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"
25 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
26 static const char ice_driver_string[] = DRV_SUMMARY;
27 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
29 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
30 #define ICE_DDP_PKG_PATH "intel/ice/ddp/"
31 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
33 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
34 MODULE_DESCRIPTION(DRV_SUMMARY);
35 MODULE_LICENSE("GPL v2");
36 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
38 static int debug = -1;
39 module_param(debug, int, 0644);
40 #ifndef CONFIG_DYNAMIC_DEBUG
41 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
43 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
44 #endif /* !CONFIG_DYNAMIC_DEBUG */
46 static DEFINE_IDA(ice_aux_ida);
47 DEFINE_STATIC_KEY_FALSE(ice_xdp_locking_key);
48 EXPORT_SYMBOL(ice_xdp_locking_key);
50 static struct workqueue_struct *ice_wq;
51 static const struct net_device_ops ice_netdev_safe_mode_ops;
52 static const struct net_device_ops ice_netdev_ops;
54 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
56 static void ice_vsi_release_all(struct ice_pf *pf);
58 static int ice_rebuild_channels(struct ice_pf *pf);
59 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_adv_fltr);
62 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
63 void *cb_priv, enum tc_setup_type type, void *type_data,
65 void (*cleanup)(struct flow_block_cb *block_cb));
67 bool netif_is_ice(struct net_device *dev)
69 return dev && (dev->netdev_ops == &ice_netdev_ops);
73 * ice_get_tx_pending - returns number of Tx descriptors not processed
74 * @ring: the ring of descriptors
76 static u16 ice_get_tx_pending(struct ice_tx_ring *ring)
80 head = ring->next_to_clean;
81 tail = ring->next_to_use;
84 return (head < tail) ?
85 tail - head : (tail + ring->count - head);
90 * ice_check_for_hang_subtask - check for and recover hung queues
91 * @pf: pointer to PF struct
93 static void ice_check_for_hang_subtask(struct ice_pf *pf)
95 struct ice_vsi *vsi = NULL;
101 ice_for_each_vsi(pf, v)
102 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
107 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
110 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
115 ice_for_each_txq(vsi, i) {
116 struct ice_tx_ring *tx_ring = vsi->tx_rings[i];
120 if (ice_ring_ch_enabled(tx_ring))
124 /* If packet counter has not changed the queue is
125 * likely stalled, so force an interrupt for this
128 * prev_pkt would be negative if there was no
131 packets = tx_ring->stats.pkts & INT_MAX;
132 if (tx_ring->tx_stats.prev_pkt == packets) {
133 /* Trigger sw interrupt to revive the queue */
134 ice_trigger_sw_intr(hw, tx_ring->q_vector);
138 /* Memory barrier between read of packet count and call
139 * to ice_get_tx_pending()
142 tx_ring->tx_stats.prev_pkt =
143 ice_get_tx_pending(tx_ring) ? packets : -1;
149 * ice_init_mac_fltr - Set initial MAC filters
150 * @pf: board private structure
152 * Set initial set of MAC filters for PF VSI; configure filters for permanent
153 * address and broadcast address. If an error is encountered, netdevice will be
156 static int ice_init_mac_fltr(struct ice_pf *pf)
161 vsi = ice_get_main_vsi(pf);
165 perm_addr = vsi->port_info->mac.perm_addr;
166 return ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
170 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
171 * @netdev: the net device on which the sync is happening
172 * @addr: MAC address to sync
174 * This is a callback function which is called by the in kernel device sync
175 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
176 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
177 * MAC filters from the hardware.
179 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
181 struct ice_netdev_priv *np = netdev_priv(netdev);
182 struct ice_vsi *vsi = np->vsi;
184 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
192 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
193 * @netdev: the net device on which the unsync is happening
194 * @addr: MAC address to unsync
196 * This is a callback function which is called by the in kernel device unsync
197 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
198 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
199 * delete the MAC filters from the hardware.
201 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
203 struct ice_netdev_priv *np = netdev_priv(netdev);
204 struct ice_vsi *vsi = np->vsi;
206 /* Under some circumstances, we might receive a request to delete our
207 * own device address from our uc list. Because we store the device
208 * address in the VSI's MAC filter list, we need to ignore such
209 * requests and not delete our device address from this list.
211 if (ether_addr_equal(addr, netdev->dev_addr))
214 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
222 * ice_vsi_fltr_changed - check if filter state changed
223 * @vsi: VSI to be checked
225 * returns true if filter state has changed, false otherwise.
227 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
229 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
230 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
231 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
235 * ice_set_promisc - Enable promiscuous mode for a given PF
236 * @vsi: the VSI being configured
237 * @promisc_m: mask of promiscuous config bits
240 static int ice_set_promisc(struct ice_vsi *vsi, u8 promisc_m)
244 if (vsi->type != ICE_VSI_PF)
247 if (vsi->num_vlan > 1)
248 status = ice_fltr_set_vlan_vsi_promisc(&vsi->back->hw, vsi, promisc_m);
250 status = ice_fltr_set_vsi_promisc(&vsi->back->hw, vsi->idx, promisc_m, 0);
255 * ice_clear_promisc - Disable promiscuous mode for a given PF
256 * @vsi: the VSI being configured
257 * @promisc_m: mask of promiscuous config bits
260 static int ice_clear_promisc(struct ice_vsi *vsi, u8 promisc_m)
264 if (vsi->type != ICE_VSI_PF)
267 if (vsi->num_vlan > 1)
268 status = ice_fltr_clear_vlan_vsi_promisc(&vsi->back->hw, vsi, promisc_m);
270 status = ice_fltr_clear_vsi_promisc(&vsi->back->hw, vsi->idx, promisc_m, 0);
275 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
276 * @vsi: ptr to the VSI
278 * Push any outstanding VSI filter changes through the AdminQ.
280 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
282 struct device *dev = ice_pf_to_dev(vsi->back);
283 struct net_device *netdev = vsi->netdev;
284 bool promisc_forced_on = false;
285 struct ice_pf *pf = vsi->back;
286 struct ice_hw *hw = &pf->hw;
287 u32 changed_flags = 0;
294 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
295 usleep_range(1000, 2000);
297 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
298 vsi->current_netdev_flags = vsi->netdev->flags;
300 INIT_LIST_HEAD(&vsi->tmp_sync_list);
301 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
303 if (ice_vsi_fltr_changed(vsi)) {
304 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
305 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
306 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
308 /* grab the netdev's addr_list_lock */
309 netif_addr_lock_bh(netdev);
310 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
311 ice_add_mac_to_unsync_list);
312 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
313 ice_add_mac_to_unsync_list);
314 /* our temp lists are populated. release lock */
315 netif_addr_unlock_bh(netdev);
318 /* Remove MAC addresses in the unsync list */
319 err = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
320 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
322 netdev_err(netdev, "Failed to delete MAC filters\n");
323 /* if we failed because of alloc failures, just bail */
328 /* Add MAC addresses in the sync list */
329 err = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
330 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
331 /* If filter is added successfully or already exists, do not go into
332 * 'if' condition and report it as error. Instead continue processing
333 * rest of the function.
335 if (err && err != -EEXIST) {
336 netdev_err(netdev, "Failed to add MAC filters\n");
337 /* If there is no more space for new umac filters, VSI
338 * should go into promiscuous mode. There should be some
339 * space reserved for promiscuous filters.
341 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
342 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
344 promisc_forced_on = true;
345 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
352 /* check for changes in promiscuous modes */
353 if (changed_flags & IFF_ALLMULTI) {
354 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
355 if (vsi->num_vlan > 1)
356 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
358 promisc_m = ICE_MCAST_PROMISC_BITS;
360 err = ice_set_promisc(vsi, promisc_m);
362 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
364 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
368 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
369 if (vsi->num_vlan > 1)
370 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
372 promisc_m = ICE_MCAST_PROMISC_BITS;
374 err = ice_clear_promisc(vsi, promisc_m);
376 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
378 vsi->current_netdev_flags |= IFF_ALLMULTI;
384 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
385 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
386 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
387 if (vsi->current_netdev_flags & IFF_PROMISC) {
388 /* Apply Rx filter rule to get traffic from wire */
389 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
390 err = ice_set_dflt_vsi(pf->first_sw, vsi);
391 if (err && err != -EEXIST) {
392 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
394 vsi->current_netdev_flags &=
399 ice_cfg_vlan_pruning(vsi, false);
402 /* Clear Rx filter to remove traffic from wire */
403 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
404 err = ice_clear_dflt_vsi(pf->first_sw);
406 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
408 vsi->current_netdev_flags |=
412 if (vsi->num_vlan > 1)
413 ice_cfg_vlan_pruning(vsi, true);
420 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
423 /* if something went wrong then set the changed flag so we try again */
424 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
425 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
427 clear_bit(ICE_CFG_BUSY, vsi->state);
432 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
433 * @pf: board private structure
435 static void ice_sync_fltr_subtask(struct ice_pf *pf)
439 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
442 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
444 ice_for_each_vsi(pf, v)
445 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
446 ice_vsi_sync_fltr(pf->vsi[v])) {
447 /* come back and try again later */
448 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
454 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
456 * @locked: is the rtnl_lock already held
458 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
463 ice_for_each_vsi(pf, v)
465 ice_dis_vsi(pf->vsi[v], locked);
467 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
468 pf->pf_agg_node[node].num_vsis = 0;
470 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
471 pf->vf_agg_node[node].num_vsis = 0;
475 * ice_clear_sw_switch_recipes - clear switch recipes
476 * @pf: board private structure
478 * Mark switch recipes as not created in sw structures. There are cases where
479 * rules (especially advanced rules) need to be restored, either re-read from
480 * hardware or added again. For example after the reset. 'recp_created' flag
481 * prevents from doing that and need to be cleared upfront.
483 static void ice_clear_sw_switch_recipes(struct ice_pf *pf)
485 struct ice_sw_recipe *recp;
488 recp = pf->hw.switch_info->recp_list;
489 for (i = 0; i < ICE_MAX_NUM_RECIPES; i++)
490 recp[i].recp_created = false;
494 * ice_prepare_for_reset - prep for reset
495 * @pf: board private structure
496 * @reset_type: reset type requested
498 * Inform or close all dependent features in prep for reset.
501 ice_prepare_for_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
503 struct ice_hw *hw = &pf->hw;
507 dev_dbg(ice_pf_to_dev(pf), "reset_type=%d\n", reset_type);
509 /* already prepared for reset */
510 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
513 ice_unplug_aux_dev(pf);
515 /* Notify VFs of impending reset */
516 if (ice_check_sq_alive(hw, &hw->mailboxq))
517 ice_vc_notify_reset(pf);
519 /* Disable VFs until reset is completed */
520 ice_for_each_vf(pf, i)
521 ice_set_vf_state_qs_dis(&pf->vf[i]);
523 if (ice_is_eswitch_mode_switchdev(pf)) {
524 if (reset_type != ICE_RESET_PFR)
525 ice_clear_sw_switch_recipes(pf);
528 /* release ADQ specific HW and SW resources */
529 vsi = ice_get_main_vsi(pf);
533 /* to be on safe side, reset orig_rss_size so that normal flow
534 * of deciding rss_size can take precedence
536 vsi->orig_rss_size = 0;
538 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
539 if (reset_type == ICE_RESET_PFR) {
540 vsi->old_ena_tc = vsi->all_enatc;
541 vsi->old_numtc = vsi->all_numtc;
543 ice_remove_q_channels(vsi, true);
545 /* for other reset type, do not support channel rebuild
546 * hence reset needed info
554 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
555 memset(&vsi->mqprio_qopt, 0, sizeof(vsi->mqprio_qopt));
560 /* clear SW filtering DB */
561 ice_clear_hw_tbls(hw);
562 /* disable the VSIs and their queues that are not already DOWN */
563 ice_pf_dis_all_vsi(pf, false);
565 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
566 ice_ptp_prepare_for_reset(pf);
569 ice_sched_clear_port(hw->port_info);
571 ice_shutdown_all_ctrlq(hw);
573 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
577 * ice_do_reset - Initiate one of many types of resets
578 * @pf: board private structure
579 * @reset_type: reset type requested before this function was called.
581 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
583 struct device *dev = ice_pf_to_dev(pf);
584 struct ice_hw *hw = &pf->hw;
586 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
588 ice_prepare_for_reset(pf, reset_type);
590 /* trigger the reset */
591 if (ice_reset(hw, reset_type)) {
592 dev_err(dev, "reset %d failed\n", reset_type);
593 set_bit(ICE_RESET_FAILED, pf->state);
594 clear_bit(ICE_RESET_OICR_RECV, pf->state);
595 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
596 clear_bit(ICE_PFR_REQ, pf->state);
597 clear_bit(ICE_CORER_REQ, pf->state);
598 clear_bit(ICE_GLOBR_REQ, pf->state);
599 wake_up(&pf->reset_wait_queue);
603 /* PFR is a bit of a special case because it doesn't result in an OICR
604 * interrupt. So for PFR, rebuild after the reset and clear the reset-
605 * associated state bits.
607 if (reset_type == ICE_RESET_PFR) {
609 ice_rebuild(pf, reset_type);
610 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
611 clear_bit(ICE_PFR_REQ, pf->state);
612 wake_up(&pf->reset_wait_queue);
613 ice_reset_all_vfs(pf, true);
618 * ice_reset_subtask - Set up for resetting the device and driver
619 * @pf: board private structure
621 static void ice_reset_subtask(struct ice_pf *pf)
623 enum ice_reset_req reset_type = ICE_RESET_INVAL;
625 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
626 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
627 * of reset is pending and sets bits in pf->state indicating the reset
628 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
629 * prepare for pending reset if not already (for PF software-initiated
630 * global resets the software should already be prepared for it as
631 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
632 * by firmware or software on other PFs, that bit is not set so prepare
633 * for the reset now), poll for reset done, rebuild and return.
635 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
636 /* Perform the largest reset requested */
637 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
638 reset_type = ICE_RESET_CORER;
639 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
640 reset_type = ICE_RESET_GLOBR;
641 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
642 reset_type = ICE_RESET_EMPR;
643 /* return if no valid reset type requested */
644 if (reset_type == ICE_RESET_INVAL)
646 ice_prepare_for_reset(pf, reset_type);
648 /* make sure we are ready to rebuild */
649 if (ice_check_reset(&pf->hw)) {
650 set_bit(ICE_RESET_FAILED, pf->state);
652 /* done with reset. start rebuild */
653 pf->hw.reset_ongoing = false;
654 ice_rebuild(pf, reset_type);
655 /* clear bit to resume normal operations, but
656 * ICE_NEEDS_RESTART bit is set in case rebuild failed
658 clear_bit(ICE_RESET_OICR_RECV, pf->state);
659 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
660 clear_bit(ICE_PFR_REQ, pf->state);
661 clear_bit(ICE_CORER_REQ, pf->state);
662 clear_bit(ICE_GLOBR_REQ, pf->state);
663 wake_up(&pf->reset_wait_queue);
664 ice_reset_all_vfs(pf, true);
670 /* No pending resets to finish processing. Check for new resets */
671 if (test_bit(ICE_PFR_REQ, pf->state))
672 reset_type = ICE_RESET_PFR;
673 if (test_bit(ICE_CORER_REQ, pf->state))
674 reset_type = ICE_RESET_CORER;
675 if (test_bit(ICE_GLOBR_REQ, pf->state))
676 reset_type = ICE_RESET_GLOBR;
677 /* If no valid reset type requested just return */
678 if (reset_type == ICE_RESET_INVAL)
681 /* reset if not already down or busy */
682 if (!test_bit(ICE_DOWN, pf->state) &&
683 !test_bit(ICE_CFG_BUSY, pf->state)) {
684 ice_do_reset(pf, reset_type);
689 * ice_print_topo_conflict - print topology conflict message
690 * @vsi: the VSI whose topology status is being checked
692 static void ice_print_topo_conflict(struct ice_vsi *vsi)
694 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
695 case ICE_AQ_LINK_TOPO_CONFLICT:
696 case ICE_AQ_LINK_MEDIA_CONFLICT:
697 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
698 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
699 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
700 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");
702 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
703 if (test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, vsi->back->flags))
704 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");
706 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");
714 * ice_print_link_msg - print link up or down message
715 * @vsi: the VSI whose link status is being queried
716 * @isup: boolean for if the link is now up or down
718 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
720 struct ice_aqc_get_phy_caps_data *caps;
721 const char *an_advertised;
732 if (vsi->current_isup == isup)
735 vsi->current_isup = isup;
738 netdev_info(vsi->netdev, "NIC Link is Down\n");
742 switch (vsi->port_info->phy.link_info.link_speed) {
743 case ICE_AQ_LINK_SPEED_100GB:
746 case ICE_AQ_LINK_SPEED_50GB:
749 case ICE_AQ_LINK_SPEED_40GB:
752 case ICE_AQ_LINK_SPEED_25GB:
755 case ICE_AQ_LINK_SPEED_20GB:
758 case ICE_AQ_LINK_SPEED_10GB:
761 case ICE_AQ_LINK_SPEED_5GB:
764 case ICE_AQ_LINK_SPEED_2500MB:
767 case ICE_AQ_LINK_SPEED_1000MB:
770 case ICE_AQ_LINK_SPEED_100MB:
778 switch (vsi->port_info->fc.current_mode) {
782 case ICE_FC_TX_PAUSE:
785 case ICE_FC_RX_PAUSE:
796 /* Get FEC mode based on negotiated link info */
797 switch (vsi->port_info->phy.link_info.fec_info) {
798 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
799 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
802 case ICE_AQ_LINK_25G_KR_FEC_EN:
803 fec = "FC-FEC/BASE-R";
810 /* check if autoneg completed, might be false due to not supported */
811 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
816 /* Get FEC mode requested based on PHY caps last SW configuration */
817 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
820 an_advertised = "Unknown";
824 status = ice_aq_get_phy_caps(vsi->port_info, false,
825 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
827 netdev_info(vsi->netdev, "Get phy capability failed.\n");
829 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
831 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
832 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
834 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
835 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
836 fec_req = "FC-FEC/BASE-R";
843 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",
844 speed, fec_req, fec, an_advertised, an, fc);
845 ice_print_topo_conflict(vsi);
849 * ice_vsi_link_event - update the VSI's netdev
850 * @vsi: the VSI on which the link event occurred
851 * @link_up: whether or not the VSI needs to be set up or down
853 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
858 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
861 if (vsi->type == ICE_VSI_PF) {
862 if (link_up == netif_carrier_ok(vsi->netdev))
866 netif_carrier_on(vsi->netdev);
867 netif_tx_wake_all_queues(vsi->netdev);
869 netif_carrier_off(vsi->netdev);
870 netif_tx_stop_all_queues(vsi->netdev);
876 * ice_set_dflt_mib - send a default config MIB to the FW
877 * @pf: private PF struct
879 * This function sends a default configuration MIB to the FW.
881 * If this function errors out at any point, the driver is still able to
882 * function. The main impact is that LFC may not operate as expected.
883 * Therefore an error state in this function should be treated with a DBG
884 * message and continue on with driver rebuild/reenable.
886 static void ice_set_dflt_mib(struct ice_pf *pf)
888 struct device *dev = ice_pf_to_dev(pf);
889 u8 mib_type, *buf, *lldpmib = NULL;
890 u16 len, typelen, offset = 0;
891 struct ice_lldp_org_tlv *tlv;
892 struct ice_hw *hw = &pf->hw;
895 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
896 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
898 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
903 /* Add ETS CFG TLV */
904 tlv = (struct ice_lldp_org_tlv *)lldpmib;
905 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
906 ICE_IEEE_ETS_TLV_LEN);
907 tlv->typelen = htons(typelen);
908 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
909 ICE_IEEE_SUBTYPE_ETS_CFG);
910 tlv->ouisubtype = htonl(ouisubtype);
915 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
916 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
917 * Octets 13 - 20 are TSA values - leave as zeros
920 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
922 tlv = (struct ice_lldp_org_tlv *)
923 ((char *)tlv + sizeof(tlv->typelen) + len);
925 /* Add ETS REC TLV */
927 tlv->typelen = htons(typelen);
929 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
930 ICE_IEEE_SUBTYPE_ETS_REC);
931 tlv->ouisubtype = htonl(ouisubtype);
933 /* First octet of buf is reserved
934 * Octets 1 - 4 map UP to TC - all UPs map to zero
935 * Octets 5 - 12 are BW values - set TC 0 to 100%.
936 * Octets 13 - 20 are TSA value - leave as zeros
940 tlv = (struct ice_lldp_org_tlv *)
941 ((char *)tlv + sizeof(tlv->typelen) + len);
943 /* Add PFC CFG TLV */
944 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
945 ICE_IEEE_PFC_TLV_LEN);
946 tlv->typelen = htons(typelen);
948 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
949 ICE_IEEE_SUBTYPE_PFC_CFG);
950 tlv->ouisubtype = htonl(ouisubtype);
952 /* Octet 1 left as all zeros - PFC disabled */
954 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
957 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
958 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
964 * ice_check_phy_fw_load - check if PHY FW load failed
965 * @pf: pointer to PF struct
966 * @link_cfg_err: bitmap from the link info structure
968 * check if external PHY FW load failed and print an error message if it did
970 static void ice_check_phy_fw_load(struct ice_pf *pf, u8 link_cfg_err)
972 if (!(link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE)) {
973 clear_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
977 if (test_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags))
980 if (link_cfg_err & ICE_AQ_LINK_EXTERNAL_PHY_LOAD_FAILURE) {
981 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");
982 set_bit(ICE_FLAG_PHY_FW_LOAD_FAILED, pf->flags);
987 * ice_check_module_power
988 * @pf: pointer to PF struct
989 * @link_cfg_err: bitmap from the link info structure
991 * check module power level returned by a previous call to aq_get_link_info
992 * and print error messages if module power level is not supported
994 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
996 /* if module power level is supported, clear the flag */
997 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
998 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
999 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1003 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
1004 * above block didn't clear this bit, there's nothing to do
1006 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
1009 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
1010 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
1011 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1012 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
1013 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
1014 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
1019 * ice_check_link_cfg_err - check if link configuration failed
1020 * @pf: pointer to the PF struct
1021 * @link_cfg_err: bitmap from the link info structure
1023 * print if any link configuration failure happens due to the value in the
1024 * link_cfg_err parameter in the link info structure
1026 static void ice_check_link_cfg_err(struct ice_pf *pf, u8 link_cfg_err)
1028 ice_check_module_power(pf, link_cfg_err);
1029 ice_check_phy_fw_load(pf, link_cfg_err);
1033 * ice_link_event - process the link event
1034 * @pf: PF that the link event is associated with
1035 * @pi: port_info for the port that the link event is associated with
1036 * @link_up: true if the physical link is up and false if it is down
1037 * @link_speed: current link speed received from the link event
1039 * Returns 0 on success and negative on failure
1042 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
1045 struct device *dev = ice_pf_to_dev(pf);
1046 struct ice_phy_info *phy_info;
1047 struct ice_vsi *vsi;
1052 phy_info = &pi->phy;
1053 phy_info->link_info_old = phy_info->link_info;
1055 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
1056 old_link_speed = phy_info->link_info_old.link_speed;
1058 /* update the link info structures and re-enable link events,
1059 * don't bail on failure due to other book keeping needed
1061 status = ice_update_link_info(pi);
1063 dev_dbg(dev, "Failed to update link status on port %d, err %d aq_err %s\n",
1065 ice_aq_str(pi->hw->adminq.sq_last_status));
1067 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
1069 /* Check if the link state is up after updating link info, and treat
1070 * this event as an UP event since the link is actually UP now.
1072 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
1075 vsi = ice_get_main_vsi(pf);
1076 if (!vsi || !vsi->port_info)
1079 /* turn off PHY if media was removed */
1080 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
1081 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
1082 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
1083 ice_set_link(vsi, false);
1086 /* if the old link up/down and speed is the same as the new */
1087 if (link_up == old_link && link_speed == old_link_speed)
1090 if (!ice_is_e810(&pf->hw))
1091 ice_ptp_link_change(pf, pf->hw.pf_id, link_up);
1093 if (ice_is_dcb_active(pf)) {
1094 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
1095 ice_dcb_rebuild(pf);
1098 ice_set_dflt_mib(pf);
1100 ice_vsi_link_event(vsi, link_up);
1101 ice_print_link_msg(vsi, link_up);
1103 ice_vc_notify_link_state(pf);
1109 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
1110 * @pf: board private structure
1112 static void ice_watchdog_subtask(struct ice_pf *pf)
1116 /* if interface is down do nothing */
1117 if (test_bit(ICE_DOWN, pf->state) ||
1118 test_bit(ICE_CFG_BUSY, pf->state))
1121 /* make sure we don't do these things too often */
1122 if (time_before(jiffies,
1123 pf->serv_tmr_prev + pf->serv_tmr_period))
1126 pf->serv_tmr_prev = jiffies;
1128 /* Update the stats for active netdevs so the network stack
1129 * can look at updated numbers whenever it cares to
1131 ice_update_pf_stats(pf);
1132 ice_for_each_vsi(pf, i)
1133 if (pf->vsi[i] && pf->vsi[i]->netdev)
1134 ice_update_vsi_stats(pf->vsi[i]);
1138 * ice_init_link_events - enable/initialize link events
1139 * @pi: pointer to the port_info instance
1141 * Returns -EIO on failure, 0 on success
1143 static int ice_init_link_events(struct ice_port_info *pi)
1147 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1148 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL |
1149 ICE_AQ_LINK_EVENT_PHY_FW_LOAD_FAIL));
1151 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1152 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1157 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1158 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1167 * ice_handle_link_event - handle link event via ARQ
1168 * @pf: PF that the link event is associated with
1169 * @event: event structure containing link status info
1172 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1174 struct ice_aqc_get_link_status_data *link_data;
1175 struct ice_port_info *port_info;
1178 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1179 port_info = pf->hw.port_info;
1183 status = ice_link_event(pf, port_info,
1184 !!(link_data->link_info & ICE_AQ_LINK_UP),
1185 le16_to_cpu(link_data->link_speed));
1187 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1193 enum ice_aq_task_state {
1194 ICE_AQ_TASK_WAITING = 0,
1195 ICE_AQ_TASK_COMPLETE,
1196 ICE_AQ_TASK_CANCELED,
1199 struct ice_aq_task {
1200 struct hlist_node entry;
1203 struct ice_rq_event_info *event;
1204 enum ice_aq_task_state state;
1208 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1209 * @pf: pointer to the PF private structure
1210 * @opcode: the opcode to wait for
1211 * @timeout: how long to wait, in jiffies
1212 * @event: storage for the event info
1214 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1215 * current thread will be put to sleep until the specified event occurs or
1216 * until the given timeout is reached.
1218 * To obtain only the descriptor contents, pass an event without an allocated
1219 * msg_buf. If the complete data buffer is desired, allocate the
1220 * event->msg_buf with enough space ahead of time.
1222 * Returns: zero on success, or a negative error code on failure.
1224 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1225 struct ice_rq_event_info *event)
1227 struct device *dev = ice_pf_to_dev(pf);
1228 struct ice_aq_task *task;
1229 unsigned long start;
1233 task = kzalloc(sizeof(*task), GFP_KERNEL);
1237 INIT_HLIST_NODE(&task->entry);
1238 task->opcode = opcode;
1239 task->event = event;
1240 task->state = ICE_AQ_TASK_WAITING;
1242 spin_lock_bh(&pf->aq_wait_lock);
1243 hlist_add_head(&task->entry, &pf->aq_wait_list);
1244 spin_unlock_bh(&pf->aq_wait_lock);
1248 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1250 switch (task->state) {
1251 case ICE_AQ_TASK_WAITING:
1252 err = ret < 0 ? ret : -ETIMEDOUT;
1254 case ICE_AQ_TASK_CANCELED:
1255 err = ret < 0 ? ret : -ECANCELED;
1257 case ICE_AQ_TASK_COMPLETE:
1258 err = ret < 0 ? ret : 0;
1261 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1266 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1267 jiffies_to_msecs(jiffies - start),
1268 jiffies_to_msecs(timeout),
1271 spin_lock_bh(&pf->aq_wait_lock);
1272 hlist_del(&task->entry);
1273 spin_unlock_bh(&pf->aq_wait_lock);
1280 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1281 * @pf: pointer to the PF private structure
1282 * @opcode: the opcode of the event
1283 * @event: the event to check
1285 * Loops over the current list of pending threads waiting for an AdminQ event.
1286 * For each matching task, copy the contents of the event into the task
1287 * structure and wake up the thread.
1289 * If multiple threads wait for the same opcode, they will all be woken up.
1291 * Note that event->msg_buf will only be duplicated if the event has a buffer
1292 * with enough space already allocated. Otherwise, only the descriptor and
1293 * message length will be copied.
1295 * Returns: true if an event was found, false otherwise
1297 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1298 struct ice_rq_event_info *event)
1300 struct ice_aq_task *task;
1303 spin_lock_bh(&pf->aq_wait_lock);
1304 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1305 if (task->state || task->opcode != opcode)
1308 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1309 task->event->msg_len = event->msg_len;
1311 /* Only copy the data buffer if a destination was set */
1312 if (task->event->msg_buf &&
1313 task->event->buf_len > event->buf_len) {
1314 memcpy(task->event->msg_buf, event->msg_buf,
1316 task->event->buf_len = event->buf_len;
1319 task->state = ICE_AQ_TASK_COMPLETE;
1322 spin_unlock_bh(&pf->aq_wait_lock);
1325 wake_up(&pf->aq_wait_queue);
1329 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1330 * @pf: the PF private structure
1332 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1333 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1335 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1337 struct ice_aq_task *task;
1339 spin_lock_bh(&pf->aq_wait_lock);
1340 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1341 task->state = ICE_AQ_TASK_CANCELED;
1342 spin_unlock_bh(&pf->aq_wait_lock);
1344 wake_up(&pf->aq_wait_queue);
1348 * __ice_clean_ctrlq - helper function to clean controlq rings
1349 * @pf: ptr to struct ice_pf
1350 * @q_type: specific Control queue type
1352 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1354 struct device *dev = ice_pf_to_dev(pf);
1355 struct ice_rq_event_info event;
1356 struct ice_hw *hw = &pf->hw;
1357 struct ice_ctl_q_info *cq;
1362 /* Do not clean control queue if/when PF reset fails */
1363 if (test_bit(ICE_RESET_FAILED, pf->state))
1367 case ICE_CTL_Q_ADMIN:
1375 case ICE_CTL_Q_MAILBOX:
1378 /* we are going to try to detect a malicious VF, so set the
1379 * state to begin detection
1381 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1384 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1388 /* check for error indications - PF_xx_AxQLEN register layout for
1389 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1391 val = rd32(hw, cq->rq.len);
1392 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1393 PF_FW_ARQLEN_ARQCRIT_M)) {
1395 if (val & PF_FW_ARQLEN_ARQVFE_M)
1396 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1398 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1399 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1402 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1403 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1405 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1406 PF_FW_ARQLEN_ARQCRIT_M);
1408 wr32(hw, cq->rq.len, val);
1411 val = rd32(hw, cq->sq.len);
1412 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1413 PF_FW_ATQLEN_ATQCRIT_M)) {
1415 if (val & PF_FW_ATQLEN_ATQVFE_M)
1416 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1418 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1419 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1422 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1423 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1425 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1426 PF_FW_ATQLEN_ATQCRIT_M);
1428 wr32(hw, cq->sq.len, val);
1431 event.buf_len = cq->rq_buf_size;
1432 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1440 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1441 if (ret == -EALREADY)
1444 dev_err(dev, "%s Receive Queue event error %d\n", qtype,
1449 opcode = le16_to_cpu(event.desc.opcode);
1451 /* Notify any thread that might be waiting for this event */
1452 ice_aq_check_events(pf, opcode, &event);
1455 case ice_aqc_opc_get_link_status:
1456 if (ice_handle_link_event(pf, &event))
1457 dev_err(dev, "Could not handle link event\n");
1459 case ice_aqc_opc_event_lan_overflow:
1460 ice_vf_lan_overflow_event(pf, &event);
1462 case ice_mbx_opc_send_msg_to_pf:
1463 if (!ice_is_malicious_vf(pf, &event, i, pending))
1464 ice_vc_process_vf_msg(pf, &event);
1466 case ice_aqc_opc_fw_logging:
1467 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1469 case ice_aqc_opc_lldp_set_mib_change:
1470 ice_dcb_process_lldp_set_mib_change(pf, &event);
1473 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1477 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1479 kfree(event.msg_buf);
1481 return pending && (i == ICE_DFLT_IRQ_WORK);
1485 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1486 * @hw: pointer to hardware info
1487 * @cq: control queue information
1489 * returns true if there are pending messages in a queue, false if there aren't
1491 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1495 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1496 return cq->rq.next_to_clean != ntu;
1500 * ice_clean_adminq_subtask - clean the AdminQ rings
1501 * @pf: board private structure
1503 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1505 struct ice_hw *hw = &pf->hw;
1507 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1510 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1513 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1515 /* There might be a situation where new messages arrive to a control
1516 * queue between processing the last message and clearing the
1517 * EVENT_PENDING bit. So before exiting, check queue head again (using
1518 * ice_ctrlq_pending) and process new messages if any.
1520 if (ice_ctrlq_pending(hw, &hw->adminq))
1521 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1527 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1528 * @pf: board private structure
1530 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1532 struct ice_hw *hw = &pf->hw;
1534 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1537 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1540 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1542 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1543 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1549 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1550 * @pf: board private structure
1552 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1554 struct ice_hw *hw = &pf->hw;
1556 /* Nothing to do here if sideband queue is not supported */
1557 if (!ice_is_sbq_supported(hw)) {
1558 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1562 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1565 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1568 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1570 if (ice_ctrlq_pending(hw, &hw->sbq))
1571 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1577 * ice_service_task_schedule - schedule the service task to wake up
1578 * @pf: board private structure
1580 * If not already scheduled, this puts the task into the work queue.
1582 void ice_service_task_schedule(struct ice_pf *pf)
1584 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1585 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1586 !test_bit(ICE_NEEDS_RESTART, pf->state))
1587 queue_work(ice_wq, &pf->serv_task);
1591 * ice_service_task_complete - finish up the service task
1592 * @pf: board private structure
1594 static void ice_service_task_complete(struct ice_pf *pf)
1596 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1598 /* force memory (pf->state) to sync before next service task */
1599 smp_mb__before_atomic();
1600 clear_bit(ICE_SERVICE_SCHED, pf->state);
1604 * ice_service_task_stop - stop service task and cancel works
1605 * @pf: board private structure
1607 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1610 static int ice_service_task_stop(struct ice_pf *pf)
1614 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1616 if (pf->serv_tmr.function)
1617 del_timer_sync(&pf->serv_tmr);
1618 if (pf->serv_task.func)
1619 cancel_work_sync(&pf->serv_task);
1621 clear_bit(ICE_SERVICE_SCHED, pf->state);
1626 * ice_service_task_restart - restart service task and schedule works
1627 * @pf: board private structure
1629 * This function is needed for suspend and resume works (e.g WoL scenario)
1631 static void ice_service_task_restart(struct ice_pf *pf)
1633 clear_bit(ICE_SERVICE_DIS, pf->state);
1634 ice_service_task_schedule(pf);
1638 * ice_service_timer - timer callback to schedule service task
1639 * @t: pointer to timer_list
1641 static void ice_service_timer(struct timer_list *t)
1643 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1645 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1646 ice_service_task_schedule(pf);
1650 * ice_handle_mdd_event - handle malicious driver detect event
1651 * @pf: pointer to the PF structure
1653 * Called from service task. OICR interrupt handler indicates MDD event.
1654 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1655 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1656 * disable the queue, the PF can be configured to reset the VF using ethtool
1657 * private flag mdd-auto-reset-vf.
1659 static void ice_handle_mdd_event(struct ice_pf *pf)
1661 struct device *dev = ice_pf_to_dev(pf);
1662 struct ice_hw *hw = &pf->hw;
1666 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1667 /* Since the VF MDD event logging is rate limited, check if
1668 * there are pending MDD events.
1670 ice_print_vfs_mdd_events(pf);
1674 /* find what triggered an MDD event */
1675 reg = rd32(hw, GL_MDET_TX_PQM);
1676 if (reg & GL_MDET_TX_PQM_VALID_M) {
1677 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1678 GL_MDET_TX_PQM_PF_NUM_S;
1679 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1680 GL_MDET_TX_PQM_VF_NUM_S;
1681 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1682 GL_MDET_TX_PQM_MAL_TYPE_S;
1683 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1684 GL_MDET_TX_PQM_QNUM_S);
1686 if (netif_msg_tx_err(pf))
1687 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1688 event, queue, pf_num, vf_num);
1689 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1692 reg = rd32(hw, GL_MDET_TX_TCLAN);
1693 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1694 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1695 GL_MDET_TX_TCLAN_PF_NUM_S;
1696 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1697 GL_MDET_TX_TCLAN_VF_NUM_S;
1698 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1699 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1700 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1701 GL_MDET_TX_TCLAN_QNUM_S);
1703 if (netif_msg_tx_err(pf))
1704 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1705 event, queue, pf_num, vf_num);
1706 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1709 reg = rd32(hw, GL_MDET_RX);
1710 if (reg & GL_MDET_RX_VALID_M) {
1711 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1712 GL_MDET_RX_PF_NUM_S;
1713 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1714 GL_MDET_RX_VF_NUM_S;
1715 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1716 GL_MDET_RX_MAL_TYPE_S;
1717 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1720 if (netif_msg_rx_err(pf))
1721 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1722 event, queue, pf_num, vf_num);
1723 wr32(hw, GL_MDET_RX, 0xffffffff);
1726 /* check to see if this PF caused an MDD event */
1727 reg = rd32(hw, PF_MDET_TX_PQM);
1728 if (reg & PF_MDET_TX_PQM_VALID_M) {
1729 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1730 if (netif_msg_tx_err(pf))
1731 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1734 reg = rd32(hw, PF_MDET_TX_TCLAN);
1735 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1736 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1737 if (netif_msg_tx_err(pf))
1738 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1741 reg = rd32(hw, PF_MDET_RX);
1742 if (reg & PF_MDET_RX_VALID_M) {
1743 wr32(hw, PF_MDET_RX, 0xFFFF);
1744 if (netif_msg_rx_err(pf))
1745 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1748 /* Check to see if one of the VFs caused an MDD event, and then
1749 * increment counters and set print pending
1751 ice_for_each_vf(pf, i) {
1752 struct ice_vf *vf = &pf->vf[i];
1754 reg = rd32(hw, VP_MDET_TX_PQM(i));
1755 if (reg & VP_MDET_TX_PQM_VALID_M) {
1756 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1757 vf->mdd_tx_events.count++;
1758 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1759 if (netif_msg_tx_err(pf))
1760 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1764 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1765 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1766 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1767 vf->mdd_tx_events.count++;
1768 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1769 if (netif_msg_tx_err(pf))
1770 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1774 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1775 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1776 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1777 vf->mdd_tx_events.count++;
1778 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1779 if (netif_msg_tx_err(pf))
1780 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1784 reg = rd32(hw, VP_MDET_RX(i));
1785 if (reg & VP_MDET_RX_VALID_M) {
1786 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1787 vf->mdd_rx_events.count++;
1788 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1789 if (netif_msg_rx_err(pf))
1790 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1793 /* Since the queue is disabled on VF Rx MDD events, the
1794 * PF can be configured to reset the VF through ethtool
1795 * private flag mdd-auto-reset-vf.
1797 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1798 /* VF MDD event counters will be cleared by
1799 * reset, so print the event prior to reset.
1801 ice_print_vf_rx_mdd_event(vf);
1802 mutex_lock(&pf->vf[i].cfg_lock);
1803 ice_reset_vf(&pf->vf[i], false);
1804 mutex_unlock(&pf->vf[i].cfg_lock);
1809 ice_print_vfs_mdd_events(pf);
1813 * ice_force_phys_link_state - Force the physical link state
1814 * @vsi: VSI to force the physical link state to up/down
1815 * @link_up: true/false indicates to set the physical link to up/down
1817 * Force the physical link state by getting the current PHY capabilities from
1818 * hardware and setting the PHY config based on the determined capabilities. If
1819 * link changes a link event will be triggered because both the Enable Automatic
1820 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1822 * Returns 0 on success, negative on failure
1824 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1826 struct ice_aqc_get_phy_caps_data *pcaps;
1827 struct ice_aqc_set_phy_cfg_data *cfg;
1828 struct ice_port_info *pi;
1832 if (!vsi || !vsi->port_info || !vsi->back)
1834 if (vsi->type != ICE_VSI_PF)
1837 dev = ice_pf_to_dev(vsi->back);
1839 pi = vsi->port_info;
1841 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1845 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1848 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1849 vsi->vsi_num, retcode);
1854 /* No change in link */
1855 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1856 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1859 /* Use the current user PHY configuration. The current user PHY
1860 * configuration is initialized during probe from PHY capabilities
1861 * software mode, and updated on set PHY configuration.
1863 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1869 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1871 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1873 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1875 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1877 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1878 vsi->vsi_num, retcode);
1889 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1890 * @pi: port info structure
1892 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1894 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1896 struct ice_aqc_get_phy_caps_data *pcaps;
1897 struct ice_pf *pf = pi->hw->back;
1900 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1904 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA,
1908 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1912 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1913 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1921 * ice_init_link_dflt_override - Initialize link default override
1922 * @pi: port info structure
1924 * Initialize link default override and PHY total port shutdown during probe
1926 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1928 struct ice_link_default_override_tlv *ldo;
1929 struct ice_pf *pf = pi->hw->back;
1931 ldo = &pf->link_dflt_override;
1932 if (ice_get_link_default_override(ldo, pi))
1935 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1938 /* Enable Total Port Shutdown (override/replace link-down-on-close
1939 * ethtool private flag) for ports with Port Disable bit set.
1941 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1942 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1946 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1947 * @pi: port info structure
1949 * If default override is enabled, initialize the user PHY cfg speed and FEC
1950 * settings using the default override mask from the NVM.
1952 * The PHY should only be configured with the default override settings the
1953 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1954 * is used to indicate that the user PHY cfg default override is initialized
1955 * and the PHY has not been configured with the default override settings. The
1956 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1959 * This function should be called only if the FW doesn't support default
1960 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1962 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1964 struct ice_link_default_override_tlv *ldo;
1965 struct ice_aqc_set_phy_cfg_data *cfg;
1966 struct ice_phy_info *phy = &pi->phy;
1967 struct ice_pf *pf = pi->hw->back;
1969 ldo = &pf->link_dflt_override;
1971 /* If link default override is enabled, use to mask NVM PHY capabilities
1972 * for speed and FEC default configuration.
1974 cfg = &phy->curr_user_phy_cfg;
1976 if (ldo->phy_type_low || ldo->phy_type_high) {
1977 cfg->phy_type_low = pf->nvm_phy_type_lo &
1978 cpu_to_le64(ldo->phy_type_low);
1979 cfg->phy_type_high = pf->nvm_phy_type_hi &
1980 cpu_to_le64(ldo->phy_type_high);
1982 cfg->link_fec_opt = ldo->fec_options;
1983 phy->curr_user_fec_req = ICE_FEC_AUTO;
1985 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1989 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1990 * @pi: port info structure
1992 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1993 * mode to default. The PHY defaults are from get PHY capabilities topology
1994 * with media so call when media is first available. An error is returned if
1995 * called when media is not available. The PHY initialization completed state is
1998 * These configurations are used when setting PHY
1999 * configuration. The user PHY configuration is updated on set PHY
2000 * configuration. Returns 0 on success, negative on failure
2002 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
2004 struct ice_aqc_get_phy_caps_data *pcaps;
2005 struct ice_phy_info *phy = &pi->phy;
2006 struct ice_pf *pf = pi->hw->back;
2009 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2012 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2016 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2017 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2020 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2023 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
2027 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
2029 /* check if lenient mode is supported and enabled */
2030 if (ice_fw_supports_link_override(pi->hw) &&
2031 !(pcaps->module_compliance_enforcement &
2032 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
2033 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
2035 /* if the FW supports default PHY configuration mode, then the driver
2036 * does not have to apply link override settings. If not,
2037 * initialize user PHY configuration with link override values
2039 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
2040 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
2041 ice_init_phy_cfg_dflt_override(pi);
2046 /* if link default override is not enabled, set user flow control and
2047 * FEC settings based on what get_phy_caps returned
2049 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
2050 pcaps->link_fec_options);
2051 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
2054 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
2055 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
2062 * ice_configure_phy - configure PHY
2065 * Set the PHY configuration. If the current PHY configuration is the same as
2066 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
2067 * configure the based get PHY capabilities for topology with media.
2069 static int ice_configure_phy(struct ice_vsi *vsi)
2071 struct device *dev = ice_pf_to_dev(vsi->back);
2072 struct ice_port_info *pi = vsi->port_info;
2073 struct ice_aqc_get_phy_caps_data *pcaps;
2074 struct ice_aqc_set_phy_cfg_data *cfg;
2075 struct ice_phy_info *phy = &pi->phy;
2076 struct ice_pf *pf = vsi->back;
2079 /* Ensure we have media as we cannot configure a medialess port */
2080 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
2083 ice_print_topo_conflict(vsi);
2085 if (!test_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags) &&
2086 phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
2089 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
2090 return ice_force_phys_link_state(vsi, true);
2092 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
2096 /* Get current PHY config */
2097 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
2100 dev_err(dev, "Failed to get PHY configuration, VSI %d error %d\n",
2105 /* If PHY enable link is configured and configuration has not changed,
2106 * there's nothing to do
2108 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
2109 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
2112 /* Use PHY topology as baseline for configuration */
2113 memset(pcaps, 0, sizeof(*pcaps));
2114 if (ice_fw_supports_report_dflt_cfg(pi->hw))
2115 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2118 err = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2121 dev_err(dev, "Failed to get PHY caps, VSI %d error %d\n",
2126 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2132 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2134 /* Speed - If default override pending, use curr_user_phy_cfg set in
2135 * ice_init_phy_user_cfg_ldo.
2137 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2138 vsi->back->state)) {
2139 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2140 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2142 u64 phy_low = 0, phy_high = 0;
2144 ice_update_phy_type(&phy_low, &phy_high,
2145 pi->phy.curr_user_speed_req);
2146 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2147 cfg->phy_type_high = pcaps->phy_type_high &
2148 cpu_to_le64(phy_high);
2151 /* Can't provide what was requested; use PHY capabilities */
2152 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2153 cfg->phy_type_low = pcaps->phy_type_low;
2154 cfg->phy_type_high = pcaps->phy_type_high;
2158 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2160 /* Can't provide what was requested; use PHY capabilities */
2161 if (cfg->link_fec_opt !=
2162 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2163 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2164 cfg->link_fec_opt = pcaps->link_fec_options;
2167 /* Flow Control - always supported; no need to check against
2170 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2172 /* Enable link and link update */
2173 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2175 err = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2177 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
2187 * ice_check_media_subtask - Check for media
2188 * @pf: pointer to PF struct
2190 * If media is available, then initialize PHY user configuration if it is not
2191 * been, and configure the PHY if the interface is up.
2193 static void ice_check_media_subtask(struct ice_pf *pf)
2195 struct ice_port_info *pi;
2196 struct ice_vsi *vsi;
2199 /* No need to check for media if it's already present */
2200 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2203 vsi = ice_get_main_vsi(pf);
2207 /* Refresh link info and check if media is present */
2208 pi = vsi->port_info;
2209 err = ice_update_link_info(pi);
2213 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
2215 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2216 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2217 ice_init_phy_user_cfg(pi);
2219 /* PHY settings are reset on media insertion, reconfigure
2220 * PHY to preserve settings.
2222 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2223 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2226 err = ice_configure_phy(vsi);
2228 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2230 /* A Link Status Event will be generated; the event handler
2231 * will complete bringing the interface up
2237 * ice_service_task - manage and run subtasks
2238 * @work: pointer to work_struct contained by the PF struct
2240 static void ice_service_task(struct work_struct *work)
2242 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2243 unsigned long start_time = jiffies;
2247 /* process reset requests first */
2248 ice_reset_subtask(pf);
2250 /* bail if a reset/recovery cycle is pending or rebuild failed */
2251 if (ice_is_reset_in_progress(pf->state) ||
2252 test_bit(ICE_SUSPENDED, pf->state) ||
2253 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2254 ice_service_task_complete(pf);
2258 if (test_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags)) {
2259 /* Plug aux device per request */
2260 ice_plug_aux_dev(pf);
2262 /* Mark plugging as done but check whether unplug was
2263 * requested during ice_plug_aux_dev() call
2264 * (e.g. from ice_clear_rdma_cap()) and if so then
2267 if (!test_and_clear_bit(ICE_FLAG_PLUG_AUX_DEV, pf->flags))
2268 ice_unplug_aux_dev(pf);
2271 if (test_and_clear_bit(ICE_FLAG_MTU_CHANGED, pf->flags)) {
2272 struct iidc_event *event;
2274 event = kzalloc(sizeof(*event), GFP_KERNEL);
2276 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
2277 ice_send_event_to_aux(pf, event);
2282 ice_clean_adminq_subtask(pf);
2283 ice_check_media_subtask(pf);
2284 ice_check_for_hang_subtask(pf);
2285 ice_sync_fltr_subtask(pf);
2286 ice_handle_mdd_event(pf);
2287 ice_watchdog_subtask(pf);
2289 if (ice_is_safe_mode(pf)) {
2290 ice_service_task_complete(pf);
2294 ice_process_vflr_event(pf);
2295 ice_clean_mailboxq_subtask(pf);
2296 ice_clean_sbq_subtask(pf);
2297 ice_sync_arfs_fltrs(pf);
2298 ice_flush_fdir_ctx(pf);
2300 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2301 ice_service_task_complete(pf);
2303 /* If the tasks have taken longer than one service timer period
2304 * or there is more work to be done, reset the service timer to
2305 * schedule the service task now.
2307 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2308 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2309 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2310 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2311 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2312 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2313 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2314 mod_timer(&pf->serv_tmr, jiffies);
2318 * ice_set_ctrlq_len - helper function to set controlq length
2319 * @hw: pointer to the HW instance
2321 static void ice_set_ctrlq_len(struct ice_hw *hw)
2323 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2324 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2325 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2326 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2327 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2328 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2329 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2330 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2331 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2332 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2333 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2334 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2338 * ice_schedule_reset - schedule a reset
2339 * @pf: board private structure
2340 * @reset: reset being requested
2342 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2344 struct device *dev = ice_pf_to_dev(pf);
2346 /* bail out if earlier reset has failed */
2347 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2348 dev_dbg(dev, "earlier reset has failed\n");
2351 /* bail if reset/recovery already in progress */
2352 if (ice_is_reset_in_progress(pf->state)) {
2353 dev_dbg(dev, "Reset already in progress\n");
2357 ice_unplug_aux_dev(pf);
2361 set_bit(ICE_PFR_REQ, pf->state);
2363 case ICE_RESET_CORER:
2364 set_bit(ICE_CORER_REQ, pf->state);
2366 case ICE_RESET_GLOBR:
2367 set_bit(ICE_GLOBR_REQ, pf->state);
2373 ice_service_task_schedule(pf);
2378 * ice_irq_affinity_notify - Callback for affinity changes
2379 * @notify: context as to what irq was changed
2380 * @mask: the new affinity mask
2382 * This is a callback function used by the irq_set_affinity_notifier function
2383 * so that we may register to receive changes to the irq affinity masks.
2386 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2387 const cpumask_t *mask)
2389 struct ice_q_vector *q_vector =
2390 container_of(notify, struct ice_q_vector, affinity_notify);
2392 cpumask_copy(&q_vector->affinity_mask, mask);
2396 * ice_irq_affinity_release - Callback for affinity notifier release
2397 * @ref: internal core kernel usage
2399 * This is a callback function used by the irq_set_affinity_notifier function
2400 * to inform the current notification subscriber that they will no longer
2401 * receive notifications.
2403 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2406 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2407 * @vsi: the VSI being configured
2409 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2411 struct ice_hw *hw = &vsi->back->hw;
2414 ice_for_each_q_vector(vsi, i)
2415 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2422 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2423 * @vsi: the VSI being configured
2424 * @basename: name for the vector
2426 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2428 int q_vectors = vsi->num_q_vectors;
2429 struct ice_pf *pf = vsi->back;
2430 int base = vsi->base_vector;
2437 dev = ice_pf_to_dev(pf);
2438 for (vector = 0; vector < q_vectors; vector++) {
2439 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2441 irq_num = pf->msix_entries[base + vector].vector;
2443 if (q_vector->tx.tx_ring && q_vector->rx.rx_ring) {
2444 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2445 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2447 } else if (q_vector->rx.rx_ring) {
2448 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2449 "%s-%s-%d", basename, "rx", rx_int_idx++);
2450 } else if (q_vector->tx.tx_ring) {
2451 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2452 "%s-%s-%d", basename, "tx", tx_int_idx++);
2454 /* skip this unused q_vector */
2457 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2458 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2459 IRQF_SHARED, q_vector->name,
2462 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2463 0, q_vector->name, q_vector);
2465 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2470 /* register for affinity change notifications */
2471 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2472 struct irq_affinity_notify *affinity_notify;
2474 affinity_notify = &q_vector->affinity_notify;
2475 affinity_notify->notify = ice_irq_affinity_notify;
2476 affinity_notify->release = ice_irq_affinity_release;
2477 irq_set_affinity_notifier(irq_num, affinity_notify);
2480 /* assign the mask for this irq */
2481 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2484 vsi->irqs_ready = true;
2490 irq_num = pf->msix_entries[base + vector].vector;
2491 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2492 irq_set_affinity_notifier(irq_num, NULL);
2493 irq_set_affinity_hint(irq_num, NULL);
2494 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2500 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2501 * @vsi: VSI to setup Tx rings used by XDP
2503 * Return 0 on success and negative value on error
2505 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2507 struct device *dev = ice_pf_to_dev(vsi->back);
2508 struct ice_tx_desc *tx_desc;
2511 ice_for_each_xdp_txq(vsi, i) {
2512 u16 xdp_q_idx = vsi->alloc_txq + i;
2513 struct ice_tx_ring *xdp_ring;
2515 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2518 goto free_xdp_rings;
2520 xdp_ring->q_index = xdp_q_idx;
2521 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2522 xdp_ring->vsi = vsi;
2523 xdp_ring->netdev = NULL;
2524 xdp_ring->next_dd = ICE_TX_THRESH - 1;
2525 xdp_ring->next_rs = ICE_TX_THRESH - 1;
2526 xdp_ring->dev = dev;
2527 xdp_ring->count = vsi->num_tx_desc;
2528 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2529 if (ice_setup_tx_ring(xdp_ring))
2530 goto free_xdp_rings;
2531 ice_set_ring_xdp(xdp_ring);
2532 xdp_ring->xsk_pool = ice_tx_xsk_pool(xdp_ring);
2533 spin_lock_init(&xdp_ring->tx_lock);
2534 for (j = 0; j < xdp_ring->count; j++) {
2535 tx_desc = ICE_TX_DESC(xdp_ring, j);
2536 tx_desc->cmd_type_offset_bsz = cpu_to_le64(ICE_TX_DESC_DTYPE_DESC_DONE);
2540 ice_for_each_rxq(vsi, i) {
2541 if (static_key_enabled(&ice_xdp_locking_key))
2542 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i % vsi->num_xdp_txq];
2544 vsi->rx_rings[i]->xdp_ring = vsi->xdp_rings[i];
2551 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2552 ice_free_tx_ring(vsi->xdp_rings[i]);
2557 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2558 * @vsi: VSI to set the bpf prog on
2559 * @prog: the bpf prog pointer
2561 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2563 struct bpf_prog *old_prog;
2566 old_prog = xchg(&vsi->xdp_prog, prog);
2568 bpf_prog_put(old_prog);
2570 ice_for_each_rxq(vsi, i)
2571 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2575 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2576 * @vsi: VSI to bring up Tx rings used by XDP
2577 * @prog: bpf program that will be assigned to VSI
2579 * Return 0 on success and negative value on error
2581 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2583 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2584 int xdp_rings_rem = vsi->num_xdp_txq;
2585 struct ice_pf *pf = vsi->back;
2586 struct ice_qs_cfg xdp_qs_cfg = {
2587 .qs_mutex = &pf->avail_q_mutex,
2588 .pf_map = pf->avail_txqs,
2589 .pf_map_size = pf->max_pf_txqs,
2590 .q_count = vsi->num_xdp_txq,
2591 .scatter_count = ICE_MAX_SCATTER_TXQS,
2592 .vsi_map = vsi->txq_map,
2593 .vsi_map_offset = vsi->alloc_txq,
2594 .mapping_mode = ICE_VSI_MAP_CONTIG
2600 dev = ice_pf_to_dev(pf);
2601 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2602 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2603 if (!vsi->xdp_rings)
2606 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2607 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2610 if (static_key_enabled(&ice_xdp_locking_key))
2611 netdev_warn(vsi->netdev,
2612 "Could not allocate one XDP Tx ring per CPU, XDP_TX/XDP_REDIRECT actions will be slower\n");
2614 if (ice_xdp_alloc_setup_rings(vsi))
2615 goto clear_xdp_rings;
2617 /* follow the logic from ice_vsi_map_rings_to_vectors */
2618 ice_for_each_q_vector(vsi, v_idx) {
2619 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2620 int xdp_rings_per_v, q_id, q_base;
2622 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2623 vsi->num_q_vectors - v_idx);
2624 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2626 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2627 struct ice_tx_ring *xdp_ring = vsi->xdp_rings[q_id];
2629 xdp_ring->q_vector = q_vector;
2630 xdp_ring->next = q_vector->tx.tx_ring;
2631 q_vector->tx.tx_ring = xdp_ring;
2633 xdp_rings_rem -= xdp_rings_per_v;
2636 /* omit the scheduler update if in reset path; XDP queues will be
2637 * taken into account at the end of ice_vsi_rebuild, where
2638 * ice_cfg_vsi_lan is being called
2640 if (ice_is_reset_in_progress(pf->state))
2643 /* tell the Tx scheduler that right now we have
2646 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2647 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2649 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2652 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %d\n",
2654 goto clear_xdp_rings;
2657 /* assign the prog only when it's not already present on VSI;
2658 * this flow is a subject of both ethtool -L and ndo_bpf flows;
2659 * VSI rebuild that happens under ethtool -L can expose us to
2660 * the bpf_prog refcount issues as we would be swapping same
2661 * bpf_prog pointers from vsi->xdp_prog and calling bpf_prog_put
2662 * on it as it would be treated as an 'old_prog'; for ndo_bpf
2663 * this is not harmful as dev_xdp_install bumps the refcount
2664 * before calling the op exposed by the driver;
2666 if (!ice_is_xdp_ena_vsi(vsi))
2667 ice_vsi_assign_bpf_prog(vsi, prog);
2671 ice_for_each_xdp_txq(vsi, i)
2672 if (vsi->xdp_rings[i]) {
2673 kfree_rcu(vsi->xdp_rings[i], rcu);
2674 vsi->xdp_rings[i] = NULL;
2678 mutex_lock(&pf->avail_q_mutex);
2679 ice_for_each_xdp_txq(vsi, i) {
2680 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2681 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2683 mutex_unlock(&pf->avail_q_mutex);
2685 devm_kfree(dev, vsi->xdp_rings);
2690 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2691 * @vsi: VSI to remove XDP rings
2693 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2696 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2698 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2699 struct ice_pf *pf = vsi->back;
2702 /* q_vectors are freed in reset path so there's no point in detaching
2703 * rings; in case of rebuild being triggered not from reset bits
2704 * in pf->state won't be set, so additionally check first q_vector
2707 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2710 ice_for_each_q_vector(vsi, v_idx) {
2711 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2712 struct ice_tx_ring *ring;
2714 ice_for_each_tx_ring(ring, q_vector->tx)
2715 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2718 /* restore the value of last node prior to XDP setup */
2719 q_vector->tx.tx_ring = ring;
2723 mutex_lock(&pf->avail_q_mutex);
2724 ice_for_each_xdp_txq(vsi, i) {
2725 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2726 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2728 mutex_unlock(&pf->avail_q_mutex);
2730 ice_for_each_xdp_txq(vsi, i)
2731 if (vsi->xdp_rings[i]) {
2732 if (vsi->xdp_rings[i]->desc)
2733 ice_free_tx_ring(vsi->xdp_rings[i]);
2734 kfree_rcu(vsi->xdp_rings[i], rcu);
2735 vsi->xdp_rings[i] = NULL;
2738 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2739 vsi->xdp_rings = NULL;
2741 if (static_key_enabled(&ice_xdp_locking_key))
2742 static_branch_dec(&ice_xdp_locking_key);
2744 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2747 ice_vsi_assign_bpf_prog(vsi, NULL);
2749 /* notify Tx scheduler that we destroyed XDP queues and bring
2750 * back the old number of child nodes
2752 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2753 max_txqs[i] = vsi->num_txq;
2755 /* change number of XDP Tx queues to 0 */
2756 vsi->num_xdp_txq = 0;
2758 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2763 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2764 * @vsi: VSI to schedule napi on
2766 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2770 ice_for_each_rxq(vsi, i) {
2771 struct ice_rx_ring *rx_ring = vsi->rx_rings[i];
2773 if (rx_ring->xsk_pool)
2774 napi_schedule(&rx_ring->q_vector->napi);
2779 * ice_vsi_determine_xdp_res - figure out how many Tx qs can XDP have
2780 * @vsi: VSI to determine the count of XDP Tx qs
2782 * returns 0 if Tx qs count is higher than at least half of CPU count,
2785 int ice_vsi_determine_xdp_res(struct ice_vsi *vsi)
2787 u16 avail = ice_get_avail_txq_count(vsi->back);
2788 u16 cpus = num_possible_cpus();
2790 if (avail < cpus / 2)
2793 vsi->num_xdp_txq = min_t(u16, avail, cpus);
2795 if (vsi->num_xdp_txq < cpus)
2796 static_branch_inc(&ice_xdp_locking_key);
2802 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2803 * @vsi: VSI to setup XDP for
2804 * @prog: XDP program
2805 * @extack: netlink extended ack
2808 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2809 struct netlink_ext_ack *extack)
2811 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2812 bool if_running = netif_running(vsi->netdev);
2813 int ret = 0, xdp_ring_err = 0;
2815 if (frame_size > vsi->rx_buf_len) {
2816 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2820 /* need to stop netdev while setting up the program for Rx rings */
2821 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2822 ret = ice_down(vsi);
2824 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2829 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2830 xdp_ring_err = ice_vsi_determine_xdp_res(vsi);
2832 NL_SET_ERR_MSG_MOD(extack, "Not enough Tx resources for XDP");
2834 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2836 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2838 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2839 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2841 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2843 /* safe to call even when prog == vsi->xdp_prog as
2844 * dev_xdp_install in net/core/dev.c incremented prog's
2845 * refcount so corresponding bpf_prog_put won't cause
2848 ice_vsi_assign_bpf_prog(vsi, prog);
2855 ice_vsi_rx_napi_schedule(vsi);
2857 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2861 * ice_xdp_safe_mode - XDP handler for safe mode
2865 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2866 struct netdev_bpf *xdp)
2868 NL_SET_ERR_MSG_MOD(xdp->extack,
2869 "Please provide working DDP firmware package in order to use XDP\n"
2870 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2875 * ice_xdp - implements XDP handler
2879 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2881 struct ice_netdev_priv *np = netdev_priv(dev);
2882 struct ice_vsi *vsi = np->vsi;
2884 if (vsi->type != ICE_VSI_PF) {
2885 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2889 switch (xdp->command) {
2890 case XDP_SETUP_PROG:
2891 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2892 case XDP_SETUP_XSK_POOL:
2893 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2901 * ice_ena_misc_vector - enable the non-queue interrupts
2902 * @pf: board private structure
2904 static void ice_ena_misc_vector(struct ice_pf *pf)
2906 struct ice_hw *hw = &pf->hw;
2909 /* Disable anti-spoof detection interrupt to prevent spurious event
2910 * interrupts during a function reset. Anti-spoof functionally is
2913 val = rd32(hw, GL_MDCK_TX_TDPU);
2914 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2915 wr32(hw, GL_MDCK_TX_TDPU, val);
2917 /* clear things first */
2918 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2919 rd32(hw, PFINT_OICR); /* read to clear */
2921 val = (PFINT_OICR_ECC_ERR_M |
2922 PFINT_OICR_MAL_DETECT_M |
2924 PFINT_OICR_PCI_EXCEPTION_M |
2926 PFINT_OICR_HMC_ERR_M |
2927 PFINT_OICR_PE_PUSH_M |
2928 PFINT_OICR_PE_CRITERR_M);
2930 wr32(hw, PFINT_OICR_ENA, val);
2932 /* SW_ITR_IDX = 0, but don't change INTENA */
2933 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2934 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2938 * ice_misc_intr - misc interrupt handler
2939 * @irq: interrupt number
2940 * @data: pointer to a q_vector
2942 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2944 struct ice_pf *pf = (struct ice_pf *)data;
2945 struct ice_hw *hw = &pf->hw;
2946 irqreturn_t ret = IRQ_NONE;
2950 dev = ice_pf_to_dev(pf);
2951 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2952 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2953 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
2955 oicr = rd32(hw, PFINT_OICR);
2956 ena_mask = rd32(hw, PFINT_OICR_ENA);
2958 if (oicr & PFINT_OICR_SWINT_M) {
2959 ena_mask &= ~PFINT_OICR_SWINT_M;
2963 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2964 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2965 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2967 if (oicr & PFINT_OICR_VFLR_M) {
2968 /* disable any further VFLR event notifications */
2969 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2970 u32 reg = rd32(hw, PFINT_OICR_ENA);
2972 reg &= ~PFINT_OICR_VFLR_M;
2973 wr32(hw, PFINT_OICR_ENA, reg);
2975 ena_mask &= ~PFINT_OICR_VFLR_M;
2976 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2980 if (oicr & PFINT_OICR_GRST_M) {
2983 /* we have a reset warning */
2984 ena_mask &= ~PFINT_OICR_GRST_M;
2985 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2986 GLGEN_RSTAT_RESET_TYPE_S;
2988 if (reset == ICE_RESET_CORER)
2990 else if (reset == ICE_RESET_GLOBR)
2992 else if (reset == ICE_RESET_EMPR)
2995 dev_dbg(dev, "Invalid reset type %d\n", reset);
2997 /* If a reset cycle isn't already in progress, we set a bit in
2998 * pf->state so that the service task can start a reset/rebuild.
3000 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
3001 if (reset == ICE_RESET_CORER)
3002 set_bit(ICE_CORER_RECV, pf->state);
3003 else if (reset == ICE_RESET_GLOBR)
3004 set_bit(ICE_GLOBR_RECV, pf->state);
3006 set_bit(ICE_EMPR_RECV, pf->state);
3008 /* There are couple of different bits at play here.
3009 * hw->reset_ongoing indicates whether the hardware is
3010 * in reset. This is set to true when a reset interrupt
3011 * is received and set back to false after the driver
3012 * has determined that the hardware is out of reset.
3014 * ICE_RESET_OICR_RECV in pf->state indicates
3015 * that a post reset rebuild is required before the
3016 * driver is operational again. This is set above.
3018 * As this is the start of the reset/rebuild cycle, set
3019 * both to indicate that.
3021 hw->reset_ongoing = true;
3025 if (oicr & PFINT_OICR_TSYN_TX_M) {
3026 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
3027 ice_ptp_process_ts(pf);
3030 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
3031 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
3032 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
3034 /* Save EVENTs from GTSYN register */
3035 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
3036 GLTSYN_STAT_EVENT1_M |
3037 GLTSYN_STAT_EVENT2_M);
3038 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
3039 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
3042 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
3043 if (oicr & ICE_AUX_CRIT_ERR) {
3044 struct iidc_event *event;
3046 ena_mask &= ~ICE_AUX_CRIT_ERR;
3047 event = kzalloc(sizeof(*event), GFP_ATOMIC);
3049 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
3050 /* report the entire OICR value to AUX driver */
3052 ice_send_event_to_aux(pf, event);
3057 /* Report any remaining unexpected interrupts */
3060 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
3061 /* If a critical error is pending there is no choice but to
3064 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
3065 PFINT_OICR_ECC_ERR_M)) {
3066 set_bit(ICE_PFR_REQ, pf->state);
3067 ice_service_task_schedule(pf);
3072 ice_service_task_schedule(pf);
3073 ice_irq_dynamic_ena(hw, NULL, NULL);
3079 * ice_dis_ctrlq_interrupts - disable control queue interrupts
3080 * @hw: pointer to HW structure
3082 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
3084 /* disable Admin queue Interrupt causes */
3085 wr32(hw, PFINT_FW_CTL,
3086 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
3088 /* disable Mailbox queue Interrupt causes */
3089 wr32(hw, PFINT_MBX_CTL,
3090 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
3092 wr32(hw, PFINT_SB_CTL,
3093 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
3095 /* disable Control queue Interrupt causes */
3096 wr32(hw, PFINT_OICR_CTL,
3097 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
3103 * ice_free_irq_msix_misc - Unroll misc vector setup
3104 * @pf: board private structure
3106 static void ice_free_irq_msix_misc(struct ice_pf *pf)
3108 struct ice_hw *hw = &pf->hw;
3110 ice_dis_ctrlq_interrupts(hw);
3112 /* disable OICR interrupt */
3113 wr32(hw, PFINT_OICR_ENA, 0);
3116 if (pf->msix_entries) {
3117 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
3118 devm_free_irq(ice_pf_to_dev(pf),
3119 pf->msix_entries[pf->oicr_idx].vector, pf);
3122 pf->num_avail_sw_msix += 1;
3123 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
3127 * ice_ena_ctrlq_interrupts - enable control queue interrupts
3128 * @hw: pointer to HW structure
3129 * @reg_idx: HW vector index to associate the control queue interrupts with
3131 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
3135 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
3136 PFINT_OICR_CTL_CAUSE_ENA_M);
3137 wr32(hw, PFINT_OICR_CTL, val);
3139 /* enable Admin queue Interrupt causes */
3140 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
3141 PFINT_FW_CTL_CAUSE_ENA_M);
3142 wr32(hw, PFINT_FW_CTL, val);
3144 /* enable Mailbox queue Interrupt causes */
3145 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
3146 PFINT_MBX_CTL_CAUSE_ENA_M);
3147 wr32(hw, PFINT_MBX_CTL, val);
3149 /* This enables Sideband queue Interrupt causes */
3150 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
3151 PFINT_SB_CTL_CAUSE_ENA_M);
3152 wr32(hw, PFINT_SB_CTL, val);
3158 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
3159 * @pf: board private structure
3161 * This sets up the handler for MSIX 0, which is used to manage the
3162 * non-queue interrupts, e.g. AdminQ and errors. This is not used
3163 * when in MSI or Legacy interrupt mode.
3165 static int ice_req_irq_msix_misc(struct ice_pf *pf)
3167 struct device *dev = ice_pf_to_dev(pf);
3168 struct ice_hw *hw = &pf->hw;
3169 int oicr_idx, err = 0;
3171 if (!pf->int_name[0])
3172 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
3173 dev_driver_string(dev), dev_name(dev));
3175 /* Do not request IRQ but do enable OICR interrupt since settings are
3176 * lost during reset. Note that this function is called only during
3177 * rebuild path and not while reset is in progress.
3179 if (ice_is_reset_in_progress(pf->state))
3182 /* reserve one vector in irq_tracker for misc interrupts */
3183 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3187 pf->num_avail_sw_msix -= 1;
3188 pf->oicr_idx = (u16)oicr_idx;
3190 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
3191 ice_misc_intr, 0, pf->int_name, pf);
3193 dev_err(dev, "devm_request_irq for %s failed: %d\n",
3195 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
3196 pf->num_avail_sw_msix += 1;
3201 ice_ena_misc_vector(pf);
3203 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3204 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3205 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3208 ice_irq_dynamic_ena(hw, NULL, NULL);
3214 * ice_napi_add - register NAPI handler for the VSI
3215 * @vsi: VSI for which NAPI handler is to be registered
3217 * This function is only called in the driver's load path. Registering the NAPI
3218 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3219 * reset/rebuild, etc.)
3221 static void ice_napi_add(struct ice_vsi *vsi)
3228 ice_for_each_q_vector(vsi, v_idx)
3229 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3230 ice_napi_poll, NAPI_POLL_WEIGHT);
3234 * ice_set_ops - set netdev and ethtools ops for the given netdev
3235 * @netdev: netdev instance
3237 static void ice_set_ops(struct net_device *netdev)
3239 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3241 if (ice_is_safe_mode(pf)) {
3242 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3243 ice_set_ethtool_safe_mode_ops(netdev);
3247 netdev->netdev_ops = &ice_netdev_ops;
3248 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3249 ice_set_ethtool_ops(netdev);
3253 * ice_set_netdev_features - set features for the given netdev
3254 * @netdev: netdev instance
3256 static void ice_set_netdev_features(struct net_device *netdev)
3258 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3259 netdev_features_t csumo_features;
3260 netdev_features_t vlano_features;
3261 netdev_features_t dflt_features;
3262 netdev_features_t tso_features;
3264 if (ice_is_safe_mode(pf)) {
3266 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3267 netdev->hw_features = netdev->features;
3271 dflt_features = NETIF_F_SG |
3276 csumo_features = NETIF_F_RXCSUM |
3281 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3282 NETIF_F_HW_VLAN_CTAG_TX |
3283 NETIF_F_HW_VLAN_CTAG_RX;
3285 tso_features = NETIF_F_TSO |
3289 NETIF_F_GSO_UDP_TUNNEL |
3290 NETIF_F_GSO_GRE_CSUM |
3291 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3292 NETIF_F_GSO_PARTIAL |
3293 NETIF_F_GSO_IPXIP4 |
3294 NETIF_F_GSO_IPXIP6 |
3297 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3298 NETIF_F_GSO_GRE_CSUM;
3299 /* set features that user can change */
3300 netdev->hw_features = dflt_features | csumo_features |
3301 vlano_features | tso_features;
3303 /* add support for HW_CSUM on packets with MPLS header */
3304 netdev->mpls_features = NETIF_F_HW_CSUM;
3306 /* enable features */
3307 netdev->features |= netdev->hw_features;
3309 netdev->hw_features |= NETIF_F_HW_TC;
3311 /* encap and VLAN devices inherit default, csumo and tso features */
3312 netdev->hw_enc_features |= dflt_features | csumo_features |
3314 netdev->vlan_features |= dflt_features | csumo_features |
3319 * ice_cfg_netdev - Allocate, configure and register a netdev
3320 * @vsi: the VSI associated with the new netdev
3322 * Returns 0 on success, negative value on failure
3324 static int ice_cfg_netdev(struct ice_vsi *vsi)
3326 struct ice_netdev_priv *np;
3327 struct net_device *netdev;
3328 u8 mac_addr[ETH_ALEN];
3330 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3335 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3336 vsi->netdev = netdev;
3337 np = netdev_priv(netdev);
3340 ice_set_netdev_features(netdev);
3342 ice_set_ops(netdev);
3344 if (vsi->type == ICE_VSI_PF) {
3345 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3346 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3347 eth_hw_addr_set(netdev, mac_addr);
3348 ether_addr_copy(netdev->perm_addr, mac_addr);
3351 netdev->priv_flags |= IFF_UNICAST_FLT;
3353 /* Setup netdev TC information */
3354 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3356 /* setup watchdog timeout value to be 5 second */
3357 netdev->watchdog_timeo = 5 * HZ;
3359 netdev->min_mtu = ETH_MIN_MTU;
3360 netdev->max_mtu = ICE_MAX_MTU;
3366 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3367 * @lut: Lookup table
3368 * @rss_table_size: Lookup table size
3369 * @rss_size: Range of queue number for hashing
3371 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3375 for (i = 0; i < rss_table_size; i++)
3376 lut[i] = i % rss_size;
3380 * ice_pf_vsi_setup - Set up a PF VSI
3381 * @pf: board private structure
3382 * @pi: pointer to the port_info instance
3384 * Returns pointer to the successfully allocated VSI software struct
3385 * on success, otherwise returns NULL on failure.
3387 static struct ice_vsi *
3388 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3390 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID, NULL);
3393 static struct ice_vsi *
3394 ice_chnl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi,
3395 struct ice_channel *ch)
3397 return ice_vsi_setup(pf, pi, ICE_VSI_CHNL, ICE_INVAL_VFID, ch);
3401 * ice_ctrl_vsi_setup - Set up a control VSI
3402 * @pf: board private structure
3403 * @pi: pointer to the port_info instance
3405 * Returns pointer to the successfully allocated VSI software struct
3406 * on success, otherwise returns NULL on failure.
3408 static struct ice_vsi *
3409 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3411 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID, NULL);
3415 * ice_lb_vsi_setup - Set up a loopback VSI
3416 * @pf: board private structure
3417 * @pi: pointer to the port_info instance
3419 * Returns pointer to the successfully allocated VSI software struct
3420 * on success, otherwise returns NULL on failure.
3423 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3425 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID, NULL);
3429 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3430 * @netdev: network interface to be adjusted
3431 * @proto: unused protocol
3432 * @vid: VLAN ID to be added
3434 * net_device_ops implementation for adding VLAN IDs
3437 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3440 struct ice_netdev_priv *np = netdev_priv(netdev);
3441 struct ice_vsi *vsi = np->vsi;
3444 /* VLAN 0 is added by default during load/reset */
3448 /* Enable VLAN pruning when a VLAN other than 0 is added */
3449 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3450 ret = ice_cfg_vlan_pruning(vsi, true);
3455 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3456 * packets aren't pruned by the device's internal switch on Rx
3458 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3460 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3466 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3467 * @netdev: network interface to be adjusted
3468 * @proto: unused protocol
3469 * @vid: VLAN ID to be removed
3471 * net_device_ops implementation for removing VLAN IDs
3474 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3477 struct ice_netdev_priv *np = netdev_priv(netdev);
3478 struct ice_vsi *vsi = np->vsi;
3481 /* don't allow removal of VLAN 0 */
3485 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3488 ret = ice_vsi_kill_vlan(vsi, vid);
3492 /* Disable pruning when VLAN 0 is the only VLAN rule */
3493 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3494 ret = ice_cfg_vlan_pruning(vsi, false);
3496 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3501 * ice_rep_indr_tc_block_unbind
3502 * @cb_priv: indirection block private data
3504 static void ice_rep_indr_tc_block_unbind(void *cb_priv)
3506 struct ice_indr_block_priv *indr_priv = cb_priv;
3508 list_del(&indr_priv->list);
3513 * ice_tc_indir_block_unregister - Unregister TC indirect block notifications
3514 * @vsi: VSI struct which has the netdev
3516 static void ice_tc_indir_block_unregister(struct ice_vsi *vsi)
3518 struct ice_netdev_priv *np = netdev_priv(vsi->netdev);
3520 flow_indr_dev_unregister(ice_indr_setup_tc_cb, np,
3521 ice_rep_indr_tc_block_unbind);
3525 * ice_tc_indir_block_remove - clean indirect TC block notifications
3528 static void ice_tc_indir_block_remove(struct ice_pf *pf)
3530 struct ice_vsi *pf_vsi = ice_get_main_vsi(pf);
3535 ice_tc_indir_block_unregister(pf_vsi);
3539 * ice_tc_indir_block_register - Register TC indirect block notifications
3540 * @vsi: VSI struct which has the netdev
3542 * Returns 0 on success, negative value on failure
3544 static int ice_tc_indir_block_register(struct ice_vsi *vsi)
3546 struct ice_netdev_priv *np;
3548 if (!vsi || !vsi->netdev)
3551 np = netdev_priv(vsi->netdev);
3553 INIT_LIST_HEAD(&np->tc_indr_block_priv_list);
3554 return flow_indr_dev_register(ice_indr_setup_tc_cb, np);
3558 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3559 * @pf: board private structure
3561 * Returns 0 on success, negative value on failure
3563 static int ice_setup_pf_sw(struct ice_pf *pf)
3565 struct device *dev = ice_pf_to_dev(pf);
3566 struct ice_vsi *vsi;
3569 if (ice_is_reset_in_progress(pf->state))
3572 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3576 /* init channel list */
3577 INIT_LIST_HEAD(&vsi->ch_list);
3579 status = ice_cfg_netdev(vsi);
3581 goto unroll_vsi_setup;
3582 /* netdev has to be configured before setting frame size */
3583 ice_vsi_cfg_frame_size(vsi);
3585 /* init indirect block notifications */
3586 status = ice_tc_indir_block_register(vsi);
3588 dev_err(dev, "Failed to register netdev notifier\n");
3589 goto unroll_cfg_netdev;
3592 /* Setup DCB netlink interface */
3593 ice_dcbnl_setup(vsi);
3595 /* registering the NAPI handler requires both the queues and
3596 * netdev to be created, which are done in ice_pf_vsi_setup()
3597 * and ice_cfg_netdev() respectively
3601 status = ice_set_cpu_rx_rmap(vsi);
3603 dev_err(dev, "Failed to set CPU Rx map VSI %d error %d\n",
3604 vsi->vsi_num, status);
3605 goto unroll_napi_add;
3607 status = ice_init_mac_fltr(pf);
3609 goto free_cpu_rx_map;
3614 ice_free_cpu_rx_rmap(vsi);
3616 ice_tc_indir_block_unregister(vsi);
3621 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3622 free_netdev(vsi->netdev);
3628 ice_vsi_release(vsi);
3633 * ice_get_avail_q_count - Get count of queues in use
3634 * @pf_qmap: bitmap to get queue use count from
3635 * @lock: pointer to a mutex that protects access to pf_qmap
3636 * @size: size of the bitmap
3639 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3645 for_each_clear_bit(bit, pf_qmap, size)
3653 * ice_get_avail_txq_count - Get count of Tx queues in use
3654 * @pf: pointer to an ice_pf instance
3656 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3658 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3663 * ice_get_avail_rxq_count - Get count of Rx queues in use
3664 * @pf: pointer to an ice_pf instance
3666 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3668 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3673 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3674 * @pf: board private structure to initialize
3676 static void ice_deinit_pf(struct ice_pf *pf)
3678 ice_service_task_stop(pf);
3679 mutex_destroy(&pf->sw_mutex);
3680 mutex_destroy(&pf->tc_mutex);
3681 mutex_destroy(&pf->avail_q_mutex);
3683 if (pf->avail_txqs) {
3684 bitmap_free(pf->avail_txqs);
3685 pf->avail_txqs = NULL;
3688 if (pf->avail_rxqs) {
3689 bitmap_free(pf->avail_rxqs);
3690 pf->avail_rxqs = NULL;
3694 ptp_clock_unregister(pf->ptp.clock);
3698 * ice_set_pf_caps - set PFs capability flags
3699 * @pf: pointer to the PF instance
3701 static void ice_set_pf_caps(struct ice_pf *pf)
3703 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3705 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3706 clear_bit(ICE_FLAG_AUX_ENA, pf->flags);
3707 if (func_caps->common_cap.rdma) {
3708 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3709 set_bit(ICE_FLAG_AUX_ENA, pf->flags);
3711 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3712 if (func_caps->common_cap.dcb)
3713 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3714 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3715 if (func_caps->common_cap.sr_iov_1_1) {
3716 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3717 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3720 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3721 if (func_caps->common_cap.rss_table_size)
3722 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3724 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3725 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3728 /* ctrl_vsi_idx will be set to a valid value when flow director
3729 * is setup by ice_init_fdir
3731 pf->ctrl_vsi_idx = ICE_NO_VSI;
3732 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3733 /* force guaranteed filter pool for PF */
3734 ice_alloc_fd_guar_item(&pf->hw, &unused,
3735 func_caps->fd_fltr_guar);
3736 /* force shared filter pool for PF */
3737 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3738 func_caps->fd_fltr_best_effort);
3741 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3742 if (func_caps->common_cap.ieee_1588)
3743 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3745 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3746 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3750 * ice_init_pf - Initialize general software structures (struct ice_pf)
3751 * @pf: board private structure to initialize
3753 static int ice_init_pf(struct ice_pf *pf)
3755 ice_set_pf_caps(pf);
3757 mutex_init(&pf->sw_mutex);
3758 mutex_init(&pf->tc_mutex);
3760 INIT_HLIST_HEAD(&pf->aq_wait_list);
3761 spin_lock_init(&pf->aq_wait_lock);
3762 init_waitqueue_head(&pf->aq_wait_queue);
3764 init_waitqueue_head(&pf->reset_wait_queue);
3766 /* setup service timer and periodic service task */
3767 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3768 pf->serv_tmr_period = HZ;
3769 INIT_WORK(&pf->serv_task, ice_service_task);
3770 clear_bit(ICE_SERVICE_SCHED, pf->state);
3772 mutex_init(&pf->avail_q_mutex);
3773 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3774 if (!pf->avail_txqs)
3777 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3778 if (!pf->avail_rxqs) {
3779 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3780 pf->avail_txqs = NULL;
3788 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3789 * @pf: board private structure
3791 * compute the number of MSIX vectors required (v_budget) and request from
3792 * the OS. Return the number of vectors reserved or negative on failure
3794 static int ice_ena_msix_range(struct ice_pf *pf)
3796 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3797 struct device *dev = ice_pf_to_dev(pf);
3800 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3801 num_cpus = num_online_cpus();
3803 /* reserve for LAN miscellaneous handler */
3804 needed = ICE_MIN_LAN_OICR_MSIX;
3805 if (v_left < needed)
3806 goto no_hw_vecs_left_err;
3810 /* reserve for flow director */
3811 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3812 needed = ICE_FDIR_MSIX;
3813 if (v_left < needed)
3814 goto no_hw_vecs_left_err;
3819 /* reserve for switchdev */
3820 needed = ICE_ESWITCH_MSIX;
3821 if (v_left < needed)
3822 goto no_hw_vecs_left_err;
3826 /* total used for non-traffic vectors */
3829 /* reserve vectors for LAN traffic */
3831 if (v_left < needed)
3832 goto no_hw_vecs_left_err;
3833 pf->num_lan_msix = needed;
3837 /* reserve vectors for RDMA auxiliary driver */
3838 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3839 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3840 if (v_left < needed)
3841 goto no_hw_vecs_left_err;
3842 pf->num_rdma_msix = needed;
3847 pf->msix_entries = devm_kcalloc(dev, v_budget,
3848 sizeof(*pf->msix_entries), GFP_KERNEL);
3849 if (!pf->msix_entries) {
3854 for (i = 0; i < v_budget; i++)
3855 pf->msix_entries[i].entry = i;
3857 /* actually reserve the vectors */
3858 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3859 ICE_MIN_MSIX, v_budget);
3861 dev_err(dev, "unable to reserve MSI-X vectors\n");
3866 if (v_actual < v_budget) {
3867 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3868 v_budget, v_actual);
3870 if (v_actual < ICE_MIN_MSIX) {
3871 /* error if we can't get minimum vectors */
3872 pci_disable_msix(pf->pdev);
3876 int v_remain = v_actual - v_other;
3877 int v_rdma = 0, v_min_rdma = 0;
3879 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3880 /* Need at least 1 interrupt in addition to
3883 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3884 v_min_rdma = ICE_MIN_RDMA_MSIX;
3887 if (v_actual == ICE_MIN_MSIX ||
3888 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
3889 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
3890 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3892 pf->num_rdma_msix = 0;
3893 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3894 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3895 (v_remain - v_rdma < v_rdma)) {
3896 /* Support minimum RDMA and give remaining
3897 * vectors to LAN MSIX
3899 pf->num_rdma_msix = v_min_rdma;
3900 pf->num_lan_msix = v_remain - v_min_rdma;
3902 /* Split remaining MSIX with RDMA after
3903 * accounting for AEQ MSIX
3905 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3906 ICE_RDMA_NUM_AEQ_MSIX;
3907 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3910 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3913 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
3914 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
3922 devm_kfree(dev, pf->msix_entries);
3925 no_hw_vecs_left_err:
3926 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3930 pf->num_rdma_msix = 0;
3931 pf->num_lan_msix = 0;
3936 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3937 * @pf: board private structure
3939 static void ice_dis_msix(struct ice_pf *pf)
3941 pci_disable_msix(pf->pdev);
3942 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3943 pf->msix_entries = NULL;
3947 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3948 * @pf: board private structure
3950 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3954 if (pf->irq_tracker) {
3955 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3956 pf->irq_tracker = NULL;
3961 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3962 * @pf: board private structure to initialize
3964 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3968 vectors = ice_ena_msix_range(pf);
3973 /* set up vector assignment tracking */
3974 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3975 struct_size(pf->irq_tracker, list, vectors),
3977 if (!pf->irq_tracker) {
3982 /* populate SW interrupts pool with number of OS granted IRQs. */
3983 pf->num_avail_sw_msix = (u16)vectors;
3984 pf->irq_tracker->num_entries = (u16)vectors;
3985 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3991 * ice_is_wol_supported - check if WoL is supported
3992 * @hw: pointer to hardware info
3994 * Check if WoL is supported based on the HW configuration.
3995 * Returns true if NVM supports and enables WoL for this port, false otherwise
3997 bool ice_is_wol_supported(struct ice_hw *hw)
4001 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
4002 * word) indicates WoL is not supported on the corresponding PF ID.
4004 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
4007 return !(BIT(hw->port_info->lport) & wol_ctrl);
4011 * ice_vsi_recfg_qs - Change the number of queues on a VSI
4012 * @vsi: VSI being changed
4013 * @new_rx: new number of Rx queues
4014 * @new_tx: new number of Tx queues
4016 * Only change the number of queues if new_tx, or new_rx is non-0.
4018 * Returns 0 on success.
4020 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
4022 struct ice_pf *pf = vsi->back;
4023 int err = 0, timeout = 50;
4025 if (!new_rx && !new_tx)
4028 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
4032 usleep_range(1000, 2000);
4036 vsi->req_txq = (u16)new_tx;
4038 vsi->req_rxq = (u16)new_rx;
4040 /* set for the next time the netdev is started */
4041 if (!netif_running(vsi->netdev)) {
4042 ice_vsi_rebuild(vsi, false);
4043 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
4048 ice_vsi_rebuild(vsi, false);
4049 ice_pf_dcb_recfg(pf);
4052 clear_bit(ICE_CFG_BUSY, pf->state);
4057 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
4058 * @pf: PF to configure
4060 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
4061 * VSI can still Tx/Rx VLAN tagged packets.
4063 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
4065 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4066 struct ice_vsi_ctx *ctxt;
4073 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
4078 ctxt->info = vsi->info;
4080 ctxt->info.valid_sections =
4081 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
4082 ICE_AQ_VSI_PROP_SECURITY_VALID |
4083 ICE_AQ_VSI_PROP_SW_VALID);
4085 /* disable VLAN anti-spoof */
4086 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
4087 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
4089 /* disable VLAN pruning and keep all other settings */
4090 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
4092 /* allow all VLANs on Tx and don't strip on Rx */
4093 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
4094 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
4096 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
4098 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %d aq_err %s\n",
4099 status, ice_aq_str(hw->adminq.sq_last_status));
4101 vsi->info.sec_flags = ctxt->info.sec_flags;
4102 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
4103 vsi->info.vlan_flags = ctxt->info.vlan_flags;
4110 * ice_log_pkg_init - log result of DDP package load
4111 * @hw: pointer to hardware info
4112 * @state: state of package load
4114 static void ice_log_pkg_init(struct ice_hw *hw, enum ice_ddp_state state)
4116 struct ice_pf *pf = hw->back;
4119 dev = ice_pf_to_dev(pf);
4122 case ICE_DDP_PKG_SUCCESS:
4123 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
4124 hw->active_pkg_name,
4125 hw->active_pkg_ver.major,
4126 hw->active_pkg_ver.minor,
4127 hw->active_pkg_ver.update,
4128 hw->active_pkg_ver.draft);
4130 case ICE_DDP_PKG_SAME_VERSION_ALREADY_LOADED:
4131 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
4132 hw->active_pkg_name,
4133 hw->active_pkg_ver.major,
4134 hw->active_pkg_ver.minor,
4135 hw->active_pkg_ver.update,
4136 hw->active_pkg_ver.draft);
4138 case ICE_DDP_PKG_ALREADY_LOADED_NOT_SUPPORTED:
4139 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",
4140 hw->active_pkg_name,
4141 hw->active_pkg_ver.major,
4142 hw->active_pkg_ver.minor,
4143 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4145 case ICE_DDP_PKG_COMPATIBLE_ALREADY_LOADED:
4146 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",
4147 hw->active_pkg_name,
4148 hw->active_pkg_ver.major,
4149 hw->active_pkg_ver.minor,
4150 hw->active_pkg_ver.update,
4151 hw->active_pkg_ver.draft,
4158 case ICE_DDP_PKG_FW_MISMATCH:
4159 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");
4161 case ICE_DDP_PKG_INVALID_FILE:
4162 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
4164 case ICE_DDP_PKG_FILE_VERSION_TOO_HIGH:
4165 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
4167 case ICE_DDP_PKG_FILE_VERSION_TOO_LOW:
4168 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",
4169 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
4171 case ICE_DDP_PKG_FILE_SIGNATURE_INVALID:
4172 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");
4174 case ICE_DDP_PKG_FILE_REVISION_TOO_LOW:
4175 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");
4177 case ICE_DDP_PKG_LOAD_ERROR:
4178 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
4179 /* poll for reset to complete */
4180 if (ice_check_reset(hw))
4181 dev_err(dev, "Error resetting device. Please reload the driver\n");
4183 case ICE_DDP_PKG_ERR:
4185 dev_err(dev, "An unknown error occurred when loading the DDP package. Entering Safe Mode.\n");
4191 * ice_load_pkg - load/reload the DDP Package file
4192 * @firmware: firmware structure when firmware requested or NULL for reload
4193 * @pf: pointer to the PF instance
4195 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
4196 * initialize HW tables.
4199 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
4201 enum ice_ddp_state state = ICE_DDP_PKG_ERR;
4202 struct device *dev = ice_pf_to_dev(pf);
4203 struct ice_hw *hw = &pf->hw;
4205 /* Load DDP Package */
4206 if (firmware && !hw->pkg_copy) {
4207 state = ice_copy_and_init_pkg(hw, firmware->data,
4209 ice_log_pkg_init(hw, state);
4210 } else if (!firmware && hw->pkg_copy) {
4211 /* Reload package during rebuild after CORER/GLOBR reset */
4212 state = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
4213 ice_log_pkg_init(hw, state);
4215 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
4218 if (!ice_is_init_pkg_successful(state)) {
4220 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4224 /* Successful download package is the precondition for advanced
4225 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
4227 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
4231 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
4232 * @pf: pointer to the PF structure
4234 * There is no error returned here because the driver should be able to handle
4235 * 128 Byte cache lines, so we only print a warning in case issues are seen,
4236 * specifically with Tx.
4238 static void ice_verify_cacheline_size(struct ice_pf *pf)
4240 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
4241 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
4242 ICE_CACHE_LINE_BYTES);
4246 * ice_send_version - update firmware with driver version
4249 * Returns 0 on success, else error code
4251 static int ice_send_version(struct ice_pf *pf)
4253 struct ice_driver_ver dv;
4255 dv.major_ver = 0xff;
4256 dv.minor_ver = 0xff;
4257 dv.build_ver = 0xff;
4258 dv.subbuild_ver = 0;
4259 strscpy((char *)dv.driver_string, UTS_RELEASE,
4260 sizeof(dv.driver_string));
4261 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4265 * ice_init_fdir - Initialize flow director VSI and configuration
4266 * @pf: pointer to the PF instance
4268 * returns 0 on success, negative on error
4270 static int ice_init_fdir(struct ice_pf *pf)
4272 struct device *dev = ice_pf_to_dev(pf);
4273 struct ice_vsi *ctrl_vsi;
4276 /* Side Band Flow Director needs to have a control VSI.
4277 * Allocate it and store it in the PF.
4279 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4281 dev_dbg(dev, "could not create control VSI\n");
4285 err = ice_vsi_open_ctrl(ctrl_vsi);
4287 dev_dbg(dev, "could not open control VSI\n");
4291 mutex_init(&pf->hw.fdir_fltr_lock);
4293 err = ice_fdir_create_dflt_rules(pf);
4300 ice_fdir_release_flows(&pf->hw);
4301 ice_vsi_close(ctrl_vsi);
4303 ice_vsi_release(ctrl_vsi);
4304 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4305 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4306 pf->ctrl_vsi_idx = ICE_NO_VSI;
4312 * ice_get_opt_fw_name - return optional firmware file name or NULL
4313 * @pf: pointer to the PF instance
4315 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4317 /* Optional firmware name same as default with additional dash
4318 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4320 struct pci_dev *pdev = pf->pdev;
4321 char *opt_fw_filename;
4324 /* Determine the name of the optional file using the DSN (two
4325 * dwords following the start of the DSN Capability).
4327 dsn = pci_get_dsn(pdev);
4331 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4332 if (!opt_fw_filename)
4335 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4336 ICE_DDP_PKG_PATH, dsn);
4338 return opt_fw_filename;
4342 * ice_request_fw - Device initialization routine
4343 * @pf: pointer to the PF instance
4345 static void ice_request_fw(struct ice_pf *pf)
4347 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4348 const struct firmware *firmware = NULL;
4349 struct device *dev = ice_pf_to_dev(pf);
4352 /* optional device-specific DDP (if present) overrides the default DDP
4353 * package file. kernel logs a debug message if the file doesn't exist,
4354 * and warning messages for other errors.
4356 if (opt_fw_filename) {
4357 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4359 kfree(opt_fw_filename);
4363 /* request for firmware was successful. Download to device */
4364 ice_load_pkg(firmware, pf);
4365 kfree(opt_fw_filename);
4366 release_firmware(firmware);
4371 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4373 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4377 /* request for firmware was successful. Download to device */
4378 ice_load_pkg(firmware, pf);
4379 release_firmware(firmware);
4383 * ice_print_wake_reason - show the wake up cause in the log
4384 * @pf: pointer to the PF struct
4386 static void ice_print_wake_reason(struct ice_pf *pf)
4388 u32 wus = pf->wakeup_reason;
4389 const char *wake_str;
4391 /* if no wake event, nothing to print */
4395 if (wus & PFPM_WUS_LNKC_M)
4396 wake_str = "Link\n";
4397 else if (wus & PFPM_WUS_MAG_M)
4398 wake_str = "Magic Packet\n";
4399 else if (wus & PFPM_WUS_MNG_M)
4400 wake_str = "Management\n";
4401 else if (wus & PFPM_WUS_FW_RST_WK_M)
4402 wake_str = "Firmware Reset\n";
4404 wake_str = "Unknown\n";
4406 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4410 * ice_register_netdev - register netdev and devlink port
4411 * @pf: pointer to the PF struct
4413 static int ice_register_netdev(struct ice_pf *pf)
4415 struct ice_vsi *vsi;
4418 vsi = ice_get_main_vsi(pf);
4419 if (!vsi || !vsi->netdev)
4422 err = register_netdev(vsi->netdev);
4424 goto err_register_netdev;
4426 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4427 netif_carrier_off(vsi->netdev);
4428 netif_tx_stop_all_queues(vsi->netdev);
4429 err = ice_devlink_create_pf_port(pf);
4431 goto err_devlink_create;
4433 devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
4437 unregister_netdev(vsi->netdev);
4438 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4439 err_register_netdev:
4440 free_netdev(vsi->netdev);
4442 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4447 * ice_probe - Device initialization routine
4448 * @pdev: PCI device information struct
4449 * @ent: entry in ice_pci_tbl
4451 * Returns 0 on success, negative on failure
4454 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4456 struct device *dev = &pdev->dev;
4461 if (pdev->is_virtfn) {
4462 dev_err(dev, "can't probe a virtual function\n");
4466 /* this driver uses devres, see
4467 * Documentation/driver-api/driver-model/devres.rst
4469 err = pcim_enable_device(pdev);
4473 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4475 dev_err(dev, "BAR0 I/O map error %d\n", err);
4479 pf = ice_allocate_pf(dev);
4483 /* initialize Auxiliary index to invalid value */
4486 /* set up for high or low DMA */
4487 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4489 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4491 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4495 pci_enable_pcie_error_reporting(pdev);
4496 pci_set_master(pdev);
4499 pci_set_drvdata(pdev, pf);
4500 set_bit(ICE_DOWN, pf->state);
4501 /* Disable service task until DOWN bit is cleared */
4502 set_bit(ICE_SERVICE_DIS, pf->state);
4505 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4506 pci_save_state(pdev);
4509 hw->vendor_id = pdev->vendor;
4510 hw->device_id = pdev->device;
4511 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4512 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4513 hw->subsystem_device_id = pdev->subsystem_device;
4514 hw->bus.device = PCI_SLOT(pdev->devfn);
4515 hw->bus.func = PCI_FUNC(pdev->devfn);
4516 ice_set_ctrlq_len(hw);
4518 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4520 #ifndef CONFIG_DYNAMIC_DEBUG
4522 hw->debug_mask = debug;
4525 err = ice_init_hw(hw);
4527 dev_err(dev, "ice_init_hw failed: %d\n", err);
4529 goto err_exit_unroll;
4532 ice_init_feature_support(pf);
4536 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4537 * set in pf->state, which will cause ice_is_safe_mode to return
4540 if (ice_is_safe_mode(pf)) {
4541 /* we already got function/device capabilities but these don't
4542 * reflect what the driver needs to do in safe mode. Instead of
4543 * adding conditional logic everywhere to ignore these
4544 * device/function capabilities, override them.
4546 ice_set_safe_mode_caps(hw);
4549 err = ice_init_pf(pf);
4551 dev_err(dev, "ice_init_pf failed: %d\n", err);
4552 goto err_init_pf_unroll;
4555 ice_devlink_init_regions(pf);
4557 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4558 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4559 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4560 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4562 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4563 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4564 pf->hw.tnl.valid_count[TNL_VXLAN];
4565 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4566 UDP_TUNNEL_TYPE_VXLAN;
4569 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4570 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4571 pf->hw.tnl.valid_count[TNL_GENEVE];
4572 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4573 UDP_TUNNEL_TYPE_GENEVE;
4577 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4578 if (!pf->num_alloc_vsi) {
4580 goto err_init_pf_unroll;
4582 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4583 dev_warn(&pf->pdev->dev,
4584 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4585 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4586 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4589 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4593 goto err_init_pf_unroll;
4596 err = ice_init_interrupt_scheme(pf);
4598 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4600 goto err_init_vsi_unroll;
4603 /* In case of MSIX we are going to setup the misc vector right here
4604 * to handle admin queue events etc. In case of legacy and MSI
4605 * the misc functionality and queue processing is combined in
4606 * the same vector and that gets setup at open.
4608 err = ice_req_irq_msix_misc(pf);
4610 dev_err(dev, "setup of misc vector failed: %d\n", err);
4611 goto err_init_interrupt_unroll;
4614 /* create switch struct for the switch element created by FW on boot */
4615 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4616 if (!pf->first_sw) {
4618 goto err_msix_misc_unroll;
4622 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4624 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4626 pf->first_sw->pf = pf;
4628 /* record the sw_id available for later use */
4629 pf->first_sw->sw_id = hw->port_info->sw_id;
4631 err = ice_setup_pf_sw(pf);
4633 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4634 goto err_alloc_sw_unroll;
4637 clear_bit(ICE_SERVICE_DIS, pf->state);
4639 /* tell the firmware we are up */
4640 err = ice_send_version(pf);
4642 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4644 goto err_send_version_unroll;
4647 /* since everything is good, start the service timer */
4648 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4650 err = ice_init_link_events(pf->hw.port_info);
4652 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4653 goto err_send_version_unroll;
4656 /* not a fatal error if this fails */
4657 err = ice_init_nvm_phy_type(pf->hw.port_info);
4659 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4661 /* not a fatal error if this fails */
4662 err = ice_update_link_info(pf->hw.port_info);
4664 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4666 ice_init_link_dflt_override(pf->hw.port_info);
4668 ice_check_link_cfg_err(pf,
4669 pf->hw.port_info->phy.link_info.link_cfg_err);
4671 /* if media available, initialize PHY settings */
4672 if (pf->hw.port_info->phy.link_info.link_info &
4673 ICE_AQ_MEDIA_AVAILABLE) {
4674 /* not a fatal error if this fails */
4675 err = ice_init_phy_user_cfg(pf->hw.port_info);
4677 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4679 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4680 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4683 ice_configure_phy(vsi);
4686 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4689 ice_verify_cacheline_size(pf);
4691 /* Save wakeup reason register for later use */
4692 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4694 /* check for a power management event */
4695 ice_print_wake_reason(pf);
4697 /* clear wake status, all bits */
4698 wr32(hw, PFPM_WUS, U32_MAX);
4700 /* Disable WoL at init, wait for user to enable */
4701 device_set_wakeup_enable(dev, false);
4703 if (ice_is_safe_mode(pf)) {
4704 ice_set_safe_mode_vlan_cfg(pf);
4708 /* initialize DDP driven features */
4709 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4712 /* Note: Flow director init failure is non-fatal to load */
4713 if (ice_init_fdir(pf))
4714 dev_err(dev, "could not initialize flow director\n");
4716 /* Note: DCB init failure is non-fatal to load */
4717 if (ice_init_pf_dcb(pf, false)) {
4718 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4719 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4721 ice_cfg_lldp_mib_change(&pf->hw, true);
4724 if (ice_init_lag(pf))
4725 dev_warn(dev, "Failed to init link aggregation support\n");
4727 /* print PCI link speed and width */
4728 pcie_print_link_status(pf->pdev);
4731 err = ice_register_netdev(pf);
4733 goto err_netdev_reg;
4735 err = ice_devlink_register_params(pf);
4737 goto err_netdev_reg;
4739 /* ready to go, so clear down state bit */
4740 clear_bit(ICE_DOWN, pf->state);
4741 if (ice_is_aux_ena(pf)) {
4742 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4743 if (pf->aux_idx < 0) {
4744 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4746 goto err_devlink_reg_param;
4749 err = ice_init_rdma(pf);
4751 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4753 goto err_init_aux_unroll;
4756 dev_warn(dev, "RDMA is not supported on this device\n");
4759 ice_devlink_register(pf);
4762 err_init_aux_unroll:
4764 ida_free(&ice_aux_ida, pf->aux_idx);
4765 err_devlink_reg_param:
4766 ice_devlink_unregister_params(pf);
4768 err_send_version_unroll:
4769 ice_vsi_release_all(pf);
4770 err_alloc_sw_unroll:
4771 set_bit(ICE_SERVICE_DIS, pf->state);
4772 set_bit(ICE_DOWN, pf->state);
4773 devm_kfree(dev, pf->first_sw);
4774 err_msix_misc_unroll:
4775 ice_free_irq_msix_misc(pf);
4776 err_init_interrupt_unroll:
4777 ice_clear_interrupt_scheme(pf);
4778 err_init_vsi_unroll:
4779 devm_kfree(dev, pf->vsi);
4782 ice_devlink_destroy_regions(pf);
4785 pci_disable_pcie_error_reporting(pdev);
4786 pci_disable_device(pdev);
4791 * ice_set_wake - enable or disable Wake on LAN
4792 * @pf: pointer to the PF struct
4794 * Simple helper for WoL control
4796 static void ice_set_wake(struct ice_pf *pf)
4798 struct ice_hw *hw = &pf->hw;
4799 bool wol = pf->wol_ena;
4801 /* clear wake state, otherwise new wake events won't fire */
4802 wr32(hw, PFPM_WUS, U32_MAX);
4804 /* enable / disable APM wake up, no RMW needed */
4805 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4807 /* set magic packet filter enabled */
4808 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4812 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4813 * @pf: pointer to the PF struct
4815 * Issue firmware command to enable multicast magic wake, making
4816 * sure that any locally administered address (LAA) is used for
4817 * wake, and that PF reset doesn't undo the LAA.
4819 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4821 struct device *dev = ice_pf_to_dev(pf);
4822 struct ice_hw *hw = &pf->hw;
4823 u8 mac_addr[ETH_ALEN];
4824 struct ice_vsi *vsi;
4831 vsi = ice_get_main_vsi(pf);
4835 /* Get current MAC address in case it's an LAA */
4837 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4839 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4841 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4842 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4843 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4845 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4847 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %d aq_err %s\n",
4848 status, ice_aq_str(hw->adminq.sq_last_status));
4852 * ice_remove - Device removal routine
4853 * @pdev: PCI device information struct
4855 static void ice_remove(struct pci_dev *pdev)
4857 struct ice_pf *pf = pci_get_drvdata(pdev);
4860 ice_devlink_unregister(pf);
4861 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4862 if (!ice_is_reset_in_progress(pf->state))
4867 ice_tc_indir_block_remove(pf);
4869 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4870 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4874 ice_service_task_stop(pf);
4876 ice_aq_cancel_waiting_tasks(pf);
4877 ice_unplug_aux_dev(pf);
4878 if (pf->aux_idx >= 0)
4879 ida_free(&ice_aux_ida, pf->aux_idx);
4880 ice_devlink_unregister_params(pf);
4881 set_bit(ICE_DOWN, pf->state);
4884 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4885 ice_ptp_release(pf);
4886 if (!ice_is_safe_mode(pf))
4887 ice_remove_arfs(pf);
4888 ice_setup_mc_magic_wake(pf);
4889 ice_vsi_release_all(pf);
4890 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4892 ice_free_irq_msix_misc(pf);
4893 ice_for_each_vsi(pf, i) {
4896 ice_vsi_free_q_vectors(pf->vsi[i]);
4899 ice_devlink_destroy_regions(pf);
4900 ice_deinit_hw(&pf->hw);
4902 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4903 * do it via ice_schedule_reset() since there is no need to rebuild
4904 * and the service task is already stopped.
4906 ice_reset(&pf->hw, ICE_RESET_PFR);
4907 pci_wait_for_pending_transaction(pdev);
4908 ice_clear_interrupt_scheme(pf);
4909 pci_disable_pcie_error_reporting(pdev);
4910 pci_disable_device(pdev);
4914 * ice_shutdown - PCI callback for shutting down device
4915 * @pdev: PCI device information struct
4917 static void ice_shutdown(struct pci_dev *pdev)
4919 struct ice_pf *pf = pci_get_drvdata(pdev);
4923 if (system_state == SYSTEM_POWER_OFF) {
4924 pci_wake_from_d3(pdev, pf->wol_ena);
4925 pci_set_power_state(pdev, PCI_D3hot);
4931 * ice_prepare_for_shutdown - prep for PCI shutdown
4932 * @pf: board private structure
4934 * Inform or close all dependent features in prep for PCI device shutdown
4936 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4938 struct ice_hw *hw = &pf->hw;
4941 /* Notify VFs of impending reset */
4942 if (ice_check_sq_alive(hw, &hw->mailboxq))
4943 ice_vc_notify_reset(pf);
4945 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4947 /* disable the VSIs and their queues that are not already DOWN */
4948 ice_pf_dis_all_vsi(pf, false);
4950 ice_for_each_vsi(pf, v)
4952 pf->vsi[v]->vsi_num = 0;
4954 ice_shutdown_all_ctrlq(hw);
4958 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4959 * @pf: board private structure to reinitialize
4961 * This routine reinitialize interrupt scheme that was cleared during
4962 * power management suspend callback.
4964 * This should be called during resume routine to re-allocate the q_vectors
4965 * and reacquire interrupts.
4967 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4969 struct device *dev = ice_pf_to_dev(pf);
4972 /* Since we clear MSIX flag during suspend, we need to
4973 * set it back during resume...
4976 ret = ice_init_interrupt_scheme(pf);
4978 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4982 /* Remap vectors and rings, after successful re-init interrupts */
4983 ice_for_each_vsi(pf, v) {
4987 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4990 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4993 ret = ice_req_irq_msix_misc(pf);
4995 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
5005 ice_vsi_free_q_vectors(pf->vsi[v]);
5012 * @dev: generic device information structure
5014 * Power Management callback to quiesce the device and prepare
5015 * for D3 transition.
5017 static int __maybe_unused ice_suspend(struct device *dev)
5019 struct pci_dev *pdev = to_pci_dev(dev);
5023 pf = pci_get_drvdata(pdev);
5025 if (!ice_pf_state_is_nominal(pf)) {
5026 dev_err(dev, "Device is not ready, no need to suspend it\n");
5030 /* Stop watchdog tasks until resume completion.
5031 * Even though it is most likely that the service task is
5032 * disabled if the device is suspended or down, the service task's
5033 * state is controlled by a different state bit, and we should
5034 * store and honor whatever state that bit is in at this point.
5036 disabled = ice_service_task_stop(pf);
5038 ice_unplug_aux_dev(pf);
5040 /* Already suspended?, then there is nothing to do */
5041 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
5043 ice_service_task_restart(pf);
5047 if (test_bit(ICE_DOWN, pf->state) ||
5048 ice_is_reset_in_progress(pf->state)) {
5049 dev_err(dev, "can't suspend device in reset or already down\n");
5051 ice_service_task_restart(pf);
5055 ice_setup_mc_magic_wake(pf);
5057 ice_prepare_for_shutdown(pf);
5061 /* Free vectors, clear the interrupt scheme and release IRQs
5062 * for proper hibernation, especially with large number of CPUs.
5063 * Otherwise hibernation might fail when mapping all the vectors back
5066 ice_free_irq_msix_misc(pf);
5067 ice_for_each_vsi(pf, v) {
5070 ice_vsi_free_q_vectors(pf->vsi[v]);
5072 ice_free_cpu_rx_rmap(ice_get_main_vsi(pf));
5073 ice_clear_interrupt_scheme(pf);
5075 pci_save_state(pdev);
5076 pci_wake_from_d3(pdev, pf->wol_ena);
5077 pci_set_power_state(pdev, PCI_D3hot);
5082 * ice_resume - PM callback for waking up from D3
5083 * @dev: generic device information structure
5085 static int __maybe_unused ice_resume(struct device *dev)
5087 struct pci_dev *pdev = to_pci_dev(dev);
5088 enum ice_reset_req reset_type;
5093 pci_set_power_state(pdev, PCI_D0);
5094 pci_restore_state(pdev);
5095 pci_save_state(pdev);
5097 if (!pci_device_is_present(pdev))
5100 ret = pci_enable_device_mem(pdev);
5102 dev_err(dev, "Cannot enable device after suspend\n");
5106 pf = pci_get_drvdata(pdev);
5109 pf->wakeup_reason = rd32(hw, PFPM_WUS);
5110 ice_print_wake_reason(pf);
5112 /* We cleared the interrupt scheme when we suspended, so we need to
5113 * restore it now to resume device functionality.
5115 ret = ice_reinit_interrupt_scheme(pf);
5117 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
5119 clear_bit(ICE_DOWN, pf->state);
5120 /* Now perform PF reset and rebuild */
5121 reset_type = ICE_RESET_PFR;
5122 /* re-enable service task for reset, but allow reset to schedule it */
5123 clear_bit(ICE_SERVICE_DIS, pf->state);
5125 if (ice_schedule_reset(pf, reset_type))
5126 dev_err(dev, "Reset during resume failed.\n");
5128 clear_bit(ICE_SUSPENDED, pf->state);
5129 ice_service_task_restart(pf);
5131 /* Restart the service task */
5132 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5136 #endif /* CONFIG_PM */
5139 * ice_pci_err_detected - warning that PCI error has been detected
5140 * @pdev: PCI device information struct
5141 * @err: the type of PCI error
5143 * Called to warn that something happened on the PCI bus and the error handling
5144 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
5146 static pci_ers_result_t
5147 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
5149 struct ice_pf *pf = pci_get_drvdata(pdev);
5152 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
5154 return PCI_ERS_RESULT_DISCONNECT;
5157 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5158 ice_service_task_stop(pf);
5160 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5161 set_bit(ICE_PFR_REQ, pf->state);
5162 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5166 return PCI_ERS_RESULT_NEED_RESET;
5170 * ice_pci_err_slot_reset - a PCI slot reset has just happened
5171 * @pdev: PCI device information struct
5173 * Called to determine if the driver can recover from the PCI slot reset by
5174 * using a register read to determine if the device is recoverable.
5176 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
5178 struct ice_pf *pf = pci_get_drvdata(pdev);
5179 pci_ers_result_t result;
5183 err = pci_enable_device_mem(pdev);
5185 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
5187 result = PCI_ERS_RESULT_DISCONNECT;
5189 pci_set_master(pdev);
5190 pci_restore_state(pdev);
5191 pci_save_state(pdev);
5192 pci_wake_from_d3(pdev, false);
5194 /* Check for life */
5195 reg = rd32(&pf->hw, GLGEN_RTRIG);
5197 result = PCI_ERS_RESULT_RECOVERED;
5199 result = PCI_ERS_RESULT_DISCONNECT;
5202 err = pci_aer_clear_nonfatal_status(pdev);
5204 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
5206 /* non-fatal, continue */
5212 * ice_pci_err_resume - restart operations after PCI error recovery
5213 * @pdev: PCI device information struct
5215 * Called to allow the driver to bring things back up after PCI error and/or
5216 * reset recovery have finished
5218 static void ice_pci_err_resume(struct pci_dev *pdev)
5220 struct ice_pf *pf = pci_get_drvdata(pdev);
5223 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
5228 if (test_bit(ICE_SUSPENDED, pf->state)) {
5229 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
5234 ice_restore_all_vfs_msi_state(pdev);
5236 ice_do_reset(pf, ICE_RESET_PFR);
5237 ice_service_task_restart(pf);
5238 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
5242 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
5243 * @pdev: PCI device information struct
5245 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
5247 struct ice_pf *pf = pci_get_drvdata(pdev);
5249 if (!test_bit(ICE_SUSPENDED, pf->state)) {
5250 ice_service_task_stop(pf);
5252 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
5253 set_bit(ICE_PFR_REQ, pf->state);
5254 ice_prepare_for_reset(pf, ICE_RESET_PFR);
5260 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5261 * @pdev: PCI device information struct
5263 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5265 ice_pci_err_resume(pdev);
5268 /* ice_pci_tbl - PCI Device ID Table
5270 * Wildcard entries (PCI_ANY_ID) should come last
5271 * Last entry must be all 0s
5273 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5274 * Class, Class Mask, private data (not used) }
5276 static const struct pci_device_id ice_pci_tbl[] = {
5277 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5278 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5279 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5280 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_BACKPLANE), 0 },
5281 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_QSFP), 0 },
5282 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5283 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5284 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5285 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5286 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5287 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5288 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5289 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5290 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5291 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5292 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5293 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5294 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5295 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5296 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5297 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5298 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5299 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5300 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5301 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5302 /* required last entry */
5305 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5307 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5309 static const struct pci_error_handlers ice_pci_err_handler = {
5310 .error_detected = ice_pci_err_detected,
5311 .slot_reset = ice_pci_err_slot_reset,
5312 .reset_prepare = ice_pci_err_reset_prepare,
5313 .reset_done = ice_pci_err_reset_done,
5314 .resume = ice_pci_err_resume
5317 static struct pci_driver ice_driver = {
5318 .name = KBUILD_MODNAME,
5319 .id_table = ice_pci_tbl,
5321 .remove = ice_remove,
5323 .driver.pm = &ice_pm_ops,
5324 #endif /* CONFIG_PM */
5325 .shutdown = ice_shutdown,
5326 .sriov_configure = ice_sriov_configure,
5327 .err_handler = &ice_pci_err_handler
5331 * ice_module_init - Driver registration routine
5333 * ice_module_init is the first routine called when the driver is
5334 * loaded. All it does is register with the PCI subsystem.
5336 static int __init ice_module_init(void)
5340 pr_info("%s\n", ice_driver_string);
5341 pr_info("%s\n", ice_copyright);
5343 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5345 pr_err("Failed to create workqueue\n");
5349 status = pci_register_driver(&ice_driver);
5351 pr_err("failed to register PCI driver, err %d\n", status);
5352 destroy_workqueue(ice_wq);
5357 module_init(ice_module_init);
5360 * ice_module_exit - Driver exit cleanup routine
5362 * ice_module_exit is called just before the driver is removed
5365 static void __exit ice_module_exit(void)
5367 pci_unregister_driver(&ice_driver);
5368 destroy_workqueue(ice_wq);
5369 pr_info("module unloaded\n");
5371 module_exit(ice_module_exit);
5374 * ice_set_mac_address - NDO callback to set MAC address
5375 * @netdev: network interface device structure
5376 * @pi: pointer to an address structure
5378 * Returns 0 on success, negative on failure
5380 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5382 struct ice_netdev_priv *np = netdev_priv(netdev);
5383 struct ice_vsi *vsi = np->vsi;
5384 struct ice_pf *pf = vsi->back;
5385 struct ice_hw *hw = &pf->hw;
5386 struct sockaddr *addr = pi;
5387 u8 old_mac[ETH_ALEN];
5392 mac = (u8 *)addr->sa_data;
5394 if (!is_valid_ether_addr(mac))
5395 return -EADDRNOTAVAIL;
5397 if (ether_addr_equal(netdev->dev_addr, mac)) {
5398 netdev_dbg(netdev, "already using mac %pM\n", mac);
5402 if (test_bit(ICE_DOWN, pf->state) ||
5403 ice_is_reset_in_progress(pf->state)) {
5404 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5409 if (ice_chnl_dmac_fltr_cnt(pf)) {
5410 netdev_err(netdev, "can't set mac %pM. Device has tc-flower filters, delete all of them and try again\n",
5415 netif_addr_lock_bh(netdev);
5416 ether_addr_copy(old_mac, netdev->dev_addr);
5417 /* change the netdev's MAC address */
5418 eth_hw_addr_set(netdev, mac);
5419 netif_addr_unlock_bh(netdev);
5421 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5422 err = ice_fltr_remove_mac(vsi, old_mac, ICE_FWD_TO_VSI);
5423 if (err && err != -ENOENT) {
5424 err = -EADDRNOTAVAIL;
5425 goto err_update_filters;
5428 /* Add filter for new MAC. If filter exists, return success */
5429 err = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5431 /* Although this MAC filter is already present in hardware it's
5432 * possible in some cases (e.g. bonding) that dev_addr was
5433 * modified outside of the driver and needs to be restored back
5436 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5438 /* error if the new filter addition failed */
5439 err = -EADDRNOTAVAIL;
5443 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5445 netif_addr_lock_bh(netdev);
5446 eth_hw_addr_set(netdev, old_mac);
5447 netif_addr_unlock_bh(netdev);
5451 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5454 /* write new MAC address to the firmware */
5455 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5456 err = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5458 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %d\n",
5465 * ice_set_rx_mode - NDO callback to set the netdev filters
5466 * @netdev: network interface device structure
5468 static void ice_set_rx_mode(struct net_device *netdev)
5470 struct ice_netdev_priv *np = netdev_priv(netdev);
5471 struct ice_vsi *vsi = np->vsi;
5476 /* Set the flags to synchronize filters
5477 * ndo_set_rx_mode may be triggered even without a change in netdev
5480 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5481 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5482 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5484 /* schedule our worker thread which will take care of
5485 * applying the new filter changes
5487 ice_service_task_schedule(vsi->back);
5491 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5492 * @netdev: network interface device structure
5493 * @queue_index: Queue ID
5494 * @maxrate: maximum bandwidth in Mbps
5497 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5499 struct ice_netdev_priv *np = netdev_priv(netdev);
5500 struct ice_vsi *vsi = np->vsi;
5505 /* Validate maxrate requested is within permitted range */
5506 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5507 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5508 maxrate, queue_index);
5512 q_handle = vsi->tx_rings[queue_index]->q_handle;
5513 tc = ice_dcb_get_tc(vsi, queue_index);
5515 /* Set BW back to default, when user set maxrate to 0 */
5517 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5518 q_handle, ICE_MAX_BW);
5520 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5521 q_handle, ICE_MAX_BW, maxrate * 1000);
5523 netdev_err(netdev, "Unable to set Tx max rate, error %d\n",
5530 * ice_fdb_add - add an entry to the hardware database
5531 * @ndm: the input from the stack
5532 * @tb: pointer to array of nladdr (unused)
5533 * @dev: the net device pointer
5534 * @addr: the MAC address entry being added
5536 * @flags: instructions from stack about fdb operation
5537 * @extack: netlink extended ack
5540 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5541 struct net_device *dev, const unsigned char *addr, u16 vid,
5542 u16 flags, struct netlink_ext_ack __always_unused *extack)
5547 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5550 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5551 netdev_err(dev, "FDB only supports static addresses\n");
5555 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5556 err = dev_uc_add_excl(dev, addr);
5557 else if (is_multicast_ether_addr(addr))
5558 err = dev_mc_add_excl(dev, addr);
5562 /* Only return duplicate errors if NLM_F_EXCL is set */
5563 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5570 * ice_fdb_del - delete an entry from the hardware database
5571 * @ndm: the input from the stack
5572 * @tb: pointer to array of nladdr (unused)
5573 * @dev: the net device pointer
5574 * @addr: the MAC address entry being added
5578 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5579 struct net_device *dev, const unsigned char *addr,
5580 __always_unused u16 vid)
5584 if (ndm->ndm_state & NUD_PERMANENT) {
5585 netdev_err(dev, "FDB only supports static addresses\n");
5589 if (is_unicast_ether_addr(addr))
5590 err = dev_uc_del(dev, addr);
5591 else if (is_multicast_ether_addr(addr))
5592 err = dev_mc_del(dev, addr);
5600 * ice_set_features - set the netdev feature flags
5601 * @netdev: ptr to the netdev being adjusted
5602 * @features: the feature set that the stack is suggesting
5605 ice_set_features(struct net_device *netdev, netdev_features_t features)
5607 struct ice_netdev_priv *np = netdev_priv(netdev);
5608 struct ice_vsi *vsi = np->vsi;
5609 struct ice_pf *pf = vsi->back;
5612 /* Don't set any netdev advanced features with device in Safe Mode */
5613 if (ice_is_safe_mode(vsi->back)) {
5614 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5618 /* Do not change setting during reset */
5619 if (ice_is_reset_in_progress(pf->state)) {
5620 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5624 /* Multiple features can be changed in one call so keep features in
5625 * separate if/else statements to guarantee each feature is checked
5627 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5628 ice_vsi_manage_rss_lut(vsi, true);
5629 else if (!(features & NETIF_F_RXHASH) &&
5630 netdev->features & NETIF_F_RXHASH)
5631 ice_vsi_manage_rss_lut(vsi, false);
5633 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5634 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5635 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5636 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5637 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5638 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5640 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5641 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5642 ret = ice_vsi_manage_vlan_insertion(vsi);
5643 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5644 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5645 ret = ice_vsi_manage_vlan_insertion(vsi);
5647 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5648 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5649 ret = ice_cfg_vlan_pruning(vsi, true);
5650 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5651 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5652 ret = ice_cfg_vlan_pruning(vsi, false);
5654 if ((features & NETIF_F_NTUPLE) &&
5655 !(netdev->features & NETIF_F_NTUPLE)) {
5656 ice_vsi_manage_fdir(vsi, true);
5658 } else if (!(features & NETIF_F_NTUPLE) &&
5659 (netdev->features & NETIF_F_NTUPLE)) {
5660 ice_vsi_manage_fdir(vsi, false);
5661 ice_clear_arfs(vsi);
5664 /* don't turn off hw_tc_offload when ADQ is already enabled */
5665 if (!(features & NETIF_F_HW_TC) && ice_is_adq_active(pf)) {
5666 dev_err(ice_pf_to_dev(pf), "ADQ is active, can't turn hw_tc_offload off\n");
5670 if ((features & NETIF_F_HW_TC) &&
5671 !(netdev->features & NETIF_F_HW_TC))
5672 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
5674 clear_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
5680 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5681 * @vsi: VSI to setup VLAN properties for
5683 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5687 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5688 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5689 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5690 ret = ice_vsi_manage_vlan_insertion(vsi);
5696 * ice_vsi_cfg - Setup the VSI
5697 * @vsi: the VSI being configured
5699 * Return 0 on success and negative value on error
5701 int ice_vsi_cfg(struct ice_vsi *vsi)
5706 ice_set_rx_mode(vsi->netdev);
5708 err = ice_vsi_vlan_setup(vsi);
5713 ice_vsi_cfg_dcb_rings(vsi);
5715 err = ice_vsi_cfg_lan_txqs(vsi);
5716 if (!err && ice_is_xdp_ena_vsi(vsi))
5717 err = ice_vsi_cfg_xdp_txqs(vsi);
5719 err = ice_vsi_cfg_rxqs(vsi);
5724 /* THEORY OF MODERATION:
5725 * The ice driver hardware works differently than the hardware that DIMLIB was
5726 * originally made for. ice hardware doesn't have packet count limits that
5727 * can trigger an interrupt, but it *does* have interrupt rate limit support,
5728 * which is hard-coded to a limit of 250,000 ints/second.
5729 * If not using dynamic moderation, the INTRL value can be modified
5730 * by ethtool rx-usecs-high.
5733 /* the throttle rate for interrupts, basically worst case delay before
5734 * an initial interrupt fires, value is stored in microseconds.
5739 /* Make a different profile for Rx that doesn't allow quite so aggressive
5740 * moderation at the high end (it maxes out at 126us or about 8k interrupts a
5743 static const struct ice_dim rx_profile[] = {
5744 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
5745 {8}, /* 125,000 ints/s */
5746 {16}, /* 62,500 ints/s */
5747 {62}, /* 16,129 ints/s */
5748 {126} /* 7,936 ints/s */
5751 /* The transmit profile, which has the same sorts of values
5752 * as the previous struct
5754 static const struct ice_dim tx_profile[] = {
5755 {2}, /* 500,000 ints/s, capped at 250K by INTRL */
5756 {8}, /* 125,000 ints/s */
5757 {40}, /* 16,125 ints/s */
5758 {128}, /* 7,812 ints/s */
5759 {256} /* 3,906 ints/s */
5762 static void ice_tx_dim_work(struct work_struct *work)
5764 struct ice_ring_container *rc;
5768 dim = container_of(work, struct dim, work);
5769 rc = (struct ice_ring_container *)dim->priv;
5771 WARN_ON(dim->profile_ix >= ARRAY_SIZE(tx_profile));
5773 /* look up the values in our local table */
5774 itr = tx_profile[dim->profile_ix].itr;
5776 ice_trace(tx_dim_work, container_of(rc, struct ice_q_vector, tx), dim);
5777 ice_write_itr(rc, itr);
5779 dim->state = DIM_START_MEASURE;
5782 static void ice_rx_dim_work(struct work_struct *work)
5784 struct ice_ring_container *rc;
5788 dim = container_of(work, struct dim, work);
5789 rc = (struct ice_ring_container *)dim->priv;
5791 WARN_ON(dim->profile_ix >= ARRAY_SIZE(rx_profile));
5793 /* look up the values in our local table */
5794 itr = rx_profile[dim->profile_ix].itr;
5796 ice_trace(rx_dim_work, container_of(rc, struct ice_q_vector, rx), dim);
5797 ice_write_itr(rc, itr);
5799 dim->state = DIM_START_MEASURE;
5802 #define ICE_DIM_DEFAULT_PROFILE_IX 1
5805 * ice_init_moderation - set up interrupt moderation
5806 * @q_vector: the vector containing rings to be configured
5808 * Set up interrupt moderation registers, with the intent to do the right thing
5809 * when called from reset or from probe, and whether or not dynamic moderation
5810 * is enabled or not. Take special care to write all the registers in both
5811 * dynamic moderation mode or not in order to make sure hardware is in a known
5814 static void ice_init_moderation(struct ice_q_vector *q_vector)
5816 struct ice_ring_container *rc;
5817 bool tx_dynamic, rx_dynamic;
5820 INIT_WORK(&rc->dim.work, ice_tx_dim_work);
5821 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5822 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
5824 tx_dynamic = ITR_IS_DYNAMIC(rc);
5826 /* set the initial TX ITR to match the above */
5827 ice_write_itr(rc, tx_dynamic ?
5828 tx_profile[rc->dim.profile_ix].itr : rc->itr_setting);
5831 INIT_WORK(&rc->dim.work, ice_rx_dim_work);
5832 rc->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5833 rc->dim.profile_ix = ICE_DIM_DEFAULT_PROFILE_IX;
5835 rx_dynamic = ITR_IS_DYNAMIC(rc);
5837 /* set the initial RX ITR to match the above */
5838 ice_write_itr(rc, rx_dynamic ? rx_profile[rc->dim.profile_ix].itr :
5841 ice_set_q_vector_intrl(q_vector);
5845 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5846 * @vsi: the VSI being configured
5848 static void ice_napi_enable_all(struct ice_vsi *vsi)
5855 ice_for_each_q_vector(vsi, q_idx) {
5856 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5858 ice_init_moderation(q_vector);
5860 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
5861 napi_enable(&q_vector->napi);
5866 * ice_up_complete - Finish the last steps of bringing up a connection
5867 * @vsi: The VSI being configured
5869 * Return 0 on success and negative value on error
5871 static int ice_up_complete(struct ice_vsi *vsi)
5873 struct ice_pf *pf = vsi->back;
5876 ice_vsi_cfg_msix(vsi);
5878 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5879 * Tx queue group list was configured and the context bits were
5880 * programmed using ice_vsi_cfg_txqs
5882 err = ice_vsi_start_all_rx_rings(vsi);
5886 clear_bit(ICE_VSI_DOWN, vsi->state);
5887 ice_napi_enable_all(vsi);
5888 ice_vsi_ena_irq(vsi);
5890 if (vsi->port_info &&
5891 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5893 ice_print_link_msg(vsi, true);
5894 netif_tx_start_all_queues(vsi->netdev);
5895 netif_carrier_on(vsi->netdev);
5896 if (!ice_is_e810(&pf->hw))
5897 ice_ptp_link_change(pf, pf->hw.pf_id, true);
5900 /* clear this now, and the first stats read will be used as baseline */
5901 vsi->stat_offsets_loaded = false;
5903 ice_service_task_schedule(pf);
5909 * ice_up - Bring the connection back up after being down
5910 * @vsi: VSI being configured
5912 int ice_up(struct ice_vsi *vsi)
5916 err = ice_vsi_cfg(vsi);
5918 err = ice_up_complete(vsi);
5924 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5925 * @syncp: pointer to u64_stats_sync
5926 * @stats: stats that pkts and bytes count will be taken from
5927 * @pkts: packets stats counter
5928 * @bytes: bytes stats counter
5930 * This function fetches stats from the ring considering the atomic operations
5931 * that needs to be performed to read u64 values in 32 bit machine.
5934 ice_fetch_u64_stats_per_ring(struct u64_stats_sync *syncp, struct ice_q_stats stats,
5935 u64 *pkts, u64 *bytes)
5940 start = u64_stats_fetch_begin_irq(syncp);
5942 *bytes = stats.bytes;
5943 } while (u64_stats_fetch_retry_irq(syncp, start));
5947 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5948 * @vsi: the VSI to be updated
5949 * @vsi_stats: the stats struct to be updated
5950 * @rings: rings to work on
5951 * @count: number of rings
5954 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi,
5955 struct rtnl_link_stats64 *vsi_stats,
5956 struct ice_tx_ring **rings, u16 count)
5960 for (i = 0; i < count; i++) {
5961 struct ice_tx_ring *ring;
5962 u64 pkts = 0, bytes = 0;
5964 ring = READ_ONCE(rings[i]);
5967 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
5968 vsi_stats->tx_packets += pkts;
5969 vsi_stats->tx_bytes += bytes;
5970 vsi->tx_restart += ring->tx_stats.restart_q;
5971 vsi->tx_busy += ring->tx_stats.tx_busy;
5972 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5977 * ice_update_vsi_ring_stats - Update VSI stats counters
5978 * @vsi: the VSI to be updated
5980 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5982 struct rtnl_link_stats64 *vsi_stats;
5986 vsi_stats = kzalloc(sizeof(*vsi_stats), GFP_ATOMIC);
5990 /* reset non-netdev (extended) stats */
5991 vsi->tx_restart = 0;
5993 vsi->tx_linearize = 0;
5994 vsi->rx_buf_failed = 0;
5995 vsi->rx_page_failed = 0;
5999 /* update Tx rings counters */
6000 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->tx_rings,
6003 /* update Rx rings counters */
6004 ice_for_each_rxq(vsi, i) {
6005 struct ice_rx_ring *ring = READ_ONCE(vsi->rx_rings[i]);
6007 ice_fetch_u64_stats_per_ring(&ring->syncp, ring->stats, &pkts, &bytes);
6008 vsi_stats->rx_packets += pkts;
6009 vsi_stats->rx_bytes += bytes;
6010 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
6011 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
6014 /* update XDP Tx rings counters */
6015 if (ice_is_xdp_ena_vsi(vsi))
6016 ice_update_vsi_tx_ring_stats(vsi, vsi_stats, vsi->xdp_rings,
6021 vsi->net_stats.tx_packets = vsi_stats->tx_packets;
6022 vsi->net_stats.tx_bytes = vsi_stats->tx_bytes;
6023 vsi->net_stats.rx_packets = vsi_stats->rx_packets;
6024 vsi->net_stats.rx_bytes = vsi_stats->rx_bytes;
6030 * ice_update_vsi_stats - Update VSI stats counters
6031 * @vsi: the VSI to be updated
6033 void ice_update_vsi_stats(struct ice_vsi *vsi)
6035 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
6036 struct ice_eth_stats *cur_es = &vsi->eth_stats;
6037 struct ice_pf *pf = vsi->back;
6039 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
6040 test_bit(ICE_CFG_BUSY, pf->state))
6043 /* get stats as recorded by Tx/Rx rings */
6044 ice_update_vsi_ring_stats(vsi);
6046 /* get VSI stats as recorded by the hardware */
6047 ice_update_eth_stats(vsi);
6049 cur_ns->tx_errors = cur_es->tx_errors;
6050 cur_ns->rx_dropped = cur_es->rx_discards;
6051 cur_ns->tx_dropped = cur_es->tx_discards;
6052 cur_ns->multicast = cur_es->rx_multicast;
6054 /* update some more netdev stats if this is main VSI */
6055 if (vsi->type == ICE_VSI_PF) {
6056 cur_ns->rx_crc_errors = pf->stats.crc_errors;
6057 cur_ns->rx_errors = pf->stats.crc_errors +
6058 pf->stats.illegal_bytes +
6059 pf->stats.rx_len_errors +
6060 pf->stats.rx_undersize +
6061 pf->hw_csum_rx_error +
6062 pf->stats.rx_jabber +
6063 pf->stats.rx_fragments +
6064 pf->stats.rx_oversize;
6065 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
6066 /* record drops from the port level */
6067 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
6072 * ice_update_pf_stats - Update PF port stats counters
6073 * @pf: PF whose stats needs to be updated
6075 void ice_update_pf_stats(struct ice_pf *pf)
6077 struct ice_hw_port_stats *prev_ps, *cur_ps;
6078 struct ice_hw *hw = &pf->hw;
6082 port = hw->port_info->lport;
6083 prev_ps = &pf->stats_prev;
6084 cur_ps = &pf->stats;
6086 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
6087 &prev_ps->eth.rx_bytes,
6088 &cur_ps->eth.rx_bytes);
6090 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
6091 &prev_ps->eth.rx_unicast,
6092 &cur_ps->eth.rx_unicast);
6094 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
6095 &prev_ps->eth.rx_multicast,
6096 &cur_ps->eth.rx_multicast);
6098 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
6099 &prev_ps->eth.rx_broadcast,
6100 &cur_ps->eth.rx_broadcast);
6102 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
6103 &prev_ps->eth.rx_discards,
6104 &cur_ps->eth.rx_discards);
6106 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
6107 &prev_ps->eth.tx_bytes,
6108 &cur_ps->eth.tx_bytes);
6110 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
6111 &prev_ps->eth.tx_unicast,
6112 &cur_ps->eth.tx_unicast);
6114 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
6115 &prev_ps->eth.tx_multicast,
6116 &cur_ps->eth.tx_multicast);
6118 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
6119 &prev_ps->eth.tx_broadcast,
6120 &cur_ps->eth.tx_broadcast);
6122 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
6123 &prev_ps->tx_dropped_link_down,
6124 &cur_ps->tx_dropped_link_down);
6126 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
6127 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
6129 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
6130 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
6132 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
6133 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
6135 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
6136 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
6138 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
6139 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
6141 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
6142 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
6144 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
6145 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
6147 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
6148 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
6150 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
6151 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
6153 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
6154 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
6156 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
6157 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
6159 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
6160 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
6162 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
6163 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
6165 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
6166 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
6168 fd_ctr_base = hw->fd_ctr_base;
6170 ice_stat_update40(hw,
6171 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
6172 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
6173 &cur_ps->fd_sb_match);
6174 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
6175 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
6177 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
6178 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
6180 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
6181 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
6183 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
6184 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
6186 ice_update_dcb_stats(pf);
6188 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
6189 &prev_ps->crc_errors, &cur_ps->crc_errors);
6191 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
6192 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
6194 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
6195 &prev_ps->mac_local_faults,
6196 &cur_ps->mac_local_faults);
6198 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
6199 &prev_ps->mac_remote_faults,
6200 &cur_ps->mac_remote_faults);
6202 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
6203 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
6205 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
6206 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
6208 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
6209 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
6211 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
6212 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
6214 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
6215 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
6217 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
6219 pf->stat_prev_loaded = true;
6223 * ice_get_stats64 - get statistics for network device structure
6224 * @netdev: network interface device structure
6225 * @stats: main device statistics structure
6228 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
6230 struct ice_netdev_priv *np = netdev_priv(netdev);
6231 struct rtnl_link_stats64 *vsi_stats;
6232 struct ice_vsi *vsi = np->vsi;
6234 vsi_stats = &vsi->net_stats;
6236 if (!vsi->num_txq || !vsi->num_rxq)
6239 /* netdev packet/byte stats come from ring counter. These are obtained
6240 * by summing up ring counters (done by ice_update_vsi_ring_stats).
6241 * But, only call the update routine and read the registers if VSI is
6244 if (!test_bit(ICE_VSI_DOWN, vsi->state))
6245 ice_update_vsi_ring_stats(vsi);
6246 stats->tx_packets = vsi_stats->tx_packets;
6247 stats->tx_bytes = vsi_stats->tx_bytes;
6248 stats->rx_packets = vsi_stats->rx_packets;
6249 stats->rx_bytes = vsi_stats->rx_bytes;
6251 /* The rest of the stats can be read from the hardware but instead we
6252 * just return values that the watchdog task has already obtained from
6255 stats->multicast = vsi_stats->multicast;
6256 stats->tx_errors = vsi_stats->tx_errors;
6257 stats->tx_dropped = vsi_stats->tx_dropped;
6258 stats->rx_errors = vsi_stats->rx_errors;
6259 stats->rx_dropped = vsi_stats->rx_dropped;
6260 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
6261 stats->rx_length_errors = vsi_stats->rx_length_errors;
6265 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
6266 * @vsi: VSI having NAPI disabled
6268 static void ice_napi_disable_all(struct ice_vsi *vsi)
6275 ice_for_each_q_vector(vsi, q_idx) {
6276 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
6278 if (q_vector->rx.rx_ring || q_vector->tx.tx_ring)
6279 napi_disable(&q_vector->napi);
6281 cancel_work_sync(&q_vector->tx.dim.work);
6282 cancel_work_sync(&q_vector->rx.dim.work);
6287 * ice_down - Shutdown the connection
6288 * @vsi: The VSI being stopped
6290 * Caller of this function is expected to set the vsi->state ICE_DOWN bit
6292 int ice_down(struct ice_vsi *vsi)
6294 int i, tx_err, rx_err, link_err = 0;
6296 WARN_ON(!test_bit(ICE_VSI_DOWN, vsi->state));
6298 if (vsi->netdev && vsi->type == ICE_VSI_PF) {
6299 if (!ice_is_e810(&vsi->back->hw))
6300 ice_ptp_link_change(vsi->back, vsi->back->hw.pf_id, false);
6301 netif_carrier_off(vsi->netdev);
6302 netif_tx_disable(vsi->netdev);
6303 } else if (vsi->type == ICE_VSI_SWITCHDEV_CTRL) {
6304 ice_eswitch_stop_all_tx_queues(vsi->back);
6307 ice_vsi_dis_irq(vsi);
6309 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
6311 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
6312 vsi->vsi_num, tx_err);
6313 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
6314 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6316 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6317 vsi->vsi_num, tx_err);
6320 rx_err = ice_vsi_stop_all_rx_rings(vsi);
6322 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6323 vsi->vsi_num, rx_err);
6325 ice_napi_disable_all(vsi);
6327 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6328 link_err = ice_force_phys_link_state(vsi, false);
6330 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6331 vsi->vsi_num, link_err);
6334 ice_for_each_txq(vsi, i)
6335 ice_clean_tx_ring(vsi->tx_rings[i]);
6337 ice_for_each_rxq(vsi, i)
6338 ice_clean_rx_ring(vsi->rx_rings[i]);
6340 if (tx_err || rx_err || link_err) {
6341 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6342 vsi->vsi_num, vsi->vsw->sw_id);
6350 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6351 * @vsi: VSI having resources allocated
6353 * Return 0 on success, negative on failure
6355 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6359 if (!vsi->num_txq) {
6360 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6365 ice_for_each_txq(vsi, i) {
6366 struct ice_tx_ring *ring = vsi->tx_rings[i];
6372 ring->netdev = vsi->netdev;
6373 err = ice_setup_tx_ring(ring);
6382 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6383 * @vsi: VSI having resources allocated
6385 * Return 0 on success, negative on failure
6387 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6391 if (!vsi->num_rxq) {
6392 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6397 ice_for_each_rxq(vsi, i) {
6398 struct ice_rx_ring *ring = vsi->rx_rings[i];
6404 ring->netdev = vsi->netdev;
6405 err = ice_setup_rx_ring(ring);
6414 * ice_vsi_open_ctrl - open control VSI for use
6415 * @vsi: the VSI to open
6417 * Initialization of the Control VSI
6419 * Returns 0 on success, negative value on error
6421 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6423 char int_name[ICE_INT_NAME_STR_LEN];
6424 struct ice_pf *pf = vsi->back;
6428 dev = ice_pf_to_dev(pf);
6429 /* allocate descriptors */
6430 err = ice_vsi_setup_tx_rings(vsi);
6434 err = ice_vsi_setup_rx_rings(vsi);
6438 err = ice_vsi_cfg(vsi);
6442 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6443 dev_driver_string(dev), dev_name(dev));
6444 err = ice_vsi_req_irq_msix(vsi, int_name);
6448 ice_vsi_cfg_msix(vsi);
6450 err = ice_vsi_start_all_rx_rings(vsi);
6452 goto err_up_complete;
6454 clear_bit(ICE_VSI_DOWN, vsi->state);
6455 ice_vsi_ena_irq(vsi);
6462 ice_vsi_free_rx_rings(vsi);
6464 ice_vsi_free_tx_rings(vsi);
6470 * ice_vsi_open - Called when a network interface is made active
6471 * @vsi: the VSI to open
6473 * Initialization of the VSI
6475 * Returns 0 on success, negative value on error
6477 int ice_vsi_open(struct ice_vsi *vsi)
6479 char int_name[ICE_INT_NAME_STR_LEN];
6480 struct ice_pf *pf = vsi->back;
6483 /* allocate descriptors */
6484 err = ice_vsi_setup_tx_rings(vsi);
6488 err = ice_vsi_setup_rx_rings(vsi);
6492 err = ice_vsi_cfg(vsi);
6496 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6497 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6498 err = ice_vsi_req_irq_msix(vsi, int_name);
6502 if (vsi->type == ICE_VSI_PF) {
6503 /* Notify the stack of the actual queue counts. */
6504 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6508 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6513 err = ice_up_complete(vsi);
6515 goto err_up_complete;
6522 ice_vsi_free_irq(vsi);
6524 ice_vsi_free_rx_rings(vsi);
6526 ice_vsi_free_tx_rings(vsi);
6532 * ice_vsi_release_all - Delete all VSIs
6533 * @pf: PF from which all VSIs are being removed
6535 static void ice_vsi_release_all(struct ice_pf *pf)
6542 ice_for_each_vsi(pf, i) {
6546 if (pf->vsi[i]->type == ICE_VSI_CHNL)
6549 err = ice_vsi_release(pf->vsi[i]);
6551 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6552 i, err, pf->vsi[i]->vsi_num);
6557 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6558 * @pf: pointer to the PF instance
6559 * @type: VSI type to rebuild
6561 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6563 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6565 struct device *dev = ice_pf_to_dev(pf);
6568 ice_for_each_vsi(pf, i) {
6569 struct ice_vsi *vsi = pf->vsi[i];
6571 if (!vsi || vsi->type != type)
6574 /* rebuild the VSI */
6575 err = ice_vsi_rebuild(vsi, true);
6577 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6578 err, vsi->idx, ice_vsi_type_str(type));
6582 /* replay filters for the VSI */
6583 err = ice_replay_vsi(&pf->hw, vsi->idx);
6585 dev_err(dev, "replay VSI failed, error %d, VSI index %d, type %s\n",
6586 err, vsi->idx, ice_vsi_type_str(type));
6590 /* Re-map HW VSI number, using VSI handle that has been
6591 * previously validated in ice_replay_vsi() call above
6593 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6595 /* enable the VSI */
6596 err = ice_ena_vsi(vsi, false);
6598 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6599 err, vsi->idx, ice_vsi_type_str(type));
6603 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6604 ice_vsi_type_str(type));
6611 * ice_update_pf_netdev_link - Update PF netdev link status
6612 * @pf: pointer to the PF instance
6614 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6619 ice_for_each_vsi(pf, i) {
6620 struct ice_vsi *vsi = pf->vsi[i];
6622 if (!vsi || vsi->type != ICE_VSI_PF)
6625 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6627 netif_carrier_on(pf->vsi[i]->netdev);
6628 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6630 netif_carrier_off(pf->vsi[i]->netdev);
6631 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6637 * ice_rebuild - rebuild after reset
6638 * @pf: PF to rebuild
6639 * @reset_type: type of reset
6641 * Do not rebuild VF VSI in this flow because that is already handled via
6642 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6643 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6644 * to reset/rebuild all the VF VSI twice.
6646 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6648 struct device *dev = ice_pf_to_dev(pf);
6649 struct ice_hw *hw = &pf->hw;
6652 if (test_bit(ICE_DOWN, pf->state))
6653 goto clear_recovery;
6655 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6657 if (reset_type == ICE_RESET_EMPR) {
6658 /* If an EMP reset has occurred, any previously pending flash
6659 * update will have completed. We no longer know whether or
6660 * not the NVM update EMP reset is restricted.
6662 pf->fw_emp_reset_disabled = false;
6665 err = ice_init_all_ctrlq(hw);
6667 dev_err(dev, "control queues init failed %d\n", err);
6668 goto err_init_ctrlq;
6671 /* if DDP was previously loaded successfully */
6672 if (!ice_is_safe_mode(pf)) {
6673 /* reload the SW DB of filter tables */
6674 if (reset_type == ICE_RESET_PFR)
6675 ice_fill_blk_tbls(hw);
6677 /* Reload DDP Package after CORER/GLOBR reset */
6678 ice_load_pkg(NULL, pf);
6681 err = ice_clear_pf_cfg(hw);
6683 dev_err(dev, "clear PF configuration failed %d\n", err);
6684 goto err_init_ctrlq;
6687 if (pf->first_sw->dflt_vsi_ena)
6688 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6689 /* clear the default VSI configuration if it exists */
6690 pf->first_sw->dflt_vsi = NULL;
6691 pf->first_sw->dflt_vsi_ena = false;
6693 ice_clear_pxe_mode(hw);
6695 err = ice_init_nvm(hw);
6697 dev_err(dev, "ice_init_nvm failed %d\n", err);
6698 goto err_init_ctrlq;
6701 err = ice_get_caps(hw);
6703 dev_err(dev, "ice_get_caps failed %d\n", err);
6704 goto err_init_ctrlq;
6707 err = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6709 dev_err(dev, "set_mac_cfg failed %d\n", err);
6710 goto err_init_ctrlq;
6713 err = ice_sched_init_port(hw->port_info);
6715 goto err_sched_init_port;
6717 /* start misc vector */
6718 err = ice_req_irq_msix_misc(pf);
6720 dev_err(dev, "misc vector setup failed: %d\n", err);
6721 goto err_sched_init_port;
6724 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6725 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6726 if (!rd32(hw, PFQF_FD_SIZE)) {
6727 u16 unused, guar, b_effort;
6729 guar = hw->func_caps.fd_fltr_guar;
6730 b_effort = hw->func_caps.fd_fltr_best_effort;
6732 /* force guaranteed filter pool for PF */
6733 ice_alloc_fd_guar_item(hw, &unused, guar);
6734 /* force shared filter pool for PF */
6735 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6739 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6740 ice_dcb_rebuild(pf);
6742 /* If the PF previously had enabled PTP, PTP init needs to happen before
6743 * the VSI rebuild. If not, this causes the PTP link status events to
6746 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
6749 /* rebuild PF VSI */
6750 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6752 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6753 goto err_vsi_rebuild;
6756 /* configure PTP timestamping after VSI rebuild */
6757 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
6758 ice_ptp_cfg_timestamp(pf, false);
6760 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_SWITCHDEV_CTRL);
6762 dev_err(dev, "Switchdev CTRL VSI rebuild failed: %d\n", err);
6763 goto err_vsi_rebuild;
6766 if (reset_type == ICE_RESET_PFR) {
6767 err = ice_rebuild_channels(pf);
6769 dev_err(dev, "failed to rebuild and replay ADQ VSIs, err %d\n",
6771 goto err_vsi_rebuild;
6775 /* If Flow Director is active */
6776 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6777 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6779 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6780 goto err_vsi_rebuild;
6783 /* replay HW Flow Director recipes */
6785 ice_fdir_replay_flows(hw);
6787 /* replay Flow Director filters */
6788 ice_fdir_replay_fltrs(pf);
6790 ice_rebuild_arfs(pf);
6793 ice_update_pf_netdev_link(pf);
6795 /* tell the firmware we are up */
6796 err = ice_send_version(pf);
6798 dev_err(dev, "Rebuild failed due to error sending driver version: %d\n",
6800 goto err_vsi_rebuild;
6803 ice_replay_post(hw);
6805 /* if we get here, reset flow is successful */
6806 clear_bit(ICE_RESET_FAILED, pf->state);
6808 ice_plug_aux_dev(pf);
6812 err_sched_init_port:
6813 ice_sched_cleanup_all(hw);
6815 ice_shutdown_all_ctrlq(hw);
6816 set_bit(ICE_RESET_FAILED, pf->state);
6818 /* set this bit in PF state to control service task scheduling */
6819 set_bit(ICE_NEEDS_RESTART, pf->state);
6820 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6824 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6825 * @vsi: Pointer to VSI structure
6827 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6829 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6830 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6832 return ICE_RXBUF_3072;
6836 * ice_change_mtu - NDO callback to change the MTU
6837 * @netdev: network interface device structure
6838 * @new_mtu: new value for maximum frame size
6840 * Returns 0 on success, negative on failure
6842 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6844 struct ice_netdev_priv *np = netdev_priv(netdev);
6845 struct ice_vsi *vsi = np->vsi;
6846 struct ice_pf *pf = vsi->back;
6850 if (new_mtu == (int)netdev->mtu) {
6851 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6855 if (ice_is_xdp_ena_vsi(vsi)) {
6856 int frame_size = ice_max_xdp_frame_size(vsi);
6858 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6859 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6860 frame_size - ICE_ETH_PKT_HDR_PAD);
6865 /* if a reset is in progress, wait for some time for it to complete */
6867 if (ice_is_reset_in_progress(pf->state)) {
6869 usleep_range(1000, 2000);
6874 } while (count < 100);
6877 netdev_err(netdev, "can't change MTU. Device is busy\n");
6881 netdev->mtu = (unsigned int)new_mtu;
6883 /* if VSI is up, bring it down and then back up */
6884 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6885 err = ice_down(vsi);
6887 netdev_err(netdev, "change MTU if_down err %d\n", err);
6893 netdev_err(netdev, "change MTU if_up err %d\n", err);
6898 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6899 set_bit(ICE_FLAG_MTU_CHANGED, pf->flags);
6905 * ice_eth_ioctl - Access the hwtstamp interface
6906 * @netdev: network interface device structure
6907 * @ifr: interface request data
6908 * @cmd: ioctl command
6910 static int ice_eth_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6912 struct ice_netdev_priv *np = netdev_priv(netdev);
6913 struct ice_pf *pf = np->vsi->back;
6917 return ice_ptp_get_ts_config(pf, ifr);
6919 return ice_ptp_set_ts_config(pf, ifr);
6926 * ice_aq_str - convert AQ err code to a string
6927 * @aq_err: the AQ error code to convert
6929 const char *ice_aq_str(enum ice_aq_err aq_err)
6934 case ICE_AQ_RC_EPERM:
6935 return "ICE_AQ_RC_EPERM";
6936 case ICE_AQ_RC_ENOENT:
6937 return "ICE_AQ_RC_ENOENT";
6938 case ICE_AQ_RC_ENOMEM:
6939 return "ICE_AQ_RC_ENOMEM";
6940 case ICE_AQ_RC_EBUSY:
6941 return "ICE_AQ_RC_EBUSY";
6942 case ICE_AQ_RC_EEXIST:
6943 return "ICE_AQ_RC_EEXIST";
6944 case ICE_AQ_RC_EINVAL:
6945 return "ICE_AQ_RC_EINVAL";
6946 case ICE_AQ_RC_ENOSPC:
6947 return "ICE_AQ_RC_ENOSPC";
6948 case ICE_AQ_RC_ENOSYS:
6949 return "ICE_AQ_RC_ENOSYS";
6950 case ICE_AQ_RC_EMODE:
6951 return "ICE_AQ_RC_EMODE";
6952 case ICE_AQ_RC_ENOSEC:
6953 return "ICE_AQ_RC_ENOSEC";
6954 case ICE_AQ_RC_EBADSIG:
6955 return "ICE_AQ_RC_EBADSIG";
6956 case ICE_AQ_RC_ESVN:
6957 return "ICE_AQ_RC_ESVN";
6958 case ICE_AQ_RC_EBADMAN:
6959 return "ICE_AQ_RC_EBADMAN";
6960 case ICE_AQ_RC_EBADBUF:
6961 return "ICE_AQ_RC_EBADBUF";
6964 return "ICE_AQ_RC_UNKNOWN";
6968 * ice_set_rss_lut - Set RSS LUT
6969 * @vsi: Pointer to VSI structure
6970 * @lut: Lookup table
6971 * @lut_size: Lookup table size
6973 * Returns 0 on success, negative on failure
6975 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6977 struct ice_aq_get_set_rss_lut_params params = {};
6978 struct ice_hw *hw = &vsi->back->hw;
6984 params.vsi_handle = vsi->idx;
6985 params.lut_size = lut_size;
6986 params.lut_type = vsi->rss_lut_type;
6989 status = ice_aq_set_rss_lut(hw, ¶ms);
6991 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %d aq_err %s\n",
6992 status, ice_aq_str(hw->adminq.sq_last_status));
6998 * ice_set_rss_key - Set RSS key
6999 * @vsi: Pointer to the VSI structure
7000 * @seed: RSS hash seed
7002 * Returns 0 on success, negative on failure
7004 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
7006 struct ice_hw *hw = &vsi->back->hw;
7012 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7014 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %d aq_err %s\n",
7015 status, ice_aq_str(hw->adminq.sq_last_status));
7021 * ice_get_rss_lut - Get RSS LUT
7022 * @vsi: Pointer to VSI structure
7023 * @lut: Buffer to store the lookup table entries
7024 * @lut_size: Size of buffer to store the lookup table entries
7026 * Returns 0 on success, negative on failure
7028 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
7030 struct ice_aq_get_set_rss_lut_params params = {};
7031 struct ice_hw *hw = &vsi->back->hw;
7037 params.vsi_handle = vsi->idx;
7038 params.lut_size = lut_size;
7039 params.lut_type = vsi->rss_lut_type;
7042 status = ice_aq_get_rss_lut(hw, ¶ms);
7044 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %d aq_err %s\n",
7045 status, ice_aq_str(hw->adminq.sq_last_status));
7051 * ice_get_rss_key - Get RSS key
7052 * @vsi: Pointer to VSI structure
7053 * @seed: Buffer to store the key in
7055 * Returns 0 on success, negative on failure
7057 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
7059 struct ice_hw *hw = &vsi->back->hw;
7065 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
7067 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %d aq_err %s\n",
7068 status, ice_aq_str(hw->adminq.sq_last_status));
7074 * ice_bridge_getlink - Get the hardware bridge mode
7077 * @seq: RTNL message seq
7078 * @dev: the netdev being configured
7079 * @filter_mask: filter mask passed in
7080 * @nlflags: netlink flags passed in
7082 * Return the bridge mode (VEB/VEPA)
7085 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
7086 struct net_device *dev, u32 filter_mask, int nlflags)
7088 struct ice_netdev_priv *np = netdev_priv(dev);
7089 struct ice_vsi *vsi = np->vsi;
7090 struct ice_pf *pf = vsi->back;
7093 bmode = pf->first_sw->bridge_mode;
7095 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
7100 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
7101 * @vsi: Pointer to VSI structure
7102 * @bmode: Hardware bridge mode (VEB/VEPA)
7104 * Returns 0 on success, negative on failure
7106 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
7108 struct ice_aqc_vsi_props *vsi_props;
7109 struct ice_hw *hw = &vsi->back->hw;
7110 struct ice_vsi_ctx *ctxt;
7113 vsi_props = &vsi->info;
7115 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
7119 ctxt->info = vsi->info;
7121 if (bmode == BRIDGE_MODE_VEB)
7122 /* change from VEPA to VEB mode */
7123 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7125 /* change from VEB to VEPA mode */
7126 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
7127 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
7129 ret = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
7131 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %d aq_err %s\n",
7132 bmode, ret, ice_aq_str(hw->adminq.sq_last_status));
7135 /* Update sw flags for book keeping */
7136 vsi_props->sw_flags = ctxt->info.sw_flags;
7144 * ice_bridge_setlink - Set the hardware bridge mode
7145 * @dev: the netdev being configured
7146 * @nlh: RTNL message
7147 * @flags: bridge setlink flags
7148 * @extack: netlink extended ack
7150 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
7151 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
7152 * not already set for all VSIs connected to this switch. And also update the
7153 * unicast switch filter rules for the corresponding switch of the netdev.
7156 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
7157 u16 __always_unused flags,
7158 struct netlink_ext_ack __always_unused *extack)
7160 struct ice_netdev_priv *np = netdev_priv(dev);
7161 struct ice_pf *pf = np->vsi->back;
7162 struct nlattr *attr, *br_spec;
7163 struct ice_hw *hw = &pf->hw;
7164 struct ice_sw *pf_sw;
7165 int rem, v, err = 0;
7167 pf_sw = pf->first_sw;
7168 /* find the attribute in the netlink message */
7169 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
7171 nla_for_each_nested(attr, br_spec, rem) {
7174 if (nla_type(attr) != IFLA_BRIDGE_MODE)
7176 mode = nla_get_u16(attr);
7177 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
7179 /* Continue if bridge mode is not being flipped */
7180 if (mode == pf_sw->bridge_mode)
7182 /* Iterates through the PF VSI list and update the loopback
7185 ice_for_each_vsi(pf, v) {
7188 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
7193 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
7194 /* Update the unicast switch filter rules for the corresponding
7195 * switch of the netdev
7197 err = ice_update_sw_rule_bridge_mode(hw);
7199 netdev_err(dev, "switch rule update failed, mode = %d err %d aq_err %s\n",
7201 ice_aq_str(hw->adminq.sq_last_status));
7202 /* revert hw->evb_veb */
7203 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
7207 pf_sw->bridge_mode = mode;
7214 * ice_tx_timeout - Respond to a Tx Hang
7215 * @netdev: network interface device structure
7216 * @txqueue: Tx queue
7218 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
7220 struct ice_netdev_priv *np = netdev_priv(netdev);
7221 struct ice_tx_ring *tx_ring = NULL;
7222 struct ice_vsi *vsi = np->vsi;
7223 struct ice_pf *pf = vsi->back;
7226 pf->tx_timeout_count++;
7228 /* Check if PFC is enabled for the TC to which the queue belongs
7229 * to. If yes then Tx timeout is not caused by a hung queue, no
7230 * need to reset and rebuild
7232 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
7233 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
7238 /* now that we have an index, find the tx_ring struct */
7239 ice_for_each_txq(vsi, i)
7240 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
7241 if (txqueue == vsi->tx_rings[i]->q_index) {
7242 tx_ring = vsi->tx_rings[i];
7246 /* Reset recovery level if enough time has elapsed after last timeout.
7247 * Also ensure no new reset action happens before next timeout period.
7249 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7250 pf->tx_timeout_recovery_level = 1;
7251 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7252 netdev->watchdog_timeo)))
7256 struct ice_hw *hw = &pf->hw;
7259 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7260 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7261 /* Read interrupt register */
7262 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7264 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7265 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7266 head, tx_ring->next_to_use, val);
7269 pf->tx_timeout_last_recovery = jiffies;
7270 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7271 pf->tx_timeout_recovery_level, txqueue);
7273 switch (pf->tx_timeout_recovery_level) {
7275 set_bit(ICE_PFR_REQ, pf->state);
7278 set_bit(ICE_CORER_REQ, pf->state);
7281 set_bit(ICE_GLOBR_REQ, pf->state);
7284 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7285 set_bit(ICE_DOWN, pf->state);
7286 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7287 set_bit(ICE_SERVICE_DIS, pf->state);
7291 ice_service_task_schedule(pf);
7292 pf->tx_timeout_recovery_level++;
7296 * ice_setup_tc_cls_flower - flower classifier offloads
7297 * @np: net device to configure
7298 * @filter_dev: device on which filter is added
7299 * @cls_flower: offload data
7302 ice_setup_tc_cls_flower(struct ice_netdev_priv *np,
7303 struct net_device *filter_dev,
7304 struct flow_cls_offload *cls_flower)
7306 struct ice_vsi *vsi = np->vsi;
7308 if (cls_flower->common.chain_index)
7311 switch (cls_flower->command) {
7312 case FLOW_CLS_REPLACE:
7313 return ice_add_cls_flower(filter_dev, vsi, cls_flower);
7314 case FLOW_CLS_DESTROY:
7315 return ice_del_cls_flower(vsi, cls_flower);
7322 * ice_setup_tc_block_cb - callback handler registered for TC block
7323 * @type: TC SETUP type
7324 * @type_data: TC flower offload data that contains user input
7325 * @cb_priv: netdev private data
7328 ice_setup_tc_block_cb(enum tc_setup_type type, void *type_data, void *cb_priv)
7330 struct ice_netdev_priv *np = cb_priv;
7333 case TC_SETUP_CLSFLOWER:
7334 return ice_setup_tc_cls_flower(np, np->vsi->netdev,
7342 * ice_validate_mqprio_qopt - Validate TCF input parameters
7343 * @vsi: Pointer to VSI
7344 * @mqprio_qopt: input parameters for mqprio queue configuration
7346 * This function validates MQPRIO params, such as qcount (power of 2 wherever
7347 * needed), and make sure user doesn't specify qcount and BW rate limit
7348 * for TCs, which are more than "num_tc"
7351 ice_validate_mqprio_qopt(struct ice_vsi *vsi,
7352 struct tc_mqprio_qopt_offload *mqprio_qopt)
7354 u64 sum_max_rate = 0, sum_min_rate = 0;
7355 int non_power_of_2_qcount = 0;
7356 struct ice_pf *pf = vsi->back;
7357 int max_rss_q_cnt = 0;
7362 if (vsi->type != ICE_VSI_PF)
7365 if (mqprio_qopt->qopt.offset[0] != 0 ||
7366 mqprio_qopt->qopt.num_tc < 1 ||
7367 mqprio_qopt->qopt.num_tc > ICE_CHNL_MAX_TC)
7370 dev = ice_pf_to_dev(pf);
7371 vsi->ch_rss_size = 0;
7372 num_tc = mqprio_qopt->qopt.num_tc;
7374 for (i = 0; num_tc; i++) {
7375 int qcount = mqprio_qopt->qopt.count[i];
7376 u64 max_rate, min_rate, rem;
7381 if (is_power_of_2(qcount)) {
7382 if (non_power_of_2_qcount &&
7383 qcount > non_power_of_2_qcount) {
7384 dev_err(dev, "qcount[%d] cannot be greater than non power of 2 qcount[%d]\n",
7385 qcount, non_power_of_2_qcount);
7388 if (qcount > max_rss_q_cnt)
7389 max_rss_q_cnt = qcount;
7391 if (non_power_of_2_qcount &&
7392 qcount != non_power_of_2_qcount) {
7393 dev_err(dev, "Only one non power of 2 qcount allowed[%d,%d]\n",
7394 qcount, non_power_of_2_qcount);
7397 if (qcount < max_rss_q_cnt) {
7398 dev_err(dev, "non power of 2 qcount[%d] cannot be less than other qcount[%d]\n",
7399 qcount, max_rss_q_cnt);
7402 max_rss_q_cnt = qcount;
7403 non_power_of_2_qcount = qcount;
7406 /* TC command takes input in K/N/Gbps or K/M/Gbit etc but
7407 * converts the bandwidth rate limit into Bytes/s when
7408 * passing it down to the driver. So convert input bandwidth
7409 * from Bytes/s to Kbps
7411 max_rate = mqprio_qopt->max_rate[i];
7412 max_rate = div_u64(max_rate, ICE_BW_KBPS_DIVISOR);
7413 sum_max_rate += max_rate;
7415 /* min_rate is minimum guaranteed rate and it can't be zero */
7416 min_rate = mqprio_qopt->min_rate[i];
7417 min_rate = div_u64(min_rate, ICE_BW_KBPS_DIVISOR);
7418 sum_min_rate += min_rate;
7420 if (min_rate && min_rate < ICE_MIN_BW_LIMIT) {
7421 dev_err(dev, "TC%d: min_rate(%llu Kbps) < %u Kbps\n", i,
7422 min_rate, ICE_MIN_BW_LIMIT);
7426 iter_div_u64_rem(min_rate, ICE_MIN_BW_LIMIT, &rem);
7428 dev_err(dev, "TC%d: Min Rate not multiple of %u Kbps",
7429 i, ICE_MIN_BW_LIMIT);
7433 iter_div_u64_rem(max_rate, ICE_MIN_BW_LIMIT, &rem);
7435 dev_err(dev, "TC%d: Max Rate not multiple of %u Kbps",
7436 i, ICE_MIN_BW_LIMIT);
7440 /* min_rate can't be more than max_rate, except when max_rate
7441 * is zero (implies max_rate sought is max line rate). In such
7442 * a case min_rate can be more than max.
7444 if (max_rate && min_rate > max_rate) {
7445 dev_err(dev, "min_rate %llu Kbps can't be more than max_rate %llu Kbps\n",
7446 min_rate, max_rate);
7450 if (i >= mqprio_qopt->qopt.num_tc - 1)
7452 if (mqprio_qopt->qopt.offset[i + 1] !=
7453 (mqprio_qopt->qopt.offset[i] + qcount))
7457 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7460 (mqprio_qopt->qopt.offset[i] + mqprio_qopt->qopt.count[i]))
7463 speed = ice_get_link_speed_kbps(vsi);
7464 if (sum_max_rate && sum_max_rate > (u64)speed) {
7465 dev_err(dev, "Invalid max Tx rate(%llu) Kbps > speed(%u) Kbps specified\n",
7466 sum_max_rate, speed);
7469 if (sum_min_rate && sum_min_rate > (u64)speed) {
7470 dev_err(dev, "Invalid min Tx rate(%llu) Kbps > speed (%u) Kbps specified\n",
7471 sum_min_rate, speed);
7475 /* make sure vsi->ch_rss_size is set correctly based on TC's qcount */
7476 vsi->ch_rss_size = max_rss_q_cnt;
7482 * ice_add_vsi_to_fdir - add a VSI to the flow director group for PF
7483 * @pf: ptr to PF device
7486 static int ice_add_vsi_to_fdir(struct ice_pf *pf, struct ice_vsi *vsi)
7488 struct device *dev = ice_pf_to_dev(pf);
7493 if (!(vsi->num_gfltr || vsi->num_bfltr))
7497 for (flow = 0; flow < ICE_FLTR_PTYPE_MAX; flow++) {
7498 struct ice_fd_hw_prof *prof;
7502 if (!(hw->fdir_prof && hw->fdir_prof[flow] &&
7503 hw->fdir_prof[flow]->cnt))
7506 for (tun = 0; tun < ICE_FD_HW_SEG_MAX; tun++) {
7507 enum ice_flow_priority prio;
7510 /* add this VSI to FDir profile for this flow */
7511 prio = ICE_FLOW_PRIO_NORMAL;
7512 prof = hw->fdir_prof[flow];
7513 prof_id = flow + tun * ICE_FLTR_PTYPE_MAX;
7514 status = ice_flow_add_entry(hw, ICE_BLK_FD, prof_id,
7515 prof->vsi_h[0], vsi->idx,
7516 prio, prof->fdir_seg[tun],
7519 dev_err(dev, "channel VSI idx %d, not able to add to group %d\n",
7524 prof->entry_h[prof->cnt][tun] = entry_h;
7527 /* store VSI for filter replay and delete */
7528 prof->vsi_h[prof->cnt] = vsi->idx;
7532 dev_dbg(dev, "VSI idx %d added to fdir group %d\n", vsi->idx,
7537 dev_dbg(dev, "VSI idx %d not added to fdir groups\n", vsi->idx);
7543 * ice_add_channel - add a channel by adding VSI
7544 * @pf: ptr to PF device
7545 * @sw_id: underlying HW switching element ID
7546 * @ch: ptr to channel structure
7548 * Add a channel (VSI) using add_vsi and queue_map
7550 static int ice_add_channel(struct ice_pf *pf, u16 sw_id, struct ice_channel *ch)
7552 struct device *dev = ice_pf_to_dev(pf);
7553 struct ice_vsi *vsi;
7555 if (ch->type != ICE_VSI_CHNL) {
7556 dev_err(dev, "add new VSI failed, ch->type %d\n", ch->type);
7560 vsi = ice_chnl_vsi_setup(pf, pf->hw.port_info, ch);
7561 if (!vsi || vsi->type != ICE_VSI_CHNL) {
7562 dev_err(dev, "create chnl VSI failure\n");
7566 ice_add_vsi_to_fdir(pf, vsi);
7569 ch->vsi_num = vsi->vsi_num;
7570 ch->info.mapping_flags = vsi->info.mapping_flags;
7572 /* set the back pointer of channel for newly created VSI */
7575 memcpy(&ch->info.q_mapping, &vsi->info.q_mapping,
7576 sizeof(vsi->info.q_mapping));
7577 memcpy(&ch->info.tc_mapping, vsi->info.tc_mapping,
7578 sizeof(vsi->info.tc_mapping));
7585 * @vsi: the VSI being setup
7586 * @ch: ptr to channel structure
7588 * Configure channel specific resources such as rings, vector.
7590 static void ice_chnl_cfg_res(struct ice_vsi *vsi, struct ice_channel *ch)
7594 for (i = 0; i < ch->num_txq; i++) {
7595 struct ice_q_vector *tx_q_vector, *rx_q_vector;
7596 struct ice_ring_container *rc;
7597 struct ice_tx_ring *tx_ring;
7598 struct ice_rx_ring *rx_ring;
7600 tx_ring = vsi->tx_rings[ch->base_q + i];
7601 rx_ring = vsi->rx_rings[ch->base_q + i];
7602 if (!tx_ring || !rx_ring)
7605 /* setup ring being channel enabled */
7609 /* following code block sets up vector specific attributes */
7610 tx_q_vector = tx_ring->q_vector;
7611 rx_q_vector = rx_ring->q_vector;
7612 if (!tx_q_vector && !rx_q_vector)
7616 tx_q_vector->ch = ch;
7617 /* setup Tx and Rx ITR setting if DIM is off */
7618 rc = &tx_q_vector->tx;
7619 if (!ITR_IS_DYNAMIC(rc))
7620 ice_write_itr(rc, rc->itr_setting);
7623 rx_q_vector->ch = ch;
7624 /* setup Tx and Rx ITR setting if DIM is off */
7625 rc = &rx_q_vector->rx;
7626 if (!ITR_IS_DYNAMIC(rc))
7627 ice_write_itr(rc, rc->itr_setting);
7631 /* it is safe to assume that, if channel has non-zero num_t[r]xq, then
7632 * GLINT_ITR register would have written to perform in-context
7633 * update, hence perform flush
7635 if (ch->num_txq || ch->num_rxq)
7636 ice_flush(&vsi->back->hw);
7640 * ice_cfg_chnl_all_res - configure channel resources
7641 * @vsi: pte to main_vsi
7642 * @ch: ptr to channel structure
7644 * This function configures channel specific resources such as flow-director
7645 * counter index, and other resources such as queues, vectors, ITR settings
7648 ice_cfg_chnl_all_res(struct ice_vsi *vsi, struct ice_channel *ch)
7650 /* configure channel (aka ADQ) resources such as queues, vectors,
7651 * ITR settings for channel specific vectors and anything else
7653 ice_chnl_cfg_res(vsi, ch);
7657 * ice_setup_hw_channel - setup new channel
7658 * @pf: ptr to PF device
7659 * @vsi: the VSI being setup
7660 * @ch: ptr to channel structure
7661 * @sw_id: underlying HW switching element ID
7662 * @type: type of channel to be created (VMDq2/VF)
7664 * Setup new channel (VSI) based on specified type (VMDq2/VF)
7665 * and configures Tx rings accordingly
7668 ice_setup_hw_channel(struct ice_pf *pf, struct ice_vsi *vsi,
7669 struct ice_channel *ch, u16 sw_id, u8 type)
7671 struct device *dev = ice_pf_to_dev(pf);
7674 ch->base_q = vsi->next_base_q;
7677 ret = ice_add_channel(pf, sw_id, ch);
7679 dev_err(dev, "failed to add_channel using sw_id %u\n", sw_id);
7683 /* configure/setup ADQ specific resources */
7684 ice_cfg_chnl_all_res(vsi, ch);
7686 /* make sure to update the next_base_q so that subsequent channel's
7687 * (aka ADQ) VSI queue map is correct
7689 vsi->next_base_q = vsi->next_base_q + ch->num_rxq;
7690 dev_dbg(dev, "added channel: vsi_num %u, num_rxq %u\n", ch->vsi_num,
7697 * ice_setup_channel - setup new channel using uplink element
7698 * @pf: ptr to PF device
7699 * @vsi: the VSI being setup
7700 * @ch: ptr to channel structure
7702 * Setup new channel (VSI) based on specified type (VMDq2/VF)
7703 * and uplink switching element
7706 ice_setup_channel(struct ice_pf *pf, struct ice_vsi *vsi,
7707 struct ice_channel *ch)
7709 struct device *dev = ice_pf_to_dev(pf);
7713 if (vsi->type != ICE_VSI_PF) {
7714 dev_err(dev, "unsupported parent VSI type(%d)\n", vsi->type);
7718 sw_id = pf->first_sw->sw_id;
7720 /* create channel (VSI) */
7721 ret = ice_setup_hw_channel(pf, vsi, ch, sw_id, ICE_VSI_CHNL);
7723 dev_err(dev, "failed to setup hw_channel\n");
7726 dev_dbg(dev, "successfully created channel()\n");
7728 return ch->ch_vsi ? true : false;
7732 * ice_set_bw_limit - setup BW limit for Tx traffic based on max_tx_rate
7733 * @vsi: VSI to be configured
7734 * @max_tx_rate: max Tx rate in Kbps to be configured as maximum BW limit
7735 * @min_tx_rate: min Tx rate in Kbps to be configured as minimum BW limit
7738 ice_set_bw_limit(struct ice_vsi *vsi, u64 max_tx_rate, u64 min_tx_rate)
7742 err = ice_set_min_bw_limit(vsi, min_tx_rate);
7746 return ice_set_max_bw_limit(vsi, max_tx_rate);
7750 * ice_create_q_channel - function to create channel
7751 * @vsi: VSI to be configured
7752 * @ch: ptr to channel (it contains channel specific params)
7754 * This function creates channel (VSI) using num_queues specified by user,
7755 * reconfigs RSS if needed.
7757 static int ice_create_q_channel(struct ice_vsi *vsi, struct ice_channel *ch)
7759 struct ice_pf *pf = vsi->back;
7765 dev = ice_pf_to_dev(pf);
7766 if (!ch->num_txq || !ch->num_rxq) {
7767 dev_err(dev, "Invalid num_queues requested: %d\n", ch->num_rxq);
7771 if (!vsi->cnt_q_avail || vsi->cnt_q_avail < ch->num_txq) {
7772 dev_err(dev, "cnt_q_avail (%u) less than num_queues %d\n",
7773 vsi->cnt_q_avail, ch->num_txq);
7777 if (!ice_setup_channel(pf, vsi, ch)) {
7778 dev_info(dev, "Failed to setup channel\n");
7781 /* configure BW rate limit */
7782 if (ch->ch_vsi && (ch->max_tx_rate || ch->min_tx_rate)) {
7785 ret = ice_set_bw_limit(ch->ch_vsi, ch->max_tx_rate,
7788 dev_err(dev, "failed to set Tx rate of %llu Kbps for VSI(%u)\n",
7789 ch->max_tx_rate, ch->ch_vsi->vsi_num);
7791 dev_dbg(dev, "set Tx rate of %llu Kbps for VSI(%u)\n",
7792 ch->max_tx_rate, ch->ch_vsi->vsi_num);
7795 vsi->cnt_q_avail -= ch->num_txq;
7801 * ice_rem_all_chnl_fltrs - removes all channel filters
7802 * @pf: ptr to PF, TC-flower based filter are tracked at PF level
7804 * Remove all advanced switch filters only if they are channel specific
7805 * tc-flower based filter
7807 static void ice_rem_all_chnl_fltrs(struct ice_pf *pf)
7809 struct ice_tc_flower_fltr *fltr;
7810 struct hlist_node *node;
7812 /* to remove all channel filters, iterate an ordered list of filters */
7813 hlist_for_each_entry_safe(fltr, node,
7814 &pf->tc_flower_fltr_list,
7816 struct ice_rule_query_data rule;
7819 /* for now process only channel specific filters */
7820 if (!ice_is_chnl_fltr(fltr))
7823 rule.rid = fltr->rid;
7824 rule.rule_id = fltr->rule_id;
7825 rule.vsi_handle = fltr->dest_id;
7826 status = ice_rem_adv_rule_by_id(&pf->hw, &rule);
7828 if (status == -ENOENT)
7829 dev_dbg(ice_pf_to_dev(pf), "TC flower filter (rule_id %u) does not exist\n",
7832 dev_err(ice_pf_to_dev(pf), "failed to delete TC flower filter, status %d\n",
7834 } else if (fltr->dest_vsi) {
7835 /* update advanced switch filter count */
7836 if (fltr->dest_vsi->type == ICE_VSI_CHNL) {
7837 u32 flags = fltr->flags;
7839 fltr->dest_vsi->num_chnl_fltr--;
7840 if (flags & (ICE_TC_FLWR_FIELD_DST_MAC |
7841 ICE_TC_FLWR_FIELD_ENC_DST_MAC))
7842 pf->num_dmac_chnl_fltrs--;
7846 hlist_del(&fltr->tc_flower_node);
7852 * ice_remove_q_channels - Remove queue channels for the TCs
7853 * @vsi: VSI to be configured
7854 * @rem_fltr: delete advanced switch filter or not
7856 * Remove queue channels for the TCs
7858 static void ice_remove_q_channels(struct ice_vsi *vsi, bool rem_fltr)
7860 struct ice_channel *ch, *ch_tmp;
7861 struct ice_pf *pf = vsi->back;
7864 /* remove all tc-flower based filter if they are channel filters only */
7866 ice_rem_all_chnl_fltrs(pf);
7868 /* remove ntuple filters since queue configuration is being changed */
7869 if (vsi->netdev->features & NETIF_F_NTUPLE) {
7870 struct ice_hw *hw = &pf->hw;
7872 mutex_lock(&hw->fdir_fltr_lock);
7873 ice_fdir_del_all_fltrs(vsi);
7874 mutex_unlock(&hw->fdir_fltr_lock);
7877 /* perform cleanup for channels if they exist */
7878 list_for_each_entry_safe(ch, ch_tmp, &vsi->ch_list, list) {
7879 struct ice_vsi *ch_vsi;
7881 list_del(&ch->list);
7882 ch_vsi = ch->ch_vsi;
7888 /* Reset queue contexts */
7889 for (i = 0; i < ch->num_rxq; i++) {
7890 struct ice_tx_ring *tx_ring;
7891 struct ice_rx_ring *rx_ring;
7893 tx_ring = vsi->tx_rings[ch->base_q + i];
7894 rx_ring = vsi->rx_rings[ch->base_q + i];
7897 if (tx_ring->q_vector)
7898 tx_ring->q_vector->ch = NULL;
7902 if (rx_ring->q_vector)
7903 rx_ring->q_vector->ch = NULL;
7907 /* Release FD resources for the channel VSI */
7908 ice_fdir_rem_adq_chnl(&pf->hw, ch->ch_vsi->idx);
7910 /* clear the VSI from scheduler tree */
7911 ice_rm_vsi_lan_cfg(ch->ch_vsi->port_info, ch->ch_vsi->idx);
7913 /* Delete VSI from FW */
7914 ice_vsi_delete(ch->ch_vsi);
7916 /* Delete VSI from PF and HW VSI arrays */
7917 ice_vsi_clear(ch->ch_vsi);
7919 /* free the channel */
7923 /* clear the channel VSI map which is stored in main VSI */
7924 ice_for_each_chnl_tc(i)
7925 vsi->tc_map_vsi[i] = NULL;
7927 /* reset main VSI's all TC information */
7933 * ice_rebuild_channels - rebuild channel
7936 * Recreate channel VSIs and replay filters
7938 static int ice_rebuild_channels(struct ice_pf *pf)
7940 struct device *dev = ice_pf_to_dev(pf);
7941 struct ice_vsi *main_vsi;
7942 bool rem_adv_fltr = true;
7943 struct ice_channel *ch;
7944 struct ice_vsi *vsi;
7948 main_vsi = ice_get_main_vsi(pf);
7952 if (!test_bit(ICE_FLAG_TC_MQPRIO, pf->flags) ||
7953 main_vsi->old_numtc == 1)
7954 return 0; /* nothing to be done */
7956 /* reconfigure main VSI based on old value of TC and cached values
7959 err = ice_vsi_cfg_tc(main_vsi, main_vsi->old_ena_tc);
7961 dev_err(dev, "failed configuring TC(ena_tc:0x%02x) for HW VSI=%u\n",
7962 main_vsi->old_ena_tc, main_vsi->vsi_num);
7966 /* rebuild ADQ VSIs */
7967 ice_for_each_vsi(pf, i) {
7968 enum ice_vsi_type type;
7971 if (!vsi || vsi->type != ICE_VSI_CHNL)
7976 /* rebuild ADQ VSI */
7977 err = ice_vsi_rebuild(vsi, true);
7979 dev_err(dev, "VSI (type:%s) at index %d rebuild failed, err %d\n",
7980 ice_vsi_type_str(type), vsi->idx, err);
7984 /* Re-map HW VSI number, using VSI handle that has been
7985 * previously validated in ice_replay_vsi() call above
7987 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
7989 /* replay filters for the VSI */
7990 err = ice_replay_vsi(&pf->hw, vsi->idx);
7992 dev_err(dev, "VSI (type:%s) replay failed, err %d, VSI index %d\n",
7993 ice_vsi_type_str(type), err, vsi->idx);
7994 rem_adv_fltr = false;
7997 dev_info(dev, "VSI (type:%s) at index %d rebuilt successfully\n",
7998 ice_vsi_type_str(type), vsi->idx);
8000 /* store ADQ VSI at correct TC index in main VSI's
8003 main_vsi->tc_map_vsi[tc_idx++] = vsi;
8006 /* ADQ VSI(s) has been rebuilt successfully, so setup
8007 * channel for main VSI's Tx and Rx rings
8009 list_for_each_entry(ch, &main_vsi->ch_list, list) {
8010 struct ice_vsi *ch_vsi;
8012 ch_vsi = ch->ch_vsi;
8016 /* reconfig channel resources */
8017 ice_cfg_chnl_all_res(main_vsi, ch);
8019 /* replay BW rate limit if it is non-zero */
8020 if (!ch->max_tx_rate && !ch->min_tx_rate)
8023 err = ice_set_bw_limit(ch_vsi, ch->max_tx_rate,
8026 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",
8027 err, ch->max_tx_rate, ch->min_tx_rate,
8030 dev_dbg(dev, "successfully rebuild BW rate limit, max_tx_rate: %llu Kbps, min_tx_rate: %llu Kbps for VSI(%u)\n",
8031 ch->max_tx_rate, ch->min_tx_rate,
8035 /* reconfig RSS for main VSI */
8036 if (main_vsi->ch_rss_size)
8037 ice_vsi_cfg_rss_lut_key(main_vsi);
8042 ice_remove_q_channels(main_vsi, rem_adv_fltr);
8047 * ice_create_q_channels - Add queue channel for the given TCs
8048 * @vsi: VSI to be configured
8050 * Configures queue channel mapping to the given TCs
8052 static int ice_create_q_channels(struct ice_vsi *vsi)
8054 struct ice_pf *pf = vsi->back;
8055 struct ice_channel *ch;
8058 ice_for_each_chnl_tc(i) {
8059 if (!(vsi->all_enatc & BIT(i)))
8062 ch = kzalloc(sizeof(*ch), GFP_KERNEL);
8067 INIT_LIST_HEAD(&ch->list);
8068 ch->num_rxq = vsi->mqprio_qopt.qopt.count[i];
8069 ch->num_txq = vsi->mqprio_qopt.qopt.count[i];
8070 ch->base_q = vsi->mqprio_qopt.qopt.offset[i];
8071 ch->max_tx_rate = vsi->mqprio_qopt.max_rate[i];
8072 ch->min_tx_rate = vsi->mqprio_qopt.min_rate[i];
8074 /* convert to Kbits/s */
8075 if (ch->max_tx_rate)
8076 ch->max_tx_rate = div_u64(ch->max_tx_rate,
8077 ICE_BW_KBPS_DIVISOR);
8078 if (ch->min_tx_rate)
8079 ch->min_tx_rate = div_u64(ch->min_tx_rate,
8080 ICE_BW_KBPS_DIVISOR);
8082 ret = ice_create_q_channel(vsi, ch);
8084 dev_err(ice_pf_to_dev(pf),
8085 "failed creating channel TC:%d\n", i);
8089 list_add_tail(&ch->list, &vsi->ch_list);
8090 vsi->tc_map_vsi[i] = ch->ch_vsi;
8091 dev_dbg(ice_pf_to_dev(pf),
8092 "successfully created channel: VSI %pK\n", ch->ch_vsi);
8097 ice_remove_q_channels(vsi, false);
8103 * ice_setup_tc_mqprio_qdisc - configure multiple traffic classes
8104 * @netdev: net device to configure
8105 * @type_data: TC offload data
8107 static int ice_setup_tc_mqprio_qdisc(struct net_device *netdev, void *type_data)
8109 struct tc_mqprio_qopt_offload *mqprio_qopt = type_data;
8110 struct ice_netdev_priv *np = netdev_priv(netdev);
8111 struct ice_vsi *vsi = np->vsi;
8112 struct ice_pf *pf = vsi->back;
8113 u16 mode, ena_tc_qdisc = 0;
8114 int cur_txq, cur_rxq;
8119 dev = ice_pf_to_dev(pf);
8120 num_tcf = mqprio_qopt->qopt.num_tc;
8121 hw = mqprio_qopt->qopt.hw;
8122 mode = mqprio_qopt->mode;
8124 clear_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8125 vsi->ch_rss_size = 0;
8126 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8130 /* Generate queue region map for number of TCF requested */
8131 for (i = 0; i < num_tcf; i++)
8132 ena_tc_qdisc |= BIT(i);
8135 case TC_MQPRIO_MODE_CHANNEL:
8137 ret = ice_validate_mqprio_qopt(vsi, mqprio_qopt);
8139 netdev_err(netdev, "failed to validate_mqprio_qopt(), ret %d\n",
8143 memcpy(&vsi->mqprio_qopt, mqprio_qopt, sizeof(*mqprio_qopt));
8144 set_bit(ICE_FLAG_TC_MQPRIO, pf->flags);
8145 /* don't assume state of hw_tc_offload during driver load
8146 * and set the flag for TC flower filter if hw_tc_offload
8149 if (vsi->netdev->features & NETIF_F_HW_TC)
8150 set_bit(ICE_FLAG_CLS_FLOWER, pf->flags);
8158 /* Requesting same TCF configuration as already enabled */
8159 if (ena_tc_qdisc == vsi->tc_cfg.ena_tc &&
8160 mode != TC_MQPRIO_MODE_CHANNEL)
8163 /* Pause VSI queues */
8164 ice_dis_vsi(vsi, true);
8166 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags))
8167 ice_remove_q_channels(vsi, true);
8169 if (!hw && !test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8170 vsi->req_txq = min_t(int, ice_get_avail_txq_count(pf),
8172 vsi->req_rxq = min_t(int, ice_get_avail_rxq_count(pf),
8175 /* logic to rebuild VSI, same like ethtool -L */
8176 u16 offset = 0, qcount_tx = 0, qcount_rx = 0;
8178 for (i = 0; i < num_tcf; i++) {
8179 if (!(ena_tc_qdisc & BIT(i)))
8182 offset = vsi->mqprio_qopt.qopt.offset[i];
8183 qcount_rx = vsi->mqprio_qopt.qopt.count[i];
8184 qcount_tx = vsi->mqprio_qopt.qopt.count[i];
8186 vsi->req_txq = offset + qcount_tx;
8187 vsi->req_rxq = offset + qcount_rx;
8189 /* store away original rss_size info, so that it gets reused
8190 * form ice_vsi_rebuild during tc-qdisc delete stage - to
8191 * determine, what should be the rss_sizefor main VSI
8193 vsi->orig_rss_size = vsi->rss_size;
8196 /* save current values of Tx and Rx queues before calling VSI rebuild
8197 * for fallback option
8199 cur_txq = vsi->num_txq;
8200 cur_rxq = vsi->num_rxq;
8202 /* proceed with rebuild main VSI using correct number of queues */
8203 ret = ice_vsi_rebuild(vsi, false);
8205 /* fallback to current number of queues */
8206 dev_info(dev, "Rebuild failed with new queues, try with current number of queues\n");
8207 vsi->req_txq = cur_txq;
8208 vsi->req_rxq = cur_rxq;
8209 clear_bit(ICE_RESET_FAILED, pf->state);
8210 if (ice_vsi_rebuild(vsi, false)) {
8211 dev_err(dev, "Rebuild of main VSI failed again\n");
8216 vsi->all_numtc = num_tcf;
8217 vsi->all_enatc = ena_tc_qdisc;
8218 ret = ice_vsi_cfg_tc(vsi, ena_tc_qdisc);
8220 netdev_err(netdev, "failed configuring TC for VSI id=%d\n",
8225 if (test_bit(ICE_FLAG_TC_MQPRIO, pf->flags)) {
8226 u64 max_tx_rate = vsi->mqprio_qopt.max_rate[0];
8227 u64 min_tx_rate = vsi->mqprio_qopt.min_rate[0];
8229 /* set TC0 rate limit if specified */
8230 if (max_tx_rate || min_tx_rate) {
8231 /* convert to Kbits/s */
8233 max_tx_rate = div_u64(max_tx_rate, ICE_BW_KBPS_DIVISOR);
8235 min_tx_rate = div_u64(min_tx_rate, ICE_BW_KBPS_DIVISOR);
8237 ret = ice_set_bw_limit(vsi, max_tx_rate, min_tx_rate);
8239 dev_dbg(dev, "set Tx rate max %llu min %llu for VSI(%u)\n",
8240 max_tx_rate, min_tx_rate, vsi->vsi_num);
8242 dev_err(dev, "failed to set Tx rate max %llu min %llu for VSI(%u)\n",
8243 max_tx_rate, min_tx_rate, vsi->vsi_num);
8247 ret = ice_create_q_channels(vsi);
8249 netdev_err(netdev, "failed configuring queue channels\n");
8252 netdev_dbg(netdev, "successfully configured channels\n");
8256 if (vsi->ch_rss_size)
8257 ice_vsi_cfg_rss_lut_key(vsi);
8260 /* if error, reset the all_numtc and all_enatc */
8266 ice_ena_vsi(vsi, true);
8271 static LIST_HEAD(ice_block_cb_list);
8274 ice_setup_tc(struct net_device *netdev, enum tc_setup_type type,
8277 struct ice_netdev_priv *np = netdev_priv(netdev);
8278 struct ice_pf *pf = np->vsi->back;
8282 case TC_SETUP_BLOCK:
8283 return flow_block_cb_setup_simple(type_data,
8285 ice_setup_tc_block_cb,
8287 case TC_SETUP_QDISC_MQPRIO:
8288 /* setup traffic classifier for receive side */
8289 mutex_lock(&pf->tc_mutex);
8290 err = ice_setup_tc_mqprio_qdisc(netdev, type_data);
8291 mutex_unlock(&pf->tc_mutex);
8299 static struct ice_indr_block_priv *
8300 ice_indr_block_priv_lookup(struct ice_netdev_priv *np,
8301 struct net_device *netdev)
8303 struct ice_indr_block_priv *cb_priv;
8305 list_for_each_entry(cb_priv, &np->tc_indr_block_priv_list, list) {
8306 if (!cb_priv->netdev)
8308 if (cb_priv->netdev == netdev)
8315 ice_indr_setup_block_cb(enum tc_setup_type type, void *type_data,
8318 struct ice_indr_block_priv *priv = indr_priv;
8319 struct ice_netdev_priv *np = priv->np;
8322 case TC_SETUP_CLSFLOWER:
8323 return ice_setup_tc_cls_flower(np, priv->netdev,
8324 (struct flow_cls_offload *)
8332 ice_indr_setup_tc_block(struct net_device *netdev, struct Qdisc *sch,
8333 struct ice_netdev_priv *np,
8334 struct flow_block_offload *f, void *data,
8335 void (*cleanup)(struct flow_block_cb *block_cb))
8337 struct ice_indr_block_priv *indr_priv;
8338 struct flow_block_cb *block_cb;
8340 if (!ice_is_tunnel_supported(netdev) &&
8341 !(is_vlan_dev(netdev) &&
8342 vlan_dev_real_dev(netdev) == np->vsi->netdev))
8345 if (f->binder_type != FLOW_BLOCK_BINDER_TYPE_CLSACT_INGRESS)
8348 switch (f->command) {
8349 case FLOW_BLOCK_BIND:
8350 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8354 indr_priv = kzalloc(sizeof(*indr_priv), GFP_KERNEL);
8358 indr_priv->netdev = netdev;
8360 list_add(&indr_priv->list, &np->tc_indr_block_priv_list);
8363 flow_indr_block_cb_alloc(ice_indr_setup_block_cb,
8364 indr_priv, indr_priv,
8365 ice_rep_indr_tc_block_unbind,
8366 f, netdev, sch, data, np,
8369 if (IS_ERR(block_cb)) {
8370 list_del(&indr_priv->list);
8372 return PTR_ERR(block_cb);
8374 flow_block_cb_add(block_cb, f);
8375 list_add_tail(&block_cb->driver_list, &ice_block_cb_list);
8377 case FLOW_BLOCK_UNBIND:
8378 indr_priv = ice_indr_block_priv_lookup(np, netdev);
8382 block_cb = flow_block_cb_lookup(f->block,
8383 ice_indr_setup_block_cb,
8388 flow_indr_block_cb_remove(block_cb, f);
8390 list_del(&block_cb->driver_list);
8399 ice_indr_setup_tc_cb(struct net_device *netdev, struct Qdisc *sch,
8400 void *cb_priv, enum tc_setup_type type, void *type_data,
8402 void (*cleanup)(struct flow_block_cb *block_cb))
8405 case TC_SETUP_BLOCK:
8406 return ice_indr_setup_tc_block(netdev, sch, cb_priv, type_data,
8415 * ice_open - Called when a network interface becomes active
8416 * @netdev: network interface device structure
8418 * The open entry point is called when a network interface is made
8419 * active by the system (IFF_UP). At this point all resources needed
8420 * for transmit and receive operations are allocated, the interrupt
8421 * handler is registered with the OS, the netdev watchdog is enabled,
8422 * and the stack is notified that the interface is ready.
8424 * Returns 0 on success, negative value on failure
8426 int ice_open(struct net_device *netdev)
8428 struct ice_netdev_priv *np = netdev_priv(netdev);
8429 struct ice_pf *pf = np->vsi->back;
8431 if (ice_is_reset_in_progress(pf->state)) {
8432 netdev_err(netdev, "can't open net device while reset is in progress");
8436 return ice_open_internal(netdev);
8440 * ice_open_internal - Called when a network interface becomes active
8441 * @netdev: network interface device structure
8443 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
8446 * Returns 0 on success, negative value on failure
8448 int ice_open_internal(struct net_device *netdev)
8450 struct ice_netdev_priv *np = netdev_priv(netdev);
8451 struct ice_vsi *vsi = np->vsi;
8452 struct ice_pf *pf = vsi->back;
8453 struct ice_port_info *pi;
8456 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
8457 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
8461 netif_carrier_off(netdev);
8463 pi = vsi->port_info;
8464 err = ice_update_link_info(pi);
8466 netdev_err(netdev, "Failed to get link info, error %d\n", err);
8470 ice_check_link_cfg_err(pf, pi->phy.link_info.link_cfg_err);
8472 /* Set PHY if there is media, otherwise, turn off PHY */
8473 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
8474 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8475 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
8476 err = ice_init_phy_user_cfg(pi);
8478 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
8484 err = ice_configure_phy(vsi);
8486 netdev_err(netdev, "Failed to set physical link up, error %d\n",
8491 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
8492 ice_set_link(vsi, false);
8495 err = ice_vsi_open(vsi);
8497 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
8498 vsi->vsi_num, vsi->vsw->sw_id);
8500 /* Update existing tunnels information */
8501 udp_tunnel_get_rx_info(netdev);
8507 * ice_stop - Disables a network interface
8508 * @netdev: network interface device structure
8510 * The stop entry point is called when an interface is de-activated by the OS,
8511 * and the netdevice enters the DOWN state. The hardware is still under the
8512 * driver's control, but the netdev interface is disabled.
8514 * Returns success only - not allowed to fail
8516 int ice_stop(struct net_device *netdev)
8518 struct ice_netdev_priv *np = netdev_priv(netdev);
8519 struct ice_vsi *vsi = np->vsi;
8520 struct ice_pf *pf = vsi->back;
8522 if (ice_is_reset_in_progress(pf->state)) {
8523 netdev_err(netdev, "can't stop net device while reset is in progress");
8533 * ice_features_check - Validate encapsulated packet conforms to limits
8535 * @netdev: This port's netdev
8536 * @features: Offload features that the stack believes apply
8538 static netdev_features_t
8539 ice_features_check(struct sk_buff *skb,
8540 struct net_device __always_unused *netdev,
8541 netdev_features_t features)
8543 bool gso = skb_is_gso(skb);
8546 /* No point in doing any of this if neither checksum nor GSO are
8547 * being requested for this frame. We can rule out both by just
8548 * checking for CHECKSUM_PARTIAL
8550 if (skb->ip_summed != CHECKSUM_PARTIAL)
8553 /* We cannot support GSO if the MSS is going to be less than
8554 * 64 bytes. If it is then we need to drop support for GSO.
8556 if (gso && (skb_shinfo(skb)->gso_size < ICE_TXD_CTX_MIN_MSS))
8557 features &= ~NETIF_F_GSO_MASK;
8559 len = skb_network_offset(skb);
8560 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
8561 goto out_rm_features;
8563 len = skb_network_header_len(skb);
8564 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8565 goto out_rm_features;
8567 if (skb->encapsulation) {
8568 /* this must work for VXLAN frames AND IPIP/SIT frames, and in
8569 * the case of IPIP frames, the transport header pointer is
8570 * after the inner header! So check to make sure that this
8571 * is a GRE or UDP_TUNNEL frame before doing that math.
8573 if (gso && (skb_shinfo(skb)->gso_type &
8574 (SKB_GSO_GRE | SKB_GSO_UDP_TUNNEL))) {
8575 len = skb_inner_network_header(skb) -
8576 skb_transport_header(skb);
8577 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
8578 goto out_rm_features;
8581 len = skb_inner_network_header_len(skb);
8582 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
8583 goto out_rm_features;
8588 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
8591 static const struct net_device_ops ice_netdev_safe_mode_ops = {
8592 .ndo_open = ice_open,
8593 .ndo_stop = ice_stop,
8594 .ndo_start_xmit = ice_start_xmit,
8595 .ndo_set_mac_address = ice_set_mac_address,
8596 .ndo_validate_addr = eth_validate_addr,
8597 .ndo_change_mtu = ice_change_mtu,
8598 .ndo_get_stats64 = ice_get_stats64,
8599 .ndo_tx_timeout = ice_tx_timeout,
8600 .ndo_bpf = ice_xdp_safe_mode,
8603 static const struct net_device_ops ice_netdev_ops = {
8604 .ndo_open = ice_open,
8605 .ndo_stop = ice_stop,
8606 .ndo_start_xmit = ice_start_xmit,
8607 .ndo_select_queue = ice_select_queue,
8608 .ndo_features_check = ice_features_check,
8609 .ndo_set_rx_mode = ice_set_rx_mode,
8610 .ndo_set_mac_address = ice_set_mac_address,
8611 .ndo_validate_addr = eth_validate_addr,
8612 .ndo_change_mtu = ice_change_mtu,
8613 .ndo_get_stats64 = ice_get_stats64,
8614 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
8615 .ndo_eth_ioctl = ice_eth_ioctl,
8616 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
8617 .ndo_set_vf_mac = ice_set_vf_mac,
8618 .ndo_get_vf_config = ice_get_vf_cfg,
8619 .ndo_set_vf_trust = ice_set_vf_trust,
8620 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
8621 .ndo_set_vf_link_state = ice_set_vf_link_state,
8622 .ndo_get_vf_stats = ice_get_vf_stats,
8623 .ndo_set_vf_rate = ice_set_vf_bw,
8624 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
8625 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
8626 .ndo_setup_tc = ice_setup_tc,
8627 .ndo_set_features = ice_set_features,
8628 .ndo_bridge_getlink = ice_bridge_getlink,
8629 .ndo_bridge_setlink = ice_bridge_setlink,
8630 .ndo_fdb_add = ice_fdb_add,
8631 .ndo_fdb_del = ice_fdb_del,
8632 #ifdef CONFIG_RFS_ACCEL
8633 .ndo_rx_flow_steer = ice_rx_flow_steer,
8635 .ndo_tx_timeout = ice_tx_timeout,
8637 .ndo_xdp_xmit = ice_xdp_xmit,
8638 .ndo_xsk_wakeup = ice_xsk_wakeup,