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"
17 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
18 static const char ice_driver_string[] = DRV_SUMMARY;
19 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
21 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
22 #define ICE_DDP_PKG_PATH "intel/ice/ddp/"
23 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
26 MODULE_DESCRIPTION(DRV_SUMMARY);
27 MODULE_LICENSE("GPL v2");
28 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
30 static int debug = -1;
31 module_param(debug, int, 0644);
32 #ifndef CONFIG_DYNAMIC_DEBUG
33 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
35 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
36 #endif /* !CONFIG_DYNAMIC_DEBUG */
38 static DEFINE_IDA(ice_aux_ida);
40 static struct workqueue_struct *ice_wq;
41 static const struct net_device_ops ice_netdev_safe_mode_ops;
42 static const struct net_device_ops ice_netdev_ops;
43 static int ice_vsi_open(struct ice_vsi *vsi);
45 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
47 static void ice_vsi_release_all(struct ice_pf *pf);
49 bool netif_is_ice(struct net_device *dev)
51 return dev && (dev->netdev_ops == &ice_netdev_ops);
55 * ice_get_tx_pending - returns number of Tx descriptors not processed
56 * @ring: the ring of descriptors
58 static u16 ice_get_tx_pending(struct ice_ring *ring)
62 head = ring->next_to_clean;
63 tail = ring->next_to_use;
66 return (head < tail) ?
67 tail - head : (tail + ring->count - head);
72 * ice_check_for_hang_subtask - check for and recover hung queues
73 * @pf: pointer to PF struct
75 static void ice_check_for_hang_subtask(struct ice_pf *pf)
77 struct ice_vsi *vsi = NULL;
83 ice_for_each_vsi(pf, v)
84 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
89 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
92 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
97 for (i = 0; i < vsi->num_txq; i++) {
98 struct ice_ring *tx_ring = vsi->tx_rings[i];
100 if (tx_ring && tx_ring->desc) {
101 /* If packet counter has not changed the queue is
102 * likely stalled, so force an interrupt for this
105 * prev_pkt would be negative if there was no
108 packets = tx_ring->stats.pkts & INT_MAX;
109 if (tx_ring->tx_stats.prev_pkt == packets) {
110 /* Trigger sw interrupt to revive the queue */
111 ice_trigger_sw_intr(hw, tx_ring->q_vector);
115 /* Memory barrier between read of packet count and call
116 * to ice_get_tx_pending()
119 tx_ring->tx_stats.prev_pkt =
120 ice_get_tx_pending(tx_ring) ? packets : -1;
126 * ice_init_mac_fltr - Set initial MAC filters
127 * @pf: board private structure
129 * Set initial set of MAC filters for PF VSI; configure filters for permanent
130 * address and broadcast address. If an error is encountered, netdevice will be
133 static int ice_init_mac_fltr(struct ice_pf *pf)
135 enum ice_status status;
139 vsi = ice_get_main_vsi(pf);
143 perm_addr = vsi->port_info->mac.perm_addr;
144 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
152 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
153 * @netdev: the net device on which the sync is happening
154 * @addr: MAC address to sync
156 * This is a callback function which is called by the in kernel device sync
157 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
158 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
159 * MAC filters from the hardware.
161 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
163 struct ice_netdev_priv *np = netdev_priv(netdev);
164 struct ice_vsi *vsi = np->vsi;
166 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
174 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
175 * @netdev: the net device on which the unsync is happening
176 * @addr: MAC address to unsync
178 * This is a callback function which is called by the in kernel device unsync
179 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
180 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
181 * delete the MAC filters from the hardware.
183 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
185 struct ice_netdev_priv *np = netdev_priv(netdev);
186 struct ice_vsi *vsi = np->vsi;
188 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
196 * ice_vsi_fltr_changed - check if filter state changed
197 * @vsi: VSI to be checked
199 * returns true if filter state has changed, false otherwise.
201 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
203 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
204 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
205 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
209 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
210 * @vsi: the VSI being configured
211 * @promisc_m: mask of promiscuous config bits
212 * @set_promisc: enable or disable promisc flag request
215 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
217 struct ice_hw *hw = &vsi->back->hw;
218 enum ice_status status = 0;
220 if (vsi->type != ICE_VSI_PF)
223 if (vsi->num_vlan > 1) {
224 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
228 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
231 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
242 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
243 * @vsi: ptr to the VSI
245 * Push any outstanding VSI filter changes through the AdminQ.
247 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
249 struct device *dev = ice_pf_to_dev(vsi->back);
250 struct net_device *netdev = vsi->netdev;
251 bool promisc_forced_on = false;
252 struct ice_pf *pf = vsi->back;
253 struct ice_hw *hw = &pf->hw;
254 enum ice_status status = 0;
255 u32 changed_flags = 0;
262 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
263 usleep_range(1000, 2000);
265 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
266 vsi->current_netdev_flags = vsi->netdev->flags;
268 INIT_LIST_HEAD(&vsi->tmp_sync_list);
269 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
271 if (ice_vsi_fltr_changed(vsi)) {
272 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
273 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
274 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
276 /* grab the netdev's addr_list_lock */
277 netif_addr_lock_bh(netdev);
278 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
279 ice_add_mac_to_unsync_list);
280 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
281 ice_add_mac_to_unsync_list);
282 /* our temp lists are populated. release lock */
283 netif_addr_unlock_bh(netdev);
286 /* Remove MAC addresses in the unsync list */
287 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
288 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
290 netdev_err(netdev, "Failed to delete MAC filters\n");
291 /* if we failed because of alloc failures, just bail */
292 if (status == ICE_ERR_NO_MEMORY) {
298 /* Add MAC addresses in the sync list */
299 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
300 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
301 /* If filter is added successfully or already exists, do not go into
302 * 'if' condition and report it as error. Instead continue processing
303 * rest of the function.
305 if (status && status != ICE_ERR_ALREADY_EXISTS) {
306 netdev_err(netdev, "Failed to add MAC filters\n");
307 /* If there is no more space for new umac filters, VSI
308 * should go into promiscuous mode. There should be some
309 * space reserved for promiscuous filters.
311 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
312 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
314 promisc_forced_on = true;
315 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
322 /* check for changes in promiscuous modes */
323 if (changed_flags & IFF_ALLMULTI) {
324 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
325 if (vsi->num_vlan > 1)
326 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
328 promisc_m = ICE_MCAST_PROMISC_BITS;
330 err = ice_cfg_promisc(vsi, promisc_m, true);
332 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
334 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
338 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
339 if (vsi->num_vlan > 1)
340 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
342 promisc_m = ICE_MCAST_PROMISC_BITS;
344 err = ice_cfg_promisc(vsi, promisc_m, false);
346 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
348 vsi->current_netdev_flags |= IFF_ALLMULTI;
354 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
355 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
356 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
357 if (vsi->current_netdev_flags & IFF_PROMISC) {
358 /* Apply Rx filter rule to get traffic from wire */
359 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
360 err = ice_set_dflt_vsi(pf->first_sw, vsi);
361 if (err && err != -EEXIST) {
362 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
364 vsi->current_netdev_flags &=
368 ice_cfg_vlan_pruning(vsi, false, false);
371 /* Clear Rx filter to remove traffic from wire */
372 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
373 err = ice_clear_dflt_vsi(pf->first_sw);
375 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
377 vsi->current_netdev_flags |=
381 if (vsi->num_vlan > 1)
382 ice_cfg_vlan_pruning(vsi, true, false);
389 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
392 /* if something went wrong then set the changed flag so we try again */
393 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
394 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
396 clear_bit(ICE_CFG_BUSY, vsi->state);
401 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
402 * @pf: board private structure
404 static void ice_sync_fltr_subtask(struct ice_pf *pf)
408 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
411 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
413 ice_for_each_vsi(pf, v)
414 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
415 ice_vsi_sync_fltr(pf->vsi[v])) {
416 /* come back and try again later */
417 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
423 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
425 * @locked: is the rtnl_lock already held
427 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
432 ice_for_each_vsi(pf, v)
434 ice_dis_vsi(pf->vsi[v], locked);
436 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
437 pf->pf_agg_node[node].num_vsis = 0;
439 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
440 pf->vf_agg_node[node].num_vsis = 0;
444 * ice_prepare_for_reset - prep for the core to reset
445 * @pf: board private structure
447 * Inform or close all dependent features in prep for reset.
450 ice_prepare_for_reset(struct ice_pf *pf)
452 struct ice_hw *hw = &pf->hw;
455 /* already prepared for reset */
456 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
459 ice_unplug_aux_dev(pf);
461 /* Notify VFs of impending reset */
462 if (ice_check_sq_alive(hw, &hw->mailboxq))
463 ice_vc_notify_reset(pf);
465 /* Disable VFs until reset is completed */
466 ice_for_each_vf(pf, i)
467 ice_set_vf_state_qs_dis(&pf->vf[i]);
469 /* clear SW filtering DB */
470 ice_clear_hw_tbls(hw);
471 /* disable the VSIs and their queues that are not already DOWN */
472 ice_pf_dis_all_vsi(pf, false);
475 ice_sched_clear_port(hw->port_info);
477 ice_shutdown_all_ctrlq(hw);
479 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
483 * ice_do_reset - Initiate one of many types of resets
484 * @pf: board private structure
485 * @reset_type: reset type requested
486 * before this function was called.
488 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
490 struct device *dev = ice_pf_to_dev(pf);
491 struct ice_hw *hw = &pf->hw;
493 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
495 ice_prepare_for_reset(pf);
497 /* trigger the reset */
498 if (ice_reset(hw, reset_type)) {
499 dev_err(dev, "reset %d failed\n", reset_type);
500 set_bit(ICE_RESET_FAILED, pf->state);
501 clear_bit(ICE_RESET_OICR_RECV, pf->state);
502 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
503 clear_bit(ICE_PFR_REQ, pf->state);
504 clear_bit(ICE_CORER_REQ, pf->state);
505 clear_bit(ICE_GLOBR_REQ, pf->state);
509 /* PFR is a bit of a special case because it doesn't result in an OICR
510 * interrupt. So for PFR, rebuild after the reset and clear the reset-
511 * associated state bits.
513 if (reset_type == ICE_RESET_PFR) {
515 ice_rebuild(pf, reset_type);
516 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
517 clear_bit(ICE_PFR_REQ, pf->state);
518 ice_reset_all_vfs(pf, true);
523 * ice_reset_subtask - Set up for resetting the device and driver
524 * @pf: board private structure
526 static void ice_reset_subtask(struct ice_pf *pf)
528 enum ice_reset_req reset_type = ICE_RESET_INVAL;
530 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
531 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
532 * of reset is pending and sets bits in pf->state indicating the reset
533 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
534 * prepare for pending reset if not already (for PF software-initiated
535 * global resets the software should already be prepared for it as
536 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
537 * by firmware or software on other PFs, that bit is not set so prepare
538 * for the reset now), poll for reset done, rebuild and return.
540 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
541 /* Perform the largest reset requested */
542 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
543 reset_type = ICE_RESET_CORER;
544 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
545 reset_type = ICE_RESET_GLOBR;
546 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
547 reset_type = ICE_RESET_EMPR;
548 /* return if no valid reset type requested */
549 if (reset_type == ICE_RESET_INVAL)
551 ice_prepare_for_reset(pf);
553 /* make sure we are ready to rebuild */
554 if (ice_check_reset(&pf->hw)) {
555 set_bit(ICE_RESET_FAILED, pf->state);
557 /* done with reset. start rebuild */
558 pf->hw.reset_ongoing = false;
559 ice_rebuild(pf, reset_type);
560 /* clear bit to resume normal operations, but
561 * ICE_NEEDS_RESTART bit is set in case rebuild failed
563 clear_bit(ICE_RESET_OICR_RECV, pf->state);
564 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
565 clear_bit(ICE_PFR_REQ, pf->state);
566 clear_bit(ICE_CORER_REQ, pf->state);
567 clear_bit(ICE_GLOBR_REQ, pf->state);
568 ice_reset_all_vfs(pf, true);
574 /* No pending resets to finish processing. Check for new resets */
575 if (test_bit(ICE_PFR_REQ, pf->state))
576 reset_type = ICE_RESET_PFR;
577 if (test_bit(ICE_CORER_REQ, pf->state))
578 reset_type = ICE_RESET_CORER;
579 if (test_bit(ICE_GLOBR_REQ, pf->state))
580 reset_type = ICE_RESET_GLOBR;
581 /* If no valid reset type requested just return */
582 if (reset_type == ICE_RESET_INVAL)
585 /* reset if not already down or busy */
586 if (!test_bit(ICE_DOWN, pf->state) &&
587 !test_bit(ICE_CFG_BUSY, pf->state)) {
588 ice_do_reset(pf, reset_type);
593 * ice_print_topo_conflict - print topology conflict message
594 * @vsi: the VSI whose topology status is being checked
596 static void ice_print_topo_conflict(struct ice_vsi *vsi)
598 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
599 case ICE_AQ_LINK_TOPO_CONFLICT:
600 case ICE_AQ_LINK_MEDIA_CONFLICT:
601 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
602 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
603 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
604 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");
606 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
607 netdev_info(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");
615 * ice_print_link_msg - print link up or down message
616 * @vsi: the VSI whose link status is being queried
617 * @isup: boolean for if the link is now up or down
619 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
621 struct ice_aqc_get_phy_caps_data *caps;
622 const char *an_advertised;
623 enum ice_status status;
633 if (vsi->current_isup == isup)
636 vsi->current_isup = isup;
639 netdev_info(vsi->netdev, "NIC Link is Down\n");
643 switch (vsi->port_info->phy.link_info.link_speed) {
644 case ICE_AQ_LINK_SPEED_100GB:
647 case ICE_AQ_LINK_SPEED_50GB:
650 case ICE_AQ_LINK_SPEED_40GB:
653 case ICE_AQ_LINK_SPEED_25GB:
656 case ICE_AQ_LINK_SPEED_20GB:
659 case ICE_AQ_LINK_SPEED_10GB:
662 case ICE_AQ_LINK_SPEED_5GB:
665 case ICE_AQ_LINK_SPEED_2500MB:
668 case ICE_AQ_LINK_SPEED_1000MB:
671 case ICE_AQ_LINK_SPEED_100MB:
679 switch (vsi->port_info->fc.current_mode) {
683 case ICE_FC_TX_PAUSE:
686 case ICE_FC_RX_PAUSE:
697 /* Get FEC mode based on negotiated link info */
698 switch (vsi->port_info->phy.link_info.fec_info) {
699 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
700 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
703 case ICE_AQ_LINK_25G_KR_FEC_EN:
704 fec = "FC-FEC/BASE-R";
711 /* check if autoneg completed, might be false due to not supported */
712 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
717 /* Get FEC mode requested based on PHY caps last SW configuration */
718 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
721 an_advertised = "Unknown";
725 status = ice_aq_get_phy_caps(vsi->port_info, false,
726 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
728 netdev_info(vsi->netdev, "Get phy capability failed.\n");
730 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
732 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
733 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
735 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
736 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
737 fec_req = "FC-FEC/BASE-R";
744 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",
745 speed, fec_req, fec, an_advertised, an, fc);
746 ice_print_topo_conflict(vsi);
750 * ice_vsi_link_event - update the VSI's netdev
751 * @vsi: the VSI on which the link event occurred
752 * @link_up: whether or not the VSI needs to be set up or down
754 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
759 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
762 if (vsi->type == ICE_VSI_PF) {
763 if (link_up == netif_carrier_ok(vsi->netdev))
767 netif_carrier_on(vsi->netdev);
768 netif_tx_wake_all_queues(vsi->netdev);
770 netif_carrier_off(vsi->netdev);
771 netif_tx_stop_all_queues(vsi->netdev);
777 * ice_set_dflt_mib - send a default config MIB to the FW
778 * @pf: private PF struct
780 * This function sends a default configuration MIB to the FW.
782 * If this function errors out at any point, the driver is still able to
783 * function. The main impact is that LFC may not operate as expected.
784 * Therefore an error state in this function should be treated with a DBG
785 * message and continue on with driver rebuild/reenable.
787 static void ice_set_dflt_mib(struct ice_pf *pf)
789 struct device *dev = ice_pf_to_dev(pf);
790 u8 mib_type, *buf, *lldpmib = NULL;
791 u16 len, typelen, offset = 0;
792 struct ice_lldp_org_tlv *tlv;
793 struct ice_hw *hw = &pf->hw;
796 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
797 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
799 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
804 /* Add ETS CFG TLV */
805 tlv = (struct ice_lldp_org_tlv *)lldpmib;
806 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
807 ICE_IEEE_ETS_TLV_LEN);
808 tlv->typelen = htons(typelen);
809 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
810 ICE_IEEE_SUBTYPE_ETS_CFG);
811 tlv->ouisubtype = htonl(ouisubtype);
816 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
817 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
818 * Octets 13 - 20 are TSA values - leave as zeros
821 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
823 tlv = (struct ice_lldp_org_tlv *)
824 ((char *)tlv + sizeof(tlv->typelen) + len);
826 /* Add ETS REC TLV */
828 tlv->typelen = htons(typelen);
830 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
831 ICE_IEEE_SUBTYPE_ETS_REC);
832 tlv->ouisubtype = htonl(ouisubtype);
834 /* First octet of buf is reserved
835 * Octets 1 - 4 map UP to TC - all UPs map to zero
836 * Octets 5 - 12 are BW values - set TC 0 to 100%.
837 * Octets 13 - 20 are TSA value - leave as zeros
841 tlv = (struct ice_lldp_org_tlv *)
842 ((char *)tlv + sizeof(tlv->typelen) + len);
844 /* Add PFC CFG TLV */
845 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
846 ICE_IEEE_PFC_TLV_LEN);
847 tlv->typelen = htons(typelen);
849 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
850 ICE_IEEE_SUBTYPE_PFC_CFG);
851 tlv->ouisubtype = htonl(ouisubtype);
853 /* Octet 1 left as all zeros - PFC disabled */
855 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
858 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
859 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
865 * ice_link_event - process the link event
866 * @pf: PF that the link event is associated with
867 * @pi: port_info for the port that the link event is associated with
868 * @link_up: true if the physical link is up and false if it is down
869 * @link_speed: current link speed received from the link event
871 * Returns 0 on success and negative on failure
874 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
877 struct device *dev = ice_pf_to_dev(pf);
878 struct ice_phy_info *phy_info;
879 enum ice_status status;
885 phy_info->link_info_old = phy_info->link_info;
887 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
888 old_link_speed = phy_info->link_info_old.link_speed;
890 /* update the link info structures and re-enable link events,
891 * don't bail on failure due to other book keeping needed
893 status = ice_update_link_info(pi);
895 dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
896 pi->lport, ice_stat_str(status),
897 ice_aq_str(pi->hw->adminq.sq_last_status));
899 /* Check if the link state is up after updating link info, and treat
900 * this event as an UP event since the link is actually UP now.
902 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
905 vsi = ice_get_main_vsi(pf);
906 if (!vsi || !vsi->port_info)
909 /* turn off PHY if media was removed */
910 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
911 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
912 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
913 ice_set_link(vsi, false);
916 /* if the old link up/down and speed is the same as the new */
917 if (link_up == old_link && link_speed == old_link_speed)
920 if (ice_is_dcb_active(pf)) {
921 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
925 ice_set_dflt_mib(pf);
927 ice_vsi_link_event(vsi, link_up);
928 ice_print_link_msg(vsi, link_up);
930 ice_vc_notify_link_state(pf);
936 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
937 * @pf: board private structure
939 static void ice_watchdog_subtask(struct ice_pf *pf)
943 /* if interface is down do nothing */
944 if (test_bit(ICE_DOWN, pf->state) ||
945 test_bit(ICE_CFG_BUSY, pf->state))
948 /* make sure we don't do these things too often */
949 if (time_before(jiffies,
950 pf->serv_tmr_prev + pf->serv_tmr_period))
953 pf->serv_tmr_prev = jiffies;
955 /* Update the stats for active netdevs so the network stack
956 * can look at updated numbers whenever it cares to
958 ice_update_pf_stats(pf);
959 ice_for_each_vsi(pf, i)
960 if (pf->vsi[i] && pf->vsi[i]->netdev)
961 ice_update_vsi_stats(pf->vsi[i]);
965 * ice_init_link_events - enable/initialize link events
966 * @pi: pointer to the port_info instance
968 * Returns -EIO on failure, 0 on success
970 static int ice_init_link_events(struct ice_port_info *pi)
974 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
975 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
977 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
978 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
983 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
984 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
993 * ice_handle_link_event - handle link event via ARQ
994 * @pf: PF that the link event is associated with
995 * @event: event structure containing link status info
998 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1000 struct ice_aqc_get_link_status_data *link_data;
1001 struct ice_port_info *port_info;
1004 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1005 port_info = pf->hw.port_info;
1009 status = ice_link_event(pf, port_info,
1010 !!(link_data->link_info & ICE_AQ_LINK_UP),
1011 le16_to_cpu(link_data->link_speed));
1013 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1019 enum ice_aq_task_state {
1020 ICE_AQ_TASK_WAITING = 0,
1021 ICE_AQ_TASK_COMPLETE,
1022 ICE_AQ_TASK_CANCELED,
1025 struct ice_aq_task {
1026 struct hlist_node entry;
1029 struct ice_rq_event_info *event;
1030 enum ice_aq_task_state state;
1034 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1035 * @pf: pointer to the PF private structure
1036 * @opcode: the opcode to wait for
1037 * @timeout: how long to wait, in jiffies
1038 * @event: storage for the event info
1040 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1041 * current thread will be put to sleep until the specified event occurs or
1042 * until the given timeout is reached.
1044 * To obtain only the descriptor contents, pass an event without an allocated
1045 * msg_buf. If the complete data buffer is desired, allocate the
1046 * event->msg_buf with enough space ahead of time.
1048 * Returns: zero on success, or a negative error code on failure.
1050 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1051 struct ice_rq_event_info *event)
1053 struct device *dev = ice_pf_to_dev(pf);
1054 struct ice_aq_task *task;
1055 unsigned long start;
1059 task = kzalloc(sizeof(*task), GFP_KERNEL);
1063 INIT_HLIST_NODE(&task->entry);
1064 task->opcode = opcode;
1065 task->event = event;
1066 task->state = ICE_AQ_TASK_WAITING;
1068 spin_lock_bh(&pf->aq_wait_lock);
1069 hlist_add_head(&task->entry, &pf->aq_wait_list);
1070 spin_unlock_bh(&pf->aq_wait_lock);
1074 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1076 switch (task->state) {
1077 case ICE_AQ_TASK_WAITING:
1078 err = ret < 0 ? ret : -ETIMEDOUT;
1080 case ICE_AQ_TASK_CANCELED:
1081 err = ret < 0 ? ret : -ECANCELED;
1083 case ICE_AQ_TASK_COMPLETE:
1084 err = ret < 0 ? ret : 0;
1087 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1092 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1093 jiffies_to_msecs(jiffies - start),
1094 jiffies_to_msecs(timeout),
1097 spin_lock_bh(&pf->aq_wait_lock);
1098 hlist_del(&task->entry);
1099 spin_unlock_bh(&pf->aq_wait_lock);
1106 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1107 * @pf: pointer to the PF private structure
1108 * @opcode: the opcode of the event
1109 * @event: the event to check
1111 * Loops over the current list of pending threads waiting for an AdminQ event.
1112 * For each matching task, copy the contents of the event into the task
1113 * structure and wake up the thread.
1115 * If multiple threads wait for the same opcode, they will all be woken up.
1117 * Note that event->msg_buf will only be duplicated if the event has a buffer
1118 * with enough space already allocated. Otherwise, only the descriptor and
1119 * message length will be copied.
1121 * Returns: true if an event was found, false otherwise
1123 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1124 struct ice_rq_event_info *event)
1126 struct ice_aq_task *task;
1129 spin_lock_bh(&pf->aq_wait_lock);
1130 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1131 if (task->state || task->opcode != opcode)
1134 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1135 task->event->msg_len = event->msg_len;
1137 /* Only copy the data buffer if a destination was set */
1138 if (task->event->msg_buf &&
1139 task->event->buf_len > event->buf_len) {
1140 memcpy(task->event->msg_buf, event->msg_buf,
1142 task->event->buf_len = event->buf_len;
1145 task->state = ICE_AQ_TASK_COMPLETE;
1148 spin_unlock_bh(&pf->aq_wait_lock);
1151 wake_up(&pf->aq_wait_queue);
1155 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1156 * @pf: the PF private structure
1158 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1159 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1161 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1163 struct ice_aq_task *task;
1165 spin_lock_bh(&pf->aq_wait_lock);
1166 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1167 task->state = ICE_AQ_TASK_CANCELED;
1168 spin_unlock_bh(&pf->aq_wait_lock);
1170 wake_up(&pf->aq_wait_queue);
1174 * __ice_clean_ctrlq - helper function to clean controlq rings
1175 * @pf: ptr to struct ice_pf
1176 * @q_type: specific Control queue type
1178 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1180 struct device *dev = ice_pf_to_dev(pf);
1181 struct ice_rq_event_info event;
1182 struct ice_hw *hw = &pf->hw;
1183 struct ice_ctl_q_info *cq;
1188 /* Do not clean control queue if/when PF reset fails */
1189 if (test_bit(ICE_RESET_FAILED, pf->state))
1193 case ICE_CTL_Q_ADMIN:
1197 case ICE_CTL_Q_MAILBOX:
1200 /* we are going to try to detect a malicious VF, so set the
1201 * state to begin detection
1203 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1206 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1210 /* check for error indications - PF_xx_AxQLEN register layout for
1211 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1213 val = rd32(hw, cq->rq.len);
1214 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1215 PF_FW_ARQLEN_ARQCRIT_M)) {
1217 if (val & PF_FW_ARQLEN_ARQVFE_M)
1218 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1220 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1221 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1224 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1225 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1227 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1228 PF_FW_ARQLEN_ARQCRIT_M);
1230 wr32(hw, cq->rq.len, val);
1233 val = rd32(hw, cq->sq.len);
1234 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1235 PF_FW_ATQLEN_ATQCRIT_M)) {
1237 if (val & PF_FW_ATQLEN_ATQVFE_M)
1238 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1240 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1241 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1244 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1245 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1247 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1248 PF_FW_ATQLEN_ATQCRIT_M);
1250 wr32(hw, cq->sq.len, val);
1253 event.buf_len = cq->rq_buf_size;
1254 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1259 enum ice_status ret;
1262 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1263 if (ret == ICE_ERR_AQ_NO_WORK)
1266 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1271 opcode = le16_to_cpu(event.desc.opcode);
1273 /* Notify any thread that might be waiting for this event */
1274 ice_aq_check_events(pf, opcode, &event);
1277 case ice_aqc_opc_get_link_status:
1278 if (ice_handle_link_event(pf, &event))
1279 dev_err(dev, "Could not handle link event\n");
1281 case ice_aqc_opc_event_lan_overflow:
1282 ice_vf_lan_overflow_event(pf, &event);
1284 case ice_mbx_opc_send_msg_to_pf:
1285 if (!ice_is_malicious_vf(pf, &event, i, pending))
1286 ice_vc_process_vf_msg(pf, &event);
1288 case ice_aqc_opc_fw_logging:
1289 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1291 case ice_aqc_opc_lldp_set_mib_change:
1292 ice_dcb_process_lldp_set_mib_change(pf, &event);
1295 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1299 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1301 kfree(event.msg_buf);
1303 return pending && (i == ICE_DFLT_IRQ_WORK);
1307 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1308 * @hw: pointer to hardware info
1309 * @cq: control queue information
1311 * returns true if there are pending messages in a queue, false if there aren't
1313 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1317 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1318 return cq->rq.next_to_clean != ntu;
1322 * ice_clean_adminq_subtask - clean the AdminQ rings
1323 * @pf: board private structure
1325 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1327 struct ice_hw *hw = &pf->hw;
1329 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1332 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1335 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1337 /* There might be a situation where new messages arrive to a control
1338 * queue between processing the last message and clearing the
1339 * EVENT_PENDING bit. So before exiting, check queue head again (using
1340 * ice_ctrlq_pending) and process new messages if any.
1342 if (ice_ctrlq_pending(hw, &hw->adminq))
1343 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1349 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1350 * @pf: board private structure
1352 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1354 struct ice_hw *hw = &pf->hw;
1356 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1359 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1362 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1364 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1365 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1371 * ice_service_task_schedule - schedule the service task to wake up
1372 * @pf: board private structure
1374 * If not already scheduled, this puts the task into the work queue.
1376 void ice_service_task_schedule(struct ice_pf *pf)
1378 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1379 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1380 !test_bit(ICE_NEEDS_RESTART, pf->state))
1381 queue_work(ice_wq, &pf->serv_task);
1385 * ice_service_task_complete - finish up the service task
1386 * @pf: board private structure
1388 static void ice_service_task_complete(struct ice_pf *pf)
1390 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1392 /* force memory (pf->state) to sync before next service task */
1393 smp_mb__before_atomic();
1394 clear_bit(ICE_SERVICE_SCHED, pf->state);
1398 * ice_service_task_stop - stop service task and cancel works
1399 * @pf: board private structure
1401 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1404 static int ice_service_task_stop(struct ice_pf *pf)
1408 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1410 if (pf->serv_tmr.function)
1411 del_timer_sync(&pf->serv_tmr);
1412 if (pf->serv_task.func)
1413 cancel_work_sync(&pf->serv_task);
1415 clear_bit(ICE_SERVICE_SCHED, pf->state);
1420 * ice_service_task_restart - restart service task and schedule works
1421 * @pf: board private structure
1423 * This function is needed for suspend and resume works (e.g WoL scenario)
1425 static void ice_service_task_restart(struct ice_pf *pf)
1427 clear_bit(ICE_SERVICE_DIS, pf->state);
1428 ice_service_task_schedule(pf);
1432 * ice_service_timer - timer callback to schedule service task
1433 * @t: pointer to timer_list
1435 static void ice_service_timer(struct timer_list *t)
1437 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1439 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1440 ice_service_task_schedule(pf);
1444 * ice_handle_mdd_event - handle malicious driver detect event
1445 * @pf: pointer to the PF structure
1447 * Called from service task. OICR interrupt handler indicates MDD event.
1448 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1449 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1450 * disable the queue, the PF can be configured to reset the VF using ethtool
1451 * private flag mdd-auto-reset-vf.
1453 static void ice_handle_mdd_event(struct ice_pf *pf)
1455 struct device *dev = ice_pf_to_dev(pf);
1456 struct ice_hw *hw = &pf->hw;
1460 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1461 /* Since the VF MDD event logging is rate limited, check if
1462 * there are pending MDD events.
1464 ice_print_vfs_mdd_events(pf);
1468 /* find what triggered an MDD event */
1469 reg = rd32(hw, GL_MDET_TX_PQM);
1470 if (reg & GL_MDET_TX_PQM_VALID_M) {
1471 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1472 GL_MDET_TX_PQM_PF_NUM_S;
1473 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1474 GL_MDET_TX_PQM_VF_NUM_S;
1475 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1476 GL_MDET_TX_PQM_MAL_TYPE_S;
1477 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1478 GL_MDET_TX_PQM_QNUM_S);
1480 if (netif_msg_tx_err(pf))
1481 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1482 event, queue, pf_num, vf_num);
1483 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1486 reg = rd32(hw, GL_MDET_TX_TCLAN);
1487 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1488 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1489 GL_MDET_TX_TCLAN_PF_NUM_S;
1490 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1491 GL_MDET_TX_TCLAN_VF_NUM_S;
1492 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1493 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1494 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1495 GL_MDET_TX_TCLAN_QNUM_S);
1497 if (netif_msg_tx_err(pf))
1498 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1499 event, queue, pf_num, vf_num);
1500 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1503 reg = rd32(hw, GL_MDET_RX);
1504 if (reg & GL_MDET_RX_VALID_M) {
1505 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1506 GL_MDET_RX_PF_NUM_S;
1507 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1508 GL_MDET_RX_VF_NUM_S;
1509 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1510 GL_MDET_RX_MAL_TYPE_S;
1511 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1514 if (netif_msg_rx_err(pf))
1515 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1516 event, queue, pf_num, vf_num);
1517 wr32(hw, GL_MDET_RX, 0xffffffff);
1520 /* check to see if this PF caused an MDD event */
1521 reg = rd32(hw, PF_MDET_TX_PQM);
1522 if (reg & PF_MDET_TX_PQM_VALID_M) {
1523 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1524 if (netif_msg_tx_err(pf))
1525 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1528 reg = rd32(hw, PF_MDET_TX_TCLAN);
1529 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1530 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1531 if (netif_msg_tx_err(pf))
1532 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1535 reg = rd32(hw, PF_MDET_RX);
1536 if (reg & PF_MDET_RX_VALID_M) {
1537 wr32(hw, PF_MDET_RX, 0xFFFF);
1538 if (netif_msg_rx_err(pf))
1539 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1542 /* Check to see if one of the VFs caused an MDD event, and then
1543 * increment counters and set print pending
1545 ice_for_each_vf(pf, i) {
1546 struct ice_vf *vf = &pf->vf[i];
1548 reg = rd32(hw, VP_MDET_TX_PQM(i));
1549 if (reg & VP_MDET_TX_PQM_VALID_M) {
1550 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1551 vf->mdd_tx_events.count++;
1552 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1553 if (netif_msg_tx_err(pf))
1554 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1558 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1559 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1560 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1561 vf->mdd_tx_events.count++;
1562 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1563 if (netif_msg_tx_err(pf))
1564 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1568 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1569 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1570 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1571 vf->mdd_tx_events.count++;
1572 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1573 if (netif_msg_tx_err(pf))
1574 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1578 reg = rd32(hw, VP_MDET_RX(i));
1579 if (reg & VP_MDET_RX_VALID_M) {
1580 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1581 vf->mdd_rx_events.count++;
1582 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1583 if (netif_msg_rx_err(pf))
1584 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1587 /* Since the queue is disabled on VF Rx MDD events, the
1588 * PF can be configured to reset the VF through ethtool
1589 * private flag mdd-auto-reset-vf.
1591 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1592 /* VF MDD event counters will be cleared by
1593 * reset, so print the event prior to reset.
1595 ice_print_vf_rx_mdd_event(vf);
1596 ice_reset_vf(&pf->vf[i], false);
1601 ice_print_vfs_mdd_events(pf);
1605 * ice_force_phys_link_state - Force the physical link state
1606 * @vsi: VSI to force the physical link state to up/down
1607 * @link_up: true/false indicates to set the physical link to up/down
1609 * Force the physical link state by getting the current PHY capabilities from
1610 * hardware and setting the PHY config based on the determined capabilities. If
1611 * link changes a link event will be triggered because both the Enable Automatic
1612 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1614 * Returns 0 on success, negative on failure
1616 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1618 struct ice_aqc_get_phy_caps_data *pcaps;
1619 struct ice_aqc_set_phy_cfg_data *cfg;
1620 struct ice_port_info *pi;
1624 if (!vsi || !vsi->port_info || !vsi->back)
1626 if (vsi->type != ICE_VSI_PF)
1629 dev = ice_pf_to_dev(vsi->back);
1631 pi = vsi->port_info;
1633 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1637 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1640 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1641 vsi->vsi_num, retcode);
1646 /* No change in link */
1647 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1648 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1651 /* Use the current user PHY configuration. The current user PHY
1652 * configuration is initialized during probe from PHY capabilities
1653 * software mode, and updated on set PHY configuration.
1655 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1661 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1663 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1665 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1667 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1669 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1670 vsi->vsi_num, retcode);
1681 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1682 * @pi: port info structure
1684 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1686 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1688 struct ice_aqc_get_phy_caps_data *pcaps;
1689 struct ice_pf *pf = pi->hw->back;
1690 enum ice_status status;
1693 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1697 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1701 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1706 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1707 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1715 * ice_init_link_dflt_override - Initialize link default override
1716 * @pi: port info structure
1718 * Initialize link default override and PHY total port shutdown during probe
1720 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1722 struct ice_link_default_override_tlv *ldo;
1723 struct ice_pf *pf = pi->hw->back;
1725 ldo = &pf->link_dflt_override;
1726 if (ice_get_link_default_override(ldo, pi))
1729 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1732 /* Enable Total Port Shutdown (override/replace link-down-on-close
1733 * ethtool private flag) for ports with Port Disable bit set.
1735 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1736 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1740 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1741 * @pi: port info structure
1743 * If default override is enabled, initialize the user PHY cfg speed and FEC
1744 * settings using the default override mask from the NVM.
1746 * The PHY should only be configured with the default override settings the
1747 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1748 * is used to indicate that the user PHY cfg default override is initialized
1749 * and the PHY has not been configured with the default override settings. The
1750 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1753 * This function should be called only if the FW doesn't support default
1754 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1756 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1758 struct ice_link_default_override_tlv *ldo;
1759 struct ice_aqc_set_phy_cfg_data *cfg;
1760 struct ice_phy_info *phy = &pi->phy;
1761 struct ice_pf *pf = pi->hw->back;
1763 ldo = &pf->link_dflt_override;
1765 /* If link default override is enabled, use to mask NVM PHY capabilities
1766 * for speed and FEC default configuration.
1768 cfg = &phy->curr_user_phy_cfg;
1770 if (ldo->phy_type_low || ldo->phy_type_high) {
1771 cfg->phy_type_low = pf->nvm_phy_type_lo &
1772 cpu_to_le64(ldo->phy_type_low);
1773 cfg->phy_type_high = pf->nvm_phy_type_hi &
1774 cpu_to_le64(ldo->phy_type_high);
1776 cfg->link_fec_opt = ldo->fec_options;
1777 phy->curr_user_fec_req = ICE_FEC_AUTO;
1779 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1783 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1784 * @pi: port info structure
1786 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1787 * mode to default. The PHY defaults are from get PHY capabilities topology
1788 * with media so call when media is first available. An error is returned if
1789 * called when media is not available. The PHY initialization completed state is
1792 * These configurations are used when setting PHY
1793 * configuration. The user PHY configuration is updated on set PHY
1794 * configuration. Returns 0 on success, negative on failure
1796 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1798 struct ice_aqc_get_phy_caps_data *pcaps;
1799 struct ice_phy_info *phy = &pi->phy;
1800 struct ice_pf *pf = pi->hw->back;
1801 enum ice_status status;
1804 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1807 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1811 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1812 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1815 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1818 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1823 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1825 /* check if lenient mode is supported and enabled */
1826 if (ice_fw_supports_link_override(pi->hw) &&
1827 !(pcaps->module_compliance_enforcement &
1828 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1829 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1831 /* if the FW supports default PHY configuration mode, then the driver
1832 * does not have to apply link override settings. If not,
1833 * initialize user PHY configuration with link override values
1835 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1836 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1837 ice_init_phy_cfg_dflt_override(pi);
1842 /* if link default override is not enabled, set user flow control and
1843 * FEC settings based on what get_phy_caps returned
1845 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1846 pcaps->link_fec_options);
1847 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1850 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1851 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
1858 * ice_configure_phy - configure PHY
1861 * Set the PHY configuration. If the current PHY configuration is the same as
1862 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1863 * configure the based get PHY capabilities for topology with media.
1865 static int ice_configure_phy(struct ice_vsi *vsi)
1867 struct device *dev = ice_pf_to_dev(vsi->back);
1868 struct ice_port_info *pi = vsi->port_info;
1869 struct ice_aqc_get_phy_caps_data *pcaps;
1870 struct ice_aqc_set_phy_cfg_data *cfg;
1871 struct ice_phy_info *phy = &pi->phy;
1872 struct ice_pf *pf = vsi->back;
1873 enum ice_status status;
1876 /* Ensure we have media as we cannot configure a medialess port */
1877 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1880 ice_print_topo_conflict(vsi);
1882 if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1885 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1886 return ice_force_phys_link_state(vsi, true);
1888 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1892 /* Get current PHY config */
1893 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1896 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1897 vsi->vsi_num, ice_stat_str(status));
1902 /* If PHY enable link is configured and configuration has not changed,
1903 * there's nothing to do
1905 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1906 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1909 /* Use PHY topology as baseline for configuration */
1910 memset(pcaps, 0, sizeof(*pcaps));
1911 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1912 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1915 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1918 dev_err(dev, "Failed to get PHY caps, VSI %d error %s\n",
1919 vsi->vsi_num, ice_stat_str(status));
1924 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1930 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
1932 /* Speed - If default override pending, use curr_user_phy_cfg set in
1933 * ice_init_phy_user_cfg_ldo.
1935 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
1936 vsi->back->state)) {
1937 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
1938 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
1940 u64 phy_low = 0, phy_high = 0;
1942 ice_update_phy_type(&phy_low, &phy_high,
1943 pi->phy.curr_user_speed_req);
1944 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
1945 cfg->phy_type_high = pcaps->phy_type_high &
1946 cpu_to_le64(phy_high);
1949 /* Can't provide what was requested; use PHY capabilities */
1950 if (!cfg->phy_type_low && !cfg->phy_type_high) {
1951 cfg->phy_type_low = pcaps->phy_type_low;
1952 cfg->phy_type_high = pcaps->phy_type_high;
1956 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
1958 /* Can't provide what was requested; use PHY capabilities */
1959 if (cfg->link_fec_opt !=
1960 (cfg->link_fec_opt & pcaps->link_fec_options)) {
1961 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
1962 cfg->link_fec_opt = pcaps->link_fec_options;
1965 /* Flow Control - always supported; no need to check against
1968 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
1970 /* Enable link and link update */
1971 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
1973 status = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
1975 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
1976 vsi->vsi_num, ice_stat_str(status));
1987 * ice_check_media_subtask - Check for media
1988 * @pf: pointer to PF struct
1990 * If media is available, then initialize PHY user configuration if it is not
1991 * been, and configure the PHY if the interface is up.
1993 static void ice_check_media_subtask(struct ice_pf *pf)
1995 struct ice_port_info *pi;
1996 struct ice_vsi *vsi;
1999 /* No need to check for media if it's already present */
2000 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2003 vsi = ice_get_main_vsi(pf);
2007 /* Refresh link info and check if media is present */
2008 pi = vsi->port_info;
2009 err = ice_update_link_info(pi);
2013 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2014 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2015 ice_init_phy_user_cfg(pi);
2017 /* PHY settings are reset on media insertion, reconfigure
2018 * PHY to preserve settings.
2020 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2021 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2024 err = ice_configure_phy(vsi);
2026 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2028 /* A Link Status Event will be generated; the event handler
2029 * will complete bringing the interface up
2035 * ice_service_task - manage and run subtasks
2036 * @work: pointer to work_struct contained by the PF struct
2038 static void ice_service_task(struct work_struct *work)
2040 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2041 unsigned long start_time = jiffies;
2045 /* process reset requests first */
2046 ice_reset_subtask(pf);
2048 /* bail if a reset/recovery cycle is pending or rebuild failed */
2049 if (ice_is_reset_in_progress(pf->state) ||
2050 test_bit(ICE_SUSPENDED, pf->state) ||
2051 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2052 ice_service_task_complete(pf);
2056 ice_clean_adminq_subtask(pf);
2057 ice_check_media_subtask(pf);
2058 ice_check_for_hang_subtask(pf);
2059 ice_sync_fltr_subtask(pf);
2060 ice_handle_mdd_event(pf);
2061 ice_watchdog_subtask(pf);
2063 if (ice_is_safe_mode(pf)) {
2064 ice_service_task_complete(pf);
2068 ice_process_vflr_event(pf);
2069 ice_clean_mailboxq_subtask(pf);
2070 ice_sync_arfs_fltrs(pf);
2071 ice_flush_fdir_ctx(pf);
2073 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2074 ice_service_task_complete(pf);
2076 /* If the tasks have taken longer than one service timer period
2077 * or there is more work to be done, reset the service timer to
2078 * schedule the service task now.
2080 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2081 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2082 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2083 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2084 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2085 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2086 mod_timer(&pf->serv_tmr, jiffies);
2090 * ice_set_ctrlq_len - helper function to set controlq length
2091 * @hw: pointer to the HW instance
2093 static void ice_set_ctrlq_len(struct ice_hw *hw)
2095 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2096 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2097 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2098 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2099 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2100 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2101 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2102 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2106 * ice_schedule_reset - schedule a reset
2107 * @pf: board private structure
2108 * @reset: reset being requested
2110 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2112 struct device *dev = ice_pf_to_dev(pf);
2114 /* bail out if earlier reset has failed */
2115 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2116 dev_dbg(dev, "earlier reset has failed\n");
2119 /* bail if reset/recovery already in progress */
2120 if (ice_is_reset_in_progress(pf->state)) {
2121 dev_dbg(dev, "Reset already in progress\n");
2125 ice_unplug_aux_dev(pf);
2129 set_bit(ICE_PFR_REQ, pf->state);
2131 case ICE_RESET_CORER:
2132 set_bit(ICE_CORER_REQ, pf->state);
2134 case ICE_RESET_GLOBR:
2135 set_bit(ICE_GLOBR_REQ, pf->state);
2141 ice_service_task_schedule(pf);
2146 * ice_irq_affinity_notify - Callback for affinity changes
2147 * @notify: context as to what irq was changed
2148 * @mask: the new affinity mask
2150 * This is a callback function used by the irq_set_affinity_notifier function
2151 * so that we may register to receive changes to the irq affinity masks.
2154 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2155 const cpumask_t *mask)
2157 struct ice_q_vector *q_vector =
2158 container_of(notify, struct ice_q_vector, affinity_notify);
2160 cpumask_copy(&q_vector->affinity_mask, mask);
2164 * ice_irq_affinity_release - Callback for affinity notifier release
2165 * @ref: internal core kernel usage
2167 * This is a callback function used by the irq_set_affinity_notifier function
2168 * to inform the current notification subscriber that they will no longer
2169 * receive notifications.
2171 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2174 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2175 * @vsi: the VSI being configured
2177 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2179 struct ice_hw *hw = &vsi->back->hw;
2182 ice_for_each_q_vector(vsi, i)
2183 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2190 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2191 * @vsi: the VSI being configured
2192 * @basename: name for the vector
2194 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2196 int q_vectors = vsi->num_q_vectors;
2197 struct ice_pf *pf = vsi->back;
2198 int base = vsi->base_vector;
2205 dev = ice_pf_to_dev(pf);
2206 for (vector = 0; vector < q_vectors; vector++) {
2207 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2209 irq_num = pf->msix_entries[base + vector].vector;
2211 if (q_vector->tx.ring && q_vector->rx.ring) {
2212 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2213 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2215 } else if (q_vector->rx.ring) {
2216 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2217 "%s-%s-%d", basename, "rx", rx_int_idx++);
2218 } else if (q_vector->tx.ring) {
2219 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2220 "%s-%s-%d", basename, "tx", tx_int_idx++);
2222 /* skip this unused q_vector */
2225 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2226 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2227 IRQF_SHARED, q_vector->name,
2230 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2231 0, q_vector->name, q_vector);
2233 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2238 /* register for affinity change notifications */
2239 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2240 struct irq_affinity_notify *affinity_notify;
2242 affinity_notify = &q_vector->affinity_notify;
2243 affinity_notify->notify = ice_irq_affinity_notify;
2244 affinity_notify->release = ice_irq_affinity_release;
2245 irq_set_affinity_notifier(irq_num, affinity_notify);
2248 /* assign the mask for this irq */
2249 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2252 vsi->irqs_ready = true;
2258 irq_num = pf->msix_entries[base + vector].vector;
2259 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2260 irq_set_affinity_notifier(irq_num, NULL);
2261 irq_set_affinity_hint(irq_num, NULL);
2262 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2268 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2269 * @vsi: VSI to setup Tx rings used by XDP
2271 * Return 0 on success and negative value on error
2273 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2275 struct device *dev = ice_pf_to_dev(vsi->back);
2278 for (i = 0; i < vsi->num_xdp_txq; i++) {
2279 u16 xdp_q_idx = vsi->alloc_txq + i;
2280 struct ice_ring *xdp_ring;
2282 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2285 goto free_xdp_rings;
2287 xdp_ring->q_index = xdp_q_idx;
2288 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2289 xdp_ring->ring_active = false;
2290 xdp_ring->vsi = vsi;
2291 xdp_ring->netdev = NULL;
2292 xdp_ring->dev = dev;
2293 xdp_ring->count = vsi->num_tx_desc;
2294 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2295 if (ice_setup_tx_ring(xdp_ring))
2296 goto free_xdp_rings;
2297 ice_set_ring_xdp(xdp_ring);
2298 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2305 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2306 ice_free_tx_ring(vsi->xdp_rings[i]);
2311 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2312 * @vsi: VSI to set the bpf prog on
2313 * @prog: the bpf prog pointer
2315 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2317 struct bpf_prog *old_prog;
2320 old_prog = xchg(&vsi->xdp_prog, prog);
2322 bpf_prog_put(old_prog);
2324 ice_for_each_rxq(vsi, i)
2325 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2329 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2330 * @vsi: VSI to bring up Tx rings used by XDP
2331 * @prog: bpf program that will be assigned to VSI
2333 * Return 0 on success and negative value on error
2335 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2337 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2338 int xdp_rings_rem = vsi->num_xdp_txq;
2339 struct ice_pf *pf = vsi->back;
2340 struct ice_qs_cfg xdp_qs_cfg = {
2341 .qs_mutex = &pf->avail_q_mutex,
2342 .pf_map = pf->avail_txqs,
2343 .pf_map_size = pf->max_pf_txqs,
2344 .q_count = vsi->num_xdp_txq,
2345 .scatter_count = ICE_MAX_SCATTER_TXQS,
2346 .vsi_map = vsi->txq_map,
2347 .vsi_map_offset = vsi->alloc_txq,
2348 .mapping_mode = ICE_VSI_MAP_CONTIG
2350 enum ice_status status;
2354 dev = ice_pf_to_dev(pf);
2355 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2356 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2357 if (!vsi->xdp_rings)
2360 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2361 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2364 if (ice_xdp_alloc_setup_rings(vsi))
2365 goto clear_xdp_rings;
2367 /* follow the logic from ice_vsi_map_rings_to_vectors */
2368 ice_for_each_q_vector(vsi, v_idx) {
2369 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2370 int xdp_rings_per_v, q_id, q_base;
2372 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2373 vsi->num_q_vectors - v_idx);
2374 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2376 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2377 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2379 xdp_ring->q_vector = q_vector;
2380 xdp_ring->next = q_vector->tx.ring;
2381 q_vector->tx.ring = xdp_ring;
2383 xdp_rings_rem -= xdp_rings_per_v;
2386 /* omit the scheduler update if in reset path; XDP queues will be
2387 * taken into account at the end of ice_vsi_rebuild, where
2388 * ice_cfg_vsi_lan is being called
2390 if (ice_is_reset_in_progress(pf->state))
2393 /* tell the Tx scheduler that right now we have
2396 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2397 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2399 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2402 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2403 ice_stat_str(status));
2404 goto clear_xdp_rings;
2406 ice_vsi_assign_bpf_prog(vsi, prog);
2410 for (i = 0; i < vsi->num_xdp_txq; i++)
2411 if (vsi->xdp_rings[i]) {
2412 kfree_rcu(vsi->xdp_rings[i], rcu);
2413 vsi->xdp_rings[i] = NULL;
2417 mutex_lock(&pf->avail_q_mutex);
2418 for (i = 0; i < vsi->num_xdp_txq; i++) {
2419 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2420 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2422 mutex_unlock(&pf->avail_q_mutex);
2424 devm_kfree(dev, vsi->xdp_rings);
2429 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2430 * @vsi: VSI to remove XDP rings
2432 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2435 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2437 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2438 struct ice_pf *pf = vsi->back;
2441 /* q_vectors are freed in reset path so there's no point in detaching
2442 * rings; in case of rebuild being triggered not from reset bits
2443 * in pf->state won't be set, so additionally check first q_vector
2446 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2449 ice_for_each_q_vector(vsi, v_idx) {
2450 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2451 struct ice_ring *ring;
2453 ice_for_each_ring(ring, q_vector->tx)
2454 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2457 /* restore the value of last node prior to XDP setup */
2458 q_vector->tx.ring = ring;
2462 mutex_lock(&pf->avail_q_mutex);
2463 for (i = 0; i < vsi->num_xdp_txq; i++) {
2464 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2465 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2467 mutex_unlock(&pf->avail_q_mutex);
2469 for (i = 0; i < vsi->num_xdp_txq; i++)
2470 if (vsi->xdp_rings[i]) {
2471 if (vsi->xdp_rings[i]->desc)
2472 ice_free_tx_ring(vsi->xdp_rings[i]);
2473 kfree_rcu(vsi->xdp_rings[i], rcu);
2474 vsi->xdp_rings[i] = NULL;
2477 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2478 vsi->xdp_rings = NULL;
2480 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2483 ice_vsi_assign_bpf_prog(vsi, NULL);
2485 /* notify Tx scheduler that we destroyed XDP queues and bring
2486 * back the old number of child nodes
2488 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2489 max_txqs[i] = vsi->num_txq;
2491 /* change number of XDP Tx queues to 0 */
2492 vsi->num_xdp_txq = 0;
2494 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2499 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2500 * @vsi: VSI to schedule napi on
2502 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2506 ice_for_each_rxq(vsi, i) {
2507 struct ice_ring *rx_ring = vsi->rx_rings[i];
2509 if (rx_ring->xsk_pool)
2510 napi_schedule(&rx_ring->q_vector->napi);
2515 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2516 * @vsi: VSI to setup XDP for
2517 * @prog: XDP program
2518 * @extack: netlink extended ack
2521 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2522 struct netlink_ext_ack *extack)
2524 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2525 bool if_running = netif_running(vsi->netdev);
2526 int ret = 0, xdp_ring_err = 0;
2528 if (frame_size > vsi->rx_buf_len) {
2529 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2533 /* need to stop netdev while setting up the program for Rx rings */
2534 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2535 ret = ice_down(vsi);
2537 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2542 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2543 vsi->num_xdp_txq = vsi->alloc_rxq;
2544 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2546 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2547 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2548 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2550 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2552 ice_vsi_assign_bpf_prog(vsi, prog);
2559 ice_vsi_rx_napi_schedule(vsi);
2561 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2565 * ice_xdp_safe_mode - XDP handler for safe mode
2569 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2570 struct netdev_bpf *xdp)
2572 NL_SET_ERR_MSG_MOD(xdp->extack,
2573 "Please provide working DDP firmware package in order to use XDP\n"
2574 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2579 * ice_xdp - implements XDP handler
2583 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2585 struct ice_netdev_priv *np = netdev_priv(dev);
2586 struct ice_vsi *vsi = np->vsi;
2588 if (vsi->type != ICE_VSI_PF) {
2589 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2593 switch (xdp->command) {
2594 case XDP_SETUP_PROG:
2595 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2596 case XDP_SETUP_XSK_POOL:
2597 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2605 * ice_ena_misc_vector - enable the non-queue interrupts
2606 * @pf: board private structure
2608 static void ice_ena_misc_vector(struct ice_pf *pf)
2610 struct ice_hw *hw = &pf->hw;
2613 /* Disable anti-spoof detection interrupt to prevent spurious event
2614 * interrupts during a function reset. Anti-spoof functionally is
2617 val = rd32(hw, GL_MDCK_TX_TDPU);
2618 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2619 wr32(hw, GL_MDCK_TX_TDPU, val);
2621 /* clear things first */
2622 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2623 rd32(hw, PFINT_OICR); /* read to clear */
2625 val = (PFINT_OICR_ECC_ERR_M |
2626 PFINT_OICR_MAL_DETECT_M |
2628 PFINT_OICR_PCI_EXCEPTION_M |
2630 PFINT_OICR_HMC_ERR_M |
2631 PFINT_OICR_PE_PUSH_M |
2632 PFINT_OICR_PE_CRITERR_M);
2634 wr32(hw, PFINT_OICR_ENA, val);
2636 /* SW_ITR_IDX = 0, but don't change INTENA */
2637 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2638 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2642 * ice_misc_intr - misc interrupt handler
2643 * @irq: interrupt number
2644 * @data: pointer to a q_vector
2646 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2648 struct ice_pf *pf = (struct ice_pf *)data;
2649 struct ice_hw *hw = &pf->hw;
2650 irqreturn_t ret = IRQ_NONE;
2654 dev = ice_pf_to_dev(pf);
2655 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2656 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2658 oicr = rd32(hw, PFINT_OICR);
2659 ena_mask = rd32(hw, PFINT_OICR_ENA);
2661 if (oicr & PFINT_OICR_SWINT_M) {
2662 ena_mask &= ~PFINT_OICR_SWINT_M;
2666 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2667 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2668 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2670 if (oicr & PFINT_OICR_VFLR_M) {
2671 /* disable any further VFLR event notifications */
2672 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2673 u32 reg = rd32(hw, PFINT_OICR_ENA);
2675 reg &= ~PFINT_OICR_VFLR_M;
2676 wr32(hw, PFINT_OICR_ENA, reg);
2678 ena_mask &= ~PFINT_OICR_VFLR_M;
2679 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2683 if (oicr & PFINT_OICR_GRST_M) {
2686 /* we have a reset warning */
2687 ena_mask &= ~PFINT_OICR_GRST_M;
2688 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2689 GLGEN_RSTAT_RESET_TYPE_S;
2691 if (reset == ICE_RESET_CORER)
2693 else if (reset == ICE_RESET_GLOBR)
2695 else if (reset == ICE_RESET_EMPR)
2698 dev_dbg(dev, "Invalid reset type %d\n", reset);
2700 /* If a reset cycle isn't already in progress, we set a bit in
2701 * pf->state so that the service task can start a reset/rebuild.
2703 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
2704 if (reset == ICE_RESET_CORER)
2705 set_bit(ICE_CORER_RECV, pf->state);
2706 else if (reset == ICE_RESET_GLOBR)
2707 set_bit(ICE_GLOBR_RECV, pf->state);
2709 set_bit(ICE_EMPR_RECV, pf->state);
2711 /* There are couple of different bits at play here.
2712 * hw->reset_ongoing indicates whether the hardware is
2713 * in reset. This is set to true when a reset interrupt
2714 * is received and set back to false after the driver
2715 * has determined that the hardware is out of reset.
2717 * ICE_RESET_OICR_RECV in pf->state indicates
2718 * that a post reset rebuild is required before the
2719 * driver is operational again. This is set above.
2721 * As this is the start of the reset/rebuild cycle, set
2722 * both to indicate that.
2724 hw->reset_ongoing = true;
2728 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
2729 if (oicr & ICE_AUX_CRIT_ERR) {
2730 struct iidc_event *event;
2732 ena_mask &= ~ICE_AUX_CRIT_ERR;
2733 event = kzalloc(sizeof(*event), GFP_KERNEL);
2735 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2736 /* report the entire OICR value to AUX driver */
2738 ice_send_event_to_aux(pf, event);
2743 /* Report any remaining unexpected interrupts */
2746 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2747 /* If a critical error is pending there is no choice but to
2750 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
2751 PFINT_OICR_ECC_ERR_M)) {
2752 set_bit(ICE_PFR_REQ, pf->state);
2753 ice_service_task_schedule(pf);
2758 ice_service_task_schedule(pf);
2759 ice_irq_dynamic_ena(hw, NULL, NULL);
2765 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2766 * @hw: pointer to HW structure
2768 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2770 /* disable Admin queue Interrupt causes */
2771 wr32(hw, PFINT_FW_CTL,
2772 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2774 /* disable Mailbox queue Interrupt causes */
2775 wr32(hw, PFINT_MBX_CTL,
2776 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2778 /* disable Control queue Interrupt causes */
2779 wr32(hw, PFINT_OICR_CTL,
2780 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2786 * ice_free_irq_msix_misc - Unroll misc vector setup
2787 * @pf: board private structure
2789 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2791 struct ice_hw *hw = &pf->hw;
2793 ice_dis_ctrlq_interrupts(hw);
2795 /* disable OICR interrupt */
2796 wr32(hw, PFINT_OICR_ENA, 0);
2799 if (pf->msix_entries) {
2800 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2801 devm_free_irq(ice_pf_to_dev(pf),
2802 pf->msix_entries[pf->oicr_idx].vector, pf);
2805 pf->num_avail_sw_msix += 1;
2806 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2810 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2811 * @hw: pointer to HW structure
2812 * @reg_idx: HW vector index to associate the control queue interrupts with
2814 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2818 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2819 PFINT_OICR_CTL_CAUSE_ENA_M);
2820 wr32(hw, PFINT_OICR_CTL, val);
2822 /* enable Admin queue Interrupt causes */
2823 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2824 PFINT_FW_CTL_CAUSE_ENA_M);
2825 wr32(hw, PFINT_FW_CTL, val);
2827 /* enable Mailbox queue Interrupt causes */
2828 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2829 PFINT_MBX_CTL_CAUSE_ENA_M);
2830 wr32(hw, PFINT_MBX_CTL, val);
2836 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2837 * @pf: board private structure
2839 * This sets up the handler for MSIX 0, which is used to manage the
2840 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2841 * when in MSI or Legacy interrupt mode.
2843 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2845 struct device *dev = ice_pf_to_dev(pf);
2846 struct ice_hw *hw = &pf->hw;
2847 int oicr_idx, err = 0;
2849 if (!pf->int_name[0])
2850 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2851 dev_driver_string(dev), dev_name(dev));
2853 /* Do not request IRQ but do enable OICR interrupt since settings are
2854 * lost during reset. Note that this function is called only during
2855 * rebuild path and not while reset is in progress.
2857 if (ice_is_reset_in_progress(pf->state))
2860 /* reserve one vector in irq_tracker for misc interrupts */
2861 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2865 pf->num_avail_sw_msix -= 1;
2866 pf->oicr_idx = (u16)oicr_idx;
2868 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2869 ice_misc_intr, 0, pf->int_name, pf);
2871 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2873 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2874 pf->num_avail_sw_msix += 1;
2879 ice_ena_misc_vector(pf);
2881 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2882 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2883 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2886 ice_irq_dynamic_ena(hw, NULL, NULL);
2892 * ice_napi_add - register NAPI handler for the VSI
2893 * @vsi: VSI for which NAPI handler is to be registered
2895 * This function is only called in the driver's load path. Registering the NAPI
2896 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2897 * reset/rebuild, etc.)
2899 static void ice_napi_add(struct ice_vsi *vsi)
2906 ice_for_each_q_vector(vsi, v_idx)
2907 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2908 ice_napi_poll, NAPI_POLL_WEIGHT);
2912 * ice_set_ops - set netdev and ethtools ops for the given netdev
2913 * @netdev: netdev instance
2915 static void ice_set_ops(struct net_device *netdev)
2917 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2919 if (ice_is_safe_mode(pf)) {
2920 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2921 ice_set_ethtool_safe_mode_ops(netdev);
2925 netdev->netdev_ops = &ice_netdev_ops;
2926 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
2927 ice_set_ethtool_ops(netdev);
2931 * ice_set_netdev_features - set features for the given netdev
2932 * @netdev: netdev instance
2934 static void ice_set_netdev_features(struct net_device *netdev)
2936 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2937 netdev_features_t csumo_features;
2938 netdev_features_t vlano_features;
2939 netdev_features_t dflt_features;
2940 netdev_features_t tso_features;
2942 if (ice_is_safe_mode(pf)) {
2944 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2945 netdev->hw_features = netdev->features;
2949 dflt_features = NETIF_F_SG |
2954 csumo_features = NETIF_F_RXCSUM |
2959 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2960 NETIF_F_HW_VLAN_CTAG_TX |
2961 NETIF_F_HW_VLAN_CTAG_RX;
2963 tso_features = NETIF_F_TSO |
2967 NETIF_F_GSO_UDP_TUNNEL |
2968 NETIF_F_GSO_GRE_CSUM |
2969 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2970 NETIF_F_GSO_PARTIAL |
2971 NETIF_F_GSO_IPXIP4 |
2972 NETIF_F_GSO_IPXIP6 |
2975 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2976 NETIF_F_GSO_GRE_CSUM;
2977 /* set features that user can change */
2978 netdev->hw_features = dflt_features | csumo_features |
2979 vlano_features | tso_features;
2981 /* add support for HW_CSUM on packets with MPLS header */
2982 netdev->mpls_features = NETIF_F_HW_CSUM;
2984 /* enable features */
2985 netdev->features |= netdev->hw_features;
2986 /* encap and VLAN devices inherit default, csumo and tso features */
2987 netdev->hw_enc_features |= dflt_features | csumo_features |
2989 netdev->vlan_features |= dflt_features | csumo_features |
2994 * ice_cfg_netdev - Allocate, configure and register a netdev
2995 * @vsi: the VSI associated with the new netdev
2997 * Returns 0 on success, negative value on failure
2999 static int ice_cfg_netdev(struct ice_vsi *vsi)
3001 struct ice_pf *pf = vsi->back;
3002 struct ice_netdev_priv *np;
3003 struct net_device *netdev;
3004 u8 mac_addr[ETH_ALEN];
3006 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3011 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3012 vsi->netdev = netdev;
3013 np = netdev_priv(netdev);
3016 ice_set_netdev_features(netdev);
3018 ice_set_ops(netdev);
3020 if (vsi->type == ICE_VSI_PF) {
3021 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
3022 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3023 ether_addr_copy(netdev->dev_addr, mac_addr);
3024 ether_addr_copy(netdev->perm_addr, mac_addr);
3027 netdev->priv_flags |= IFF_UNICAST_FLT;
3029 /* Setup netdev TC information */
3030 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3032 /* setup watchdog timeout value to be 5 second */
3033 netdev->watchdog_timeo = 5 * HZ;
3035 netdev->min_mtu = ETH_MIN_MTU;
3036 netdev->max_mtu = ICE_MAX_MTU;
3042 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3043 * @lut: Lookup table
3044 * @rss_table_size: Lookup table size
3045 * @rss_size: Range of queue number for hashing
3047 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3051 for (i = 0; i < rss_table_size; i++)
3052 lut[i] = i % rss_size;
3056 * ice_pf_vsi_setup - Set up a PF VSI
3057 * @pf: board private structure
3058 * @pi: pointer to the port_info instance
3060 * Returns pointer to the successfully allocated VSI software struct
3061 * on success, otherwise returns NULL on failure.
3063 static struct ice_vsi *
3064 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3066 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3070 * ice_ctrl_vsi_setup - Set up a control VSI
3071 * @pf: board private structure
3072 * @pi: pointer to the port_info instance
3074 * Returns pointer to the successfully allocated VSI software struct
3075 * on success, otherwise returns NULL on failure.
3077 static struct ice_vsi *
3078 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3080 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3084 * ice_lb_vsi_setup - Set up a loopback VSI
3085 * @pf: board private structure
3086 * @pi: pointer to the port_info instance
3088 * Returns pointer to the successfully allocated VSI software struct
3089 * on success, otherwise returns NULL on failure.
3092 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3094 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3098 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3099 * @netdev: network interface to be adjusted
3100 * @proto: unused protocol
3101 * @vid: VLAN ID to be added
3103 * net_device_ops implementation for adding VLAN IDs
3106 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3109 struct ice_netdev_priv *np = netdev_priv(netdev);
3110 struct ice_vsi *vsi = np->vsi;
3113 /* VLAN 0 is added by default during load/reset */
3117 /* Enable VLAN pruning when a VLAN other than 0 is added */
3118 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3119 ret = ice_cfg_vlan_pruning(vsi, true, false);
3124 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3125 * packets aren't pruned by the device's internal switch on Rx
3127 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3129 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3135 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3136 * @netdev: network interface to be adjusted
3137 * @proto: unused protocol
3138 * @vid: VLAN ID to be removed
3140 * net_device_ops implementation for removing VLAN IDs
3143 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3146 struct ice_netdev_priv *np = netdev_priv(netdev);
3147 struct ice_vsi *vsi = np->vsi;
3150 /* don't allow removal of VLAN 0 */
3154 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3157 ret = ice_vsi_kill_vlan(vsi, vid);
3161 /* Disable pruning when VLAN 0 is the only VLAN rule */
3162 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3163 ret = ice_cfg_vlan_pruning(vsi, false, false);
3165 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3170 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3171 * @pf: board private structure
3173 * Returns 0 on success, negative value on failure
3175 static int ice_setup_pf_sw(struct ice_pf *pf)
3177 struct ice_vsi *vsi;
3180 if (ice_is_reset_in_progress(pf->state))
3183 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3187 status = ice_cfg_netdev(vsi);
3190 goto unroll_vsi_setup;
3192 /* netdev has to be configured before setting frame size */
3193 ice_vsi_cfg_frame_size(vsi);
3195 /* Setup DCB netlink interface */
3196 ice_dcbnl_setup(vsi);
3198 /* registering the NAPI handler requires both the queues and
3199 * netdev to be created, which are done in ice_pf_vsi_setup()
3200 * and ice_cfg_netdev() respectively
3204 status = ice_set_cpu_rx_rmap(vsi);
3206 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3207 vsi->vsi_num, status);
3209 goto unroll_napi_add;
3211 status = ice_init_mac_fltr(pf);
3213 goto free_cpu_rx_map;
3218 ice_free_cpu_rx_rmap(vsi);
3224 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3225 free_netdev(vsi->netdev);
3231 ice_vsi_release(vsi);
3236 * ice_get_avail_q_count - Get count of queues in use
3237 * @pf_qmap: bitmap to get queue use count from
3238 * @lock: pointer to a mutex that protects access to pf_qmap
3239 * @size: size of the bitmap
3242 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3248 for_each_clear_bit(bit, pf_qmap, size)
3256 * ice_get_avail_txq_count - Get count of Tx queues in use
3257 * @pf: pointer to an ice_pf instance
3259 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3261 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3266 * ice_get_avail_rxq_count - Get count of Rx queues in use
3267 * @pf: pointer to an ice_pf instance
3269 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3271 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3276 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3277 * @pf: board private structure to initialize
3279 static void ice_deinit_pf(struct ice_pf *pf)
3281 ice_service_task_stop(pf);
3282 mutex_destroy(&pf->sw_mutex);
3283 mutex_destroy(&pf->tc_mutex);
3284 mutex_destroy(&pf->avail_q_mutex);
3286 if (pf->avail_txqs) {
3287 bitmap_free(pf->avail_txqs);
3288 pf->avail_txqs = NULL;
3291 if (pf->avail_rxqs) {
3292 bitmap_free(pf->avail_rxqs);
3293 pf->avail_rxqs = NULL;
3298 * ice_set_pf_caps - set PFs capability flags
3299 * @pf: pointer to the PF instance
3301 static void ice_set_pf_caps(struct ice_pf *pf)
3303 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3305 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3306 clear_bit(ICE_FLAG_AUX_ENA, pf->flags);
3307 if (func_caps->common_cap.rdma) {
3308 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3309 set_bit(ICE_FLAG_AUX_ENA, pf->flags);
3311 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3312 if (func_caps->common_cap.dcb)
3313 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3314 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3315 if (func_caps->common_cap.sr_iov_1_1) {
3316 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3317 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3320 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3321 if (func_caps->common_cap.rss_table_size)
3322 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3324 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3325 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3328 /* ctrl_vsi_idx will be set to a valid value when flow director
3329 * is setup by ice_init_fdir
3331 pf->ctrl_vsi_idx = ICE_NO_VSI;
3332 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3333 /* force guaranteed filter pool for PF */
3334 ice_alloc_fd_guar_item(&pf->hw, &unused,
3335 func_caps->fd_fltr_guar);
3336 /* force shared filter pool for PF */
3337 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3338 func_caps->fd_fltr_best_effort);
3341 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3342 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3346 * ice_init_pf - Initialize general software structures (struct ice_pf)
3347 * @pf: board private structure to initialize
3349 static int ice_init_pf(struct ice_pf *pf)
3351 ice_set_pf_caps(pf);
3353 mutex_init(&pf->sw_mutex);
3354 mutex_init(&pf->tc_mutex);
3356 INIT_HLIST_HEAD(&pf->aq_wait_list);
3357 spin_lock_init(&pf->aq_wait_lock);
3358 init_waitqueue_head(&pf->aq_wait_queue);
3360 /* setup service timer and periodic service task */
3361 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3362 pf->serv_tmr_period = HZ;
3363 INIT_WORK(&pf->serv_task, ice_service_task);
3364 clear_bit(ICE_SERVICE_SCHED, pf->state);
3366 mutex_init(&pf->avail_q_mutex);
3367 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3368 if (!pf->avail_txqs)
3371 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3372 if (!pf->avail_rxqs) {
3373 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3374 pf->avail_txqs = NULL;
3382 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3383 * @pf: board private structure
3385 * compute the number of MSIX vectors required (v_budget) and request from
3386 * the OS. Return the number of vectors reserved or negative on failure
3388 static int ice_ena_msix_range(struct ice_pf *pf)
3390 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3391 struct device *dev = ice_pf_to_dev(pf);
3394 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3395 num_cpus = num_online_cpus();
3397 /* reserve for LAN miscellaneous handler */
3398 needed = ICE_MIN_LAN_OICR_MSIX;
3399 if (v_left < needed)
3400 goto no_hw_vecs_left_err;
3404 /* reserve for flow director */
3405 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3406 needed = ICE_FDIR_MSIX;
3407 if (v_left < needed)
3408 goto no_hw_vecs_left_err;
3413 /* total used for non-traffic vectors */
3416 /* reserve vectors for LAN traffic */
3418 if (v_left < needed)
3419 goto no_hw_vecs_left_err;
3420 pf->num_lan_msix = needed;
3424 /* reserve vectors for RDMA auxiliary driver */
3425 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3426 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3427 if (v_left < needed)
3428 goto no_hw_vecs_left_err;
3429 pf->num_rdma_msix = needed;
3434 pf->msix_entries = devm_kcalloc(dev, v_budget,
3435 sizeof(*pf->msix_entries), GFP_KERNEL);
3436 if (!pf->msix_entries) {
3441 for (i = 0; i < v_budget; i++)
3442 pf->msix_entries[i].entry = i;
3444 /* actually reserve the vectors */
3445 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3446 ICE_MIN_MSIX, v_budget);
3448 dev_err(dev, "unable to reserve MSI-X vectors\n");
3453 if (v_actual < v_budget) {
3454 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3455 v_budget, v_actual);
3457 if (v_actual < ICE_MIN_MSIX) {
3458 /* error if we can't get minimum vectors */
3459 pci_disable_msix(pf->pdev);
3463 int v_remain = v_actual - v_other;
3464 int v_rdma = 0, v_min_rdma = 0;
3466 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3467 /* Need at least 1 interrupt in addition to
3470 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3471 v_min_rdma = ICE_MIN_RDMA_MSIX;
3474 if (v_actual == ICE_MIN_MSIX ||
3475 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
3476 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
3477 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3479 pf->num_rdma_msix = 0;
3480 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3481 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3482 (v_remain - v_rdma < v_rdma)) {
3483 /* Support minimum RDMA and give remaining
3484 * vectors to LAN MSIX
3486 pf->num_rdma_msix = v_min_rdma;
3487 pf->num_lan_msix = v_remain - v_min_rdma;
3489 /* Split remaining MSIX with RDMA after
3490 * accounting for AEQ MSIX
3492 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3493 ICE_RDMA_NUM_AEQ_MSIX;
3494 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3497 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3500 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
3501 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
3509 devm_kfree(dev, pf->msix_entries);
3512 no_hw_vecs_left_err:
3513 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3517 pf->num_rdma_msix = 0;
3518 pf->num_lan_msix = 0;
3523 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3524 * @pf: board private structure
3526 static void ice_dis_msix(struct ice_pf *pf)
3528 pci_disable_msix(pf->pdev);
3529 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3530 pf->msix_entries = NULL;
3534 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3535 * @pf: board private structure
3537 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3541 if (pf->irq_tracker) {
3542 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3543 pf->irq_tracker = NULL;
3548 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3549 * @pf: board private structure to initialize
3551 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3555 vectors = ice_ena_msix_range(pf);
3560 /* set up vector assignment tracking */
3561 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3562 struct_size(pf->irq_tracker, list, vectors),
3564 if (!pf->irq_tracker) {
3569 /* populate SW interrupts pool with number of OS granted IRQs. */
3570 pf->num_avail_sw_msix = (u16)vectors;
3571 pf->irq_tracker->num_entries = (u16)vectors;
3572 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3578 * ice_is_wol_supported - check if WoL is supported
3579 * @hw: pointer to hardware info
3581 * Check if WoL is supported based on the HW configuration.
3582 * Returns true if NVM supports and enables WoL for this port, false otherwise
3584 bool ice_is_wol_supported(struct ice_hw *hw)
3588 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3589 * word) indicates WoL is not supported on the corresponding PF ID.
3591 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3594 return !(BIT(hw->port_info->lport) & wol_ctrl);
3598 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3599 * @vsi: VSI being changed
3600 * @new_rx: new number of Rx queues
3601 * @new_tx: new number of Tx queues
3603 * Only change the number of queues if new_tx, or new_rx is non-0.
3605 * Returns 0 on success.
3607 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3609 struct ice_pf *pf = vsi->back;
3610 int err = 0, timeout = 50;
3612 if (!new_rx && !new_tx)
3615 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3619 usleep_range(1000, 2000);
3623 vsi->req_txq = (u16)new_tx;
3625 vsi->req_rxq = (u16)new_rx;
3627 /* set for the next time the netdev is started */
3628 if (!netif_running(vsi->netdev)) {
3629 ice_vsi_rebuild(vsi, false);
3630 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3635 ice_vsi_rebuild(vsi, false);
3636 ice_pf_dcb_recfg(pf);
3639 clear_bit(ICE_CFG_BUSY, pf->state);
3644 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3645 * @pf: PF to configure
3647 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3648 * VSI can still Tx/Rx VLAN tagged packets.
3650 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3652 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3653 struct ice_vsi_ctx *ctxt;
3654 enum ice_status status;
3660 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3665 ctxt->info = vsi->info;
3667 ctxt->info.valid_sections =
3668 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3669 ICE_AQ_VSI_PROP_SECURITY_VALID |
3670 ICE_AQ_VSI_PROP_SW_VALID);
3672 /* disable VLAN anti-spoof */
3673 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3674 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3676 /* disable VLAN pruning and keep all other settings */
3677 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3679 /* allow all VLANs on Tx and don't strip on Rx */
3680 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3681 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3683 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3685 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3686 ice_stat_str(status),
3687 ice_aq_str(hw->adminq.sq_last_status));
3689 vsi->info.sec_flags = ctxt->info.sec_flags;
3690 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3691 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3698 * ice_log_pkg_init - log result of DDP package load
3699 * @hw: pointer to hardware info
3700 * @status: status of package load
3703 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3705 struct ice_pf *pf = (struct ice_pf *)hw->back;
3706 struct device *dev = ice_pf_to_dev(pf);
3710 /* The package download AdminQ command returned success because
3711 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3712 * already a package loaded on the device.
3714 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3715 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3716 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3717 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3718 !memcmp(hw->pkg_name, hw->active_pkg_name,
3719 sizeof(hw->pkg_name))) {
3720 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3721 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3722 hw->active_pkg_name,
3723 hw->active_pkg_ver.major,
3724 hw->active_pkg_ver.minor,
3725 hw->active_pkg_ver.update,
3726 hw->active_pkg_ver.draft);
3728 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3729 hw->active_pkg_name,
3730 hw->active_pkg_ver.major,
3731 hw->active_pkg_ver.minor,
3732 hw->active_pkg_ver.update,
3733 hw->active_pkg_ver.draft);
3734 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3735 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3736 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",
3737 hw->active_pkg_name,
3738 hw->active_pkg_ver.major,
3739 hw->active_pkg_ver.minor,
3740 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3741 *status = ICE_ERR_NOT_SUPPORTED;
3742 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3743 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3744 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",
3745 hw->active_pkg_name,
3746 hw->active_pkg_ver.major,
3747 hw->active_pkg_ver.minor,
3748 hw->active_pkg_ver.update,
3749 hw->active_pkg_ver.draft,
3756 dev_err(dev, "An unknown error occurred when loading the DDP package, please reboot the system. If the problem persists, update the NVM. Entering Safe Mode.\n");
3757 *status = ICE_ERR_NOT_SUPPORTED;
3760 case ICE_ERR_FW_DDP_MISMATCH:
3761 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");
3763 case ICE_ERR_BUF_TOO_SHORT:
3765 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3767 case ICE_ERR_NOT_SUPPORTED:
3768 /* Package File version not supported */
3769 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3770 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3771 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3772 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3773 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3774 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3775 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3776 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",
3777 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3779 case ICE_ERR_AQ_ERROR:
3780 switch (hw->pkg_dwnld_status) {
3781 case ICE_AQ_RC_ENOSEC:
3782 case ICE_AQ_RC_EBADSIG:
3783 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");
3785 case ICE_AQ_RC_ESVN:
3786 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");
3788 case ICE_AQ_RC_EBADMAN:
3789 case ICE_AQ_RC_EBADBUF:
3790 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3791 /* poll for reset to complete */
3792 if (ice_check_reset(hw))
3793 dev_err(dev, "Error resetting device. Please reload the driver\n");
3800 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3807 * ice_load_pkg - load/reload the DDP Package file
3808 * @firmware: firmware structure when firmware requested or NULL for reload
3809 * @pf: pointer to the PF instance
3811 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3812 * initialize HW tables.
3815 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3817 enum ice_status status = ICE_ERR_PARAM;
3818 struct device *dev = ice_pf_to_dev(pf);
3819 struct ice_hw *hw = &pf->hw;
3821 /* Load DDP Package */
3822 if (firmware && !hw->pkg_copy) {
3823 status = ice_copy_and_init_pkg(hw, firmware->data,
3825 ice_log_pkg_init(hw, &status);
3826 } else if (!firmware && hw->pkg_copy) {
3827 /* Reload package during rebuild after CORER/GLOBR reset */
3828 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3829 ice_log_pkg_init(hw, &status);
3831 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3836 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3840 /* Successful download package is the precondition for advanced
3841 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3843 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3847 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3848 * @pf: pointer to the PF structure
3850 * There is no error returned here because the driver should be able to handle
3851 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3852 * specifically with Tx.
3854 static void ice_verify_cacheline_size(struct ice_pf *pf)
3856 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3857 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3858 ICE_CACHE_LINE_BYTES);
3862 * ice_send_version - update firmware with driver version
3865 * Returns ICE_SUCCESS on success, else error code
3867 static enum ice_status ice_send_version(struct ice_pf *pf)
3869 struct ice_driver_ver dv;
3871 dv.major_ver = 0xff;
3872 dv.minor_ver = 0xff;
3873 dv.build_ver = 0xff;
3874 dv.subbuild_ver = 0;
3875 strscpy((char *)dv.driver_string, UTS_RELEASE,
3876 sizeof(dv.driver_string));
3877 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3881 * ice_init_fdir - Initialize flow director VSI and configuration
3882 * @pf: pointer to the PF instance
3884 * returns 0 on success, negative on error
3886 static int ice_init_fdir(struct ice_pf *pf)
3888 struct device *dev = ice_pf_to_dev(pf);
3889 struct ice_vsi *ctrl_vsi;
3892 /* Side Band Flow Director needs to have a control VSI.
3893 * Allocate it and store it in the PF.
3895 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3897 dev_dbg(dev, "could not create control VSI\n");
3901 err = ice_vsi_open_ctrl(ctrl_vsi);
3903 dev_dbg(dev, "could not open control VSI\n");
3907 mutex_init(&pf->hw.fdir_fltr_lock);
3909 err = ice_fdir_create_dflt_rules(pf);
3916 ice_fdir_release_flows(&pf->hw);
3917 ice_vsi_close(ctrl_vsi);
3919 ice_vsi_release(ctrl_vsi);
3920 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3921 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3922 pf->ctrl_vsi_idx = ICE_NO_VSI;
3928 * ice_get_opt_fw_name - return optional firmware file name or NULL
3929 * @pf: pointer to the PF instance
3931 static char *ice_get_opt_fw_name(struct ice_pf *pf)
3933 /* Optional firmware name same as default with additional dash
3934 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3936 struct pci_dev *pdev = pf->pdev;
3937 char *opt_fw_filename;
3940 /* Determine the name of the optional file using the DSN (two
3941 * dwords following the start of the DSN Capability).
3943 dsn = pci_get_dsn(pdev);
3947 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3948 if (!opt_fw_filename)
3951 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3952 ICE_DDP_PKG_PATH, dsn);
3954 return opt_fw_filename;
3958 * ice_request_fw - Device initialization routine
3959 * @pf: pointer to the PF instance
3961 static void ice_request_fw(struct ice_pf *pf)
3963 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3964 const struct firmware *firmware = NULL;
3965 struct device *dev = ice_pf_to_dev(pf);
3968 /* optional device-specific DDP (if present) overrides the default DDP
3969 * package file. kernel logs a debug message if the file doesn't exist,
3970 * and warning messages for other errors.
3972 if (opt_fw_filename) {
3973 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3975 kfree(opt_fw_filename);
3979 /* request for firmware was successful. Download to device */
3980 ice_load_pkg(firmware, pf);
3981 kfree(opt_fw_filename);
3982 release_firmware(firmware);
3987 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3989 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3993 /* request for firmware was successful. Download to device */
3994 ice_load_pkg(firmware, pf);
3995 release_firmware(firmware);
3999 * ice_print_wake_reason - show the wake up cause in the log
4000 * @pf: pointer to the PF struct
4002 static void ice_print_wake_reason(struct ice_pf *pf)
4004 u32 wus = pf->wakeup_reason;
4005 const char *wake_str;
4007 /* if no wake event, nothing to print */
4011 if (wus & PFPM_WUS_LNKC_M)
4012 wake_str = "Link\n";
4013 else if (wus & PFPM_WUS_MAG_M)
4014 wake_str = "Magic Packet\n";
4015 else if (wus & PFPM_WUS_MNG_M)
4016 wake_str = "Management\n";
4017 else if (wus & PFPM_WUS_FW_RST_WK_M)
4018 wake_str = "Firmware Reset\n";
4020 wake_str = "Unknown\n";
4022 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4026 * ice_register_netdev - register netdev and devlink port
4027 * @pf: pointer to the PF struct
4029 static int ice_register_netdev(struct ice_pf *pf)
4031 struct ice_vsi *vsi;
4034 vsi = ice_get_main_vsi(pf);
4035 if (!vsi || !vsi->netdev)
4038 err = register_netdev(vsi->netdev);
4040 goto err_register_netdev;
4042 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4043 netif_carrier_off(vsi->netdev);
4044 netif_tx_stop_all_queues(vsi->netdev);
4045 err = ice_devlink_create_port(vsi);
4047 goto err_devlink_create;
4049 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
4053 unregister_netdev(vsi->netdev);
4054 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4055 err_register_netdev:
4056 free_netdev(vsi->netdev);
4058 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4063 * ice_probe - Device initialization routine
4064 * @pdev: PCI device information struct
4065 * @ent: entry in ice_pci_tbl
4067 * Returns 0 on success, negative on failure
4070 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4072 struct device *dev = &pdev->dev;
4077 /* this driver uses devres, see
4078 * Documentation/driver-api/driver-model/devres.rst
4080 err = pcim_enable_device(pdev);
4084 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4086 dev_err(dev, "BAR0 I/O map error %d\n", err);
4090 pf = ice_allocate_pf(dev);
4094 /* set up for high or low DMA */
4095 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4097 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4099 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4103 pci_enable_pcie_error_reporting(pdev);
4104 pci_set_master(pdev);
4107 pci_set_drvdata(pdev, pf);
4108 set_bit(ICE_DOWN, pf->state);
4109 /* Disable service task until DOWN bit is cleared */
4110 set_bit(ICE_SERVICE_DIS, pf->state);
4113 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4114 pci_save_state(pdev);
4117 hw->vendor_id = pdev->vendor;
4118 hw->device_id = pdev->device;
4119 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4120 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4121 hw->subsystem_device_id = pdev->subsystem_device;
4122 hw->bus.device = PCI_SLOT(pdev->devfn);
4123 hw->bus.func = PCI_FUNC(pdev->devfn);
4124 ice_set_ctrlq_len(hw);
4126 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4128 err = ice_devlink_register(pf);
4130 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4131 goto err_exit_unroll;
4134 #ifndef CONFIG_DYNAMIC_DEBUG
4136 hw->debug_mask = debug;
4139 err = ice_init_hw(hw);
4141 dev_err(dev, "ice_init_hw failed: %d\n", err);
4143 goto err_exit_unroll;
4148 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4149 * set in pf->state, which will cause ice_is_safe_mode to return
4152 if (ice_is_safe_mode(pf)) {
4153 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4154 /* we already got function/device capabilities but these don't
4155 * reflect what the driver needs to do in safe mode. Instead of
4156 * adding conditional logic everywhere to ignore these
4157 * device/function capabilities, override them.
4159 ice_set_safe_mode_caps(hw);
4162 err = ice_init_pf(pf);
4164 dev_err(dev, "ice_init_pf failed: %d\n", err);
4165 goto err_init_pf_unroll;
4168 ice_devlink_init_regions(pf);
4170 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4171 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4172 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4173 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4175 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4176 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4177 pf->hw.tnl.valid_count[TNL_VXLAN];
4178 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4179 UDP_TUNNEL_TYPE_VXLAN;
4182 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4183 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4184 pf->hw.tnl.valid_count[TNL_GENEVE];
4185 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4186 UDP_TUNNEL_TYPE_GENEVE;
4190 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4191 if (!pf->num_alloc_vsi) {
4193 goto err_init_pf_unroll;
4195 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4196 dev_warn(&pf->pdev->dev,
4197 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4198 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4199 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4202 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4206 goto err_init_pf_unroll;
4209 err = ice_init_interrupt_scheme(pf);
4211 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4213 goto err_init_vsi_unroll;
4216 /* In case of MSIX we are going to setup the misc vector right here
4217 * to handle admin queue events etc. In case of legacy and MSI
4218 * the misc functionality and queue processing is combined in
4219 * the same vector and that gets setup at open.
4221 err = ice_req_irq_msix_misc(pf);
4223 dev_err(dev, "setup of misc vector failed: %d\n", err);
4224 goto err_init_interrupt_unroll;
4227 /* create switch struct for the switch element created by FW on boot */
4228 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4229 if (!pf->first_sw) {
4231 goto err_msix_misc_unroll;
4235 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4237 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4239 pf->first_sw->pf = pf;
4241 /* record the sw_id available for later use */
4242 pf->first_sw->sw_id = hw->port_info->sw_id;
4244 err = ice_setup_pf_sw(pf);
4246 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4247 goto err_alloc_sw_unroll;
4250 clear_bit(ICE_SERVICE_DIS, pf->state);
4252 /* tell the firmware we are up */
4253 err = ice_send_version(pf);
4255 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4257 goto err_send_version_unroll;
4260 /* since everything is good, start the service timer */
4261 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4263 err = ice_init_link_events(pf->hw.port_info);
4265 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4266 goto err_send_version_unroll;
4269 /* not a fatal error if this fails */
4270 err = ice_init_nvm_phy_type(pf->hw.port_info);
4272 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4274 /* not a fatal error if this fails */
4275 err = ice_update_link_info(pf->hw.port_info);
4277 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4279 ice_init_link_dflt_override(pf->hw.port_info);
4281 /* if media available, initialize PHY settings */
4282 if (pf->hw.port_info->phy.link_info.link_info &
4283 ICE_AQ_MEDIA_AVAILABLE) {
4284 /* not a fatal error if this fails */
4285 err = ice_init_phy_user_cfg(pf->hw.port_info);
4287 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4289 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4290 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4293 ice_configure_phy(vsi);
4296 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4299 ice_verify_cacheline_size(pf);
4301 /* Save wakeup reason register for later use */
4302 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4304 /* check for a power management event */
4305 ice_print_wake_reason(pf);
4307 /* clear wake status, all bits */
4308 wr32(hw, PFPM_WUS, U32_MAX);
4310 /* Disable WoL at init, wait for user to enable */
4311 device_set_wakeup_enable(dev, false);
4313 if (ice_is_safe_mode(pf)) {
4314 ice_set_safe_mode_vlan_cfg(pf);
4318 /* initialize DDP driven features */
4320 /* Note: Flow director init failure is non-fatal to load */
4321 if (ice_init_fdir(pf))
4322 dev_err(dev, "could not initialize flow director\n");
4324 /* Note: DCB init failure is non-fatal to load */
4325 if (ice_init_pf_dcb(pf, false)) {
4326 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4327 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4329 ice_cfg_lldp_mib_change(&pf->hw, true);
4332 if (ice_init_lag(pf))
4333 dev_warn(dev, "Failed to init link aggregation support\n");
4335 /* print PCI link speed and width */
4336 pcie_print_link_status(pf->pdev);
4339 err = ice_register_netdev(pf);
4341 goto err_netdev_reg;
4343 /* ready to go, so clear down state bit */
4344 clear_bit(ICE_DOWN, pf->state);
4345 if (ice_is_aux_ena(pf)) {
4346 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4347 if (pf->aux_idx < 0) {
4348 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4350 goto err_netdev_reg;
4353 err = ice_init_rdma(pf);
4355 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4357 goto err_init_aux_unroll;
4360 dev_warn(dev, "RDMA is not supported on this device\n");
4365 err_init_aux_unroll:
4367 ida_free(&ice_aux_ida, pf->aux_idx);
4369 err_send_version_unroll:
4370 ice_vsi_release_all(pf);
4371 err_alloc_sw_unroll:
4372 set_bit(ICE_SERVICE_DIS, pf->state);
4373 set_bit(ICE_DOWN, pf->state);
4374 devm_kfree(dev, pf->first_sw);
4375 err_msix_misc_unroll:
4376 ice_free_irq_msix_misc(pf);
4377 err_init_interrupt_unroll:
4378 ice_clear_interrupt_scheme(pf);
4379 err_init_vsi_unroll:
4380 devm_kfree(dev, pf->vsi);
4383 ice_devlink_destroy_regions(pf);
4386 ice_devlink_unregister(pf);
4387 pci_disable_pcie_error_reporting(pdev);
4388 pci_disable_device(pdev);
4393 * ice_set_wake - enable or disable Wake on LAN
4394 * @pf: pointer to the PF struct
4396 * Simple helper for WoL control
4398 static void ice_set_wake(struct ice_pf *pf)
4400 struct ice_hw *hw = &pf->hw;
4401 bool wol = pf->wol_ena;
4403 /* clear wake state, otherwise new wake events won't fire */
4404 wr32(hw, PFPM_WUS, U32_MAX);
4406 /* enable / disable APM wake up, no RMW needed */
4407 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4409 /* set magic packet filter enabled */
4410 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4414 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4415 * @pf: pointer to the PF struct
4417 * Issue firmware command to enable multicast magic wake, making
4418 * sure that any locally administered address (LAA) is used for
4419 * wake, and that PF reset doesn't undo the LAA.
4421 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4423 struct device *dev = ice_pf_to_dev(pf);
4424 struct ice_hw *hw = &pf->hw;
4425 enum ice_status status;
4426 u8 mac_addr[ETH_ALEN];
4427 struct ice_vsi *vsi;
4433 vsi = ice_get_main_vsi(pf);
4437 /* Get current MAC address in case it's an LAA */
4439 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4441 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4443 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4444 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4445 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4447 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4449 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4450 ice_stat_str(status),
4451 ice_aq_str(hw->adminq.sq_last_status));
4455 * ice_remove - Device removal routine
4456 * @pdev: PCI device information struct
4458 static void ice_remove(struct pci_dev *pdev)
4460 struct ice_pf *pf = pci_get_drvdata(pdev);
4466 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4467 if (!ice_is_reset_in_progress(pf->state))
4472 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4473 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4477 ice_service_task_stop(pf);
4479 ice_aq_cancel_waiting_tasks(pf);
4480 ice_unplug_aux_dev(pf);
4481 ida_free(&ice_aux_ida, pf->aux_idx);
4482 set_bit(ICE_DOWN, pf->state);
4484 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4486 if (!ice_is_safe_mode(pf))
4487 ice_remove_arfs(pf);
4488 ice_setup_mc_magic_wake(pf);
4489 ice_vsi_release_all(pf);
4491 ice_free_irq_msix_misc(pf);
4492 ice_for_each_vsi(pf, i) {
4495 ice_vsi_free_q_vectors(pf->vsi[i]);
4498 ice_devlink_destroy_regions(pf);
4499 ice_deinit_hw(&pf->hw);
4500 ice_devlink_unregister(pf);
4502 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4503 * do it via ice_schedule_reset() since there is no need to rebuild
4504 * and the service task is already stopped.
4506 ice_reset(&pf->hw, ICE_RESET_PFR);
4507 pci_wait_for_pending_transaction(pdev);
4508 ice_clear_interrupt_scheme(pf);
4509 pci_disable_pcie_error_reporting(pdev);
4510 pci_disable_device(pdev);
4514 * ice_shutdown - PCI callback for shutting down device
4515 * @pdev: PCI device information struct
4517 static void ice_shutdown(struct pci_dev *pdev)
4519 struct ice_pf *pf = pci_get_drvdata(pdev);
4523 if (system_state == SYSTEM_POWER_OFF) {
4524 pci_wake_from_d3(pdev, pf->wol_ena);
4525 pci_set_power_state(pdev, PCI_D3hot);
4531 * ice_prepare_for_shutdown - prep for PCI shutdown
4532 * @pf: board private structure
4534 * Inform or close all dependent features in prep for PCI device shutdown
4536 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4538 struct ice_hw *hw = &pf->hw;
4541 /* Notify VFs of impending reset */
4542 if (ice_check_sq_alive(hw, &hw->mailboxq))
4543 ice_vc_notify_reset(pf);
4545 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4547 /* disable the VSIs and their queues that are not already DOWN */
4548 ice_pf_dis_all_vsi(pf, false);
4550 ice_for_each_vsi(pf, v)
4552 pf->vsi[v]->vsi_num = 0;
4554 ice_shutdown_all_ctrlq(hw);
4558 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4559 * @pf: board private structure to reinitialize
4561 * This routine reinitialize interrupt scheme that was cleared during
4562 * power management suspend callback.
4564 * This should be called during resume routine to re-allocate the q_vectors
4565 * and reacquire interrupts.
4567 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4569 struct device *dev = ice_pf_to_dev(pf);
4572 /* Since we clear MSIX flag during suspend, we need to
4573 * set it back during resume...
4576 ret = ice_init_interrupt_scheme(pf);
4578 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4582 /* Remap vectors and rings, after successful re-init interrupts */
4583 ice_for_each_vsi(pf, v) {
4587 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4590 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4593 ret = ice_req_irq_msix_misc(pf);
4595 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4605 ice_vsi_free_q_vectors(pf->vsi[v]);
4612 * @dev: generic device information structure
4614 * Power Management callback to quiesce the device and prepare
4615 * for D3 transition.
4617 static int __maybe_unused ice_suspend(struct device *dev)
4619 struct pci_dev *pdev = to_pci_dev(dev);
4623 pf = pci_get_drvdata(pdev);
4625 if (!ice_pf_state_is_nominal(pf)) {
4626 dev_err(dev, "Device is not ready, no need to suspend it\n");
4630 /* Stop watchdog tasks until resume completion.
4631 * Even though it is most likely that the service task is
4632 * disabled if the device is suspended or down, the service task's
4633 * state is controlled by a different state bit, and we should
4634 * store and honor whatever state that bit is in at this point.
4636 disabled = ice_service_task_stop(pf);
4638 ice_unplug_aux_dev(pf);
4640 /* Already suspended?, then there is nothing to do */
4641 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
4643 ice_service_task_restart(pf);
4647 if (test_bit(ICE_DOWN, pf->state) ||
4648 ice_is_reset_in_progress(pf->state)) {
4649 dev_err(dev, "can't suspend device in reset or already down\n");
4651 ice_service_task_restart(pf);
4655 ice_setup_mc_magic_wake(pf);
4657 ice_prepare_for_shutdown(pf);
4661 /* Free vectors, clear the interrupt scheme and release IRQs
4662 * for proper hibernation, especially with large number of CPUs.
4663 * Otherwise hibernation might fail when mapping all the vectors back
4666 ice_free_irq_msix_misc(pf);
4667 ice_for_each_vsi(pf, v) {
4670 ice_vsi_free_q_vectors(pf->vsi[v]);
4672 ice_free_cpu_rx_rmap(ice_get_main_vsi(pf));
4673 ice_clear_interrupt_scheme(pf);
4675 pci_save_state(pdev);
4676 pci_wake_from_d3(pdev, pf->wol_ena);
4677 pci_set_power_state(pdev, PCI_D3hot);
4682 * ice_resume - PM callback for waking up from D3
4683 * @dev: generic device information structure
4685 static int __maybe_unused ice_resume(struct device *dev)
4687 struct pci_dev *pdev = to_pci_dev(dev);
4688 enum ice_reset_req reset_type;
4693 pci_set_power_state(pdev, PCI_D0);
4694 pci_restore_state(pdev);
4695 pci_save_state(pdev);
4697 if (!pci_device_is_present(pdev))
4700 ret = pci_enable_device_mem(pdev);
4702 dev_err(dev, "Cannot enable device after suspend\n");
4706 pf = pci_get_drvdata(pdev);
4709 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4710 ice_print_wake_reason(pf);
4712 /* We cleared the interrupt scheme when we suspended, so we need to
4713 * restore it now to resume device functionality.
4715 ret = ice_reinit_interrupt_scheme(pf);
4717 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4719 clear_bit(ICE_DOWN, pf->state);
4720 /* Now perform PF reset and rebuild */
4721 reset_type = ICE_RESET_PFR;
4722 /* re-enable service task for reset, but allow reset to schedule it */
4723 clear_bit(ICE_SERVICE_DIS, pf->state);
4725 if (ice_schedule_reset(pf, reset_type))
4726 dev_err(dev, "Reset during resume failed.\n");
4728 clear_bit(ICE_SUSPENDED, pf->state);
4729 ice_service_task_restart(pf);
4731 /* Restart the service task */
4732 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4736 #endif /* CONFIG_PM */
4739 * ice_pci_err_detected - warning that PCI error has been detected
4740 * @pdev: PCI device information struct
4741 * @err: the type of PCI error
4743 * Called to warn that something happened on the PCI bus and the error handling
4744 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4746 static pci_ers_result_t
4747 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4749 struct ice_pf *pf = pci_get_drvdata(pdev);
4752 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4754 return PCI_ERS_RESULT_DISCONNECT;
4757 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4758 ice_service_task_stop(pf);
4760 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4761 set_bit(ICE_PFR_REQ, pf->state);
4762 ice_prepare_for_reset(pf);
4766 return PCI_ERS_RESULT_NEED_RESET;
4770 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4771 * @pdev: PCI device information struct
4773 * Called to determine if the driver can recover from the PCI slot reset by
4774 * using a register read to determine if the device is recoverable.
4776 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4778 struct ice_pf *pf = pci_get_drvdata(pdev);
4779 pci_ers_result_t result;
4783 err = pci_enable_device_mem(pdev);
4785 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4787 result = PCI_ERS_RESULT_DISCONNECT;
4789 pci_set_master(pdev);
4790 pci_restore_state(pdev);
4791 pci_save_state(pdev);
4792 pci_wake_from_d3(pdev, false);
4794 /* Check for life */
4795 reg = rd32(&pf->hw, GLGEN_RTRIG);
4797 result = PCI_ERS_RESULT_RECOVERED;
4799 result = PCI_ERS_RESULT_DISCONNECT;
4802 err = pci_aer_clear_nonfatal_status(pdev);
4804 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4806 /* non-fatal, continue */
4812 * ice_pci_err_resume - restart operations after PCI error recovery
4813 * @pdev: PCI device information struct
4815 * Called to allow the driver to bring things back up after PCI error and/or
4816 * reset recovery have finished
4818 static void ice_pci_err_resume(struct pci_dev *pdev)
4820 struct ice_pf *pf = pci_get_drvdata(pdev);
4823 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4828 if (test_bit(ICE_SUSPENDED, pf->state)) {
4829 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4834 ice_restore_all_vfs_msi_state(pdev);
4836 ice_do_reset(pf, ICE_RESET_PFR);
4837 ice_service_task_restart(pf);
4838 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4842 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4843 * @pdev: PCI device information struct
4845 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4847 struct ice_pf *pf = pci_get_drvdata(pdev);
4849 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4850 ice_service_task_stop(pf);
4852 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4853 set_bit(ICE_PFR_REQ, pf->state);
4854 ice_prepare_for_reset(pf);
4860 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4861 * @pdev: PCI device information struct
4863 static void ice_pci_err_reset_done(struct pci_dev *pdev)
4865 ice_pci_err_resume(pdev);
4868 /* ice_pci_tbl - PCI Device ID Table
4870 * Wildcard entries (PCI_ANY_ID) should come last
4871 * Last entry must be all 0s
4873 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4874 * Class, Class Mask, private data (not used) }
4876 static const struct pci_device_id ice_pci_tbl[] = {
4877 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4878 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4879 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4880 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4881 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4882 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4883 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4884 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4885 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4886 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4887 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4888 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4889 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4890 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4891 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4892 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4893 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4894 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4895 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4896 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4897 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4898 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4899 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4900 /* required last entry */
4903 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4905 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4907 static const struct pci_error_handlers ice_pci_err_handler = {
4908 .error_detected = ice_pci_err_detected,
4909 .slot_reset = ice_pci_err_slot_reset,
4910 .reset_prepare = ice_pci_err_reset_prepare,
4911 .reset_done = ice_pci_err_reset_done,
4912 .resume = ice_pci_err_resume
4915 static struct pci_driver ice_driver = {
4916 .name = KBUILD_MODNAME,
4917 .id_table = ice_pci_tbl,
4919 .remove = ice_remove,
4921 .driver.pm = &ice_pm_ops,
4922 #endif /* CONFIG_PM */
4923 .shutdown = ice_shutdown,
4924 .sriov_configure = ice_sriov_configure,
4925 .err_handler = &ice_pci_err_handler
4929 * ice_module_init - Driver registration routine
4931 * ice_module_init is the first routine called when the driver is
4932 * loaded. All it does is register with the PCI subsystem.
4934 static int __init ice_module_init(void)
4938 pr_info("%s\n", ice_driver_string);
4939 pr_info("%s\n", ice_copyright);
4941 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4943 pr_err("Failed to create workqueue\n");
4947 status = pci_register_driver(&ice_driver);
4949 pr_err("failed to register PCI driver, err %d\n", status);
4950 destroy_workqueue(ice_wq);
4955 module_init(ice_module_init);
4958 * ice_module_exit - Driver exit cleanup routine
4960 * ice_module_exit is called just before the driver is removed
4963 static void __exit ice_module_exit(void)
4965 pci_unregister_driver(&ice_driver);
4966 destroy_workqueue(ice_wq);
4967 pr_info("module unloaded\n");
4969 module_exit(ice_module_exit);
4972 * ice_set_mac_address - NDO callback to set MAC address
4973 * @netdev: network interface device structure
4974 * @pi: pointer to an address structure
4976 * Returns 0 on success, negative on failure
4978 static int ice_set_mac_address(struct net_device *netdev, void *pi)
4980 struct ice_netdev_priv *np = netdev_priv(netdev);
4981 struct ice_vsi *vsi = np->vsi;
4982 struct ice_pf *pf = vsi->back;
4983 struct ice_hw *hw = &pf->hw;
4984 struct sockaddr *addr = pi;
4985 enum ice_status status;
4990 mac = (u8 *)addr->sa_data;
4992 if (!is_valid_ether_addr(mac))
4993 return -EADDRNOTAVAIL;
4995 if (ether_addr_equal(netdev->dev_addr, mac)) {
4996 netdev_warn(netdev, "already using mac %pM\n", mac);
5000 if (test_bit(ICE_DOWN, pf->state) ||
5001 ice_is_reset_in_progress(pf->state)) {
5002 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5007 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5008 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
5009 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
5010 err = -EADDRNOTAVAIL;
5011 goto err_update_filters;
5014 /* Add filter for new MAC. If filter exists, return success */
5015 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5016 if (status == ICE_ERR_ALREADY_EXISTS) {
5017 /* Although this MAC filter is already present in hardware it's
5018 * possible in some cases (e.g. bonding) that dev_addr was
5019 * modified outside of the driver and needs to be restored back
5022 memcpy(netdev->dev_addr, mac, netdev->addr_len);
5023 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5027 /* error if the new filter addition failed */
5029 err = -EADDRNOTAVAIL;
5033 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5038 /* change the netdev's MAC address */
5039 memcpy(netdev->dev_addr, mac, netdev->addr_len);
5040 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5043 /* write new MAC address to the firmware */
5044 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5045 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5047 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
5048 mac, ice_stat_str(status));
5054 * ice_set_rx_mode - NDO callback to set the netdev filters
5055 * @netdev: network interface device structure
5057 static void ice_set_rx_mode(struct net_device *netdev)
5059 struct ice_netdev_priv *np = netdev_priv(netdev);
5060 struct ice_vsi *vsi = np->vsi;
5065 /* Set the flags to synchronize filters
5066 * ndo_set_rx_mode may be triggered even without a change in netdev
5069 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5070 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5071 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5073 /* schedule our worker thread which will take care of
5074 * applying the new filter changes
5076 ice_service_task_schedule(vsi->back);
5080 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5081 * @netdev: network interface device structure
5082 * @queue_index: Queue ID
5083 * @maxrate: maximum bandwidth in Mbps
5086 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5088 struct ice_netdev_priv *np = netdev_priv(netdev);
5089 struct ice_vsi *vsi = np->vsi;
5090 enum ice_status status;
5094 /* Validate maxrate requested is within permitted range */
5095 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5096 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5097 maxrate, queue_index);
5101 q_handle = vsi->tx_rings[queue_index]->q_handle;
5102 tc = ice_dcb_get_tc(vsi, queue_index);
5104 /* Set BW back to default, when user set maxrate to 0 */
5106 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5107 q_handle, ICE_MAX_BW);
5109 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5110 q_handle, ICE_MAX_BW, maxrate * 1000);
5112 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
5113 ice_stat_str(status));
5121 * ice_fdb_add - add an entry to the hardware database
5122 * @ndm: the input from the stack
5123 * @tb: pointer to array of nladdr (unused)
5124 * @dev: the net device pointer
5125 * @addr: the MAC address entry being added
5127 * @flags: instructions from stack about fdb operation
5128 * @extack: netlink extended ack
5131 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5132 struct net_device *dev, const unsigned char *addr, u16 vid,
5133 u16 flags, struct netlink_ext_ack __always_unused *extack)
5138 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5141 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5142 netdev_err(dev, "FDB only supports static addresses\n");
5146 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5147 err = dev_uc_add_excl(dev, addr);
5148 else if (is_multicast_ether_addr(addr))
5149 err = dev_mc_add_excl(dev, addr);
5153 /* Only return duplicate errors if NLM_F_EXCL is set */
5154 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5161 * ice_fdb_del - delete an entry from the hardware database
5162 * @ndm: the input from the stack
5163 * @tb: pointer to array of nladdr (unused)
5164 * @dev: the net device pointer
5165 * @addr: the MAC address entry being added
5169 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5170 struct net_device *dev, const unsigned char *addr,
5171 __always_unused u16 vid)
5175 if (ndm->ndm_state & NUD_PERMANENT) {
5176 netdev_err(dev, "FDB only supports static addresses\n");
5180 if (is_unicast_ether_addr(addr))
5181 err = dev_uc_del(dev, addr);
5182 else if (is_multicast_ether_addr(addr))
5183 err = dev_mc_del(dev, addr);
5191 * ice_set_features - set the netdev feature flags
5192 * @netdev: ptr to the netdev being adjusted
5193 * @features: the feature set that the stack is suggesting
5196 ice_set_features(struct net_device *netdev, netdev_features_t features)
5198 struct ice_netdev_priv *np = netdev_priv(netdev);
5199 struct ice_vsi *vsi = np->vsi;
5200 struct ice_pf *pf = vsi->back;
5203 /* Don't set any netdev advanced features with device in Safe Mode */
5204 if (ice_is_safe_mode(vsi->back)) {
5205 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5209 /* Do not change setting during reset */
5210 if (ice_is_reset_in_progress(pf->state)) {
5211 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5215 /* Multiple features can be changed in one call so keep features in
5216 * separate if/else statements to guarantee each feature is checked
5218 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5219 ice_vsi_manage_rss_lut(vsi, true);
5220 else if (!(features & NETIF_F_RXHASH) &&
5221 netdev->features & NETIF_F_RXHASH)
5222 ice_vsi_manage_rss_lut(vsi, false);
5224 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5225 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5226 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5227 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5228 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5229 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5231 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5232 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5233 ret = ice_vsi_manage_vlan_insertion(vsi);
5234 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5235 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5236 ret = ice_vsi_manage_vlan_insertion(vsi);
5238 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5239 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5240 ret = ice_cfg_vlan_pruning(vsi, true, false);
5241 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5242 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5243 ret = ice_cfg_vlan_pruning(vsi, false, false);
5245 if ((features & NETIF_F_NTUPLE) &&
5246 !(netdev->features & NETIF_F_NTUPLE)) {
5247 ice_vsi_manage_fdir(vsi, true);
5249 } else if (!(features & NETIF_F_NTUPLE) &&
5250 (netdev->features & NETIF_F_NTUPLE)) {
5251 ice_vsi_manage_fdir(vsi, false);
5252 ice_clear_arfs(vsi);
5259 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5260 * @vsi: VSI to setup VLAN properties for
5262 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5266 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5267 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5268 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5269 ret = ice_vsi_manage_vlan_insertion(vsi);
5275 * ice_vsi_cfg - Setup the VSI
5276 * @vsi: the VSI being configured
5278 * Return 0 on success and negative value on error
5280 int ice_vsi_cfg(struct ice_vsi *vsi)
5285 ice_set_rx_mode(vsi->netdev);
5287 err = ice_vsi_vlan_setup(vsi);
5292 ice_vsi_cfg_dcb_rings(vsi);
5294 err = ice_vsi_cfg_lan_txqs(vsi);
5295 if (!err && ice_is_xdp_ena_vsi(vsi))
5296 err = ice_vsi_cfg_xdp_txqs(vsi);
5298 err = ice_vsi_cfg_rxqs(vsi);
5303 /* THEORY OF MODERATION:
5304 * The below code creates custom DIM profiles for use by this driver, because
5305 * the ice driver hardware works differently than the hardware that DIMLIB was
5306 * originally made for. ice hardware doesn't have packet count limits that
5307 * can trigger an interrupt, but it *does* have interrupt rate limit support,
5308 * and this code adds that capability to be used by the driver when it's using
5309 * DIMLIB. The DIMLIB code was always designed to be a suggestion to the driver
5310 * for how to "respond" to traffic and interrupts, so this driver uses a
5311 * slightly different set of moderation parameters to get best performance.
5314 /* the throttle rate for interrupts, basically worst case delay before
5315 * an initial interrupt fires, value is stored in microseconds.
5318 /* the rate limit for interrupts, which can cap a delay from a small
5319 * ITR at a certain amount of interrupts per second. f.e. a 2us ITR
5320 * could yield as much as 500,000 interrupts per second, but with a
5321 * 10us rate limit, it limits to 100,000 interrupts per second. Value
5322 * is stored in microseconds.
5327 /* Make a different profile for Rx that doesn't allow quite so aggressive
5328 * moderation at the high end (it maxes out at 128us or about 8k interrupts a
5329 * second. The INTRL/rate parameters here are only useful to cap small ITR
5330 * values, which is why for larger ITR's - like 128, which can only generate
5331 * 8k interrupts per second, there is no point to rate limit and the values
5332 * are set to zero. The rate limit values do affect latency, and so must
5333 * be reasonably small so to not impact latency sensitive tests.
5335 static const struct ice_dim rx_profile[] = {
5343 /* The transmit profile, which has the same sorts of values
5344 * as the previous struct
5346 static const struct ice_dim tx_profile[] = {
5354 static void ice_tx_dim_work(struct work_struct *work)
5356 struct ice_ring_container *rc;
5357 struct ice_q_vector *q_vector;
5361 dim = container_of(work, struct dim, work);
5362 rc = container_of(dim, struct ice_ring_container, dim);
5363 q_vector = container_of(rc, struct ice_q_vector, tx);
5365 if (dim->profile_ix >= ARRAY_SIZE(tx_profile))
5366 dim->profile_ix = ARRAY_SIZE(tx_profile) - 1;
5368 /* look up the values in our local table */
5369 itr = tx_profile[dim->profile_ix].itr;
5370 intrl = tx_profile[dim->profile_ix].intrl;
5372 ice_write_itr(rc, itr);
5373 ice_write_intrl(q_vector, intrl);
5375 dim->state = DIM_START_MEASURE;
5378 static void ice_rx_dim_work(struct work_struct *work)
5380 struct ice_ring_container *rc;
5381 struct ice_q_vector *q_vector;
5385 dim = container_of(work, struct dim, work);
5386 rc = container_of(dim, struct ice_ring_container, dim);
5387 q_vector = container_of(rc, struct ice_q_vector, rx);
5389 if (dim->profile_ix >= ARRAY_SIZE(rx_profile))
5390 dim->profile_ix = ARRAY_SIZE(rx_profile) - 1;
5392 /* look up the values in our local table */
5393 itr = rx_profile[dim->profile_ix].itr;
5394 intrl = rx_profile[dim->profile_ix].intrl;
5396 ice_write_itr(rc, itr);
5397 ice_write_intrl(q_vector, intrl);
5399 dim->state = DIM_START_MEASURE;
5403 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5404 * @vsi: the VSI being configured
5406 static void ice_napi_enable_all(struct ice_vsi *vsi)
5413 ice_for_each_q_vector(vsi, q_idx) {
5414 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5416 INIT_WORK(&q_vector->tx.dim.work, ice_tx_dim_work);
5417 q_vector->tx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5419 INIT_WORK(&q_vector->rx.dim.work, ice_rx_dim_work);
5420 q_vector->rx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5422 if (q_vector->rx.ring || q_vector->tx.ring)
5423 napi_enable(&q_vector->napi);
5428 * ice_up_complete - Finish the last steps of bringing up a connection
5429 * @vsi: The VSI being configured
5431 * Return 0 on success and negative value on error
5433 static int ice_up_complete(struct ice_vsi *vsi)
5435 struct ice_pf *pf = vsi->back;
5438 ice_vsi_cfg_msix(vsi);
5440 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5441 * Tx queue group list was configured and the context bits were
5442 * programmed using ice_vsi_cfg_txqs
5444 err = ice_vsi_start_all_rx_rings(vsi);
5448 clear_bit(ICE_VSI_DOWN, vsi->state);
5449 ice_napi_enable_all(vsi);
5450 ice_vsi_ena_irq(vsi);
5452 if (vsi->port_info &&
5453 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5455 ice_print_link_msg(vsi, true);
5456 netif_tx_start_all_queues(vsi->netdev);
5457 netif_carrier_on(vsi->netdev);
5460 ice_service_task_schedule(pf);
5466 * ice_up - Bring the connection back up after being down
5467 * @vsi: VSI being configured
5469 int ice_up(struct ice_vsi *vsi)
5473 err = ice_vsi_cfg(vsi);
5475 err = ice_up_complete(vsi);
5481 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5482 * @ring: Tx or Rx ring to read stats from
5483 * @pkts: packets stats counter
5484 * @bytes: bytes stats counter
5486 * This function fetches stats from the ring considering the atomic operations
5487 * that needs to be performed to read u64 values in 32 bit machine.
5490 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5499 start = u64_stats_fetch_begin_irq(&ring->syncp);
5500 *pkts = ring->stats.pkts;
5501 *bytes = ring->stats.bytes;
5502 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5506 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5507 * @vsi: the VSI to be updated
5508 * @rings: rings to work on
5509 * @count: number of rings
5512 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5515 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5518 for (i = 0; i < count; i++) {
5519 struct ice_ring *ring;
5522 ring = READ_ONCE(rings[i]);
5523 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5524 vsi_stats->tx_packets += pkts;
5525 vsi_stats->tx_bytes += bytes;
5526 vsi->tx_restart += ring->tx_stats.restart_q;
5527 vsi->tx_busy += ring->tx_stats.tx_busy;
5528 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5533 * ice_update_vsi_ring_stats - Update VSI stats counters
5534 * @vsi: the VSI to be updated
5536 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5538 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5539 struct ice_ring *ring;
5543 /* reset netdev stats */
5544 vsi_stats->tx_packets = 0;
5545 vsi_stats->tx_bytes = 0;
5546 vsi_stats->rx_packets = 0;
5547 vsi_stats->rx_bytes = 0;
5549 /* reset non-netdev (extended) stats */
5550 vsi->tx_restart = 0;
5552 vsi->tx_linearize = 0;
5553 vsi->rx_buf_failed = 0;
5554 vsi->rx_page_failed = 0;
5558 /* update Tx rings counters */
5559 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5561 /* update Rx rings counters */
5562 ice_for_each_rxq(vsi, i) {
5563 ring = READ_ONCE(vsi->rx_rings[i]);
5564 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5565 vsi_stats->rx_packets += pkts;
5566 vsi_stats->rx_bytes += bytes;
5567 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5568 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5571 /* update XDP Tx rings counters */
5572 if (ice_is_xdp_ena_vsi(vsi))
5573 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5580 * ice_update_vsi_stats - Update VSI stats counters
5581 * @vsi: the VSI to be updated
5583 void ice_update_vsi_stats(struct ice_vsi *vsi)
5585 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5586 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5587 struct ice_pf *pf = vsi->back;
5589 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
5590 test_bit(ICE_CFG_BUSY, pf->state))
5593 /* get stats as recorded by Tx/Rx rings */
5594 ice_update_vsi_ring_stats(vsi);
5596 /* get VSI stats as recorded by the hardware */
5597 ice_update_eth_stats(vsi);
5599 cur_ns->tx_errors = cur_es->tx_errors;
5600 cur_ns->rx_dropped = cur_es->rx_discards;
5601 cur_ns->tx_dropped = cur_es->tx_discards;
5602 cur_ns->multicast = cur_es->rx_multicast;
5604 /* update some more netdev stats if this is main VSI */
5605 if (vsi->type == ICE_VSI_PF) {
5606 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5607 cur_ns->rx_errors = pf->stats.crc_errors +
5608 pf->stats.illegal_bytes +
5609 pf->stats.rx_len_errors +
5610 pf->stats.rx_undersize +
5611 pf->hw_csum_rx_error +
5612 pf->stats.rx_jabber +
5613 pf->stats.rx_fragments +
5614 pf->stats.rx_oversize;
5615 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5616 /* record drops from the port level */
5617 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5622 * ice_update_pf_stats - Update PF port stats counters
5623 * @pf: PF whose stats needs to be updated
5625 void ice_update_pf_stats(struct ice_pf *pf)
5627 struct ice_hw_port_stats *prev_ps, *cur_ps;
5628 struct ice_hw *hw = &pf->hw;
5632 port = hw->port_info->lport;
5633 prev_ps = &pf->stats_prev;
5634 cur_ps = &pf->stats;
5636 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5637 &prev_ps->eth.rx_bytes,
5638 &cur_ps->eth.rx_bytes);
5640 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5641 &prev_ps->eth.rx_unicast,
5642 &cur_ps->eth.rx_unicast);
5644 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5645 &prev_ps->eth.rx_multicast,
5646 &cur_ps->eth.rx_multicast);
5648 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5649 &prev_ps->eth.rx_broadcast,
5650 &cur_ps->eth.rx_broadcast);
5652 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5653 &prev_ps->eth.rx_discards,
5654 &cur_ps->eth.rx_discards);
5656 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5657 &prev_ps->eth.tx_bytes,
5658 &cur_ps->eth.tx_bytes);
5660 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5661 &prev_ps->eth.tx_unicast,
5662 &cur_ps->eth.tx_unicast);
5664 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5665 &prev_ps->eth.tx_multicast,
5666 &cur_ps->eth.tx_multicast);
5668 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5669 &prev_ps->eth.tx_broadcast,
5670 &cur_ps->eth.tx_broadcast);
5672 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5673 &prev_ps->tx_dropped_link_down,
5674 &cur_ps->tx_dropped_link_down);
5676 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5677 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5679 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5680 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5682 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5683 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5685 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5686 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5688 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5689 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5691 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5692 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5694 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5695 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5697 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5698 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5700 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5701 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5703 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5704 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5706 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5707 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5709 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5710 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5712 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5713 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5715 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5716 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5718 fd_ctr_base = hw->fd_ctr_base;
5720 ice_stat_update40(hw,
5721 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5722 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5723 &cur_ps->fd_sb_match);
5724 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5725 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5727 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5728 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5730 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5731 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5733 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5734 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5736 ice_update_dcb_stats(pf);
5738 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5739 &prev_ps->crc_errors, &cur_ps->crc_errors);
5741 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5742 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5744 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5745 &prev_ps->mac_local_faults,
5746 &cur_ps->mac_local_faults);
5748 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5749 &prev_ps->mac_remote_faults,
5750 &cur_ps->mac_remote_faults);
5752 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5753 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5755 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5756 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5758 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5759 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5761 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5762 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5764 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5765 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5767 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5769 pf->stat_prev_loaded = true;
5773 * ice_get_stats64 - get statistics for network device structure
5774 * @netdev: network interface device structure
5775 * @stats: main device statistics structure
5778 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5780 struct ice_netdev_priv *np = netdev_priv(netdev);
5781 struct rtnl_link_stats64 *vsi_stats;
5782 struct ice_vsi *vsi = np->vsi;
5784 vsi_stats = &vsi->net_stats;
5786 if (!vsi->num_txq || !vsi->num_rxq)
5789 /* netdev packet/byte stats come from ring counter. These are obtained
5790 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5791 * But, only call the update routine and read the registers if VSI is
5794 if (!test_bit(ICE_VSI_DOWN, vsi->state))
5795 ice_update_vsi_ring_stats(vsi);
5796 stats->tx_packets = vsi_stats->tx_packets;
5797 stats->tx_bytes = vsi_stats->tx_bytes;
5798 stats->rx_packets = vsi_stats->rx_packets;
5799 stats->rx_bytes = vsi_stats->rx_bytes;
5801 /* The rest of the stats can be read from the hardware but instead we
5802 * just return values that the watchdog task has already obtained from
5805 stats->multicast = vsi_stats->multicast;
5806 stats->tx_errors = vsi_stats->tx_errors;
5807 stats->tx_dropped = vsi_stats->tx_dropped;
5808 stats->rx_errors = vsi_stats->rx_errors;
5809 stats->rx_dropped = vsi_stats->rx_dropped;
5810 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5811 stats->rx_length_errors = vsi_stats->rx_length_errors;
5815 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5816 * @vsi: VSI having NAPI disabled
5818 static void ice_napi_disable_all(struct ice_vsi *vsi)
5825 ice_for_each_q_vector(vsi, q_idx) {
5826 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5828 if (q_vector->rx.ring || q_vector->tx.ring)
5829 napi_disable(&q_vector->napi);
5831 cancel_work_sync(&q_vector->tx.dim.work);
5832 cancel_work_sync(&q_vector->rx.dim.work);
5837 * ice_down - Shutdown the connection
5838 * @vsi: The VSI being stopped
5840 int ice_down(struct ice_vsi *vsi)
5842 int i, tx_err, rx_err, link_err = 0;
5844 /* Caller of this function is expected to set the
5845 * vsi->state ICE_DOWN bit
5848 netif_carrier_off(vsi->netdev);
5849 netif_tx_disable(vsi->netdev);
5852 ice_vsi_dis_irq(vsi);
5854 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5856 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5857 vsi->vsi_num, tx_err);
5858 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5859 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5861 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5862 vsi->vsi_num, tx_err);
5865 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5867 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5868 vsi->vsi_num, rx_err);
5870 ice_napi_disable_all(vsi);
5872 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5873 link_err = ice_force_phys_link_state(vsi, false);
5875 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5876 vsi->vsi_num, link_err);
5879 ice_for_each_txq(vsi, i)
5880 ice_clean_tx_ring(vsi->tx_rings[i]);
5882 ice_for_each_rxq(vsi, i)
5883 ice_clean_rx_ring(vsi->rx_rings[i]);
5885 if (tx_err || rx_err || link_err) {
5886 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5887 vsi->vsi_num, vsi->vsw->sw_id);
5895 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5896 * @vsi: VSI having resources allocated
5898 * Return 0 on success, negative on failure
5900 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5904 if (!vsi->num_txq) {
5905 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5910 ice_for_each_txq(vsi, i) {
5911 struct ice_ring *ring = vsi->tx_rings[i];
5916 ring->netdev = vsi->netdev;
5917 err = ice_setup_tx_ring(ring);
5926 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5927 * @vsi: VSI having resources allocated
5929 * Return 0 on success, negative on failure
5931 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5935 if (!vsi->num_rxq) {
5936 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5941 ice_for_each_rxq(vsi, i) {
5942 struct ice_ring *ring = vsi->rx_rings[i];
5947 ring->netdev = vsi->netdev;
5948 err = ice_setup_rx_ring(ring);
5957 * ice_vsi_open_ctrl - open control VSI for use
5958 * @vsi: the VSI to open
5960 * Initialization of the Control VSI
5962 * Returns 0 on success, negative value on error
5964 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5966 char int_name[ICE_INT_NAME_STR_LEN];
5967 struct ice_pf *pf = vsi->back;
5971 dev = ice_pf_to_dev(pf);
5972 /* allocate descriptors */
5973 err = ice_vsi_setup_tx_rings(vsi);
5977 err = ice_vsi_setup_rx_rings(vsi);
5981 err = ice_vsi_cfg(vsi);
5985 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5986 dev_driver_string(dev), dev_name(dev));
5987 err = ice_vsi_req_irq_msix(vsi, int_name);
5991 ice_vsi_cfg_msix(vsi);
5993 err = ice_vsi_start_all_rx_rings(vsi);
5995 goto err_up_complete;
5997 clear_bit(ICE_VSI_DOWN, vsi->state);
5998 ice_vsi_ena_irq(vsi);
6005 ice_vsi_free_rx_rings(vsi);
6007 ice_vsi_free_tx_rings(vsi);
6013 * ice_vsi_open - Called when a network interface is made active
6014 * @vsi: the VSI to open
6016 * Initialization of the VSI
6018 * Returns 0 on success, negative value on error
6020 static int ice_vsi_open(struct ice_vsi *vsi)
6022 char int_name[ICE_INT_NAME_STR_LEN];
6023 struct ice_pf *pf = vsi->back;
6026 /* allocate descriptors */
6027 err = ice_vsi_setup_tx_rings(vsi);
6031 err = ice_vsi_setup_rx_rings(vsi);
6035 err = ice_vsi_cfg(vsi);
6039 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6040 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6041 err = ice_vsi_req_irq_msix(vsi, int_name);
6045 /* Notify the stack of the actual queue counts. */
6046 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6050 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6054 err = ice_up_complete(vsi);
6056 goto err_up_complete;
6063 ice_vsi_free_irq(vsi);
6065 ice_vsi_free_rx_rings(vsi);
6067 ice_vsi_free_tx_rings(vsi);
6073 * ice_vsi_release_all - Delete all VSIs
6074 * @pf: PF from which all VSIs are being removed
6076 static void ice_vsi_release_all(struct ice_pf *pf)
6083 ice_for_each_vsi(pf, i) {
6087 err = ice_vsi_release(pf->vsi[i]);
6089 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6090 i, err, pf->vsi[i]->vsi_num);
6095 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6096 * @pf: pointer to the PF instance
6097 * @type: VSI type to rebuild
6099 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6101 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6103 struct device *dev = ice_pf_to_dev(pf);
6104 enum ice_status status;
6107 ice_for_each_vsi(pf, i) {
6108 struct ice_vsi *vsi = pf->vsi[i];
6110 if (!vsi || vsi->type != type)
6113 /* rebuild the VSI */
6114 err = ice_vsi_rebuild(vsi, true);
6116 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6117 err, vsi->idx, ice_vsi_type_str(type));
6121 /* replay filters for the VSI */
6122 status = ice_replay_vsi(&pf->hw, vsi->idx);
6124 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
6125 ice_stat_str(status), vsi->idx,
6126 ice_vsi_type_str(type));
6130 /* Re-map HW VSI number, using VSI handle that has been
6131 * previously validated in ice_replay_vsi() call above
6133 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6135 /* enable the VSI */
6136 err = ice_ena_vsi(vsi, false);
6138 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6139 err, vsi->idx, ice_vsi_type_str(type));
6143 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6144 ice_vsi_type_str(type));
6151 * ice_update_pf_netdev_link - Update PF netdev link status
6152 * @pf: pointer to the PF instance
6154 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6159 ice_for_each_vsi(pf, i) {
6160 struct ice_vsi *vsi = pf->vsi[i];
6162 if (!vsi || vsi->type != ICE_VSI_PF)
6165 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6167 netif_carrier_on(pf->vsi[i]->netdev);
6168 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6170 netif_carrier_off(pf->vsi[i]->netdev);
6171 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6177 * ice_rebuild - rebuild after reset
6178 * @pf: PF to rebuild
6179 * @reset_type: type of reset
6181 * Do not rebuild VF VSI in this flow because that is already handled via
6182 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6183 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6184 * to reset/rebuild all the VF VSI twice.
6186 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6188 struct device *dev = ice_pf_to_dev(pf);
6189 struct ice_hw *hw = &pf->hw;
6190 enum ice_status ret;
6193 if (test_bit(ICE_DOWN, pf->state))
6194 goto clear_recovery;
6196 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6198 ret = ice_init_all_ctrlq(hw);
6200 dev_err(dev, "control queues init failed %s\n",
6202 goto err_init_ctrlq;
6205 /* if DDP was previously loaded successfully */
6206 if (!ice_is_safe_mode(pf)) {
6207 /* reload the SW DB of filter tables */
6208 if (reset_type == ICE_RESET_PFR)
6209 ice_fill_blk_tbls(hw);
6211 /* Reload DDP Package after CORER/GLOBR reset */
6212 ice_load_pkg(NULL, pf);
6215 ret = ice_clear_pf_cfg(hw);
6217 dev_err(dev, "clear PF configuration failed %s\n",
6219 goto err_init_ctrlq;
6222 if (pf->first_sw->dflt_vsi_ena)
6223 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6224 /* clear the default VSI configuration if it exists */
6225 pf->first_sw->dflt_vsi = NULL;
6226 pf->first_sw->dflt_vsi_ena = false;
6228 ice_clear_pxe_mode(hw);
6230 ret = ice_get_caps(hw);
6232 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
6233 goto err_init_ctrlq;
6236 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6238 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6239 goto err_init_ctrlq;
6242 err = ice_sched_init_port(hw->port_info);
6244 goto err_sched_init_port;
6246 /* start misc vector */
6247 err = ice_req_irq_msix_misc(pf);
6249 dev_err(dev, "misc vector setup failed: %d\n", err);
6250 goto err_sched_init_port;
6253 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6254 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6255 if (!rd32(hw, PFQF_FD_SIZE)) {
6256 u16 unused, guar, b_effort;
6258 guar = hw->func_caps.fd_fltr_guar;
6259 b_effort = hw->func_caps.fd_fltr_best_effort;
6261 /* force guaranteed filter pool for PF */
6262 ice_alloc_fd_guar_item(hw, &unused, guar);
6263 /* force shared filter pool for PF */
6264 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6268 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6269 ice_dcb_rebuild(pf);
6271 /* rebuild PF VSI */
6272 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6274 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6275 goto err_vsi_rebuild;
6278 /* If Flow Director is active */
6279 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6280 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6282 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6283 goto err_vsi_rebuild;
6286 /* replay HW Flow Director recipes */
6288 ice_fdir_replay_flows(hw);
6290 /* replay Flow Director filters */
6291 ice_fdir_replay_fltrs(pf);
6293 ice_rebuild_arfs(pf);
6296 ice_update_pf_netdev_link(pf);
6298 /* tell the firmware we are up */
6299 ret = ice_send_version(pf);
6301 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6303 goto err_vsi_rebuild;
6306 ice_replay_post(hw);
6308 /* if we get here, reset flow is successful */
6309 clear_bit(ICE_RESET_FAILED, pf->state);
6311 ice_plug_aux_dev(pf);
6315 err_sched_init_port:
6316 ice_sched_cleanup_all(hw);
6318 ice_shutdown_all_ctrlq(hw);
6319 set_bit(ICE_RESET_FAILED, pf->state);
6321 /* set this bit in PF state to control service task scheduling */
6322 set_bit(ICE_NEEDS_RESTART, pf->state);
6323 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6327 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6328 * @vsi: Pointer to VSI structure
6330 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6332 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6333 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6335 return ICE_RXBUF_3072;
6339 * ice_change_mtu - NDO callback to change the MTU
6340 * @netdev: network interface device structure
6341 * @new_mtu: new value for maximum frame size
6343 * Returns 0 on success, negative on failure
6345 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6347 struct ice_netdev_priv *np = netdev_priv(netdev);
6348 struct ice_vsi *vsi = np->vsi;
6349 struct ice_pf *pf = vsi->back;
6350 struct iidc_event *event;
6354 if (new_mtu == (int)netdev->mtu) {
6355 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6359 if (ice_is_xdp_ena_vsi(vsi)) {
6360 int frame_size = ice_max_xdp_frame_size(vsi);
6362 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6363 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6364 frame_size - ICE_ETH_PKT_HDR_PAD);
6369 /* if a reset is in progress, wait for some time for it to complete */
6371 if (ice_is_reset_in_progress(pf->state)) {
6373 usleep_range(1000, 2000);
6378 } while (count < 100);
6381 netdev_err(netdev, "can't change MTU. Device is busy\n");
6385 event = kzalloc(sizeof(*event), GFP_KERNEL);
6389 set_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6390 ice_send_event_to_aux(pf, event);
6391 clear_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6393 netdev->mtu = (unsigned int)new_mtu;
6395 /* if VSI is up, bring it down and then back up */
6396 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6397 err = ice_down(vsi);
6399 netdev_err(netdev, "change MTU if_down err %d\n", err);
6405 netdev_err(netdev, "change MTU if_up err %d\n", err);
6410 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6412 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
6413 ice_send_event_to_aux(pf, event);
6420 * ice_aq_str - convert AQ err code to a string
6421 * @aq_err: the AQ error code to convert
6423 const char *ice_aq_str(enum ice_aq_err aq_err)
6428 case ICE_AQ_RC_EPERM:
6429 return "ICE_AQ_RC_EPERM";
6430 case ICE_AQ_RC_ENOENT:
6431 return "ICE_AQ_RC_ENOENT";
6432 case ICE_AQ_RC_ENOMEM:
6433 return "ICE_AQ_RC_ENOMEM";
6434 case ICE_AQ_RC_EBUSY:
6435 return "ICE_AQ_RC_EBUSY";
6436 case ICE_AQ_RC_EEXIST:
6437 return "ICE_AQ_RC_EEXIST";
6438 case ICE_AQ_RC_EINVAL:
6439 return "ICE_AQ_RC_EINVAL";
6440 case ICE_AQ_RC_ENOSPC:
6441 return "ICE_AQ_RC_ENOSPC";
6442 case ICE_AQ_RC_ENOSYS:
6443 return "ICE_AQ_RC_ENOSYS";
6444 case ICE_AQ_RC_EMODE:
6445 return "ICE_AQ_RC_EMODE";
6446 case ICE_AQ_RC_ENOSEC:
6447 return "ICE_AQ_RC_ENOSEC";
6448 case ICE_AQ_RC_EBADSIG:
6449 return "ICE_AQ_RC_EBADSIG";
6450 case ICE_AQ_RC_ESVN:
6451 return "ICE_AQ_RC_ESVN";
6452 case ICE_AQ_RC_EBADMAN:
6453 return "ICE_AQ_RC_EBADMAN";
6454 case ICE_AQ_RC_EBADBUF:
6455 return "ICE_AQ_RC_EBADBUF";
6458 return "ICE_AQ_RC_UNKNOWN";
6462 * ice_stat_str - convert status err code to a string
6463 * @stat_err: the status error code to convert
6465 const char *ice_stat_str(enum ice_status stat_err)
6471 return "ICE_ERR_PARAM";
6472 case ICE_ERR_NOT_IMPL:
6473 return "ICE_ERR_NOT_IMPL";
6474 case ICE_ERR_NOT_READY:
6475 return "ICE_ERR_NOT_READY";
6476 case ICE_ERR_NOT_SUPPORTED:
6477 return "ICE_ERR_NOT_SUPPORTED";
6478 case ICE_ERR_BAD_PTR:
6479 return "ICE_ERR_BAD_PTR";
6480 case ICE_ERR_INVAL_SIZE:
6481 return "ICE_ERR_INVAL_SIZE";
6482 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6483 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6484 case ICE_ERR_RESET_FAILED:
6485 return "ICE_ERR_RESET_FAILED";
6486 case ICE_ERR_FW_API_VER:
6487 return "ICE_ERR_FW_API_VER";
6488 case ICE_ERR_NO_MEMORY:
6489 return "ICE_ERR_NO_MEMORY";
6491 return "ICE_ERR_CFG";
6492 case ICE_ERR_OUT_OF_RANGE:
6493 return "ICE_ERR_OUT_OF_RANGE";
6494 case ICE_ERR_ALREADY_EXISTS:
6495 return "ICE_ERR_ALREADY_EXISTS";
6497 return "ICE_ERR_NVM";
6498 case ICE_ERR_NVM_CHECKSUM:
6499 return "ICE_ERR_NVM_CHECKSUM";
6500 case ICE_ERR_BUF_TOO_SHORT:
6501 return "ICE_ERR_BUF_TOO_SHORT";
6502 case ICE_ERR_NVM_BLANK_MODE:
6503 return "ICE_ERR_NVM_BLANK_MODE";
6504 case ICE_ERR_IN_USE:
6505 return "ICE_ERR_IN_USE";
6506 case ICE_ERR_MAX_LIMIT:
6507 return "ICE_ERR_MAX_LIMIT";
6508 case ICE_ERR_RESET_ONGOING:
6509 return "ICE_ERR_RESET_ONGOING";
6510 case ICE_ERR_HW_TABLE:
6511 return "ICE_ERR_HW_TABLE";
6512 case ICE_ERR_DOES_NOT_EXIST:
6513 return "ICE_ERR_DOES_NOT_EXIST";
6514 case ICE_ERR_FW_DDP_MISMATCH:
6515 return "ICE_ERR_FW_DDP_MISMATCH";
6516 case ICE_ERR_AQ_ERROR:
6517 return "ICE_ERR_AQ_ERROR";
6518 case ICE_ERR_AQ_TIMEOUT:
6519 return "ICE_ERR_AQ_TIMEOUT";
6520 case ICE_ERR_AQ_FULL:
6521 return "ICE_ERR_AQ_FULL";
6522 case ICE_ERR_AQ_NO_WORK:
6523 return "ICE_ERR_AQ_NO_WORK";
6524 case ICE_ERR_AQ_EMPTY:
6525 return "ICE_ERR_AQ_EMPTY";
6526 case ICE_ERR_AQ_FW_CRITICAL:
6527 return "ICE_ERR_AQ_FW_CRITICAL";
6530 return "ICE_ERR_UNKNOWN";
6534 * ice_set_rss_lut - Set RSS LUT
6535 * @vsi: Pointer to VSI structure
6536 * @lut: Lookup table
6537 * @lut_size: Lookup table size
6539 * Returns 0 on success, negative on failure
6541 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6543 struct ice_aq_get_set_rss_lut_params params = {};
6544 struct ice_hw *hw = &vsi->back->hw;
6545 enum ice_status status;
6550 params.vsi_handle = vsi->idx;
6551 params.lut_size = lut_size;
6552 params.lut_type = vsi->rss_lut_type;
6555 status = ice_aq_set_rss_lut(hw, ¶ms);
6557 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %s aq_err %s\n",
6558 ice_stat_str(status),
6559 ice_aq_str(hw->adminq.sq_last_status));
6567 * ice_set_rss_key - Set RSS key
6568 * @vsi: Pointer to the VSI structure
6569 * @seed: RSS hash seed
6571 * Returns 0 on success, negative on failure
6573 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
6575 struct ice_hw *hw = &vsi->back->hw;
6576 enum ice_status status;
6581 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6583 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %s aq_err %s\n",
6584 ice_stat_str(status),
6585 ice_aq_str(hw->adminq.sq_last_status));
6593 * ice_get_rss_lut - Get RSS LUT
6594 * @vsi: Pointer to VSI structure
6595 * @lut: Buffer to store the lookup table entries
6596 * @lut_size: Size of buffer to store the lookup table entries
6598 * Returns 0 on success, negative on failure
6600 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6602 struct ice_aq_get_set_rss_lut_params params = {};
6603 struct ice_hw *hw = &vsi->back->hw;
6604 enum ice_status status;
6609 params.vsi_handle = vsi->idx;
6610 params.lut_size = lut_size;
6611 params.lut_type = vsi->rss_lut_type;
6614 status = ice_aq_get_rss_lut(hw, ¶ms);
6616 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %s aq_err %s\n",
6617 ice_stat_str(status),
6618 ice_aq_str(hw->adminq.sq_last_status));
6626 * ice_get_rss_key - Get RSS key
6627 * @vsi: Pointer to VSI structure
6628 * @seed: Buffer to store the key in
6630 * Returns 0 on success, negative on failure
6632 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
6634 struct ice_hw *hw = &vsi->back->hw;
6635 enum ice_status status;
6640 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6642 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %s aq_err %s\n",
6643 ice_stat_str(status),
6644 ice_aq_str(hw->adminq.sq_last_status));
6652 * ice_bridge_getlink - Get the hardware bridge mode
6655 * @seq: RTNL message seq
6656 * @dev: the netdev being configured
6657 * @filter_mask: filter mask passed in
6658 * @nlflags: netlink flags passed in
6660 * Return the bridge mode (VEB/VEPA)
6663 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6664 struct net_device *dev, u32 filter_mask, int nlflags)
6666 struct ice_netdev_priv *np = netdev_priv(dev);
6667 struct ice_vsi *vsi = np->vsi;
6668 struct ice_pf *pf = vsi->back;
6671 bmode = pf->first_sw->bridge_mode;
6673 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6678 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6679 * @vsi: Pointer to VSI structure
6680 * @bmode: Hardware bridge mode (VEB/VEPA)
6682 * Returns 0 on success, negative on failure
6684 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6686 struct ice_aqc_vsi_props *vsi_props;
6687 struct ice_hw *hw = &vsi->back->hw;
6688 struct ice_vsi_ctx *ctxt;
6689 enum ice_status status;
6692 vsi_props = &vsi->info;
6694 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6698 ctxt->info = vsi->info;
6700 if (bmode == BRIDGE_MODE_VEB)
6701 /* change from VEPA to VEB mode */
6702 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6704 /* change from VEB to VEPA mode */
6705 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6706 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6708 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6710 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6711 bmode, ice_stat_str(status),
6712 ice_aq_str(hw->adminq.sq_last_status));
6716 /* Update sw flags for book keeping */
6717 vsi_props->sw_flags = ctxt->info.sw_flags;
6725 * ice_bridge_setlink - Set the hardware bridge mode
6726 * @dev: the netdev being configured
6727 * @nlh: RTNL message
6728 * @flags: bridge setlink flags
6729 * @extack: netlink extended ack
6731 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6732 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6733 * not already set for all VSIs connected to this switch. And also update the
6734 * unicast switch filter rules for the corresponding switch of the netdev.
6737 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6738 u16 __always_unused flags,
6739 struct netlink_ext_ack __always_unused *extack)
6741 struct ice_netdev_priv *np = netdev_priv(dev);
6742 struct ice_pf *pf = np->vsi->back;
6743 struct nlattr *attr, *br_spec;
6744 struct ice_hw *hw = &pf->hw;
6745 enum ice_status status;
6746 struct ice_sw *pf_sw;
6747 int rem, v, err = 0;
6749 pf_sw = pf->first_sw;
6750 /* find the attribute in the netlink message */
6751 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6753 nla_for_each_nested(attr, br_spec, rem) {
6756 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6758 mode = nla_get_u16(attr);
6759 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6761 /* Continue if bridge mode is not being flipped */
6762 if (mode == pf_sw->bridge_mode)
6764 /* Iterates through the PF VSI list and update the loopback
6767 ice_for_each_vsi(pf, v) {
6770 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6775 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6776 /* Update the unicast switch filter rules for the corresponding
6777 * switch of the netdev
6779 status = ice_update_sw_rule_bridge_mode(hw);
6781 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6782 mode, ice_stat_str(status),
6783 ice_aq_str(hw->adminq.sq_last_status));
6784 /* revert hw->evb_veb */
6785 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6789 pf_sw->bridge_mode = mode;
6796 * ice_tx_timeout - Respond to a Tx Hang
6797 * @netdev: network interface device structure
6798 * @txqueue: Tx queue
6800 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6802 struct ice_netdev_priv *np = netdev_priv(netdev);
6803 struct ice_ring *tx_ring = NULL;
6804 struct ice_vsi *vsi = np->vsi;
6805 struct ice_pf *pf = vsi->back;
6808 pf->tx_timeout_count++;
6810 /* Check if PFC is enabled for the TC to which the queue belongs
6811 * to. If yes then Tx timeout is not caused by a hung queue, no
6812 * need to reset and rebuild
6814 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6815 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6820 /* now that we have an index, find the tx_ring struct */
6821 for (i = 0; i < vsi->num_txq; i++)
6822 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6823 if (txqueue == vsi->tx_rings[i]->q_index) {
6824 tx_ring = vsi->tx_rings[i];
6828 /* Reset recovery level if enough time has elapsed after last timeout.
6829 * Also ensure no new reset action happens before next timeout period.
6831 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6832 pf->tx_timeout_recovery_level = 1;
6833 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6834 netdev->watchdog_timeo)))
6838 struct ice_hw *hw = &pf->hw;
6841 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6842 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6843 /* Read interrupt register */
6844 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6846 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6847 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6848 head, tx_ring->next_to_use, val);
6851 pf->tx_timeout_last_recovery = jiffies;
6852 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6853 pf->tx_timeout_recovery_level, txqueue);
6855 switch (pf->tx_timeout_recovery_level) {
6857 set_bit(ICE_PFR_REQ, pf->state);
6860 set_bit(ICE_CORER_REQ, pf->state);
6863 set_bit(ICE_GLOBR_REQ, pf->state);
6866 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6867 set_bit(ICE_DOWN, pf->state);
6868 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
6869 set_bit(ICE_SERVICE_DIS, pf->state);
6873 ice_service_task_schedule(pf);
6874 pf->tx_timeout_recovery_level++;
6878 * ice_open - Called when a network interface becomes active
6879 * @netdev: network interface device structure
6881 * The open entry point is called when a network interface is made
6882 * active by the system (IFF_UP). At this point all resources needed
6883 * for transmit and receive operations are allocated, the interrupt
6884 * handler is registered with the OS, the netdev watchdog is enabled,
6885 * and the stack is notified that the interface is ready.
6887 * Returns 0 on success, negative value on failure
6889 int ice_open(struct net_device *netdev)
6891 struct ice_netdev_priv *np = netdev_priv(netdev);
6892 struct ice_pf *pf = np->vsi->back;
6894 if (ice_is_reset_in_progress(pf->state)) {
6895 netdev_err(netdev, "can't open net device while reset is in progress");
6899 return ice_open_internal(netdev);
6903 * ice_open_internal - Called when a network interface becomes active
6904 * @netdev: network interface device structure
6906 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
6909 * Returns 0 on success, negative value on failure
6911 int ice_open_internal(struct net_device *netdev)
6913 struct ice_netdev_priv *np = netdev_priv(netdev);
6914 struct ice_vsi *vsi = np->vsi;
6915 struct ice_pf *pf = vsi->back;
6916 struct ice_port_info *pi;
6917 enum ice_status status;
6920 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
6921 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6925 netif_carrier_off(netdev);
6927 pi = vsi->port_info;
6928 status = ice_update_link_info(pi);
6930 netdev_err(netdev, "Failed to get link info, error %s\n",
6931 ice_stat_str(status));
6935 /* Set PHY if there is media, otherwise, turn off PHY */
6936 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6937 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6938 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
6939 err = ice_init_phy_user_cfg(pi);
6941 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6947 err = ice_configure_phy(vsi);
6949 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6954 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6955 ice_set_link(vsi, false);
6958 err = ice_vsi_open(vsi);
6960 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6961 vsi->vsi_num, vsi->vsw->sw_id);
6963 /* Update existing tunnels information */
6964 udp_tunnel_get_rx_info(netdev);
6970 * ice_stop - Disables a network interface
6971 * @netdev: network interface device structure
6973 * The stop entry point is called when an interface is de-activated by the OS,
6974 * and the netdevice enters the DOWN state. The hardware is still under the
6975 * driver's control, but the netdev interface is disabled.
6977 * Returns success only - not allowed to fail
6979 int ice_stop(struct net_device *netdev)
6981 struct ice_netdev_priv *np = netdev_priv(netdev);
6982 struct ice_vsi *vsi = np->vsi;
6983 struct ice_pf *pf = vsi->back;
6985 if (ice_is_reset_in_progress(pf->state)) {
6986 netdev_err(netdev, "can't stop net device while reset is in progress");
6996 * ice_features_check - Validate encapsulated packet conforms to limits
6998 * @netdev: This port's netdev
6999 * @features: Offload features that the stack believes apply
7001 static netdev_features_t
7002 ice_features_check(struct sk_buff *skb,
7003 struct net_device __always_unused *netdev,
7004 netdev_features_t features)
7008 /* No point in doing any of this if neither checksum nor GSO are
7009 * being requested for this frame. We can rule out both by just
7010 * checking for CHECKSUM_PARTIAL
7012 if (skb->ip_summed != CHECKSUM_PARTIAL)
7015 /* We cannot support GSO if the MSS is going to be less than
7016 * 64 bytes. If it is then we need to drop support for GSO.
7018 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
7019 features &= ~NETIF_F_GSO_MASK;
7021 len = skb_network_header(skb) - skb->data;
7022 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
7023 goto out_rm_features;
7025 len = skb_transport_header(skb) - skb_network_header(skb);
7026 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7027 goto out_rm_features;
7029 if (skb->encapsulation) {
7030 len = skb_inner_network_header(skb) - skb_transport_header(skb);
7031 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
7032 goto out_rm_features;
7034 len = skb_inner_transport_header(skb) -
7035 skb_inner_network_header(skb);
7036 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7037 goto out_rm_features;
7042 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
7045 static const struct net_device_ops ice_netdev_safe_mode_ops = {
7046 .ndo_open = ice_open,
7047 .ndo_stop = ice_stop,
7048 .ndo_start_xmit = ice_start_xmit,
7049 .ndo_set_mac_address = ice_set_mac_address,
7050 .ndo_validate_addr = eth_validate_addr,
7051 .ndo_change_mtu = ice_change_mtu,
7052 .ndo_get_stats64 = ice_get_stats64,
7053 .ndo_tx_timeout = ice_tx_timeout,
7054 .ndo_bpf = ice_xdp_safe_mode,
7057 static const struct net_device_ops ice_netdev_ops = {
7058 .ndo_open = ice_open,
7059 .ndo_stop = ice_stop,
7060 .ndo_start_xmit = ice_start_xmit,
7061 .ndo_features_check = ice_features_check,
7062 .ndo_set_rx_mode = ice_set_rx_mode,
7063 .ndo_set_mac_address = ice_set_mac_address,
7064 .ndo_validate_addr = eth_validate_addr,
7065 .ndo_change_mtu = ice_change_mtu,
7066 .ndo_get_stats64 = ice_get_stats64,
7067 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
7068 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
7069 .ndo_set_vf_mac = ice_set_vf_mac,
7070 .ndo_get_vf_config = ice_get_vf_cfg,
7071 .ndo_set_vf_trust = ice_set_vf_trust,
7072 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
7073 .ndo_set_vf_link_state = ice_set_vf_link_state,
7074 .ndo_get_vf_stats = ice_get_vf_stats,
7075 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
7076 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
7077 .ndo_set_features = ice_set_features,
7078 .ndo_bridge_getlink = ice_bridge_getlink,
7079 .ndo_bridge_setlink = ice_bridge_setlink,
7080 .ndo_fdb_add = ice_fdb_add,
7081 .ndo_fdb_del = ice_fdb_del,
7082 #ifdef CONFIG_RFS_ACCEL
7083 .ndo_rx_flow_steer = ice_rx_flow_steer,
7085 .ndo_tx_timeout = ice_tx_timeout,
7087 .ndo_xdp_xmit = ice_xdp_xmit,
7088 .ndo_xsk_wakeup = ice_xsk_wakeup,