_remote_faults,
6543 &cur_ps->mac_remote_faults);
6545 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6546 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6548 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6549 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6551 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6552 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6554 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6555 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6557 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6558 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6560 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6562 pf->stat_prev_loaded = true;
6566 * ice_get_stats64 - get statistics for network device structure
6567 * @netdev: network interface device structure
6568 * @stats: main device statistics structure
6571 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6573 struct ice_netdev_priv *np = netdev_priv(netdev);
6574 struct rtnl_link_stats64 *vsi_stats;
6575 struct ice_vsi *vsi = np->vsi;
6577 vsi_stats = &vsi->net_stats;
6579 if (!vsi->num_txq || !vsi->num_rxq)
6582 /* netdev packet/byte stats come from ring counter. These are obtained
6583 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6584 * But, only call the update routine and read the registers if VSI is
6587 if (!test_bit(ICE_VSI_DOWN, vsi->state))
6588 ice_update_vsi_ring_stats(vsi);
6589 stats->tx_packets = vsi_stats->tx_packets;
6590 stats->tx_bytes = vsi_stats->tx_bytes;
6591 stats->rx_packets = vsi_stats->rx_packets;
6592 stats->rx_bytes = vsi_stats->rx_bytes;
6594 /* The rest of the stats can be read from the hardware but instead we
6595 * just return values that the watchdog task has already obtained from
6598 stats->multicast = vsi_stats->multicast;
6599 stats->tx_errors = vsi_stats->tx_errors;
6600 stats->tx_dropped = vsi_stats->tx_dropped;
6601 stats->rx_errors = vsi_stats->rx_errors;
6602 stats->rx_dropped = vsi_stats->rx_dropped;
6603 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6604 stats->rx_length_errors = vsi_stats->rx_length_errors;
6608 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6609 * @vsi: VSI having NAPI disabled
6611 static void ice_napi_disable_all(struct ice_vsi *vsi)
6618 ice_for_each_q_vector(vsi, q_idx) {
6619 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6621 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6622 napi_disable(&q_vector->napi);
6624 cancel_work_sync(&q_vector->tx.dim.work);
6625 cancel_work_sync(&q_vector->rx.dim.work);
6630 * ice_down - Shutdown the connection
6631 * @vsi: The VSI being stopped
6633 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
6635 int ice_down(struct ice_vsi *vsi)
6637 int i, tx_err, rx_err, link_err = 0, vlan_err = 0;
6639 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
6641 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6642 vlan_err = ice_vsi_del_vlan_zero(vsi);
6643 if (!ice_is_e810(&vsi->back->hw))
6644 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
6645 netif_carrier_off(vsi->netdev);
6646 netif_tx_disable(vsi->netdev);
6647 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
6648 ice_eswitch_stop_all_tx_queues(vsi->back);
6651 ice_vsi_dis_irq(vsi);
6653 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6655 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6656 vsi->vsi_num, tx_err);
6657 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6658 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6660 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6661 vsi->vsi_num, tx_err);
6664 rx_err = ice_vsi_stop_all_rx_rings(vsi);
6666 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6667 vsi->vsi_num, rx_err);
6669 ice_napi_disable_all(vsi);
6671 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6672 link_err = ice_force_phys_link_state(vsi, false);
6674 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6675 vsi->vsi_num, link_err);
6678 ice_for_each_txq(vsi, i)
6679 ice_clean_tx_ring(vsi->tx_rings[i]);
6681 ice_for_each_rxq(vsi, i)
6682 ice_clean_rx_ring(vsi->rx_rings[i]);
6684 if (tx_err || rx_err || link_err || vlan_err) {
6685 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6686 vsi->vsi_num, vsi->vsw->sw_id);
6694 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6695 * @vsi: VSI having resources allocated
6697 * Return 0 on success, negative on failure
6699 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6703 if (!vsi->num_txq) {
6704 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6709 ice_for_each_txq(vsi, i) {
6710 struct ice_tx_ring *ring = vsi->tx_rings[i];
6716 ring->netdev = vsi->netdev;
6717 err = ice_setup_tx_ring(ring);
6726 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6727 * @vsi: VSI having resources allocated
6729 * Return 0 on success, negative on failure
6731 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6735 if (!vsi->num_rxq) {
6736 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6741 ice_for_each_rxq(vsi, i) {
6742 struct ice_rx_ring *ring = vsi->rx_rings[i];
6748 ring->netdev = vsi->netdev;
6749 err = ice_setup_rx_ring(ring);
6758 * ice_vsi_open_ctrl - open control VSI for use
6759 * @vsi: the VSI to open
6761 * Initialization of the Control VSI
6763 * Returns 0 on success, negative value on error
6765 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6767 char int_name[ICE_INT_NAME_STR_LEN];
6768 struct ice_pf *pf = vsi->back;
6772 dev = ice_pf_to_dev(pf);
6773 /* allocate descriptors */
6774 err = ice_vsi_setup_tx_rings(vsi);
6778 err = ice_vsi_setup_rx_rings(vsi);
6782 err = ice_vsi_cfg(vsi);
6786 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6787 dev_driver_string(dev), dev_name(dev));
6788 err = ice_vsi_req_irq_msix(vsi, int_name);
6792 ice_vsi_cfg_msix(vsi);
6794 err = ice_vsi_start_all_rx_rings(vsi);
6796 goto err_up_complete;
6798 clear_bit(ICE_VSI_DOWN, vsi->state);
6799 ice_vsi_ena_irq(vsi);
6806 ice_vsi_free_rx_rings(vsi);
6808 ice_vsi_free_tx_rings(vsi);
6814 * ice_vsi_open - Called when a network interface is made active
6815 * @vsi: the VSI to open
6817 * Initialization of the VSI
6819 * Returns 0 on success, negative value on error
6821 int ice_vsi_open(struct ice_vsi *vsi)
6823 char int_name[ICE_INT_NAME_STR_LEN];
6824 struct ice_pf *pf = vsi->back;
6827 /* allocate descriptors */
6828 err = ice_vsi_setup_tx_rings(vsi);
6832 err = ice_vsi_setup_rx_rings(vsi);
6836 err = ice_vsi_cfg(vsi);
6840 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6841 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6842 err = ice_vsi_req_irq_msix(vsi, int_name);
6846 if (vsi->type == ICE_VSI_PF) {
6847 /* Notify the stack of the actual queue counts. */
6848 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6852 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6857 err = ice_up_complete(vsi);
6859 goto err_up_complete;
6866 ice_vsi_free_irq(vsi);
6868 ice_vsi_free_rx_rings(vsi);
6870 ice_vsi_free_tx_rings(vsi);
6876 * ice_vsi_release_all - Delete all VSIs
6877 * @pf: PF from which all VSIs are being removed
6879 static void ice_vsi_release_all(struct ice_pf *pf)
6886 ice_for_each_vsi(pf, i) {
6890 if (pf->vsi[i]->type == ICE_VSI_CHNL)
6893 err = ice_vsi_release(pf->vsi[i]);
6895 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6896 i, err, pf->vsi[i]->vsi_num);
6901 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6902 * @pf: pointer to the PF instance
6903 * @type: VSI type to rebuild
6905 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6907 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6909 struct device *dev = ice_pf_to_dev(pf);
6912 ice_for_each_vsi(pf, i) {
6913 struct ice_vsi *vsi = pf->vsi[i];
6915 if (!vsi || vsi->type != type)
6918 /* rebuild the VSI */
6919 err = ice_vsi_rebuild(vsi, true);
6921 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6922 err, vsi->idx, ice_vsi_type_str(type));
6926 /* replay filters for the VSI */
6927 err = ice_replay_vsi(&pf->hw, vsi->idx);
6929 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
6930 err, vsi->idx, ice_vsi_type_str(type));
6934 /* Re-map HW VSI number, using VSI handle that has been
6935 * previously validated in ice_replay_vsi() call above
6937 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6939 /* enable the VSI */
6940 err = ice_ena_vsi(vsi, false);
6942 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6943 err, vsi->idx, ice_vsi_type_str(type));
6947 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6948 ice_vsi_type_str(type));
6955 * ice_update_pf_netdev_link - Update PF netdev link status
6956 * @pf: pointer to the PF instance
6958 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6963 ice_for_each_vsi(pf, i) {
6964 struct ice_vsi *vsi = pf->vsi[i];
6966 if (!vsi || vsi->type != ICE_VSI_PF)
6969 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6971 netif_carrier_on(pf->vsi[i]->netdev);
6972 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6974 netif_carrier_off(pf->vsi[i]->netdev);
6975 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6981 * ice_rebuild - rebuild after reset
6982 * @pf: PF to rebuild
6983 * @reset_type: type of reset
6985 * Do not rebuild VF VSI in this flow because that is already handled via
6986 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6987 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6988 * to reset/rebuild all the VF VSI twice.
6990 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6992 struct device *dev = ice_pf_to_dev(pf);
6993 struct ice_hw *hw = &pf->hw;
6997 if (test_bit(ICE_DOWN, pf->state))
6998 goto clear_recovery;
7000 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
7002 #define ICE_EMP_RESET_SLEEP_MS 5000
7003 if (reset_type == ICE_RESET_EMPR) {
7004 /* If an EMP reset has occurred, any previously pending flash
7005 * update will have completed. We no longer know whether or
7006 * not the NVM update EMP reset is restricted.
7008 pf->fw_emp_reset_disabled = false;
7010 msleep(ICE_EMP_RESET_SLEEP_MS);
7013 err = ice_init_all_ctrlq(hw);
7015 dev_err(dev, "control queues init failed %d\n", err);
7016 goto err_init_ctrlq;
7019 /* if DDP was previously loaded successfully */
7020 if (!ice_is_safe_mode(pf)) {
7021 /* reload the SW DB of filter tables */
7022 if (reset_type == ICE_RESET_PFR)
7023 ice_fill_blk_tbls(hw);
7025 /* Reload DDP Package after CORER/GLOBR reset */
7026 ice_load_pkg(NULL, pf);
7029 err = ice_clear_pf_cfg(hw);
7031 dev_err(dev, "clear PF configuration failed %d\n", err);
7032 goto err_init_ctrlq;
7035 ice_clear_pxe_mode(hw);
7037 err = ice_init_nvm(hw);
7039 dev_err(dev, "ice_init_nvm failed %d\n", err);
7040 goto err_init_ctrlq;
7043 err = ice_get_caps(hw);
7045 dev_err(dev, "ice_get_caps failed %d\n", err);
7046 goto err_init_ctrlq;
7049 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7051 dev_err(dev, "set_mac_cfg failed %d\n", err);
7052 goto err_init_ctrlq;
7055 dvm = ice_is_dvm_ena(hw);
7057 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7059 goto err_init_ctrlq;
7061 err = ice_sched_init_port(hw->port_info);
7063 goto err_sched_init_port;
7065 /* start misc vector */
7066 err = ice_req_irq_msix_misc(pf);
7068 dev_err(dev, "misc vector setup failed: %d\n", err);
7069 goto err_sched_init_port;
7072 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7073 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7074 if (!rd32(hw, PFQF_FD_SIZE)) {
7075 u16 unused, guar, b_effort;
7077 guar = hw->func_caps.fd_fltr_guar;
7078 b_effort = hw->func_caps.fd_fltr_best_effort;
7080 /* force guaranteed filter pool for PF */
7081 ice_alloc_fd_guar_item(hw, &unused, guar);
7082 /* force shared filter pool for PF */
7083 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7087 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7088 ice_dcb_rebuild(pf);
7090 /* If the PF previously had enabled PTP, PTP init needs to happen before
7091 * the VSI rebuild. If not, this causes the PTP link status events to
7094 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7097 if (ice_is_feature_supported(pf, ICE_F_GNSS))
7100 /* rebuild PF VSI */
7101 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7103 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7104 goto err_vsi_rebuild;
7107 /* configure PTP timestamping after VSI rebuild */
7108 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7109 ice_ptp_cfg_timestamp(pf, false);
7111 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
7113 dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
7114 goto err_vsi_rebuild;
7117 if (reset_type == ICE_RESET_PFR) {
7118 err = ice_rebuild_channels(pf);
7120 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7122 goto err_vsi_rebuild;
7126 /* If Flow Director is active */
7127 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7128 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7130 dev_err(dev, "control VSI rebuild failed: %d\n", err);
7131 goto err_vsi_rebuild;
7134 /* replay HW Flow Director recipes */
7136 ice_fdir_replay_flows(hw);
7138 /* replay Flow Director filters */
7139 ice_fdir_replay_fltrs(pf);
7141 ice_rebuild_arfs(pf);
7144 ice_update_pf_netdev_link(pf);
7146 /* tell the firmware we are up */
7147 err = ice_send_version(pf);
7149 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7151 goto err_vsi_rebuild;
7154 ice_replay_post(hw);
7156 /* if we get here, reset flow is successful */
7157 clear_bit(ICE_RESET_FAILED, pf->state);
7159 ice_plug_aux_dev(pf);
7163 err_sched_init_port:
7164 ice_sched_cleanup_all(hw);
7166 ice_shutdown_all_ctrlq(hw);
7167 set_bit(ICE_RESET_FAILED, pf->state);
7169 /* set this bit in PF state to control service task scheduling */
7170 set_bit(ICE_NEEDS_RESTART, pf->state);
7171 dev_err(dev, "Rebuild failed, unload and reload driver\n");
7175 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
7176 * @vsi: Pointer to VSI structure
7178 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
7180 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
7181 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
7183 return ICE_RXBUF_3072;
7187 * ice_change_mtu - NDO callback to change the MTU
7188 * @netdev: network interface device structure
7189 * @new_mtu: new value for maximum frame size
7191 * Returns 0 on success, negative on failure
7193 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7195 struct ice_netdev_priv *np = netdev_priv(netdev);
7196 struct ice_vsi *vsi = np->vsi;
7197 struct ice_pf *pf = vsi->back;
7201 if (new_mtu == (int)netdev->mtu) {
7202 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7206 if (ice_is_xdp_ena_vsi(vsi)) {
7207 int frame_size = ice_max_xdp_frame_size(vsi);
7209 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7210 netdev_err(netdev, "max MTU for XDP usage is %d\n",
7211 frame_size - ICE_ETH_PKT_HDR_PAD);
7216 /* if a reset is in progress, wait for some time for it to complete */
7218 if (ice_is_reset_in_progress(pf->state)) {
7220 usleep_range(1000, 2000);
7225 } while (count < 100);
7228 netdev_err(netdev, "can't change MTU. Device is busy\n");
7232 netdev->mtu = (unsigned int)new_mtu;
7234 /* if VSI is up, bring it down and then back up */
7235 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
7236 err = ice_down(vsi);
7238 netdev_err(netdev, "change MTU if_down err %d\n", err);
7244 netdev_err(netdev, "change MTU if_up err %d\n", err);
7249 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7250 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7256 * ice_eth_ioctl - Access the hwtstamp interface
7257 * @netdev: network interface device structure
7258 * @ifr: interface request data
7259 * @cmd: ioctl command
7261 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7263 struct ice_netdev_priv *np = netdev_priv(netdev);
7264 struct ice_pf *pf = np->vsi->back;
7268 return ice_ptp_get_ts_config(pf, ifr);
7270 return ice_ptp_set_ts_config(pf, ifr);
7277 * ice_aq_str - convert AQ err code to a string
7278 * @aq_err: the AQ error code to convert
7280 const char *ice_aq_str(enum ice_aq_err aq_err)
7285 case ICE_AQ_RC_EPERM:
7286 return "ICE_AQ_RC_EPERM";
7287 case ICE_AQ_RC_ENOENT:
7288 return "ICE_AQ_RC_ENOENT";
7289 case ICE_AQ_RC_ENOMEM:
7290 return "ICE_AQ_RC_ENOMEM";
7291 case ICE_AQ_RC_EBUSY:
7292 return "ICE_AQ_RC_EBUSY";
7293 case ICE_AQ_RC_EEXIST:
7294 return "ICE_AQ_RC_EEXIST";
7295 case ICE_AQ_RC_EINVAL:
7296 return "ICE_AQ_RC_EINVAL";
7297 case ICE_AQ_RC_ENOSPC:
7298 return "ICE_AQ_RC_ENOSPC";
7299 case ICE_AQ_RC_ENOSYS:
7300 return "ICE_AQ_RC_ENOSYS";
7301 case ICE_AQ_RC_EMODE:
7302 return "ICE_AQ_RC_EMODE";
7303 case ICE_AQ_RC_ENOSEC:
7304 return "ICE_AQ_RC_ENOSEC";
7305 case ICE_AQ_RC_EBADSIG:
7306 return "ICE_AQ_RC_EBADSIG";
7307 case ICE_AQ_RC_ESVN:
7308 return "ICE_AQ_RC_ESVN";
7309 case ICE_AQ_RC_EBADMAN:
7310 return "ICE_AQ_RC_EBADMAN";
7311 case ICE_AQ_RC_EBADBUF:
7312 return "ICE_AQ_RC_EBADBUF";
7315 return "ICE_AQ_RC_UNKNOWN";
7319 * ice_set_rss_lut - Set RSS LUT
7320 * @vsi: Pointer to VSI structure
7321 * @lut: Lookup table
7322 * @lut_size: Lookup table size
7324 * Returns 0 on success, negative on failure
7326 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7328 struct ice_aq_get_set_rss_lut_params params = {};
7329 struct ice_hw *hw = &vsi->back->hw;
7335 params.vsi_handle = vsi->idx;
7336 params.lut_size = lut_size;
7337 params.lut_type = vsi->rss_lut_type;
7340 status = ice_aq_set_rss_lut(hw, ¶ms);
7342 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7343 status, ice_aq_str(hw->adminq.sq_last_status));
7349 * ice_set_rss_key - Set RSS key
7350 * @vsi: Pointer to the VSI structure
7351 * @seed: RSS hash seed
7353 * Returns 0 on success, negative on failure
7355 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7357 struct ice_hw *hw = &vsi->back->hw;
7363 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7365 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7366 status, ice_aq_str(hw->adminq.sq_last_status));
7372 * ice_get_rss_lut - Get RSS LUT
7373 * @vsi: Pointer to VSI structure
7374 * @lut: Buffer to store the lookup table entries
7375 * @lut_size: Size of buffer to store the lookup table entries
7377 * Returns 0 on success, negative on failure
7379 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7381 struct ice_aq_get_set_rss_lut_params params = {};
7382 struct ice_hw *hw = &vsi->back->hw;
7388 params.vsi_handle = vsi->idx;
7389 params.lut_size = lut_size;
7390 params.lut_type = vsi->rss_lut_type;
7393 status = ice_aq_get_rss_lut(hw, ¶ms);
7395 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7396 status, ice_aq_str(hw->adminq.sq_last_status));
7402 * ice_get_rss_key - Get RSS key
7403 * @vsi: Pointer to VSI structure
7404 * @seed: Buffer to store the key in
7406 * Returns 0 on success, negative on failure
7408 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7410 struct ice_hw *hw = &vsi->back->hw;
7416 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7418 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7419 status, ice_aq_str(hw->adminq.sq_last_status));
7425 * ice_bridge_getlink - Get the hardware bridge mode
7428 * @seq: RTNL message seq
7429 * @dev: the netdev being configured
7430 * @filter_mask: filter mask passed in
7431 * @nlflags: netlink flags passed in
7433 * Return the bridge mode (VEB/VEPA)
7436 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7437 struct net_device *dev, u32 filter_mask, int nlflags)
7439 struct ice_netdev_priv *np = netdev_priv(dev);
7440 struct ice_vsi *vsi = np->vsi;
7441 struct ice_pf *pf = vsi->back;
7444 bmode = pf->first_sw->bridge_mode;
7446 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7451 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7452 * @vsi: Pointer to VSI structure
7453 * @bmode: Hardware bridge mode (VEB/VEPA)
7455 * Returns 0 on success, negative on failure
7457 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7459 struct ice_aqc_vsi_props *vsi_props;
7460 struct ice_hw *hw = &vsi->back->hw;
7461 struct ice_vsi_ctx *ctxt;
7464 vsi_props = &vsi->info;
7466 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7470 ctxt->info = vsi->info;
7472 if (bmode == BRIDGE_MODE_VEB)
7473 /* change from VEPA to VEB mode */
7474 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7476 /* change from VEB to VEPA mode */
7477 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7478 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7480 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7482 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7483 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7486 /* Update sw flags for book keeping */
7487 vsi_props->sw_flags = ctxt->info.sw_flags;
7495 * ice_bridge_setlink - Set the hardware bridge mode
7496 * @dev: the netdev being configured
7497 * @nlh: RTNL message
7498 * @flags: bridge setlink flags
7499 * @extack: netlink extended ack
7501 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7502 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7503 * not already set for all VSIs connected to this switch. And also update the
7504 * unicast switch filter rules for the corresponding switch of the netdev.
7507 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7508 u16 __always_unused flags,
7509 struct netlink_ext_ack __always_unused *extack)
7511 struct ice_netdev_priv *np = netdev_priv(dev);
7512 struct ice_pf *pf = np->vsi->back;
7513 struct nlattr *attr, *br_spec;
7514 struct ice_hw *hw = &pf->hw;
7515 struct ice_sw *pf_sw;
7516 int rem, v, err = 0;
7518 pf_sw = pf->first_sw;
7519 /* find the attribute in the netlink message */
7520 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7522 nla_for_each_nested(attr, br_spec, rem) {
7525 if (nla_type(attr) != IFLA_BRIDGE_MODE)
7527 mode = nla_get_u16(attr);
7528 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7530 /* Continue if bridge mode is not being flipped */
7531 if (mode == pf_sw->bridge_mode)
7533 /* Iterates through the PF VSI list and update the loopback
7536 ice_for_each_vsi(pf, v) {
7539 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7544 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7545 /* Update the unicast switch filter rules for the corresponding
7546 * switch of the netdev
7548 err = ice_update_sw_rule_bridge_mode(hw);
7550 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7552 ice_aq_str(hw->adminq.sq_last_status));
7553 /* revert hw->evb_veb */
7554 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7558 pf_sw->bridge_mode = mode;
7565 * ice_tx_timeout - Respond to a Tx Hang
7566 * @netdev: network interface device structure
7567 * @txqueue: Tx queue
7569 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7571 struct ice_netdev_priv *np = netdev_priv(netdev);
7572 struct ice_tx_ring *tx_ring = NULL;
7573 struct ice_vsi *vsi = np->vsi;
7574 struct ice_pf *pf = vsi->back;
7577 pf->tx_timeout_count++;
7579 /* Check if PFC is enabled for the TC to which the queue belongs
7580 * to. If yes then Tx timeout is not caused by a hung queue, no
7581 * need to reset and rebuild
7583 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7584 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7589 /* now that we have an index, find the tx_ring struct */
7590 ice_for_each_txq(vsi, i)
7591 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7592 if (txqueue == vsi->tx_rings[i]->q_index) {
7593 tx_ring = vsi->tx_rings[i];
7597 /* Reset recovery level if enough time has elapsed after last timeout.
7598 * Also ensure no new reset action happens before next timeout period.
7600 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7601 pf->tx_timeout_recovery_level = 1;
7602 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7603 netdev->watchdog_timeo)))
7607 struct ice_hw *hw = &pf->hw;
7610 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7611 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7612 /* Read interrupt register */
7613 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7615 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7616 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7617 head, tx_ring->next_to_use, val);
7620 pf->tx_timeout_last_recovery = jiffies;
7621 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7622 pf->tx_timeout_recovery_level, txqueue);
7624 switch (pf->tx_timeout_recovery_level) {
7626 set_bit(ICE_PFR_REQ, pf->state);
7629 set_bit(ICE_CORER_REQ, pf->state);
7632 set_bit(ICE_GLOBR_REQ, pf->state);
7635 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7636 set_bit(ICE_DOWN, pf->state);
7637 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7638 set_bit(ICE_SERVICE_DIS, pf->state);
7642 ice_service_task_schedule(pf);
7643 pf->tx_timeout_recovery_level++;
7647 * ice_setup_tc_cls_flower - flower classifier offloads
7648 * @np: net device to configure
7649 * @filter_dev: device on which filter is added
7650 * @cls_flower: offload data
7653 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
7654 struct net_device *filter_dev,
7655 struct flow_cls_offload *cls_flower)
7657 struct ice_vsi *vsi = np->vsi;
7659 if (cls_flower->common.chain_index)
7662 switch (cls_flower->command) {
7663 case FLOW_CLS_REPLACE:
7664 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
7665 case FLOW_CLS_DESTROY:
7666 return ice_del_cls_flower(vsi, cls_flower);
7673 * ice_setup_tc_block_cb - callback handler registered for TC block
7674 * @type: TC SETUP type
7675 * @type_data: TC flower offload data that contains user input
7676 * @cb_priv: netdev private data
7679 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
7681 struct ice_netdev_priv *np = cb_priv;
7684 case TC_SETUP_CLSFLOWER:
7685 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
7693 * ice_validate_mqprio_qopt - Validate TCF input parameters
7694 * @vsi: Pointer to VSI
7695 * @mqprio_qopt: input parameters for mqprio queue configuration
7697 * This function validates MQPRIO params, such as qcount (power of 2 wherever
7698 * needed), and make sure user doesn't specify qcount and BW rate limit
7699 * for TCs, which are more than "num_tc"
7702 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
7703 struct tc_mqprio_qopt_offload *mqprio_qopt)
7705 u64 sum_max_rate = 0, sum_min_rate = 0;
7706 int non_power_of_2_qcount = 0;
7707 struct ice_pf *pf = vsi->back;
7708 int max_rss_q_cnt = 0;
7713 if (vsi->type != ICE_VSI_PF)
7716 if (mqprio_qopt->qopt.offset[0] != 0 ||
7717 mqprio_qopt->qopt.num_tc < 1 ||
7718 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
7721 dev = ice_pf_to_dev(pf);
7722 vsi->ch_rss_size = 0;
7723 num_tc = mqprio_qopt->qopt.num_tc;
7725 for (i = 0; num_tc; i++) {
7726 int qcount = mqprio_qopt->qopt.count[i];
7727 u64 max_rate, min_rate, rem;
7732 if (is_power_of_2(qcount)) {
7733 if (non_power_of_2_qcount &&
7734 qcount > non_power_of_2_qcount) {
7735 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
7736 qcount, non_power_of_2_qcount);
7739 if (qcount > max_rss_q_cnt)
7740 max_rss_q_cnt = qcount;
7742 if (non_power_of_2_qcount &&
7743 qcount != non_power_of_2_qcount) {
7744 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
7745 qcount, non_power_of_2_qcount);
7748 if (qcount < max_rss_q_cnt) {
7749 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
7750 qcount, max_rss_q_cnt);
7753 max_rss_q_cnt = qcount;
7754 non_power_of_2_qcount = qcount;
7757 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
7758 * converts the bandwidth rate limit into Bytes/s when
7759 * passing it down to the driver. So convert input bandwidth
7760 * from Bytes/s to Kbps
7762 max_rate = mqprio_qopt->max_rate[i];
7763 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
7764 sum_max_rate += max_rate;
7766 /* min_rate is minimum guaranteed rate and it can't be zero */
7767 min_rate = mqprio_qopt->min_rate[i];
7768 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
7769 sum_min_rate += min_rate;
7771 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
7772 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
7773 min_rate, ICE_MIN_BW_LIMIT);
7777 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
7779 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
7780 i, ICE_MIN_BW_LIMIT);
7784 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
7786 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
7787 i, ICE_MIN_BW_LIMIT);
7791 /* min_rate can't be more than max_rate, except when max_rate
7792 * is zero (implies max_rate sought is max line rate). In such
7793 * a case min_rate can be more than max.
7795 if (max_rate && min_rate > max_rate) {
7796 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
7797 min_rate, max_rate);
7801 if (i >= mqprio_qopt->qopt.num_tc - 1)
7803 if (mqprio_qopt->qopt.offset[i + 1] !=
7804 (mqprio_qopt->qopt.offset[i] + qcount))
7808 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7811 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7814 speed = ice_get_link_speed_kbps(vsi);
7815 if (sum_max_rate && sum_max_rate > (u64)speed) {
7816 dev_err(dev, "Invalid max Tx rate(%llu) Kbps > speed(%u) Kbps specified\n",
7817 sum_max_rate, speed);
7820 if (sum_min_rate && sum_min_rate > (u64)speed) {
7821 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
7822 sum_min_rate, speed);
7826 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
7827 vsi->ch_rss_size = max_rss_q_cnt;
7833 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
7834 * @pf: ptr to PF device
7837 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
7839 struct device *dev = ice_pf_to_dev(pf);
7844 if (!(vsi->num_gfltr || vsi->num_bfltr))
7848 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
7849 struct ice_fd_hw_prof *prof;
7853 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
7854 hw->fdir_prof[flow]->cnt))
7857 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
7858 enum ice_flow_priority prio;
7861 /* add this VSI to FDir profile for this flow */
7862 prio = ICE_FLOW_PRIO_NORMAL;
7863 prof = hw->fdir_prof[flow];
7864 prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
7865 status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
7866 prof->vsi_h[0], vsi->idx,
7867 prio, prof->fdir_seg[tun],
7870 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
7875 prof->entry_h[prof->cnt][tun] = entry_h;
7878 /* store VSI for filter replay and delete */
7879 prof->vsi_h[prof->cnt] = vsi->idx;
7883 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
7888 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
7894 * ice_add_channel - add a channel by adding VSI
7895 * @pf: ptr to PF device
7896 * @sw_id: underlying HW switching element ID
7897 * @ch: ptr to channel structure
7899 * Add a channel (VSI) using add_vsi and queue_map
7901 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
7903 struct device *dev = ice_pf_to_dev(pf);
7904 struct ice_vsi *vsi;
7906 if (ch->type != ICE_VSI_CHNL) {
7907 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
7911 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
7912 if (!vsi || vsi->type != ICE_VSI_CHNL) {
7913 dev_err(dev, "create chnl VSI failure\n");
7917 ice_add_vsi_to_fdir(pf, vsi);
7920 ch->vsi_num = vsi->vsi_num;
7921 ch->info.mapping_flags = vsi->info.mapping_flags;
7923 /* set the back pointer of channel for newly created VSI */
7926 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
7927 sizeof(vsi->info.q_mapping));
7928 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
7929 sizeof(vsi->info.tc_mapping));
7936 * @vsi: the VSI being setup
7937 * @ch: ptr to channel structure
7939 * Configure channel specific resources such as rings, vector.
7941 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
7945 for (i = 0; i < ch->num_txq; i++) {
7946 struct ice_q_vector *tx_q_vector, *rx_q_vector;
7947 struct ice_ring_container *rc;
7948 struct ice_tx_ring *tx_ring;
7949 struct ice_rx_ring *rx_ring;
7951 tx_ring = vsi->tx_rings[ch->base_q + i];
7952 rx_ring = vsi->rx_rings[ch->base_q + i];
7953 if (!tx_ring || !rx_ring)
7956 /* setup ring being channel enabled */
7960 /* following code block sets up vector specific attributes */
7961 tx_q_vector = tx_ring->q_vector;
7962 rx_q_vector = rx_ring->q_vector;
7963 if (!tx_q_vector && !rx_q_vector)
7967 tx_q_vector->ch = ch;
7968 /* setup Tx and Rx ITR setting if DIM is off */
7969 rc = &tx_q_vector->tx;
7970 if (!ITR_IS_DYNAMIC(rc))
7971 ice_write_itr(rc, rc->itr_setting);
7974 rx_q_vector->ch = ch;
7975 /* setup Tx and Rx ITR setting if DIM is off */
7976 rc = &rx_q_vector->rx;
7977 if (!ITR_IS_DYNAMIC(rc))
7978 ice_write_itr(rc, rc->itr_setting);
7982 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
7983 * GLINT_ITR register would have written to perform in-context
7984 * update, hence perform flush
7986 if (ch->num_txq || ch->num_rxq)
7987 ice_flush(&vsi->back->hw);
7991 * ice_cfg_chnl_all_res - configure channel resources
7992 * @vsi: pte to main_vsi
7993 * @ch: ptr to channel structure
7995 * This function configures channel specific resources such as flow-director
7996 * counter index, and other resources such as queues, vectors, ITR settings
7999 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8001 /* configure channel (aka ADQ) resources such as queues, vectors,
8002 * ITR settings for channel specific vectors and anything else
8004 ice_chnl_cfg_res(vsi, ch);
8008 * ice_setup_hw_channel - setup new channel
8009 * @pf: ptr to PF device
8010 * @vsi: the VSI being setup
8011 * @ch: ptr to channel structure
8012 * @sw_id: underlying HW switching element ID
8013 * @type: type of channel to be created (VMDq2/VF)
8015 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8016 * and configures Tx rings accordingly
8019 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8020 struct ice_channel *ch, u16 sw_id, u8 type)
8022 struct device *dev = ice_pf_to_dev(pf);
8025 ch->base_q = vsi->next_base_q;
8028 ret = ice_add_channel(pf, sw_id, ch);
8030 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8034 /* configure/setup ADQ specific resources */
8035 ice_cfg_chnl_all_res(vsi, ch);
8037 /* make sure to update the next_base_q so that subsequent channel's
8038 * (aka ADQ) VSI queue map is correct
8040 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8041 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8048 * ice_setup_channel - setup new channel using uplink element
8049 * @pf: ptr to PF device
8050 * @vsi: the VSI being setup
8051 * @ch: ptr to channel structure
8053 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8054 * and uplink switching element
8057 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8058 struct ice_channel *ch)
8060 struct device *dev = ice_pf_to_dev(pf);
8064 if (vsi->type != ICE_VSI_PF) {
8065 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8069 sw_id = pf->first_sw->sw_id;
8071 /* create channel (VSI) */
8072 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8074 dev_err(dev, "failed to setup hw_channel\n");
8077 dev_dbg(dev, "successfully created channel()\n");
8079 return ch->ch_vsi ? true : false;
8083 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8084 * @vsi: VSI to be configured
8085 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8086 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8089 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8093 err = ice_set_min_bw_limit(vsi, min_tx_rate);
8097 return ice_set_max_bw_limit(vsi, max_tx_rate);
8101 * ice_create_q_channel - function to create channel
8102 * @vsi: VSI to be configured
8103 * @ch: ptr to channel (it contains channel specific params)
8105 * This function creates channel (VSI) using num_queues specified by user,
8106 * reconfigs RSS if needed.
8108 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8110 struct ice_pf *pf = vsi->back;
8116 dev = ice_pf_to_dev(pf);
8117 if (!ch->num_txq || !ch->num_rxq) {
8118 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8122 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8123 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8124 vsi->cnt_q_avail, ch->num_txq);
8128 if (!ice_setup_channel(pf, vsi, ch)) {
8129 dev_info(dev, "Failed to setup channel\n");
8132 /* configure BW rate limit */
8133 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8136 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8139 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8140 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8142 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8143 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8146 vsi->cnt_q_avail -= ch->num_txq;
8152 * ice_rem_all_chnl_fltrs - removes all channel filters
8153 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8155 * Remove all advanced switch filters only if they are channel specific
8156 * tc-flower based filter
8158 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8160 struct ice_tc_flower_fltr *fltr;
8161 struct hlist_node *node;
8163 /* to remove all channel filters, iterate an ordered list of filters */
8164 hlist_for_each_entry_safe(fltr, node,
8165 &pf->tc_flower_fltr_list,
8167 struct ice_rule_query_data rule;
8170 /* for now process only channel specific filters */
8171 if (!ice_is_chnl_fltr(fltr))
8174 rule.rid = fltr->rid;
8175 rule.rule_id = fltr->rule_id;
8176 rule.vsi_handle = fltr->dest_id;
8177 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8179 if (status == -ENOENT)
8180 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8183 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8185 } else if (fltr->dest_vsi) {
8186 /* update advanced switch filter count */
8187 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8188 u32 flags = fltr->flags;
8190 fltr->dest_vsi->num_chnl_fltr--;
8191 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8192 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8193 pf->num_dmac_chnl_fltrs--;
8197 hlist_del(&fltr->tc_flower_node);
8203 * ice_remove_q_channels - Remove queue channels for the TCs
8204 * @vsi: VSI to be configured
8205 * @rem_fltr: delete advanced switch filter or not
8207 * Remove queue channels for the TCs
8209 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8211 struct ice_channel *ch, *ch_tmp;
8212 struct ice_pf *pf = vsi->back;
8215 /* remove all tc-flower based filter if they are channel filters only */
8217 ice_rem_all_chnl_fltrs(pf);
8219 /* remove ntuple filters since queue configuration is being changed */
8220 if (vsi->netdev->features & NETIF_F_NTUPLE) {
8221 struct ice_hw *hw = &pf->hw;
8223 mutex_lock(&hw->fdir_fltr_lock);
8224 ice_fdir_del_all_fltrs(vsi);
8225 mutex_unlock(&hw->fdir_fltr_lock);
8228 /* perform cleanup for channels if they exist */
8229 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8230 struct ice_vsi *ch_vsi;
8232 list_del(&ch->list);
8233 ch_vsi = ch->ch_vsi;
8239 /* Reset queue contexts */
8240 for (i = 0; i < ch->num_rxq; i++) {
8241 struct ice_tx_ring *tx_ring;
8242 struct ice_rx_ring *rx_ring;
8244 tx_ring = vsi->tx_rings[ch->base_q + i];
8245 rx_ring = vsi->rx_rings[ch->base_q + i];
8248 if (tx_ring->q_vector)
8249 tx_ring->q_vector->ch = NULL;
8253 if (rx_ring->q_vector)
8254 rx_ring->q_vector->ch = NULL;
8258 /* Release FD resources for the channel VSI */
8259 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8261 /* clear the VSI from scheduler tree */
8262 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8264 /* Delete VSI from FW */
8265 ice_vsi_delete(ch->ch_vsi);
8267 /* Delete VSI from PF and HW VSI arrays */
8268 ice_vsi_clear(ch->ch_vsi);
8270 /* free the channel */
8274 /* clear the channel VSI map which is stored in main VSI */
8275 ice_for_each_chnl_tc(i)
8276 vsi->tc_map_vsi[i] = NULL;
8278 /* reset main VSI's all TC information */
8284 * ice_rebuild_channels - rebuild channel
8287 * Recreate channel VSIs and replay filters
8289 static int ice_rebuild_channels(struct ice_pf *pf)
8291 struct device *dev = ice_pf_to_dev(pf);
8292 struct ice_vsi *main_vsi;
8293 bool rem_adv_fltr = true;
8294 struct ice_channel *ch;
8295 struct ice_vsi *vsi;
8299 main_vsi = ice_get_main_vsi(pf);
8303 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8304 main_vsi->old_numtc == 1)
8305 return 0; /* nothing to be done */
8307 /* reconfigure main VSI based on old value of TC and cached values
8310 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8312 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8313 main_vsi->old_ena_tc, main_vsi->vsi_num);
8317 /* rebuild ADQ VSIs */
8318 ice_for_each_vsi(pf, i) {
8319 enum ice_vsi_type type;
8322 if (!vsi || vsi->type != ICE_VSI_CHNL)
8327 /* rebuild ADQ VSI */
8328 err = ice_vsi_rebuild(vsi, true);
8330 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8331 ice_vsi_type_str(type), vsi->idx, err);
8335 /* Re-map HW VSI number, using VSI handle that has been
8336 * previously validated in ice_replay_vsi() call above
8338 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8340 /* replay filters for the VSI */
8341 err = ice_replay_vsi(&pf->hw, vsi->idx);
8343 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8344 ice_vsi_type_str(type), err, vsi->idx);
8345 rem_adv_fltr = false;
8348 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8349 ice_vsi_type_str(type), vsi->idx);
8351 /* store ADQ VSI at correct TC index in main VSI's
8354 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8357 /* ADQ VSI(s) has been rebuilt successfully, so setup
8358 * channel for main VSI's Tx and Rx rings
8360 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8361 struct ice_vsi *ch_vsi;
8363 ch_vsi = ch->ch_vsi;
8367 /* reconfig channel resources */
8368 ice_cfg_chnl_all_res(main_vsi, ch);
8370 /* replay BW rate limit if it is non-zero */
8371 if (!ch->max_tx_rate && !ch->min_tx_rate)
8374 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8377 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",
8378 err, ch->max_tx_rate, ch->min_tx_rate,
8381 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8382 ch->max_tx_rate, ch->min_tx_rate,
8386 /* reconfig RSS for main VSI */
8387 if (main_vsi->ch_rss_size)
8388 ice_vsi_cfg_rss_lut_key(main_vsi);
8393 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8398 * ice_create_q_channels - Add queue channel for the given TCs
8399 * @vsi: VSI to be configured
8401 * Configures queue channel mapping to the given TCs
8403 static int ice_create_q_channels(struct ice_vsi *vsi)
8405 struct ice_pf *pf = vsi->back;
8406 struct ice_channel *ch;
8409 ice_for_each_chnl_tc(i) {
8410 if (!(vsi->all_enatc & BIT(i)))
8413 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8418 INIT_LIST_HEAD(&ch->list);
8419 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8420 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8421 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8422 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8423 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8425 /* convert to Kbits/s */
8426 if (ch->max_tx_rate)
8427 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8428 ICE_BW_KBPS_DIVISOR);
8429 if (ch->min_tx_rate)
8430 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8431 ICE_BW_KBPS_DIVISOR);
8433 ret = ice_create_q_channel(vsi, ch);
8435 dev_err(ice_pf_to_dev(pf),
8436 "failed creating channel TC:%d\n", i);
8440 list_add_tail(&ch->list, &vsi->ch_list);
8441 vsi->tc_map_vsi[i] = ch->ch_vsi;
8442 dev_dbg(ice_pf_to_dev(pf),
8443 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8448 ice_remove_q_channels(vsi, false);
8454 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8455 * @netdev: net device to configure
8456 * @type_data: TC offload data
8458 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8460 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8461 struct ice_netdev_priv *np = netdev_priv(netdev);
8462 struct ice_vsi *vsi = np->vsi;
8463 struct ice_pf *pf = vsi->back;
8464 u16 mode, ena_tc_qdisc = 0;
8465 int cur_txq, cur_rxq;
8470 dev = ice_pf_to_dev(pf);
8471 num_tcf = mqprio_qopt->qopt.num_tc;
8472 hw = mqprio_qopt->qopt.hw;
8473 mode = mqprio_qopt->mode;
8475 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8476 vsi->ch_rss_size = 0;
8477 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8481 /* Generate queue region map for number of TCF requested */
8482 for (i = 0; i < num_tcf; i++)
8483 ena_tc_qdisc |= BIT(i);
8486 case TC_MQPRIO_MODE_CHANNEL:
8488 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8490 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8494 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8495 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8496 /* don't assume state of hw_tc_offload during driver load
8497 * and set the flag for TC flower filter if hw_tc_offload
8500 if (vsi->netdev->features & NETIF_F_HW_TC)
8501 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8509 /* Requesting same TCF configuration as already enabled */
8510 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8511 mode != TC_MQPRIO_MODE_CHANNEL)
8514 /* Pause VSI queues */
8515 ice_dis_vsi(vsi, true);
8517 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8518 ice_remove_q_channels(vsi, true);
8520 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8521 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8523 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8526 /* logic to rebuild VSI, same like ethtool -L */
8527 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8529 for (i = 0; i < num_tcf; i++) {
8530 if (!(ena_tc_qdisc & BIT(i)))
8533 offset = vsi->mqprio_qopt.qopt.offset[i];
8534 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8535 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8537 vsi->req_txq = offset + qcount_tx;
8538 vsi->req_rxq = offset + qcount_rx;
8540 /* store away original rss_size info, so that it gets reused
8541 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8542 * determine, what should be the rss_sizefor main VSI
8544 vsi->orig_rss_size = vsi->rss_size;
8547 /* save current values of Tx and Rx queues before calling VSI rebuild
8548 * for fallback option
8550 cur_txq = vsi->num_txq;
8551 cur_rxq = vsi->num_rxq;
8553 /* proceed with rebuild main VSI using correct number of queues */
8554 ret = ice_vsi_rebuild(vsi, false);
8556 /* fallback to current number of queues */
8557 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8558 vsi->req_txq = cur_txq;
8559 vsi->req_rxq = cur_rxq;
8560 clear_bit(ICE_RESET_FAILED, pf->state);
8561 if (ice_vsi_rebuild(vsi, false)) {
8562 dev_err(dev, "Rebuild of main VSI failed again\n");
8567 vsi->all_numtc = num_tcf;
8568 vsi->all_enatc = ena_tc_qdisc;
8569 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8571 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8576 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8577 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8578 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8580 /* set TC0 rate limit if specified */
8581 if (max_tx_rate || min_tx_rate) {
8582 /* convert to Kbits/s */
8584 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8586 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8588 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8590 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8591 max_tx_rate, min_tx_rate, vsi->vsi_num);
8593 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8594 max_tx_rate, min_tx_rate, vsi->vsi_num);
8598 ret = ice_create_q_channels(vsi);
8600 netdev_err(netdev, "failed configuring queue channels\n");
8603 netdev_dbg(netdev, "successfully configured channels\n");
8607 if (vsi->ch_rss_size)
8608 ice_vsi_cfg_rss_lut_key(vsi);
8611 /* if error, reset the all_numtc and all_enatc */
8617 ice_ena_vsi(vsi, true);
8622 static LIST_HEAD(ice_block_cb_list);
8625 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8628 struct ice_netdev_priv *np = netdev_priv(netdev);
8629 struct ice_pf *pf = np->vsi->back;
8633 case TC_SETUP_BLOCK:
8634 return flow_block_cb_setup_simple(type_data,
8636 ice_setup_tc_block_cb,
8638 case TC_SETUP_QDISC_MQPRIO:
8639 /* setup traffic classifier for receive side */
8640 mutex_lock(&pf->tc_mutex);
8641 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
8642 mutex_unlock(&pf->tc_mutex);
8650 static struct ice_indr_block_priv *
8651 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
8652 struct net_device *netdev)
8654 struct ice_indr_block_priv *cb_priv;
8656 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
8657 if (!cb_priv->netdev)
8659 if (cb_priv->netdev == netdev)
8666 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
8669 struct ice_indr_block_priv *priv = indr_priv;
8670 struct ice_netdev_priv *np = priv->np;
8673 case TC_SETUP_CLSFLOWER:
8674 return ice_setup_tc_cls_flower(np, priv->netdev,
8675 (struct flow_cls_offload *)
8683 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
8684 struct ice_netdev_priv *np,
8685 struct flow_block_offload *f, void *data,
8686 void (*cleanup)(struct flow_block_cb *block_cb))
8688 struct ice_indr_block_priv *indr_priv;
8689 struct flow_block_cb *block_cb;
8691 if (!ice_is_tunnel_supported(netdev) &&
8692 !(is_vlan_dev(netdev) &&
8693 vlan_dev_real_dev(netdev) == np->vsi->netdev))
8696 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
8699 switch (f->command) {
8700 case FLOW_BLOCK_BIND:
8701 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8705 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
8709 indr_priv->netdev = netdev;
8711 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
8714 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
8715 indr_priv, indr_priv,
8716 ice_rep_indr_tc_block_unbind,
8717 f, netdev, sch, data, np,
8720 if (IS_ERR(block_cb)) {
8721 list_del(&indr_priv->list);
8723 return PTR_ERR(block_cb);
8725 flow_block_cb_add(block_cb, f);
8726 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
8728 case FLOW_BLOCK_UNBIND:
8729 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8733 block_cb = flow_block_cb_lookup(f->block,
8734 ice_indr_setup_block_cb,
8739 flow_indr_block_cb_remove(block_cb, f);
8741 list_del(&block_cb->driver_list);
8750 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
8751 void *cb_priv, enum tc_setup_type type, void *type_data,
8753 void (*cleanup)(struct flow_block_cb *block_cb))
8756 case TC_SETUP_BLOCK:
8757 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
8766 * ice_open - Called when a network interface becomes active
8767 * @netdev: network interface device structure
8769 * The open entry point is called when a network interface is made
8770 * active by the system (IFF_UP). At this point all resources needed
8771 * for transmit and receive operations are allocated, the interrupt
8772 * handler is registered with the OS, the netdev watchdog is enabled,
8773 * and the stack is notified that the interface is ready.
8775 * Returns 0 on success, negative value on failure
8777 int ice_open(struct net_device *netdev)
8779 struct ice_netdev_priv *np = netdev_priv(netdev);
8780 struct ice_pf *pf = np->vsi->back;
8782 if (ice_is_reset_in_progress(pf->state)) {
8783 netdev_err(netdev, "can't open net device while reset is in progress");
8787 return ice_open_internal(netdev);
8791 * ice_open_internal - Called when a network interface becomes active
8792 * @netdev: network interface device structure
8794 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
8797 * Returns 0 on success, negative value on failure
8799 int ice_open_internal(struct net_device *netdev)
8801 struct ice_netdev_priv *np = netdev_priv(netdev);
8802 struct ice_vsi *vsi = np->vsi;
8803 struct ice_pf *pf = vsi->back;
8804 struct ice_port_info *pi;
8807 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
8808 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
8812 netif_carrier_off(netdev);
8814 pi = vsi->port_info;
8815 err = ice_update_link_info(pi);
8817 netdev_err(netdev, "Failed to get link info, error %d\n", err);
8821 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
8823 /* Set PHY if there is media, otherwise, turn off PHY */
8824 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
8825 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8826 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
8827 err = ice_init_phy_user_cfg(pi);
8829 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
8835 err = ice_configure_phy(vsi);
8837 netdev_err(netdev, "Failed to set physical link up, error %d\n",
8842 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8843 ice_set_link(vsi, false);
8846 err = ice_vsi_open(vsi);
8848 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
8849 vsi->vsi_num, vsi->vsw->sw_id);
8851 /* Update existing tunnels information */
8852 udp_tunnel_get_rx_info(netdev);
8858 * ice_stop - Disables a network interface
8859 * @netdev: network interface device structure
8861 * The stop entry point is called when an interface is de-activated by the OS,
8862 * and the netdevice enters the DOWN state. The hardware is still under the
8863 * driver's control, but the netdev interface is disabled.
8865 * Returns success only - not allowed to fail
8867 int ice_stop(struct net_device *netdev)
8869 struct ice_netdev_priv *np = netdev_priv(netdev);
8870 struct ice_vsi *vsi = np->vsi;
8871 struct ice_pf *pf = vsi->back;
8873 if (ice_is_reset_in_progress(pf->state)) {
8874 netdev_err(netdev, "can't stop net device while reset is in progress");
8884 * ice_features_check - Validate encapsulated packet conforms to limits
8886 * @netdev: This port's netdev
8887 * @features: Offload features that the stack believes apply
8889 static netdev_features_t
8890 ice_features_check(struct sk_buff *skb,
8891 struct net_device __always_unused *netdev,
8892 netdev_features_t features)
8894 bool gso = skb_is_gso(skb);
8897 /* No point in doing any of this if neither checksum nor GSO are
8898 * being requested for this frame. We can rule out both by just
8899 * checking for CHECKSUM_PARTIAL
8901 if (skb->ip_summed != CHECKSUM_PARTIAL)
8904 /* We cannot support GSO if the MSS is going to be less than
8905 * 64 bytes. If it is then we need to drop support for GSO.
8907 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
8908 features &= ~NETIF_F_GSO_MASK;
8910 len = skb_network_offset(skb);
8911 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
8912 goto out_rm_features;
8914 len = skb_network_header_len(skb);
8915 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8916 goto out_rm_features;
8918 if (skb->encapsulation) {
8919 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
8920 * the case of IPIP frames, the transport header pointer is
8921 * after the inner header! So check to make sure that this
8922 * is a GRE or UDP_TUNNEL frame before doing that math.
8924 if (gso && (skb_shinfo(skb)->gso_type &
8925 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
8926 len = skb_inner_network_header(skb) -
8927 skb_transport_header(skb);
8928 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
8929 goto out_rm_features;
8932 len = skb_inner_network_header_len(skb);
8933 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8934 goto out_rm_features;
8939 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8942 static const struct net_device_ops ice_netdev_safe_mode_ops = {
8943 .ndo_open = ice_open,
8944 .ndo_stop = ice_stop,
8945 .ndo_start_xmit = ice_start_xmit,
8946 .ndo_set_mac_address = ice_set_mac_address,
8947 .ndo_validate_addr = eth_validate_addr,
8948 .ndo_change_mtu = ice_change_mtu,
8949 .ndo_get_stats64 = ice_get_stats64,
8950 .ndo_tx_timeout = ice_tx_timeout,
8951 .ndo_bpf = ice_xdp_safe_mode,
8954 static const struct net_device_ops ice_netdev_ops = {
8955 .ndo_open = ice_open,
8956 .ndo_stop = ice_stop,
8957 .ndo_start_xmit = ice_start_xmit,
8958 .ndo_select_queue = ice_select_queue,
8959 .ndo_features_check = ice_features_check,
8960 .ndo_fix_features = ice_fix_features,
8961 .ndo_set_rx_mode = ice_set_rx_mode,
8962 .ndo_set_mac_address = ice_set_mac_address,
8963 .ndo_validate_addr = eth_validate_addr,
8964 .ndo_change_mtu = ice_change_mtu,
8965 .ndo_get_stats64 = ice_get_stats64,
8966 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
8967 .ndo_eth_ioctl = ice_eth_ioctl,
8968 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
8969 .ndo_set_vf_mac = ice_set_vf_mac,
8970 .ndo_get_vf_config = ice_get_vf_cfg,
8971 .ndo_set_vf_trust = ice_set_vf_trust,
8972 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
8973 .ndo_set_vf_link_state = ice_set_vf_link_state,
8974 .ndo_get_vf_stats = ice_get_vf_stats,
8975 .ndo_set_vf_rate = ice_set_vf_bw,
8976 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
8977 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
8978 .ndo_setup_tc = ice_setup_tc,
8979 .ndo_set_features = ice_set_features,
8980 .ndo_bridge_getlink = ice_bridge_getlink,
8981 .ndo_bridge_setlink = ice_bridge_setlink,
8982 .ndo_fdb_add = ice_fdb_add,
8983 .ndo_fdb_del = ice_fdb_del,
8984 #ifdef CONFIG_RFS_ACCEL
8985 .ndo_rx_flow_steer = ice_rx_flow_steer,
8987 .ndo_tx_timeout = ice_tx_timeout,
8989 .ndo_xdp_xmit = ice_xdp_xmit,
8990 .ndo_xsk_wakeup = ice_xsk_wakeup,
8991 .ndo_get_devlink_port = ice_get_devlink_port,