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(pf->first_sw)) {
416 err = ice_set_dflt_vsi(pf->first_sw, 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(pf->first_sw, vsi)) {
430 err = ice_clear_dflt_vsi(pf->first_sw);
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");
2404 set_bit(ICE_PFR_REQ, pf->state);
2406 case ICE_RESET_CORER:
2407 set_bit(ICE_CORER_REQ, pf->state);
2409 case ICE_RESET_GLOBR:
2410 set_bit(ICE_GLOBR_REQ, pf->state);
2416 ice_service_task_schedule(pf);
2421 * ice_irq_affinity_notify - Callback for affinity changes
2422 * @notify: context as to what irq was changed
2423 * @mask: the new affinity mask
2425 * This is a callback function used by the irq_set_affinity_notifier function
2426 * so that we may register to receive changes to the irq affinity masks.
2429 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2430 const cpumask_t *mask)
2432 struct ice_q_vector *q_vector =
2433 container_of(notify, struct ice_q_vector, affinity_notify);
2435 cpumask_copy(&q_vector->affinity_mask, mask);
2439 * ice_irq_affinity_release - Callback for affinity notifier release
2440 * @ref: internal core kernel usage
2442 * This is a callback function used by the irq_set_affinity_notifier function
2443 * to inform the current notification subscriber that they will no longer
2444 * receive notifications.
2446 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2449 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2450 * @vsi: the VSI being configured
2452 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2454 struct ice_hw *hw = &vsi->back->hw;
2457 ice_for_each_q_vector(vsi, i)
2458 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2465 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2466 * @vsi: the VSI being configured
2467 * @basename: name for the vector
2469 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2471 int q_vectors = vsi->num_q_vectors;
2472 struct ice_pf *pf = vsi->back;
2473 int base = vsi->base_vector;
2480 dev = ice_pf_to_dev(pf);
2481 for (vector = 0; vector < q_vectors; vector++) {
2482 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2484 irq_num = pf->msix_entries[base + vector].vector;
2486 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2487 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2488 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2490 } else if (q_vector->rx.rx_ring) {
2491 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2492 "%s-%s-%d", basename, "rx", rx_int_idx++);
2493 } else if (q_vector->tx.tx_ring) {
2494 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2495 "%s-%s-%d", basename, "tx", tx_int_idx++);
2497 /* skip this unused q_vector */
2500 if (vsi->type == ICE_VSI_CTRL && vsi->vf)
2501 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2502 IRQF_SHARED, q_vector->name,
2505 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2506 0, q_vector->name, q_vector);
2508 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2513 /* register for affinity change notifications */
2514 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2515 struct irq_affinity_notify *affinity_notify;
2517 affinity_notify = &q_vector->affinity_notify;
2518 affinity_notify->notify = ice_irq_affinity_notify;
2519 affinity_notify->release = ice_irq_affinity_release;
2520 irq_set_affinity_notifier(irq_num, affinity_notify);
2523 /* assign the mask for this irq */
2524 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2527 err = ice_set_cpu_rx_rmap(vsi);
2529 netdev_err(vsi->netdev, "Failed to setup CPU RMAP on VSI %u: %pe\n",
2530 vsi->vsi_num, ERR_PTR(err));
2534 vsi->irqs_ready = true;
2540 irq_num = pf->msix_entries[base + vector].vector;
2541 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2542 irq_set_affinity_notifier(irq_num, NULL);
2543 irq_set_affinity_hint(irq_num, NULL);
2544 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2550 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2551 * @vsi: VSI to setup Tx rings used by XDP
2553 * Return 0 on success and negative value on error
2555 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2557 struct device *dev = ice_pf_to_dev(vsi->back);
2558 struct ice_tx_desc *tx_desc;
2561 ice_for_each_xdp_txq(vsi, i) {
2562 u16 xdp_q_idx = vsi->alloc_txq + i;
2563 struct ice_tx_ring *xdp_ring;
2565 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2568 goto free_xdp_rings;
2570 xdp_ring->q_index = xdp_q_idx;
2571 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2572 xdp_ring->vsi = vsi;
2573 xdp_ring->netdev = NULL;
2574 xdp_ring->dev = dev;
2575 xdp_ring->count = vsi->num_tx_desc;
2576 xdp_ring->next_dd = ICE_RING_QUARTER(xdp_ring) - 1;
2577 xdp_ring->next_rs = ICE_RING_QUARTER(xdp_ring) - 1;
2578 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2579 if (ice_setup_tx_ring(xdp_ring))
2580 goto free_xdp_rings;
2581 ice_set_ring_xdp(xdp_ring);
2582 spin_lock_init(&xdp_ring->tx_lock);
2583 for (j = 0; j < xdp_ring->count; j++) {
2584 tx_desc = ICE_TX_DESC(xdp_ring, j);
2585 tx_desc->cmd_type_offset_bsz = 0;
2593 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2594 ice_free_tx_ring(vsi->xdp_rings[i]);
2599 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2600 * @vsi: VSI to set the bpf prog on
2601 * @prog: the bpf prog pointer
2603 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2605 struct bpf_prog *old_prog;
2608 old_prog = xchg(&vsi->xdp_prog, prog);
2610 bpf_prog_put(old_prog);
2612 ice_for_each_rxq(vsi, i)
2613 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2617 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2618 * @vsi: VSI to bring up Tx rings used by XDP
2619 * @prog: bpf program that will be assigned to VSI
2621 * Return 0 on success and negative value on error
2623 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2625 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2626 int xdp_rings_rem = vsi->num_xdp_txq;
2627 struct ice_pf *pf = vsi->back;
2628 struct ice_qs_cfg xdp_qs_cfg = {
2629 .qs_mutex = &pf->avail_q_mutex,
2630 .pf_map = pf->avail_txqs,
2631 .pf_map_size = pf->max_pf_txqs,
2632 .q_count = vsi->num_xdp_txq,
2633 .scatter_count = ICE_MAX_SCATTER_TXQS,
2634 .vsi_map = vsi->txq_map,
2635 .vsi_map_offset = vsi->alloc_txq,
2636 .mapping_mode = ICE_VSI_MAP_CONTIG
2642 dev = ice_pf_to_dev(pf);
2643 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2644 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2645 if (!vsi->xdp_rings)
2648 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2649 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2652 if (static_key_enabled(&ice_xdp_locking_key))
2653 netdev_warn(vsi->netdev,
2654 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2656 if (ice_xdp_alloc_setup_rings(vsi))
2657 goto clear_xdp_rings;
2659 /* follow the logic from ice_vsi_map_rings_to_vectors */
2660 ice_for_each_q_vector(vsi, v_idx) {
2661 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2662 int xdp_rings_per_v, q_id, q_base;
2664 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2665 vsi->num_q_vectors - v_idx);
2666 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2668 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2669 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2671 xdp_ring->q_vector = q_vector;
2672 xdp_ring->next = q_vector->tx.tx_ring;
2673 q_vector->tx.tx_ring = xdp_ring;
2675 xdp_rings_rem -= xdp_rings_per_v;
2678 ice_for_each_rxq(vsi, i) {
2679 if (static_key_enabled(&ice_xdp_locking_key)) {
2680 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2682 struct ice_q_vector *q_vector = vsi->rx_rings[i]->q_vector;
2683 struct ice_tx_ring *ring;
2685 ice_for_each_tx_ring(ring, q_vector->tx) {
2686 if (ice_ring_is_xdp(ring)) {
2687 vsi->rx_rings[i]->xdp_ring = ring;
2692 ice_tx_xsk_pool(vsi, i);
2695 /* omit the scheduler update if in reset path; XDP queues will be
2696 * taken into account at the end of ice_vsi_rebuild, where
2697 * ice_cfg_vsi_lan is being called
2699 if (ice_is_reset_in_progress(pf->state))
2702 /* tell the Tx scheduler that right now we have
2705 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2706 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2708 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2711 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2713 goto clear_xdp_rings;
2716 /* assign the prog only when it's not already present on VSI;
2717 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2718 * VSI rebuild that happens under ethtool -L can expose us to
2719 * the bpf_prog refcount issues as we would be swapping same
2720 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2721 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2722 * this is not harmful as dev_xdp_install bumps the refcount
2723 * before calling the op exposed by the driver;
2725 if (!ice_is_xdp_ena_vsi(vsi))
2726 ice_vsi_assign_bpf_prog(vsi, prog);
2730 ice_for_each_xdp_txq(vsi, i)
2731 if (vsi->xdp_rings[i]) {
2732 kfree_rcu(vsi->xdp_rings[i], rcu);
2733 vsi->xdp_rings[i] = NULL;
2737 mutex_lock(&pf->avail_q_mutex);
2738 ice_for_each_xdp_txq(vsi, i) {
2739 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2740 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2742 mutex_unlock(&pf->avail_q_mutex);
2744 devm_kfree(dev, vsi->xdp_rings);
2749 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2750 * @vsi: VSI to remove XDP rings
2752 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2755 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2757 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2758 struct ice_pf *pf = vsi->back;
2761 /* q_vectors are freed in reset path so there's no point in detaching
2762 * rings; in case of rebuild being triggered not from reset bits
2763 * in pf->state won't be set, so additionally check first q_vector
2766 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2769 ice_for_each_q_vector(vsi, v_idx) {
2770 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2771 struct ice_tx_ring *ring;
2773 ice_for_each_tx_ring(ring, q_vector->tx)
2774 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2777 /* restore the value of last node prior to XDP setup */
2778 q_vector->tx.tx_ring = ring;
2782 mutex_lock(&pf->avail_q_mutex);
2783 ice_for_each_xdp_txq(vsi, i) {
2784 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2785 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2787 mutex_unlock(&pf->avail_q_mutex);
2789 ice_for_each_xdp_txq(vsi, i)
2790 if (vsi->xdp_rings[i]) {
2791 if (vsi->xdp_rings[i]->desc) {
2793 ice_free_tx_ring(vsi->xdp_rings[i]);
2795 kfree_rcu(vsi->xdp_rings[i], rcu);
2796 vsi->xdp_rings[i] = NULL;
2799 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2800 vsi->xdp_rings = NULL;
2802 if (static_key_enabled(&ice_xdp_locking_key))
2803 static_branch_dec(&ice_xdp_locking_key);
2805 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2808 ice_vsi_assign_bpf_prog(vsi, NULL);
2810 /* notify Tx scheduler that we destroyed XDP queues and bring
2811 * back the old number of child nodes
2813 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2814 max_txqs[i] = vsi->num_txq;
2816 /* change number of XDP Tx queues to 0 */
2817 vsi->num_xdp_txq = 0;
2819 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2824 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2825 * @vsi: VSI to schedule napi on
2827 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2831 ice_for_each_rxq(vsi, i) {
2832 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2834 if (rx_ring->xsk_pool)
2835 napi_schedule(&rx_ring->q_vector->napi);
2840 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2841 * @vsi: VSI to determine the count of XDP Tx qs
2843 * returns 0 if Tx qs count is higher than at least half of CPU count,
2846 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2848 u16 avail = ice_get_avail_txq_count(vsi->back);
2849 u16 cpus = num_possible_cpus();
2851 if (avail < cpus / 2)
2854 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2856 if (vsi->num_xdp_txq < cpus)
2857 static_branch_inc(&ice_xdp_locking_key);
2863 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2864 * @vsi: VSI to setup XDP for
2865 * @prog: XDP program
2866 * @extack: netlink extended ack
2869 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2870 struct netlink_ext_ack *extack)
2872 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2873 bool if_running = netif_running(vsi->netdev);
2874 int ret = 0, xdp_ring_err = 0;
2876 if (frame_size > vsi->rx_buf_len) {
2877 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2881 /* need to stop netdev while setting up the program for Rx rings */
2882 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2883 ret = ice_down(vsi);
2885 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2890 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2891 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2893 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2895 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2897 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2899 /* reallocate Rx queues that are used for zero-copy */
2900 xdp_ring_err = ice_realloc_zc_buf(vsi, true);
2902 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Rx resources failed");
2903 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2904 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2906 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2907 /* reallocate Rx queues that were used for zero-copy */
2908 xdp_ring_err = ice_realloc_zc_buf(vsi, false);
2910 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Rx resources failed");
2912 /* safe to call even when prog == vsi->xdp_prog as
2913 * dev_xdp_install in net/core/dev.c incremented prog's
2914 * refcount so corresponding bpf_prog_put won't cause
2917 ice_vsi_assign_bpf_prog(vsi, prog);
2924 ice_vsi_rx_napi_schedule(vsi);
2926 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2930 * ice_xdp_safe_mode - XDP handler for safe mode
2934 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2935 struct netdev_bpf *xdp)
2937 NL_SET_ERR_MSG_MOD(xdp->extack,
2938 "Please provide working DDP firmware package in order to use XDP\n"
2939 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2944 * ice_xdp - implements XDP handler
2948 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2950 struct ice_netdev_priv *np = netdev_priv(dev);
2951 struct ice_vsi *vsi = np->vsi;
2953 if (vsi->type != ICE_VSI_PF) {
2954 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2958 switch (xdp->command) {
2959 case XDP_SETUP_PROG:
2960 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2961 case XDP_SETUP_XSK_POOL:
2962 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2970 * ice_ena_misc_vector - enable the non-queue interrupts
2971 * @pf: board private structure
2973 static void ice_ena_misc_vector(struct ice_pf *pf)
2975 struct ice_hw *hw = &pf->hw;
2978 /* Disable anti-spoof detection interrupt to prevent spurious event
2979 * interrupts during a function reset. Anti-spoof functionally is
2982 val = rd32(hw, GL_MDCK_TX_TDPU);
2983 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2984 wr32(hw, GL_MDCK_TX_TDPU, val);
2986 /* clear things first */
2987 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2988 rd32(hw, PFINT_OICR); /* read to clear */
2990 val = (PFINT_OICR_ECC_ERR_M |
2991 PFINT_OICR_MAL_DETECT_M |
2993 PFINT_OICR_PCI_EXCEPTION_M |
2995 PFINT_OICR_HMC_ERR_M |
2996 PFINT_OICR_PE_PUSH_M |
2997 PFINT_OICR_PE_CRITERR_M);
2999 wr32(hw, PFINT_OICR_ENA, val);
3001 /* SW_ITR_IDX = 0, but don't change INTENA */
3002 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
3003 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
3007 * ice_misc_intr - misc interrupt handler
3008 * @irq: interrupt number
3009 * @data: pointer to a q_vector
3011 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
3013 struct ice_pf *pf = (struct ice_pf *)data;
3014 struct ice_hw *hw = &pf->hw;
3015 irqreturn_t ret = IRQ_NONE;
3019 dev = ice_pf_to_dev(pf);
3020 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
3021 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
3022 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
3024 oicr = rd32(hw, PFINT_OICR);
3025 ena_mask = rd32(hw, PFINT_OICR_ENA);
3027 if (oicr & PFINT_OICR_SWINT_M) {
3028 ena_mask &= ~PFINT_OICR_SWINT_M;
3032 if (oicr & PFINT_OICR_MAL_DETECT_M) {
3033 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
3034 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
3036 if (oicr & PFINT_OICR_VFLR_M) {
3037 /* disable any further VFLR event notifications */
3038 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
3039 u32 reg = rd32(hw, PFINT_OICR_ENA);
3041 reg &= ~PFINT_OICR_VFLR_M;
3042 wr32(hw, PFINT_OICR_ENA, reg);
3044 ena_mask &= ~PFINT_OICR_VFLR_M;
3045 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
3049 if (oicr & PFINT_OICR_GRST_M) {
3052 /* we have a reset warning */
3053 ena_mask &= ~PFINT_OICR_GRST_M;
3054 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
3055 GLGEN_RSTAT_RESET_TYPE_S;
3057 if (reset == ICE_RESET_CORER)
3059 else if (reset == ICE_RESET_GLOBR)
3061 else if (reset == ICE_RESET_EMPR)
3064 dev_dbg(dev, "Invalid reset type %d\n", reset);
3066 /* If a reset cycle isn't already in progress, we set a bit in
3067 * pf->state so that the service task can start a reset/rebuild.
3069 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3070 if (reset == ICE_RESET_CORER)
3071 set_bit(ICE_CORER_RECV, pf->state);
3072 else if (reset == ICE_RESET_GLOBR)
3073 set_bit(ICE_GLOBR_RECV, pf->state);
3075 set_bit(ICE_EMPR_RECV, pf->state);
3077 /* There are couple of different bits at play here.
3078 * hw->reset_ongoing indicates whether the hardware is
3079 * in reset. This is set to true when a reset interrupt
3080 * is received and set back to false after the driver
3081 * has determined that the hardware is out of reset.
3083 * ICE_RESET_OICR_RECV in pf->state indicates
3084 * that a post reset rebuild is required before the
3085 * driver is operational again. This is set above.
3087 * As this is the start of the reset/rebuild cycle, set
3088 * both to indicate that.
3090 hw->reset_ongoing = true;
3094 if (oicr & PFINT_OICR_TSYN_TX_M) {
3095 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3096 ice_ptp_process_ts(pf);
3099 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3100 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3101 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3103 /* Save EVENTs from GTSYN register */
3104 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
3105 GLTSYN_STAT_EVENT1_M |
3106 GLTSYN_STAT_EVENT2_M);
3107 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3108 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
3111 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3112 if (oicr & ICE_AUX_CRIT_ERR) {
3113 pf->oicr_err_reg |= oicr;
3114 set_bit(ICE_AUX_ERR_PENDING, pf->state);
3115 ena_mask &= ~ICE_AUX_CRIT_ERR;
3118 /* Report any remaining unexpected interrupts */
3121 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3122 /* If a critical error is pending there is no choice but to
3125 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3126 PFINT_OICR_ECC_ERR_M)) {
3127 set_bit(ICE_PFR_REQ, pf->state);
3128 ice_service_task_schedule(pf);
3133 ice_service_task_schedule(pf);
3134 ice_irq_dynamic_ena(hw, NULL, NULL);
3140 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3141 * @hw: pointer to HW structure
3143 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3145 /* disable Admin queue Interrupt causes */
3146 wr32(hw, PFINT_FW_CTL,
3147 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3149 /* disable Mailbox queue Interrupt causes */
3150 wr32(hw, PFINT_MBX_CTL,
3151 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3153 wr32(hw, PFINT_SB_CTL,
3154 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3156 /* disable Control queue Interrupt causes */
3157 wr32(hw, PFINT_OICR_CTL,
3158 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3164 * ice_free_irq_msix_misc - Unroll misc vector setup
3165 * @pf: board private structure
3167 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3169 struct ice_hw *hw = &pf->hw;
3171 ice_dis_ctrlq_interrupts(hw);
3173 /* disable OICR interrupt */
3174 wr32(hw, PFINT_OICR_ENA, 0);
3177 if (pf->msix_entries) {
3178 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
3179 devm_free_irq(ice_pf_to_dev(pf),
3180 pf->msix_entries[pf->oicr_idx].vector, pf);
3183 pf->num_avail_sw_msix += 1;
3184 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
3188 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3189 * @hw: pointer to HW structure
3190 * @reg_idx: HW vector index to associate the control queue interrupts with
3192 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3196 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3197 PFINT_OICR_CTL_CAUSE_ENA_M);
3198 wr32(hw, PFINT_OICR_CTL, val);
3200 /* enable Admin queue Interrupt causes */
3201 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3202 PFINT_FW_CTL_CAUSE_ENA_M);
3203 wr32(hw, PFINT_FW_CTL, val);
3205 /* enable Mailbox queue Interrupt causes */
3206 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3207 PFINT_MBX_CTL_CAUSE_ENA_M);
3208 wr32(hw, PFINT_MBX_CTL, val);
3210 /* This enables Sideband queue Interrupt causes */
3211 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3212 PFINT_SB_CTL_CAUSE_ENA_M);
3213 wr32(hw, PFINT_SB_CTL, val);
3219 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3220 * @pf: board private structure
3222 * This sets up the handler for MSIX 0, which is used to manage the
3223 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3224 * when in MSI or Legacy interrupt mode.
3226 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3228 struct device *dev = ice_pf_to_dev(pf);
3229 struct ice_hw *hw = &pf->hw;
3230 int oicr_idx, err = 0;
3232 if (!pf->int_name[0])
3233 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3234 dev_driver_string(dev), dev_name(dev));
3236 /* Do not request IRQ but do enable OICR interrupt since settings are
3237 * lost during reset. Note that this function is called only during
3238 * rebuild path and not while reset is in progress.
3240 if (ice_is_reset_in_progress(pf->state))
3243 /* reserve one vector in irq_tracker for misc interrupts */
3244 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3248 pf->num_avail_sw_msix -= 1;
3249 pf->oicr_idx = (u16)oicr_idx;
3251 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
3252 ice_misc_intr, 0, pf->int_name, pf);
3254 dev_err(dev, "devm_request_irq for %s failed: %d\n",
3256 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3257 pf->num_avail_sw_msix += 1;
3262 ice_ena_misc_vector(pf);
3264 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3265 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3266 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3269 ice_irq_dynamic_ena(hw, NULL, NULL);
3275 * ice_napi_add - register NAPI handler for the VSI
3276 * @vsi: VSI for which NAPI handler is to be registered
3278 * This function is only called in the driver's load path. Registering the NAPI
3279 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3280 * reset/rebuild, etc.)
3282 static void ice_napi_add(struct ice_vsi *vsi)
3289 ice_for_each_q_vector(vsi, v_idx)
3290 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3291 ice_napi_poll, NAPI_POLL_WEIGHT);
3295 * ice_set_ops - set netdev and ethtools ops for the given netdev
3296 * @netdev: netdev instance
3298 static void ice_set_ops(struct net_device *netdev)
3300 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3302 if (ice_is_safe_mode(pf)) {
3303 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3304 ice_set_ethtool_safe_mode_ops(netdev);
3308 netdev->netdev_ops = &ice_netdev_ops;
3309 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3310 ice_set_ethtool_ops(netdev);
3314 * ice_set_netdev_features - set features for the given netdev
3315 * @netdev: netdev instance
3317 static void ice_set_netdev_features(struct net_device *netdev)
3319 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3320 bool is_dvm_ena = ice_is_dvm_ena(&pf->hw);
3321 netdev_features_t csumo_features;
3322 netdev_features_t vlano_features;
3323 netdev_features_t dflt_features;
3324 netdev_features_t tso_features;
3326 if (ice_is_safe_mode(pf)) {
3328 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3329 netdev->hw_features = netdev->features;
3333 dflt_features = NETIF_F_SG |
3338 csumo_features = NETIF_F_RXCSUM |
3343 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3344 NETIF_F_HW_VLAN_CTAG_TX |
3345 NETIF_F_HW_VLAN_CTAG_RX;
3347 /* Enable CTAG/STAG filtering by default in Double VLAN Mode (DVM) */
3349 vlano_features |= NETIF_F_HW_VLAN_STAG_FILTER;
3351 tso_features = NETIF_F_TSO |
3355 NETIF_F_GSO_UDP_TUNNEL |
3356 NETIF_F_GSO_GRE_CSUM |
3357 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3358 NETIF_F_GSO_PARTIAL |
3359 NETIF_F_GSO_IPXIP4 |
3360 NETIF_F_GSO_IPXIP6 |
3363 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3364 NETIF_F_GSO_GRE_CSUM;
3365 /* set features that user can change */
3366 netdev->hw_features = dflt_features | csumo_features |
3367 vlano_features | tso_features;
3369 /* add support for HW_CSUM on packets with MPLS header */
3370 netdev->mpls_features = NETIF_F_HW_CSUM |
3374 /* enable features */
3375 netdev->features |= netdev->hw_features;
3377 netdev->hw_features |= NETIF_F_HW_TC;
3379 /* encap and VLAN devices inherit default, csumo and tso features */
3380 netdev->hw_enc_features |= dflt_features | csumo_features |
3382 netdev->vlan_features |= dflt_features | csumo_features |
3385 /* advertise support but don't enable by default since only one type of
3386 * VLAN offload can be enabled at a time (i.e. CTAG or STAG). When one
3387 * type turns on the other has to be turned off. This is enforced by the
3388 * ice_fix_features() ndo callback.
3391 netdev->hw_features |= NETIF_F_HW_VLAN_STAG_RX |
3392 NETIF_F_HW_VLAN_STAG_TX;
3396 * ice_cfg_netdev - Allocate, configure and register a netdev
3397 * @vsi: the VSI associated with the new netdev
3399 * Returns 0 on success, negative value on failure
3401 static int ice_cfg_netdev(struct ice_vsi *vsi)
3403 struct ice_netdev_priv *np;
3404 struct net_device *netdev;
3405 u8 mac_addr[ETH_ALEN];
3407 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3412 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3413 vsi->netdev = netdev;
3414 np = netdev_priv(netdev);
3417 ice_set_netdev_features(netdev);
3419 ice_set_ops(netdev);
3421 if (vsi->type == ICE_VSI_PF) {
3422 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3423 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3424 eth_hw_addr_set(netdev, mac_addr);
3425 ether_addr_copy(netdev->perm_addr, mac_addr);
3428 netdev->priv_flags |= IFF_UNICAST_FLT;
3430 /* Setup netdev TC information */
3431 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3433 /* setup watchdog timeout value to be 5 second */
3434 netdev->watchdog_timeo = 5 * HZ;
3436 netdev->min_mtu = ETH_MIN_MTU;
3437 netdev->max_mtu = ICE_MAX_MTU;
3443 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3444 * @lut: Lookup table
3445 * @rss_table_size: Lookup table size
3446 * @rss_size: Range of queue number for hashing
3448 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3452 for (i = 0; i < rss_table_size; i++)
3453 lut[i] = i % rss_size;
3457 * ice_pf_vsi_setup - Set up a PF VSI
3458 * @pf: board private structure
3459 * @pi: pointer to the port_info instance
3461 * Returns pointer to the successfully allocated VSI software struct
3462 * on success, otherwise returns NULL on failure.
3464 static struct ice_vsi *
3465 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3467 return ice_vsi_setup(pf, pi, ICE_VSI_PF, NULL, NULL);
3470 static struct ice_vsi *
3471 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3472 struct ice_channel *ch)
3474 return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, NULL, ch);
3478 * ice_ctrl_vsi_setup - Set up a control 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.
3485 static struct ice_vsi *
3486 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3488 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, NULL, NULL);
3492 * ice_lb_vsi_setup - Set up a loopback VSI
3493 * @pf: board private structure
3494 * @pi: pointer to the port_info instance
3496 * Returns pointer to the successfully allocated VSI software struct
3497 * on success, otherwise returns NULL on failure.
3500 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3502 return ice_vsi_setup(pf, pi, ICE_VSI_LB, NULL, NULL);
3506 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3507 * @netdev: network interface to be adjusted
3509 * @vid: VLAN ID to be added
3511 * net_device_ops implementation for adding VLAN IDs
3514 ice_vlan_rx_add_vid(struct net_device *netdev, __be16 proto, u16 vid)
3516 struct ice_netdev_priv *np = netdev_priv(netdev);
3517 struct ice_vsi_vlan_ops *vlan_ops;
3518 struct ice_vsi *vsi = np->vsi;
3519 struct ice_vlan vlan;
3522 /* VLAN 0 is added by default during load/reset */
3526 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3527 usleep_range(1000, 2000);
3529 /* Add multicast promisc rule for the VLAN ID to be added if
3530 * all-multicast is currently enabled.
3532 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3533 ret = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3534 ICE_MCAST_VLAN_PROMISC_BITS,
3540 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3542 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3543 * packets aren't pruned by the device's internal switch on Rx
3545 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3546 ret = vlan_ops->add_vlan(vsi, &vlan);
3550 /* If all-multicast is currently enabled and this VLAN ID is only one
3551 * besides VLAN-0 we have to update look-up type of multicast promisc
3552 * rule for VLAN-0 from ICE_SW_LKUP_PROMISC to ICE_SW_LKUP_PROMISC_VLAN.
3554 if ((vsi->current_netdev_flags & IFF_ALLMULTI) &&
3555 ice_vsi_num_non_zero_vlans(vsi) == 1) {
3556 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3557 ICE_MCAST_PROMISC_BITS, 0);
3558 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3559 ICE_MCAST_VLAN_PROMISC_BITS, 0);
3563 clear_bit(ICE_CFG_BUSY, vsi->state);
3569 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3570 * @netdev: network interface to be adjusted
3572 * @vid: VLAN ID to be removed
3574 * net_device_ops implementation for removing VLAN IDs
3577 ice_vlan_rx_kill_vid(struct net_device *netdev, __be16 proto, u16 vid)
3579 struct ice_netdev_priv *np = netdev_priv(netdev);
3580 struct ice_vsi_vlan_ops *vlan_ops;
3581 struct ice_vsi *vsi = np->vsi;
3582 struct ice_vlan vlan;
3585 /* don't allow removal of VLAN 0 */
3589 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
3590 usleep_range(1000, 2000);
3592 ret = ice_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3593 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3595 netdev_err(netdev, "Error clearing multicast promiscuous mode on VSI %i\n",
3597 vsi->current_netdev_flags |= IFF_ALLMULTI;
3600 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
3602 /* Make sure VLAN delete is successful before updating VLAN
3605 vlan = ICE_VLAN(be16_to_cpu(proto), vid, 0);
3606 ret = vlan_ops->del_vlan(vsi, &vlan);
3610 /* Remove multicast promisc rule for the removed VLAN ID if
3611 * all-multicast is enabled.
3613 if (vsi->current_netdev_flags & IFF_ALLMULTI)
3614 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3615 ICE_MCAST_VLAN_PROMISC_BITS, vid);
3617 if (!ice_vsi_has_non_zero_vlans(vsi)) {
3618 /* Update look-up type of multicast promisc rule for VLAN 0
3619 * from ICE_SW_LKUP_PROMISC_VLAN to ICE_SW_LKUP_PROMISC when
3620 * all-multicast is enabled and VLAN 0 is the only VLAN rule.
3622 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
3623 ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx,
3624 ICE_MCAST_VLAN_PROMISC_BITS,
3626 ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx,
3627 ICE_MCAST_PROMISC_BITS, 0);
3632 clear_bit(ICE_CFG_BUSY, vsi->state);
3638 * ice_rep_indr_tc_block_unbind
3639 * @cb_priv: indirection block private data
3641 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3643 struct ice_indr_block_priv *indr_priv = cb_priv;
3645 list_del(&indr_priv->list);
3650 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3651 * @vsi: VSI struct which has the netdev
3653 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3655 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3657 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3658 ice_rep_indr_tc_block_unbind);
3662 * ice_tc_indir_block_remove - clean indirect TC block notifications
3665 static void ice_tc_indir_block_remove(struct ice_pf *pf)
3667 struct ice_vsi *pf_vsi = ice_get_main_vsi(pf);
3672 ice_tc_indir_block_unregister(pf_vsi);
3676 * ice_tc_indir_block_register - Register TC indirect block notifications
3677 * @vsi: VSI struct which has the netdev
3679 * Returns 0 on success, negative value on failure
3681 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3683 struct ice_netdev_priv *np;
3685 if (!vsi || !vsi->netdev)
3688 np = netdev_priv(vsi->netdev);
3690 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3691 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3695 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3696 * @pf: board private structure
3698 * Returns 0 on success, negative value on failure
3700 static int ice_setup_pf_sw(struct ice_pf *pf)
3702 struct device *dev = ice_pf_to_dev(pf);
3703 bool dvm = ice_is_dvm_ena(&pf->hw);
3704 struct ice_vsi *vsi;
3707 if (ice_is_reset_in_progress(pf->state))
3710 status = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
3714 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3718 /* init channel list */
3719 INIT_LIST_HEAD(&vsi->ch_list);
3721 status = ice_cfg_netdev(vsi);
3723 goto unroll_vsi_setup;
3724 /* netdev has to be configured before setting frame size */
3725 ice_vsi_cfg_frame_size(vsi);
3727 /* init indirect block notifications */
3728 status = ice_tc_indir_block_register(vsi);
3730 dev_err(dev, "Failed to register netdev notifier\n");
3731 goto unroll_cfg_netdev;
3734 /* Setup DCB netlink interface */
3735 ice_dcbnl_setup(vsi);
3737 /* registering the NAPI handler requires both the queues and
3738 * netdev to be created, which are done in ice_pf_vsi_setup()
3739 * and ice_cfg_netdev() respectively
3743 status = ice_init_mac_fltr(pf);
3745 goto unroll_napi_add;
3750 ice_tc_indir_block_unregister(vsi);
3755 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3756 free_netdev(vsi->netdev);
3762 ice_vsi_release(vsi);
3767 * ice_get_avail_q_count - Get count of queues in use
3768 * @pf_qmap: bitmap to get queue use count from
3769 * @lock: pointer to a mutex that protects access to pf_qmap
3770 * @size: size of the bitmap
3773 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3779 for_each_clear_bit(bit, pf_qmap, size)
3787 * ice_get_avail_txq_count - Get count of Tx queues in use
3788 * @pf: pointer to an ice_pf instance
3790 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3792 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3797 * ice_get_avail_rxq_count - Get count of Rx queues in use
3798 * @pf: pointer to an ice_pf instance
3800 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3802 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3807 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3808 * @pf: board private structure to initialize
3810 static void ice_deinit_pf(struct ice_pf *pf)
3812 ice_service_task_stop(pf);
3813 mutex_destroy(&pf->adev_mutex);
3814 mutex_destroy(&pf->sw_mutex);
3815 mutex_destroy(&pf->tc_mutex);
3816 mutex_destroy(&pf->avail_q_mutex);
3817 mutex_destroy(&pf->vfs.table_lock);
3819 if (pf->avail_txqs) {
3820 bitmap_free(pf->avail_txqs);
3821 pf->avail_txqs = NULL;
3824 if (pf->avail_rxqs) {
3825 bitmap_free(pf->avail_rxqs);
3826 pf->avail_rxqs = NULL;
3830 ptp_clock_unregister(pf->ptp.clock);
3834 * ice_set_pf_caps - set PFs capability flags
3835 * @pf: pointer to the PF instance
3837 static void ice_set_pf_caps(struct ice_pf *pf)
3839 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3841 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3842 if (func_caps->common_cap.rdma)
3843 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3844 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3845 if (func_caps->common_cap.dcb)
3846 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3847 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3848 if (func_caps->common_cap.sr_iov_1_1) {
3849 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3850 pf->vfs.num_supported = min_t(int, func_caps->num_allocd_vfs,
3853 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3854 if (func_caps->common_cap.rss_table_size)
3855 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3857 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3858 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3861 /* ctrl_vsi_idx will be set to a valid value when flow director
3862 * is setup by ice_init_fdir
3864 pf->ctrl_vsi_idx = ICE_NO_VSI;
3865 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3866 /* force guaranteed filter pool for PF */
3867 ice_alloc_fd_guar_item(&pf->hw, &unused,
3868 func_caps->fd_fltr_guar);
3869 /* force shared filter pool for PF */
3870 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3871 func_caps->fd_fltr_best_effort);
3874 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3875 if (func_caps->common_cap.ieee_1588)
3876 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3878 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3879 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3883 * ice_init_pf - Initialize general software structures (struct ice_pf)
3884 * @pf: board private structure to initialize
3886 static int ice_init_pf(struct ice_pf *pf)
3888 ice_set_pf_caps(pf);
3890 mutex_init(&pf->sw_mutex);
3891 mutex_init(&pf->tc_mutex);
3892 mutex_init(&pf->adev_mutex);
3894 INIT_HLIST_HEAD(&pf->aq_wait_list);
3895 spin_lock_init(&pf->aq_wait_lock);
3896 init_waitqueue_head(&pf->aq_wait_queue);
3898 init_waitqueue_head(&pf->reset_wait_queue);
3900 /* setup service timer and periodic service task */
3901 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3902 pf->serv_tmr_period = HZ;
3903 INIT_WORK(&pf->serv_task, ice_service_task);
3904 clear_bit(ICE_SERVICE_SCHED, pf->state);
3906 mutex_init(&pf->avail_q_mutex);
3907 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3908 if (!pf->avail_txqs)
3911 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3912 if (!pf->avail_rxqs) {
3913 bitmap_free(pf->avail_txqs);
3914 pf->avail_txqs = NULL;
3918 mutex_init(&pf->vfs.table_lock);
3919 hash_init(pf->vfs.table);
3925 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3926 * @pf: board private structure
3928 * compute the number of MSIX vectors required (v_budget) and request from
3929 * the OS. Return the number of vectors reserved or negative on failure
3931 static int ice_ena_msix_range(struct ice_pf *pf)
3933 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3934 struct device *dev = ice_pf_to_dev(pf);
3937 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3938 num_cpus = num_online_cpus();
3940 /* reserve for LAN miscellaneous handler */
3941 needed = ICE_MIN_LAN_OICR_MSIX;
3942 if (v_left < needed)
3943 goto no_hw_vecs_left_err;
3947 /* reserve for flow director */
3948 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3949 needed = ICE_FDIR_MSIX;
3950 if (v_left < needed)
3951 goto no_hw_vecs_left_err;
3956 /* reserve for switchdev */
3957 needed = ICE_ESWITCH_MSIX;
3958 if (v_left < needed)
3959 goto no_hw_vecs_left_err;
3963 /* total used for non-traffic vectors */
3966 /* reserve vectors for LAN traffic */
3968 if (v_left < needed)
3969 goto no_hw_vecs_left_err;
3970 pf->num_lan_msix = needed;
3974 /* reserve vectors for RDMA auxiliary driver */
3975 if (ice_is_rdma_ena(pf)) {
3976 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3977 if (v_left < needed)
3978 goto no_hw_vecs_left_err;
3979 pf->num_rdma_msix = needed;
3984 pf->msix_entries = devm_kcalloc(dev, v_budget,
3985 sizeof(*pf->msix_entries), GFP_KERNEL);
3986 if (!pf->msix_entries) {
3991 for (i = 0; i < v_budget; i++)
3992 pf->msix_entries[i].entry = i;
3994 /* actually reserve the vectors */
3995 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3996 ICE_MIN_MSIX, v_budget);
3998 dev_err(dev, "unable to reserve MSI-X vectors\n");
4003 if (v_actual < v_budget) {
4004 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
4005 v_budget, v_actual);
4007 if (v_actual < ICE_MIN_MSIX) {
4008 /* error if we can't get minimum vectors */
4009 pci_disable_msix(pf->pdev);
4013 int v_remain = v_actual - v_other;
4014 int v_rdma = 0, v_min_rdma = 0;
4016 if (ice_is_rdma_ena(pf)) {
4017 /* Need at least 1 interrupt in addition to
4020 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
4021 v_min_rdma = ICE_MIN_RDMA_MSIX;
4024 if (v_actual == ICE_MIN_MSIX ||
4025 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
4026 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
4027 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
4029 pf->num_rdma_msix = 0;
4030 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
4031 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
4032 (v_remain - v_rdma < v_rdma)) {
4033 /* Support minimum RDMA and give remaining
4034 * vectors to LAN MSIX
4036 pf->num_rdma_msix = v_min_rdma;
4037 pf->num_lan_msix = v_remain - v_min_rdma;
4039 /* Split remaining MSIX with RDMA after
4040 * accounting for AEQ MSIX
4042 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
4043 ICE_RDMA_NUM_AEQ_MSIX;
4044 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
4047 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
4050 if (ice_is_rdma_ena(pf))
4051 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
4059 devm_kfree(dev, pf->msix_entries);
4062 no_hw_vecs_left_err:
4063 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
4067 pf->num_rdma_msix = 0;
4068 pf->num_lan_msix = 0;
4073 * ice_dis_msix - Disable MSI-X interrupt setup in OS
4074 * @pf: board private structure
4076 static void ice_dis_msix(struct ice_pf *pf)
4078 pci_disable_msix(pf->pdev);
4079 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
4080 pf->msix_entries = NULL;
4084 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
4085 * @pf: board private structure
4087 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
4091 if (pf->irq_tracker) {
4092 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
4093 pf->irq_tracker = NULL;
4098 * ice_init_interrupt_scheme - Determine proper interrupt scheme
4099 * @pf: board private structure to initialize
4101 static int ice_init_interrupt_scheme(struct ice_pf *pf)
4105 vectors = ice_ena_msix_range(pf);
4110 /* set up vector assignment tracking */
4111 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
4112 struct_size(pf->irq_tracker, list, vectors),
4114 if (!pf->irq_tracker) {
4119 /* populate SW interrupts pool with number of OS granted IRQs. */
4120 pf->num_avail_sw_msix = (u16)vectors;
4121 pf->irq_tracker->num_entries = (u16)vectors;
4122 pf->irq_tracker->end = pf->irq_tracker->num_entries;
4128 * ice_is_wol_supported - check if WoL is supported
4129 * @hw: pointer to hardware info
4131 * Check if WoL is supported based on the HW configuration.
4132 * Returns true if NVM supports and enables WoL for this port, false otherwise
4134 bool ice_is_wol_supported(struct ice_hw *hw)
4138 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4139 * word) indicates WoL is not supported on the corresponding PF ID.
4141 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4144 return !(BIT(hw->port_info->lport) & wol_ctrl);
4148 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4149 * @vsi: VSI being changed
4150 * @new_rx: new number of Rx queues
4151 * @new_tx: new number of Tx queues
4153 * Only change the number of queues if new_tx, or new_rx is non-0.
4155 * Returns 0 on success.
4157 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
4159 struct ice_pf *pf = vsi->back;
4160 int err = 0, timeout = 50;
4162 if (!new_rx && !new_tx)
4165 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4169 usleep_range(1000, 2000);
4173 vsi->req_txq = (u16)new_tx;
4175 vsi->req_rxq = (u16)new_rx;
4177 /* set for the next time the netdev is started */
4178 if (!netif_running(vsi->netdev)) {
4179 ice_vsi_rebuild(vsi, false);
4180 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4185 ice_vsi_rebuild(vsi, false);
4186 ice_pf_dcb_recfg(pf);
4189 clear_bit(ICE_CFG_BUSY, pf->state);
4194 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4195 * @pf: PF to configure
4197 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4198 * VSI can still Tx/Rx VLAN tagged packets.
4200 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4202 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4203 struct ice_vsi_ctx *ctxt;
4210 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4215 ctxt->info = vsi->info;
4217 ctxt->info.valid_sections =
4218 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4219 ICE_AQ_VSI_PROP_SECURITY_VALID |
4220 ICE_AQ_VSI_PROP_SW_VALID);
4222 /* disable VLAN anti-spoof */
4223 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4224 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4226 /* disable VLAN pruning and keep all other settings */
4227 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4229 /* allow all VLANs on Tx and don't strip on Rx */
4230 ctxt->info.inner_vlan_flags = ICE_AQ_VSI_INNER_VLAN_TX_MODE_ALL |
4231 ICE_AQ_VSI_INNER_VLAN_EMODE_NOTHING;
4233 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4235 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4236 status, ice_aq_str(hw->adminq.sq_last_status));
4238 vsi->info.sec_flags = ctxt->info.sec_flags;
4239 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4240 vsi->info.inner_vlan_flags = ctxt->info.inner_vlan_flags;
4247 * ice_log_pkg_init - log result of DDP package load
4248 * @hw: pointer to hardware info
4249 * @state: state of package load
4251 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4253 struct ice_pf *pf = hw->back;
4256 dev = ice_pf_to_dev(pf);
4259 case ICE_DDP_PKG_SUCCESS:
4260 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4261 hw->active_pkg_name,
4262 hw->active_pkg_ver.major,
4263 hw->active_pkg_ver.minor,
4264 hw->active_pkg_ver.update,
4265 hw->active_pkg_ver.draft);
4267 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4268 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4269 hw->active_pkg_name,
4270 hw->active_pkg_ver.major,
4271 hw->active_pkg_ver.minor,
4272 hw->active_pkg_ver.update,
4273 hw->active_pkg_ver.draft);
4275 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4276 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",
4277 hw->active_pkg_name,
4278 hw->active_pkg_ver.major,
4279 hw->active_pkg_ver.minor,
4280 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4282 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4283 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",
4284 hw->active_pkg_name,
4285 hw->active_pkg_ver.major,
4286 hw->active_pkg_ver.minor,
4287 hw->active_pkg_ver.update,
4288 hw->active_pkg_ver.draft,
4295 case ICE_DDP_PKG_FW_MISMATCH:
4296 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");
4298 case ICE_DDP_PKG_INVALID_FILE:
4299 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4301 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4302 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4304 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4305 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",
4306 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4308 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4309 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");
4311 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4312 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");
4314 case ICE_DDP_PKG_LOAD_ERROR:
4315 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4316 /* poll for reset to complete */
4317 if (ice_check_reset(hw))
4318 dev_err(dev, "Error resetting device. Please reload the driver\n");
4320 case ICE_DDP_PKG_ERR:
4322 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4328 * ice_load_pkg - load/reload the DDP Package file
4329 * @firmware: firmware structure when firmware requested or NULL for reload
4330 * @pf: pointer to the PF instance
4332 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4333 * initialize HW tables.
4336 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4338 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4339 struct device *dev = ice_pf_to_dev(pf);
4340 struct ice_hw *hw = &pf->hw;
4342 /* Load DDP Package */
4343 if (firmware && !hw->pkg_copy) {
4344 state = ice_copy_and_init_pkg(hw, firmware->data,
4346 ice_log_pkg_init(hw, state);
4347 } else if (!firmware && hw->pkg_copy) {
4348 /* Reload package during rebuild after CORER/GLOBR reset */
4349 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4350 ice_log_pkg_init(hw, state);
4352 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4355 if (!ice_is_init_pkg_successful(state)) {
4357 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4361 /* Successful download package is the precondition for advanced
4362 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4364 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4368 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4369 * @pf: pointer to the PF structure
4371 * There is no error returned here because the driver should be able to handle
4372 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4373 * specifically with Tx.
4375 static void ice_verify_cacheline_size(struct ice_pf *pf)
4377 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4378 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4379 ICE_CACHE_LINE_BYTES);
4383 * ice_send_version - update firmware with driver version
4386 * Returns 0 on success, else error code
4388 static int ice_send_version(struct ice_pf *pf)
4390 struct ice_driver_ver dv;
4392 dv.major_ver = 0xff;
4393 dv.minor_ver = 0xff;
4394 dv.build_ver = 0xff;
4395 dv.subbuild_ver = 0;
4396 strscpy((char *)dv.driver_string, UTS_RELEASE,
4397 sizeof(dv.driver_string));
4398 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4402 * ice_init_fdir - Initialize flow director VSI and configuration
4403 * @pf: pointer to the PF instance
4405 * returns 0 on success, negative on error
4407 static int ice_init_fdir(struct ice_pf *pf)
4409 struct device *dev = ice_pf_to_dev(pf);
4410 struct ice_vsi *ctrl_vsi;
4413 /* Side Band Flow Director needs to have a control VSI.
4414 * Allocate it and store it in the PF.
4416 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4418 dev_dbg(dev, "could not create control VSI\n");
4422 err = ice_vsi_open_ctrl(ctrl_vsi);
4424 dev_dbg(dev, "could not open control VSI\n");
4428 mutex_init(&pf->hw.fdir_fltr_lock);
4430 err = ice_fdir_create_dflt_rules(pf);
4437 ice_fdir_release_flows(&pf->hw);
4438 ice_vsi_close(ctrl_vsi);
4440 ice_vsi_release(ctrl_vsi);
4441 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4442 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4443 pf->ctrl_vsi_idx = ICE_NO_VSI;
4449 * ice_get_opt_fw_name - return optional firmware file name or NULL
4450 * @pf: pointer to the PF instance
4452 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4454 /* Optional firmware name same as default with additional dash
4455 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4457 struct pci_dev *pdev = pf->pdev;
4458 char *opt_fw_filename;
4461 /* Determine the name of the optional file using the DSN (two
4462 * dwords following the start of the DSN Capability).
4464 dsn = pci_get_dsn(pdev);
4468 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4469 if (!opt_fw_filename)
4472 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4473 ICE_DDP_PKG_PATH, dsn);
4475 return opt_fw_filename;
4479 * ice_request_fw - Device initialization routine
4480 * @pf: pointer to the PF instance
4482 static void ice_request_fw(struct ice_pf *pf)
4484 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4485 const struct firmware *firmware = NULL;
4486 struct device *dev = ice_pf_to_dev(pf);
4489 /* optional device-specific DDP (if present) overrides the default DDP
4490 * package file. kernel logs a debug message if the file doesn't exist,
4491 * and warning messages for other errors.
4493 if (opt_fw_filename) {
4494 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4496 kfree(opt_fw_filename);
4500 /* request for firmware was successful. Download to device */
4501 ice_load_pkg(firmware, pf);
4502 kfree(opt_fw_filename);
4503 release_firmware(firmware);
4508 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4510 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4514 /* request for firmware was successful. Download to device */
4515 ice_load_pkg(firmware, pf);
4516 release_firmware(firmware);
4520 * ice_print_wake_reason - show the wake up cause in the log
4521 * @pf: pointer to the PF struct
4523 static void ice_print_wake_reason(struct ice_pf *pf)
4525 u32 wus = pf->wakeup_reason;
4526 const char *wake_str;
4528 /* if no wake event, nothing to print */
4532 if (wus & PFPM_WUS_LNKC_M)
4533 wake_str = "Link\n";
4534 else if (wus & PFPM_WUS_MAG_M)
4535 wake_str = "Magic Packet\n";
4536 else if (wus & PFPM_WUS_MNG_M)
4537 wake_str = "Management\n";
4538 else if (wus & PFPM_WUS_FW_RST_WK_M)
4539 wake_str = "Firmware Reset\n";
4541 wake_str = "Unknown\n";
4543 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4547 * ice_register_netdev - register netdev and devlink port
4548 * @pf: pointer to the PF struct
4550 static int ice_register_netdev(struct ice_pf *pf)
4552 struct ice_vsi *vsi;
4555 vsi = ice_get_main_vsi(pf);
4556 if (!vsi || !vsi->netdev)
4559 err = register_netdev(vsi->netdev);
4561 goto err_register_netdev;
4563 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4564 netif_carrier_off(vsi->netdev);
4565 netif_tx_stop_all_queues(vsi->netdev);
4566 err = ice_devlink_create_pf_port(pf);
4568 goto err_devlink_create;
4570 devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
4574 unregister_netdev(vsi->netdev);
4575 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4576 err_register_netdev:
4577 free_netdev(vsi->netdev);
4579 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4584 * ice_probe - Device initialization routine
4585 * @pdev: PCI device information struct
4586 * @ent: entry in ice_pci_tbl
4588 * Returns 0 on success, negative on failure
4591 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4593 struct device *dev = &pdev->dev;
4598 if (pdev->is_virtfn) {
4599 dev_err(dev, "can't probe a virtual function\n");
4603 /* this driver uses devres, see
4604 * Documentation/driver-api/driver-model/devres.rst
4606 err = pcim_enable_device(pdev);
4610 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4612 dev_err(dev, "BAR0 I/O map error %d\n", err);
4616 pf = ice_allocate_pf(dev);
4620 /* initialize Auxiliary index to invalid value */
4623 /* set up for high or low DMA */
4624 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4626 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4630 pci_enable_pcie_error_reporting(pdev);
4631 pci_set_master(pdev);
4634 pci_set_drvdata(pdev, pf);
4635 set_bit(ICE_DOWN, pf->state);
4636 /* Disable service task until DOWN bit is cleared */
4637 set_bit(ICE_SERVICE_DIS, pf->state);
4640 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4641 pci_save_state(pdev);
4644 hw->vendor_id = pdev->vendor;
4645 hw->device_id = pdev->device;
4646 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4647 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4648 hw->subsystem_device_id = pdev->subsystem_device;
4649 hw->bus.device = PCI_SLOT(pdev->devfn);
4650 hw->bus.func = PCI_FUNC(pdev->devfn);
4651 ice_set_ctrlq_len(hw);
4653 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4655 #ifndef CONFIG_DYNAMIC_DEBUG
4657 hw->debug_mask = debug;
4660 err = ice_init_hw(hw);
4662 dev_err(dev, "ice_init_hw failed: %d\n", err);
4664 goto err_exit_unroll;
4667 ice_init_feature_support(pf);
4671 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4672 * set in pf->state, which will cause ice_is_safe_mode to return
4675 if (ice_is_safe_mode(pf)) {
4676 /* we already got function/device capabilities but these don't
4677 * reflect what the driver needs to do in safe mode. Instead of
4678 * adding conditional logic everywhere to ignore these
4679 * device/function capabilities, override them.
4681 ice_set_safe_mode_caps(hw);
4684 hw->ucast_shared = true;
4686 err = ice_init_pf(pf);
4688 dev_err(dev, "ice_init_pf failed: %d\n", err);
4689 goto err_init_pf_unroll;
4692 ice_devlink_init_regions(pf);
4694 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4695 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4696 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4697 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4699 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4700 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4701 pf->hw.tnl.valid_count[TNL_VXLAN];
4702 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4703 UDP_TUNNEL_TYPE_VXLAN;
4706 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4707 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4708 pf->hw.tnl.valid_count[TNL_GENEVE];
4709 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4710 UDP_TUNNEL_TYPE_GENEVE;
4714 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4715 if (!pf->num_alloc_vsi) {
4717 goto err_init_pf_unroll;
4719 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4720 dev_warn(&pf->pdev->dev,
4721 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4722 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4723 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4726 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4730 goto err_init_pf_unroll;
4733 err = ice_init_interrupt_scheme(pf);
4735 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4737 goto err_init_vsi_unroll;
4740 /* In case of MSIX we are going to setup the misc vector right here
4741 * to handle admin queue events etc. In case of legacy and MSI
4742 * the misc functionality and queue processing is combined in
4743 * the same vector and that gets setup at open.
4745 err = ice_req_irq_msix_misc(pf);
4747 dev_err(dev, "setup of misc vector failed: %d\n", err);
4748 goto err_init_interrupt_unroll;
4751 /* create switch struct for the switch element created by FW on boot */
4752 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4753 if (!pf->first_sw) {
4755 goto err_msix_misc_unroll;
4759 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4761 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4763 pf->first_sw->pf = pf;
4765 /* record the sw_id available for later use */
4766 pf->first_sw->sw_id = hw->port_info->sw_id;
4768 err = ice_setup_pf_sw(pf);
4770 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4771 goto err_alloc_sw_unroll;
4774 clear_bit(ICE_SERVICE_DIS, pf->state);
4776 /* tell the firmware we are up */
4777 err = ice_send_version(pf);
4779 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4781 goto err_send_version_unroll;
4784 /* since everything is good, start the service timer */
4785 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4787 err = ice_init_link_events(pf->hw.port_info);
4789 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4790 goto err_send_version_unroll;
4793 /* not a fatal error if this fails */
4794 err = ice_init_nvm_phy_type(pf->hw.port_info);
4796 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4798 /* not a fatal error if this fails */
4799 err = ice_update_link_info(pf->hw.port_info);
4801 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4803 ice_init_link_dflt_override(pf->hw.port_info);
4805 ice_check_link_cfg_err(pf,
4806 pf->hw.port_info->phy.link_info.link_cfg_err);
4808 /* if media available, initialize PHY settings */
4809 if (pf->hw.port_info->phy.link_info.link_info &
4810 ICE_AQ_MEDIA_AVAILABLE) {
4811 /* not a fatal error if this fails */
4812 err = ice_init_phy_user_cfg(pf->hw.port_info);
4814 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4816 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4817 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4820 ice_configure_phy(vsi);
4823 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4826 ice_verify_cacheline_size(pf);
4828 /* Save wakeup reason register for later use */
4829 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4831 /* check for a power management event */
4832 ice_print_wake_reason(pf);
4834 /* clear wake status, all bits */
4835 wr32(hw, PFPM_WUS, U32_MAX);
4837 /* Disable WoL at init, wait for user to enable */
4838 device_set_wakeup_enable(dev, false);
4840 if (ice_is_safe_mode(pf)) {
4841 ice_set_safe_mode_vlan_cfg(pf);
4845 /* initialize DDP driven features */
4846 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4849 if (ice_is_feature_supported(pf, ICE_F_GNSS))
4852 /* Note: Flow director init failure is non-fatal to load */
4853 if (ice_init_fdir(pf))
4854 dev_err(dev, "could not initialize flow director\n");
4856 /* Note: DCB init failure is non-fatal to load */
4857 if (ice_init_pf_dcb(pf, false)) {
4858 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4859 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4861 ice_cfg_lldp_mib_change(&pf->hw, true);
4864 if (ice_init_lag(pf))
4865 dev_warn(dev, "Failed to init link aggregation support\n");
4867 /* print PCI link speed and width */
4868 pcie_print_link_status(pf->pdev);
4871 err = ice_register_netdev(pf);
4873 goto err_netdev_reg;
4875 err = ice_devlink_register_params(pf);
4877 goto err_netdev_reg;
4879 /* ready to go, so clear down state bit */
4880 clear_bit(ICE_DOWN, pf->state);
4881 if (ice_is_rdma_ena(pf)) {
4882 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4883 if (pf->aux_idx < 0) {
4884 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4886 goto err_devlink_reg_param;
4889 err = ice_init_rdma(pf);
4891 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4893 goto err_init_aux_unroll;
4896 dev_warn(dev, "RDMA is not supported on this device\n");
4899 ice_devlink_register(pf);
4902 err_init_aux_unroll:
4904 ida_free(&ice_aux_ida, pf->aux_idx);
4905 err_devlink_reg_param:
4906 ice_devlink_unregister_params(pf);
4908 err_send_version_unroll:
4909 ice_vsi_release_all(pf);
4910 err_alloc_sw_unroll:
4911 set_bit(ICE_SERVICE_DIS, pf->state);
4912 set_bit(ICE_DOWN, pf->state);
4913 devm_kfree(dev, pf->first_sw);
4914 err_msix_misc_unroll:
4915 ice_free_irq_msix_misc(pf);
4916 err_init_interrupt_unroll:
4917 ice_clear_interrupt_scheme(pf);
4918 err_init_vsi_unroll:
4919 devm_kfree(dev, pf->vsi);
4922 ice_devlink_destroy_regions(pf);
4925 pci_disable_pcie_error_reporting(pdev);
4926 pci_disable_device(pdev);
4931 * ice_set_wake - enable or disable Wake on LAN
4932 * @pf: pointer to the PF struct
4934 * Simple helper for WoL control
4936 static void ice_set_wake(struct ice_pf *pf)
4938 struct ice_hw *hw = &pf->hw;
4939 bool wol = pf->wol_ena;
4941 /* clear wake state, otherwise new wake events won't fire */
4942 wr32(hw, PFPM_WUS, U32_MAX);
4944 /* enable / disable APM wake up, no RMW needed */
4945 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4947 /* set magic packet filter enabled */
4948 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4952 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4953 * @pf: pointer to the PF struct
4955 * Issue firmware command to enable multicast magic wake, making
4956 * sure that any locally administered address (LAA) is used for
4957 * wake, and that PF reset doesn't undo the LAA.
4959 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4961 struct device *dev = ice_pf_to_dev(pf);
4962 struct ice_hw *hw = &pf->hw;
4963 u8 mac_addr[ETH_ALEN];
4964 struct ice_vsi *vsi;
4971 vsi = ice_get_main_vsi(pf);
4975 /* Get current MAC address in case it's an LAA */
4977 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4979 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4981 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4982 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4983 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4985 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4987 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
4988 status, ice_aq_str(hw->adminq.sq_last_status));
4992 * ice_remove - Device removal routine
4993 * @pdev: PCI device information struct
4995 static void ice_remove(struct pci_dev *pdev)
4997 struct ice_pf *pf = pci_get_drvdata(pdev);
5000 ice_devlink_unregister(pf);
5001 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
5002 if (!ice_is_reset_in_progress(pf->state))
5007 ice_tc_indir_block_remove(pf);
5009 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
5010 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
5014 ice_service_task_stop(pf);
5016 ice_aq_cancel_waiting_tasks(pf);
5017 ice_unplug_aux_dev(pf);
5018 if (pf->aux_idx >= 0)
5019 ida_free(&ice_aux_ida, pf->aux_idx);
5020 ice_devlink_unregister_params(pf);
5021 set_bit(ICE_DOWN, pf->state);
5024 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
5025 ice_ptp_release(pf);
5026 if (ice_is_feature_supported(pf, ICE_F_GNSS))
5028 if (!ice_is_safe_mode(pf))
5029 ice_remove_arfs(pf);
5030 ice_setup_mc_magic_wake(pf);
5031 ice_vsi_release_all(pf);
5032 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
5034 ice_free_irq_msix_misc(pf);
5035 ice_for_each_vsi(pf, i) {
5038 ice_vsi_free_q_vectors(pf->vsi[i]);
5041 ice_devlink_destroy_regions(pf);
5042 ice_deinit_hw(&pf->hw);
5044 /* Issue a PFR as part of the prescribed driver unload flow. Do not
5045 * do it via ice_schedule_reset() since there is no need to rebuild
5046 * and the service task is already stopped.
5048 ice_reset(&pf->hw, ICE_RESET_PFR);
5049 pci_wait_for_pending_transaction(pdev);
5050 ice_clear_interrupt_scheme(pf);
5051 pci_disable_pcie_error_reporting(pdev);
5052 pci_disable_device(pdev);
5056 * ice_shutdown - PCI callback for shutting down device
5057 * @pdev: PCI device information struct
5059 static void ice_shutdown(struct pci_dev *pdev)
5061 struct ice_pf *pf = pci_get_drvdata(pdev);
5065 if (system_state == SYSTEM_POWER_OFF) {
5066 pci_wake_from_d3(pdev, pf->wol_ena);
5067 pci_set_power_state(pdev, PCI_D3hot);
5073 * ice_prepare_for_shutdown - prep for PCI shutdown
5074 * @pf: board private structure
5076 * Inform or close all dependent features in prep for PCI device shutdown
5078 static void ice_prepare_for_shutdown(struct ice_pf *pf)
5080 struct ice_hw *hw = &pf->hw;
5083 /* Notify VFs of impending reset */
5084 if (ice_check_sq_alive(hw, &hw->mailboxq))
5085 ice_vc_notify_reset(pf);
5087 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
5089 /* disable the VSIs and their queues that are not already DOWN */
5090 ice_pf_dis_all_vsi(pf, false);
5092 ice_for_each_vsi(pf, v)
5094 pf->vsi[v]->vsi_num = 0;
5096 ice_shutdown_all_ctrlq(hw);
5100 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
5101 * @pf: board private structure to reinitialize
5103 * This routine reinitialize interrupt scheme that was cleared during
5104 * power management suspend callback.
5106 * This should be called during resume routine to re-allocate the q_vectors
5107 * and reacquire interrupts.
5109 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
5111 struct device *dev = ice_pf_to_dev(pf);
5114 /* Since we clear MSIX flag during suspend, we need to
5115 * set it back during resume...
5118 ret = ice_init_interrupt_scheme(pf);
5120 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
5124 /* Remap vectors and rings, after successful re-init interrupts */
5125 ice_for_each_vsi(pf, v) {
5129 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
5132 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
5135 ret = ice_req_irq_msix_misc(pf);
5137 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5147 ice_vsi_free_q_vectors(pf->vsi[v]);
5154 * @dev: generic device information structure
5156 * Power Management callback to quiesce the device and prepare
5157 * for D3 transition.
5159 static int __maybe_unused ice_suspend(struct device *dev)
5161 struct pci_dev *pdev = to_pci_dev(dev);
5165 pf = pci_get_drvdata(pdev);
5167 if (!ice_pf_state_is_nominal(pf)) {
5168 dev_err(dev, "Device is not ready, no need to suspend it\n");
5172 /* Stop watchdog tasks until resume completion.
5173 * Even though it is most likely that the service task is
5174 * disabled if the device is suspended or down, the service task's
5175 * state is controlled by a different state bit, and we should
5176 * store and honor whatever state that bit is in at this point.
5178 disabled = ice_service_task_stop(pf);
5180 ice_unplug_aux_dev(pf);
5182 /* Already suspended?, then there is nothing to do */
5183 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5185 ice_service_task_restart(pf);
5189 if (test_bit(ICE_DOWN, pf->state) ||
5190 ice_is_reset_in_progress(pf->state)) {
5191 dev_err(dev, "can't suspend device in reset or already down\n");
5193 ice_service_task_restart(pf);
5197 ice_setup_mc_magic_wake(pf);
5199 ice_prepare_for_shutdown(pf);
5203 /* Free vectors, clear the interrupt scheme and release IRQs
5204 * for proper hibernation, especially with large number of CPUs.
5205 * Otherwise hibernation might fail when mapping all the vectors back
5208 ice_free_irq_msix_misc(pf);
5209 ice_for_each_vsi(pf, v) {
5212 ice_vsi_free_q_vectors(pf->vsi[v]);
5214 ice_clear_interrupt_scheme(pf);
5216 pci_save_state(pdev);
5217 pci_wake_from_d3(pdev, pf->wol_ena);
5218 pci_set_power_state(pdev, PCI_D3hot);
5223 * ice_resume - PM callback for waking up from D3
5224 * @dev: generic device information structure
5226 static int __maybe_unused ice_resume(struct device *dev)
5228 struct pci_dev *pdev = to_pci_dev(dev);
5229 enum ice_reset_req reset_type;
5234 pci_set_power_state(pdev, PCI_D0);
5235 pci_restore_state(pdev);
5236 pci_save_state(pdev);
5238 if (!pci_device_is_present(pdev))
5241 ret = pci_enable_device_mem(pdev);
5243 dev_err(dev, "Cannot enable device after suspend\n");
5247 pf = pci_get_drvdata(pdev);
5250 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5251 ice_print_wake_reason(pf);
5253 /* We cleared the interrupt scheme when we suspended, so we need to
5254 * restore it now to resume device functionality.
5256 ret = ice_reinit_interrupt_scheme(pf);
5258 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5260 clear_bit(ICE_DOWN, pf->state);
5261 /* Now perform PF reset and rebuild */
5262 reset_type = ICE_RESET_PFR;
5263 /* re-enable service task for reset, but allow reset to schedule it */
5264 clear_bit(ICE_SERVICE_DIS, pf->state);
5266 if (ice_schedule_reset(pf, reset_type))
5267 dev_err(dev, "Reset during resume failed.\n");
5269 clear_bit(ICE_SUSPENDED, pf->state);
5270 ice_service_task_restart(pf);
5272 /* Restart the service task */
5273 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5277 #endif /* CONFIG_PM */
5280 * ice_pci_err_detected - warning that PCI error has been detected
5281 * @pdev: PCI device information struct
5282 * @err: the type of PCI error
5284 * Called to warn that something happened on the PCI bus and the error handling
5285 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5287 static pci_ers_result_t
5288 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5290 struct ice_pf *pf = pci_get_drvdata(pdev);
5293 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5295 return PCI_ERS_RESULT_DISCONNECT;
5298 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5299 ice_service_task_stop(pf);
5301 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5302 set_bit(ICE_PFR_REQ, pf->state);
5303 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5307 return PCI_ERS_RESULT_NEED_RESET;
5311 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5312 * @pdev: PCI device information struct
5314 * Called to determine if the driver can recover from the PCI slot reset by
5315 * using a register read to determine if the device is recoverable.
5317 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5319 struct ice_pf *pf = pci_get_drvdata(pdev);
5320 pci_ers_result_t result;
5324 err = pci_enable_device_mem(pdev);
5326 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5328 result = PCI_ERS_RESULT_DISCONNECT;
5330 pci_set_master(pdev);
5331 pci_restore_state(pdev);
5332 pci_save_state(pdev);
5333 pci_wake_from_d3(pdev, false);
5335 /* Check for life */
5336 reg = rd32(&pf->hw, GLGEN_RTRIG);
5338 result = PCI_ERS_RESULT_RECOVERED;
5340 result = PCI_ERS_RESULT_DISCONNECT;
5343 err = pci_aer_clear_nonfatal_status(pdev);
5345 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
5347 /* non-fatal, continue */
5353 * ice_pci_err_resume - restart operations after PCI error recovery
5354 * @pdev: PCI device information struct
5356 * Called to allow the driver to bring things back up after PCI error and/or
5357 * reset recovery have finished
5359 static void ice_pci_err_resume(struct pci_dev *pdev)
5361 struct ice_pf *pf = pci_get_drvdata(pdev);
5364 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5369 if (test_bit(ICE_SUSPENDED, pf->state)) {
5370 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5375 ice_restore_all_vfs_msi_state(pdev);
5377 ice_do_reset(pf, ICE_RESET_PFR);
5378 ice_service_task_restart(pf);
5379 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5383 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5384 * @pdev: PCI device information struct
5386 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5388 struct ice_pf *pf = pci_get_drvdata(pdev);
5390 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5391 ice_service_task_stop(pf);
5393 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5394 set_bit(ICE_PFR_REQ, pf->state);
5395 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5401 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5402 * @pdev: PCI device information struct
5404 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5406 ice_pci_err_resume(pdev);
5409 /* ice_pci_tbl - PCI Device ID Table
5411 * Wildcard entries (PCI_ANY_ID) should come last
5412 * Last entry must be all 0s
5414 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5415 * Class, Class Mask, private data (not used) }
5417 static const struct pci_device_id ice_pci_tbl[] = {
5418 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5419 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5420 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5421 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5422 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5423 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5424 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5425 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5426 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5427 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5428 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5429 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5430 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5431 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5432 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5433 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5434 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5435 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5436 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5437 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5438 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5439 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5440 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5441 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5442 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5443 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822_SI_DFLT), 0 },
5444 /* required last entry */
5447 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5449 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5451 static const struct pci_error_handlers ice_pci_err_handler = {
5452 .error_detected = ice_pci_err_detected,
5453 .slot_reset = ice_pci_err_slot_reset,
5454 .reset_prepare = ice_pci_err_reset_prepare,
5455 .reset_done = ice_pci_err_reset_done,
5456 .resume = ice_pci_err_resume
5459 static struct pci_driver ice_driver = {
5460 .name = KBUILD_MODNAME,
5461 .id_table = ice_pci_tbl,
5463 .remove = ice_remove,
5465 .driver.pm = &ice_pm_ops,
5466 #endif /* CONFIG_PM */
5467 .shutdown = ice_shutdown,
5468 .sriov_configure = ice_sriov_configure,
5469 .err_handler = &ice_pci_err_handler
5473 * ice_module_init - Driver registration routine
5475 * ice_module_init is the first routine called when the driver is
5476 * loaded. All it does is register with the PCI subsystem.
5478 static int __init ice_module_init(void)
5482 pr_info("%s\n", ice_driver_string);
5483 pr_info("%s\n", ice_copyright);
5485 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5487 pr_err("Failed to create workqueue\n");
5491 status = pci_register_driver(&ice_driver);
5493 pr_err("failed to register PCI driver, err %d\n", status);
5494 destroy_workqueue(ice_wq);
5499 module_init(ice_module_init);
5502 * ice_module_exit - Driver exit cleanup routine
5504 * ice_module_exit is called just before the driver is removed
5507 static void __exit ice_module_exit(void)
5509 pci_unregister_driver(&ice_driver);
5510 destroy_workqueue(ice_wq);
5511 pr_info("module unloaded\n");
5513 module_exit(ice_module_exit);
5516 * ice_set_mac_address - NDO callback to set MAC address
5517 * @netdev: network interface device structure
5518 * @pi: pointer to an address structure
5520 * Returns 0 on success, negative on failure
5522 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5524 struct ice_netdev_priv *np = netdev_priv(netdev);
5525 struct ice_vsi *vsi = np->vsi;
5526 struct ice_pf *pf = vsi->back;
5527 struct ice_hw *hw = &pf->hw;
5528 struct sockaddr *addr = pi;
5529 u8 old_mac[ETH_ALEN];
5534 mac = (u8 *)addr->sa_data;
5536 if (!is_valid_ether_addr(mac))
5537 return -EADDRNOTAVAIL;
5539 if (ether_addr_equal(netdev->dev_addr, mac)) {
5540 netdev_dbg(netdev, "already using mac %pM\n", mac);
5544 if (test_bit(ICE_DOWN, pf->state) ||
5545 ice_is_reset_in_progress(pf->state)) {
5546 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5551 if (ice_chnl_dmac_fltr_cnt(pf)) {
5552 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5557 netif_addr_lock_bh(netdev);
5558 ether_addr_copy(old_mac, netdev->dev_addr);
5559 /* change the netdev's MAC address */
5560 eth_hw_addr_set(netdev, mac);
5561 netif_addr_unlock_bh(netdev);
5563 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5564 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5565 if (err && err != -ENOENT) {
5566 err = -EADDRNOTAVAIL;
5567 goto err_update_filters;
5570 /* Add filter for new MAC. If filter exists, return success */
5571 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5572 if (err == -EEXIST) {
5573 /* Although this MAC filter is already present in hardware it's
5574 * possible in some cases (e.g. bonding) that dev_addr was
5575 * modified outside of the driver and needs to be restored back
5578 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5582 /* error if the new filter addition failed */
5583 err = -EADDRNOTAVAIL;
5588 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5590 netif_addr_lock_bh(netdev);
5591 eth_hw_addr_set(netdev, old_mac);
5592 netif_addr_unlock_bh(netdev);
5596 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5599 /* write new MAC address to the firmware */
5600 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5601 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5603 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5610 * ice_set_rx_mode - NDO callback to set the netdev filters
5611 * @netdev: network interface device structure
5613 static void ice_set_rx_mode(struct net_device *netdev)
5615 struct ice_netdev_priv *np = netdev_priv(netdev);
5616 struct ice_vsi *vsi = np->vsi;
5621 /* Set the flags to synchronize filters
5622 * ndo_set_rx_mode may be triggered even without a change in netdev
5625 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5626 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5627 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5629 /* schedule our worker thread which will take care of
5630 * applying the new filter changes
5632 ice_service_task_schedule(vsi->back);
5636 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5637 * @netdev: network interface device structure
5638 * @queue_index: Queue ID
5639 * @maxrate: maximum bandwidth in Mbps
5642 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5644 struct ice_netdev_priv *np = netdev_priv(netdev);
5645 struct ice_vsi *vsi = np->vsi;
5650 /* Validate maxrate requested is within permitted range */
5651 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5652 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5653 maxrate, queue_index);
5657 q_handle = vsi->tx_rings[queue_index]->q_handle;
5658 tc = ice_dcb_get_tc(vsi, queue_index);
5660 /* Set BW back to default, when user set maxrate to 0 */
5662 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5663 q_handle, ICE_MAX_BW);
5665 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5666 q_handle, ICE_MAX_BW, maxrate * 1000);
5668 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5675 * ice_fdb_add - add an entry to the hardware database
5676 * @ndm: the input from the stack
5677 * @tb: pointer to array of nladdr (unused)
5678 * @dev: the net device pointer
5679 * @addr: the MAC address entry being added
5681 * @flags: instructions from stack about fdb operation
5682 * @extack: netlink extended ack
5685 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5686 struct net_device *dev, const unsigned char *addr, u16 vid,
5687 u16 flags, struct netlink_ext_ack __always_unused *extack)
5692 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5695 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5696 netdev_err(dev, "FDB only supports static addresses\n");
5700 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5701 err = dev_uc_add_excl(dev, addr);
5702 else if (is_multicast_ether_addr(addr))
5703 err = dev_mc_add_excl(dev, addr);
5707 /* Only return duplicate errors if NLM_F_EXCL is set */
5708 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5715 * ice_fdb_del - delete an entry from the hardware database
5716 * @ndm: the input from the stack
5717 * @tb: pointer to array of nladdr (unused)
5718 * @dev: the net device pointer
5719 * @addr: the MAC address entry being added
5721 * @extack: netlink extended ack
5724 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5725 struct net_device *dev, const unsigned char *addr,
5726 __always_unused u16 vid, struct netlink_ext_ack *extack)
5730 if (ndm->ndm_state & NUD_PERMANENT) {
5731 netdev_err(dev, "FDB only supports static addresses\n");
5735 if (is_unicast_ether_addr(addr))
5736 err = dev_uc_del(dev, addr);
5737 else if (is_multicast_ether_addr(addr))
5738 err = dev_mc_del(dev, addr);
5745 #define NETIF_VLAN_OFFLOAD_FEATURES (NETIF_F_HW_VLAN_CTAG_RX | \
5746 NETIF_F_HW_VLAN_CTAG_TX | \
5747 NETIF_F_HW_VLAN_STAG_RX | \
5748 NETIF_F_HW_VLAN_STAG_TX)
5750 #define NETIF_VLAN_FILTERING_FEATURES (NETIF_F_HW_VLAN_CTAG_FILTER | \
5751 NETIF_F_HW_VLAN_STAG_FILTER)
5754 * ice_fix_features - fix the netdev features flags based on device limitations
5755 * @netdev: ptr to the netdev that flags are being fixed on
5756 * @features: features that need to be checked and possibly fixed
5758 * Make sure any fixups are made to features in this callback. This enables the
5759 * driver to not have to check unsupported configurations throughout the driver
5760 * because that's the responsiblity of this callback.
5762 * Single VLAN Mode (SVM) Supported Features:
5763 * NETIF_F_HW_VLAN_CTAG_FILTER
5764 * NETIF_F_HW_VLAN_CTAG_RX
5765 * NETIF_F_HW_VLAN_CTAG_TX
5767 * Double VLAN Mode (DVM) Supported Features:
5768 * NETIF_F_HW_VLAN_CTAG_FILTER
5769 * NETIF_F_HW_VLAN_CTAG_RX
5770 * NETIF_F_HW_VLAN_CTAG_TX
5772 * NETIF_F_HW_VLAN_STAG_FILTER
5773 * NETIF_HW_VLAN_STAG_RX
5774 * NETIF_HW_VLAN_STAG_TX
5776 * Features that need fixing:
5777 * Cannot simultaneously enable CTAG and STAG stripping and/or insertion.
5778 * These are mutually exlusive as the VSI context cannot support multiple
5779 * VLAN ethertypes simultaneously for stripping and/or insertion. If this
5780 * is not done, then default to clearing the requested STAG offload
5783 * All supported filtering has to be enabled or disabled together. For
5784 * example, in DVM, CTAG and STAG filtering have to be enabled and disabled
5785 * together. If this is not done, then default to VLAN filtering disabled.
5786 * These are mutually exclusive as there is currently no way to
5787 * enable/disable VLAN filtering based on VLAN ethertype when using VLAN
5790 static netdev_features_t
5791 ice_fix_features(struct net_device *netdev, netdev_features_t features)
5793 struct ice_netdev_priv *np = netdev_priv(netdev);
5794 netdev_features_t req_vlan_fltr, cur_vlan_fltr;
5795 bool cur_ctag, cur_stag, req_ctag, req_stag;
5797 cur_vlan_fltr = netdev->features & NETIF_VLAN_FILTERING_FEATURES;
5798 cur_ctag = cur_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5799 cur_stag = cur_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
5801 req_vlan_fltr = features & NETIF_VLAN_FILTERING_FEATURES;
5802 req_ctag = req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER;
5803 req_stag = req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER;
5805 if (req_vlan_fltr != cur_vlan_fltr) {
5806 if (ice_is_dvm_ena(&np->vsi->back->hw)) {
5807 if (req_ctag && req_stag) {
5808 features |= NETIF_VLAN_FILTERING_FEATURES;
5809 } else if (!req_ctag && !req_stag) {
5810 features &= ~NETIF_VLAN_FILTERING_FEATURES;
5811 } else if ((!cur_ctag && req_ctag && !cur_stag) ||
5812 (!cur_stag && req_stag && !cur_ctag)) {
5813 features |= NETIF_VLAN_FILTERING_FEATURES;
5814 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");
5815 } else if ((cur_ctag && !req_ctag && cur_stag) ||
5816 (cur_stag && !req_stag && cur_ctag)) {
5817 features &= ~NETIF_VLAN_FILTERING_FEATURES;
5818 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");
5821 if (req_vlan_fltr & NETIF_F_HW_VLAN_STAG_FILTER)
5822 netdev_warn(netdev, "cannot support requested 802.1ad filtering setting in SVM mode\n");
5824 if (req_vlan_fltr & NETIF_F_HW_VLAN_CTAG_FILTER)
5825 features |= NETIF_F_HW_VLAN_CTAG_FILTER;
5829 if ((features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX)) &&
5830 (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))) {
5831 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");
5832 features &= ~(NETIF_F_HW_VLAN_STAG_RX |
5833 NETIF_F_HW_VLAN_STAG_TX);
5840 * ice_set_vlan_offload_features - set VLAN offload features for the PF VSI
5842 * @features: features used to determine VLAN offload settings
5844 * First, determine the vlan_ethertype based on the VLAN offload bits in
5845 * features. Then determine if stripping and insertion should be enabled or
5846 * disabled. Finally enable or disable VLAN stripping and insertion.
5849 ice_set_vlan_offload_features(struct ice_vsi *vsi, netdev_features_t features)
5851 bool enable_stripping = true, enable_insertion = true;
5852 struct ice_vsi_vlan_ops *vlan_ops;
5853 int strip_err = 0, insert_err = 0;
5854 u16 vlan_ethertype = 0;
5856 vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5858 if (features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_STAG_TX))
5859 vlan_ethertype = ETH_P_8021AD;
5860 else if (features & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX))
5861 vlan_ethertype = ETH_P_8021Q;
5863 if (!(features & (NETIF_F_HW_VLAN_STAG_RX | NETIF_F_HW_VLAN_CTAG_RX)))
5864 enable_stripping = false;
5865 if (!(features & (NETIF_F_HW_VLAN_STAG_TX | NETIF_F_HW_VLAN_CTAG_TX)))
5866 enable_insertion = false;
5868 if (enable_stripping)
5869 strip_err = vlan_ops->ena_stripping(vsi, vlan_ethertype);
5871 strip_err = vlan_ops->dis_stripping(vsi);
5873 if (enable_insertion)
5874 insert_err = vlan_ops->ena_insertion(vsi, vlan_ethertype);
5876 insert_err = vlan_ops->dis_insertion(vsi);
5878 if (strip_err || insert_err)
5885 * ice_set_vlan_filtering_features - set VLAN filtering features for the PF VSI
5887 * @features: features used to determine VLAN filtering settings
5889 * Enable or disable Rx VLAN filtering based on the VLAN filtering bits in the
5893 ice_set_vlan_filtering_features(struct ice_vsi *vsi, netdev_features_t features)
5895 struct ice_vsi_vlan_ops *vlan_ops = ice_get_compat_vsi_vlan_ops(vsi);
5898 /* support Single VLAN Mode (SVM) and Double VLAN Mode (DVM) by checking
5899 * if either bit is set
5902 (NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_HW_VLAN_STAG_FILTER))
5903 err = vlan_ops->ena_rx_filtering(vsi);
5905 err = vlan_ops->dis_rx_filtering(vsi);
5911 * ice_set_vlan_features - set VLAN settings based on suggested feature set
5912 * @netdev: ptr to the netdev being adjusted
5913 * @features: the feature set that the stack is suggesting
5915 * Only update VLAN settings if the requested_vlan_features are different than
5916 * the current_vlan_features.
5919 ice_set_vlan_features(struct net_device *netdev, netdev_features_t features)
5921 netdev_features_t current_vlan_features, requested_vlan_features;
5922 struct ice_netdev_priv *np = netdev_priv(netdev);
5923 struct ice_vsi *vsi = np->vsi;
5926 current_vlan_features = netdev->features & NETIF_VLAN_OFFLOAD_FEATURES;
5927 requested_vlan_features = features & NETIF_VLAN_OFFLOAD_FEATURES;
5928 if (current_vlan_features ^ requested_vlan_features) {
5929 err = ice_set_vlan_offload_features(vsi, features);
5934 current_vlan_features = netdev->features &
5935 NETIF_VLAN_FILTERING_FEATURES;
5936 requested_vlan_features = features & NETIF_VLAN_FILTERING_FEATURES;
5937 if (current_vlan_features ^ requested_vlan_features) {
5938 err = ice_set_vlan_filtering_features(vsi, features);
5947 * ice_set_features - set the netdev feature flags
5948 * @netdev: ptr to the netdev being adjusted
5949 * @features: the feature set that the stack is suggesting
5952 ice_set_features(struct net_device *netdev, netdev_features_t features)
5954 struct ice_netdev_priv *np = netdev_priv(netdev);
5955 struct ice_vsi *vsi = np->vsi;
5956 struct ice_pf *pf = vsi->back;
5959 /* Don't set any netdev advanced features with device in Safe Mode */
5960 if (ice_is_safe_mode(vsi->back)) {
5961 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5965 /* Do not change setting during reset */
5966 if (ice_is_reset_in_progress(pf->state)) {
5967 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5971 /* Multiple features can be changed in one call so keep features in
5972 * separate if/else statements to guarantee each feature is checked
5974 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5975 ice_vsi_manage_rss_lut(vsi, true);
5976 else if (!(features & NETIF_F_RXHASH) &&
5977 netdev->features & NETIF_F_RXHASH)
5978 ice_vsi_manage_rss_lut(vsi, false);
5980 ret = ice_set_vlan_features(netdev, features);
5984 if ((features & NETIF_F_NTUPLE) &&
5985 !(netdev->features & NETIF_F_NTUPLE)) {
5986 ice_vsi_manage_fdir(vsi, true);
5988 } else if (!(features & NETIF_F_NTUPLE) &&
5989 (netdev->features & NETIF_F_NTUPLE)) {
5990 ice_vsi_manage_fdir(vsi, false);
5991 ice_clear_arfs(vsi);
5994 /* don't turn off hw_tc_offload when ADQ is already enabled */
5995 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
5996 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
6000 if ((features & NETIF_F_HW_TC) &&
6001 !(netdev->features & NETIF_F_HW_TC))
6002 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6004 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
6010 * ice_vsi_vlan_setup - Setup VLAN offload properties on a PF VSI
6011 * @vsi: VSI to setup VLAN properties for
6013 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
6017 err = ice_set_vlan_offload_features(vsi, vsi->netdev->features);
6021 err = ice_set_vlan_filtering_features(vsi, vsi->netdev->features);
6025 return ice_vsi_add_vlan_zero(vsi);
6029 * ice_vsi_cfg - Setup the VSI
6030 * @vsi: the VSI being configured
6032 * Return 0 on success and negative value on error
6034 int ice_vsi_cfg(struct ice_vsi *vsi)
6039 ice_set_rx_mode(vsi->netdev);
6041 if (vsi->type != ICE_VSI_LB) {
6042 err = ice_vsi_vlan_setup(vsi);
6048 ice_vsi_cfg_dcb_rings(vsi);
6050 err = ice_vsi_cfg_lan_txqs(vsi);
6051 if (!err && ice_is_xdp_ena_vsi(vsi))
6052 err = ice_vsi_cfg_xdp_txqs(vsi);
6054 err = ice_vsi_cfg_rxqs(vsi);
6059 /* THEORY OF MODERATION:
6060 * The ice driver hardware works differently than the hardware that DIMLIB was
6061 * originally made for. ice hardware doesn't have packet count limits that
6062 * can trigger an interrupt, but it *does* have interrupt rate limit support,
6063 * which is hard-coded to a limit of 250,000 ints/second.
6064 * If not using dynamic moderation, the INTRL value can be modified
6065 * by ethtool rx-usecs-high.
6068 /* the throttle rate for interrupts, basically worst case delay before
6069 * an initial interrupt fires, value is stored in microseconds.
6074 /* Make a different profile for Rx that doesn't allow quite so aggressive
6075 * moderation at the high end (it maxes out at 126us or about 8k interrupts a
6078 static const struct ice_dim rx_profile[] = {
6079 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6080 {8}, /* 125,000 ints/s */
6081 {16}, /* 62,500 ints/s */
6082 {62}, /* 16,129 ints/s */
6083 {126} /* 7,936 ints/s */
6086 /* The transmit profile, which has the same sorts of values
6087 * as the previous struct
6089 static const struct ice_dim tx_profile[] = {
6090 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
6091 {8}, /* 125,000 ints/s */
6092 {40}, /* 16,125 ints/s */
6093 {128}, /* 7,812 ints/s */
6094 {256} /* 3,906 ints/s */
6097 static void ice_tx_dim_work(struct work_struct *work)
6099 struct ice_ring_container *rc;
6103 dim = container_of(work, struct dim, work);
6104 rc = (struct ice_ring_container *)dim->priv;
6106 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
6108 /* look up the values in our local table */
6109 itr = tx_profile[dim->profile_ix].itr;
6111 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
6112 ice_write_itr(rc, itr);
6114 dim->state = DIM_START_MEASURE;
6117 static void ice_rx_dim_work(struct work_struct *work)
6119 struct ice_ring_container *rc;
6123 dim = container_of(work, struct dim, work);
6124 rc = (struct ice_ring_container *)dim->priv;
6126 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
6128 /* look up the values in our local table */
6129 itr = rx_profile[dim->profile_ix].itr;
6131 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
6132 ice_write_itr(rc, itr);
6134 dim->state = DIM_START_MEASURE;
6137 #define ICE_DIM_DEFAULT_PROFILE_IX 1
6140 * ice_init_moderation - set up interrupt moderation
6141 * @q_vector: the vector containing rings to be configured
6143 * Set up interrupt moderation registers, with the intent to do the right thing
6144 * when called from reset or from probe, and whether or not dynamic moderation
6145 * is enabled or not. Take special care to write all the registers in both
6146 * dynamic moderation mode or not in order to make sure hardware is in a known
6149 static void ice_init_moderation(struct ice_q_vector *q_vector)
6151 struct ice_ring_container *rc;
6152 bool tx_dynamic, rx_dynamic;
6155 INIT_WORK(&rc->dim.work, ice_tx_dim_work);
6156 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6157 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6159 tx_dynamic = ITR_IS_DYNAMIC(rc);
6161 /* set the initial TX ITR to match the above */
6162 ice_write_itr(rc, tx_dynamic ?
6163 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
6166 INIT_WORK(&rc->dim.work, ice_rx_dim_work);
6167 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
6168 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
6170 rx_dynamic = ITR_IS_DYNAMIC(rc);
6172 /* set the initial RX ITR to match the above */
6173 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
6176 ice_set_q_vector_intrl(q_vector);
6180 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
6181 * @vsi: the VSI being configured
6183 static void ice_napi_enable_all(struct ice_vsi *vsi)
6190 ice_for_each_q_vector(vsi, q_idx) {
6191 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6193 ice_init_moderation(q_vector);
6195 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6196 napi_enable(&q_vector->napi);
6201 * ice_up_complete - Finish the last steps of bringing up a connection
6202 * @vsi: The VSI being configured
6204 * Return 0 on success and negative value on error
6206 static int ice_up_complete(struct ice_vsi *vsi)
6208 struct ice_pf *pf = vsi->back;
6211 ice_vsi_cfg_msix(vsi);
6213 /* Enable only Rx rings, Tx rings were enabled by the FW when the
6214 * Tx queue group list was configured and the context bits were
6215 * programmed using ice_vsi_cfg_txqs
6217 err = ice_vsi_start_all_rx_rings(vsi);
6221 clear_bit(ICE_VSI_DOWN, vsi->state);
6222 ice_napi_enable_all(vsi);
6223 ice_vsi_ena_irq(vsi);
6225 if (vsi->port_info &&
6226 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
6228 ice_print_link_msg(vsi, true);
6229 netif_tx_start_all_queues(vsi->netdev);
6230 netif_carrier_on(vsi->netdev);
6231 if (!ice_is_e810(&pf->hw))
6232 ice_ptp_link_change(pf, pf->hw.pf_id, true);
6235 /* Perform an initial read of the statistics registers now to
6236 * set the baseline so counters are ready when interface is up
6238 ice_update_eth_stats(vsi);
6239 ice_service_task_schedule(pf);
6245 * ice_up - Bring the connection back up after being down
6246 * @vsi: VSI being configured
6248 int ice_up(struct ice_vsi *vsi)
6252 err = ice_vsi_cfg(vsi);
6254 err = ice_up_complete(vsi);
6260 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
6261 * @syncp: pointer to u64_stats_sync
6262 * @stats: stats that pkts and bytes count will be taken from
6263 * @pkts: packets stats counter
6264 * @bytes: bytes stats counter
6266 * This function fetches stats from the ring considering the atomic operations
6267 * that needs to be performed to read u64 values in 32 bit machine.
6270 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp,
6271 struct ice_q_stats stats, u64 *pkts, u64 *bytes)
6276 start = u64_stats_fetch_begin_irq(syncp);
6278 *bytes = stats.bytes;
6279 } while (u64_stats_fetch_retry_irq(syncp, start));
6283 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
6284 * @vsi: the VSI to be updated
6285 * @vsi_stats: the stats struct to be updated
6286 * @rings: rings to work on
6287 * @count: number of rings
6290 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
6291 struct rtnl_link_stats64 *vsi_stats,
6292 struct ice_tx_ring **rings, u16 count)
6296 for (i = 0; i < count; i++) {
6297 struct ice_tx_ring *ring;
6298 u64 pkts = 0, bytes = 0;
6300 ring = READ_ONCE(rings[i]);
6303 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6304 vsi_stats->tx_packets += pkts;
6305 vsi_stats->tx_bytes += bytes;
6306 vsi->tx_restart += ring->tx_stats.restart_q;
6307 vsi->tx_busy += ring->tx_stats.tx_busy;
6308 vsi->tx_linearize += ring->tx_stats.tx_linearize;
6313 * ice_update_vsi_ring_stats - Update VSI stats counters
6314 * @vsi: the VSI to be updated
6316 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
6318 struct rtnl_link_stats64 *vsi_stats;
6322 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
6326 /* reset non-netdev (extended) stats */
6327 vsi->tx_restart = 0;
6329 vsi->tx_linearize = 0;
6330 vsi->rx_buf_failed = 0;
6331 vsi->rx_page_failed = 0;
6335 /* update Tx rings counters */
6336 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6339 /* update Rx rings counters */
6340 ice_for_each_rxq(vsi, i) {
6341 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6343 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6344 vsi_stats->rx_packets += pkts;
6345 vsi_stats->rx_bytes += bytes;
6346 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
6347 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
6350 /* update XDP Tx rings counters */
6351 if (ice_is_xdp_ena_vsi(vsi))
6352 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6357 vsi->net_stats.tx_packets = vsi_stats->tx_packets;
6358 vsi->net_stats.tx_bytes = vsi_stats->tx_bytes;
6359 vsi->net_stats.rx_packets = vsi_stats->rx_packets;
6360 vsi->net_stats.rx_bytes = vsi_stats->rx_bytes;
6366 * ice_update_vsi_stats - Update VSI stats counters
6367 * @vsi: the VSI to be updated
6369 void ice_update_vsi_stats(struct ice_vsi *vsi)
6371 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6372 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6373 struct ice_pf *pf = vsi->back;
6375 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6376 test_bit(ICE_CFG_BUSY, pf->state))
6379 /* get stats as recorded by Tx/Rx rings */
6380 ice_update_vsi_ring_stats(vsi);
6382 /* get VSI stats as recorded by the hardware */
6383 ice_update_eth_stats(vsi);
6385 cur_ns->tx_errors = cur_es->tx_errors;
6386 cur_ns->rx_dropped = cur_es->rx_discards;
6387 cur_ns->tx_dropped = cur_es->tx_discards;
6388 cur_ns->multicast = cur_es->rx_multicast;
6390 /* update some more netdev stats if this is main VSI */
6391 if (vsi->type == ICE_VSI_PF) {
6392 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6393 cur_ns->rx_errors = pf->stats.crc_errors +
6394 pf->stats.illegal_bytes +
6395 pf->stats.rx_len_errors +
6396 pf->stats.rx_undersize +
6397 pf->hw_csum_rx_error +
6398 pf->stats.rx_jabber +
6399 pf->stats.rx_fragments +
6400 pf->stats.rx_oversize;
6401 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6402 /* record drops from the port level */
6403 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6408 * ice_update_pf_stats - Update PF port stats counters
6409 * @pf: PF whose stats needs to be updated
6411 void ice_update_pf_stats(struct ice_pf *pf)
6413 struct ice_hw_port_stats *prev_ps, *cur_ps;
6414 struct ice_hw *hw = &pf->hw;
6418 port = hw->port_info->lport;
6419 prev_ps = &pf->stats_prev;
6420 cur_ps = &pf->stats;
6422 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6423 &prev_ps->eth.rx_bytes,
6424 &cur_ps->eth.rx_bytes);
6426 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6427 &prev_ps->eth.rx_unicast,
6428 &cur_ps->eth.rx_unicast);
6430 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6431 &prev_ps->eth.rx_multicast,
6432 &cur_ps->eth.rx_multicast);
6434 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6435 &prev_ps->eth.rx_broadcast,
6436 &cur_ps->eth.rx_broadcast);
6438 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6439 &prev_ps->eth.rx_discards,
6440 &cur_ps->eth.rx_discards);
6442 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6443 &prev_ps->eth.tx_bytes,
6444 &cur_ps->eth.tx_bytes);
6446 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6447 &prev_ps->eth.tx_unicast,
6448 &cur_ps->eth.tx_unicast);
6450 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6451 &prev_ps->eth.tx_multicast,
6452 &cur_ps->eth.tx_multicast);
6454 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6455 &prev_ps->eth.tx_broadcast,
6456 &cur_ps->eth.tx_broadcast);
6458 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6459 &prev_ps->tx_dropped_link_down,
6460 &cur_ps->tx_dropped_link_down);
6462 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6463 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6465 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6466 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6468 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6469 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6471 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6472 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6474 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6475 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6477 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6478 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6480 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6481 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6483 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6484 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6486 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6487 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6489 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6490 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6492 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6493 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6495 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6496 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6498 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6499 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6501 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6502 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6504 fd_ctr_base = hw->fd_ctr_base;
6506 ice_stat_update40(hw,
6507 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6508 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6509 &cur_ps->fd_sb_match);
6510 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6511 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6513 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6514 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6516 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6517 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6519 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6520 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6522 ice_update_dcb_stats(pf);
6524 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6525 &prev_ps->crc_errors, &cur_ps->crc_errors);
6527 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6528 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6530 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6531 &prev_ps->mac_local_faults,
6532 &cur_ps->mac_local_faults);
6534 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6535 &prev_ps->mac_remote_faults,
6536 &cur_ps->mac_remote_faults);
6538 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6539 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6541 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6542 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6544 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6545 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6547 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6548 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6550 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6551 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6553 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6555 pf->stat_prev_loaded = true;
6559 * ice_get_stats64 - get statistics for network device structure
6560 * @netdev: network interface device structure
6561 * @stats: main device statistics structure
6564 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6566 struct ice_netdev_priv *np = netdev_priv(netdev);
6567 struct rtnl_link_stats64 *vsi_stats;
6568 struct ice_vsi *vsi = np->vsi;
6570 vsi_stats = &vsi->net_stats;
6572 if (!vsi->num_txq || !vsi->num_rxq)
6575 /* netdev packet/byte stats come from ring counter. These are obtained
6576 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6577 * But, only call the update routine and read the registers if VSI is
6580 if (!test_bit(ICE_VSI_DOWN, vsi->state))
6581 ice_update_vsi_ring_stats(vsi);
6582 stats->tx_packets = vsi_stats->tx_packets;
6583 stats->tx_bytes = vsi_stats->tx_bytes;
6584 stats->rx_packets = vsi_stats->rx_packets;
6585 stats->rx_bytes = vsi_stats->rx_bytes;
6587 /* The rest of the stats can be read from the hardware but instead we
6588 * just return values that the watchdog task has already obtained from
6591 stats->multicast = vsi_stats->multicast;
6592 stats->tx_errors = vsi_stats->tx_errors;
6593 stats->tx_dropped = vsi_stats->tx_dropped;
6594 stats->rx_errors = vsi_stats->rx_errors;
6595 stats->rx_dropped = vsi_stats->rx_dropped;
6596 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6597 stats->rx_length_errors = vsi_stats->rx_length_errors;
6601 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6602 * @vsi: VSI having NAPI disabled
6604 static void ice_napi_disable_all(struct ice_vsi *vsi)
6611 ice_for_each_q_vector(vsi, q_idx) {
6612 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6614 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6615 napi_disable(&q_vector->napi);
6617 cancel_work_sync(&q_vector->tx.dim.work);
6618 cancel_work_sync(&q_vector->rx.dim.work);
6623 * ice_down - Shutdown the connection
6624 * @vsi: The VSI being stopped
6626 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
6628 int ice_down(struct ice_vsi *vsi)
6630 int i, tx_err, rx_err, link_err = 0, vlan_err = 0;
6632 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
6634 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6635 vlan_err = ice_vsi_del_vlan_zero(vsi);
6636 if (!ice_is_e810(&vsi->back->hw))
6637 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
6638 netif_carrier_off(vsi->netdev);
6639 netif_tx_disable(vsi->netdev);
6640 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
6641 ice_eswitch_stop_all_tx_queues(vsi->back);
6644 ice_vsi_dis_irq(vsi);
6646 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6648 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6649 vsi->vsi_num, tx_err);
6650 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6651 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6653 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6654 vsi->vsi_num, tx_err);
6657 rx_err = ice_vsi_stop_all_rx_rings(vsi);
6659 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6660 vsi->vsi_num, rx_err);
6662 ice_napi_disable_all(vsi);
6664 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6665 link_err = ice_force_phys_link_state(vsi, false);
6667 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6668 vsi->vsi_num, link_err);
6671 ice_for_each_txq(vsi, i)
6672 ice_clean_tx_ring(vsi->tx_rings[i]);
6674 ice_for_each_rxq(vsi, i)
6675 ice_clean_rx_ring(vsi->rx_rings[i]);
6677 if (tx_err || rx_err || link_err || vlan_err) {
6678 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6679 vsi->vsi_num, vsi->vsw->sw_id);
6687 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6688 * @vsi: VSI having resources allocated
6690 * Return 0 on success, negative on failure
6692 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6696 if (!vsi->num_txq) {
6697 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6702 ice_for_each_txq(vsi, i) {
6703 struct ice_tx_ring *ring = vsi->tx_rings[i];
6709 ring->netdev = vsi->netdev;
6710 err = ice_setup_tx_ring(ring);
6719 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6720 * @vsi: VSI having resources allocated
6722 * Return 0 on success, negative on failure
6724 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6728 if (!vsi->num_rxq) {
6729 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6734 ice_for_each_rxq(vsi, i) {
6735 struct ice_rx_ring *ring = vsi->rx_rings[i];
6741 ring->netdev = vsi->netdev;
6742 err = ice_setup_rx_ring(ring);
6751 * ice_vsi_open_ctrl - open control VSI for use
6752 * @vsi: the VSI to open
6754 * Initialization of the Control VSI
6756 * Returns 0 on success, negative value on error
6758 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6760 char int_name[ICE_INT_NAME_STR_LEN];
6761 struct ice_pf *pf = vsi->back;
6765 dev = ice_pf_to_dev(pf);
6766 /* allocate descriptors */
6767 err = ice_vsi_setup_tx_rings(vsi);
6771 err = ice_vsi_setup_rx_rings(vsi);
6775 err = ice_vsi_cfg(vsi);
6779 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6780 dev_driver_string(dev), dev_name(dev));
6781 err = ice_vsi_req_irq_msix(vsi, int_name);
6785 ice_vsi_cfg_msix(vsi);
6787 err = ice_vsi_start_all_rx_rings(vsi);
6789 goto err_up_complete;
6791 clear_bit(ICE_VSI_DOWN, vsi->state);
6792 ice_vsi_ena_irq(vsi);
6799 ice_vsi_free_rx_rings(vsi);
6801 ice_vsi_free_tx_rings(vsi);
6807 * ice_vsi_open - Called when a network interface is made active
6808 * @vsi: the VSI to open
6810 * Initialization of the VSI
6812 * Returns 0 on success, negative value on error
6814 int ice_vsi_open(struct ice_vsi *vsi)
6816 char int_name[ICE_INT_NAME_STR_LEN];
6817 struct ice_pf *pf = vsi->back;
6820 /* allocate descriptors */
6821 err = ice_vsi_setup_tx_rings(vsi);
6825 err = ice_vsi_setup_rx_rings(vsi);
6829 err = ice_vsi_cfg(vsi);
6833 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6834 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6835 err = ice_vsi_req_irq_msix(vsi, int_name);
6839 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
6841 if (vsi->type == ICE_VSI_PF) {
6842 /* Notify the stack of the actual queue counts. */
6843 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6847 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6852 err = ice_up_complete(vsi);
6854 goto err_up_complete;
6861 ice_vsi_free_irq(vsi);
6863 ice_vsi_free_rx_rings(vsi);
6865 ice_vsi_free_tx_rings(vsi);
6871 * ice_vsi_release_all - Delete all VSIs
6872 * @pf: PF from which all VSIs are being removed
6874 static void ice_vsi_release_all(struct ice_pf *pf)
6881 ice_for_each_vsi(pf, i) {
6885 if (pf->vsi[i]->type == ICE_VSI_CHNL)
6888 err = ice_vsi_release(pf->vsi[i]);
6890 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6891 i, err, pf->vsi[i]->vsi_num);
6896 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6897 * @pf: pointer to the PF instance
6898 * @type: VSI type to rebuild
6900 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6902 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6904 struct device *dev = ice_pf_to_dev(pf);
6907 ice_for_each_vsi(pf, i) {
6908 struct ice_vsi *vsi = pf->vsi[i];
6910 if (!vsi || vsi->type != type)
6913 /* rebuild the VSI */
6914 err = ice_vsi_rebuild(vsi, true);
6916 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6917 err, vsi->idx, ice_vsi_type_str(type));
6921 /* replay filters for the VSI */
6922 err = ice_replay_vsi(&pf->hw, vsi->idx);
6924 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
6925 err, vsi->idx, ice_vsi_type_str(type));
6929 /* Re-map HW VSI number, using VSI handle that has been
6930 * previously validated in ice_replay_vsi() call above
6932 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6934 /* enable the VSI */
6935 err = ice_ena_vsi(vsi, false);
6937 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6938 err, vsi->idx, ice_vsi_type_str(type));
6942 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6943 ice_vsi_type_str(type));
6950 * ice_update_pf_netdev_link - Update PF netdev link status
6951 * @pf: pointer to the PF instance
6953 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6958 ice_for_each_vsi(pf, i) {
6959 struct ice_vsi *vsi = pf->vsi[i];
6961 if (!vsi || vsi->type != ICE_VSI_PF)
6964 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6966 netif_carrier_on(pf->vsi[i]->netdev);
6967 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6969 netif_carrier_off(pf->vsi[i]->netdev);
6970 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6976 * ice_rebuild - rebuild after reset
6977 * @pf: PF to rebuild
6978 * @reset_type: type of reset
6980 * Do not rebuild VF VSI in this flow because that is already handled via
6981 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6982 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6983 * to reset/rebuild all the VF VSI twice.
6985 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6987 struct device *dev = ice_pf_to_dev(pf);
6988 struct ice_hw *hw = &pf->hw;
6992 if (test_bit(ICE_DOWN, pf->state))
6993 goto clear_recovery;
6995 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6997 #define ICE_EMP_RESET_SLEEP_MS 5000
6998 if (reset_type == ICE_RESET_EMPR) {
6999 /* If an EMP reset has occurred, any previously pending flash
7000 * update will have completed. We no longer know whether or
7001 * not the NVM update EMP reset is restricted.
7003 pf->fw_emp_reset_disabled = false;
7005 msleep(ICE_EMP_RESET_SLEEP_MS);
7008 err = ice_init_all_ctrlq(hw);
7010 dev_err(dev, "control queues init failed %d\n", err);
7011 goto err_init_ctrlq;
7014 /* if DDP was previously loaded successfully */
7015 if (!ice_is_safe_mode(pf)) {
7016 /* reload the SW DB of filter tables */
7017 if (reset_type == ICE_RESET_PFR)
7018 ice_fill_blk_tbls(hw);
7020 /* Reload DDP Package after CORER/GLOBR reset */
7021 ice_load_pkg(NULL, pf);
7024 err = ice_clear_pf_cfg(hw);
7026 dev_err(dev, "clear PF configuration failed %d\n", err);
7027 goto err_init_ctrlq;
7030 if (pf->first_sw->dflt_vsi_ena)
7031 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
7032 /* clear the default VSI configuration if it exists */
7033 pf->first_sw->dflt_vsi = NULL;
7034 pf->first_sw->dflt_vsi_ena = false;
7036 ice_clear_pxe_mode(hw);
7038 err = ice_init_nvm(hw);
7040 dev_err(dev, "ice_init_nvm failed %d\n", err);
7041 goto err_init_ctrlq;
7044 err = ice_get_caps(hw);
7046 dev_err(dev, "ice_get_caps failed %d\n", err);
7047 goto err_init_ctrlq;
7050 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
7052 dev_err(dev, "set_mac_cfg failed %d\n", err);
7053 goto err_init_ctrlq;
7056 dvm = ice_is_dvm_ena(hw);
7058 err = ice_aq_set_port_params(pf->hw.port_info, dvm, NULL);
7060 goto err_init_ctrlq;
7062 err = ice_sched_init_port(hw->port_info);
7064 goto err_sched_init_port;
7066 /* start misc vector */
7067 err = ice_req_irq_msix_misc(pf);
7069 dev_err(dev, "misc vector setup failed: %d\n", err);
7070 goto err_sched_init_port;
7073 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7074 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
7075 if (!rd32(hw, PFQF_FD_SIZE)) {
7076 u16 unused, guar, b_effort;
7078 guar = hw->func_caps.fd_fltr_guar;
7079 b_effort = hw->func_caps.fd_fltr_best_effort;
7081 /* force guaranteed filter pool for PF */
7082 ice_alloc_fd_guar_item(hw, &unused, guar);
7083 /* force shared filter pool for PF */
7084 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
7088 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
7089 ice_dcb_rebuild(pf);
7091 /* If the PF previously had enabled PTP, PTP init needs to happen before
7092 * the VSI rebuild. If not, this causes the PTP link status events to
7095 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7098 if (ice_is_feature_supported(pf, ICE_F_GNSS))
7101 /* rebuild PF VSI */
7102 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
7104 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
7105 goto err_vsi_rebuild;
7108 /* configure PTP timestamping after VSI rebuild */
7109 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
7110 ice_ptp_cfg_timestamp(pf, false);
7112 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
7114 dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
7115 goto err_vsi_rebuild;
7118 if (reset_type == ICE_RESET_PFR) {
7119 err = ice_rebuild_channels(pf);
7121 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
7123 goto err_vsi_rebuild;
7127 /* If Flow Director is active */
7128 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
7129 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
7131 dev_err(dev, "control VSI rebuild failed: %d\n", err);
7132 goto err_vsi_rebuild;
7135 /* replay HW Flow Director recipes */
7137 ice_fdir_replay_flows(hw);
7139 /* replay Flow Director filters */
7140 ice_fdir_replay_fltrs(pf);
7142 ice_rebuild_arfs(pf);
7145 ice_update_pf_netdev_link(pf);
7147 /* tell the firmware we are up */
7148 err = ice_send_version(pf);
7150 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
7152 goto err_vsi_rebuild;
7155 ice_replay_post(hw);
7157 /* if we get here, reset flow is successful */
7158 clear_bit(ICE_RESET_FAILED, pf->state);
7160 ice_plug_aux_dev(pf);
7164 err_sched_init_port:
7165 ice_sched_cleanup_all(hw);
7167 ice_shutdown_all_ctrlq(hw);
7168 set_bit(ICE_RESET_FAILED, pf->state);
7170 /* set this bit in PF state to control service task scheduling */
7171 set_bit(ICE_NEEDS_RESTART, pf->state);
7172 dev_err(dev, "Rebuild failed, unload and reload driver\n");
7176 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
7177 * @vsi: Pointer to VSI structure
7179 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
7181 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
7182 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
7184 return ICE_RXBUF_3072;
7188 * ice_change_mtu - NDO callback to change the MTU
7189 * @netdev: network interface device structure
7190 * @new_mtu: new value for maximum frame size
7192 * Returns 0 on success, negative on failure
7194 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
7196 struct ice_netdev_priv *np = netdev_priv(netdev);
7197 struct ice_vsi *vsi = np->vsi;
7198 struct ice_pf *pf = vsi->back;
7202 if (new_mtu == (int)netdev->mtu) {
7203 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
7207 if (ice_is_xdp_ena_vsi(vsi)) {
7208 int frame_size = ice_max_xdp_frame_size(vsi);
7210 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
7211 netdev_err(netdev, "max MTU for XDP usage is %d\n",
7212 frame_size - ICE_ETH_PKT_HDR_PAD);
7217 /* if a reset is in progress, wait for some time for it to complete */
7219 if (ice_is_reset_in_progress(pf->state)) {
7221 usleep_range(1000, 2000);
7226 } while (count < 100);
7229 netdev_err(netdev, "can't change MTU. Device is busy\n");
7233 netdev->mtu = (unsigned int)new_mtu;
7235 /* if VSI is up, bring it down and then back up */
7236 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
7237 err = ice_down(vsi);
7239 netdev_err(netdev, "change MTU if_down err %d\n", err);
7245 netdev_err(netdev, "change MTU if_up err %d\n", err);
7250 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
7251 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
7257 * ice_eth_ioctl - Access the hwtstamp interface
7258 * @netdev: network interface device structure
7259 * @ifr: interface request data
7260 * @cmd: ioctl command
7262 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
7264 struct ice_netdev_priv *np = netdev_priv(netdev);
7265 struct ice_pf *pf = np->vsi->back;
7269 return ice_ptp_get_ts_config(pf, ifr);
7271 return ice_ptp_set_ts_config(pf, ifr);
7278 * ice_aq_str - convert AQ err code to a string
7279 * @aq_err: the AQ error code to convert
7281 const char *ice_aq_str(enum ice_aq_err aq_err)
7286 case ICE_AQ_RC_EPERM:
7287 return "ICE_AQ_RC_EPERM";
7288 case ICE_AQ_RC_ENOENT:
7289 return "ICE_AQ_RC_ENOENT";
7290 case ICE_AQ_RC_ENOMEM:
7291 return "ICE_AQ_RC_ENOMEM";
7292 case ICE_AQ_RC_EBUSY:
7293 return "ICE_AQ_RC_EBUSY";
7294 case ICE_AQ_RC_EEXIST:
7295 return "ICE_AQ_RC_EEXIST";
7296 case ICE_AQ_RC_EINVAL:
7297 return "ICE_AQ_RC_EINVAL";
7298 case ICE_AQ_RC_ENOSPC:
7299 return "ICE_AQ_RC_ENOSPC";
7300 case ICE_AQ_RC_ENOSYS:
7301 return "ICE_AQ_RC_ENOSYS";
7302 case ICE_AQ_RC_EMODE:
7303 return "ICE_AQ_RC_EMODE";
7304 case ICE_AQ_RC_ENOSEC:
7305 return "ICE_AQ_RC_ENOSEC";
7306 case ICE_AQ_RC_EBADSIG:
7307 return "ICE_AQ_RC_EBADSIG";
7308 case ICE_AQ_RC_ESVN:
7309 return "ICE_AQ_RC_ESVN";
7310 case ICE_AQ_RC_EBADMAN:
7311 return "ICE_AQ_RC_EBADMAN";
7312 case ICE_AQ_RC_EBADBUF:
7313 return "ICE_AQ_RC_EBADBUF";
7316 return "ICE_AQ_RC_UNKNOWN";
7320 * ice_set_rss_lut - Set RSS LUT
7321 * @vsi: Pointer to VSI structure
7322 * @lut: Lookup table
7323 * @lut_size: Lookup table size
7325 * Returns 0 on success, negative on failure
7327 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7329 struct ice_aq_get_set_rss_lut_params params = {};
7330 struct ice_hw *hw = &vsi->back->hw;
7336 params.vsi_handle = vsi->idx;
7337 params.lut_size = lut_size;
7338 params.lut_type = vsi->rss_lut_type;
7341 status = ice_aq_set_rss_lut(hw, ¶ms);
7343 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
7344 status, ice_aq_str(hw->adminq.sq_last_status));
7350 * ice_set_rss_key - Set RSS key
7351 * @vsi: Pointer to the VSI structure
7352 * @seed: RSS hash seed
7354 * Returns 0 on success, negative on failure
7356 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7358 struct ice_hw *hw = &vsi->back->hw;
7364 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7366 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7367 status, ice_aq_str(hw->adminq.sq_last_status));
7373 * ice_get_rss_lut - Get RSS LUT
7374 * @vsi: Pointer to VSI structure
7375 * @lut: Buffer to store the lookup table entries
7376 * @lut_size: Size of buffer to store the lookup table entries
7378 * Returns 0 on success, negative on failure
7380 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7382 struct ice_aq_get_set_rss_lut_params params = {};
7383 struct ice_hw *hw = &vsi->back->hw;
7389 params.vsi_handle = vsi->idx;
7390 params.lut_size = lut_size;
7391 params.lut_type = vsi->rss_lut_type;
7394 status = ice_aq_get_rss_lut(hw, ¶ms);
7396 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7397 status, ice_aq_str(hw->adminq.sq_last_status));
7403 * ice_get_rss_key - Get RSS key
7404 * @vsi: Pointer to VSI structure
7405 * @seed: Buffer to store the key in
7407 * Returns 0 on success, negative on failure
7409 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7411 struct ice_hw *hw = &vsi->back->hw;
7417 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7419 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7420 status, ice_aq_str(hw->adminq.sq_last_status));
7426 * ice_bridge_getlink - Get the hardware bridge mode
7429 * @seq: RTNL message seq
7430 * @dev: the netdev being configured
7431 * @filter_mask: filter mask passed in
7432 * @nlflags: netlink flags passed in
7434 * Return the bridge mode (VEB/VEPA)
7437 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7438 struct net_device *dev, u32 filter_mask, int nlflags)
7440 struct ice_netdev_priv *np = netdev_priv(dev);
7441 struct ice_vsi *vsi = np->vsi;
7442 struct ice_pf *pf = vsi->back;
7445 bmode = pf->first_sw->bridge_mode;
7447 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7452 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7453 * @vsi: Pointer to VSI structure
7454 * @bmode: Hardware bridge mode (VEB/VEPA)
7456 * Returns 0 on success, negative on failure
7458 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7460 struct ice_aqc_vsi_props *vsi_props;
7461 struct ice_hw *hw = &vsi->back->hw;
7462 struct ice_vsi_ctx *ctxt;
7465 vsi_props = &vsi->info;
7467 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7471 ctxt->info = vsi->info;
7473 if (bmode == BRIDGE_MODE_VEB)
7474 /* change from VEPA to VEB mode */
7475 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7477 /* change from VEB to VEPA mode */
7478 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7479 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7481 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7483 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7484 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7487 /* Update sw flags for book keeping */
7488 vsi_props->sw_flags = ctxt->info.sw_flags;
7496 * ice_bridge_setlink - Set the hardware bridge mode
7497 * @dev: the netdev being configured
7498 * @nlh: RTNL message
7499 * @flags: bridge setlink flags
7500 * @extack: netlink extended ack
7502 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7503 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7504 * not already set for all VSIs connected to this switch. And also update the
7505 * unicast switch filter rules for the corresponding switch of the netdev.
7508 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7509 u16 __always_unused flags,
7510 struct netlink_ext_ack __always_unused *extack)
7512 struct ice_netdev_priv *np = netdev_priv(dev);
7513 struct ice_pf *pf = np->vsi->back;
7514 struct nlattr *attr, *br_spec;
7515 struct ice_hw *hw = &pf->hw;
7516 struct ice_sw *pf_sw;
7517 int rem, v, err = 0;
7519 pf_sw = pf->first_sw;
7520 /* find the attribute in the netlink message */
7521 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7523 nla_for_each_nested(attr, br_spec, rem) {
7526 if (nla_type(attr) != IFLA_BRIDGE_MODE)
7528 mode = nla_get_u16(attr);
7529 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7531 /* Continue if bridge mode is not being flipped */
7532 if (mode == pf_sw->bridge_mode)
7534 /* Iterates through the PF VSI list and update the loopback
7537 ice_for_each_vsi(pf, v) {
7540 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7545 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7546 /* Update the unicast switch filter rules for the corresponding
7547 * switch of the netdev
7549 err = ice_update_sw_rule_bridge_mode(hw);
7551 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7553 ice_aq_str(hw->adminq.sq_last_status));
7554 /* revert hw->evb_veb */
7555 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7559 pf_sw->bridge_mode = mode;
7566 * ice_tx_timeout - Respond to a Tx Hang
7567 * @netdev: network interface device structure
7568 * @txqueue: Tx queue
7570 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7572 struct ice_netdev_priv *np = netdev_priv(netdev);
7573 struct ice_tx_ring *tx_ring = NULL;
7574 struct ice_vsi *vsi = np->vsi;
7575 struct ice_pf *pf = vsi->back;
7578 pf->tx_timeout_count++;
7580 /* Check if PFC is enabled for the TC to which the queue belongs
7581 * to. If yes then Tx timeout is not caused by a hung queue, no
7582 * need to reset and rebuild
7584 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7585 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7590 /* now that we have an index, find the tx_ring struct */
7591 ice_for_each_txq(vsi, i)
7592 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7593 if (txqueue == vsi->tx_rings[i]->q_index) {
7594 tx_ring = vsi->tx_rings[i];
7598 /* Reset recovery level if enough time has elapsed after last timeout.
7599 * Also ensure no new reset action happens before next timeout period.
7601 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7602 pf->tx_timeout_recovery_level = 1;
7603 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7604 netdev->watchdog_timeo)))
7608 struct ice_hw *hw = &pf->hw;
7611 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7612 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7613 /* Read interrupt register */
7614 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7616 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7617 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7618 head, tx_ring->next_to_use, val);
7621 pf->tx_timeout_last_recovery = jiffies;
7622 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7623 pf->tx_timeout_recovery_level, txqueue);
7625 switch (pf->tx_timeout_recovery_level) {
7627 set_bit(ICE_PFR_REQ, pf->state);
7630 set_bit(ICE_CORER_REQ, pf->state);
7633 set_bit(ICE_GLOBR_REQ, pf->state);
7636 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7637 set_bit(ICE_DOWN, pf->state);
7638 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7639 set_bit(ICE_SERVICE_DIS, pf->state);
7643 ice_service_task_schedule(pf);
7644 pf->tx_timeout_recovery_level++;
7648 * ice_setup_tc_cls_flower - flower classifier offloads
7649 * @np: net device to configure
7650 * @filter_dev: device on which filter is added
7651 * @cls_flower: offload data
7654 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
7655 struct net_device *filter_dev,
7656 struct flow_cls_offload *cls_flower)
7658 struct ice_vsi *vsi = np->vsi;
7660 if (cls_flower->common.chain_index)
7663 switch (cls_flower->command) {
7664 case FLOW_CLS_REPLACE:
7665 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
7666 case FLOW_CLS_DESTROY:
7667 return ice_del_cls_flower(vsi, cls_flower);
7674 * ice_setup_tc_block_cb - callback handler registered for TC block
7675 * @type: TC SETUP type
7676 * @type_data: TC flower offload data that contains user input
7677 * @cb_priv: netdev private data
7680 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
7682 struct ice_netdev_priv *np = cb_priv;
7685 case TC_SETUP_CLSFLOWER:
7686 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
7694 * ice_validate_mqprio_qopt - Validate TCF input parameters
7695 * @vsi: Pointer to VSI
7696 * @mqprio_qopt: input parameters for mqprio queue configuration
7698 * This function validates MQPRIO params, such as qcount (power of 2 wherever
7699 * needed), and make sure user doesn't specify qcount and BW rate limit
7700 * for TCs, which are more than "num_tc"
7703 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
7704 struct tc_mqprio_qopt_offload *mqprio_qopt)
7706 u64 sum_max_rate = 0, sum_min_rate = 0;
7707 int non_power_of_2_qcount = 0;
7708 struct ice_pf *pf = vsi->back;
7709 int max_rss_q_cnt = 0;
7714 if (vsi->type != ICE_VSI_PF)
7717 if (mqprio_qopt->qopt.offset[0] != 0 ||
7718 mqprio_qopt->qopt.num_tc < 1 ||
7719 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
7722 dev = ice_pf_to_dev(pf);
7723 vsi->ch_rss_size = 0;
7724 num_tc = mqprio_qopt->qopt.num_tc;
7726 for (i = 0; num_tc; i++) {
7727 int qcount = mqprio_qopt->qopt.count[i];
7728 u64 max_rate, min_rate, rem;
7733 if (is_power_of_2(qcount)) {
7734 if (non_power_of_2_qcount &&
7735 qcount > non_power_of_2_qcount) {
7736 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
7737 qcount, non_power_of_2_qcount);
7740 if (qcount > max_rss_q_cnt)
7741 max_rss_q_cnt = qcount;
7743 if (non_power_of_2_qcount &&
7744 qcount != non_power_of_2_qcount) {
7745 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
7746 qcount, non_power_of_2_qcount);
7749 if (qcount < max_rss_q_cnt) {
7750 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
7751 qcount, max_rss_q_cnt);
7754 max_rss_q_cnt = qcount;
7755 non_power_of_2_qcount = qcount;
7758 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
7759 * converts the bandwidth rate limit into Bytes/s when
7760 * passing it down to the driver. So convert input bandwidth
7761 * from Bytes/s to Kbps
7763 max_rate = mqprio_qopt->max_rate[i];
7764 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
7765 sum_max_rate += max_rate;
7767 /* min_rate is minimum guaranteed rate and it can't be zero */
7768 min_rate = mqprio_qopt->min_rate[i];
7769 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
7770 sum_min_rate += min_rate;
7772 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
7773 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
7774 min_rate, ICE_MIN_BW_LIMIT);
7778 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
7780 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
7781 i, ICE_MIN_BW_LIMIT);
7785 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
7787 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
7788 i, ICE_MIN_BW_LIMIT);
7792 /* min_rate can't be more than max_rate, except when max_rate
7793 * is zero (implies max_rate sought is max line rate). In such
7794 * a case min_rate can be more than max.
7796 if (max_rate && min_rate > max_rate) {
7797 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
7798 min_rate, max_rate);
7802 if (i >= mqprio_qopt->qopt.num_tc - 1)
7804 if (mqprio_qopt->qopt.offset[i + 1] !=
7805 (mqprio_qopt->qopt.offset[i] + qcount))
7809 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7812 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7815 speed = ice_get_link_speed_kbps(vsi);
7816 if (sum_max_rate && sum_max_rate > (u64)speed) {
7817 dev_err(dev, "Invalid max Tx rate(%llu) Kbps > speed(%u) Kbps specified\n",
7818 sum_max_rate, speed);
7821 if (sum_min_rate && sum_min_rate > (u64)speed) {
7822 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
7823 sum_min_rate, speed);
7827 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
7828 vsi->ch_rss_size = max_rss_q_cnt;
7834 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
7835 * @pf: ptr to PF device
7838 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
7840 struct device *dev = ice_pf_to_dev(pf);
7845 if (!(vsi->num_gfltr || vsi->num_bfltr))
7849 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
7850 struct ice_fd_hw_prof *prof;
7854 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
7855 hw->fdir_prof[flow]->cnt))
7858 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
7859 enum ice_flow_priority prio;
7862 /* add this VSI to FDir profile for this flow */
7863 prio = ICE_FLOW_PRIO_NORMAL;
7864 prof = hw->fdir_prof[flow];
7865 prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
7866 status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
7867 prof->vsi_h[0], vsi->idx,
7868 prio, prof->fdir_seg[tun],
7871 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
7876 prof->entry_h[prof->cnt][tun] = entry_h;
7879 /* store VSI for filter replay and delete */
7880 prof->vsi_h[prof->cnt] = vsi->idx;
7884 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
7889 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
7895 * ice_add_channel - add a channel by adding VSI
7896 * @pf: ptr to PF device
7897 * @sw_id: underlying HW switching element ID
7898 * @ch: ptr to channel structure
7900 * Add a channel (VSI) using add_vsi and queue_map
7902 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
7904 struct device *dev = ice_pf_to_dev(pf);
7905 struct ice_vsi *vsi;
7907 if (ch->type != ICE_VSI_CHNL) {
7908 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
7912 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
7913 if (!vsi || vsi->type != ICE_VSI_CHNL) {
7914 dev_err(dev, "create chnl VSI failure\n");
7918 ice_add_vsi_to_fdir(pf, vsi);
7921 ch->vsi_num = vsi->vsi_num;
7922 ch->info.mapping_flags = vsi->info.mapping_flags;
7924 /* set the back pointer of channel for newly created VSI */
7927 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
7928 sizeof(vsi->info.q_mapping));
7929 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
7930 sizeof(vsi->info.tc_mapping));
7937 * @vsi: the VSI being setup
7938 * @ch: ptr to channel structure
7940 * Configure channel specific resources such as rings, vector.
7942 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
7946 for (i = 0; i < ch->num_txq; i++) {
7947 struct ice_q_vector *tx_q_vector, *rx_q_vector;
7948 struct ice_ring_container *rc;
7949 struct ice_tx_ring *tx_ring;
7950 struct ice_rx_ring *rx_ring;
7952 tx_ring = vsi->tx_rings[ch->base_q + i];
7953 rx_ring = vsi->rx_rings[ch->base_q + i];
7954 if (!tx_ring || !rx_ring)
7957 /* setup ring being channel enabled */
7961 /* following code block sets up vector specific attributes */
7962 tx_q_vector = tx_ring->q_vector;
7963 rx_q_vector = rx_ring->q_vector;
7964 if (!tx_q_vector && !rx_q_vector)
7968 tx_q_vector->ch = ch;
7969 /* setup Tx and Rx ITR setting if DIM is off */
7970 rc = &tx_q_vector->tx;
7971 if (!ITR_IS_DYNAMIC(rc))
7972 ice_write_itr(rc, rc->itr_setting);
7975 rx_q_vector->ch = ch;
7976 /* setup Tx and Rx ITR setting if DIM is off */
7977 rc = &rx_q_vector->rx;
7978 if (!ITR_IS_DYNAMIC(rc))
7979 ice_write_itr(rc, rc->itr_setting);
7983 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
7984 * GLINT_ITR register would have written to perform in-context
7985 * update, hence perform flush
7987 if (ch->num_txq || ch->num_rxq)
7988 ice_flush(&vsi->back->hw);
7992 * ice_cfg_chnl_all_res - configure channel resources
7993 * @vsi: pte to main_vsi
7994 * @ch: ptr to channel structure
7996 * This function configures channel specific resources such as flow-director
7997 * counter index, and other resources such as queues, vectors, ITR settings
8000 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
8002 /* configure channel (aka ADQ) resources such as queues, vectors,
8003 * ITR settings for channel specific vectors and anything else
8005 ice_chnl_cfg_res(vsi, ch);
8009 * ice_setup_hw_channel - setup new channel
8010 * @pf: ptr to PF device
8011 * @vsi: the VSI being setup
8012 * @ch: ptr to channel structure
8013 * @sw_id: underlying HW switching element ID
8014 * @type: type of channel to be created (VMDq2/VF)
8016 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8017 * and configures Tx rings accordingly
8020 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8021 struct ice_channel *ch, u16 sw_id, u8 type)
8023 struct device *dev = ice_pf_to_dev(pf);
8026 ch->base_q = vsi->next_base_q;
8029 ret = ice_add_channel(pf, sw_id, ch);
8031 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
8035 /* configure/setup ADQ specific resources */
8036 ice_cfg_chnl_all_res(vsi, ch);
8038 /* make sure to update the next_base_q so that subsequent channel's
8039 * (aka ADQ) VSI queue map is correct
8041 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
8042 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
8049 * ice_setup_channel - setup new channel using uplink element
8050 * @pf: ptr to PF device
8051 * @vsi: the VSI being setup
8052 * @ch: ptr to channel structure
8054 * Setup new channel (VSI) based on specified type (VMDq2/VF)
8055 * and uplink switching element
8058 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
8059 struct ice_channel *ch)
8061 struct device *dev = ice_pf_to_dev(pf);
8065 if (vsi->type != ICE_VSI_PF) {
8066 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
8070 sw_id = pf->first_sw->sw_id;
8072 /* create channel (VSI) */
8073 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
8075 dev_err(dev, "failed to setup hw_channel\n");
8078 dev_dbg(dev, "successfully created channel()\n");
8080 return ch->ch_vsi ? true : false;
8084 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
8085 * @vsi: VSI to be configured
8086 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
8087 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
8090 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
8094 err = ice_set_min_bw_limit(vsi, min_tx_rate);
8098 return ice_set_max_bw_limit(vsi, max_tx_rate);
8102 * ice_create_q_channel - function to create channel
8103 * @vsi: VSI to be configured
8104 * @ch: ptr to channel (it contains channel specific params)
8106 * This function creates channel (VSI) using num_queues specified by user,
8107 * reconfigs RSS if needed.
8109 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
8111 struct ice_pf *pf = vsi->back;
8117 dev = ice_pf_to_dev(pf);
8118 if (!ch->num_txq || !ch->num_rxq) {
8119 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
8123 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
8124 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
8125 vsi->cnt_q_avail, ch->num_txq);
8129 if (!ice_setup_channel(pf, vsi, ch)) {
8130 dev_info(dev, "Failed to setup channel\n");
8133 /* configure BW rate limit */
8134 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
8137 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
8140 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
8141 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8143 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
8144 ch->max_tx_rate, ch->ch_vsi->vsi_num);
8147 vsi->cnt_q_avail -= ch->num_txq;
8153 * ice_rem_all_chnl_fltrs - removes all channel filters
8154 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
8156 * Remove all advanced switch filters only if they are channel specific
8157 * tc-flower based filter
8159 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
8161 struct ice_tc_flower_fltr *fltr;
8162 struct hlist_node *node;
8164 /* to remove all channel filters, iterate an ordered list of filters */
8165 hlist_for_each_entry_safe(fltr, node,
8166 &pf->tc_flower_fltr_list,
8168 struct ice_rule_query_data rule;
8171 /* for now process only channel specific filters */
8172 if (!ice_is_chnl_fltr(fltr))
8175 rule.rid = fltr->rid;
8176 rule.rule_id = fltr->rule_id;
8177 rule.vsi_handle = fltr->dest_id;
8178 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
8180 if (status == -ENOENT)
8181 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
8184 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
8186 } else if (fltr->dest_vsi) {
8187 /* update advanced switch filter count */
8188 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
8189 u32 flags = fltr->flags;
8191 fltr->dest_vsi->num_chnl_fltr--;
8192 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
8193 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
8194 pf->num_dmac_chnl_fltrs--;
8198 hlist_del(&fltr->tc_flower_node);
8204 * ice_remove_q_channels - Remove queue channels for the TCs
8205 * @vsi: VSI to be configured
8206 * @rem_fltr: delete advanced switch filter or not
8208 * Remove queue channels for the TCs
8210 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
8212 struct ice_channel *ch, *ch_tmp;
8213 struct ice_pf *pf = vsi->back;
8216 /* remove all tc-flower based filter if they are channel filters only */
8218 ice_rem_all_chnl_fltrs(pf);
8220 /* remove ntuple filters since queue configuration is being changed */
8221 if (vsi->netdev->features & NETIF_F_NTUPLE) {
8222 struct ice_hw *hw = &pf->hw;
8224 mutex_lock(&hw->fdir_fltr_lock);
8225 ice_fdir_del_all_fltrs(vsi);
8226 mutex_unlock(&hw->fdir_fltr_lock);
8229 /* perform cleanup for channels if they exist */
8230 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
8231 struct ice_vsi *ch_vsi;
8233 list_del(&ch->list);
8234 ch_vsi = ch->ch_vsi;
8240 /* Reset queue contexts */
8241 for (i = 0; i < ch->num_rxq; i++) {
8242 struct ice_tx_ring *tx_ring;
8243 struct ice_rx_ring *rx_ring;
8245 tx_ring = vsi->tx_rings[ch->base_q + i];
8246 rx_ring = vsi->rx_rings[ch->base_q + i];
8249 if (tx_ring->q_vector)
8250 tx_ring->q_vector->ch = NULL;
8254 if (rx_ring->q_vector)
8255 rx_ring->q_vector->ch = NULL;
8259 /* Release FD resources for the channel VSI */
8260 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
8262 /* clear the VSI from scheduler tree */
8263 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
8265 /* Delete VSI from FW */
8266 ice_vsi_delete(ch->ch_vsi);
8268 /* Delete VSI from PF and HW VSI arrays */
8269 ice_vsi_clear(ch->ch_vsi);
8271 /* free the channel */
8275 /* clear the channel VSI map which is stored in main VSI */
8276 ice_for_each_chnl_tc(i)
8277 vsi->tc_map_vsi[i] = NULL;
8279 /* reset main VSI's all TC information */
8285 * ice_rebuild_channels - rebuild channel
8288 * Recreate channel VSIs and replay filters
8290 static int ice_rebuild_channels(struct ice_pf *pf)
8292 struct device *dev = ice_pf_to_dev(pf);
8293 struct ice_vsi *main_vsi;
8294 bool rem_adv_fltr = true;
8295 struct ice_channel *ch;
8296 struct ice_vsi *vsi;
8300 main_vsi = ice_get_main_vsi(pf);
8304 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
8305 main_vsi->old_numtc == 1)
8306 return 0; /* nothing to be done */
8308 /* reconfigure main VSI based on old value of TC and cached values
8311 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
8313 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
8314 main_vsi->old_ena_tc, main_vsi->vsi_num);
8318 /* rebuild ADQ VSIs */
8319 ice_for_each_vsi(pf, i) {
8320 enum ice_vsi_type type;
8323 if (!vsi || vsi->type != ICE_VSI_CHNL)
8328 /* rebuild ADQ VSI */
8329 err = ice_vsi_rebuild(vsi, true);
8331 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
8332 ice_vsi_type_str(type), vsi->idx, err);
8336 /* Re-map HW VSI number, using VSI handle that has been
8337 * previously validated in ice_replay_vsi() call above
8339 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
8341 /* replay filters for the VSI */
8342 err = ice_replay_vsi(&pf->hw, vsi->idx);
8344 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
8345 ice_vsi_type_str(type), err, vsi->idx);
8346 rem_adv_fltr = false;
8349 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
8350 ice_vsi_type_str(type), vsi->idx);
8352 /* store ADQ VSI at correct TC index in main VSI's
8355 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8358 /* ADQ VSI(s) has been rebuilt successfully, so setup
8359 * channel for main VSI's Tx and Rx rings
8361 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8362 struct ice_vsi *ch_vsi;
8364 ch_vsi = ch->ch_vsi;
8368 /* reconfig channel resources */
8369 ice_cfg_chnl_all_res(main_vsi, ch);
8371 /* replay BW rate limit if it is non-zero */
8372 if (!ch->max_tx_rate && !ch->min_tx_rate)
8375 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8378 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",
8379 err, ch->max_tx_rate, ch->min_tx_rate,
8382 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8383 ch->max_tx_rate, ch->min_tx_rate,
8387 /* reconfig RSS for main VSI */
8388 if (main_vsi->ch_rss_size)
8389 ice_vsi_cfg_rss_lut_key(main_vsi);
8394 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8399 * ice_create_q_channels - Add queue channel for the given TCs
8400 * @vsi: VSI to be configured
8402 * Configures queue channel mapping to the given TCs
8404 static int ice_create_q_channels(struct ice_vsi *vsi)
8406 struct ice_pf *pf = vsi->back;
8407 struct ice_channel *ch;
8410 ice_for_each_chnl_tc(i) {
8411 if (!(vsi->all_enatc & BIT(i)))
8414 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8419 INIT_LIST_HEAD(&ch->list);
8420 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8421 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8422 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8423 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8424 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8426 /* convert to Kbits/s */
8427 if (ch->max_tx_rate)
8428 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8429 ICE_BW_KBPS_DIVISOR);
8430 if (ch->min_tx_rate)
8431 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8432 ICE_BW_KBPS_DIVISOR);
8434 ret = ice_create_q_channel(vsi, ch);
8436 dev_err(ice_pf_to_dev(pf),
8437 "failed creating channel TC:%d\n", i);
8441 list_add_tail(&ch->list, &vsi->ch_list);
8442 vsi->tc_map_vsi[i] = ch->ch_vsi;
8443 dev_dbg(ice_pf_to_dev(pf),
8444 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8449 ice_remove_q_channels(vsi, false);
8455 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8456 * @netdev: net device to configure
8457 * @type_data: TC offload data
8459 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8461 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8462 struct ice_netdev_priv *np = netdev_priv(netdev);
8463 struct ice_vsi *vsi = np->vsi;
8464 struct ice_pf *pf = vsi->back;
8465 u16 mode, ena_tc_qdisc = 0;
8466 int cur_txq, cur_rxq;
8471 dev = ice_pf_to_dev(pf);
8472 num_tcf = mqprio_qopt->qopt.num_tc;
8473 hw = mqprio_qopt->qopt.hw;
8474 mode = mqprio_qopt->mode;
8476 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8477 vsi->ch_rss_size = 0;
8478 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8482 /* Generate queue region map for number of TCF requested */
8483 for (i = 0; i < num_tcf; i++)
8484 ena_tc_qdisc |= BIT(i);
8487 case TC_MQPRIO_MODE_CHANNEL:
8489 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8491 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8495 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8496 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8497 /* don't assume state of hw_tc_offload during driver load
8498 * and set the flag for TC flower filter if hw_tc_offload
8501 if (vsi->netdev->features & NETIF_F_HW_TC)
8502 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8510 /* Requesting same TCF configuration as already enabled */
8511 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8512 mode != TC_MQPRIO_MODE_CHANNEL)
8515 /* Pause VSI queues */
8516 ice_dis_vsi(vsi, true);
8518 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8519 ice_remove_q_channels(vsi, true);
8521 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8522 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8524 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8527 /* logic to rebuild VSI, same like ethtool -L */
8528 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8530 for (i = 0; i < num_tcf; i++) {
8531 if (!(ena_tc_qdisc & BIT(i)))
8534 offset = vsi->mqprio_qopt.qopt.offset[i];
8535 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8536 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8538 vsi->req_txq = offset + qcount_tx;
8539 vsi->req_rxq = offset + qcount_rx;
8541 /* store away original rss_size info, so that it gets reused
8542 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8543 * determine, what should be the rss_sizefor main VSI
8545 vsi->orig_rss_size = vsi->rss_size;
8548 /* save current values of Tx and Rx queues before calling VSI rebuild
8549 * for fallback option
8551 cur_txq = vsi->num_txq;
8552 cur_rxq = vsi->num_rxq;
8554 /* proceed with rebuild main VSI using correct number of queues */
8555 ret = ice_vsi_rebuild(vsi, false);
8557 /* fallback to current number of queues */
8558 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8559 vsi->req_txq = cur_txq;
8560 vsi->req_rxq = cur_rxq;
8561 clear_bit(ICE_RESET_FAILED, pf->state);
8562 if (ice_vsi_rebuild(vsi, false)) {
8563 dev_err(dev, "Rebuild of main VSI failed again\n");
8568 vsi->all_numtc = num_tcf;
8569 vsi->all_enatc = ena_tc_qdisc;
8570 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8572 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8577 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8578 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8579 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8581 /* set TC0 rate limit if specified */
8582 if (max_tx_rate || min_tx_rate) {
8583 /* convert to Kbits/s */
8585 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8587 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8589 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8591 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8592 max_tx_rate, min_tx_rate, vsi->vsi_num);
8594 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8595 max_tx_rate, min_tx_rate, vsi->vsi_num);
8599 ret = ice_create_q_channels(vsi);
8601 netdev_err(netdev, "failed configuring queue channels\n");
8604 netdev_dbg(netdev, "successfully configured channels\n");
8608 if (vsi->ch_rss_size)
8609 ice_vsi_cfg_rss_lut_key(vsi);
8612 /* if error, reset the all_numtc and all_enatc */
8618 ice_ena_vsi(vsi, true);
8623 static LIST_HEAD(ice_block_cb_list);
8626 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8629 struct ice_netdev_priv *np = netdev_priv(netdev);
8630 struct ice_pf *pf = np->vsi->back;
8634 case TC_SETUP_BLOCK:
8635 return flow_block_cb_setup_simple(type_data,
8637 ice_setup_tc_block_cb,
8639 case TC_SETUP_QDISC_MQPRIO:
8640 /* setup traffic classifier for receive side */
8641 mutex_lock(&pf->tc_mutex);
8642 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
8643 mutex_unlock(&pf->tc_mutex);
8651 static struct ice_indr_block_priv *
8652 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
8653 struct net_device *netdev)
8655 struct ice_indr_block_priv *cb_priv;
8657 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
8658 if (!cb_priv->netdev)
8660 if (cb_priv->netdev == netdev)
8667 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
8670 struct ice_indr_block_priv *priv = indr_priv;
8671 struct ice_netdev_priv *np = priv->np;
8674 case TC_SETUP_CLSFLOWER:
8675 return ice_setup_tc_cls_flower(np, priv->netdev,
8676 (struct flow_cls_offload *)
8684 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
8685 struct ice_netdev_priv *np,
8686 struct flow_block_offload *f, void *data,
8687 void (*cleanup)(struct flow_block_cb *block_cb))
8689 struct ice_indr_block_priv *indr_priv;
8690 struct flow_block_cb *block_cb;
8692 if (!ice_is_tunnel_supported(netdev) &&
8693 !(is_vlan_dev(netdev) &&
8694 vlan_dev_real_dev(netdev) == np->vsi->netdev))
8697 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
8700 switch (f->command) {
8701 case FLOW_BLOCK_BIND:
8702 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8706 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
8710 indr_priv->netdev = netdev;
8712 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
8715 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
8716 indr_priv, indr_priv,
8717 ice_rep_indr_tc_block_unbind,
8718 f, netdev, sch, data, np,
8721 if (IS_ERR(block_cb)) {
8722 list_del(&indr_priv->list);
8724 return PTR_ERR(block_cb);
8726 flow_block_cb_add(block_cb, f);
8727 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
8729 case FLOW_BLOCK_UNBIND:
8730 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8734 block_cb = flow_block_cb_lookup(f->block,
8735 ice_indr_setup_block_cb,
8740 flow_indr_block_cb_remove(block_cb, f);
8742 list_del(&block_cb->driver_list);
8751 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
8752 void *cb_priv, enum tc_setup_type type, void *type_data,
8754 void (*cleanup)(struct flow_block_cb *block_cb))
8757 case TC_SETUP_BLOCK:
8758 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
8767 * ice_open - Called when a network interface becomes active
8768 * @netdev: network interface device structure
8770 * The open entry point is called when a network interface is made
8771 * active by the system (IFF_UP). At this point all resources needed
8772 * for transmit and receive operations are allocated, the interrupt
8773 * handler is registered with the OS, the netdev watchdog is enabled,
8774 * and the stack is notified that the interface is ready.
8776 * Returns 0 on success, negative value on failure
8778 int ice_open(struct net_device *netdev)
8780 struct ice_netdev_priv *np = netdev_priv(netdev);
8781 struct ice_pf *pf = np->vsi->back;
8783 if (ice_is_reset_in_progress(pf->state)) {
8784 netdev_err(netdev, "can't open net device while reset is in progress");
8788 return ice_open_internal(netdev);
8792 * ice_open_internal - Called when a network interface becomes active
8793 * @netdev: network interface device structure
8795 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
8798 * Returns 0 on success, negative value on failure
8800 int ice_open_internal(struct net_device *netdev)
8802 struct ice_netdev_priv *np = netdev_priv(netdev);
8803 struct ice_vsi *vsi = np->vsi;
8804 struct ice_pf *pf = vsi->back;
8805 struct ice_port_info *pi;
8808 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
8809 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
8813 netif_carrier_off(netdev);
8815 pi = vsi->port_info;
8816 err = ice_update_link_info(pi);
8818 netdev_err(netdev, "Failed to get link info, error %d\n", err);
8822 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
8824 /* Set PHY if there is media, otherwise, turn off PHY */
8825 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
8826 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8827 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
8828 err = ice_init_phy_user_cfg(pi);
8830 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
8836 err = ice_configure_phy(vsi);
8838 netdev_err(netdev, "Failed to set physical link up, error %d\n",
8843 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8844 ice_set_link(vsi, false);
8847 err = ice_vsi_open(vsi);
8849 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
8850 vsi->vsi_num, vsi->vsw->sw_id);
8852 /* Update existing tunnels information */
8853 udp_tunnel_get_rx_info(netdev);
8859 * ice_stop - Disables a network interface
8860 * @netdev: network interface device structure
8862 * The stop entry point is called when an interface is de-activated by the OS,
8863 * and the netdevice enters the DOWN state. The hardware is still under the
8864 * driver's control, but the netdev interface is disabled.
8866 * Returns success only - not allowed to fail
8868 int ice_stop(struct net_device *netdev)
8870 struct ice_netdev_priv *np = netdev_priv(netdev);
8871 struct ice_vsi *vsi = np->vsi;
8872 struct ice_pf *pf = vsi->back;
8874 if (ice_is_reset_in_progress(pf->state)) {
8875 netdev_err(netdev, "can't stop net device while reset is in progress");
8885 * ice_features_check - Validate encapsulated packet conforms to limits
8887 * @netdev: This port's netdev
8888 * @features: Offload features that the stack believes apply
8890 static netdev_features_t
8891 ice_features_check(struct sk_buff *skb,
8892 struct net_device __always_unused *netdev,
8893 netdev_features_t features)
8895 bool gso = skb_is_gso(skb);
8898 /* No point in doing any of this if neither checksum nor GSO are
8899 * being requested for this frame. We can rule out both by just
8900 * checking for CHECKSUM_PARTIAL
8902 if (skb->ip_summed != CHECKSUM_PARTIAL)
8905 /* We cannot support GSO if the MSS is going to be less than
8906 * 64 bytes. If it is then we need to drop support for GSO.
8908 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
8909 features &= ~NETIF_F_GSO_MASK;
8911 len = skb_network_offset(skb);
8912 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
8913 goto out_rm_features;
8915 len = skb_network_header_len(skb);
8916 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8917 goto out_rm_features;
8919 if (skb->encapsulation) {
8920 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
8921 * the case of IPIP frames, the transport header pointer is
8922 * after the inner header! So check to make sure that this
8923 * is a GRE or UDP_TUNNEL frame before doing that math.
8925 if (gso && (skb_shinfo(skb)->gso_type &
8926 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
8927 len = skb_inner_network_header(skb) -
8928 skb_transport_header(skb);
8929 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
8930 goto out_rm_features;
8933 len = skb_inner_network_header_len(skb);
8934 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8935 goto out_rm_features;
8940 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8943 static const struct net_device_ops ice_netdev_safe_mode_ops = {
8944 .ndo_open = ice_open,
8945 .ndo_stop = ice_stop,
8946 .ndo_start_xmit = ice_start_xmit,
8947 .ndo_set_mac_address = ice_set_mac_address,
8948 .ndo_validate_addr = eth_validate_addr,
8949 .ndo_change_mtu = ice_change_mtu,
8950 .ndo_get_stats64 = ice_get_stats64,
8951 .ndo_tx_timeout = ice_tx_timeout,
8952 .ndo_bpf = ice_xdp_safe_mode,
8955 static const struct net_device_ops ice_netdev_ops = {
8956 .ndo_open = ice_open,
8957 .ndo_stop = ice_stop,
8958 .ndo_start_xmit = ice_start_xmit,
8959 .ndo_select_queue = ice_select_queue,
8960 .ndo_features_check = ice_features_check,
8961 .ndo_fix_features = ice_fix_features,
8962 .ndo_set_rx_mode = ice_set_rx_mode,
8963 .ndo_set_mac_address = ice_set_mac_address,
8964 .ndo_validate_addr = eth_validate_addr,
8965 .ndo_change_mtu = ice_change_mtu,
8966 .ndo_get_stats64 = ice_get_stats64,
8967 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
8968 .ndo_eth_ioctl = ice_eth_ioctl,
8969 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
8970 .ndo_set_vf_mac = ice_set_vf_mac,
8971 .ndo_get_vf_config = ice_get_vf_cfg,
8972 .ndo_set_vf_trust = ice_set_vf_trust,
8973 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
8974 .ndo_set_vf_link_state = ice_set_vf_link_state,
8975 .ndo_get_vf_stats = ice_get_vf_stats,
8976 .ndo_set_vf_rate = ice_set_vf_bw,
8977 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
8978 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
8979 .ndo_setup_tc = ice_setup_tc,
8980 .ndo_set_features = ice_set_features,
8981 .ndo_bridge_getlink = ice_bridge_getlink,
8982 .ndo_bridge_setlink = ice_bridge_setlink,
8983 .ndo_fdb_add = ice_fdb_add,
8984 .ndo_fdb_del = ice_fdb_del,
8985 #ifdef CONFIG_RFS_ACCEL
8986 .ndo_rx_flow_steer = ice_rx_flow_steer,
8988 .ndo_tx_timeout = ice_tx_timeout,
8990 .ndo_xdp_xmit = ice_xdp_xmit,
8991 .ndo_xsk_wakeup = ice_xsk_wakeup,
8992 .ndo_get_devlink_port = ice_get_devlink_port,