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 struct workqueue_struct *ice_wq;
39 static const struct net_device_ops ice_netdev_safe_mode_ops;
40 static const struct net_device_ops ice_netdev_ops;
41 static int ice_vsi_open(struct ice_vsi *vsi);
43 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
45 static void ice_vsi_release_all(struct ice_pf *pf);
47 bool netif_is_ice(struct net_device *dev)
49 return dev && (dev->netdev_ops == &ice_netdev_ops);
53 * ice_get_tx_pending - returns number of Tx descriptors not processed
54 * @ring: the ring of descriptors
56 static u16 ice_get_tx_pending(struct ice_ring *ring)
60 head = ring->next_to_clean;
61 tail = ring->next_to_use;
64 return (head < tail) ?
65 tail - head : (tail + ring->count - head);
70 * ice_check_for_hang_subtask - check for and recover hung queues
71 * @pf: pointer to PF struct
73 static void ice_check_for_hang_subtask(struct ice_pf *pf)
75 struct ice_vsi *vsi = NULL;
81 ice_for_each_vsi(pf, v)
82 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
87 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
90 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
95 for (i = 0; i < vsi->num_txq; i++) {
96 struct ice_ring *tx_ring = vsi->tx_rings[i];
98 if (tx_ring && tx_ring->desc) {
99 /* If packet counter has not changed the queue is
100 * likely stalled, so force an interrupt for this
103 * prev_pkt would be negative if there was no
106 packets = tx_ring->stats.pkts & INT_MAX;
107 if (tx_ring->tx_stats.prev_pkt == packets) {
108 /* Trigger sw interrupt to revive the queue */
109 ice_trigger_sw_intr(hw, tx_ring->q_vector);
113 /* Memory barrier between read of packet count and call
114 * to ice_get_tx_pending()
117 tx_ring->tx_stats.prev_pkt =
118 ice_get_tx_pending(tx_ring) ? packets : -1;
124 * ice_init_mac_fltr - Set initial MAC filters
125 * @pf: board private structure
127 * Set initial set of MAC filters for PF VSI; configure filters for permanent
128 * address and broadcast address. If an error is encountered, netdevice will be
131 static int ice_init_mac_fltr(struct ice_pf *pf)
133 enum ice_status status;
137 vsi = ice_get_main_vsi(pf);
141 perm_addr = vsi->port_info->mac.perm_addr;
142 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
150 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
151 * @netdev: the net device on which the sync is happening
152 * @addr: MAC address to sync
154 * This is a callback function which is called by the in kernel device sync
155 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
156 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
157 * MAC filters from the hardware.
159 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
161 struct ice_netdev_priv *np = netdev_priv(netdev);
162 struct ice_vsi *vsi = np->vsi;
164 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
172 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
173 * @netdev: the net device on which the unsync is happening
174 * @addr: MAC address to unsync
176 * This is a callback function which is called by the in kernel device unsync
177 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
178 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
179 * delete the MAC filters from the hardware.
181 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
183 struct ice_netdev_priv *np = netdev_priv(netdev);
184 struct ice_vsi *vsi = np->vsi;
186 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
194 * ice_vsi_fltr_changed - check if filter state changed
195 * @vsi: VSI to be checked
197 * returns true if filter state has changed, false otherwise.
199 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
201 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
202 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
203 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
207 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
208 * @vsi: the VSI being configured
209 * @promisc_m: mask of promiscuous config bits
210 * @set_promisc: enable or disable promisc flag request
213 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
215 struct ice_hw *hw = &vsi->back->hw;
216 enum ice_status status = 0;
218 if (vsi->type != ICE_VSI_PF)
221 if (vsi->num_vlan > 1) {
222 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
226 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
229 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
240 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
241 * @vsi: ptr to the VSI
243 * Push any outstanding VSI filter changes through the AdminQ.
245 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
247 struct device *dev = ice_pf_to_dev(vsi->back);
248 struct net_device *netdev = vsi->netdev;
249 bool promisc_forced_on = false;
250 struct ice_pf *pf = vsi->back;
251 struct ice_hw *hw = &pf->hw;
252 enum ice_status status = 0;
253 u32 changed_flags = 0;
260 while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state))
261 usleep_range(1000, 2000);
263 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
264 vsi->current_netdev_flags = vsi->netdev->flags;
266 INIT_LIST_HEAD(&vsi->tmp_sync_list);
267 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
269 if (ice_vsi_fltr_changed(vsi)) {
270 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
271 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
272 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
274 /* grab the netdev's addr_list_lock */
275 netif_addr_lock_bh(netdev);
276 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
277 ice_add_mac_to_unsync_list);
278 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
279 ice_add_mac_to_unsync_list);
280 /* our temp lists are populated. release lock */
281 netif_addr_unlock_bh(netdev);
284 /* Remove MAC addresses in the unsync list */
285 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
286 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
288 netdev_err(netdev, "Failed to delete MAC filters\n");
289 /* if we failed because of alloc failures, just bail */
290 if (status == ICE_ERR_NO_MEMORY) {
296 /* Add MAC addresses in the sync list */
297 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
298 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
299 /* If filter is added successfully or already exists, do not go into
300 * 'if' condition and report it as error. Instead continue processing
301 * rest of the function.
303 if (status && status != ICE_ERR_ALREADY_EXISTS) {
304 netdev_err(netdev, "Failed to add MAC filters\n");
305 /* If there is no more space for new umac filters, VSI
306 * should go into promiscuous mode. There should be some
307 * space reserved for promiscuous filters.
309 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
310 !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC,
312 promisc_forced_on = true;
313 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
320 /* check for changes in promiscuous modes */
321 if (changed_flags & IFF_ALLMULTI) {
322 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
323 if (vsi->num_vlan > 1)
324 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
326 promisc_m = ICE_MCAST_PROMISC_BITS;
328 err = ice_cfg_promisc(vsi, promisc_m, true);
330 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
332 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
336 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
337 if (vsi->num_vlan > 1)
338 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
340 promisc_m = ICE_MCAST_PROMISC_BITS;
342 err = ice_cfg_promisc(vsi, promisc_m, false);
344 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
346 vsi->current_netdev_flags |= IFF_ALLMULTI;
352 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
353 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
354 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
355 if (vsi->current_netdev_flags & IFF_PROMISC) {
356 /* Apply Rx filter rule to get traffic from wire */
357 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
358 err = ice_set_dflt_vsi(pf->first_sw, vsi);
359 if (err && err != -EEXIST) {
360 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
362 vsi->current_netdev_flags &=
366 ice_cfg_vlan_pruning(vsi, false, false);
369 /* Clear Rx filter to remove traffic from wire */
370 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
371 err = ice_clear_dflt_vsi(pf->first_sw);
373 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
375 vsi->current_netdev_flags |=
379 if (vsi->num_vlan > 1)
380 ice_cfg_vlan_pruning(vsi, true, false);
387 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
390 /* if something went wrong then set the changed flag so we try again */
391 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
392 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
394 clear_bit(__ICE_CFG_BUSY, vsi->state);
399 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
400 * @pf: board private structure
402 static void ice_sync_fltr_subtask(struct ice_pf *pf)
406 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
409 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
411 ice_for_each_vsi(pf, v)
412 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
413 ice_vsi_sync_fltr(pf->vsi[v])) {
414 /* come back and try again later */
415 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
421 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
423 * @locked: is the rtnl_lock already held
425 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
430 ice_for_each_vsi(pf, v)
432 ice_dis_vsi(pf->vsi[v], locked);
434 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
435 pf->pf_agg_node[node].num_vsis = 0;
437 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
438 pf->vf_agg_node[node].num_vsis = 0;
442 * ice_prepare_for_reset - prep for the core to reset
443 * @pf: board private structure
445 * Inform or close all dependent features in prep for reset.
448 ice_prepare_for_reset(struct ice_pf *pf)
450 struct ice_hw *hw = &pf->hw;
453 /* already prepared for reset */
454 if (test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
457 /* Notify VFs of impending reset */
458 if (ice_check_sq_alive(hw, &hw->mailboxq))
459 ice_vc_notify_reset(pf);
461 /* Disable VFs until reset is completed */
462 ice_for_each_vf(pf, i)
463 ice_set_vf_state_qs_dis(&pf->vf[i]);
465 /* clear SW filtering DB */
466 ice_clear_hw_tbls(hw);
467 /* disable the VSIs and their queues that are not already DOWN */
468 ice_pf_dis_all_vsi(pf, false);
471 ice_sched_clear_port(hw->port_info);
473 ice_shutdown_all_ctrlq(hw);
475 set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
479 * ice_do_reset - Initiate one of many types of resets
480 * @pf: board private structure
481 * @reset_type: reset type requested
482 * before this function was called.
484 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
486 struct device *dev = ice_pf_to_dev(pf);
487 struct ice_hw *hw = &pf->hw;
489 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
491 ice_prepare_for_reset(pf);
493 /* trigger the reset */
494 if (ice_reset(hw, reset_type)) {
495 dev_err(dev, "reset %d failed\n", reset_type);
496 set_bit(__ICE_RESET_FAILED, pf->state);
497 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
498 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
499 clear_bit(__ICE_PFR_REQ, pf->state);
500 clear_bit(__ICE_CORER_REQ, pf->state);
501 clear_bit(__ICE_GLOBR_REQ, pf->state);
505 /* PFR is a bit of a special case because it doesn't result in an OICR
506 * interrupt. So for PFR, rebuild after the reset and clear the reset-
507 * associated state bits.
509 if (reset_type == ICE_RESET_PFR) {
511 ice_rebuild(pf, reset_type);
512 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
513 clear_bit(__ICE_PFR_REQ, pf->state);
514 ice_reset_all_vfs(pf, true);
519 * ice_reset_subtask - Set up for resetting the device and driver
520 * @pf: board private structure
522 static void ice_reset_subtask(struct ice_pf *pf)
524 enum ice_reset_req reset_type = ICE_RESET_INVAL;
526 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
527 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
528 * of reset is pending and sets bits in pf->state indicating the reset
529 * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set
530 * prepare for pending reset if not already (for PF software-initiated
531 * global resets the software should already be prepared for it as
532 * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
533 * by firmware or software on other PFs, that bit is not set so prepare
534 * for the reset now), poll for reset done, rebuild and return.
536 if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
537 /* Perform the largest reset requested */
538 if (test_and_clear_bit(__ICE_CORER_RECV, pf->state))
539 reset_type = ICE_RESET_CORER;
540 if (test_and_clear_bit(__ICE_GLOBR_RECV, pf->state))
541 reset_type = ICE_RESET_GLOBR;
542 if (test_and_clear_bit(__ICE_EMPR_RECV, pf->state))
543 reset_type = ICE_RESET_EMPR;
544 /* return if no valid reset type requested */
545 if (reset_type == ICE_RESET_INVAL)
547 ice_prepare_for_reset(pf);
549 /* make sure we are ready to rebuild */
550 if (ice_check_reset(&pf->hw)) {
551 set_bit(__ICE_RESET_FAILED, pf->state);
553 /* done with reset. start rebuild */
554 pf->hw.reset_ongoing = false;
555 ice_rebuild(pf, reset_type);
556 /* clear bit to resume normal operations, but
557 * ICE_NEEDS_RESTART bit is set in case rebuild failed
559 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
560 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
561 clear_bit(__ICE_PFR_REQ, pf->state);
562 clear_bit(__ICE_CORER_REQ, pf->state);
563 clear_bit(__ICE_GLOBR_REQ, pf->state);
564 ice_reset_all_vfs(pf, true);
570 /* No pending resets to finish processing. Check for new resets */
571 if (test_bit(__ICE_PFR_REQ, pf->state))
572 reset_type = ICE_RESET_PFR;
573 if (test_bit(__ICE_CORER_REQ, pf->state))
574 reset_type = ICE_RESET_CORER;
575 if (test_bit(__ICE_GLOBR_REQ, pf->state))
576 reset_type = ICE_RESET_GLOBR;
577 /* If no valid reset type requested just return */
578 if (reset_type == ICE_RESET_INVAL)
581 /* reset if not already down or busy */
582 if (!test_bit(__ICE_DOWN, pf->state) &&
583 !test_bit(__ICE_CFG_BUSY, pf->state)) {
584 ice_do_reset(pf, reset_type);
589 * ice_print_topo_conflict - print topology conflict message
590 * @vsi: the VSI whose topology status is being checked
592 static void ice_print_topo_conflict(struct ice_vsi *vsi)
594 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
595 case ICE_AQ_LINK_TOPO_CONFLICT:
596 case ICE_AQ_LINK_MEDIA_CONFLICT:
597 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
598 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
599 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
600 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");
602 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
603 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");
611 * ice_print_link_msg - print link up or down message
612 * @vsi: the VSI whose link status is being queried
613 * @isup: boolean for if the link is now up or down
615 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
617 struct ice_aqc_get_phy_caps_data *caps;
618 const char *an_advertised;
619 enum ice_status status;
629 if (vsi->current_isup == isup)
632 vsi->current_isup = isup;
635 netdev_info(vsi->netdev, "NIC Link is Down\n");
639 switch (vsi->port_info->phy.link_info.link_speed) {
640 case ICE_AQ_LINK_SPEED_100GB:
643 case ICE_AQ_LINK_SPEED_50GB:
646 case ICE_AQ_LINK_SPEED_40GB:
649 case ICE_AQ_LINK_SPEED_25GB:
652 case ICE_AQ_LINK_SPEED_20GB:
655 case ICE_AQ_LINK_SPEED_10GB:
658 case ICE_AQ_LINK_SPEED_5GB:
661 case ICE_AQ_LINK_SPEED_2500MB:
664 case ICE_AQ_LINK_SPEED_1000MB:
667 case ICE_AQ_LINK_SPEED_100MB:
675 switch (vsi->port_info->fc.current_mode) {
679 case ICE_FC_TX_PAUSE:
682 case ICE_FC_RX_PAUSE:
693 /* Get FEC mode based on negotiated link info */
694 switch (vsi->port_info->phy.link_info.fec_info) {
695 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
696 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
699 case ICE_AQ_LINK_25G_KR_FEC_EN:
700 fec = "FC-FEC/BASE-R";
707 /* check if autoneg completed, might be false due to not supported */
708 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
713 /* Get FEC mode requested based on PHY caps last SW configuration */
714 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
717 an_advertised = "Unknown";
721 status = ice_aq_get_phy_caps(vsi->port_info, false,
722 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
724 netdev_info(vsi->netdev, "Get phy capability failed.\n");
726 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
728 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
729 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
731 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
732 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
733 fec_req = "FC-FEC/BASE-R";
740 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",
741 speed, fec_req, fec, an_advertised, an, fc);
742 ice_print_topo_conflict(vsi);
746 * ice_vsi_link_event - update the VSI's netdev
747 * @vsi: the VSI on which the link event occurred
748 * @link_up: whether or not the VSI needs to be set up or down
750 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
755 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
758 if (vsi->type == ICE_VSI_PF) {
759 if (link_up == netif_carrier_ok(vsi->netdev))
763 netif_carrier_on(vsi->netdev);
764 netif_tx_wake_all_queues(vsi->netdev);
766 netif_carrier_off(vsi->netdev);
767 netif_tx_stop_all_queues(vsi->netdev);
773 * ice_set_dflt_mib - send a default config MIB to the FW
774 * @pf: private PF struct
776 * This function sends a default configuration MIB to the FW.
778 * If this function errors out at any point, the driver is still able to
779 * function. The main impact is that LFC may not operate as expected.
780 * Therefore an error state in this function should be treated with a DBG
781 * message and continue on with driver rebuild/reenable.
783 static void ice_set_dflt_mib(struct ice_pf *pf)
785 struct device *dev = ice_pf_to_dev(pf);
786 u8 mib_type, *buf, *lldpmib = NULL;
787 u16 len, typelen, offset = 0;
788 struct ice_lldp_org_tlv *tlv;
789 struct ice_hw *hw = &pf->hw;
792 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
793 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
795 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
800 /* Add ETS CFG TLV */
801 tlv = (struct ice_lldp_org_tlv *)lldpmib;
802 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
803 ICE_IEEE_ETS_TLV_LEN);
804 tlv->typelen = htons(typelen);
805 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
806 ICE_IEEE_SUBTYPE_ETS_CFG);
807 tlv->ouisubtype = htonl(ouisubtype);
812 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
813 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
814 * Octets 13 - 20 are TSA values - leave as zeros
817 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
819 tlv = (struct ice_lldp_org_tlv *)
820 ((char *)tlv + sizeof(tlv->typelen) + len);
822 /* Add ETS REC TLV */
824 tlv->typelen = htons(typelen);
826 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
827 ICE_IEEE_SUBTYPE_ETS_REC);
828 tlv->ouisubtype = htonl(ouisubtype);
830 /* First octet of buf is reserved
831 * Octets 1 - 4 map UP to TC - all UPs map to zero
832 * Octets 5 - 12 are BW values - set TC 0 to 100%.
833 * Octets 13 - 20 are TSA value - leave as zeros
837 tlv = (struct ice_lldp_org_tlv *)
838 ((char *)tlv + sizeof(tlv->typelen) + len);
840 /* Add PFC CFG TLV */
841 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
842 ICE_IEEE_PFC_TLV_LEN);
843 tlv->typelen = htons(typelen);
845 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
846 ICE_IEEE_SUBTYPE_PFC_CFG);
847 tlv->ouisubtype = htonl(ouisubtype);
849 /* Octet 1 left as all zeros - PFC disabled */
851 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
854 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
855 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
861 * ice_link_event - process the link event
862 * @pf: PF that the link event is associated with
863 * @pi: port_info for the port that the link event is associated with
864 * @link_up: true if the physical link is up and false if it is down
865 * @link_speed: current link speed received from the link event
867 * Returns 0 on success and negative on failure
870 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
873 struct device *dev = ice_pf_to_dev(pf);
874 struct ice_phy_info *phy_info;
875 enum ice_status status;
881 phy_info->link_info_old = phy_info->link_info;
883 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
884 old_link_speed = phy_info->link_info_old.link_speed;
886 /* update the link info structures and re-enable link events,
887 * don't bail on failure due to other book keeping needed
889 status = ice_update_link_info(pi);
891 dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
892 pi->lport, ice_stat_str(status),
893 ice_aq_str(pi->hw->adminq.sq_last_status));
895 /* Check if the link state is up after updating link info, and treat
896 * this event as an UP event since the link is actually UP now.
898 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
901 vsi = ice_get_main_vsi(pf);
902 if (!vsi || !vsi->port_info)
905 /* turn off PHY if media was removed */
906 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
907 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
908 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
909 ice_set_link(vsi, false);
912 /* if the old link up/down and speed is the same as the new */
913 if (link_up == old_link && link_speed == old_link_speed)
916 if (ice_is_dcb_active(pf)) {
917 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
921 ice_set_dflt_mib(pf);
923 ice_vsi_link_event(vsi, link_up);
924 ice_print_link_msg(vsi, link_up);
926 ice_vc_notify_link_state(pf);
932 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
933 * @pf: board private structure
935 static void ice_watchdog_subtask(struct ice_pf *pf)
939 /* if interface is down do nothing */
940 if (test_bit(__ICE_DOWN, pf->state) ||
941 test_bit(__ICE_CFG_BUSY, pf->state))
944 /* make sure we don't do these things too often */
945 if (time_before(jiffies,
946 pf->serv_tmr_prev + pf->serv_tmr_period))
949 pf->serv_tmr_prev = jiffies;
951 /* Update the stats for active netdevs so the network stack
952 * can look at updated numbers whenever it cares to
954 ice_update_pf_stats(pf);
955 ice_for_each_vsi(pf, i)
956 if (pf->vsi[i] && pf->vsi[i]->netdev)
957 ice_update_vsi_stats(pf->vsi[i]);
961 * ice_init_link_events - enable/initialize link events
962 * @pi: pointer to the port_info instance
964 * Returns -EIO on failure, 0 on success
966 static int ice_init_link_events(struct ice_port_info *pi)
970 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
971 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
973 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
974 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
979 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
980 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
989 * ice_handle_link_event - handle link event via ARQ
990 * @pf: PF that the link event is associated with
991 * @event: event structure containing link status info
994 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
996 struct ice_aqc_get_link_status_data *link_data;
997 struct ice_port_info *port_info;
1000 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1001 port_info = pf->hw.port_info;
1005 status = ice_link_event(pf, port_info,
1006 !!(link_data->link_info & ICE_AQ_LINK_UP),
1007 le16_to_cpu(link_data->link_speed));
1009 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1015 enum ice_aq_task_state {
1016 ICE_AQ_TASK_WAITING = 0,
1017 ICE_AQ_TASK_COMPLETE,
1018 ICE_AQ_TASK_CANCELED,
1021 struct ice_aq_task {
1022 struct hlist_node entry;
1025 struct ice_rq_event_info *event;
1026 enum ice_aq_task_state state;
1030 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1031 * @pf: pointer to the PF private structure
1032 * @opcode: the opcode to wait for
1033 * @timeout: how long to wait, in jiffies
1034 * @event: storage for the event info
1036 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1037 * current thread will be put to sleep until the specified event occurs or
1038 * until the given timeout is reached.
1040 * To obtain only the descriptor contents, pass an event without an allocated
1041 * msg_buf. If the complete data buffer is desired, allocate the
1042 * event->msg_buf with enough space ahead of time.
1044 * Returns: zero on success, or a negative error code on failure.
1046 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1047 struct ice_rq_event_info *event)
1049 struct device *dev = ice_pf_to_dev(pf);
1050 struct ice_aq_task *task;
1051 unsigned long start;
1055 task = kzalloc(sizeof(*task), GFP_KERNEL);
1059 INIT_HLIST_NODE(&task->entry);
1060 task->opcode = opcode;
1061 task->event = event;
1062 task->state = ICE_AQ_TASK_WAITING;
1064 spin_lock_bh(&pf->aq_wait_lock);
1065 hlist_add_head(&task->entry, &pf->aq_wait_list);
1066 spin_unlock_bh(&pf->aq_wait_lock);
1070 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1072 switch (task->state) {
1073 case ICE_AQ_TASK_WAITING:
1074 err = ret < 0 ? ret : -ETIMEDOUT;
1076 case ICE_AQ_TASK_CANCELED:
1077 err = ret < 0 ? ret : -ECANCELED;
1079 case ICE_AQ_TASK_COMPLETE:
1080 err = ret < 0 ? ret : 0;
1083 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1088 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1089 jiffies_to_msecs(jiffies - start),
1090 jiffies_to_msecs(timeout),
1093 spin_lock_bh(&pf->aq_wait_lock);
1094 hlist_del(&task->entry);
1095 spin_unlock_bh(&pf->aq_wait_lock);
1102 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1103 * @pf: pointer to the PF private structure
1104 * @opcode: the opcode of the event
1105 * @event: the event to check
1107 * Loops over the current list of pending threads waiting for an AdminQ event.
1108 * For each matching task, copy the contents of the event into the task
1109 * structure and wake up the thread.
1111 * If multiple threads wait for the same opcode, they will all be woken up.
1113 * Note that event->msg_buf will only be duplicated if the event has a buffer
1114 * with enough space already allocated. Otherwise, only the descriptor and
1115 * message length will be copied.
1117 * Returns: true if an event was found, false otherwise
1119 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1120 struct ice_rq_event_info *event)
1122 struct ice_aq_task *task;
1125 spin_lock_bh(&pf->aq_wait_lock);
1126 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1127 if (task->state || task->opcode != opcode)
1130 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1131 task->event->msg_len = event->msg_len;
1133 /* Only copy the data buffer if a destination was set */
1134 if (task->event->msg_buf &&
1135 task->event->buf_len > event->buf_len) {
1136 memcpy(task->event->msg_buf, event->msg_buf,
1138 task->event->buf_len = event->buf_len;
1141 task->state = ICE_AQ_TASK_COMPLETE;
1144 spin_unlock_bh(&pf->aq_wait_lock);
1147 wake_up(&pf->aq_wait_queue);
1151 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1152 * @pf: the PF private structure
1154 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1155 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1157 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1159 struct ice_aq_task *task;
1161 spin_lock_bh(&pf->aq_wait_lock);
1162 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1163 task->state = ICE_AQ_TASK_CANCELED;
1164 spin_unlock_bh(&pf->aq_wait_lock);
1166 wake_up(&pf->aq_wait_queue);
1170 * __ice_clean_ctrlq - helper function to clean controlq rings
1171 * @pf: ptr to struct ice_pf
1172 * @q_type: specific Control queue type
1174 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1176 struct device *dev = ice_pf_to_dev(pf);
1177 struct ice_rq_event_info event;
1178 struct ice_hw *hw = &pf->hw;
1179 struct ice_ctl_q_info *cq;
1184 /* Do not clean control queue if/when PF reset fails */
1185 if (test_bit(__ICE_RESET_FAILED, pf->state))
1189 case ICE_CTL_Q_ADMIN:
1193 case ICE_CTL_Q_MAILBOX:
1198 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1202 /* check for error indications - PF_xx_AxQLEN register layout for
1203 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1205 val = rd32(hw, cq->rq.len);
1206 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1207 PF_FW_ARQLEN_ARQCRIT_M)) {
1209 if (val & PF_FW_ARQLEN_ARQVFE_M)
1210 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1212 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1213 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1216 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1217 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1219 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1220 PF_FW_ARQLEN_ARQCRIT_M);
1222 wr32(hw, cq->rq.len, val);
1225 val = rd32(hw, cq->sq.len);
1226 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1227 PF_FW_ATQLEN_ATQCRIT_M)) {
1229 if (val & PF_FW_ATQLEN_ATQVFE_M)
1230 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1232 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1233 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1236 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1237 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1239 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1240 PF_FW_ATQLEN_ATQCRIT_M);
1242 wr32(hw, cq->sq.len, val);
1245 event.buf_len = cq->rq_buf_size;
1246 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1251 enum ice_status ret;
1254 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1255 if (ret == ICE_ERR_AQ_NO_WORK)
1258 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1263 opcode = le16_to_cpu(event.desc.opcode);
1265 /* Notify any thread that might be waiting for this event */
1266 ice_aq_check_events(pf, opcode, &event);
1269 case ice_aqc_opc_get_link_status:
1270 if (ice_handle_link_event(pf, &event))
1271 dev_err(dev, "Could not handle link event\n");
1273 case ice_aqc_opc_event_lan_overflow:
1274 ice_vf_lan_overflow_event(pf, &event);
1276 case ice_mbx_opc_send_msg_to_pf:
1277 ice_vc_process_vf_msg(pf, &event);
1279 case ice_aqc_opc_fw_logging:
1280 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1282 case ice_aqc_opc_lldp_set_mib_change:
1283 ice_dcb_process_lldp_set_mib_change(pf, &event);
1286 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1290 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1292 kfree(event.msg_buf);
1294 return pending && (i == ICE_DFLT_IRQ_WORK);
1298 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1299 * @hw: pointer to hardware info
1300 * @cq: control queue information
1302 * returns true if there are pending messages in a queue, false if there aren't
1304 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1308 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1309 return cq->rq.next_to_clean != ntu;
1313 * ice_clean_adminq_subtask - clean the AdminQ rings
1314 * @pf: board private structure
1316 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1318 struct ice_hw *hw = &pf->hw;
1320 if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1323 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1326 clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1328 /* There might be a situation where new messages arrive to a control
1329 * queue between processing the last message and clearing the
1330 * EVENT_PENDING bit. So before exiting, check queue head again (using
1331 * ice_ctrlq_pending) and process new messages if any.
1333 if (ice_ctrlq_pending(hw, &hw->adminq))
1334 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1340 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1341 * @pf: board private structure
1343 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1345 struct ice_hw *hw = &pf->hw;
1347 if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1350 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1353 clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1355 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1356 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1362 * ice_service_task_schedule - schedule the service task to wake up
1363 * @pf: board private structure
1365 * If not already scheduled, this puts the task into the work queue.
1367 void ice_service_task_schedule(struct ice_pf *pf)
1369 if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
1370 !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
1371 !test_bit(__ICE_NEEDS_RESTART, pf->state))
1372 queue_work(ice_wq, &pf->serv_task);
1376 * ice_service_task_complete - finish up the service task
1377 * @pf: board private structure
1379 static void ice_service_task_complete(struct ice_pf *pf)
1381 WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
1383 /* force memory (pf->state) to sync before next service task */
1384 smp_mb__before_atomic();
1385 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1389 * ice_service_task_stop - stop service task and cancel works
1390 * @pf: board private structure
1392 * Return 0 if the __ICE_SERVICE_DIS bit was not already set,
1395 static int ice_service_task_stop(struct ice_pf *pf)
1399 ret = test_and_set_bit(__ICE_SERVICE_DIS, pf->state);
1401 if (pf->serv_tmr.function)
1402 del_timer_sync(&pf->serv_tmr);
1403 if (pf->serv_task.func)
1404 cancel_work_sync(&pf->serv_task);
1406 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1411 * ice_service_task_restart - restart service task and schedule works
1412 * @pf: board private structure
1414 * This function is needed for suspend and resume works (e.g WoL scenario)
1416 static void ice_service_task_restart(struct ice_pf *pf)
1418 clear_bit(__ICE_SERVICE_DIS, pf->state);
1419 ice_service_task_schedule(pf);
1423 * ice_service_timer - timer callback to schedule service task
1424 * @t: pointer to timer_list
1426 static void ice_service_timer(struct timer_list *t)
1428 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1430 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1431 ice_service_task_schedule(pf);
1435 * ice_handle_mdd_event - handle malicious driver detect event
1436 * @pf: pointer to the PF structure
1438 * Called from service task. OICR interrupt handler indicates MDD event.
1439 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1440 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1441 * disable the queue, the PF can be configured to reset the VF using ethtool
1442 * private flag mdd-auto-reset-vf.
1444 static void ice_handle_mdd_event(struct ice_pf *pf)
1446 struct device *dev = ice_pf_to_dev(pf);
1447 struct ice_hw *hw = &pf->hw;
1451 if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state)) {
1452 /* Since the VF MDD event logging is rate limited, check if
1453 * there are pending MDD events.
1455 ice_print_vfs_mdd_events(pf);
1459 /* find what triggered an MDD event */
1460 reg = rd32(hw, GL_MDET_TX_PQM);
1461 if (reg & GL_MDET_TX_PQM_VALID_M) {
1462 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1463 GL_MDET_TX_PQM_PF_NUM_S;
1464 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1465 GL_MDET_TX_PQM_VF_NUM_S;
1466 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1467 GL_MDET_TX_PQM_MAL_TYPE_S;
1468 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1469 GL_MDET_TX_PQM_QNUM_S);
1471 if (netif_msg_tx_err(pf))
1472 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1473 event, queue, pf_num, vf_num);
1474 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1477 reg = rd32(hw, GL_MDET_TX_TCLAN);
1478 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1479 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1480 GL_MDET_TX_TCLAN_PF_NUM_S;
1481 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1482 GL_MDET_TX_TCLAN_VF_NUM_S;
1483 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1484 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1485 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1486 GL_MDET_TX_TCLAN_QNUM_S);
1488 if (netif_msg_tx_err(pf))
1489 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1490 event, queue, pf_num, vf_num);
1491 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1494 reg = rd32(hw, GL_MDET_RX);
1495 if (reg & GL_MDET_RX_VALID_M) {
1496 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1497 GL_MDET_RX_PF_NUM_S;
1498 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1499 GL_MDET_RX_VF_NUM_S;
1500 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1501 GL_MDET_RX_MAL_TYPE_S;
1502 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1505 if (netif_msg_rx_err(pf))
1506 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1507 event, queue, pf_num, vf_num);
1508 wr32(hw, GL_MDET_RX, 0xffffffff);
1511 /* check to see if this PF caused an MDD event */
1512 reg = rd32(hw, PF_MDET_TX_PQM);
1513 if (reg & PF_MDET_TX_PQM_VALID_M) {
1514 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1515 if (netif_msg_tx_err(pf))
1516 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1519 reg = rd32(hw, PF_MDET_TX_TCLAN);
1520 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1521 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1522 if (netif_msg_tx_err(pf))
1523 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1526 reg = rd32(hw, PF_MDET_RX);
1527 if (reg & PF_MDET_RX_VALID_M) {
1528 wr32(hw, PF_MDET_RX, 0xFFFF);
1529 if (netif_msg_rx_err(pf))
1530 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1533 /* Check to see if one of the VFs caused an MDD event, and then
1534 * increment counters and set print pending
1536 ice_for_each_vf(pf, i) {
1537 struct ice_vf *vf = &pf->vf[i];
1539 reg = rd32(hw, VP_MDET_TX_PQM(i));
1540 if (reg & VP_MDET_TX_PQM_VALID_M) {
1541 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1542 vf->mdd_tx_events.count++;
1543 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1544 if (netif_msg_tx_err(pf))
1545 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1549 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1550 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1551 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1552 vf->mdd_tx_events.count++;
1553 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1554 if (netif_msg_tx_err(pf))
1555 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1559 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1560 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1561 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1562 vf->mdd_tx_events.count++;
1563 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1564 if (netif_msg_tx_err(pf))
1565 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1569 reg = rd32(hw, VP_MDET_RX(i));
1570 if (reg & VP_MDET_RX_VALID_M) {
1571 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1572 vf->mdd_rx_events.count++;
1573 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1574 if (netif_msg_rx_err(pf))
1575 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1578 /* Since the queue is disabled on VF Rx MDD events, the
1579 * PF can be configured to reset the VF through ethtool
1580 * private flag mdd-auto-reset-vf.
1582 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1583 /* VF MDD event counters will be cleared by
1584 * reset, so print the event prior to reset.
1586 ice_print_vf_rx_mdd_event(vf);
1587 ice_reset_vf(&pf->vf[i], false);
1592 ice_print_vfs_mdd_events(pf);
1596 * ice_force_phys_link_state - Force the physical link state
1597 * @vsi: VSI to force the physical link state to up/down
1598 * @link_up: true/false indicates to set the physical link to up/down
1600 * Force the physical link state by getting the current PHY capabilities from
1601 * hardware and setting the PHY config based on the determined capabilities. If
1602 * link changes a link event will be triggered because both the Enable Automatic
1603 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1605 * Returns 0 on success, negative on failure
1607 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1609 struct ice_aqc_get_phy_caps_data *pcaps;
1610 struct ice_aqc_set_phy_cfg_data *cfg;
1611 struct ice_port_info *pi;
1615 if (!vsi || !vsi->port_info || !vsi->back)
1617 if (vsi->type != ICE_VSI_PF)
1620 dev = ice_pf_to_dev(vsi->back);
1622 pi = vsi->port_info;
1624 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1628 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1631 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1632 vsi->vsi_num, retcode);
1637 /* No change in link */
1638 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1639 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1642 /* Use the current user PHY configuration. The current user PHY
1643 * configuration is initialized during probe from PHY capabilities
1644 * software mode, and updated on set PHY configuration.
1646 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1652 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1654 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1656 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1658 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1660 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1661 vsi->vsi_num, retcode);
1672 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1673 * @pi: port info structure
1675 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1677 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1679 struct ice_aqc_get_phy_caps_data *pcaps;
1680 struct ice_pf *pf = pi->hw->back;
1681 enum ice_status status;
1684 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1688 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1692 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1697 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1698 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1706 * ice_init_link_dflt_override - Initialize link default override
1707 * @pi: port info structure
1709 * Initialize link default override and PHY total port shutdown during probe
1711 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1713 struct ice_link_default_override_tlv *ldo;
1714 struct ice_pf *pf = pi->hw->back;
1716 ldo = &pf->link_dflt_override;
1717 if (ice_get_link_default_override(ldo, pi))
1720 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1723 /* Enable Total Port Shutdown (override/replace link-down-on-close
1724 * ethtool private flag) for ports with Port Disable bit set.
1726 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1727 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1731 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1732 * @pi: port info structure
1734 * If default override is enabled, initialize the user PHY cfg speed and FEC
1735 * settings using the default override mask from the NVM.
1737 * The PHY should only be configured with the default override settings the
1738 * first time media is available. The __ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1739 * is used to indicate that the user PHY cfg default override is initialized
1740 * and the PHY has not been configured with the default override settings. The
1741 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1744 * This function should be called only if the FW doesn't support default
1745 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1747 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1749 struct ice_link_default_override_tlv *ldo;
1750 struct ice_aqc_set_phy_cfg_data *cfg;
1751 struct ice_phy_info *phy = &pi->phy;
1752 struct ice_pf *pf = pi->hw->back;
1754 ldo = &pf->link_dflt_override;
1756 /* If link default override is enabled, use to mask NVM PHY capabilities
1757 * for speed and FEC default configuration.
1759 cfg = &phy->curr_user_phy_cfg;
1761 if (ldo->phy_type_low || ldo->phy_type_high) {
1762 cfg->phy_type_low = pf->nvm_phy_type_lo &
1763 cpu_to_le64(ldo->phy_type_low);
1764 cfg->phy_type_high = pf->nvm_phy_type_hi &
1765 cpu_to_le64(ldo->phy_type_high);
1767 cfg->link_fec_opt = ldo->fec_options;
1768 phy->curr_user_fec_req = ICE_FEC_AUTO;
1770 set_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1774 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1775 * @pi: port info structure
1777 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1778 * mode to default. The PHY defaults are from get PHY capabilities topology
1779 * with media so call when media is first available. An error is returned if
1780 * called when media is not available. The PHY initialization completed state is
1783 * These configurations are used when setting PHY
1784 * configuration. The user PHY configuration is updated on set PHY
1785 * configuration. Returns 0 on success, negative on failure
1787 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1789 struct ice_aqc_get_phy_caps_data *pcaps;
1790 struct ice_phy_info *phy = &pi->phy;
1791 struct ice_pf *pf = pi->hw->back;
1792 enum ice_status status;
1795 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1798 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1802 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1803 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1806 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1809 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1814 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1816 /* check if lenient mode is supported and enabled */
1817 if (ice_fw_supports_link_override(pi->hw) &&
1818 !(pcaps->module_compliance_enforcement &
1819 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1820 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1822 /* if the FW supports default PHY configuration mode, then the driver
1823 * does not have to apply link override settings. If not,
1824 * initialize user PHY configuration with link override values
1826 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1827 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1828 ice_init_phy_cfg_dflt_override(pi);
1833 /* if link default override is not enabled, set user flow control and
1834 * FEC settings based on what get_phy_caps returned
1836 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1837 pcaps->link_fec_options);
1838 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1841 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1842 set_bit(__ICE_PHY_INIT_COMPLETE, pf->state);
1849 * ice_configure_phy - configure PHY
1852 * Set the PHY configuration. If the current PHY configuration is the same as
1853 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1854 * configure the based get PHY capabilities for topology with media.
1856 static int ice_configure_phy(struct ice_vsi *vsi)
1858 struct device *dev = ice_pf_to_dev(vsi->back);
1859 struct ice_port_info *pi = vsi->port_info;
1860 struct ice_aqc_get_phy_caps_data *pcaps;
1861 struct ice_aqc_set_phy_cfg_data *cfg;
1862 struct ice_phy_info *phy = &pi->phy;
1863 struct ice_pf *pf = vsi->back;
1864 enum ice_status status;
1867 /* Ensure we have media as we cannot configure a medialess port */
1868 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1871 ice_print_topo_conflict(vsi);
1873 if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1876 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1877 return ice_force_phys_link_state(vsi, true);
1879 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1883 /* Get current PHY config */
1884 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1887 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1888 vsi->vsi_num, ice_stat_str(status));
1893 /* If PHY enable link is configured and configuration has not changed,
1894 * there's nothing to do
1896 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1897 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1900 /* Use PHY topology as baseline for configuration */
1901 memset(pcaps, 0, sizeof(*pcaps));
1902 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1903 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1906 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1909 dev_err(dev, "Failed to get PHY caps, VSI %d error %s\n",
1910 vsi->vsi_num, ice_stat_str(status));
1915 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1921 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
1923 /* Speed - If default override pending, use curr_user_phy_cfg set in
1924 * ice_init_phy_user_cfg_ldo.
1926 if (test_and_clear_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING,
1927 vsi->back->state)) {
1928 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
1929 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
1931 u64 phy_low = 0, phy_high = 0;
1933 ice_update_phy_type(&phy_low, &phy_high,
1934 pi->phy.curr_user_speed_req);
1935 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
1936 cfg->phy_type_high = pcaps->phy_type_high &
1937 cpu_to_le64(phy_high);
1940 /* Can't provide what was requested; use PHY capabilities */
1941 if (!cfg->phy_type_low && !cfg->phy_type_high) {
1942 cfg->phy_type_low = pcaps->phy_type_low;
1943 cfg->phy_type_high = pcaps->phy_type_high;
1947 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
1949 /* Can't provide what was requested; use PHY capabilities */
1950 if (cfg->link_fec_opt !=
1951 (cfg->link_fec_opt & pcaps->link_fec_options)) {
1952 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
1953 cfg->link_fec_opt = pcaps->link_fec_options;
1956 /* Flow Control - always supported; no need to check against
1959 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
1961 /* Enable link and link update */
1962 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
1964 status = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
1966 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
1967 vsi->vsi_num, ice_stat_str(status));
1978 * ice_check_media_subtask - Check for media
1979 * @pf: pointer to PF struct
1981 * If media is available, then initialize PHY user configuration if it is not
1982 * been, and configure the PHY if the interface is up.
1984 static void ice_check_media_subtask(struct ice_pf *pf)
1986 struct ice_port_info *pi;
1987 struct ice_vsi *vsi;
1990 /* No need to check for media if it's already present */
1991 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
1994 vsi = ice_get_main_vsi(pf);
1998 /* Refresh link info and check if media is present */
1999 pi = vsi->port_info;
2000 err = ice_update_link_info(pi);
2004 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2005 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state))
2006 ice_init_phy_user_cfg(pi);
2008 /* PHY settings are reset on media insertion, reconfigure
2009 * PHY to preserve settings.
2011 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2012 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2015 err = ice_configure_phy(vsi);
2017 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2019 /* A Link Status Event will be generated; the event handler
2020 * will complete bringing the interface up
2026 * ice_service_task - manage and run subtasks
2027 * @work: pointer to work_struct contained by the PF struct
2029 static void ice_service_task(struct work_struct *work)
2031 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2032 unsigned long start_time = jiffies;
2036 /* process reset requests first */
2037 ice_reset_subtask(pf);
2039 /* bail if a reset/recovery cycle is pending or rebuild failed */
2040 if (ice_is_reset_in_progress(pf->state) ||
2041 test_bit(__ICE_SUSPENDED, pf->state) ||
2042 test_bit(__ICE_NEEDS_RESTART, pf->state)) {
2043 ice_service_task_complete(pf);
2047 ice_clean_adminq_subtask(pf);
2048 ice_check_media_subtask(pf);
2049 ice_check_for_hang_subtask(pf);
2050 ice_sync_fltr_subtask(pf);
2051 ice_handle_mdd_event(pf);
2052 ice_watchdog_subtask(pf);
2054 if (ice_is_safe_mode(pf)) {
2055 ice_service_task_complete(pf);
2059 ice_process_vflr_event(pf);
2060 ice_clean_mailboxq_subtask(pf);
2061 ice_sync_arfs_fltrs(pf);
2062 ice_flush_fdir_ctx(pf);
2063 /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
2064 ice_service_task_complete(pf);
2066 /* If the tasks have taken longer than one service timer period
2067 * or there is more work to be done, reset the service timer to
2068 * schedule the service task now.
2070 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2071 test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
2072 test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
2073 test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2074 test_bit(__ICE_FD_VF_FLUSH_CTX, pf->state) ||
2075 test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
2076 mod_timer(&pf->serv_tmr, jiffies);
2080 * ice_set_ctrlq_len - helper function to set controlq length
2081 * @hw: pointer to the HW instance
2083 static void ice_set_ctrlq_len(struct ice_hw *hw)
2085 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2086 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2087 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2088 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2089 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2090 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2091 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2092 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2096 * ice_schedule_reset - schedule a reset
2097 * @pf: board private structure
2098 * @reset: reset being requested
2100 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2102 struct device *dev = ice_pf_to_dev(pf);
2104 /* bail out if earlier reset has failed */
2105 if (test_bit(__ICE_RESET_FAILED, pf->state)) {
2106 dev_dbg(dev, "earlier reset has failed\n");
2109 /* bail if reset/recovery already in progress */
2110 if (ice_is_reset_in_progress(pf->state)) {
2111 dev_dbg(dev, "Reset already in progress\n");
2117 set_bit(__ICE_PFR_REQ, pf->state);
2119 case ICE_RESET_CORER:
2120 set_bit(__ICE_CORER_REQ, pf->state);
2122 case ICE_RESET_GLOBR:
2123 set_bit(__ICE_GLOBR_REQ, pf->state);
2129 ice_service_task_schedule(pf);
2134 * ice_irq_affinity_notify - Callback for affinity changes
2135 * @notify: context as to what irq was changed
2136 * @mask: the new affinity mask
2138 * This is a callback function used by the irq_set_affinity_notifier function
2139 * so that we may register to receive changes to the irq affinity masks.
2142 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2143 const cpumask_t *mask)
2145 struct ice_q_vector *q_vector =
2146 container_of(notify, struct ice_q_vector, affinity_notify);
2148 cpumask_copy(&q_vector->affinity_mask, mask);
2152 * ice_irq_affinity_release - Callback for affinity notifier release
2153 * @ref: internal core kernel usage
2155 * This is a callback function used by the irq_set_affinity_notifier function
2156 * to inform the current notification subscriber that they will no longer
2157 * receive notifications.
2159 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2162 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2163 * @vsi: the VSI being configured
2165 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2167 struct ice_hw *hw = &vsi->back->hw;
2170 ice_for_each_q_vector(vsi, i)
2171 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2178 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2179 * @vsi: the VSI being configured
2180 * @basename: name for the vector
2182 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2184 int q_vectors = vsi->num_q_vectors;
2185 struct ice_pf *pf = vsi->back;
2186 int base = vsi->base_vector;
2193 dev = ice_pf_to_dev(pf);
2194 for (vector = 0; vector < q_vectors; vector++) {
2195 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2197 irq_num = pf->msix_entries[base + vector].vector;
2199 if (q_vector->tx.ring && q_vector->rx.ring) {
2200 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2201 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2203 } else if (q_vector->rx.ring) {
2204 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2205 "%s-%s-%d", basename, "rx", rx_int_idx++);
2206 } else if (q_vector->tx.ring) {
2207 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2208 "%s-%s-%d", basename, "tx", tx_int_idx++);
2210 /* skip this unused q_vector */
2213 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2214 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2215 IRQF_SHARED, q_vector->name,
2218 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2219 0, q_vector->name, q_vector);
2221 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2226 /* register for affinity change notifications */
2227 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2228 struct irq_affinity_notify *affinity_notify;
2230 affinity_notify = &q_vector->affinity_notify;
2231 affinity_notify->notify = ice_irq_affinity_notify;
2232 affinity_notify->release = ice_irq_affinity_release;
2233 irq_set_affinity_notifier(irq_num, affinity_notify);
2236 /* assign the mask for this irq */
2237 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2240 vsi->irqs_ready = true;
2246 irq_num = pf->msix_entries[base + vector].vector;
2247 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2248 irq_set_affinity_notifier(irq_num, NULL);
2249 irq_set_affinity_hint(irq_num, NULL);
2250 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2256 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2257 * @vsi: VSI to setup Tx rings used by XDP
2259 * Return 0 on success and negative value on error
2261 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2263 struct device *dev = ice_pf_to_dev(vsi->back);
2266 for (i = 0; i < vsi->num_xdp_txq; i++) {
2267 u16 xdp_q_idx = vsi->alloc_txq + i;
2268 struct ice_ring *xdp_ring;
2270 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2273 goto free_xdp_rings;
2275 xdp_ring->q_index = xdp_q_idx;
2276 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2277 xdp_ring->ring_active = false;
2278 xdp_ring->vsi = vsi;
2279 xdp_ring->netdev = NULL;
2280 xdp_ring->dev = dev;
2281 xdp_ring->count = vsi->num_tx_desc;
2282 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2283 if (ice_setup_tx_ring(xdp_ring))
2284 goto free_xdp_rings;
2285 ice_set_ring_xdp(xdp_ring);
2286 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2293 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2294 ice_free_tx_ring(vsi->xdp_rings[i]);
2299 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2300 * @vsi: VSI to set the bpf prog on
2301 * @prog: the bpf prog pointer
2303 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2305 struct bpf_prog *old_prog;
2308 old_prog = xchg(&vsi->xdp_prog, prog);
2310 bpf_prog_put(old_prog);
2312 ice_for_each_rxq(vsi, i)
2313 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2317 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2318 * @vsi: VSI to bring up Tx rings used by XDP
2319 * @prog: bpf program that will be assigned to VSI
2321 * Return 0 on success and negative value on error
2323 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2325 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2326 int xdp_rings_rem = vsi->num_xdp_txq;
2327 struct ice_pf *pf = vsi->back;
2328 struct ice_qs_cfg xdp_qs_cfg = {
2329 .qs_mutex = &pf->avail_q_mutex,
2330 .pf_map = pf->avail_txqs,
2331 .pf_map_size = pf->max_pf_txqs,
2332 .q_count = vsi->num_xdp_txq,
2333 .scatter_count = ICE_MAX_SCATTER_TXQS,
2334 .vsi_map = vsi->txq_map,
2335 .vsi_map_offset = vsi->alloc_txq,
2336 .mapping_mode = ICE_VSI_MAP_CONTIG
2338 enum ice_status status;
2342 dev = ice_pf_to_dev(pf);
2343 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2344 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2345 if (!vsi->xdp_rings)
2348 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2349 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2352 if (ice_xdp_alloc_setup_rings(vsi))
2353 goto clear_xdp_rings;
2355 /* follow the logic from ice_vsi_map_rings_to_vectors */
2356 ice_for_each_q_vector(vsi, v_idx) {
2357 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2358 int xdp_rings_per_v, q_id, q_base;
2360 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2361 vsi->num_q_vectors - v_idx);
2362 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2364 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2365 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2367 xdp_ring->q_vector = q_vector;
2368 xdp_ring->next = q_vector->tx.ring;
2369 q_vector->tx.ring = xdp_ring;
2371 xdp_rings_rem -= xdp_rings_per_v;
2374 /* omit the scheduler update if in reset path; XDP queues will be
2375 * taken into account at the end of ice_vsi_rebuild, where
2376 * ice_cfg_vsi_lan is being called
2378 if (ice_is_reset_in_progress(pf->state))
2381 /* tell the Tx scheduler that right now we have
2384 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2385 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2387 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2390 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2391 ice_stat_str(status));
2392 goto clear_xdp_rings;
2394 ice_vsi_assign_bpf_prog(vsi, prog);
2398 for (i = 0; i < vsi->num_xdp_txq; i++)
2399 if (vsi->xdp_rings[i]) {
2400 kfree_rcu(vsi->xdp_rings[i], rcu);
2401 vsi->xdp_rings[i] = NULL;
2405 mutex_lock(&pf->avail_q_mutex);
2406 for (i = 0; i < vsi->num_xdp_txq; i++) {
2407 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2408 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2410 mutex_unlock(&pf->avail_q_mutex);
2412 devm_kfree(dev, vsi->xdp_rings);
2417 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2418 * @vsi: VSI to remove XDP rings
2420 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2423 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2425 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2426 struct ice_pf *pf = vsi->back;
2429 /* q_vectors are freed in reset path so there's no point in detaching
2430 * rings; in case of rebuild being triggered not from reset bits
2431 * in pf->state won't be set, so additionally check first q_vector
2434 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2437 ice_for_each_q_vector(vsi, v_idx) {
2438 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2439 struct ice_ring *ring;
2441 ice_for_each_ring(ring, q_vector->tx)
2442 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2445 /* restore the value of last node prior to XDP setup */
2446 q_vector->tx.ring = ring;
2450 mutex_lock(&pf->avail_q_mutex);
2451 for (i = 0; i < vsi->num_xdp_txq; i++) {
2452 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2453 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2455 mutex_unlock(&pf->avail_q_mutex);
2457 for (i = 0; i < vsi->num_xdp_txq; i++)
2458 if (vsi->xdp_rings[i]) {
2459 if (vsi->xdp_rings[i]->desc)
2460 ice_free_tx_ring(vsi->xdp_rings[i]);
2461 kfree_rcu(vsi->xdp_rings[i], rcu);
2462 vsi->xdp_rings[i] = NULL;
2465 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2466 vsi->xdp_rings = NULL;
2468 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2471 ice_vsi_assign_bpf_prog(vsi, NULL);
2473 /* notify Tx scheduler that we destroyed XDP queues and bring
2474 * back the old number of child nodes
2476 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2477 max_txqs[i] = vsi->num_txq;
2479 /* change number of XDP Tx queues to 0 */
2480 vsi->num_xdp_txq = 0;
2482 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2487 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2488 * @vsi: VSI to schedule napi on
2490 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2494 ice_for_each_rxq(vsi, i) {
2495 struct ice_ring *rx_ring = vsi->rx_rings[i];
2497 if (rx_ring->xsk_pool)
2498 napi_schedule(&rx_ring->q_vector->napi);
2503 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2504 * @vsi: VSI to setup XDP for
2505 * @prog: XDP program
2506 * @extack: netlink extended ack
2509 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2510 struct netlink_ext_ack *extack)
2512 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2513 bool if_running = netif_running(vsi->netdev);
2514 int ret = 0, xdp_ring_err = 0;
2516 if (frame_size > vsi->rx_buf_len) {
2517 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2521 /* need to stop netdev while setting up the program for Rx rings */
2522 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2523 ret = ice_down(vsi);
2525 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2530 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2531 vsi->num_xdp_txq = vsi->alloc_rxq;
2532 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2534 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2535 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2536 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2538 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2540 ice_vsi_assign_bpf_prog(vsi, prog);
2547 ice_vsi_rx_napi_schedule(vsi);
2549 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2553 * ice_xdp - implements XDP handler
2557 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2559 struct ice_netdev_priv *np = netdev_priv(dev);
2560 struct ice_vsi *vsi = np->vsi;
2562 if (vsi->type != ICE_VSI_PF) {
2563 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2567 switch (xdp->command) {
2568 case XDP_SETUP_PROG:
2569 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2570 case XDP_SETUP_XSK_POOL:
2571 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2579 * ice_ena_misc_vector - enable the non-queue interrupts
2580 * @pf: board private structure
2582 static void ice_ena_misc_vector(struct ice_pf *pf)
2584 struct ice_hw *hw = &pf->hw;
2587 /* Disable anti-spoof detection interrupt to prevent spurious event
2588 * interrupts during a function reset. Anti-spoof functionally is
2591 val = rd32(hw, GL_MDCK_TX_TDPU);
2592 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2593 wr32(hw, GL_MDCK_TX_TDPU, val);
2595 /* clear things first */
2596 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2597 rd32(hw, PFINT_OICR); /* read to clear */
2599 val = (PFINT_OICR_ECC_ERR_M |
2600 PFINT_OICR_MAL_DETECT_M |
2602 PFINT_OICR_PCI_EXCEPTION_M |
2604 PFINT_OICR_HMC_ERR_M |
2605 PFINT_OICR_PE_CRITERR_M);
2607 wr32(hw, PFINT_OICR_ENA, val);
2609 /* SW_ITR_IDX = 0, but don't change INTENA */
2610 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2611 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2615 * ice_misc_intr - misc interrupt handler
2616 * @irq: interrupt number
2617 * @data: pointer to a q_vector
2619 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2621 struct ice_pf *pf = (struct ice_pf *)data;
2622 struct ice_hw *hw = &pf->hw;
2623 irqreturn_t ret = IRQ_NONE;
2627 dev = ice_pf_to_dev(pf);
2628 set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
2629 set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2631 oicr = rd32(hw, PFINT_OICR);
2632 ena_mask = rd32(hw, PFINT_OICR_ENA);
2634 if (oicr & PFINT_OICR_SWINT_M) {
2635 ena_mask &= ~PFINT_OICR_SWINT_M;
2639 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2640 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2641 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
2643 if (oicr & PFINT_OICR_VFLR_M) {
2644 /* disable any further VFLR event notifications */
2645 if (test_bit(__ICE_VF_RESETS_DISABLED, pf->state)) {
2646 u32 reg = rd32(hw, PFINT_OICR_ENA);
2648 reg &= ~PFINT_OICR_VFLR_M;
2649 wr32(hw, PFINT_OICR_ENA, reg);
2651 ena_mask &= ~PFINT_OICR_VFLR_M;
2652 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
2656 if (oicr & PFINT_OICR_GRST_M) {
2659 /* we have a reset warning */
2660 ena_mask &= ~PFINT_OICR_GRST_M;
2661 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2662 GLGEN_RSTAT_RESET_TYPE_S;
2664 if (reset == ICE_RESET_CORER)
2666 else if (reset == ICE_RESET_GLOBR)
2668 else if (reset == ICE_RESET_EMPR)
2671 dev_dbg(dev, "Invalid reset type %d\n", reset);
2673 /* If a reset cycle isn't already in progress, we set a bit in
2674 * pf->state so that the service task can start a reset/rebuild.
2675 * We also make note of which reset happened so that peer
2676 * devices/drivers can be informed.
2678 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
2679 if (reset == ICE_RESET_CORER)
2680 set_bit(__ICE_CORER_RECV, pf->state);
2681 else if (reset == ICE_RESET_GLOBR)
2682 set_bit(__ICE_GLOBR_RECV, pf->state);
2684 set_bit(__ICE_EMPR_RECV, pf->state);
2686 /* There are couple of different bits at play here.
2687 * hw->reset_ongoing indicates whether the hardware is
2688 * in reset. This is set to true when a reset interrupt
2689 * is received and set back to false after the driver
2690 * has determined that the hardware is out of reset.
2692 * __ICE_RESET_OICR_RECV in pf->state indicates
2693 * that a post reset rebuild is required before the
2694 * driver is operational again. This is set above.
2696 * As this is the start of the reset/rebuild cycle, set
2697 * both to indicate that.
2699 hw->reset_ongoing = true;
2703 if (oicr & PFINT_OICR_HMC_ERR_M) {
2704 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
2705 dev_dbg(dev, "HMC Error interrupt - info 0x%x, data 0x%x\n",
2706 rd32(hw, PFHMC_ERRORINFO),
2707 rd32(hw, PFHMC_ERRORDATA));
2710 /* Report any remaining unexpected interrupts */
2713 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2714 /* If a critical error is pending there is no choice but to
2717 if (oicr & (PFINT_OICR_PE_CRITERR_M |
2718 PFINT_OICR_PCI_EXCEPTION_M |
2719 PFINT_OICR_ECC_ERR_M)) {
2720 set_bit(__ICE_PFR_REQ, pf->state);
2721 ice_service_task_schedule(pf);
2726 ice_service_task_schedule(pf);
2727 ice_irq_dynamic_ena(hw, NULL, NULL);
2733 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2734 * @hw: pointer to HW structure
2736 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2738 /* disable Admin queue Interrupt causes */
2739 wr32(hw, PFINT_FW_CTL,
2740 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2742 /* disable Mailbox queue Interrupt causes */
2743 wr32(hw, PFINT_MBX_CTL,
2744 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2746 /* disable Control queue Interrupt causes */
2747 wr32(hw, PFINT_OICR_CTL,
2748 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2754 * ice_free_irq_msix_misc - Unroll misc vector setup
2755 * @pf: board private structure
2757 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2759 struct ice_hw *hw = &pf->hw;
2761 ice_dis_ctrlq_interrupts(hw);
2763 /* disable OICR interrupt */
2764 wr32(hw, PFINT_OICR_ENA, 0);
2767 if (pf->msix_entries) {
2768 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2769 devm_free_irq(ice_pf_to_dev(pf),
2770 pf->msix_entries[pf->oicr_idx].vector, pf);
2773 pf->num_avail_sw_msix += 1;
2774 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2778 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2779 * @hw: pointer to HW structure
2780 * @reg_idx: HW vector index to associate the control queue interrupts with
2782 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2786 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2787 PFINT_OICR_CTL_CAUSE_ENA_M);
2788 wr32(hw, PFINT_OICR_CTL, val);
2790 /* enable Admin queue Interrupt causes */
2791 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2792 PFINT_FW_CTL_CAUSE_ENA_M);
2793 wr32(hw, PFINT_FW_CTL, val);
2795 /* enable Mailbox queue Interrupt causes */
2796 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2797 PFINT_MBX_CTL_CAUSE_ENA_M);
2798 wr32(hw, PFINT_MBX_CTL, val);
2804 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2805 * @pf: board private structure
2807 * This sets up the handler for MSIX 0, which is used to manage the
2808 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2809 * when in MSI or Legacy interrupt mode.
2811 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2813 struct device *dev = ice_pf_to_dev(pf);
2814 struct ice_hw *hw = &pf->hw;
2815 int oicr_idx, err = 0;
2817 if (!pf->int_name[0])
2818 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2819 dev_driver_string(dev), dev_name(dev));
2821 /* Do not request IRQ but do enable OICR interrupt since settings are
2822 * lost during reset. Note that this function is called only during
2823 * rebuild path and not while reset is in progress.
2825 if (ice_is_reset_in_progress(pf->state))
2828 /* reserve one vector in irq_tracker for misc interrupts */
2829 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2833 pf->num_avail_sw_msix -= 1;
2834 pf->oicr_idx = (u16)oicr_idx;
2836 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2837 ice_misc_intr, 0, pf->int_name, pf);
2839 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2841 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2842 pf->num_avail_sw_msix += 1;
2847 ice_ena_misc_vector(pf);
2849 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2850 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2851 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2854 ice_irq_dynamic_ena(hw, NULL, NULL);
2860 * ice_napi_add - register NAPI handler for the VSI
2861 * @vsi: VSI for which NAPI handler is to be registered
2863 * This function is only called in the driver's load path. Registering the NAPI
2864 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2865 * reset/rebuild, etc.)
2867 static void ice_napi_add(struct ice_vsi *vsi)
2874 ice_for_each_q_vector(vsi, v_idx)
2875 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2876 ice_napi_poll, NAPI_POLL_WEIGHT);
2880 * ice_set_ops - set netdev and ethtools ops for the given netdev
2881 * @netdev: netdev instance
2883 static void ice_set_ops(struct net_device *netdev)
2885 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2887 if (ice_is_safe_mode(pf)) {
2888 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2889 ice_set_ethtool_safe_mode_ops(netdev);
2893 netdev->netdev_ops = &ice_netdev_ops;
2894 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
2895 ice_set_ethtool_ops(netdev);
2899 * ice_set_netdev_features - set features for the given netdev
2900 * @netdev: netdev instance
2902 static void ice_set_netdev_features(struct net_device *netdev)
2904 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2905 netdev_features_t csumo_features;
2906 netdev_features_t vlano_features;
2907 netdev_features_t dflt_features;
2908 netdev_features_t tso_features;
2910 if (ice_is_safe_mode(pf)) {
2912 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2913 netdev->hw_features = netdev->features;
2917 dflt_features = NETIF_F_SG |
2922 csumo_features = NETIF_F_RXCSUM |
2927 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2928 NETIF_F_HW_VLAN_CTAG_TX |
2929 NETIF_F_HW_VLAN_CTAG_RX;
2931 tso_features = NETIF_F_TSO |
2935 NETIF_F_GSO_UDP_TUNNEL |
2936 NETIF_F_GSO_GRE_CSUM |
2937 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2938 NETIF_F_GSO_PARTIAL |
2939 NETIF_F_GSO_IPXIP4 |
2940 NETIF_F_GSO_IPXIP6 |
2943 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2944 NETIF_F_GSO_GRE_CSUM;
2945 /* set features that user can change */
2946 netdev->hw_features = dflt_features | csumo_features |
2947 vlano_features | tso_features;
2949 /* add support for HW_CSUM on packets with MPLS header */
2950 netdev->mpls_features = NETIF_F_HW_CSUM;
2952 /* enable features */
2953 netdev->features |= netdev->hw_features;
2954 /* encap and VLAN devices inherit default, csumo and tso features */
2955 netdev->hw_enc_features |= dflt_features | csumo_features |
2957 netdev->vlan_features |= dflt_features | csumo_features |
2962 * ice_cfg_netdev - Allocate, configure and register a netdev
2963 * @vsi: the VSI associated with the new netdev
2965 * Returns 0 on success, negative value on failure
2967 static int ice_cfg_netdev(struct ice_vsi *vsi)
2969 struct ice_pf *pf = vsi->back;
2970 struct ice_netdev_priv *np;
2971 struct net_device *netdev;
2972 u8 mac_addr[ETH_ALEN];
2974 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
2979 vsi->netdev = netdev;
2980 np = netdev_priv(netdev);
2983 ice_set_netdev_features(netdev);
2985 ice_set_ops(netdev);
2987 if (vsi->type == ICE_VSI_PF) {
2988 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
2989 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
2990 ether_addr_copy(netdev->dev_addr, mac_addr);
2991 ether_addr_copy(netdev->perm_addr, mac_addr);
2994 netdev->priv_flags |= IFF_UNICAST_FLT;
2996 /* Setup netdev TC information */
2997 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
2999 /* setup watchdog timeout value to be 5 second */
3000 netdev->watchdog_timeo = 5 * HZ;
3002 netdev->min_mtu = ETH_MIN_MTU;
3003 netdev->max_mtu = ICE_MAX_MTU;
3009 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3010 * @lut: Lookup table
3011 * @rss_table_size: Lookup table size
3012 * @rss_size: Range of queue number for hashing
3014 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3018 for (i = 0; i < rss_table_size; i++)
3019 lut[i] = i % rss_size;
3023 * ice_pf_vsi_setup - Set up a PF VSI
3024 * @pf: board private structure
3025 * @pi: pointer to the port_info instance
3027 * Returns pointer to the successfully allocated VSI software struct
3028 * on success, otherwise returns NULL on failure.
3030 static struct ice_vsi *
3031 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3033 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3037 * ice_ctrl_vsi_setup - Set up a control VSI
3038 * @pf: board private structure
3039 * @pi: pointer to the port_info instance
3041 * Returns pointer to the successfully allocated VSI software struct
3042 * on success, otherwise returns NULL on failure.
3044 static struct ice_vsi *
3045 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3047 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3051 * ice_lb_vsi_setup - Set up a loopback VSI
3052 * @pf: board private structure
3053 * @pi: pointer to the port_info instance
3055 * Returns pointer to the successfully allocated VSI software struct
3056 * on success, otherwise returns NULL on failure.
3059 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3061 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3065 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3066 * @netdev: network interface to be adjusted
3067 * @proto: unused protocol
3068 * @vid: VLAN ID to be added
3070 * net_device_ops implementation for adding VLAN IDs
3073 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3076 struct ice_netdev_priv *np = netdev_priv(netdev);
3077 struct ice_vsi *vsi = np->vsi;
3080 /* VLAN 0 is added by default during load/reset */
3084 /* Enable VLAN pruning when a VLAN other than 0 is added */
3085 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3086 ret = ice_cfg_vlan_pruning(vsi, true, false);
3091 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3092 * packets aren't pruned by the device's internal switch on Rx
3094 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3096 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3102 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3103 * @netdev: network interface to be adjusted
3104 * @proto: unused protocol
3105 * @vid: VLAN ID to be removed
3107 * net_device_ops implementation for removing VLAN IDs
3110 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3113 struct ice_netdev_priv *np = netdev_priv(netdev);
3114 struct ice_vsi *vsi = np->vsi;
3117 /* don't allow removal of VLAN 0 */
3121 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3124 ret = ice_vsi_kill_vlan(vsi, vid);
3128 /* Disable pruning when VLAN 0 is the only VLAN rule */
3129 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3130 ret = ice_cfg_vlan_pruning(vsi, false, false);
3132 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3137 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3138 * @pf: board private structure
3140 * Returns 0 on success, negative value on failure
3142 static int ice_setup_pf_sw(struct ice_pf *pf)
3144 struct ice_vsi *vsi;
3147 if (ice_is_reset_in_progress(pf->state))
3150 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3154 status = ice_cfg_netdev(vsi);
3157 goto unroll_vsi_setup;
3159 /* netdev has to be configured before setting frame size */
3160 ice_vsi_cfg_frame_size(vsi);
3162 /* Setup DCB netlink interface */
3163 ice_dcbnl_setup(vsi);
3165 /* registering the NAPI handler requires both the queues and
3166 * netdev to be created, which are done in ice_pf_vsi_setup()
3167 * and ice_cfg_netdev() respectively
3171 status = ice_set_cpu_rx_rmap(vsi);
3173 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3174 vsi->vsi_num, status);
3176 goto unroll_napi_add;
3178 status = ice_init_mac_fltr(pf);
3180 goto free_cpu_rx_map;
3185 ice_free_cpu_rx_rmap(vsi);
3191 free_netdev(vsi->netdev);
3197 ice_vsi_release(vsi);
3202 * ice_get_avail_q_count - Get count of queues in use
3203 * @pf_qmap: bitmap to get queue use count from
3204 * @lock: pointer to a mutex that protects access to pf_qmap
3205 * @size: size of the bitmap
3208 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3214 for_each_clear_bit(bit, pf_qmap, size)
3222 * ice_get_avail_txq_count - Get count of Tx queues in use
3223 * @pf: pointer to an ice_pf instance
3225 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3227 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3232 * ice_get_avail_rxq_count - Get count of Rx queues in use
3233 * @pf: pointer to an ice_pf instance
3235 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3237 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3242 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3243 * @pf: board private structure to initialize
3245 static void ice_deinit_pf(struct ice_pf *pf)
3247 ice_service_task_stop(pf);
3248 mutex_destroy(&pf->sw_mutex);
3249 mutex_destroy(&pf->tc_mutex);
3250 mutex_destroy(&pf->avail_q_mutex);
3252 if (pf->avail_txqs) {
3253 bitmap_free(pf->avail_txqs);
3254 pf->avail_txqs = NULL;
3257 if (pf->avail_rxqs) {
3258 bitmap_free(pf->avail_rxqs);
3259 pf->avail_rxqs = NULL;
3264 * ice_set_pf_caps - set PFs capability flags
3265 * @pf: pointer to the PF instance
3267 static void ice_set_pf_caps(struct ice_pf *pf)
3269 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3271 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3272 if (func_caps->common_cap.dcb)
3273 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3274 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3275 if (func_caps->common_cap.sr_iov_1_1) {
3276 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3277 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3280 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3281 if (func_caps->common_cap.rss_table_size)
3282 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3284 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3285 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3288 /* ctrl_vsi_idx will be set to a valid value when flow director
3289 * is setup by ice_init_fdir
3291 pf->ctrl_vsi_idx = ICE_NO_VSI;
3292 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3293 /* force guaranteed filter pool for PF */
3294 ice_alloc_fd_guar_item(&pf->hw, &unused,
3295 func_caps->fd_fltr_guar);
3296 /* force shared filter pool for PF */
3297 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3298 func_caps->fd_fltr_best_effort);
3301 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3302 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3306 * ice_init_pf - Initialize general software structures (struct ice_pf)
3307 * @pf: board private structure to initialize
3309 static int ice_init_pf(struct ice_pf *pf)
3311 ice_set_pf_caps(pf);
3313 mutex_init(&pf->sw_mutex);
3314 mutex_init(&pf->tc_mutex);
3316 INIT_HLIST_HEAD(&pf->aq_wait_list);
3317 spin_lock_init(&pf->aq_wait_lock);
3318 init_waitqueue_head(&pf->aq_wait_queue);
3320 /* setup service timer and periodic service task */
3321 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3322 pf->serv_tmr_period = HZ;
3323 INIT_WORK(&pf->serv_task, ice_service_task);
3324 clear_bit(__ICE_SERVICE_SCHED, pf->state);
3326 mutex_init(&pf->avail_q_mutex);
3327 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3328 if (!pf->avail_txqs)
3331 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3332 if (!pf->avail_rxqs) {
3333 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3334 pf->avail_txqs = NULL;
3342 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3343 * @pf: board private structure
3345 * compute the number of MSIX vectors required (v_budget) and request from
3346 * the OS. Return the number of vectors reserved or negative on failure
3348 static int ice_ena_msix_range(struct ice_pf *pf)
3350 int v_left, v_actual, v_other, v_budget = 0;
3351 struct device *dev = ice_pf_to_dev(pf);
3354 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3356 /* reserve for LAN miscellaneous handler */
3357 needed = ICE_MIN_LAN_OICR_MSIX;
3358 if (v_left < needed)
3359 goto no_hw_vecs_left_err;
3363 /* reserve for flow director */
3364 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3365 needed = ICE_FDIR_MSIX;
3366 if (v_left < needed)
3367 goto no_hw_vecs_left_err;
3372 /* total used for non-traffic vectors */
3375 /* reserve vectors for LAN traffic */
3376 needed = min_t(int, num_online_cpus(), v_left);
3377 if (v_left < needed)
3378 goto no_hw_vecs_left_err;
3379 pf->num_lan_msix = needed;
3383 pf->msix_entries = devm_kcalloc(dev, v_budget,
3384 sizeof(*pf->msix_entries), GFP_KERNEL);
3385 if (!pf->msix_entries) {
3390 for (i = 0; i < v_budget; i++)
3391 pf->msix_entries[i].entry = i;
3393 /* actually reserve the vectors */
3394 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3395 ICE_MIN_MSIX, v_budget);
3397 dev_err(dev, "unable to reserve MSI-X vectors\n");
3402 if (v_actual < v_budget) {
3403 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3404 v_budget, v_actual);
3406 if (v_actual < ICE_MIN_MSIX) {
3407 /* error if we can't get minimum vectors */
3408 pci_disable_msix(pf->pdev);
3412 int v_traffic = v_actual - v_other;
3414 if (v_actual == ICE_MIN_MSIX ||
3415 v_traffic < ICE_MIN_LAN_TXRX_MSIX)
3416 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3418 pf->num_lan_msix = v_traffic;
3420 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3428 devm_kfree(dev, pf->msix_entries);
3431 no_hw_vecs_left_err:
3432 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3436 pf->num_lan_msix = 0;
3441 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3442 * @pf: board private structure
3444 static void ice_dis_msix(struct ice_pf *pf)
3446 pci_disable_msix(pf->pdev);
3447 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3448 pf->msix_entries = NULL;
3452 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3453 * @pf: board private structure
3455 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3459 if (pf->irq_tracker) {
3460 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3461 pf->irq_tracker = NULL;
3466 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3467 * @pf: board private structure to initialize
3469 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3473 vectors = ice_ena_msix_range(pf);
3478 /* set up vector assignment tracking */
3479 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3480 struct_size(pf->irq_tracker, list, vectors),
3482 if (!pf->irq_tracker) {
3487 /* populate SW interrupts pool with number of OS granted IRQs. */
3488 pf->num_avail_sw_msix = (u16)vectors;
3489 pf->irq_tracker->num_entries = (u16)vectors;
3490 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3496 * ice_is_wol_supported - get NVM state of WoL
3497 * @pf: board private structure
3499 * Check if WoL is supported based on the HW configuration.
3500 * Returns true if NVM supports and enables WoL for this port, false otherwise
3502 bool ice_is_wol_supported(struct ice_pf *pf)
3504 struct ice_hw *hw = &pf->hw;
3507 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3508 * word) indicates WoL is not supported on the corresponding PF ID.
3510 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3513 return !(BIT(hw->pf_id) & wol_ctrl);
3517 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3518 * @vsi: VSI being changed
3519 * @new_rx: new number of Rx queues
3520 * @new_tx: new number of Tx queues
3522 * Only change the number of queues if new_tx, or new_rx is non-0.
3524 * Returns 0 on success.
3526 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3528 struct ice_pf *pf = vsi->back;
3529 int err = 0, timeout = 50;
3531 if (!new_rx && !new_tx)
3534 while (test_and_set_bit(__ICE_CFG_BUSY, pf->state)) {
3538 usleep_range(1000, 2000);
3542 vsi->req_txq = (u16)new_tx;
3544 vsi->req_rxq = (u16)new_rx;
3546 /* set for the next time the netdev is started */
3547 if (!netif_running(vsi->netdev)) {
3548 ice_vsi_rebuild(vsi, false);
3549 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3554 ice_vsi_rebuild(vsi, false);
3555 ice_pf_dcb_recfg(pf);
3558 clear_bit(__ICE_CFG_BUSY, pf->state);
3563 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3564 * @pf: PF to configure
3566 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3567 * VSI can still Tx/Rx VLAN tagged packets.
3569 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3571 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3572 struct ice_vsi_ctx *ctxt;
3573 enum ice_status status;
3579 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3584 ctxt->info = vsi->info;
3586 ctxt->info.valid_sections =
3587 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3588 ICE_AQ_VSI_PROP_SECURITY_VALID |
3589 ICE_AQ_VSI_PROP_SW_VALID);
3591 /* disable VLAN anti-spoof */
3592 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3593 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3595 /* disable VLAN pruning and keep all other settings */
3596 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3598 /* allow all VLANs on Tx and don't strip on Rx */
3599 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3600 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3602 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3604 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3605 ice_stat_str(status),
3606 ice_aq_str(hw->adminq.sq_last_status));
3608 vsi->info.sec_flags = ctxt->info.sec_flags;
3609 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3610 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3617 * ice_log_pkg_init - log result of DDP package load
3618 * @hw: pointer to hardware info
3619 * @status: status of package load
3622 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3624 struct ice_pf *pf = (struct ice_pf *)hw->back;
3625 struct device *dev = ice_pf_to_dev(pf);
3629 /* The package download AdminQ command returned success because
3630 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3631 * already a package loaded on the device.
3633 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3634 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3635 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3636 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3637 !memcmp(hw->pkg_name, hw->active_pkg_name,
3638 sizeof(hw->pkg_name))) {
3639 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3640 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3641 hw->active_pkg_name,
3642 hw->active_pkg_ver.major,
3643 hw->active_pkg_ver.minor,
3644 hw->active_pkg_ver.update,
3645 hw->active_pkg_ver.draft);
3647 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3648 hw->active_pkg_name,
3649 hw->active_pkg_ver.major,
3650 hw->active_pkg_ver.minor,
3651 hw->active_pkg_ver.update,
3652 hw->active_pkg_ver.draft);
3653 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3654 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3655 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",
3656 hw->active_pkg_name,
3657 hw->active_pkg_ver.major,
3658 hw->active_pkg_ver.minor,
3659 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3660 *status = ICE_ERR_NOT_SUPPORTED;
3661 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3662 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3663 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",
3664 hw->active_pkg_name,
3665 hw->active_pkg_ver.major,
3666 hw->active_pkg_ver.minor,
3667 hw->active_pkg_ver.update,
3668 hw->active_pkg_ver.draft,
3675 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");
3676 *status = ICE_ERR_NOT_SUPPORTED;
3679 case ICE_ERR_FW_DDP_MISMATCH:
3680 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");
3682 case ICE_ERR_BUF_TOO_SHORT:
3684 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3686 case ICE_ERR_NOT_SUPPORTED:
3687 /* Package File version not supported */
3688 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3689 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3690 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3691 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3692 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3693 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3694 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3695 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",
3696 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3698 case ICE_ERR_AQ_ERROR:
3699 switch (hw->pkg_dwnld_status) {
3700 case ICE_AQ_RC_ENOSEC:
3701 case ICE_AQ_RC_EBADSIG:
3702 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");
3704 case ICE_AQ_RC_ESVN:
3705 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");
3707 case ICE_AQ_RC_EBADMAN:
3708 case ICE_AQ_RC_EBADBUF:
3709 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3710 /* poll for reset to complete */
3711 if (ice_check_reset(hw))
3712 dev_err(dev, "Error resetting device. Please reload the driver\n");
3719 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3726 * ice_load_pkg - load/reload the DDP Package file
3727 * @firmware: firmware structure when firmware requested or NULL for reload
3728 * @pf: pointer to the PF instance
3730 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3731 * initialize HW tables.
3734 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3736 enum ice_status status = ICE_ERR_PARAM;
3737 struct device *dev = ice_pf_to_dev(pf);
3738 struct ice_hw *hw = &pf->hw;
3740 /* Load DDP Package */
3741 if (firmware && !hw->pkg_copy) {
3742 status = ice_copy_and_init_pkg(hw, firmware->data,
3744 ice_log_pkg_init(hw, &status);
3745 } else if (!firmware && hw->pkg_copy) {
3746 /* Reload package during rebuild after CORER/GLOBR reset */
3747 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3748 ice_log_pkg_init(hw, &status);
3750 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3755 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3759 /* Successful download package is the precondition for advanced
3760 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3762 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3766 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3767 * @pf: pointer to the PF structure
3769 * There is no error returned here because the driver should be able to handle
3770 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3771 * specifically with Tx.
3773 static void ice_verify_cacheline_size(struct ice_pf *pf)
3775 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3776 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3777 ICE_CACHE_LINE_BYTES);
3781 * ice_send_version - update firmware with driver version
3784 * Returns ICE_SUCCESS on success, else error code
3786 static enum ice_status ice_send_version(struct ice_pf *pf)
3788 struct ice_driver_ver dv;
3790 dv.major_ver = 0xff;
3791 dv.minor_ver = 0xff;
3792 dv.build_ver = 0xff;
3793 dv.subbuild_ver = 0;
3794 strscpy((char *)dv.driver_string, UTS_RELEASE,
3795 sizeof(dv.driver_string));
3796 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3800 * ice_init_fdir - Initialize flow director VSI and configuration
3801 * @pf: pointer to the PF instance
3803 * returns 0 on success, negative on error
3805 static int ice_init_fdir(struct ice_pf *pf)
3807 struct device *dev = ice_pf_to_dev(pf);
3808 struct ice_vsi *ctrl_vsi;
3811 /* Side Band Flow Director needs to have a control VSI.
3812 * Allocate it and store it in the PF.
3814 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3816 dev_dbg(dev, "could not create control VSI\n");
3820 err = ice_vsi_open_ctrl(ctrl_vsi);
3822 dev_dbg(dev, "could not open control VSI\n");
3826 mutex_init(&pf->hw.fdir_fltr_lock);
3828 err = ice_fdir_create_dflt_rules(pf);
3835 ice_fdir_release_flows(&pf->hw);
3836 ice_vsi_close(ctrl_vsi);
3838 ice_vsi_release(ctrl_vsi);
3839 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3840 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3841 pf->ctrl_vsi_idx = ICE_NO_VSI;
3847 * ice_get_opt_fw_name - return optional firmware file name or NULL
3848 * @pf: pointer to the PF instance
3850 static char *ice_get_opt_fw_name(struct ice_pf *pf)
3852 /* Optional firmware name same as default with additional dash
3853 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3855 struct pci_dev *pdev = pf->pdev;
3856 char *opt_fw_filename;
3859 /* Determine the name of the optional file using the DSN (two
3860 * dwords following the start of the DSN Capability).
3862 dsn = pci_get_dsn(pdev);
3866 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3867 if (!opt_fw_filename)
3870 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3871 ICE_DDP_PKG_PATH, dsn);
3873 return opt_fw_filename;
3877 * ice_request_fw - Device initialization routine
3878 * @pf: pointer to the PF instance
3880 static void ice_request_fw(struct ice_pf *pf)
3882 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3883 const struct firmware *firmware = NULL;
3884 struct device *dev = ice_pf_to_dev(pf);
3887 /* optional device-specific DDP (if present) overrides the default DDP
3888 * package file. kernel logs a debug message if the file doesn't exist,
3889 * and warning messages for other errors.
3891 if (opt_fw_filename) {
3892 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3894 kfree(opt_fw_filename);
3898 /* request for firmware was successful. Download to device */
3899 ice_load_pkg(firmware, pf);
3900 kfree(opt_fw_filename);
3901 release_firmware(firmware);
3906 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3908 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3912 /* request for firmware was successful. Download to device */
3913 ice_load_pkg(firmware, pf);
3914 release_firmware(firmware);
3918 * ice_print_wake_reason - show the wake up cause in the log
3919 * @pf: pointer to the PF struct
3921 static void ice_print_wake_reason(struct ice_pf *pf)
3923 u32 wus = pf->wakeup_reason;
3924 const char *wake_str;
3926 /* if no wake event, nothing to print */
3930 if (wus & PFPM_WUS_LNKC_M)
3931 wake_str = "Link\n";
3932 else if (wus & PFPM_WUS_MAG_M)
3933 wake_str = "Magic Packet\n";
3934 else if (wus & PFPM_WUS_MNG_M)
3935 wake_str = "Management\n";
3936 else if (wus & PFPM_WUS_FW_RST_WK_M)
3937 wake_str = "Firmware Reset\n";
3939 wake_str = "Unknown\n";
3941 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
3945 * ice_register_netdev - register netdev and devlink port
3946 * @pf: pointer to the PF struct
3948 static int ice_register_netdev(struct ice_pf *pf)
3950 struct ice_vsi *vsi;
3953 vsi = ice_get_main_vsi(pf);
3954 if (!vsi || !vsi->netdev)
3957 err = register_netdev(vsi->netdev);
3959 goto err_register_netdev;
3961 netif_carrier_off(vsi->netdev);
3962 netif_tx_stop_all_queues(vsi->netdev);
3963 err = ice_devlink_create_port(vsi);
3965 goto err_devlink_create;
3967 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
3971 unregister_netdev(vsi->netdev);
3972 err_register_netdev:
3973 free_netdev(vsi->netdev);
3979 * ice_probe - Device initialization routine
3980 * @pdev: PCI device information struct
3981 * @ent: entry in ice_pci_tbl
3983 * Returns 0 on success, negative on failure
3986 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
3988 struct device *dev = &pdev->dev;
3993 /* this driver uses devres, see
3994 * Documentation/driver-api/driver-model/devres.rst
3996 err = pcim_enable_device(pdev);
4000 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
4002 dev_err(dev, "BAR0 I/O map error %d\n", err);
4006 pf = ice_allocate_pf(dev);
4010 /* set up for high or low DMA */
4011 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4013 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4015 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4019 pci_enable_pcie_error_reporting(pdev);
4020 pci_set_master(pdev);
4023 pci_set_drvdata(pdev, pf);
4024 set_bit(__ICE_DOWN, pf->state);
4025 /* Disable service task until DOWN bit is cleared */
4026 set_bit(__ICE_SERVICE_DIS, pf->state);
4029 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4030 pci_save_state(pdev);
4033 hw->vendor_id = pdev->vendor;
4034 hw->device_id = pdev->device;
4035 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4036 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4037 hw->subsystem_device_id = pdev->subsystem_device;
4038 hw->bus.device = PCI_SLOT(pdev->devfn);
4039 hw->bus.func = PCI_FUNC(pdev->devfn);
4040 ice_set_ctrlq_len(hw);
4042 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4044 err = ice_devlink_register(pf);
4046 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4047 goto err_exit_unroll;
4050 #ifndef CONFIG_DYNAMIC_DEBUG
4052 hw->debug_mask = debug;
4055 err = ice_init_hw(hw);
4057 dev_err(dev, "ice_init_hw failed: %d\n", err);
4059 goto err_exit_unroll;
4064 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4065 * set in pf->state, which will cause ice_is_safe_mode to return
4068 if (ice_is_safe_mode(pf)) {
4069 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4070 /* we already got function/device capabilities but these don't
4071 * reflect what the driver needs to do in safe mode. Instead of
4072 * adding conditional logic everywhere to ignore these
4073 * device/function capabilities, override them.
4075 ice_set_safe_mode_caps(hw);
4078 err = ice_init_pf(pf);
4080 dev_err(dev, "ice_init_pf failed: %d\n", err);
4081 goto err_init_pf_unroll;
4084 ice_devlink_init_regions(pf);
4086 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4087 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4088 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4089 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4091 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4092 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4093 pf->hw.tnl.valid_count[TNL_VXLAN];
4094 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4095 UDP_TUNNEL_TYPE_VXLAN;
4098 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4099 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4100 pf->hw.tnl.valid_count[TNL_GENEVE];
4101 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4102 UDP_TUNNEL_TYPE_GENEVE;
4106 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4107 if (!pf->num_alloc_vsi) {
4109 goto err_init_pf_unroll;
4111 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4112 dev_warn(&pf->pdev->dev,
4113 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4114 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4115 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4118 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4122 goto err_init_pf_unroll;
4125 err = ice_init_interrupt_scheme(pf);
4127 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4129 goto err_init_vsi_unroll;
4132 /* In case of MSIX we are going to setup the misc vector right here
4133 * to handle admin queue events etc. In case of legacy and MSI
4134 * the misc functionality and queue processing is combined in
4135 * the same vector and that gets setup at open.
4137 err = ice_req_irq_msix_misc(pf);
4139 dev_err(dev, "setup of misc vector failed: %d\n", err);
4140 goto err_init_interrupt_unroll;
4143 /* create switch struct for the switch element created by FW on boot */
4144 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4145 if (!pf->first_sw) {
4147 goto err_msix_misc_unroll;
4151 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4153 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4155 pf->first_sw->pf = pf;
4157 /* record the sw_id available for later use */
4158 pf->first_sw->sw_id = hw->port_info->sw_id;
4160 err = ice_setup_pf_sw(pf);
4162 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4163 goto err_alloc_sw_unroll;
4166 clear_bit(__ICE_SERVICE_DIS, pf->state);
4168 /* tell the firmware we are up */
4169 err = ice_send_version(pf);
4171 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4173 goto err_send_version_unroll;
4176 /* since everything is good, start the service timer */
4177 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4179 err = ice_init_link_events(pf->hw.port_info);
4181 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4182 goto err_send_version_unroll;
4185 err = ice_init_nvm_phy_type(pf->hw.port_info);
4187 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4188 goto err_send_version_unroll;
4191 err = ice_update_link_info(pf->hw.port_info);
4193 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4194 goto err_send_version_unroll;
4197 ice_init_link_dflt_override(pf->hw.port_info);
4199 /* if media available, initialize PHY settings */
4200 if (pf->hw.port_info->phy.link_info.link_info &
4201 ICE_AQ_MEDIA_AVAILABLE) {
4202 err = ice_init_phy_user_cfg(pf->hw.port_info);
4204 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4205 goto err_send_version_unroll;
4208 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4209 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4212 ice_configure_phy(vsi);
4215 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4218 ice_verify_cacheline_size(pf);
4220 /* Save wakeup reason register for later use */
4221 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4223 /* check for a power management event */
4224 ice_print_wake_reason(pf);
4226 /* clear wake status, all bits */
4227 wr32(hw, PFPM_WUS, U32_MAX);
4229 /* Disable WoL at init, wait for user to enable */
4230 device_set_wakeup_enable(dev, false);
4232 if (ice_is_safe_mode(pf)) {
4233 ice_set_safe_mode_vlan_cfg(pf);
4237 /* initialize DDP driven features */
4239 /* Note: Flow director init failure is non-fatal to load */
4240 if (ice_init_fdir(pf))
4241 dev_err(dev, "could not initialize flow director\n");
4243 /* Note: DCB init failure is non-fatal to load */
4244 if (ice_init_pf_dcb(pf, false)) {
4245 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4246 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4248 ice_cfg_lldp_mib_change(&pf->hw, true);
4251 if (ice_init_lag(pf))
4252 dev_warn(dev, "Failed to init link aggregation support\n");
4254 /* print PCI link speed and width */
4255 pcie_print_link_status(pf->pdev);
4258 err = ice_register_netdev(pf);
4260 goto err_netdev_reg;
4262 /* ready to go, so clear down state bit */
4263 clear_bit(__ICE_DOWN, pf->state);
4268 err_send_version_unroll:
4269 ice_vsi_release_all(pf);
4270 err_alloc_sw_unroll:
4271 set_bit(__ICE_SERVICE_DIS, pf->state);
4272 set_bit(__ICE_DOWN, pf->state);
4273 devm_kfree(dev, pf->first_sw);
4274 err_msix_misc_unroll:
4275 ice_free_irq_msix_misc(pf);
4276 err_init_interrupt_unroll:
4277 ice_clear_interrupt_scheme(pf);
4278 err_init_vsi_unroll:
4279 devm_kfree(dev, pf->vsi);
4282 ice_devlink_destroy_regions(pf);
4285 ice_devlink_unregister(pf);
4286 pci_disable_pcie_error_reporting(pdev);
4287 pci_disable_device(pdev);
4292 * ice_set_wake - enable or disable Wake on LAN
4293 * @pf: pointer to the PF struct
4295 * Simple helper for WoL control
4297 static void ice_set_wake(struct ice_pf *pf)
4299 struct ice_hw *hw = &pf->hw;
4300 bool wol = pf->wol_ena;
4302 /* clear wake state, otherwise new wake events won't fire */
4303 wr32(hw, PFPM_WUS, U32_MAX);
4305 /* enable / disable APM wake up, no RMW needed */
4306 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4308 /* set magic packet filter enabled */
4309 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4313 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4314 * @pf: pointer to the PF struct
4316 * Issue firmware command to enable multicast magic wake, making
4317 * sure that any locally administered address (LAA) is used for
4318 * wake, and that PF reset doesn't undo the LAA.
4320 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4322 struct device *dev = ice_pf_to_dev(pf);
4323 struct ice_hw *hw = &pf->hw;
4324 enum ice_status status;
4325 u8 mac_addr[ETH_ALEN];
4326 struct ice_vsi *vsi;
4332 vsi = ice_get_main_vsi(pf);
4336 /* Get current MAC address in case it's an LAA */
4338 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4340 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4342 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4343 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4344 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4346 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4348 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4349 ice_stat_str(status),
4350 ice_aq_str(hw->adminq.sq_last_status));
4354 * ice_remove - Device removal routine
4355 * @pdev: PCI device information struct
4357 static void ice_remove(struct pci_dev *pdev)
4359 struct ice_pf *pf = pci_get_drvdata(pdev);
4365 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4366 if (!ice_is_reset_in_progress(pf->state))
4371 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4372 set_bit(__ICE_VF_RESETS_DISABLED, pf->state);
4376 set_bit(__ICE_DOWN, pf->state);
4377 ice_service_task_stop(pf);
4379 ice_aq_cancel_waiting_tasks(pf);
4381 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4383 if (!ice_is_safe_mode(pf))
4384 ice_remove_arfs(pf);
4385 ice_setup_mc_magic_wake(pf);
4386 ice_vsi_release_all(pf);
4388 ice_free_irq_msix_misc(pf);
4389 ice_for_each_vsi(pf, i) {
4392 ice_vsi_free_q_vectors(pf->vsi[i]);
4395 ice_devlink_destroy_regions(pf);
4396 ice_deinit_hw(&pf->hw);
4397 ice_devlink_unregister(pf);
4399 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4400 * do it via ice_schedule_reset() since there is no need to rebuild
4401 * and the service task is already stopped.
4403 ice_reset(&pf->hw, ICE_RESET_PFR);
4404 pci_wait_for_pending_transaction(pdev);
4405 ice_clear_interrupt_scheme(pf);
4406 pci_disable_pcie_error_reporting(pdev);
4407 pci_disable_device(pdev);
4411 * ice_shutdown - PCI callback for shutting down device
4412 * @pdev: PCI device information struct
4414 static void ice_shutdown(struct pci_dev *pdev)
4416 struct ice_pf *pf = pci_get_drvdata(pdev);
4420 if (system_state == SYSTEM_POWER_OFF) {
4421 pci_wake_from_d3(pdev, pf->wol_ena);
4422 pci_set_power_state(pdev, PCI_D3hot);
4428 * ice_prepare_for_shutdown - prep for PCI shutdown
4429 * @pf: board private structure
4431 * Inform or close all dependent features in prep for PCI device shutdown
4433 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4435 struct ice_hw *hw = &pf->hw;
4438 /* Notify VFs of impending reset */
4439 if (ice_check_sq_alive(hw, &hw->mailboxq))
4440 ice_vc_notify_reset(pf);
4442 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4444 /* disable the VSIs and their queues that are not already DOWN */
4445 ice_pf_dis_all_vsi(pf, false);
4447 ice_for_each_vsi(pf, v)
4449 pf->vsi[v]->vsi_num = 0;
4451 ice_shutdown_all_ctrlq(hw);
4455 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4456 * @pf: board private structure to reinitialize
4458 * This routine reinitialize interrupt scheme that was cleared during
4459 * power management suspend callback.
4461 * This should be called during resume routine to re-allocate the q_vectors
4462 * and reacquire interrupts.
4464 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4466 struct device *dev = ice_pf_to_dev(pf);
4469 /* Since we clear MSIX flag during suspend, we need to
4470 * set it back during resume...
4473 ret = ice_init_interrupt_scheme(pf);
4475 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4479 /* Remap vectors and rings, after successful re-init interrupts */
4480 ice_for_each_vsi(pf, v) {
4484 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4487 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4490 ret = ice_req_irq_msix_misc(pf);
4492 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4502 ice_vsi_free_q_vectors(pf->vsi[v]);
4509 * @dev: generic device information structure
4511 * Power Management callback to quiesce the device and prepare
4512 * for D3 transition.
4514 static int __maybe_unused ice_suspend(struct device *dev)
4516 struct pci_dev *pdev = to_pci_dev(dev);
4520 pf = pci_get_drvdata(pdev);
4522 if (!ice_pf_state_is_nominal(pf)) {
4523 dev_err(dev, "Device is not ready, no need to suspend it\n");
4527 /* Stop watchdog tasks until resume completion.
4528 * Even though it is most likely that the service task is
4529 * disabled if the device is suspended or down, the service task's
4530 * state is controlled by a different state bit, and we should
4531 * store and honor whatever state that bit is in at this point.
4533 disabled = ice_service_task_stop(pf);
4535 /* Already suspended?, then there is nothing to do */
4536 if (test_and_set_bit(__ICE_SUSPENDED, pf->state)) {
4538 ice_service_task_restart(pf);
4542 if (test_bit(__ICE_DOWN, pf->state) ||
4543 ice_is_reset_in_progress(pf->state)) {
4544 dev_err(dev, "can't suspend device in reset or already down\n");
4546 ice_service_task_restart(pf);
4550 ice_setup_mc_magic_wake(pf);
4552 ice_prepare_for_shutdown(pf);
4556 /* Free vectors, clear the interrupt scheme and release IRQs
4557 * for proper hibernation, especially with large number of CPUs.
4558 * Otherwise hibernation might fail when mapping all the vectors back
4561 ice_free_irq_msix_misc(pf);
4562 ice_for_each_vsi(pf, v) {
4565 ice_vsi_free_q_vectors(pf->vsi[v]);
4567 ice_clear_interrupt_scheme(pf);
4569 pci_save_state(pdev);
4570 pci_wake_from_d3(pdev, pf->wol_ena);
4571 pci_set_power_state(pdev, PCI_D3hot);
4576 * ice_resume - PM callback for waking up from D3
4577 * @dev: generic device information structure
4579 static int __maybe_unused ice_resume(struct device *dev)
4581 struct pci_dev *pdev = to_pci_dev(dev);
4582 enum ice_reset_req reset_type;
4587 pci_set_power_state(pdev, PCI_D0);
4588 pci_restore_state(pdev);
4589 pci_save_state(pdev);
4591 if (!pci_device_is_present(pdev))
4594 ret = pci_enable_device_mem(pdev);
4596 dev_err(dev, "Cannot enable device after suspend\n");
4600 pf = pci_get_drvdata(pdev);
4603 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4604 ice_print_wake_reason(pf);
4606 /* We cleared the interrupt scheme when we suspended, so we need to
4607 * restore it now to resume device functionality.
4609 ret = ice_reinit_interrupt_scheme(pf);
4611 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4613 clear_bit(__ICE_DOWN, pf->state);
4614 /* Now perform PF reset and rebuild */
4615 reset_type = ICE_RESET_PFR;
4616 /* re-enable service task for reset, but allow reset to schedule it */
4617 clear_bit(__ICE_SERVICE_DIS, pf->state);
4619 if (ice_schedule_reset(pf, reset_type))
4620 dev_err(dev, "Reset during resume failed.\n");
4622 clear_bit(__ICE_SUSPENDED, pf->state);
4623 ice_service_task_restart(pf);
4625 /* Restart the service task */
4626 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4630 #endif /* CONFIG_PM */
4633 * ice_pci_err_detected - warning that PCI error has been detected
4634 * @pdev: PCI device information struct
4635 * @err: the type of PCI error
4637 * Called to warn that something happened on the PCI bus and the error handling
4638 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4640 static pci_ers_result_t
4641 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4643 struct ice_pf *pf = pci_get_drvdata(pdev);
4646 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4648 return PCI_ERS_RESULT_DISCONNECT;
4651 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4652 ice_service_task_stop(pf);
4654 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4655 set_bit(__ICE_PFR_REQ, pf->state);
4656 ice_prepare_for_reset(pf);
4660 return PCI_ERS_RESULT_NEED_RESET;
4664 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4665 * @pdev: PCI device information struct
4667 * Called to determine if the driver can recover from the PCI slot reset by
4668 * using a register read to determine if the device is recoverable.
4670 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4672 struct ice_pf *pf = pci_get_drvdata(pdev);
4673 pci_ers_result_t result;
4677 err = pci_enable_device_mem(pdev);
4679 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4681 result = PCI_ERS_RESULT_DISCONNECT;
4683 pci_set_master(pdev);
4684 pci_restore_state(pdev);
4685 pci_save_state(pdev);
4686 pci_wake_from_d3(pdev, false);
4688 /* Check for life */
4689 reg = rd32(&pf->hw, GLGEN_RTRIG);
4691 result = PCI_ERS_RESULT_RECOVERED;
4693 result = PCI_ERS_RESULT_DISCONNECT;
4696 err = pci_aer_clear_nonfatal_status(pdev);
4698 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4700 /* non-fatal, continue */
4706 * ice_pci_err_resume - restart operations after PCI error recovery
4707 * @pdev: PCI device information struct
4709 * Called to allow the driver to bring things back up after PCI error and/or
4710 * reset recovery have finished
4712 static void ice_pci_err_resume(struct pci_dev *pdev)
4714 struct ice_pf *pf = pci_get_drvdata(pdev);
4717 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4722 if (test_bit(__ICE_SUSPENDED, pf->state)) {
4723 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4728 ice_restore_all_vfs_msi_state(pdev);
4730 ice_do_reset(pf, ICE_RESET_PFR);
4731 ice_service_task_restart(pf);
4732 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4736 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4737 * @pdev: PCI device information struct
4739 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4741 struct ice_pf *pf = pci_get_drvdata(pdev);
4743 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4744 ice_service_task_stop(pf);
4746 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4747 set_bit(__ICE_PFR_REQ, pf->state);
4748 ice_prepare_for_reset(pf);
4754 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4755 * @pdev: PCI device information struct
4757 static void ice_pci_err_reset_done(struct pci_dev *pdev)
4759 ice_pci_err_resume(pdev);
4762 /* ice_pci_tbl - PCI Device ID Table
4764 * Wildcard entries (PCI_ANY_ID) should come last
4765 * Last entry must be all 0s
4767 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4768 * Class, Class Mask, private data (not used) }
4770 static const struct pci_device_id ice_pci_tbl[] = {
4771 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4772 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4773 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4774 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4775 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4776 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4777 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4778 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4779 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4780 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4781 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4782 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4783 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4784 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4785 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4786 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4787 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4788 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4789 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4790 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4791 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4792 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4793 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4794 /* required last entry */
4797 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4799 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4801 static const struct pci_error_handlers ice_pci_err_handler = {
4802 .error_detected = ice_pci_err_detected,
4803 .slot_reset = ice_pci_err_slot_reset,
4804 .reset_prepare = ice_pci_err_reset_prepare,
4805 .reset_done = ice_pci_err_reset_done,
4806 .resume = ice_pci_err_resume
4809 static struct pci_driver ice_driver = {
4810 .name = KBUILD_MODNAME,
4811 .id_table = ice_pci_tbl,
4813 .remove = ice_remove,
4815 .driver.pm = &ice_pm_ops,
4816 #endif /* CONFIG_PM */
4817 .shutdown = ice_shutdown,
4818 .sriov_configure = ice_sriov_configure,
4819 .err_handler = &ice_pci_err_handler
4823 * ice_module_init - Driver registration routine
4825 * ice_module_init is the first routine called when the driver is
4826 * loaded. All it does is register with the PCI subsystem.
4828 static int __init ice_module_init(void)
4832 pr_info("%s\n", ice_driver_string);
4833 pr_info("%s\n", ice_copyright);
4835 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4837 pr_err("Failed to create workqueue\n");
4841 status = pci_register_driver(&ice_driver);
4843 pr_err("failed to register PCI driver, err %d\n", status);
4844 destroy_workqueue(ice_wq);
4849 module_init(ice_module_init);
4852 * ice_module_exit - Driver exit cleanup routine
4854 * ice_module_exit is called just before the driver is removed
4857 static void __exit ice_module_exit(void)
4859 pci_unregister_driver(&ice_driver);
4860 destroy_workqueue(ice_wq);
4861 pr_info("module unloaded\n");
4863 module_exit(ice_module_exit);
4866 * ice_set_mac_address - NDO callback to set MAC address
4867 * @netdev: network interface device structure
4868 * @pi: pointer to an address structure
4870 * Returns 0 on success, negative on failure
4872 static int ice_set_mac_address(struct net_device *netdev, void *pi)
4874 struct ice_netdev_priv *np = netdev_priv(netdev);
4875 struct ice_vsi *vsi = np->vsi;
4876 struct ice_pf *pf = vsi->back;
4877 struct ice_hw *hw = &pf->hw;
4878 struct sockaddr *addr = pi;
4879 enum ice_status status;
4884 mac = (u8 *)addr->sa_data;
4886 if (!is_valid_ether_addr(mac))
4887 return -EADDRNOTAVAIL;
4889 if (ether_addr_equal(netdev->dev_addr, mac)) {
4890 netdev_warn(netdev, "already using mac %pM\n", mac);
4894 if (test_bit(__ICE_DOWN, pf->state) ||
4895 ice_is_reset_in_progress(pf->state)) {
4896 netdev_err(netdev, "can't set mac %pM. device not ready\n",
4901 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
4902 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
4903 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
4904 err = -EADDRNOTAVAIL;
4905 goto err_update_filters;
4908 /* Add filter for new MAC. If filter exists, return success */
4909 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
4910 if (status == ICE_ERR_ALREADY_EXISTS) {
4911 /* Although this MAC filter is already present in hardware it's
4912 * possible in some cases (e.g. bonding) that dev_addr was
4913 * modified outside of the driver and needs to be restored back
4916 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4917 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
4921 /* error if the new filter addition failed */
4923 err = -EADDRNOTAVAIL;
4927 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
4932 /* change the netdev's MAC address */
4933 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4934 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
4937 /* write new MAC address to the firmware */
4938 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
4939 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
4941 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
4942 mac, ice_stat_str(status));
4948 * ice_set_rx_mode - NDO callback to set the netdev filters
4949 * @netdev: network interface device structure
4951 static void ice_set_rx_mode(struct net_device *netdev)
4953 struct ice_netdev_priv *np = netdev_priv(netdev);
4954 struct ice_vsi *vsi = np->vsi;
4959 /* Set the flags to synchronize filters
4960 * ndo_set_rx_mode may be triggered even without a change in netdev
4963 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
4964 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
4965 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
4967 /* schedule our worker thread which will take care of
4968 * applying the new filter changes
4970 ice_service_task_schedule(vsi->back);
4974 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
4975 * @netdev: network interface device structure
4976 * @queue_index: Queue ID
4977 * @maxrate: maximum bandwidth in Mbps
4980 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
4982 struct ice_netdev_priv *np = netdev_priv(netdev);
4983 struct ice_vsi *vsi = np->vsi;
4984 enum ice_status status;
4988 /* Validate maxrate requested is within permitted range */
4989 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
4990 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
4991 maxrate, queue_index);
4995 q_handle = vsi->tx_rings[queue_index]->q_handle;
4996 tc = ice_dcb_get_tc(vsi, queue_index);
4998 /* Set BW back to default, when user set maxrate to 0 */
5000 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5001 q_handle, ICE_MAX_BW);
5003 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5004 q_handle, ICE_MAX_BW, maxrate * 1000);
5006 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
5007 ice_stat_str(status));
5015 * ice_fdb_add - add an entry to the hardware database
5016 * @ndm: the input from the stack
5017 * @tb: pointer to array of nladdr (unused)
5018 * @dev: the net device pointer
5019 * @addr: the MAC address entry being added
5021 * @flags: instructions from stack about fdb operation
5022 * @extack: netlink extended ack
5025 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5026 struct net_device *dev, const unsigned char *addr, u16 vid,
5027 u16 flags, struct netlink_ext_ack __always_unused *extack)
5032 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5035 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5036 netdev_err(dev, "FDB only supports static addresses\n");
5040 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5041 err = dev_uc_add_excl(dev, addr);
5042 else if (is_multicast_ether_addr(addr))
5043 err = dev_mc_add_excl(dev, addr);
5047 /* Only return duplicate errors if NLM_F_EXCL is set */
5048 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5055 * ice_fdb_del - delete an entry from the hardware database
5056 * @ndm: the input from the stack
5057 * @tb: pointer to array of nladdr (unused)
5058 * @dev: the net device pointer
5059 * @addr: the MAC address entry being added
5063 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5064 struct net_device *dev, const unsigned char *addr,
5065 __always_unused u16 vid)
5069 if (ndm->ndm_state & NUD_PERMANENT) {
5070 netdev_err(dev, "FDB only supports static addresses\n");
5074 if (is_unicast_ether_addr(addr))
5075 err = dev_uc_del(dev, addr);
5076 else if (is_multicast_ether_addr(addr))
5077 err = dev_mc_del(dev, addr);
5085 * ice_set_features - set the netdev feature flags
5086 * @netdev: ptr to the netdev being adjusted
5087 * @features: the feature set that the stack is suggesting
5090 ice_set_features(struct net_device *netdev, netdev_features_t features)
5092 struct ice_netdev_priv *np = netdev_priv(netdev);
5093 struct ice_vsi *vsi = np->vsi;
5094 struct ice_pf *pf = vsi->back;
5097 /* Don't set any netdev advanced features with device in Safe Mode */
5098 if (ice_is_safe_mode(vsi->back)) {
5099 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5103 /* Do not change setting during reset */
5104 if (ice_is_reset_in_progress(pf->state)) {
5105 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5109 /* Multiple features can be changed in one call so keep features in
5110 * separate if/else statements to guarantee each feature is checked
5112 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5113 ret = ice_vsi_manage_rss_lut(vsi, true);
5114 else if (!(features & NETIF_F_RXHASH) &&
5115 netdev->features & NETIF_F_RXHASH)
5116 ret = ice_vsi_manage_rss_lut(vsi, false);
5118 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5119 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5120 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5121 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5122 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5123 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5125 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5126 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5127 ret = ice_vsi_manage_vlan_insertion(vsi);
5128 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5129 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5130 ret = ice_vsi_manage_vlan_insertion(vsi);
5132 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5133 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5134 ret = ice_cfg_vlan_pruning(vsi, true, false);
5135 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5136 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5137 ret = ice_cfg_vlan_pruning(vsi, false, false);
5139 if ((features & NETIF_F_NTUPLE) &&
5140 !(netdev->features & NETIF_F_NTUPLE)) {
5141 ice_vsi_manage_fdir(vsi, true);
5143 } else if (!(features & NETIF_F_NTUPLE) &&
5144 (netdev->features & NETIF_F_NTUPLE)) {
5145 ice_vsi_manage_fdir(vsi, false);
5146 ice_clear_arfs(vsi);
5153 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5154 * @vsi: VSI to setup VLAN properties for
5156 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5160 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5161 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5162 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5163 ret = ice_vsi_manage_vlan_insertion(vsi);
5169 * ice_vsi_cfg - Setup the VSI
5170 * @vsi: the VSI being configured
5172 * Return 0 on success and negative value on error
5174 int ice_vsi_cfg(struct ice_vsi *vsi)
5179 ice_set_rx_mode(vsi->netdev);
5181 err = ice_vsi_vlan_setup(vsi);
5186 ice_vsi_cfg_dcb_rings(vsi);
5188 err = ice_vsi_cfg_lan_txqs(vsi);
5189 if (!err && ice_is_xdp_ena_vsi(vsi))
5190 err = ice_vsi_cfg_xdp_txqs(vsi);
5192 err = ice_vsi_cfg_rxqs(vsi);
5198 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5199 * @vsi: the VSI being configured
5201 static void ice_napi_enable_all(struct ice_vsi *vsi)
5208 ice_for_each_q_vector(vsi, q_idx) {
5209 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5211 if (q_vector->rx.ring || q_vector->tx.ring)
5212 napi_enable(&q_vector->napi);
5217 * ice_up_complete - Finish the last steps of bringing up a connection
5218 * @vsi: The VSI being configured
5220 * Return 0 on success and negative value on error
5222 static int ice_up_complete(struct ice_vsi *vsi)
5224 struct ice_pf *pf = vsi->back;
5227 ice_vsi_cfg_msix(vsi);
5229 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5230 * Tx queue group list was configured and the context bits were
5231 * programmed using ice_vsi_cfg_txqs
5233 err = ice_vsi_start_all_rx_rings(vsi);
5237 clear_bit(ICE_VSI_DOWN, vsi->state);
5238 ice_napi_enable_all(vsi);
5239 ice_vsi_ena_irq(vsi);
5241 if (vsi->port_info &&
5242 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5244 ice_print_link_msg(vsi, true);
5245 netif_tx_start_all_queues(vsi->netdev);
5246 netif_carrier_on(vsi->netdev);
5249 ice_service_task_schedule(pf);
5255 * ice_up - Bring the connection back up after being down
5256 * @vsi: VSI being configured
5258 int ice_up(struct ice_vsi *vsi)
5262 err = ice_vsi_cfg(vsi);
5264 err = ice_up_complete(vsi);
5270 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5271 * @ring: Tx or Rx ring to read stats from
5272 * @pkts: packets stats counter
5273 * @bytes: bytes stats counter
5275 * This function fetches stats from the ring considering the atomic operations
5276 * that needs to be performed to read u64 values in 32 bit machine.
5279 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5288 start = u64_stats_fetch_begin_irq(&ring->syncp);
5289 *pkts = ring->stats.pkts;
5290 *bytes = ring->stats.bytes;
5291 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5295 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5296 * @vsi: the VSI to be updated
5297 * @rings: rings to work on
5298 * @count: number of rings
5301 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5304 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5307 for (i = 0; i < count; i++) {
5308 struct ice_ring *ring;
5311 ring = READ_ONCE(rings[i]);
5312 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5313 vsi_stats->tx_packets += pkts;
5314 vsi_stats->tx_bytes += bytes;
5315 vsi->tx_restart += ring->tx_stats.restart_q;
5316 vsi->tx_busy += ring->tx_stats.tx_busy;
5317 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5322 * ice_update_vsi_ring_stats - Update VSI stats counters
5323 * @vsi: the VSI to be updated
5325 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5327 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5328 struct ice_ring *ring;
5332 /* reset netdev stats */
5333 vsi_stats->tx_packets = 0;
5334 vsi_stats->tx_bytes = 0;
5335 vsi_stats->rx_packets = 0;
5336 vsi_stats->rx_bytes = 0;
5338 /* reset non-netdev (extended) stats */
5339 vsi->tx_restart = 0;
5341 vsi->tx_linearize = 0;
5342 vsi->rx_buf_failed = 0;
5343 vsi->rx_page_failed = 0;
5347 /* update Tx rings counters */
5348 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5350 /* update Rx rings counters */
5351 ice_for_each_rxq(vsi, i) {
5352 ring = READ_ONCE(vsi->rx_rings[i]);
5353 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5354 vsi_stats->rx_packets += pkts;
5355 vsi_stats->rx_bytes += bytes;
5356 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5357 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5360 /* update XDP Tx rings counters */
5361 if (ice_is_xdp_ena_vsi(vsi))
5362 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5369 * ice_update_vsi_stats - Update VSI stats counters
5370 * @vsi: the VSI to be updated
5372 void ice_update_vsi_stats(struct ice_vsi *vsi)
5374 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5375 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5376 struct ice_pf *pf = vsi->back;
5378 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
5379 test_bit(__ICE_CFG_BUSY, pf->state))
5382 /* get stats as recorded by Tx/Rx rings */
5383 ice_update_vsi_ring_stats(vsi);
5385 /* get VSI stats as recorded by the hardware */
5386 ice_update_eth_stats(vsi);
5388 cur_ns->tx_errors = cur_es->tx_errors;
5389 cur_ns->rx_dropped = cur_es->rx_discards;
5390 cur_ns->tx_dropped = cur_es->tx_discards;
5391 cur_ns->multicast = cur_es->rx_multicast;
5393 /* update some more netdev stats if this is main VSI */
5394 if (vsi->type == ICE_VSI_PF) {
5395 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5396 cur_ns->rx_errors = pf->stats.crc_errors +
5397 pf->stats.illegal_bytes +
5398 pf->stats.rx_len_errors +
5399 pf->stats.rx_undersize +
5400 pf->hw_csum_rx_error +
5401 pf->stats.rx_jabber +
5402 pf->stats.rx_fragments +
5403 pf->stats.rx_oversize;
5404 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5405 /* record drops from the port level */
5406 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5411 * ice_update_pf_stats - Update PF port stats counters
5412 * @pf: PF whose stats needs to be updated
5414 void ice_update_pf_stats(struct ice_pf *pf)
5416 struct ice_hw_port_stats *prev_ps, *cur_ps;
5417 struct ice_hw *hw = &pf->hw;
5421 port = hw->port_info->lport;
5422 prev_ps = &pf->stats_prev;
5423 cur_ps = &pf->stats;
5425 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5426 &prev_ps->eth.rx_bytes,
5427 &cur_ps->eth.rx_bytes);
5429 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5430 &prev_ps->eth.rx_unicast,
5431 &cur_ps->eth.rx_unicast);
5433 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5434 &prev_ps->eth.rx_multicast,
5435 &cur_ps->eth.rx_multicast);
5437 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5438 &prev_ps->eth.rx_broadcast,
5439 &cur_ps->eth.rx_broadcast);
5441 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5442 &prev_ps->eth.rx_discards,
5443 &cur_ps->eth.rx_discards);
5445 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5446 &prev_ps->eth.tx_bytes,
5447 &cur_ps->eth.tx_bytes);
5449 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5450 &prev_ps->eth.tx_unicast,
5451 &cur_ps->eth.tx_unicast);
5453 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5454 &prev_ps->eth.tx_multicast,
5455 &cur_ps->eth.tx_multicast);
5457 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5458 &prev_ps->eth.tx_broadcast,
5459 &cur_ps->eth.tx_broadcast);
5461 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5462 &prev_ps->tx_dropped_link_down,
5463 &cur_ps->tx_dropped_link_down);
5465 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5466 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5468 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5469 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5471 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5472 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5474 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5475 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5477 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5478 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5480 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5481 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5483 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5484 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5486 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5487 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5489 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5490 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5492 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5493 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5495 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5496 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5498 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5499 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5501 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5502 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5504 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5505 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5507 fd_ctr_base = hw->fd_ctr_base;
5509 ice_stat_update40(hw,
5510 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5511 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5512 &cur_ps->fd_sb_match);
5513 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5514 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5516 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5517 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5519 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5520 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5522 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5523 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5525 ice_update_dcb_stats(pf);
5527 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5528 &prev_ps->crc_errors, &cur_ps->crc_errors);
5530 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5531 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5533 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5534 &prev_ps->mac_local_faults,
5535 &cur_ps->mac_local_faults);
5537 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5538 &prev_ps->mac_remote_faults,
5539 &cur_ps->mac_remote_faults);
5541 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5542 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5544 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5545 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5547 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5548 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5550 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5551 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5553 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5554 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5556 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5558 pf->stat_prev_loaded = true;
5562 * ice_get_stats64 - get statistics for network device structure
5563 * @netdev: network interface device structure
5564 * @stats: main device statistics structure
5567 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5569 struct ice_netdev_priv *np = netdev_priv(netdev);
5570 struct rtnl_link_stats64 *vsi_stats;
5571 struct ice_vsi *vsi = np->vsi;
5573 vsi_stats = &vsi->net_stats;
5575 if (!vsi->num_txq || !vsi->num_rxq)
5578 /* netdev packet/byte stats come from ring counter. These are obtained
5579 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5580 * But, only call the update routine and read the registers if VSI is
5583 if (!test_bit(ICE_VSI_DOWN, vsi->state))
5584 ice_update_vsi_ring_stats(vsi);
5585 stats->tx_packets = vsi_stats->tx_packets;
5586 stats->tx_bytes = vsi_stats->tx_bytes;
5587 stats->rx_packets = vsi_stats->rx_packets;
5588 stats->rx_bytes = vsi_stats->rx_bytes;
5590 /* The rest of the stats can be read from the hardware but instead we
5591 * just return values that the watchdog task has already obtained from
5594 stats->multicast = vsi_stats->multicast;
5595 stats->tx_errors = vsi_stats->tx_errors;
5596 stats->tx_dropped = vsi_stats->tx_dropped;
5597 stats->rx_errors = vsi_stats->rx_errors;
5598 stats->rx_dropped = vsi_stats->rx_dropped;
5599 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5600 stats->rx_length_errors = vsi_stats->rx_length_errors;
5604 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5605 * @vsi: VSI having NAPI disabled
5607 static void ice_napi_disable_all(struct ice_vsi *vsi)
5614 ice_for_each_q_vector(vsi, q_idx) {
5615 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5617 if (q_vector->rx.ring || q_vector->tx.ring)
5618 napi_disable(&q_vector->napi);
5623 * ice_down - Shutdown the connection
5624 * @vsi: The VSI being stopped
5626 int ice_down(struct ice_vsi *vsi)
5628 int i, tx_err, rx_err, link_err = 0;
5630 /* Caller of this function is expected to set the
5631 * vsi->state __ICE_DOWN bit
5634 netif_carrier_off(vsi->netdev);
5635 netif_tx_disable(vsi->netdev);
5638 ice_vsi_dis_irq(vsi);
5640 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5642 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5643 vsi->vsi_num, tx_err);
5644 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5645 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5647 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5648 vsi->vsi_num, tx_err);
5651 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5653 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5654 vsi->vsi_num, rx_err);
5656 ice_napi_disable_all(vsi);
5658 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5659 link_err = ice_force_phys_link_state(vsi, false);
5661 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5662 vsi->vsi_num, link_err);
5665 ice_for_each_txq(vsi, i)
5666 ice_clean_tx_ring(vsi->tx_rings[i]);
5668 ice_for_each_rxq(vsi, i)
5669 ice_clean_rx_ring(vsi->rx_rings[i]);
5671 if (tx_err || rx_err || link_err) {
5672 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5673 vsi->vsi_num, vsi->vsw->sw_id);
5681 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5682 * @vsi: VSI having resources allocated
5684 * Return 0 on success, negative on failure
5686 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5690 if (!vsi->num_txq) {
5691 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5696 ice_for_each_txq(vsi, i) {
5697 struct ice_ring *ring = vsi->tx_rings[i];
5702 ring->netdev = vsi->netdev;
5703 err = ice_setup_tx_ring(ring);
5712 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5713 * @vsi: VSI having resources allocated
5715 * Return 0 on success, negative on failure
5717 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5721 if (!vsi->num_rxq) {
5722 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5727 ice_for_each_rxq(vsi, i) {
5728 struct ice_ring *ring = vsi->rx_rings[i];
5733 ring->netdev = vsi->netdev;
5734 err = ice_setup_rx_ring(ring);
5743 * ice_vsi_open_ctrl - open control VSI for use
5744 * @vsi: the VSI to open
5746 * Initialization of the Control VSI
5748 * Returns 0 on success, negative value on error
5750 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5752 char int_name[ICE_INT_NAME_STR_LEN];
5753 struct ice_pf *pf = vsi->back;
5757 dev = ice_pf_to_dev(pf);
5758 /* allocate descriptors */
5759 err = ice_vsi_setup_tx_rings(vsi);
5763 err = ice_vsi_setup_rx_rings(vsi);
5767 err = ice_vsi_cfg(vsi);
5771 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5772 dev_driver_string(dev), dev_name(dev));
5773 err = ice_vsi_req_irq_msix(vsi, int_name);
5777 ice_vsi_cfg_msix(vsi);
5779 err = ice_vsi_start_all_rx_rings(vsi);
5781 goto err_up_complete;
5783 clear_bit(ICE_VSI_DOWN, vsi->state);
5784 ice_vsi_ena_irq(vsi);
5791 ice_vsi_free_rx_rings(vsi);
5793 ice_vsi_free_tx_rings(vsi);
5799 * ice_vsi_open - Called when a network interface is made active
5800 * @vsi: the VSI to open
5802 * Initialization of the VSI
5804 * Returns 0 on success, negative value on error
5806 static int ice_vsi_open(struct ice_vsi *vsi)
5808 char int_name[ICE_INT_NAME_STR_LEN];
5809 struct ice_pf *pf = vsi->back;
5812 /* allocate descriptors */
5813 err = ice_vsi_setup_tx_rings(vsi);
5817 err = ice_vsi_setup_rx_rings(vsi);
5821 err = ice_vsi_cfg(vsi);
5825 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5826 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
5827 err = ice_vsi_req_irq_msix(vsi, int_name);
5831 /* Notify the stack of the actual queue counts. */
5832 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
5836 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
5840 err = ice_up_complete(vsi);
5842 goto err_up_complete;
5849 ice_vsi_free_irq(vsi);
5851 ice_vsi_free_rx_rings(vsi);
5853 ice_vsi_free_tx_rings(vsi);
5859 * ice_vsi_release_all - Delete all VSIs
5860 * @pf: PF from which all VSIs are being removed
5862 static void ice_vsi_release_all(struct ice_pf *pf)
5869 ice_for_each_vsi(pf, i) {
5873 err = ice_vsi_release(pf->vsi[i]);
5875 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
5876 i, err, pf->vsi[i]->vsi_num);
5881 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
5882 * @pf: pointer to the PF instance
5883 * @type: VSI type to rebuild
5885 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
5887 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
5889 struct device *dev = ice_pf_to_dev(pf);
5890 enum ice_status status;
5893 ice_for_each_vsi(pf, i) {
5894 struct ice_vsi *vsi = pf->vsi[i];
5896 if (!vsi || vsi->type != type)
5899 /* rebuild the VSI */
5900 err = ice_vsi_rebuild(vsi, true);
5902 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
5903 err, vsi->idx, ice_vsi_type_str(type));
5907 /* replay filters for the VSI */
5908 status = ice_replay_vsi(&pf->hw, vsi->idx);
5910 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
5911 ice_stat_str(status), vsi->idx,
5912 ice_vsi_type_str(type));
5916 /* Re-map HW VSI number, using VSI handle that has been
5917 * previously validated in ice_replay_vsi() call above
5919 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
5921 /* enable the VSI */
5922 err = ice_ena_vsi(vsi, false);
5924 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
5925 err, vsi->idx, ice_vsi_type_str(type));
5929 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
5930 ice_vsi_type_str(type));
5937 * ice_update_pf_netdev_link - Update PF netdev link status
5938 * @pf: pointer to the PF instance
5940 static void ice_update_pf_netdev_link(struct ice_pf *pf)
5945 ice_for_each_vsi(pf, i) {
5946 struct ice_vsi *vsi = pf->vsi[i];
5948 if (!vsi || vsi->type != ICE_VSI_PF)
5951 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
5953 netif_carrier_on(pf->vsi[i]->netdev);
5954 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
5956 netif_carrier_off(pf->vsi[i]->netdev);
5957 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
5963 * ice_rebuild - rebuild after reset
5964 * @pf: PF to rebuild
5965 * @reset_type: type of reset
5967 * Do not rebuild VF VSI in this flow because that is already handled via
5968 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
5969 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
5970 * to reset/rebuild all the VF VSI twice.
5972 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
5974 struct device *dev = ice_pf_to_dev(pf);
5975 struct ice_hw *hw = &pf->hw;
5976 enum ice_status ret;
5979 if (test_bit(__ICE_DOWN, pf->state))
5980 goto clear_recovery;
5982 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
5984 ret = ice_init_all_ctrlq(hw);
5986 dev_err(dev, "control queues init failed %s\n",
5988 goto err_init_ctrlq;
5991 /* if DDP was previously loaded successfully */
5992 if (!ice_is_safe_mode(pf)) {
5993 /* reload the SW DB of filter tables */
5994 if (reset_type == ICE_RESET_PFR)
5995 ice_fill_blk_tbls(hw);
5997 /* Reload DDP Package after CORER/GLOBR reset */
5998 ice_load_pkg(NULL, pf);
6001 ret = ice_clear_pf_cfg(hw);
6003 dev_err(dev, "clear PF configuration failed %s\n",
6005 goto err_init_ctrlq;
6008 if (pf->first_sw->dflt_vsi_ena)
6009 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6010 /* clear the default VSI configuration if it exists */
6011 pf->first_sw->dflt_vsi = NULL;
6012 pf->first_sw->dflt_vsi_ena = false;
6014 ice_clear_pxe_mode(hw);
6016 ret = ice_get_caps(hw);
6018 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
6019 goto err_init_ctrlq;
6022 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6024 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6025 goto err_init_ctrlq;
6028 err = ice_sched_init_port(hw->port_info);
6030 goto err_sched_init_port;
6032 /* start misc vector */
6033 err = ice_req_irq_msix_misc(pf);
6035 dev_err(dev, "misc vector setup failed: %d\n", err);
6036 goto err_sched_init_port;
6039 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6040 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6041 if (!rd32(hw, PFQF_FD_SIZE)) {
6042 u16 unused, guar, b_effort;
6044 guar = hw->func_caps.fd_fltr_guar;
6045 b_effort = hw->func_caps.fd_fltr_best_effort;
6047 /* force guaranteed filter pool for PF */
6048 ice_alloc_fd_guar_item(hw, &unused, guar);
6049 /* force shared filter pool for PF */
6050 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6054 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6055 ice_dcb_rebuild(pf);
6057 /* rebuild PF VSI */
6058 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6060 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6061 goto err_vsi_rebuild;
6064 /* If Flow Director is active */
6065 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6066 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6068 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6069 goto err_vsi_rebuild;
6072 /* replay HW Flow Director recipes */
6074 ice_fdir_replay_flows(hw);
6076 /* replay Flow Director filters */
6077 ice_fdir_replay_fltrs(pf);
6079 ice_rebuild_arfs(pf);
6082 ice_update_pf_netdev_link(pf);
6084 /* tell the firmware we are up */
6085 ret = ice_send_version(pf);
6087 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6089 goto err_vsi_rebuild;
6092 ice_replay_post(hw);
6094 /* if we get here, reset flow is successful */
6095 clear_bit(__ICE_RESET_FAILED, pf->state);
6099 err_sched_init_port:
6100 ice_sched_cleanup_all(hw);
6102 ice_shutdown_all_ctrlq(hw);
6103 set_bit(__ICE_RESET_FAILED, pf->state);
6105 /* set this bit in PF state to control service task scheduling */
6106 set_bit(__ICE_NEEDS_RESTART, pf->state);
6107 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6111 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6112 * @vsi: Pointer to VSI structure
6114 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6116 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6117 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6119 return ICE_RXBUF_3072;
6123 * ice_change_mtu - NDO callback to change the MTU
6124 * @netdev: network interface device structure
6125 * @new_mtu: new value for maximum frame size
6127 * Returns 0 on success, negative on failure
6129 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6131 struct ice_netdev_priv *np = netdev_priv(netdev);
6132 struct ice_vsi *vsi = np->vsi;
6133 struct ice_pf *pf = vsi->back;
6136 if (new_mtu == (int)netdev->mtu) {
6137 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6141 if (ice_is_xdp_ena_vsi(vsi)) {
6142 int frame_size = ice_max_xdp_frame_size(vsi);
6144 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6145 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6146 frame_size - ICE_ETH_PKT_HDR_PAD);
6151 /* if a reset is in progress, wait for some time for it to complete */
6153 if (ice_is_reset_in_progress(pf->state)) {
6155 usleep_range(1000, 2000);
6160 } while (count < 100);
6163 netdev_err(netdev, "can't change MTU. Device is busy\n");
6167 netdev->mtu = (unsigned int)new_mtu;
6169 /* if VSI is up, bring it down and then back up */
6170 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6173 err = ice_down(vsi);
6175 netdev_err(netdev, "change MTU if_down err %d\n", err);
6181 netdev_err(netdev, "change MTU if_up err %d\n", err);
6186 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6191 * ice_aq_str - convert AQ err code to a string
6192 * @aq_err: the AQ error code to convert
6194 const char *ice_aq_str(enum ice_aq_err aq_err)
6199 case ICE_AQ_RC_EPERM:
6200 return "ICE_AQ_RC_EPERM";
6201 case ICE_AQ_RC_ENOENT:
6202 return "ICE_AQ_RC_ENOENT";
6203 case ICE_AQ_RC_ENOMEM:
6204 return "ICE_AQ_RC_ENOMEM";
6205 case ICE_AQ_RC_EBUSY:
6206 return "ICE_AQ_RC_EBUSY";
6207 case ICE_AQ_RC_EEXIST:
6208 return "ICE_AQ_RC_EEXIST";
6209 case ICE_AQ_RC_EINVAL:
6210 return "ICE_AQ_RC_EINVAL";
6211 case ICE_AQ_RC_ENOSPC:
6212 return "ICE_AQ_RC_ENOSPC";
6213 case ICE_AQ_RC_ENOSYS:
6214 return "ICE_AQ_RC_ENOSYS";
6215 case ICE_AQ_RC_EMODE:
6216 return "ICE_AQ_RC_EMODE";
6217 case ICE_AQ_RC_ENOSEC:
6218 return "ICE_AQ_RC_ENOSEC";
6219 case ICE_AQ_RC_EBADSIG:
6220 return "ICE_AQ_RC_EBADSIG";
6221 case ICE_AQ_RC_ESVN:
6222 return "ICE_AQ_RC_ESVN";
6223 case ICE_AQ_RC_EBADMAN:
6224 return "ICE_AQ_RC_EBADMAN";
6225 case ICE_AQ_RC_EBADBUF:
6226 return "ICE_AQ_RC_EBADBUF";
6229 return "ICE_AQ_RC_UNKNOWN";
6233 * ice_stat_str - convert status err code to a string
6234 * @stat_err: the status error code to convert
6236 const char *ice_stat_str(enum ice_status stat_err)
6242 return "ICE_ERR_PARAM";
6243 case ICE_ERR_NOT_IMPL:
6244 return "ICE_ERR_NOT_IMPL";
6245 case ICE_ERR_NOT_READY:
6246 return "ICE_ERR_NOT_READY";
6247 case ICE_ERR_NOT_SUPPORTED:
6248 return "ICE_ERR_NOT_SUPPORTED";
6249 case ICE_ERR_BAD_PTR:
6250 return "ICE_ERR_BAD_PTR";
6251 case ICE_ERR_INVAL_SIZE:
6252 return "ICE_ERR_INVAL_SIZE";
6253 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6254 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6255 case ICE_ERR_RESET_FAILED:
6256 return "ICE_ERR_RESET_FAILED";
6257 case ICE_ERR_FW_API_VER:
6258 return "ICE_ERR_FW_API_VER";
6259 case ICE_ERR_NO_MEMORY:
6260 return "ICE_ERR_NO_MEMORY";
6262 return "ICE_ERR_CFG";
6263 case ICE_ERR_OUT_OF_RANGE:
6264 return "ICE_ERR_OUT_OF_RANGE";
6265 case ICE_ERR_ALREADY_EXISTS:
6266 return "ICE_ERR_ALREADY_EXISTS";
6268 return "ICE_ERR_NVM";
6269 case ICE_ERR_NVM_CHECKSUM:
6270 return "ICE_ERR_NVM_CHECKSUM";
6271 case ICE_ERR_BUF_TOO_SHORT:
6272 return "ICE_ERR_BUF_TOO_SHORT";
6273 case ICE_ERR_NVM_BLANK_MODE:
6274 return "ICE_ERR_NVM_BLANK_MODE";
6275 case ICE_ERR_IN_USE:
6276 return "ICE_ERR_IN_USE";
6277 case ICE_ERR_MAX_LIMIT:
6278 return "ICE_ERR_MAX_LIMIT";
6279 case ICE_ERR_RESET_ONGOING:
6280 return "ICE_ERR_RESET_ONGOING";
6281 case ICE_ERR_HW_TABLE:
6282 return "ICE_ERR_HW_TABLE";
6283 case ICE_ERR_DOES_NOT_EXIST:
6284 return "ICE_ERR_DOES_NOT_EXIST";
6285 case ICE_ERR_FW_DDP_MISMATCH:
6286 return "ICE_ERR_FW_DDP_MISMATCH";
6287 case ICE_ERR_AQ_ERROR:
6288 return "ICE_ERR_AQ_ERROR";
6289 case ICE_ERR_AQ_TIMEOUT:
6290 return "ICE_ERR_AQ_TIMEOUT";
6291 case ICE_ERR_AQ_FULL:
6292 return "ICE_ERR_AQ_FULL";
6293 case ICE_ERR_AQ_NO_WORK:
6294 return "ICE_ERR_AQ_NO_WORK";
6295 case ICE_ERR_AQ_EMPTY:
6296 return "ICE_ERR_AQ_EMPTY";
6297 case ICE_ERR_AQ_FW_CRITICAL:
6298 return "ICE_ERR_AQ_FW_CRITICAL";
6301 return "ICE_ERR_UNKNOWN";
6305 * ice_set_rss_lut - Set RSS LUT
6306 * @vsi: Pointer to VSI structure
6307 * @lut: Lookup table
6308 * @lut_size: Lookup table size
6310 * Returns 0 on success, negative on failure
6312 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6314 struct ice_aq_get_set_rss_lut_params params = {};
6315 struct ice_hw *hw = &vsi->back->hw;
6316 enum ice_status status;
6321 params.vsi_handle = vsi->idx;
6322 params.lut_size = lut_size;
6323 params.lut_type = vsi->rss_lut_type;
6326 status = ice_aq_set_rss_lut(hw, ¶ms);
6328 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %s aq_err %s\n",
6329 ice_stat_str(status),
6330 ice_aq_str(hw->adminq.sq_last_status));
6338 * ice_set_rss_key - Set RSS key
6339 * @vsi: Pointer to the VSI structure
6340 * @seed: RSS hash seed
6342 * Returns 0 on success, negative on failure
6344 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
6346 struct ice_hw *hw = &vsi->back->hw;
6347 enum ice_status status;
6352 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6354 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %s aq_err %s\n",
6355 ice_stat_str(status),
6356 ice_aq_str(hw->adminq.sq_last_status));
6364 * ice_get_rss_lut - Get RSS LUT
6365 * @vsi: Pointer to VSI structure
6366 * @lut: Buffer to store the lookup table entries
6367 * @lut_size: Size of buffer to store the lookup table entries
6369 * Returns 0 on success, negative on failure
6371 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6373 struct ice_aq_get_set_rss_lut_params params = {};
6374 struct ice_hw *hw = &vsi->back->hw;
6375 enum ice_status status;
6380 params.vsi_handle = vsi->idx;
6381 params.lut_size = lut_size;
6382 params.lut_type = vsi->rss_lut_type;
6385 status = ice_aq_get_rss_lut(hw, ¶ms);
6387 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %s aq_err %s\n",
6388 ice_stat_str(status),
6389 ice_aq_str(hw->adminq.sq_last_status));
6397 * ice_get_rss_key - Get RSS key
6398 * @vsi: Pointer to VSI structure
6399 * @seed: Buffer to store the key in
6401 * Returns 0 on success, negative on failure
6403 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
6405 struct ice_hw *hw = &vsi->back->hw;
6406 enum ice_status status;
6411 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6413 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %s aq_err %s\n",
6414 ice_stat_str(status),
6415 ice_aq_str(hw->adminq.sq_last_status));
6423 * ice_bridge_getlink - Get the hardware bridge mode
6426 * @seq: RTNL message seq
6427 * @dev: the netdev being configured
6428 * @filter_mask: filter mask passed in
6429 * @nlflags: netlink flags passed in
6431 * Return the bridge mode (VEB/VEPA)
6434 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6435 struct net_device *dev, u32 filter_mask, int nlflags)
6437 struct ice_netdev_priv *np = netdev_priv(dev);
6438 struct ice_vsi *vsi = np->vsi;
6439 struct ice_pf *pf = vsi->back;
6442 bmode = pf->first_sw->bridge_mode;
6444 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6449 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6450 * @vsi: Pointer to VSI structure
6451 * @bmode: Hardware bridge mode (VEB/VEPA)
6453 * Returns 0 on success, negative on failure
6455 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6457 struct ice_aqc_vsi_props *vsi_props;
6458 struct ice_hw *hw = &vsi->back->hw;
6459 struct ice_vsi_ctx *ctxt;
6460 enum ice_status status;
6463 vsi_props = &vsi->info;
6465 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6469 ctxt->info = vsi->info;
6471 if (bmode == BRIDGE_MODE_VEB)
6472 /* change from VEPA to VEB mode */
6473 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6475 /* change from VEB to VEPA mode */
6476 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6477 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6479 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6481 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6482 bmode, ice_stat_str(status),
6483 ice_aq_str(hw->adminq.sq_last_status));
6487 /* Update sw flags for book keeping */
6488 vsi_props->sw_flags = ctxt->info.sw_flags;
6496 * ice_bridge_setlink - Set the hardware bridge mode
6497 * @dev: the netdev being configured
6498 * @nlh: RTNL message
6499 * @flags: bridge setlink flags
6500 * @extack: netlink extended ack
6502 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6503 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6504 * not already set for all VSIs connected to this switch. And also update the
6505 * unicast switch filter rules for the corresponding switch of the netdev.
6508 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6509 u16 __always_unused flags,
6510 struct netlink_ext_ack __always_unused *extack)
6512 struct ice_netdev_priv *np = netdev_priv(dev);
6513 struct ice_pf *pf = np->vsi->back;
6514 struct nlattr *attr, *br_spec;
6515 struct ice_hw *hw = &pf->hw;
6516 enum ice_status status;
6517 struct ice_sw *pf_sw;
6518 int rem, v, err = 0;
6520 pf_sw = pf->first_sw;
6521 /* find the attribute in the netlink message */
6522 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6524 nla_for_each_nested(attr, br_spec, rem) {
6527 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6529 mode = nla_get_u16(attr);
6530 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6532 /* Continue if bridge mode is not being flipped */
6533 if (mode == pf_sw->bridge_mode)
6535 /* Iterates through the PF VSI list and update the loopback
6538 ice_for_each_vsi(pf, v) {
6541 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6546 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6547 /* Update the unicast switch filter rules for the corresponding
6548 * switch of the netdev
6550 status = ice_update_sw_rule_bridge_mode(hw);
6552 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6553 mode, ice_stat_str(status),
6554 ice_aq_str(hw->adminq.sq_last_status));
6555 /* revert hw->evb_veb */
6556 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6560 pf_sw->bridge_mode = mode;
6567 * ice_tx_timeout - Respond to a Tx Hang
6568 * @netdev: network interface device structure
6569 * @txqueue: Tx queue
6571 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6573 struct ice_netdev_priv *np = netdev_priv(netdev);
6574 struct ice_ring *tx_ring = NULL;
6575 struct ice_vsi *vsi = np->vsi;
6576 struct ice_pf *pf = vsi->back;
6579 pf->tx_timeout_count++;
6581 /* Check if PFC is enabled for the TC to which the queue belongs
6582 * to. If yes then Tx timeout is not caused by a hung queue, no
6583 * need to reset and rebuild
6585 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6586 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6591 /* now that we have an index, find the tx_ring struct */
6592 for (i = 0; i < vsi->num_txq; i++)
6593 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6594 if (txqueue == vsi->tx_rings[i]->q_index) {
6595 tx_ring = vsi->tx_rings[i];
6599 /* Reset recovery level if enough time has elapsed after last timeout.
6600 * Also ensure no new reset action happens before next timeout period.
6602 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6603 pf->tx_timeout_recovery_level = 1;
6604 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6605 netdev->watchdog_timeo)))
6609 struct ice_hw *hw = &pf->hw;
6612 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6613 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6614 /* Read interrupt register */
6615 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6617 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6618 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6619 head, tx_ring->next_to_use, val);
6622 pf->tx_timeout_last_recovery = jiffies;
6623 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6624 pf->tx_timeout_recovery_level, txqueue);
6626 switch (pf->tx_timeout_recovery_level) {
6628 set_bit(__ICE_PFR_REQ, pf->state);
6631 set_bit(__ICE_CORER_REQ, pf->state);
6634 set_bit(__ICE_GLOBR_REQ, pf->state);
6637 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6638 set_bit(__ICE_DOWN, pf->state);
6639 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
6640 set_bit(__ICE_SERVICE_DIS, pf->state);
6644 ice_service_task_schedule(pf);
6645 pf->tx_timeout_recovery_level++;
6649 * ice_open - Called when a network interface becomes active
6650 * @netdev: network interface device structure
6652 * The open entry point is called when a network interface is made
6653 * active by the system (IFF_UP). At this point all resources needed
6654 * for transmit and receive operations are allocated, the interrupt
6655 * handler is registered with the OS, the netdev watchdog is enabled,
6656 * and the stack is notified that the interface is ready.
6658 * Returns 0 on success, negative value on failure
6660 int ice_open(struct net_device *netdev)
6662 struct ice_netdev_priv *np = netdev_priv(netdev);
6663 struct ice_vsi *vsi = np->vsi;
6664 struct ice_pf *pf = vsi->back;
6665 struct ice_port_info *pi;
6666 enum ice_status status;
6669 if (test_bit(__ICE_NEEDS_RESTART, pf->state)) {
6670 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6674 netif_carrier_off(netdev);
6676 pi = vsi->port_info;
6677 status = ice_update_link_info(pi);
6679 netdev_err(netdev, "Failed to get link info, error %s\n",
6680 ice_stat_str(status));
6684 /* Set PHY if there is media, otherwise, turn off PHY */
6685 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6686 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6687 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state)) {
6688 err = ice_init_phy_user_cfg(pi);
6690 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6696 err = ice_configure_phy(vsi);
6698 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6703 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6704 ice_set_link(vsi, false);
6707 err = ice_vsi_open(vsi);
6709 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6710 vsi->vsi_num, vsi->vsw->sw_id);
6712 /* Update existing tunnels information */
6713 udp_tunnel_get_rx_info(netdev);
6719 * ice_stop - Disables a network interface
6720 * @netdev: network interface device structure
6722 * The stop entry point is called when an interface is de-activated by the OS,
6723 * and the netdevice enters the DOWN state. The hardware is still under the
6724 * driver's control, but the netdev interface is disabled.
6726 * Returns success only - not allowed to fail
6728 int ice_stop(struct net_device *netdev)
6730 struct ice_netdev_priv *np = netdev_priv(netdev);
6731 struct ice_vsi *vsi = np->vsi;
6739 * ice_features_check - Validate encapsulated packet conforms to limits
6741 * @netdev: This port's netdev
6742 * @features: Offload features that the stack believes apply
6744 static netdev_features_t
6745 ice_features_check(struct sk_buff *skb,
6746 struct net_device __always_unused *netdev,
6747 netdev_features_t features)
6751 /* No point in doing any of this if neither checksum nor GSO are
6752 * being requested for this frame. We can rule out both by just
6753 * checking for CHECKSUM_PARTIAL
6755 if (skb->ip_summed != CHECKSUM_PARTIAL)
6758 /* We cannot support GSO if the MSS is going to be less than
6759 * 64 bytes. If it is then we need to drop support for GSO.
6761 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
6762 features &= ~NETIF_F_GSO_MASK;
6764 len = skb_network_header(skb) - skb->data;
6765 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
6766 goto out_rm_features;
6768 len = skb_transport_header(skb) - skb_network_header(skb);
6769 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6770 goto out_rm_features;
6772 if (skb->encapsulation) {
6773 len = skb_inner_network_header(skb) - skb_transport_header(skb);
6774 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
6775 goto out_rm_features;
6777 len = skb_inner_transport_header(skb) -
6778 skb_inner_network_header(skb);
6779 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6780 goto out_rm_features;
6785 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
6788 static const struct net_device_ops ice_netdev_safe_mode_ops = {
6789 .ndo_open = ice_open,
6790 .ndo_stop = ice_stop,
6791 .ndo_start_xmit = ice_start_xmit,
6792 .ndo_set_mac_address = ice_set_mac_address,
6793 .ndo_validate_addr = eth_validate_addr,
6794 .ndo_change_mtu = ice_change_mtu,
6795 .ndo_get_stats64 = ice_get_stats64,
6796 .ndo_tx_timeout = ice_tx_timeout,
6799 static const struct net_device_ops ice_netdev_ops = {
6800 .ndo_open = ice_open,
6801 .ndo_stop = ice_stop,
6802 .ndo_start_xmit = ice_start_xmit,
6803 .ndo_features_check = ice_features_check,
6804 .ndo_set_rx_mode = ice_set_rx_mode,
6805 .ndo_set_mac_address = ice_set_mac_address,
6806 .ndo_validate_addr = eth_validate_addr,
6807 .ndo_change_mtu = ice_change_mtu,
6808 .ndo_get_stats64 = ice_get_stats64,
6809 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
6810 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
6811 .ndo_set_vf_mac = ice_set_vf_mac,
6812 .ndo_get_vf_config = ice_get_vf_cfg,
6813 .ndo_set_vf_trust = ice_set_vf_trust,
6814 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
6815 .ndo_set_vf_link_state = ice_set_vf_link_state,
6816 .ndo_get_vf_stats = ice_get_vf_stats,
6817 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
6818 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
6819 .ndo_set_features = ice_set_features,
6820 .ndo_bridge_getlink = ice_bridge_getlink,
6821 .ndo_bridge_setlink = ice_bridge_setlink,
6822 .ndo_fdb_add = ice_fdb_add,
6823 .ndo_fdb_del = ice_fdb_del,
6824 #ifdef CONFIG_RFS_ACCEL
6825 .ndo_rx_flow_steer = ice_rx_flow_steer,
6827 .ndo_tx_timeout = ice_tx_timeout,
6829 .ndo_xdp_xmit = ice_xdp_xmit,
6830 .ndo_xsk_wakeup = ice_xsk_wakeup,