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);
48 * ice_get_tx_pending - returns number of Tx descriptors not processed
49 * @ring: the ring of descriptors
51 static u16 ice_get_tx_pending(struct ice_ring *ring)
55 head = ring->next_to_clean;
56 tail = ring->next_to_use;
59 return (head < tail) ?
60 tail - head : (tail + ring->count - head);
65 * ice_check_for_hang_subtask - check for and recover hung queues
66 * @pf: pointer to PF struct
68 static void ice_check_for_hang_subtask(struct ice_pf *pf)
70 struct ice_vsi *vsi = NULL;
76 ice_for_each_vsi(pf, v)
77 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
82 if (!vsi || test_bit(__ICE_DOWN, vsi->state))
85 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
90 for (i = 0; i < vsi->num_txq; i++) {
91 struct ice_ring *tx_ring = vsi->tx_rings[i];
93 if (tx_ring && tx_ring->desc) {
94 /* If packet counter has not changed the queue is
95 * likely stalled, so force an interrupt for this
98 * prev_pkt would be negative if there was no
101 packets = tx_ring->stats.pkts & INT_MAX;
102 if (tx_ring->tx_stats.prev_pkt == packets) {
103 /* Trigger sw interrupt to revive the queue */
104 ice_trigger_sw_intr(hw, tx_ring->q_vector);
108 /* Memory barrier between read of packet count and call
109 * to ice_get_tx_pending()
112 tx_ring->tx_stats.prev_pkt =
113 ice_get_tx_pending(tx_ring) ? packets : -1;
119 * ice_init_mac_fltr - Set initial MAC filters
120 * @pf: board private structure
122 * Set initial set of MAC filters for PF VSI; configure filters for permanent
123 * address and broadcast address. If an error is encountered, netdevice will be
126 static int ice_init_mac_fltr(struct ice_pf *pf)
128 enum ice_status status;
132 vsi = ice_get_main_vsi(pf);
136 perm_addr = vsi->port_info->mac.perm_addr;
137 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
141 /* We aren't useful with no MAC filters, so unregister if we
144 if (vsi->netdev->reg_state == NETREG_REGISTERED) {
145 dev_err(ice_pf_to_dev(pf), "Could not add MAC filters error %s. Unregistering device\n",
146 ice_stat_str(status));
147 unregister_netdev(vsi->netdev);
148 free_netdev(vsi->netdev);
156 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
157 * @netdev: the net device on which the sync is happening
158 * @addr: MAC address to sync
160 * This is a callback function which is called by the in kernel device sync
161 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
162 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
163 * MAC filters from the hardware.
165 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
167 struct ice_netdev_priv *np = netdev_priv(netdev);
168 struct ice_vsi *vsi = np->vsi;
170 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
178 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
179 * @netdev: the net device on which the unsync is happening
180 * @addr: MAC address to unsync
182 * This is a callback function which is called by the in kernel device unsync
183 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
184 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
185 * delete the MAC filters from the hardware.
187 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
189 struct ice_netdev_priv *np = netdev_priv(netdev);
190 struct ice_vsi *vsi = np->vsi;
192 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
200 * ice_vsi_fltr_changed - check if filter state changed
201 * @vsi: VSI to be checked
203 * returns true if filter state has changed, false otherwise.
205 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
207 return test_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags) ||
208 test_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags) ||
209 test_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
213 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
214 * @vsi: the VSI being configured
215 * @promisc_m: mask of promiscuous config bits
216 * @set_promisc: enable or disable promisc flag request
219 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
221 struct ice_hw *hw = &vsi->back->hw;
222 enum ice_status status = 0;
224 if (vsi->type != ICE_VSI_PF)
228 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
232 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
235 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
246 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
247 * @vsi: ptr to the VSI
249 * Push any outstanding VSI filter changes through the AdminQ.
251 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
253 struct device *dev = ice_pf_to_dev(vsi->back);
254 struct net_device *netdev = vsi->netdev;
255 bool promisc_forced_on = false;
256 struct ice_pf *pf = vsi->back;
257 struct ice_hw *hw = &pf->hw;
258 enum ice_status status = 0;
259 u32 changed_flags = 0;
266 while (test_and_set_bit(__ICE_CFG_BUSY, vsi->state))
267 usleep_range(1000, 2000);
269 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
270 vsi->current_netdev_flags = vsi->netdev->flags;
272 INIT_LIST_HEAD(&vsi->tmp_sync_list);
273 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
275 if (ice_vsi_fltr_changed(vsi)) {
276 clear_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
277 clear_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
278 clear_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
280 /* grab the netdev's addr_list_lock */
281 netif_addr_lock_bh(netdev);
282 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
283 ice_add_mac_to_unsync_list);
284 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
285 ice_add_mac_to_unsync_list);
286 /* our temp lists are populated. release lock */
287 netif_addr_unlock_bh(netdev);
290 /* Remove MAC addresses in the unsync list */
291 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
292 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
294 netdev_err(netdev, "Failed to delete MAC filters\n");
295 /* if we failed because of alloc failures, just bail */
296 if (status == ICE_ERR_NO_MEMORY) {
302 /* Add MAC addresses in the sync list */
303 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
304 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
305 /* If filter is added successfully or already exists, do not go into
306 * 'if' condition and report it as error. Instead continue processing
307 * rest of the function.
309 if (status && status != ICE_ERR_ALREADY_EXISTS) {
310 netdev_err(netdev, "Failed to add MAC filters\n");
311 /* If there is no more space for new umac filters, VSI
312 * should go into promiscuous mode. There should be some
313 * space reserved for promiscuous filters.
315 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
316 !test_and_set_bit(__ICE_FLTR_OVERFLOW_PROMISC,
318 promisc_forced_on = true;
319 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
326 /* check for changes in promiscuous modes */
327 if (changed_flags & IFF_ALLMULTI) {
328 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
330 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
332 promisc_m = ICE_MCAST_PROMISC_BITS;
334 err = ice_cfg_promisc(vsi, promisc_m, true);
336 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
338 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
342 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
344 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
346 promisc_m = ICE_MCAST_PROMISC_BITS;
348 err = ice_cfg_promisc(vsi, promisc_m, false);
350 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
352 vsi->current_netdev_flags |= IFF_ALLMULTI;
358 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
359 test_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags)) {
360 clear_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
361 if (vsi->current_netdev_flags & IFF_PROMISC) {
362 /* Apply Rx filter rule to get traffic from wire */
363 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
364 err = ice_set_dflt_vsi(pf->first_sw, vsi);
365 if (err && err != -EEXIST) {
366 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
368 vsi->current_netdev_flags &=
372 ice_cfg_vlan_pruning(vsi, false, false);
375 /* Clear Rx filter to remove traffic from wire */
376 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
377 err = ice_clear_dflt_vsi(pf->first_sw);
379 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
381 vsi->current_netdev_flags |=
385 if (vsi->num_vlan > 1)
386 ice_cfg_vlan_pruning(vsi, true, false);
393 set_bit(ICE_VSI_FLAG_PROMISC_CHANGED, vsi->flags);
396 /* if something went wrong then set the changed flag so we try again */
397 set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
398 set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
400 clear_bit(__ICE_CFG_BUSY, vsi->state);
405 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
406 * @pf: board private structure
408 static void ice_sync_fltr_subtask(struct ice_pf *pf)
412 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
415 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
417 ice_for_each_vsi(pf, v)
418 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
419 ice_vsi_sync_fltr(pf->vsi[v])) {
420 /* come back and try again later */
421 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
427 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
429 * @locked: is the rtnl_lock already held
431 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
435 ice_for_each_vsi(pf, v)
437 ice_dis_vsi(pf->vsi[v], locked);
441 * ice_prepare_for_reset - prep for the core to reset
442 * @pf: board private structure
444 * Inform or close all dependent features in prep for reset.
447 ice_prepare_for_reset(struct ice_pf *pf)
449 struct ice_hw *hw = &pf->hw;
452 /* already prepared for reset */
453 if (test_bit(__ICE_PREPARED_FOR_RESET, pf->state))
456 /* Notify VFs of impending reset */
457 if (ice_check_sq_alive(hw, &hw->mailboxq))
458 ice_vc_notify_reset(pf);
460 /* Disable VFs until reset is completed */
461 ice_for_each_vf(pf, i)
462 ice_set_vf_state_qs_dis(&pf->vf[i]);
464 /* clear SW filtering DB */
465 ice_clear_hw_tbls(hw);
466 /* disable the VSIs and their queues that are not already DOWN */
467 ice_pf_dis_all_vsi(pf, false);
470 ice_sched_clear_port(hw->port_info);
472 ice_shutdown_all_ctrlq(hw);
474 set_bit(__ICE_PREPARED_FOR_RESET, pf->state);
478 * ice_do_reset - Initiate one of many types of resets
479 * @pf: board private structure
480 * @reset_type: reset type requested
481 * before this function was called.
483 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
485 struct device *dev = ice_pf_to_dev(pf);
486 struct ice_hw *hw = &pf->hw;
488 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
490 ice_prepare_for_reset(pf);
492 /* trigger the reset */
493 if (ice_reset(hw, reset_type)) {
494 dev_err(dev, "reset %d failed\n", reset_type);
495 set_bit(__ICE_RESET_FAILED, pf->state);
496 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
497 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
498 clear_bit(__ICE_PFR_REQ, pf->state);
499 clear_bit(__ICE_CORER_REQ, pf->state);
500 clear_bit(__ICE_GLOBR_REQ, pf->state);
504 /* PFR is a bit of a special case because it doesn't result in an OICR
505 * interrupt. So for PFR, rebuild after the reset and clear the reset-
506 * associated state bits.
508 if (reset_type == ICE_RESET_PFR) {
510 ice_rebuild(pf, reset_type);
511 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
512 clear_bit(__ICE_PFR_REQ, pf->state);
513 ice_reset_all_vfs(pf, true);
518 * ice_reset_subtask - Set up for resetting the device and driver
519 * @pf: board private structure
521 static void ice_reset_subtask(struct ice_pf *pf)
523 enum ice_reset_req reset_type = ICE_RESET_INVAL;
525 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
526 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
527 * of reset is pending and sets bits in pf->state indicating the reset
528 * type and __ICE_RESET_OICR_RECV. So, if the latter bit is set
529 * prepare for pending reset if not already (for PF software-initiated
530 * global resets the software should already be prepared for it as
531 * indicated by __ICE_PREPARED_FOR_RESET; for global resets initiated
532 * by firmware or software on other PFs, that bit is not set so prepare
533 * for the reset now), poll for reset done, rebuild and return.
535 if (test_bit(__ICE_RESET_OICR_RECV, pf->state)) {
536 /* Perform the largest reset requested */
537 if (test_and_clear_bit(__ICE_CORER_RECV, pf->state))
538 reset_type = ICE_RESET_CORER;
539 if (test_and_clear_bit(__ICE_GLOBR_RECV, pf->state))
540 reset_type = ICE_RESET_GLOBR;
541 if (test_and_clear_bit(__ICE_EMPR_RECV, pf->state))
542 reset_type = ICE_RESET_EMPR;
543 /* return if no valid reset type requested */
544 if (reset_type == ICE_RESET_INVAL)
546 ice_prepare_for_reset(pf);
548 /* make sure we are ready to rebuild */
549 if (ice_check_reset(&pf->hw)) {
550 set_bit(__ICE_RESET_FAILED, pf->state);
552 /* done with reset. start rebuild */
553 pf->hw.reset_ongoing = false;
554 ice_rebuild(pf, reset_type);
555 /* clear bit to resume normal operations, but
556 * ICE_NEEDS_RESTART bit is set in case rebuild failed
558 clear_bit(__ICE_RESET_OICR_RECV, pf->state);
559 clear_bit(__ICE_PREPARED_FOR_RESET, pf->state);
560 clear_bit(__ICE_PFR_REQ, pf->state);
561 clear_bit(__ICE_CORER_REQ, pf->state);
562 clear_bit(__ICE_GLOBR_REQ, pf->state);
563 ice_reset_all_vfs(pf, true);
569 /* No pending resets to finish processing. Check for new resets */
570 if (test_bit(__ICE_PFR_REQ, pf->state))
571 reset_type = ICE_RESET_PFR;
572 if (test_bit(__ICE_CORER_REQ, pf->state))
573 reset_type = ICE_RESET_CORER;
574 if (test_bit(__ICE_GLOBR_REQ, pf->state))
575 reset_type = ICE_RESET_GLOBR;
576 /* If no valid reset type requested just return */
577 if (reset_type == ICE_RESET_INVAL)
580 /* reset if not already down or busy */
581 if (!test_bit(__ICE_DOWN, pf->state) &&
582 !test_bit(__ICE_CFG_BUSY, pf->state)) {
583 ice_do_reset(pf, reset_type);
588 * ice_print_topo_conflict - print topology conflict message
589 * @vsi: the VSI whose topology status is being checked
591 static void ice_print_topo_conflict(struct ice_vsi *vsi)
593 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
594 case ICE_AQ_LINK_TOPO_CONFLICT:
595 case ICE_AQ_LINK_MEDIA_CONFLICT:
596 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
597 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
598 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
599 netdev_info(vsi->netdev, "Possible mis-configuration of the Ethernet port detected, please use the Intel(R) Ethernet Port Configuration Tool application to address the issue.\n");
601 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
602 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");
610 * ice_print_link_msg - print link up or down message
611 * @vsi: the VSI whose link status is being queried
612 * @isup: boolean for if the link is now up or down
614 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
616 struct ice_aqc_get_phy_caps_data *caps;
617 const char *an_advertised;
618 enum ice_status status;
628 if (vsi->current_isup == isup)
631 vsi->current_isup = isup;
634 netdev_info(vsi->netdev, "NIC Link is Down\n");
638 switch (vsi->port_info->phy.link_info.link_speed) {
639 case ICE_AQ_LINK_SPEED_100GB:
642 case ICE_AQ_LINK_SPEED_50GB:
645 case ICE_AQ_LINK_SPEED_40GB:
648 case ICE_AQ_LINK_SPEED_25GB:
651 case ICE_AQ_LINK_SPEED_20GB:
654 case ICE_AQ_LINK_SPEED_10GB:
657 case ICE_AQ_LINK_SPEED_5GB:
660 case ICE_AQ_LINK_SPEED_2500MB:
663 case ICE_AQ_LINK_SPEED_1000MB:
666 case ICE_AQ_LINK_SPEED_100MB:
674 switch (vsi->port_info->fc.current_mode) {
678 case ICE_FC_TX_PAUSE:
681 case ICE_FC_RX_PAUSE:
692 /* Get FEC mode based on negotiated link info */
693 switch (vsi->port_info->phy.link_info.fec_info) {
694 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
695 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
698 case ICE_AQ_LINK_25G_KR_FEC_EN:
699 fec = "FC-FEC/BASE-R";
706 /* check if autoneg completed, might be false due to not supported */
707 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
712 /* Get FEC mode requested based on PHY caps last SW configuration */
713 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
716 an_advertised = "Unknown";
720 status = ice_aq_get_phy_caps(vsi->port_info, false,
721 ICE_AQC_REPORT_SW_CFG, caps, NULL);
723 netdev_info(vsi->netdev, "Get phy capability failed.\n");
725 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
727 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
728 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
730 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
731 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
732 fec_req = "FC-FEC/BASE-R";
739 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",
740 speed, fec_req, fec, an_advertised, an, fc);
741 ice_print_topo_conflict(vsi);
745 * ice_vsi_link_event - update the VSI's netdev
746 * @vsi: the VSI on which the link event occurred
747 * @link_up: whether or not the VSI needs to be set up or down
749 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
754 if (test_bit(__ICE_DOWN, vsi->state) || !vsi->netdev)
757 if (vsi->type == ICE_VSI_PF) {
758 if (link_up == netif_carrier_ok(vsi->netdev))
762 netif_carrier_on(vsi->netdev);
763 netif_tx_wake_all_queues(vsi->netdev);
765 netif_carrier_off(vsi->netdev);
766 netif_tx_stop_all_queues(vsi->netdev);
772 * ice_set_dflt_mib - send a default config MIB to the FW
773 * @pf: private PF struct
775 * This function sends a default configuration MIB to the FW.
777 * If this function errors out at any point, the driver is still able to
778 * function. The main impact is that LFC may not operate as expected.
779 * Therefore an error state in this function should be treated with a DBG
780 * message and continue on with driver rebuild/reenable.
782 static void ice_set_dflt_mib(struct ice_pf *pf)
784 struct device *dev = ice_pf_to_dev(pf);
785 u8 mib_type, *buf, *lldpmib = NULL;
786 u16 len, typelen, offset = 0;
787 struct ice_lldp_org_tlv *tlv;
792 dev_dbg(dev, "%s NULL pf pointer\n", __func__);
797 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
798 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
800 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
805 /* Add ETS CFG TLV */
806 tlv = (struct ice_lldp_org_tlv *)lldpmib;
807 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
808 ICE_IEEE_ETS_TLV_LEN);
809 tlv->typelen = htons(typelen);
810 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
811 ICE_IEEE_SUBTYPE_ETS_CFG);
812 tlv->ouisubtype = htonl(ouisubtype);
817 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
818 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
819 * Octets 13 - 20 are TSA values - leave as zeros
822 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
824 tlv = (struct ice_lldp_org_tlv *)
825 ((char *)tlv + sizeof(tlv->typelen) + len);
827 /* Add ETS REC TLV */
829 tlv->typelen = htons(typelen);
831 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
832 ICE_IEEE_SUBTYPE_ETS_REC);
833 tlv->ouisubtype = htonl(ouisubtype);
835 /* First octet of buf is reserved
836 * Octets 1 - 4 map UP to TC - all UPs map to zero
837 * Octets 5 - 12 are BW values - set TC 0 to 100%.
838 * Octets 13 - 20 are TSA value - leave as zeros
842 tlv = (struct ice_lldp_org_tlv *)
843 ((char *)tlv + sizeof(tlv->typelen) + len);
845 /* Add PFC CFG TLV */
846 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
847 ICE_IEEE_PFC_TLV_LEN);
848 tlv->typelen = htons(typelen);
850 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
851 ICE_IEEE_SUBTYPE_PFC_CFG);
852 tlv->ouisubtype = htonl(ouisubtype);
854 /* Octet 1 left as all zeros - PFC disabled */
856 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
859 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
860 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
866 * ice_link_event - process the link event
867 * @pf: PF that the link event is associated with
868 * @pi: port_info for the port that the link event is associated with
869 * @link_up: true if the physical link is up and false if it is down
870 * @link_speed: current link speed received from the link event
872 * Returns 0 on success and negative on failure
875 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
878 struct device *dev = ice_pf_to_dev(pf);
879 struct ice_phy_info *phy_info;
886 phy_info->link_info_old = phy_info->link_info;
888 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
889 old_link_speed = phy_info->link_info_old.link_speed;
891 /* update the link info structures and re-enable link events,
892 * don't bail on failure due to other book keeping needed
894 result = ice_update_link_info(pi);
896 dev_dbg(dev, "Failed to update link status and re-enable link events for port %d\n",
899 /* Check if the link state is up after updating link info, and treat
900 * this event as an UP event since the link is actually UP now.
902 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
905 vsi = ice_get_main_vsi(pf);
906 if (!vsi || !vsi->port_info)
909 /* turn off PHY if media was removed */
910 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
911 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
912 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
914 result = ice_aq_set_link_restart_an(pi, false, NULL);
916 dev_dbg(dev, "Failed to set link down, VSI %d error %d\n",
917 vsi->vsi_num, result);
922 /* if the old link up/down and speed is the same as the new */
923 if (link_up == old_link && link_speed == old_link_speed)
926 if (ice_is_dcb_active(pf)) {
927 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
931 ice_set_dflt_mib(pf);
933 ice_vsi_link_event(vsi, link_up);
934 ice_print_link_msg(vsi, link_up);
936 ice_vc_notify_link_state(pf);
942 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
943 * @pf: board private structure
945 static void ice_watchdog_subtask(struct ice_pf *pf)
949 /* if interface is down do nothing */
950 if (test_bit(__ICE_DOWN, pf->state) ||
951 test_bit(__ICE_CFG_BUSY, pf->state))
954 /* make sure we don't do these things too often */
955 if (time_before(jiffies,
956 pf->serv_tmr_prev + pf->serv_tmr_period))
959 pf->serv_tmr_prev = jiffies;
961 /* Update the stats for active netdevs so the network stack
962 * can look at updated numbers whenever it cares to
964 ice_update_pf_stats(pf);
965 ice_for_each_vsi(pf, i)
966 if (pf->vsi[i] && pf->vsi[i]->netdev)
967 ice_update_vsi_stats(pf->vsi[i]);
971 * ice_init_link_events - enable/initialize link events
972 * @pi: pointer to the port_info instance
974 * Returns -EIO on failure, 0 on success
976 static int ice_init_link_events(struct ice_port_info *pi)
980 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
981 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
983 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
984 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
989 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
990 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
999 * ice_handle_link_event - handle link event via ARQ
1000 * @pf: PF that the link event is associated with
1001 * @event: event structure containing link status info
1004 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1006 struct ice_aqc_get_link_status_data *link_data;
1007 struct ice_port_info *port_info;
1010 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1011 port_info = pf->hw.port_info;
1015 status = ice_link_event(pf, port_info,
1016 !!(link_data->link_info & ICE_AQ_LINK_UP),
1017 le16_to_cpu(link_data->link_speed));
1019 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1025 enum ice_aq_task_state {
1026 ICE_AQ_TASK_WAITING = 0,
1027 ICE_AQ_TASK_COMPLETE,
1028 ICE_AQ_TASK_CANCELED,
1031 struct ice_aq_task {
1032 struct hlist_node entry;
1035 struct ice_rq_event_info *event;
1036 enum ice_aq_task_state state;
1040 * ice_wait_for_aq_event - Wait for an AdminQ event from firmware
1041 * @pf: pointer to the PF private structure
1042 * @opcode: the opcode to wait for
1043 * @timeout: how long to wait, in jiffies
1044 * @event: storage for the event info
1046 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1047 * current thread will be put to sleep until the specified event occurs or
1048 * until the given timeout is reached.
1050 * To obtain only the descriptor contents, pass an event without an allocated
1051 * msg_buf. If the complete data buffer is desired, allocate the
1052 * event->msg_buf with enough space ahead of time.
1054 * Returns: zero on success, or a negative error code on failure.
1056 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1057 struct ice_rq_event_info *event)
1059 struct ice_aq_task *task;
1063 task = kzalloc(sizeof(*task), GFP_KERNEL);
1067 INIT_HLIST_NODE(&task->entry);
1068 task->opcode = opcode;
1069 task->event = event;
1070 task->state = ICE_AQ_TASK_WAITING;
1072 spin_lock_bh(&pf->aq_wait_lock);
1073 hlist_add_head(&task->entry, &pf->aq_wait_list);
1074 spin_unlock_bh(&pf->aq_wait_lock);
1076 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1078 switch (task->state) {
1079 case ICE_AQ_TASK_WAITING:
1080 err = ret < 0 ? ret : -ETIMEDOUT;
1082 case ICE_AQ_TASK_CANCELED:
1083 err = ret < 0 ? ret : -ECANCELED;
1085 case ICE_AQ_TASK_COMPLETE:
1086 err = ret < 0 ? ret : 0;
1089 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1094 spin_lock_bh(&pf->aq_wait_lock);
1095 hlist_del(&task->entry);
1096 spin_unlock_bh(&pf->aq_wait_lock);
1103 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1104 * @pf: pointer to the PF private structure
1105 * @opcode: the opcode of the event
1106 * @event: the event to check
1108 * Loops over the current list of pending threads waiting for an AdminQ event.
1109 * For each matching task, copy the contents of the event into the task
1110 * structure and wake up the thread.
1112 * If multiple threads wait for the same opcode, they will all be woken up.
1114 * Note that event->msg_buf will only be duplicated if the event has a buffer
1115 * with enough space already allocated. Otherwise, only the descriptor and
1116 * message length will be copied.
1118 * Returns: true if an event was found, false otherwise
1120 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1121 struct ice_rq_event_info *event)
1123 struct ice_aq_task *task;
1126 spin_lock_bh(&pf->aq_wait_lock);
1127 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1128 if (task->state || task->opcode != opcode)
1131 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1132 task->event->msg_len = event->msg_len;
1134 /* Only copy the data buffer if a destination was set */
1135 if (task->event->msg_buf &&
1136 task->event->buf_len > event->buf_len) {
1137 memcpy(task->event->msg_buf, event->msg_buf,
1139 task->event->buf_len = event->buf_len;
1142 task->state = ICE_AQ_TASK_COMPLETE;
1145 spin_unlock_bh(&pf->aq_wait_lock);
1148 wake_up(&pf->aq_wait_queue);
1152 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1153 * @pf: the PF private structure
1155 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1156 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1158 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1160 struct ice_aq_task *task;
1162 spin_lock_bh(&pf->aq_wait_lock);
1163 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1164 task->state = ICE_AQ_TASK_CANCELED;
1165 spin_unlock_bh(&pf->aq_wait_lock);
1167 wake_up(&pf->aq_wait_queue);
1171 * __ice_clean_ctrlq - helper function to clean controlq rings
1172 * @pf: ptr to struct ice_pf
1173 * @q_type: specific Control queue type
1175 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1177 struct device *dev = ice_pf_to_dev(pf);
1178 struct ice_rq_event_info event;
1179 struct ice_hw *hw = &pf->hw;
1180 struct ice_ctl_q_info *cq;
1185 /* Do not clean control queue if/when PF reset fails */
1186 if (test_bit(__ICE_RESET_FAILED, pf->state))
1190 case ICE_CTL_Q_ADMIN:
1194 case ICE_CTL_Q_MAILBOX:
1199 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1203 /* check for error indications - PF_xx_AxQLEN register layout for
1204 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1206 val = rd32(hw, cq->rq.len);
1207 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1208 PF_FW_ARQLEN_ARQCRIT_M)) {
1210 if (val & PF_FW_ARQLEN_ARQVFE_M)
1211 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1213 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1214 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1217 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1218 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1220 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1221 PF_FW_ARQLEN_ARQCRIT_M);
1223 wr32(hw, cq->rq.len, val);
1226 val = rd32(hw, cq->sq.len);
1227 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1228 PF_FW_ATQLEN_ATQCRIT_M)) {
1230 if (val & PF_FW_ATQLEN_ATQVFE_M)
1231 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1233 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1234 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1237 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1238 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1240 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1241 PF_FW_ATQLEN_ATQCRIT_M);
1243 wr32(hw, cq->sq.len, val);
1246 event.buf_len = cq->rq_buf_size;
1247 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1252 enum ice_status ret;
1255 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1256 if (ret == ICE_ERR_AQ_NO_WORK)
1259 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1264 opcode = le16_to_cpu(event.desc.opcode);
1266 /* Notify any thread that might be waiting for this event */
1267 ice_aq_check_events(pf, opcode, &event);
1270 case ice_aqc_opc_get_link_status:
1271 if (ice_handle_link_event(pf, &event))
1272 dev_err(dev, "Could not handle link event\n");
1274 case ice_aqc_opc_event_lan_overflow:
1275 ice_vf_lan_overflow_event(pf, &event);
1277 case ice_mbx_opc_send_msg_to_pf:
1278 ice_vc_process_vf_msg(pf, &event);
1280 case ice_aqc_opc_fw_logging:
1281 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1283 case ice_aqc_opc_lldp_set_mib_change:
1284 ice_dcb_process_lldp_set_mib_change(pf, &event);
1287 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1291 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1293 kfree(event.msg_buf);
1295 return pending && (i == ICE_DFLT_IRQ_WORK);
1299 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1300 * @hw: pointer to hardware info
1301 * @cq: control queue information
1303 * returns true if there are pending messages in a queue, false if there aren't
1305 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1309 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1310 return cq->rq.next_to_clean != ntu;
1314 * ice_clean_adminq_subtask - clean the AdminQ rings
1315 * @pf: board private structure
1317 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1319 struct ice_hw *hw = &pf->hw;
1321 if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1324 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1327 clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1329 /* There might be a situation where new messages arrive to a control
1330 * queue between processing the last message and clearing the
1331 * EVENT_PENDING bit. So before exiting, check queue head again (using
1332 * ice_ctrlq_pending) and process new messages if any.
1334 if (ice_ctrlq_pending(hw, &hw->adminq))
1335 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1341 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1342 * @pf: board private structure
1344 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1346 struct ice_hw *hw = &pf->hw;
1348 if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1351 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1354 clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1356 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1357 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1363 * ice_service_task_schedule - schedule the service task to wake up
1364 * @pf: board private structure
1366 * If not already scheduled, this puts the task into the work queue.
1368 void ice_service_task_schedule(struct ice_pf *pf)
1370 if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
1371 !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
1372 !test_bit(__ICE_NEEDS_RESTART, pf->state))
1373 queue_work(ice_wq, &pf->serv_task);
1377 * ice_service_task_complete - finish up the service task
1378 * @pf: board private structure
1380 static void ice_service_task_complete(struct ice_pf *pf)
1382 WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
1384 /* force memory (pf->state) to sync before next service task */
1385 smp_mb__before_atomic();
1386 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1390 * ice_service_task_stop - stop service task and cancel works
1391 * @pf: board private structure
1393 * Return 0 if the __ICE_SERVICE_DIS bit was not already set,
1396 static int ice_service_task_stop(struct ice_pf *pf)
1400 ret = test_and_set_bit(__ICE_SERVICE_DIS, pf->state);
1402 if (pf->serv_tmr.function)
1403 del_timer_sync(&pf->serv_tmr);
1404 if (pf->serv_task.func)
1405 cancel_work_sync(&pf->serv_task);
1407 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1412 * ice_service_task_restart - restart service task and schedule works
1413 * @pf: board private structure
1415 * This function is needed for suspend and resume works (e.g WoL scenario)
1417 static void ice_service_task_restart(struct ice_pf *pf)
1419 clear_bit(__ICE_SERVICE_DIS, pf->state);
1420 ice_service_task_schedule(pf);
1424 * ice_service_timer - timer callback to schedule service task
1425 * @t: pointer to timer_list
1427 static void ice_service_timer(struct timer_list *t)
1429 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1431 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1432 ice_service_task_schedule(pf);
1436 * ice_handle_mdd_event - handle malicious driver detect event
1437 * @pf: pointer to the PF structure
1439 * Called from service task. OICR interrupt handler indicates MDD event.
1440 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1441 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1442 * disable the queue, the PF can be configured to reset the VF using ethtool
1443 * private flag mdd-auto-reset-vf.
1445 static void ice_handle_mdd_event(struct ice_pf *pf)
1447 struct device *dev = ice_pf_to_dev(pf);
1448 struct ice_hw *hw = &pf->hw;
1452 if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state)) {
1453 /* Since the VF MDD event logging is rate limited, check if
1454 * there are pending MDD events.
1456 ice_print_vfs_mdd_events(pf);
1460 /* find what triggered an MDD event */
1461 reg = rd32(hw, GL_MDET_TX_PQM);
1462 if (reg & GL_MDET_TX_PQM_VALID_M) {
1463 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1464 GL_MDET_TX_PQM_PF_NUM_S;
1465 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1466 GL_MDET_TX_PQM_VF_NUM_S;
1467 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1468 GL_MDET_TX_PQM_MAL_TYPE_S;
1469 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1470 GL_MDET_TX_PQM_QNUM_S);
1472 if (netif_msg_tx_err(pf))
1473 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1474 event, queue, pf_num, vf_num);
1475 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1478 reg = rd32(hw, GL_MDET_TX_TCLAN);
1479 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1480 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1481 GL_MDET_TX_TCLAN_PF_NUM_S;
1482 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1483 GL_MDET_TX_TCLAN_VF_NUM_S;
1484 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1485 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1486 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1487 GL_MDET_TX_TCLAN_QNUM_S);
1489 if (netif_msg_tx_err(pf))
1490 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1491 event, queue, pf_num, vf_num);
1492 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1495 reg = rd32(hw, GL_MDET_RX);
1496 if (reg & GL_MDET_RX_VALID_M) {
1497 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1498 GL_MDET_RX_PF_NUM_S;
1499 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1500 GL_MDET_RX_VF_NUM_S;
1501 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1502 GL_MDET_RX_MAL_TYPE_S;
1503 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1506 if (netif_msg_rx_err(pf))
1507 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1508 event, queue, pf_num, vf_num);
1509 wr32(hw, GL_MDET_RX, 0xffffffff);
1512 /* check to see if this PF caused an MDD event */
1513 reg = rd32(hw, PF_MDET_TX_PQM);
1514 if (reg & PF_MDET_TX_PQM_VALID_M) {
1515 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1516 if (netif_msg_tx_err(pf))
1517 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1520 reg = rd32(hw, PF_MDET_TX_TCLAN);
1521 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1522 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1523 if (netif_msg_tx_err(pf))
1524 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1527 reg = rd32(hw, PF_MDET_RX);
1528 if (reg & PF_MDET_RX_VALID_M) {
1529 wr32(hw, PF_MDET_RX, 0xFFFF);
1530 if (netif_msg_rx_err(pf))
1531 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1534 /* Check to see if one of the VFs caused an MDD event, and then
1535 * increment counters and set print pending
1537 ice_for_each_vf(pf, i) {
1538 struct ice_vf *vf = &pf->vf[i];
1540 reg = rd32(hw, VP_MDET_TX_PQM(i));
1541 if (reg & VP_MDET_TX_PQM_VALID_M) {
1542 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1543 vf->mdd_tx_events.count++;
1544 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1545 if (netif_msg_tx_err(pf))
1546 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1550 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1551 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1552 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1553 vf->mdd_tx_events.count++;
1554 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1555 if (netif_msg_tx_err(pf))
1556 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1560 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1561 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1562 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1563 vf->mdd_tx_events.count++;
1564 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1565 if (netif_msg_tx_err(pf))
1566 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1570 reg = rd32(hw, VP_MDET_RX(i));
1571 if (reg & VP_MDET_RX_VALID_M) {
1572 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1573 vf->mdd_rx_events.count++;
1574 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1575 if (netif_msg_rx_err(pf))
1576 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1579 /* Since the queue is disabled on VF Rx MDD events, the
1580 * PF can be configured to reset the VF through ethtool
1581 * private flag mdd-auto-reset-vf.
1583 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1584 /* VF MDD event counters will be cleared by
1585 * reset, so print the event prior to reset.
1587 ice_print_vf_rx_mdd_event(vf);
1588 ice_reset_vf(&pf->vf[i], false);
1593 ice_print_vfs_mdd_events(pf);
1597 * ice_force_phys_link_state - Force the physical link state
1598 * @vsi: VSI to force the physical link state to up/down
1599 * @link_up: true/false indicates to set the physical link to up/down
1601 * Force the physical link state by getting the current PHY capabilities from
1602 * hardware and setting the PHY config based on the determined capabilities. If
1603 * link changes a link event will be triggered because both the Enable Automatic
1604 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1606 * Returns 0 on success, negative on failure
1608 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1610 struct ice_aqc_get_phy_caps_data *pcaps;
1611 struct ice_aqc_set_phy_cfg_data *cfg;
1612 struct ice_port_info *pi;
1616 if (!vsi || !vsi->port_info || !vsi->back)
1618 if (vsi->type != ICE_VSI_PF)
1621 dev = ice_pf_to_dev(vsi->back);
1623 pi = vsi->port_info;
1625 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1629 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
1632 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1633 vsi->vsi_num, retcode);
1638 /* No change in link */
1639 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1640 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1643 /* Use the current user PHY configuration. The current user PHY
1644 * configuration is initialized during probe from PHY capabilities
1645 * software mode, and updated on set PHY configuration.
1647 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1653 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1655 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1657 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1659 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1661 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1662 vsi->vsi_num, retcode);
1673 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1674 * @pi: port info structure
1676 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1678 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1680 struct ice_aqc_get_phy_caps_data *pcaps;
1681 struct ice_pf *pf = pi->hw->back;
1682 enum ice_status status;
1685 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1689 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_NVM_CAP, pcaps,
1693 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1698 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1699 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1707 * ice_init_link_dflt_override - Initialize link default override
1708 * @pi: port info structure
1710 * Initialize link default override and PHY total port shutdown during probe
1712 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1714 struct ice_link_default_override_tlv *ldo;
1715 struct ice_pf *pf = pi->hw->back;
1717 ldo = &pf->link_dflt_override;
1718 if (ice_get_link_default_override(ldo, pi))
1721 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1724 /* Enable Total Port Shutdown (override/replace link-down-on-close
1725 * ethtool private flag) for ports with Port Disable bit set.
1727 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1728 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1732 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1733 * @pi: port info structure
1735 * If default override is enabled, initialized the user PHY cfg speed and FEC
1736 * settings using the default override mask from the NVM.
1738 * The PHY should only be configured with the default override settings the
1739 * first time media is available. The __ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1740 * is used to indicate that the user PHY cfg default override is initialized
1741 * and the PHY has not been configured with the default override settings. The
1742 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1745 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1747 struct ice_link_default_override_tlv *ldo;
1748 struct ice_aqc_set_phy_cfg_data *cfg;
1749 struct ice_phy_info *phy = &pi->phy;
1750 struct ice_pf *pf = pi->hw->back;
1752 ldo = &pf->link_dflt_override;
1754 /* If link default override is enabled, use to mask NVM PHY capabilities
1755 * for speed and FEC default configuration.
1757 cfg = &phy->curr_user_phy_cfg;
1759 if (ldo->phy_type_low || ldo->phy_type_high) {
1760 cfg->phy_type_low = pf->nvm_phy_type_lo &
1761 cpu_to_le64(ldo->phy_type_low);
1762 cfg->phy_type_high = pf->nvm_phy_type_hi &
1763 cpu_to_le64(ldo->phy_type_high);
1765 cfg->link_fec_opt = ldo->fec_options;
1766 phy->curr_user_fec_req = ICE_FEC_AUTO;
1768 set_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1772 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1773 * @pi: port info structure
1775 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1776 * mode to default. The PHY defaults are from get PHY capabilities topology
1777 * with media so call when media is first available. An error is returned if
1778 * called when media is not available. The PHY initialization completed state is
1781 * These configurations are used when setting PHY
1782 * configuration. The user PHY configuration is updated on set PHY
1783 * configuration. Returns 0 on success, negative on failure
1785 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1787 struct ice_aqc_get_phy_caps_data *pcaps;
1788 struct ice_phy_info *phy = &pi->phy;
1789 struct ice_pf *pf = pi->hw->back;
1790 enum ice_status status;
1791 struct ice_vsi *vsi;
1794 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1797 vsi = ice_get_main_vsi(pf);
1801 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1805 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1808 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1813 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1815 /* check if lenient mode is supported and enabled */
1816 if (ice_fw_supports_link_override(&vsi->back->hw) &&
1817 !(pcaps->module_compliance_enforcement &
1818 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1819 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1821 /* if link default override is enabled, initialize user PHY
1822 * configuration with link default override values
1824 if (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN) {
1825 ice_init_phy_cfg_dflt_override(pi);
1830 /* if link default override is not enabled, initialize PHY using
1831 * topology with media
1833 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1834 pcaps->link_fec_options);
1835 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1838 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1839 set_bit(__ICE_PHY_INIT_COMPLETE, pf->state);
1846 * ice_configure_phy - configure PHY
1849 * Set the PHY configuration. If the current PHY configuration is the same as
1850 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1851 * configure the based get PHY capabilities for topology with media.
1853 static int ice_configure_phy(struct ice_vsi *vsi)
1855 struct device *dev = ice_pf_to_dev(vsi->back);
1856 struct ice_aqc_get_phy_caps_data *pcaps;
1857 struct ice_aqc_set_phy_cfg_data *cfg;
1858 struct ice_port_info *pi;
1859 enum ice_status status;
1862 pi = vsi->port_info;
1866 /* Ensure we have media as we cannot configure a medialess port */
1867 if (!(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1870 ice_print_topo_conflict(vsi);
1872 if (vsi->port_info->phy.link_info.topo_media_conflict ==
1873 ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1876 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->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_SW_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, &pi->phy.curr_user_phy_cfg))
1900 /* Use PHY topology as baseline for configuration */
1901 memset(pcaps, 0, sizeof(*pcaps));
1902 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1905 dev_err(dev, "Failed to get PHY topology, VSI %d error %s\n",
1906 vsi->vsi_num, ice_stat_str(status));
1911 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1917 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
1919 /* Speed - If default override pending, use curr_user_phy_cfg set in
1920 * ice_init_phy_user_cfg_ldo.
1922 if (test_and_clear_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING,
1923 vsi->back->state)) {
1924 cfg->phy_type_low = pi->phy.curr_user_phy_cfg.phy_type_low;
1925 cfg->phy_type_high = pi->phy.curr_user_phy_cfg.phy_type_high;
1927 u64 phy_low = 0, phy_high = 0;
1929 ice_update_phy_type(&phy_low, &phy_high,
1930 pi->phy.curr_user_speed_req);
1931 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
1932 cfg->phy_type_high = pcaps->phy_type_high &
1933 cpu_to_le64(phy_high);
1936 /* Can't provide what was requested; use PHY capabilities */
1937 if (!cfg->phy_type_low && !cfg->phy_type_high) {
1938 cfg->phy_type_low = pcaps->phy_type_low;
1939 cfg->phy_type_high = pcaps->phy_type_high;
1943 ice_cfg_phy_fec(pi, cfg, pi->phy.curr_user_fec_req);
1945 /* Can't provide what was requested; use PHY capabilities */
1946 if (cfg->link_fec_opt !=
1947 (cfg->link_fec_opt & pcaps->link_fec_options)) {
1948 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
1949 cfg->link_fec_opt = pcaps->link_fec_options;
1952 /* Flow Control - always supported; no need to check against
1955 ice_cfg_phy_fc(pi, cfg, pi->phy.curr_user_fc_req);
1957 /* Enable link and link update */
1958 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
1960 status = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1962 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
1963 vsi->vsi_num, ice_stat_str(status));
1974 * ice_check_media_subtask - Check for media
1975 * @pf: pointer to PF struct
1977 * If media is available, then initialize PHY user configuration if it is not
1978 * been, and configure the PHY if the interface is up.
1980 static void ice_check_media_subtask(struct ice_pf *pf)
1982 struct ice_port_info *pi;
1983 struct ice_vsi *vsi;
1986 /* No need to check for media if it's already present */
1987 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
1990 vsi = ice_get_main_vsi(pf);
1994 /* Refresh link info and check if media is present */
1995 pi = vsi->port_info;
1996 err = ice_update_link_info(pi);
2000 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2001 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state))
2002 ice_init_phy_user_cfg(pi);
2004 /* PHY settings are reset on media insertion, reconfigure
2005 * PHY to preserve settings.
2007 if (test_bit(__ICE_DOWN, vsi->state) &&
2008 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2011 err = ice_configure_phy(vsi);
2013 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2015 /* A Link Status Event will be generated; the event handler
2016 * will complete bringing the interface up
2022 * ice_service_task - manage and run subtasks
2023 * @work: pointer to work_struct contained by the PF struct
2025 static void ice_service_task(struct work_struct *work)
2027 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2028 unsigned long start_time = jiffies;
2032 /* process reset requests first */
2033 ice_reset_subtask(pf);
2035 /* bail if a reset/recovery cycle is pending or rebuild failed */
2036 if (ice_is_reset_in_progress(pf->state) ||
2037 test_bit(__ICE_SUSPENDED, pf->state) ||
2038 test_bit(__ICE_NEEDS_RESTART, pf->state)) {
2039 ice_service_task_complete(pf);
2043 ice_clean_adminq_subtask(pf);
2044 ice_check_media_subtask(pf);
2045 ice_check_for_hang_subtask(pf);
2046 ice_sync_fltr_subtask(pf);
2047 ice_handle_mdd_event(pf);
2048 ice_watchdog_subtask(pf);
2050 if (ice_is_safe_mode(pf)) {
2051 ice_service_task_complete(pf);
2055 ice_process_vflr_event(pf);
2056 ice_clean_mailboxq_subtask(pf);
2057 ice_sync_arfs_fltrs(pf);
2058 /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
2059 ice_service_task_complete(pf);
2061 /* If the tasks have taken longer than one service timer period
2062 * or there is more work to be done, reset the service timer to
2063 * schedule the service task now.
2065 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2066 test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
2067 test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
2068 test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2069 test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
2070 mod_timer(&pf->serv_tmr, jiffies);
2074 * ice_set_ctrlq_len - helper function to set controlq length
2075 * @hw: pointer to the HW instance
2077 static void ice_set_ctrlq_len(struct ice_hw *hw)
2079 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2080 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2081 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2082 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2083 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2084 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2085 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2086 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2090 * ice_schedule_reset - schedule a reset
2091 * @pf: board private structure
2092 * @reset: reset being requested
2094 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2096 struct device *dev = ice_pf_to_dev(pf);
2098 /* bail out if earlier reset has failed */
2099 if (test_bit(__ICE_RESET_FAILED, pf->state)) {
2100 dev_dbg(dev, "earlier reset has failed\n");
2103 /* bail if reset/recovery already in progress */
2104 if (ice_is_reset_in_progress(pf->state)) {
2105 dev_dbg(dev, "Reset already in progress\n");
2111 set_bit(__ICE_PFR_REQ, pf->state);
2113 case ICE_RESET_CORER:
2114 set_bit(__ICE_CORER_REQ, pf->state);
2116 case ICE_RESET_GLOBR:
2117 set_bit(__ICE_GLOBR_REQ, pf->state);
2123 ice_service_task_schedule(pf);
2128 * ice_irq_affinity_notify - Callback for affinity changes
2129 * @notify: context as to what irq was changed
2130 * @mask: the new affinity mask
2132 * This is a callback function used by the irq_set_affinity_notifier function
2133 * so that we may register to receive changes to the irq affinity masks.
2136 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2137 const cpumask_t *mask)
2139 struct ice_q_vector *q_vector =
2140 container_of(notify, struct ice_q_vector, affinity_notify);
2142 cpumask_copy(&q_vector->affinity_mask, mask);
2146 * ice_irq_affinity_release - Callback for affinity notifier release
2147 * @ref: internal core kernel usage
2149 * This is a callback function used by the irq_set_affinity_notifier function
2150 * to inform the current notification subscriber that they will no longer
2151 * receive notifications.
2153 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2156 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2157 * @vsi: the VSI being configured
2159 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2161 struct ice_hw *hw = &vsi->back->hw;
2164 ice_for_each_q_vector(vsi, i)
2165 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2172 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2173 * @vsi: the VSI being configured
2174 * @basename: name for the vector
2176 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2178 int q_vectors = vsi->num_q_vectors;
2179 struct ice_pf *pf = vsi->back;
2180 int base = vsi->base_vector;
2187 dev = ice_pf_to_dev(pf);
2188 for (vector = 0; vector < q_vectors; vector++) {
2189 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2191 irq_num = pf->msix_entries[base + vector].vector;
2193 if (q_vector->tx.ring && q_vector->rx.ring) {
2194 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2195 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2197 } else if (q_vector->rx.ring) {
2198 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2199 "%s-%s-%d", basename, "rx", rx_int_idx++);
2200 } else if (q_vector->tx.ring) {
2201 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2202 "%s-%s-%d", basename, "tx", tx_int_idx++);
2204 /* skip this unused q_vector */
2207 err = devm_request_irq(dev, irq_num, vsi->irq_handler, 0,
2208 q_vector->name, q_vector);
2210 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2215 /* register for affinity change notifications */
2216 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2217 struct irq_affinity_notify *affinity_notify;
2219 affinity_notify = &q_vector->affinity_notify;
2220 affinity_notify->notify = ice_irq_affinity_notify;
2221 affinity_notify->release = ice_irq_affinity_release;
2222 irq_set_affinity_notifier(irq_num, affinity_notify);
2225 /* assign the mask for this irq */
2226 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2229 vsi->irqs_ready = true;
2235 irq_num = pf->msix_entries[base + vector].vector;
2236 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2237 irq_set_affinity_notifier(irq_num, NULL);
2238 irq_set_affinity_hint(irq_num, NULL);
2239 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2245 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2246 * @vsi: VSI to setup Tx rings used by XDP
2248 * Return 0 on success and negative value on error
2250 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2252 struct device *dev = ice_pf_to_dev(vsi->back);
2255 for (i = 0; i < vsi->num_xdp_txq; i++) {
2256 u16 xdp_q_idx = vsi->alloc_txq + i;
2257 struct ice_ring *xdp_ring;
2259 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2262 goto free_xdp_rings;
2264 xdp_ring->q_index = xdp_q_idx;
2265 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2266 xdp_ring->ring_active = false;
2267 xdp_ring->vsi = vsi;
2268 xdp_ring->netdev = NULL;
2269 xdp_ring->dev = dev;
2270 xdp_ring->count = vsi->num_tx_desc;
2271 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2272 if (ice_setup_tx_ring(xdp_ring))
2273 goto free_xdp_rings;
2274 ice_set_ring_xdp(xdp_ring);
2275 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2282 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2283 ice_free_tx_ring(vsi->xdp_rings[i]);
2288 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2289 * @vsi: VSI to set the bpf prog on
2290 * @prog: the bpf prog pointer
2292 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2294 struct bpf_prog *old_prog;
2297 old_prog = xchg(&vsi->xdp_prog, prog);
2299 bpf_prog_put(old_prog);
2301 ice_for_each_rxq(vsi, i)
2302 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2306 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2307 * @vsi: VSI to bring up Tx rings used by XDP
2308 * @prog: bpf program that will be assigned to VSI
2310 * Return 0 on success and negative value on error
2312 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2314 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2315 int xdp_rings_rem = vsi->num_xdp_txq;
2316 struct ice_pf *pf = vsi->back;
2317 struct ice_qs_cfg xdp_qs_cfg = {
2318 .qs_mutex = &pf->avail_q_mutex,
2319 .pf_map = pf->avail_txqs,
2320 .pf_map_size = pf->max_pf_txqs,
2321 .q_count = vsi->num_xdp_txq,
2322 .scatter_count = ICE_MAX_SCATTER_TXQS,
2323 .vsi_map = vsi->txq_map,
2324 .vsi_map_offset = vsi->alloc_txq,
2325 .mapping_mode = ICE_VSI_MAP_CONTIG
2327 enum ice_status status;
2331 dev = ice_pf_to_dev(pf);
2332 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2333 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2334 if (!vsi->xdp_rings)
2337 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2338 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2341 if (ice_xdp_alloc_setup_rings(vsi))
2342 goto clear_xdp_rings;
2344 /* follow the logic from ice_vsi_map_rings_to_vectors */
2345 ice_for_each_q_vector(vsi, v_idx) {
2346 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2347 int xdp_rings_per_v, q_id, q_base;
2349 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2350 vsi->num_q_vectors - v_idx);
2351 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2353 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2354 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2356 xdp_ring->q_vector = q_vector;
2357 xdp_ring->next = q_vector->tx.ring;
2358 q_vector->tx.ring = xdp_ring;
2360 xdp_rings_rem -= xdp_rings_per_v;
2363 /* omit the scheduler update if in reset path; XDP queues will be
2364 * taken into account at the end of ice_vsi_rebuild, where
2365 * ice_cfg_vsi_lan is being called
2367 if (ice_is_reset_in_progress(pf->state))
2370 /* tell the Tx scheduler that right now we have
2373 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2374 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2376 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2379 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2380 ice_stat_str(status));
2381 goto clear_xdp_rings;
2383 ice_vsi_assign_bpf_prog(vsi, prog);
2387 for (i = 0; i < vsi->num_xdp_txq; i++)
2388 if (vsi->xdp_rings[i]) {
2389 kfree_rcu(vsi->xdp_rings[i], rcu);
2390 vsi->xdp_rings[i] = NULL;
2394 mutex_lock(&pf->avail_q_mutex);
2395 for (i = 0; i < vsi->num_xdp_txq; i++) {
2396 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2397 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2399 mutex_unlock(&pf->avail_q_mutex);
2401 devm_kfree(dev, vsi->xdp_rings);
2406 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2407 * @vsi: VSI to remove XDP rings
2409 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2412 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2414 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2415 struct ice_pf *pf = vsi->back;
2418 /* q_vectors are freed in reset path so there's no point in detaching
2419 * rings; in case of rebuild being triggered not from reset reset bits
2420 * in pf->state won't be set, so additionally check first q_vector
2423 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2426 ice_for_each_q_vector(vsi, v_idx) {
2427 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2428 struct ice_ring *ring;
2430 ice_for_each_ring(ring, q_vector->tx)
2431 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2434 /* restore the value of last node prior to XDP setup */
2435 q_vector->tx.ring = ring;
2439 mutex_lock(&pf->avail_q_mutex);
2440 for (i = 0; i < vsi->num_xdp_txq; i++) {
2441 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2442 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2444 mutex_unlock(&pf->avail_q_mutex);
2446 for (i = 0; i < vsi->num_xdp_txq; i++)
2447 if (vsi->xdp_rings[i]) {
2448 if (vsi->xdp_rings[i]->desc)
2449 ice_free_tx_ring(vsi->xdp_rings[i]);
2450 kfree_rcu(vsi->xdp_rings[i], rcu);
2451 vsi->xdp_rings[i] = NULL;
2454 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2455 vsi->xdp_rings = NULL;
2457 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2460 ice_vsi_assign_bpf_prog(vsi, NULL);
2462 /* notify Tx scheduler that we destroyed XDP queues and bring
2463 * back the old number of child nodes
2465 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2466 max_txqs[i] = vsi->num_txq;
2468 /* change number of XDP Tx queues to 0 */
2469 vsi->num_xdp_txq = 0;
2471 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2476 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2477 * @vsi: VSI to setup XDP for
2478 * @prog: XDP program
2479 * @extack: netlink extended ack
2482 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2483 struct netlink_ext_ack *extack)
2485 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2486 bool if_running = netif_running(vsi->netdev);
2487 int ret = 0, xdp_ring_err = 0;
2489 if (frame_size > vsi->rx_buf_len) {
2490 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2494 /* need to stop netdev while setting up the program for Rx rings */
2495 if (if_running && !test_and_set_bit(__ICE_DOWN, vsi->state)) {
2496 ret = ice_down(vsi);
2498 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2503 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2504 vsi->num_xdp_txq = vsi->alloc_rxq;
2505 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2507 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2508 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2509 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2511 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2513 ice_vsi_assign_bpf_prog(vsi, prog);
2519 if (!ret && prog && vsi->xsk_pools) {
2522 ice_for_each_rxq(vsi, i) {
2523 struct ice_ring *rx_ring = vsi->rx_rings[i];
2525 if (rx_ring->xsk_pool)
2526 napi_schedule(&rx_ring->q_vector->napi);
2530 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2534 * ice_xdp - implements XDP handler
2538 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2540 struct ice_netdev_priv *np = netdev_priv(dev);
2541 struct ice_vsi *vsi = np->vsi;
2543 if (vsi->type != ICE_VSI_PF) {
2544 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2548 switch (xdp->command) {
2549 case XDP_SETUP_PROG:
2550 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2551 case XDP_SETUP_XSK_POOL:
2552 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2560 * ice_ena_misc_vector - enable the non-queue interrupts
2561 * @pf: board private structure
2563 static void ice_ena_misc_vector(struct ice_pf *pf)
2565 struct ice_hw *hw = &pf->hw;
2568 /* Disable anti-spoof detection interrupt to prevent spurious event
2569 * interrupts during a function reset. Anti-spoof functionally is
2572 val = rd32(hw, GL_MDCK_TX_TDPU);
2573 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2574 wr32(hw, GL_MDCK_TX_TDPU, val);
2576 /* clear things first */
2577 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2578 rd32(hw, PFINT_OICR); /* read to clear */
2580 val = (PFINT_OICR_ECC_ERR_M |
2581 PFINT_OICR_MAL_DETECT_M |
2583 PFINT_OICR_PCI_EXCEPTION_M |
2585 PFINT_OICR_HMC_ERR_M |
2586 PFINT_OICR_PE_CRITERR_M);
2588 wr32(hw, PFINT_OICR_ENA, val);
2590 /* SW_ITR_IDX = 0, but don't change INTENA */
2591 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2592 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2596 * ice_misc_intr - misc interrupt handler
2597 * @irq: interrupt number
2598 * @data: pointer to a q_vector
2600 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2602 struct ice_pf *pf = (struct ice_pf *)data;
2603 struct ice_hw *hw = &pf->hw;
2604 irqreturn_t ret = IRQ_NONE;
2608 dev = ice_pf_to_dev(pf);
2609 set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
2610 set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2612 oicr = rd32(hw, PFINT_OICR);
2613 ena_mask = rd32(hw, PFINT_OICR_ENA);
2615 if (oicr & PFINT_OICR_SWINT_M) {
2616 ena_mask &= ~PFINT_OICR_SWINT_M;
2620 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2621 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2622 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
2624 if (oicr & PFINT_OICR_VFLR_M) {
2625 /* disable any further VFLR event notifications */
2626 if (test_bit(__ICE_VF_RESETS_DISABLED, pf->state)) {
2627 u32 reg = rd32(hw, PFINT_OICR_ENA);
2629 reg &= ~PFINT_OICR_VFLR_M;
2630 wr32(hw, PFINT_OICR_ENA, reg);
2632 ena_mask &= ~PFINT_OICR_VFLR_M;
2633 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
2637 if (oicr & PFINT_OICR_GRST_M) {
2640 /* we have a reset warning */
2641 ena_mask &= ~PFINT_OICR_GRST_M;
2642 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2643 GLGEN_RSTAT_RESET_TYPE_S;
2645 if (reset == ICE_RESET_CORER)
2647 else if (reset == ICE_RESET_GLOBR)
2649 else if (reset == ICE_RESET_EMPR)
2652 dev_dbg(dev, "Invalid reset type %d\n", reset);
2654 /* If a reset cycle isn't already in progress, we set a bit in
2655 * pf->state so that the service task can start a reset/rebuild.
2656 * We also make note of which reset happened so that peer
2657 * devices/drivers can be informed.
2659 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
2660 if (reset == ICE_RESET_CORER)
2661 set_bit(__ICE_CORER_RECV, pf->state);
2662 else if (reset == ICE_RESET_GLOBR)
2663 set_bit(__ICE_GLOBR_RECV, pf->state);
2665 set_bit(__ICE_EMPR_RECV, pf->state);
2667 /* There are couple of different bits at play here.
2668 * hw->reset_ongoing indicates whether the hardware is
2669 * in reset. This is set to true when a reset interrupt
2670 * is received and set back to false after the driver
2671 * has determined that the hardware is out of reset.
2673 * __ICE_RESET_OICR_RECV in pf->state indicates
2674 * that a post reset rebuild is required before the
2675 * driver is operational again. This is set above.
2677 * As this is the start of the reset/rebuild cycle, set
2678 * both to indicate that.
2680 hw->reset_ongoing = true;
2684 if (oicr & PFINT_OICR_HMC_ERR_M) {
2685 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
2686 dev_dbg(dev, "HMC Error interrupt - info 0x%x, data 0x%x\n",
2687 rd32(hw, PFHMC_ERRORINFO),
2688 rd32(hw, PFHMC_ERRORDATA));
2691 /* Report any remaining unexpected interrupts */
2694 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2695 /* If a critical error is pending there is no choice but to
2698 if (oicr & (PFINT_OICR_PE_CRITERR_M |
2699 PFINT_OICR_PCI_EXCEPTION_M |
2700 PFINT_OICR_ECC_ERR_M)) {
2701 set_bit(__ICE_PFR_REQ, pf->state);
2702 ice_service_task_schedule(pf);
2707 ice_service_task_schedule(pf);
2708 ice_irq_dynamic_ena(hw, NULL, NULL);
2714 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2715 * @hw: pointer to HW structure
2717 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2719 /* disable Admin queue Interrupt causes */
2720 wr32(hw, PFINT_FW_CTL,
2721 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2723 /* disable Mailbox queue Interrupt causes */
2724 wr32(hw, PFINT_MBX_CTL,
2725 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2727 /* disable Control queue Interrupt causes */
2728 wr32(hw, PFINT_OICR_CTL,
2729 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2735 * ice_free_irq_msix_misc - Unroll misc vector setup
2736 * @pf: board private structure
2738 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2740 struct ice_hw *hw = &pf->hw;
2742 ice_dis_ctrlq_interrupts(hw);
2744 /* disable OICR interrupt */
2745 wr32(hw, PFINT_OICR_ENA, 0);
2748 if (pf->msix_entries) {
2749 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2750 devm_free_irq(ice_pf_to_dev(pf),
2751 pf->msix_entries[pf->oicr_idx].vector, pf);
2754 pf->num_avail_sw_msix += 1;
2755 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2759 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2760 * @hw: pointer to HW structure
2761 * @reg_idx: HW vector index to associate the control queue interrupts with
2763 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2767 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2768 PFINT_OICR_CTL_CAUSE_ENA_M);
2769 wr32(hw, PFINT_OICR_CTL, val);
2771 /* enable Admin queue Interrupt causes */
2772 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2773 PFINT_FW_CTL_CAUSE_ENA_M);
2774 wr32(hw, PFINT_FW_CTL, val);
2776 /* enable Mailbox queue Interrupt causes */
2777 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2778 PFINT_MBX_CTL_CAUSE_ENA_M);
2779 wr32(hw, PFINT_MBX_CTL, val);
2785 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2786 * @pf: board private structure
2788 * This sets up the handler for MSIX 0, which is used to manage the
2789 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2790 * when in MSI or Legacy interrupt mode.
2792 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2794 struct device *dev = ice_pf_to_dev(pf);
2795 struct ice_hw *hw = &pf->hw;
2796 int oicr_idx, err = 0;
2798 if (!pf->int_name[0])
2799 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2800 dev_driver_string(dev), dev_name(dev));
2802 /* Do not request IRQ but do enable OICR interrupt since settings are
2803 * lost during reset. Note that this function is called only during
2804 * rebuild path and not while reset is in progress.
2806 if (ice_is_reset_in_progress(pf->state))
2809 /* reserve one vector in irq_tracker for misc interrupts */
2810 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2814 pf->num_avail_sw_msix -= 1;
2815 pf->oicr_idx = (u16)oicr_idx;
2817 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2818 ice_misc_intr, 0, pf->int_name, pf);
2820 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2822 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2823 pf->num_avail_sw_msix += 1;
2828 ice_ena_misc_vector(pf);
2830 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2831 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2832 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2835 ice_irq_dynamic_ena(hw, NULL, NULL);
2841 * ice_napi_add - register NAPI handler for the VSI
2842 * @vsi: VSI for which NAPI handler is to be registered
2844 * This function is only called in the driver's load path. Registering the NAPI
2845 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2846 * reset/rebuild, etc.)
2848 static void ice_napi_add(struct ice_vsi *vsi)
2855 ice_for_each_q_vector(vsi, v_idx)
2856 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2857 ice_napi_poll, NAPI_POLL_WEIGHT);
2861 * ice_set_ops - set netdev and ethtools ops for the given netdev
2862 * @netdev: netdev instance
2864 static void ice_set_ops(struct net_device *netdev)
2866 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2868 if (ice_is_safe_mode(pf)) {
2869 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2870 ice_set_ethtool_safe_mode_ops(netdev);
2874 netdev->netdev_ops = &ice_netdev_ops;
2875 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
2876 ice_set_ethtool_ops(netdev);
2880 * ice_set_netdev_features - set features for the given netdev
2881 * @netdev: netdev instance
2883 static void ice_set_netdev_features(struct net_device *netdev)
2885 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2886 netdev_features_t csumo_features;
2887 netdev_features_t vlano_features;
2888 netdev_features_t dflt_features;
2889 netdev_features_t tso_features;
2891 if (ice_is_safe_mode(pf)) {
2893 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2894 netdev->hw_features = netdev->features;
2898 dflt_features = NETIF_F_SG |
2903 csumo_features = NETIF_F_RXCSUM |
2908 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2909 NETIF_F_HW_VLAN_CTAG_TX |
2910 NETIF_F_HW_VLAN_CTAG_RX;
2912 tso_features = NETIF_F_TSO |
2916 NETIF_F_GSO_UDP_TUNNEL |
2917 NETIF_F_GSO_GRE_CSUM |
2918 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2919 NETIF_F_GSO_PARTIAL |
2920 NETIF_F_GSO_IPXIP4 |
2921 NETIF_F_GSO_IPXIP6 |
2924 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2925 NETIF_F_GSO_GRE_CSUM;
2926 /* set features that user can change */
2927 netdev->hw_features = dflt_features | csumo_features |
2928 vlano_features | tso_features;
2930 /* add support for HW_CSUM on packets with MPLS header */
2931 netdev->mpls_features = NETIF_F_HW_CSUM;
2933 /* enable features */
2934 netdev->features |= netdev->hw_features;
2935 /* encap and VLAN devices inherit default, csumo and tso features */
2936 netdev->hw_enc_features |= dflt_features | csumo_features |
2938 netdev->vlan_features |= dflt_features | csumo_features |
2943 * ice_cfg_netdev - Allocate, configure and register a netdev
2944 * @vsi: the VSI associated with the new netdev
2946 * Returns 0 on success, negative value on failure
2948 static int ice_cfg_netdev(struct ice_vsi *vsi)
2950 struct ice_pf *pf = vsi->back;
2951 struct ice_netdev_priv *np;
2952 struct net_device *netdev;
2953 u8 mac_addr[ETH_ALEN];
2956 err = ice_devlink_create_port(pf);
2960 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
2964 goto err_destroy_devlink_port;
2967 vsi->netdev = netdev;
2968 np = netdev_priv(netdev);
2971 ice_set_netdev_features(netdev);
2973 ice_set_ops(netdev);
2975 if (vsi->type == ICE_VSI_PF) {
2976 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
2977 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
2978 ether_addr_copy(netdev->dev_addr, mac_addr);
2979 ether_addr_copy(netdev->perm_addr, mac_addr);
2982 netdev->priv_flags |= IFF_UNICAST_FLT;
2984 /* Setup netdev TC information */
2985 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
2987 /* setup watchdog timeout value to be 5 second */
2988 netdev->watchdog_timeo = 5 * HZ;
2990 netdev->min_mtu = ETH_MIN_MTU;
2991 netdev->max_mtu = ICE_MAX_MTU;
2993 err = register_netdev(vsi->netdev);
2995 goto err_free_netdev;
2997 devlink_port_type_eth_set(&pf->devlink_port, vsi->netdev);
2999 netif_carrier_off(vsi->netdev);
3001 /* make sure transmit queues start off as stopped */
3002 netif_tx_stop_all_queues(vsi->netdev);
3007 free_netdev(vsi->netdev);
3009 err_destroy_devlink_port:
3010 ice_devlink_destroy_port(pf);
3015 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3016 * @lut: Lookup table
3017 * @rss_table_size: Lookup table size
3018 * @rss_size: Range of queue number for hashing
3020 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3024 for (i = 0; i < rss_table_size; i++)
3025 lut[i] = i % rss_size;
3029 * ice_pf_vsi_setup - Set up a PF VSI
3030 * @pf: board private structure
3031 * @pi: pointer to the port_info instance
3033 * Returns pointer to the successfully allocated VSI software struct
3034 * on success, otherwise returns NULL on failure.
3036 static struct ice_vsi *
3037 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3039 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3043 * ice_ctrl_vsi_setup - Set up a control VSI
3044 * @pf: board private structure
3045 * @pi: pointer to the port_info instance
3047 * Returns pointer to the successfully allocated VSI software struct
3048 * on success, otherwise returns NULL on failure.
3050 static struct ice_vsi *
3051 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3053 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3057 * ice_lb_vsi_setup - Set up a loopback VSI
3058 * @pf: board private structure
3059 * @pi: pointer to the port_info instance
3061 * Returns pointer to the successfully allocated VSI software struct
3062 * on success, otherwise returns NULL on failure.
3065 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3067 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3071 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3072 * @netdev: network interface to be adjusted
3073 * @proto: unused protocol
3074 * @vid: VLAN ID to be added
3076 * net_device_ops implementation for adding VLAN IDs
3079 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3082 struct ice_netdev_priv *np = netdev_priv(netdev);
3083 struct ice_vsi *vsi = np->vsi;
3086 if (vid >= VLAN_N_VID) {
3087 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
3095 /* VLAN 0 is added by default during load/reset */
3099 /* Enable VLAN pruning when a VLAN other than 0 is added */
3100 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3101 ret = ice_cfg_vlan_pruning(vsi, true, false);
3106 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3107 * packets aren't pruned by the device's internal switch on Rx
3109 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3111 vsi->vlan_ena = true;
3112 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3119 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3120 * @netdev: network interface to be adjusted
3121 * @proto: unused protocol
3122 * @vid: VLAN ID to be removed
3124 * net_device_ops implementation for removing VLAN IDs
3127 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3130 struct ice_netdev_priv *np = netdev_priv(netdev);
3131 struct ice_vsi *vsi = np->vsi;
3137 /* don't allow removal of VLAN 0 */
3141 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3144 ret = ice_vsi_kill_vlan(vsi, vid);
3148 /* Disable pruning when VLAN 0 is the only VLAN rule */
3149 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3150 ret = ice_cfg_vlan_pruning(vsi, false, false);
3152 vsi->vlan_ena = false;
3153 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3158 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3159 * @pf: board private structure
3161 * Returns 0 on success, negative value on failure
3163 static int ice_setup_pf_sw(struct ice_pf *pf)
3165 struct ice_vsi *vsi;
3168 if (ice_is_reset_in_progress(pf->state))
3171 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3175 status = ice_cfg_netdev(vsi);
3178 goto unroll_vsi_setup;
3180 /* netdev has to be configured before setting frame size */
3181 ice_vsi_cfg_frame_size(vsi);
3183 /* Setup DCB netlink interface */
3184 ice_dcbnl_setup(vsi);
3186 /* registering the NAPI handler requires both the queues and
3187 * netdev to be created, which are done in ice_pf_vsi_setup()
3188 * and ice_cfg_netdev() respectively
3192 status = ice_set_cpu_rx_rmap(vsi);
3194 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3195 vsi->vsi_num, status);
3197 goto unroll_napi_add;
3199 status = ice_init_mac_fltr(pf);
3201 goto free_cpu_rx_map;
3206 ice_free_cpu_rx_rmap(vsi);
3212 if (vsi->netdev->reg_state == NETREG_REGISTERED)
3213 unregister_netdev(vsi->netdev);
3214 free_netdev(vsi->netdev);
3220 ice_vsi_release(vsi);
3225 * ice_get_avail_q_count - Get count of queues in use
3226 * @pf_qmap: bitmap to get queue use count from
3227 * @lock: pointer to a mutex that protects access to pf_qmap
3228 * @size: size of the bitmap
3231 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3237 for_each_clear_bit(bit, pf_qmap, size)
3245 * ice_get_avail_txq_count - Get count of Tx queues in use
3246 * @pf: pointer to an ice_pf instance
3248 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3250 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3255 * ice_get_avail_rxq_count - Get count of Rx queues in use
3256 * @pf: pointer to an ice_pf instance
3258 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3260 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3265 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3266 * @pf: board private structure to initialize
3268 static void ice_deinit_pf(struct ice_pf *pf)
3270 ice_service_task_stop(pf);
3271 mutex_destroy(&pf->sw_mutex);
3272 mutex_destroy(&pf->tc_mutex);
3273 mutex_destroy(&pf->avail_q_mutex);
3275 if (pf->avail_txqs) {
3276 bitmap_free(pf->avail_txqs);
3277 pf->avail_txqs = NULL;
3280 if (pf->avail_rxqs) {
3281 bitmap_free(pf->avail_rxqs);
3282 pf->avail_rxqs = NULL;
3287 * ice_set_pf_caps - set PFs capability flags
3288 * @pf: pointer to the PF instance
3290 static void ice_set_pf_caps(struct ice_pf *pf)
3292 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3294 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3295 if (func_caps->common_cap.dcb)
3296 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3297 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3298 if (func_caps->common_cap.sr_iov_1_1) {
3299 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3300 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3303 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3304 if (func_caps->common_cap.rss_table_size)
3305 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3307 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3308 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3311 /* ctrl_vsi_idx will be set to a valid value when flow director
3312 * is setup by ice_init_fdir
3314 pf->ctrl_vsi_idx = ICE_NO_VSI;
3315 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3316 /* force guaranteed filter pool for PF */
3317 ice_alloc_fd_guar_item(&pf->hw, &unused,
3318 func_caps->fd_fltr_guar);
3319 /* force shared filter pool for PF */
3320 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3321 func_caps->fd_fltr_best_effort);
3324 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3325 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3329 * ice_init_pf - Initialize general software structures (struct ice_pf)
3330 * @pf: board private structure to initialize
3332 static int ice_init_pf(struct ice_pf *pf)
3334 ice_set_pf_caps(pf);
3336 mutex_init(&pf->sw_mutex);
3337 mutex_init(&pf->tc_mutex);
3339 INIT_HLIST_HEAD(&pf->aq_wait_list);
3340 spin_lock_init(&pf->aq_wait_lock);
3341 init_waitqueue_head(&pf->aq_wait_queue);
3343 /* setup service timer and periodic service task */
3344 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3345 pf->serv_tmr_period = HZ;
3346 INIT_WORK(&pf->serv_task, ice_service_task);
3347 clear_bit(__ICE_SERVICE_SCHED, pf->state);
3349 mutex_init(&pf->avail_q_mutex);
3350 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3351 if (!pf->avail_txqs)
3354 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3355 if (!pf->avail_rxqs) {
3356 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3357 pf->avail_txqs = NULL;
3365 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3366 * @pf: board private structure
3368 * compute the number of MSIX vectors required (v_budget) and request from
3369 * the OS. Return the number of vectors reserved or negative on failure
3371 static int ice_ena_msix_range(struct ice_pf *pf)
3373 struct device *dev = ice_pf_to_dev(pf);
3374 int v_left, v_actual, v_budget = 0;
3377 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3379 /* reserve one vector for miscellaneous handler */
3381 if (v_left < needed)
3382 goto no_hw_vecs_left_err;
3386 /* reserve vectors for LAN traffic */
3387 needed = min_t(int, num_online_cpus(), v_left);
3388 if (v_left < needed)
3389 goto no_hw_vecs_left_err;
3390 pf->num_lan_msix = needed;
3394 /* reserve one vector for flow director */
3395 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3396 needed = ICE_FDIR_MSIX;
3397 if (v_left < needed)
3398 goto no_hw_vecs_left_err;
3403 pf->msix_entries = devm_kcalloc(dev, v_budget,
3404 sizeof(*pf->msix_entries), GFP_KERNEL);
3406 if (!pf->msix_entries) {
3411 for (i = 0; i < v_budget; i++)
3412 pf->msix_entries[i].entry = i;
3414 /* actually reserve the vectors */
3415 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3416 ICE_MIN_MSIX, v_budget);
3419 dev_err(dev, "unable to reserve MSI-X vectors\n");
3424 if (v_actual < v_budget) {
3425 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3426 v_budget, v_actual);
3427 /* 2 vectors each for LAN and RDMA (traffic + OICR), one for flow director */
3428 #define ICE_MIN_LAN_VECS 2
3429 #define ICE_MIN_RDMA_VECS 2
3430 #define ICE_MIN_VECS (ICE_MIN_LAN_VECS + ICE_MIN_RDMA_VECS + 1)
3432 if (v_actual < ICE_MIN_LAN_VECS) {
3433 /* error if we can't get minimum vectors */
3434 pci_disable_msix(pf->pdev);
3438 pf->num_lan_msix = ICE_MIN_LAN_VECS;
3445 devm_kfree(dev, pf->msix_entries);
3448 no_hw_vecs_left_err:
3449 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3453 pf->num_lan_msix = 0;
3458 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3459 * @pf: board private structure
3461 static void ice_dis_msix(struct ice_pf *pf)
3463 pci_disable_msix(pf->pdev);
3464 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3465 pf->msix_entries = NULL;
3469 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3470 * @pf: board private structure
3472 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3476 if (pf->irq_tracker) {
3477 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3478 pf->irq_tracker = NULL;
3483 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3484 * @pf: board private structure to initialize
3486 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3490 vectors = ice_ena_msix_range(pf);
3495 /* set up vector assignment tracking */
3497 devm_kzalloc(ice_pf_to_dev(pf), sizeof(*pf->irq_tracker) +
3498 (sizeof(u16) * vectors), GFP_KERNEL);
3499 if (!pf->irq_tracker) {
3504 /* populate SW interrupts pool with number of OS granted IRQs. */
3505 pf->num_avail_sw_msix = (u16)vectors;
3506 pf->irq_tracker->num_entries = (u16)vectors;
3507 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3513 * ice_is_wol_supported - get NVM state of WoL
3514 * @pf: board private structure
3516 * Check if WoL is supported based on the HW configuration.
3517 * Returns true if NVM supports and enables WoL for this port, false otherwise
3519 bool ice_is_wol_supported(struct ice_pf *pf)
3521 struct ice_hw *hw = &pf->hw;
3524 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3525 * word) indicates WoL is not supported on the corresponding PF ID.
3527 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3530 return !(BIT(hw->pf_id) & wol_ctrl);
3534 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3535 * @vsi: VSI being changed
3536 * @new_rx: new number of Rx queues
3537 * @new_tx: new number of Tx queues
3539 * Only change the number of queues if new_tx, or new_rx is non-0.
3541 * Returns 0 on success.
3543 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3545 struct ice_pf *pf = vsi->back;
3546 int err = 0, timeout = 50;
3548 if (!new_rx && !new_tx)
3551 while (test_and_set_bit(__ICE_CFG_BUSY, pf->state)) {
3555 usleep_range(1000, 2000);
3559 vsi->req_txq = (u16)new_tx;
3561 vsi->req_rxq = (u16)new_rx;
3563 /* set for the next time the netdev is started */
3564 if (!netif_running(vsi->netdev)) {
3565 ice_vsi_rebuild(vsi, false);
3566 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3571 ice_vsi_rebuild(vsi, false);
3572 ice_pf_dcb_recfg(pf);
3575 clear_bit(__ICE_CFG_BUSY, pf->state);
3580 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3581 * @pf: PF to configure
3583 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3584 * VSI can still Tx/Rx VLAN tagged packets.
3586 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3588 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3589 struct ice_vsi_ctx *ctxt;
3590 enum ice_status status;
3596 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3601 ctxt->info = vsi->info;
3603 ctxt->info.valid_sections =
3604 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3605 ICE_AQ_VSI_PROP_SECURITY_VALID |
3606 ICE_AQ_VSI_PROP_SW_VALID);
3608 /* disable VLAN anti-spoof */
3609 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3610 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3612 /* disable VLAN pruning and keep all other settings */
3613 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3615 /* allow all VLANs on Tx and don't strip on Rx */
3616 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3617 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3619 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3621 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3622 ice_stat_str(status),
3623 ice_aq_str(hw->adminq.sq_last_status));
3625 vsi->info.sec_flags = ctxt->info.sec_flags;
3626 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3627 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3634 * ice_log_pkg_init - log result of DDP package load
3635 * @hw: pointer to hardware info
3636 * @status: status of package load
3639 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3641 struct ice_pf *pf = (struct ice_pf *)hw->back;
3642 struct device *dev = ice_pf_to_dev(pf);
3646 /* The package download AdminQ command returned success because
3647 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3648 * already a package loaded on the device.
3650 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3651 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3652 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3653 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3654 !memcmp(hw->pkg_name, hw->active_pkg_name,
3655 sizeof(hw->pkg_name))) {
3656 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3657 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3658 hw->active_pkg_name,
3659 hw->active_pkg_ver.major,
3660 hw->active_pkg_ver.minor,
3661 hw->active_pkg_ver.update,
3662 hw->active_pkg_ver.draft);
3664 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3665 hw->active_pkg_name,
3666 hw->active_pkg_ver.major,
3667 hw->active_pkg_ver.minor,
3668 hw->active_pkg_ver.update,
3669 hw->active_pkg_ver.draft);
3670 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3671 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3672 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",
3673 hw->active_pkg_name,
3674 hw->active_pkg_ver.major,
3675 hw->active_pkg_ver.minor,
3676 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3677 *status = ICE_ERR_NOT_SUPPORTED;
3678 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3679 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3680 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",
3681 hw->active_pkg_name,
3682 hw->active_pkg_ver.major,
3683 hw->active_pkg_ver.minor,
3684 hw->active_pkg_ver.update,
3685 hw->active_pkg_ver.draft,
3692 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");
3693 *status = ICE_ERR_NOT_SUPPORTED;
3696 case ICE_ERR_FW_DDP_MISMATCH:
3697 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");
3699 case ICE_ERR_BUF_TOO_SHORT:
3701 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3703 case ICE_ERR_NOT_SUPPORTED:
3704 /* Package File version not supported */
3705 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3706 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3707 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3708 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3709 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3710 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3711 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3712 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",
3713 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3715 case ICE_ERR_AQ_ERROR:
3716 switch (hw->pkg_dwnld_status) {
3717 case ICE_AQ_RC_ENOSEC:
3718 case ICE_AQ_RC_EBADSIG:
3719 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");
3721 case ICE_AQ_RC_ESVN:
3722 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");
3724 case ICE_AQ_RC_EBADMAN:
3725 case ICE_AQ_RC_EBADBUF:
3726 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3727 /* poll for reset to complete */
3728 if (ice_check_reset(hw))
3729 dev_err(dev, "Error resetting device. Please reload the driver\n");
3736 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3743 * ice_load_pkg - load/reload the DDP Package file
3744 * @firmware: firmware structure when firmware requested or NULL for reload
3745 * @pf: pointer to the PF instance
3747 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3748 * initialize HW tables.
3751 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3753 enum ice_status status = ICE_ERR_PARAM;
3754 struct device *dev = ice_pf_to_dev(pf);
3755 struct ice_hw *hw = &pf->hw;
3757 /* Load DDP Package */
3758 if (firmware && !hw->pkg_copy) {
3759 status = ice_copy_and_init_pkg(hw, firmware->data,
3761 ice_log_pkg_init(hw, &status);
3762 } else if (!firmware && hw->pkg_copy) {
3763 /* Reload package during rebuild after CORER/GLOBR reset */
3764 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3765 ice_log_pkg_init(hw, &status);
3767 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3772 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3776 /* Successful download package is the precondition for advanced
3777 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3779 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3783 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3784 * @pf: pointer to the PF structure
3786 * There is no error returned here because the driver should be able to handle
3787 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3788 * specifically with Tx.
3790 static void ice_verify_cacheline_size(struct ice_pf *pf)
3792 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3793 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3794 ICE_CACHE_LINE_BYTES);
3798 * ice_send_version - update firmware with driver version
3801 * Returns ICE_SUCCESS on success, else error code
3803 static enum ice_status ice_send_version(struct ice_pf *pf)
3805 struct ice_driver_ver dv;
3807 dv.major_ver = 0xff;
3808 dv.minor_ver = 0xff;
3809 dv.build_ver = 0xff;
3810 dv.subbuild_ver = 0;
3811 strscpy((char *)dv.driver_string, UTS_RELEASE,
3812 sizeof(dv.driver_string));
3813 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3817 * ice_init_fdir - Initialize flow director VSI and configuration
3818 * @pf: pointer to the PF instance
3820 * returns 0 on success, negative on error
3822 static int ice_init_fdir(struct ice_pf *pf)
3824 struct device *dev = ice_pf_to_dev(pf);
3825 struct ice_vsi *ctrl_vsi;
3828 /* Side Band Flow Director needs to have a control VSI.
3829 * Allocate it and store it in the PF.
3831 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3833 dev_dbg(dev, "could not create control VSI\n");
3837 err = ice_vsi_open_ctrl(ctrl_vsi);
3839 dev_dbg(dev, "could not open control VSI\n");
3843 mutex_init(&pf->hw.fdir_fltr_lock);
3845 err = ice_fdir_create_dflt_rules(pf);
3852 ice_fdir_release_flows(&pf->hw);
3853 ice_vsi_close(ctrl_vsi);
3855 ice_vsi_release(ctrl_vsi);
3856 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3857 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3858 pf->ctrl_vsi_idx = ICE_NO_VSI;
3864 * ice_get_opt_fw_name - return optional firmware file name or NULL
3865 * @pf: pointer to the PF instance
3867 static char *ice_get_opt_fw_name(struct ice_pf *pf)
3869 /* Optional firmware name same as default with additional dash
3870 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3872 struct pci_dev *pdev = pf->pdev;
3873 char *opt_fw_filename;
3876 /* Determine the name of the optional file using the DSN (two
3877 * dwords following the start of the DSN Capability).
3879 dsn = pci_get_dsn(pdev);
3883 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3884 if (!opt_fw_filename)
3887 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3888 ICE_DDP_PKG_PATH, dsn);
3890 return opt_fw_filename;
3894 * ice_request_fw - Device initialization routine
3895 * @pf: pointer to the PF instance
3897 static void ice_request_fw(struct ice_pf *pf)
3899 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3900 const struct firmware *firmware = NULL;
3901 struct device *dev = ice_pf_to_dev(pf);
3904 /* optional device-specific DDP (if present) overrides the default DDP
3905 * package file. kernel logs a debug message if the file doesn't exist,
3906 * and warning messages for other errors.
3908 if (opt_fw_filename) {
3909 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3911 kfree(opt_fw_filename);
3915 /* request for firmware was successful. Download to device */
3916 ice_load_pkg(firmware, pf);
3917 kfree(opt_fw_filename);
3918 release_firmware(firmware);
3923 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3925 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3929 /* request for firmware was successful. Download to device */
3930 ice_load_pkg(firmware, pf);
3931 release_firmware(firmware);
3935 * ice_print_wake_reason - show the wake up cause in the log
3936 * @pf: pointer to the PF struct
3938 static void ice_print_wake_reason(struct ice_pf *pf)
3940 u32 wus = pf->wakeup_reason;
3941 const char *wake_str;
3943 /* if no wake event, nothing to print */
3947 if (wus & PFPM_WUS_LNKC_M)
3948 wake_str = "Link\n";
3949 else if (wus & PFPM_WUS_MAG_M)
3950 wake_str = "Magic Packet\n";
3951 else if (wus & PFPM_WUS_MNG_M)
3952 wake_str = "Management\n";
3953 else if (wus & PFPM_WUS_FW_RST_WK_M)
3954 wake_str = "Firmware Reset\n";
3956 wake_str = "Unknown\n";
3958 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
3962 * ice_probe - Device initialization routine
3963 * @pdev: PCI device information struct
3964 * @ent: entry in ice_pci_tbl
3966 * Returns 0 on success, negative on failure
3969 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
3971 struct device *dev = &pdev->dev;
3976 /* this driver uses devres, see
3977 * Documentation/driver-api/driver-model/devres.rst
3979 err = pcim_enable_device(pdev);
3983 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
3985 dev_err(dev, "BAR0 I/O map error %d\n", err);
3989 pf = ice_allocate_pf(dev);
3993 /* set up for high or low DMA */
3994 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
3996 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
3998 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4002 pci_enable_pcie_error_reporting(pdev);
4003 pci_set_master(pdev);
4006 pci_set_drvdata(pdev, pf);
4007 set_bit(__ICE_DOWN, pf->state);
4008 /* Disable service task until DOWN bit is cleared */
4009 set_bit(__ICE_SERVICE_DIS, pf->state);
4012 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4013 pci_save_state(pdev);
4016 hw->vendor_id = pdev->vendor;
4017 hw->device_id = pdev->device;
4018 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4019 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4020 hw->subsystem_device_id = pdev->subsystem_device;
4021 hw->bus.device = PCI_SLOT(pdev->devfn);
4022 hw->bus.func = PCI_FUNC(pdev->devfn);
4023 ice_set_ctrlq_len(hw);
4025 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4027 err = ice_devlink_register(pf);
4029 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4030 goto err_exit_unroll;
4033 #ifndef CONFIG_DYNAMIC_DEBUG
4035 hw->debug_mask = debug;
4038 err = ice_init_hw(hw);
4040 dev_err(dev, "ice_init_hw failed: %d\n", err);
4042 goto err_exit_unroll;
4047 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4048 * set in pf->state, which will cause ice_is_safe_mode to return
4051 if (ice_is_safe_mode(pf)) {
4052 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4053 /* we already got function/device capabilities but these don't
4054 * reflect what the driver needs to do in safe mode. Instead of
4055 * adding conditional logic everywhere to ignore these
4056 * device/function capabilities, override them.
4058 ice_set_safe_mode_caps(hw);
4061 err = ice_init_pf(pf);
4063 dev_err(dev, "ice_init_pf failed: %d\n", err);
4064 goto err_init_pf_unroll;
4067 ice_devlink_init_regions(pf);
4069 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4070 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4071 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4072 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4074 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4075 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4076 pf->hw.tnl.valid_count[TNL_VXLAN];
4077 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4078 UDP_TUNNEL_TYPE_VXLAN;
4081 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4082 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4083 pf->hw.tnl.valid_count[TNL_GENEVE];
4084 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4085 UDP_TUNNEL_TYPE_GENEVE;
4089 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4090 if (!pf->num_alloc_vsi) {
4092 goto err_init_pf_unroll;
4094 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4095 dev_warn(&pf->pdev->dev,
4096 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4097 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4098 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4101 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4105 goto err_init_pf_unroll;
4108 err = ice_init_interrupt_scheme(pf);
4110 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4112 goto err_init_vsi_unroll;
4115 /* In case of MSIX we are going to setup the misc vector right here
4116 * to handle admin queue events etc. In case of legacy and MSI
4117 * the misc functionality and queue processing is combined in
4118 * the same vector and that gets setup at open.
4120 err = ice_req_irq_msix_misc(pf);
4122 dev_err(dev, "setup of misc vector failed: %d\n", err);
4123 goto err_init_interrupt_unroll;
4126 /* create switch struct for the switch element created by FW on boot */
4127 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4128 if (!pf->first_sw) {
4130 goto err_msix_misc_unroll;
4134 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4136 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4138 pf->first_sw->pf = pf;
4140 /* record the sw_id available for later use */
4141 pf->first_sw->sw_id = hw->port_info->sw_id;
4143 err = ice_setup_pf_sw(pf);
4145 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4146 goto err_alloc_sw_unroll;
4149 clear_bit(__ICE_SERVICE_DIS, pf->state);
4151 /* tell the firmware we are up */
4152 err = ice_send_version(pf);
4154 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4156 goto err_send_version_unroll;
4159 /* since everything is good, start the service timer */
4160 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4162 err = ice_init_link_events(pf->hw.port_info);
4164 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4165 goto err_send_version_unroll;
4168 err = ice_init_nvm_phy_type(pf->hw.port_info);
4170 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4171 goto err_send_version_unroll;
4174 err = ice_update_link_info(pf->hw.port_info);
4176 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4177 goto err_send_version_unroll;
4180 ice_init_link_dflt_override(pf->hw.port_info);
4182 /* if media available, initialize PHY settings */
4183 if (pf->hw.port_info->phy.link_info.link_info &
4184 ICE_AQ_MEDIA_AVAILABLE) {
4185 err = ice_init_phy_user_cfg(pf->hw.port_info);
4187 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4188 goto err_send_version_unroll;
4191 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4192 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4195 ice_configure_phy(vsi);
4198 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4201 ice_verify_cacheline_size(pf);
4203 /* Save wakeup reason register for later use */
4204 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4206 /* check for a power management event */
4207 ice_print_wake_reason(pf);
4209 /* clear wake status, all bits */
4210 wr32(hw, PFPM_WUS, U32_MAX);
4212 /* Disable WoL at init, wait for user to enable */
4213 device_set_wakeup_enable(dev, false);
4215 if (ice_is_safe_mode(pf)) {
4216 ice_set_safe_mode_vlan_cfg(pf);
4220 /* initialize DDP driven features */
4222 /* Note: Flow director init failure is non-fatal to load */
4223 if (ice_init_fdir(pf))
4224 dev_err(dev, "could not initialize flow director\n");
4226 /* Note: DCB init failure is non-fatal to load */
4227 if (ice_init_pf_dcb(pf, false)) {
4228 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4229 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4231 ice_cfg_lldp_mib_change(&pf->hw, true);
4234 /* print PCI link speed and width */
4235 pcie_print_link_status(pf->pdev);
4238 /* ready to go, so clear down state bit */
4239 clear_bit(__ICE_DOWN, pf->state);
4242 err_send_version_unroll:
4243 ice_vsi_release_all(pf);
4244 err_alloc_sw_unroll:
4245 ice_devlink_destroy_port(pf);
4246 set_bit(__ICE_SERVICE_DIS, pf->state);
4247 set_bit(__ICE_DOWN, pf->state);
4248 devm_kfree(dev, pf->first_sw);
4249 err_msix_misc_unroll:
4250 ice_free_irq_msix_misc(pf);
4251 err_init_interrupt_unroll:
4252 ice_clear_interrupt_scheme(pf);
4253 err_init_vsi_unroll:
4254 devm_kfree(dev, pf->vsi);
4257 ice_devlink_destroy_regions(pf);
4260 ice_devlink_unregister(pf);
4261 pci_disable_pcie_error_reporting(pdev);
4262 pci_disable_device(pdev);
4267 * ice_set_wake - enable or disable Wake on LAN
4268 * @pf: pointer to the PF struct
4270 * Simple helper for WoL control
4272 static void ice_set_wake(struct ice_pf *pf)
4274 struct ice_hw *hw = &pf->hw;
4275 bool wol = pf->wol_ena;
4277 /* clear wake state, otherwise new wake events won't fire */
4278 wr32(hw, PFPM_WUS, U32_MAX);
4280 /* enable / disable APM wake up, no RMW needed */
4281 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4283 /* set magic packet filter enabled */
4284 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4288 * ice_setup_magic_mc_wake - setup device to wake on multicast magic packet
4289 * @pf: pointer to the PF struct
4291 * Issue firmware command to enable multicast magic wake, making
4292 * sure that any locally administered address (LAA) is used for
4293 * wake, and that PF reset doesn't undo the LAA.
4295 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4297 struct device *dev = ice_pf_to_dev(pf);
4298 struct ice_hw *hw = &pf->hw;
4299 enum ice_status status;
4300 u8 mac_addr[ETH_ALEN];
4301 struct ice_vsi *vsi;
4307 vsi = ice_get_main_vsi(pf);
4311 /* Get current MAC address in case it's an LAA */
4313 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4315 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4317 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4318 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4319 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4321 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4323 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4324 ice_stat_str(status),
4325 ice_aq_str(hw->adminq.sq_last_status));
4329 * ice_remove - Device removal routine
4330 * @pdev: PCI device information struct
4332 static void ice_remove(struct pci_dev *pdev)
4334 struct ice_pf *pf = pci_get_drvdata(pdev);
4340 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4341 if (!ice_is_reset_in_progress(pf->state))
4346 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4347 set_bit(__ICE_VF_RESETS_DISABLED, pf->state);
4351 set_bit(__ICE_DOWN, pf->state);
4352 ice_service_task_stop(pf);
4354 ice_aq_cancel_waiting_tasks(pf);
4356 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4357 if (!ice_is_safe_mode(pf))
4358 ice_remove_arfs(pf);
4359 ice_setup_mc_magic_wake(pf);
4360 ice_devlink_destroy_port(pf);
4361 ice_vsi_release_all(pf);
4363 ice_free_irq_msix_misc(pf);
4364 ice_for_each_vsi(pf, i) {
4367 ice_vsi_free_q_vectors(pf->vsi[i]);
4370 ice_devlink_destroy_regions(pf);
4371 ice_deinit_hw(&pf->hw);
4372 ice_devlink_unregister(pf);
4374 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4375 * do it via ice_schedule_reset() since there is no need to rebuild
4376 * and the service task is already stopped.
4378 ice_reset(&pf->hw, ICE_RESET_PFR);
4379 pci_wait_for_pending_transaction(pdev);
4380 ice_clear_interrupt_scheme(pf);
4381 pci_disable_pcie_error_reporting(pdev);
4382 pci_disable_device(pdev);
4386 * ice_shutdown - PCI callback for shutting down device
4387 * @pdev: PCI device information struct
4389 static void ice_shutdown(struct pci_dev *pdev)
4391 struct ice_pf *pf = pci_get_drvdata(pdev);
4395 if (system_state == SYSTEM_POWER_OFF) {
4396 pci_wake_from_d3(pdev, pf->wol_ena);
4397 pci_set_power_state(pdev, PCI_D3hot);
4403 * ice_prepare_for_shutdown - prep for PCI shutdown
4404 * @pf: board private structure
4406 * Inform or close all dependent features in prep for PCI device shutdown
4408 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4410 struct ice_hw *hw = &pf->hw;
4413 /* Notify VFs of impending reset */
4414 if (ice_check_sq_alive(hw, &hw->mailboxq))
4415 ice_vc_notify_reset(pf);
4417 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4419 /* disable the VSIs and their queues that are not already DOWN */
4420 ice_pf_dis_all_vsi(pf, false);
4422 ice_for_each_vsi(pf, v)
4424 pf->vsi[v]->vsi_num = 0;
4426 ice_shutdown_all_ctrlq(hw);
4430 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4431 * @pf: board private structure to reinitialize
4433 * This routine reinitialize interrupt scheme that was cleared during
4434 * power management suspend callback.
4436 * This should be called during resume routine to re-allocate the q_vectors
4437 * and reacquire interrupts.
4439 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4441 struct device *dev = ice_pf_to_dev(pf);
4444 /* Since we clear MSIX flag during suspend, we need to
4445 * set it back during resume...
4448 ret = ice_init_interrupt_scheme(pf);
4450 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4454 /* Remap vectors and rings, after successful re-init interrupts */
4455 ice_for_each_vsi(pf, v) {
4459 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4462 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4465 ret = ice_req_irq_msix_misc(pf);
4467 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4477 ice_vsi_free_q_vectors(pf->vsi[v]);
4484 * @dev: generic device information structure
4486 * Power Management callback to quiesce the device and prepare
4487 * for D3 transition.
4489 static int __maybe_unused ice_suspend(struct device *dev)
4491 struct pci_dev *pdev = to_pci_dev(dev);
4495 pf = pci_get_drvdata(pdev);
4497 if (!ice_pf_state_is_nominal(pf)) {
4498 dev_err(dev, "Device is not ready, no need to suspend it\n");
4502 /* Stop watchdog tasks until resume completion.
4503 * Even though it is most likely that the service task is
4504 * disabled if the device is suspended or down, the service task's
4505 * state is controlled by a different state bit, and we should
4506 * store and honor whatever state that bit is in at this point.
4508 disabled = ice_service_task_stop(pf);
4510 /* Already suspended?, then there is nothing to do */
4511 if (test_and_set_bit(__ICE_SUSPENDED, pf->state)) {
4513 ice_service_task_restart(pf);
4517 if (test_bit(__ICE_DOWN, pf->state) ||
4518 ice_is_reset_in_progress(pf->state)) {
4519 dev_err(dev, "can't suspend device in reset or already down\n");
4521 ice_service_task_restart(pf);
4525 ice_setup_mc_magic_wake(pf);
4527 ice_prepare_for_shutdown(pf);
4531 /* Free vectors, clear the interrupt scheme and release IRQs
4532 * for proper hibernation, especially with large number of CPUs.
4533 * Otherwise hibernation might fail when mapping all the vectors back
4536 ice_free_irq_msix_misc(pf);
4537 ice_for_each_vsi(pf, v) {
4540 ice_vsi_free_q_vectors(pf->vsi[v]);
4542 ice_clear_interrupt_scheme(pf);
4544 pci_save_state(pdev);
4545 pci_wake_from_d3(pdev, pf->wol_ena);
4546 pci_set_power_state(pdev, PCI_D3hot);
4551 * ice_resume - PM callback for waking up from D3
4552 * @dev: generic device information structure
4554 static int __maybe_unused ice_resume(struct device *dev)
4556 struct pci_dev *pdev = to_pci_dev(dev);
4557 enum ice_reset_req reset_type;
4562 pci_set_power_state(pdev, PCI_D0);
4563 pci_restore_state(pdev);
4564 pci_save_state(pdev);
4566 if (!pci_device_is_present(pdev))
4569 ret = pci_enable_device_mem(pdev);
4571 dev_err(dev, "Cannot enable device after suspend\n");
4575 pf = pci_get_drvdata(pdev);
4578 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4579 ice_print_wake_reason(pf);
4581 /* We cleared the interrupt scheme when we suspended, so we need to
4582 * restore it now to resume device functionality.
4584 ret = ice_reinit_interrupt_scheme(pf);
4586 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4588 clear_bit(__ICE_DOWN, pf->state);
4589 /* Now perform PF reset and rebuild */
4590 reset_type = ICE_RESET_PFR;
4591 /* re-enable service task for reset, but allow reset to schedule it */
4592 clear_bit(__ICE_SERVICE_DIS, pf->state);
4594 if (ice_schedule_reset(pf, reset_type))
4595 dev_err(dev, "Reset during resume failed.\n");
4597 clear_bit(__ICE_SUSPENDED, pf->state);
4598 ice_service_task_restart(pf);
4600 /* Restart the service task */
4601 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4605 #endif /* CONFIG_PM */
4608 * ice_pci_err_detected - warning that PCI error has been detected
4609 * @pdev: PCI device information struct
4610 * @err: the type of PCI error
4612 * Called to warn that something happened on the PCI bus and the error handling
4613 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4615 static pci_ers_result_t
4616 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4618 struct ice_pf *pf = pci_get_drvdata(pdev);
4621 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4623 return PCI_ERS_RESULT_DISCONNECT;
4626 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4627 ice_service_task_stop(pf);
4629 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4630 set_bit(__ICE_PFR_REQ, pf->state);
4631 ice_prepare_for_reset(pf);
4635 return PCI_ERS_RESULT_NEED_RESET;
4639 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4640 * @pdev: PCI device information struct
4642 * Called to determine if the driver can recover from the PCI slot reset by
4643 * using a register read to determine if the device is recoverable.
4645 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4647 struct ice_pf *pf = pci_get_drvdata(pdev);
4648 pci_ers_result_t result;
4652 err = pci_enable_device_mem(pdev);
4654 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4656 result = PCI_ERS_RESULT_DISCONNECT;
4658 pci_set_master(pdev);
4659 pci_restore_state(pdev);
4660 pci_save_state(pdev);
4661 pci_wake_from_d3(pdev, false);
4663 /* Check for life */
4664 reg = rd32(&pf->hw, GLGEN_RTRIG);
4666 result = PCI_ERS_RESULT_RECOVERED;
4668 result = PCI_ERS_RESULT_DISCONNECT;
4671 err = pci_aer_clear_nonfatal_status(pdev);
4673 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4675 /* non-fatal, continue */
4681 * ice_pci_err_resume - restart operations after PCI error recovery
4682 * @pdev: PCI device information struct
4684 * Called to allow the driver to bring things back up after PCI error and/or
4685 * reset recovery have finished
4687 static void ice_pci_err_resume(struct pci_dev *pdev)
4689 struct ice_pf *pf = pci_get_drvdata(pdev);
4692 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4697 if (test_bit(__ICE_SUSPENDED, pf->state)) {
4698 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4703 ice_restore_all_vfs_msi_state(pdev);
4705 ice_do_reset(pf, ICE_RESET_PFR);
4706 ice_service_task_restart(pf);
4707 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4711 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4712 * @pdev: PCI device information struct
4714 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4716 struct ice_pf *pf = pci_get_drvdata(pdev);
4718 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4719 ice_service_task_stop(pf);
4721 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4722 set_bit(__ICE_PFR_REQ, pf->state);
4723 ice_prepare_for_reset(pf);
4729 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4730 * @pdev: PCI device information struct
4732 static void ice_pci_err_reset_done(struct pci_dev *pdev)
4734 ice_pci_err_resume(pdev);
4737 /* ice_pci_tbl - PCI Device ID Table
4739 * Wildcard entries (PCI_ANY_ID) should come last
4740 * Last entry must be all 0s
4742 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4743 * Class, Class Mask, private data (not used) }
4745 static const struct pci_device_id ice_pci_tbl[] = {
4746 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4747 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4748 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4749 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4750 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4751 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4752 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4753 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4754 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4755 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4756 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4757 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4758 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4759 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4760 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4761 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4762 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4763 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4764 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4765 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4766 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4767 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4768 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4769 /* required last entry */
4772 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4774 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4776 static const struct pci_error_handlers ice_pci_err_handler = {
4777 .error_detected = ice_pci_err_detected,
4778 .slot_reset = ice_pci_err_slot_reset,
4779 .reset_prepare = ice_pci_err_reset_prepare,
4780 .reset_done = ice_pci_err_reset_done,
4781 .resume = ice_pci_err_resume
4784 static struct pci_driver ice_driver = {
4785 .name = KBUILD_MODNAME,
4786 .id_table = ice_pci_tbl,
4788 .remove = ice_remove,
4790 .driver.pm = &ice_pm_ops,
4791 #endif /* CONFIG_PM */
4792 .shutdown = ice_shutdown,
4793 .sriov_configure = ice_sriov_configure,
4794 .err_handler = &ice_pci_err_handler
4798 * ice_module_init - Driver registration routine
4800 * ice_module_init is the first routine called when the driver is
4801 * loaded. All it does is register with the PCI subsystem.
4803 static int __init ice_module_init(void)
4807 pr_info("%s\n", ice_driver_string);
4808 pr_info("%s\n", ice_copyright);
4810 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4812 pr_err("Failed to create workqueue\n");
4816 status = pci_register_driver(&ice_driver);
4818 pr_err("failed to register PCI driver, err %d\n", status);
4819 destroy_workqueue(ice_wq);
4824 module_init(ice_module_init);
4827 * ice_module_exit - Driver exit cleanup routine
4829 * ice_module_exit is called just before the driver is removed
4832 static void __exit ice_module_exit(void)
4834 pci_unregister_driver(&ice_driver);
4835 destroy_workqueue(ice_wq);
4836 pr_info("module unloaded\n");
4838 module_exit(ice_module_exit);
4841 * ice_set_mac_address - NDO callback to set MAC address
4842 * @netdev: network interface device structure
4843 * @pi: pointer to an address structure
4845 * Returns 0 on success, negative on failure
4847 static int ice_set_mac_address(struct net_device *netdev, void *pi)
4849 struct ice_netdev_priv *np = netdev_priv(netdev);
4850 struct ice_vsi *vsi = np->vsi;
4851 struct ice_pf *pf = vsi->back;
4852 struct ice_hw *hw = &pf->hw;
4853 struct sockaddr *addr = pi;
4854 enum ice_status status;
4859 mac = (u8 *)addr->sa_data;
4861 if (!is_valid_ether_addr(mac))
4862 return -EADDRNOTAVAIL;
4864 if (ether_addr_equal(netdev->dev_addr, mac)) {
4865 netdev_warn(netdev, "already using mac %pM\n", mac);
4869 if (test_bit(__ICE_DOWN, pf->state) ||
4870 ice_is_reset_in_progress(pf->state)) {
4871 netdev_err(netdev, "can't set mac %pM. device not ready\n",
4876 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
4877 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
4878 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
4879 err = -EADDRNOTAVAIL;
4880 goto err_update_filters;
4883 /* Add filter for new MAC. If filter exists, just return success */
4884 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
4885 if (status == ICE_ERR_ALREADY_EXISTS) {
4886 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
4890 /* error if the new filter addition failed */
4892 err = -EADDRNOTAVAIL;
4896 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
4901 /* change the netdev's MAC address */
4902 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4903 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
4906 /* write new MAC address to the firmware */
4907 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
4908 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
4910 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
4911 mac, ice_stat_str(status));
4917 * ice_set_rx_mode - NDO callback to set the netdev filters
4918 * @netdev: network interface device structure
4920 static void ice_set_rx_mode(struct net_device *netdev)
4922 struct ice_netdev_priv *np = netdev_priv(netdev);
4923 struct ice_vsi *vsi = np->vsi;
4928 /* Set the flags to synchronize filters
4929 * ndo_set_rx_mode may be triggered even without a change in netdev
4932 set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
4933 set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
4934 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
4936 /* schedule our worker thread which will take care of
4937 * applying the new filter changes
4939 ice_service_task_schedule(vsi->back);
4943 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
4944 * @netdev: network interface device structure
4945 * @queue_index: Queue ID
4946 * @maxrate: maximum bandwidth in Mbps
4949 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
4951 struct ice_netdev_priv *np = netdev_priv(netdev);
4952 struct ice_vsi *vsi = np->vsi;
4953 enum ice_status status;
4957 /* Validate maxrate requested is within permitted range */
4958 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
4959 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
4960 maxrate, queue_index);
4964 q_handle = vsi->tx_rings[queue_index]->q_handle;
4965 tc = ice_dcb_get_tc(vsi, queue_index);
4967 /* Set BW back to default, when user set maxrate to 0 */
4969 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
4970 q_handle, ICE_MAX_BW);
4972 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
4973 q_handle, ICE_MAX_BW, maxrate * 1000);
4975 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
4976 ice_stat_str(status));
4984 * ice_fdb_add - add an entry to the hardware database
4985 * @ndm: the input from the stack
4986 * @tb: pointer to array of nladdr (unused)
4987 * @dev: the net device pointer
4988 * @addr: the MAC address entry being added
4990 * @flags: instructions from stack about fdb operation
4991 * @extack: netlink extended ack
4994 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
4995 struct net_device *dev, const unsigned char *addr, u16 vid,
4996 u16 flags, struct netlink_ext_ack __always_unused *extack)
5001 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5004 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5005 netdev_err(dev, "FDB only supports static addresses\n");
5009 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5010 err = dev_uc_add_excl(dev, addr);
5011 else if (is_multicast_ether_addr(addr))
5012 err = dev_mc_add_excl(dev, addr);
5016 /* Only return duplicate errors if NLM_F_EXCL is set */
5017 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5024 * ice_fdb_del - delete an entry from the hardware database
5025 * @ndm: the input from the stack
5026 * @tb: pointer to array of nladdr (unused)
5027 * @dev: the net device pointer
5028 * @addr: the MAC address entry being added
5032 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5033 struct net_device *dev, const unsigned char *addr,
5034 __always_unused u16 vid)
5038 if (ndm->ndm_state & NUD_PERMANENT) {
5039 netdev_err(dev, "FDB only supports static addresses\n");
5043 if (is_unicast_ether_addr(addr))
5044 err = dev_uc_del(dev, addr);
5045 else if (is_multicast_ether_addr(addr))
5046 err = dev_mc_del(dev, addr);
5054 * ice_set_features - set the netdev feature flags
5055 * @netdev: ptr to the netdev being adjusted
5056 * @features: the feature set that the stack is suggesting
5059 ice_set_features(struct net_device *netdev, netdev_features_t features)
5061 struct ice_netdev_priv *np = netdev_priv(netdev);
5062 struct ice_vsi *vsi = np->vsi;
5063 struct ice_pf *pf = vsi->back;
5066 /* Don't set any netdev advanced features with device in Safe Mode */
5067 if (ice_is_safe_mode(vsi->back)) {
5068 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5072 /* Do not change setting during reset */
5073 if (ice_is_reset_in_progress(pf->state)) {
5074 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5078 /* Multiple features can be changed in one call so keep features in
5079 * separate if/else statements to guarantee each feature is checked
5081 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5082 ret = ice_vsi_manage_rss_lut(vsi, true);
5083 else if (!(features & NETIF_F_RXHASH) &&
5084 netdev->features & NETIF_F_RXHASH)
5085 ret = ice_vsi_manage_rss_lut(vsi, false);
5087 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5088 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5089 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5090 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5091 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5092 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5094 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5095 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5096 ret = ice_vsi_manage_vlan_insertion(vsi);
5097 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5098 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5099 ret = ice_vsi_manage_vlan_insertion(vsi);
5101 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5102 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5103 ret = ice_cfg_vlan_pruning(vsi, true, false);
5104 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5105 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5106 ret = ice_cfg_vlan_pruning(vsi, false, false);
5108 if ((features & NETIF_F_NTUPLE) &&
5109 !(netdev->features & NETIF_F_NTUPLE)) {
5110 ice_vsi_manage_fdir(vsi, true);
5112 } else if (!(features & NETIF_F_NTUPLE) &&
5113 (netdev->features & NETIF_F_NTUPLE)) {
5114 ice_vsi_manage_fdir(vsi, false);
5115 ice_clear_arfs(vsi);
5122 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5123 * @vsi: VSI to setup VLAN properties for
5125 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5129 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5130 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5131 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5132 ret = ice_vsi_manage_vlan_insertion(vsi);
5138 * ice_vsi_cfg - Setup the VSI
5139 * @vsi: the VSI being configured
5141 * Return 0 on success and negative value on error
5143 int ice_vsi_cfg(struct ice_vsi *vsi)
5148 ice_set_rx_mode(vsi->netdev);
5150 err = ice_vsi_vlan_setup(vsi);
5155 ice_vsi_cfg_dcb_rings(vsi);
5157 err = ice_vsi_cfg_lan_txqs(vsi);
5158 if (!err && ice_is_xdp_ena_vsi(vsi))
5159 err = ice_vsi_cfg_xdp_txqs(vsi);
5161 err = ice_vsi_cfg_rxqs(vsi);
5167 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5168 * @vsi: the VSI being configured
5170 static void ice_napi_enable_all(struct ice_vsi *vsi)
5177 ice_for_each_q_vector(vsi, q_idx) {
5178 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5180 if (q_vector->rx.ring || q_vector->tx.ring)
5181 napi_enable(&q_vector->napi);
5186 * ice_up_complete - Finish the last steps of bringing up a connection
5187 * @vsi: The VSI being configured
5189 * Return 0 on success and negative value on error
5191 static int ice_up_complete(struct ice_vsi *vsi)
5193 struct ice_pf *pf = vsi->back;
5196 ice_vsi_cfg_msix(vsi);
5198 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5199 * Tx queue group list was configured and the context bits were
5200 * programmed using ice_vsi_cfg_txqs
5202 err = ice_vsi_start_all_rx_rings(vsi);
5206 clear_bit(__ICE_DOWN, vsi->state);
5207 ice_napi_enable_all(vsi);
5208 ice_vsi_ena_irq(vsi);
5210 if (vsi->port_info &&
5211 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5213 ice_print_link_msg(vsi, true);
5214 netif_tx_start_all_queues(vsi->netdev);
5215 netif_carrier_on(vsi->netdev);
5218 ice_service_task_schedule(pf);
5224 * ice_up - Bring the connection back up after being down
5225 * @vsi: VSI being configured
5227 int ice_up(struct ice_vsi *vsi)
5231 err = ice_vsi_cfg(vsi);
5233 err = ice_up_complete(vsi);
5239 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5240 * @ring: Tx or Rx ring to read stats from
5241 * @pkts: packets stats counter
5242 * @bytes: bytes stats counter
5244 * This function fetches stats from the ring considering the atomic operations
5245 * that needs to be performed to read u64 values in 32 bit machine.
5248 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5257 start = u64_stats_fetch_begin_irq(&ring->syncp);
5258 *pkts = ring->stats.pkts;
5259 *bytes = ring->stats.bytes;
5260 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5264 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5265 * @vsi: the VSI to be updated
5266 * @rings: rings to work on
5267 * @count: number of rings
5270 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5273 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5276 for (i = 0; i < count; i++) {
5277 struct ice_ring *ring;
5280 ring = READ_ONCE(rings[i]);
5281 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5282 vsi_stats->tx_packets += pkts;
5283 vsi_stats->tx_bytes += bytes;
5284 vsi->tx_restart += ring->tx_stats.restart_q;
5285 vsi->tx_busy += ring->tx_stats.tx_busy;
5286 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5291 * ice_update_vsi_ring_stats - Update VSI stats counters
5292 * @vsi: the VSI to be updated
5294 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5296 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5297 struct ice_ring *ring;
5301 /* reset netdev stats */
5302 vsi_stats->tx_packets = 0;
5303 vsi_stats->tx_bytes = 0;
5304 vsi_stats->rx_packets = 0;
5305 vsi_stats->rx_bytes = 0;
5307 /* reset non-netdev (extended) stats */
5308 vsi->tx_restart = 0;
5310 vsi->tx_linearize = 0;
5311 vsi->rx_buf_failed = 0;
5312 vsi->rx_page_failed = 0;
5313 vsi->rx_gro_dropped = 0;
5317 /* update Tx rings counters */
5318 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5320 /* update Rx rings counters */
5321 ice_for_each_rxq(vsi, i) {
5322 ring = READ_ONCE(vsi->rx_rings[i]);
5323 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5324 vsi_stats->rx_packets += pkts;
5325 vsi_stats->rx_bytes += bytes;
5326 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5327 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5328 vsi->rx_gro_dropped += ring->rx_stats.gro_dropped;
5331 /* update XDP Tx rings counters */
5332 if (ice_is_xdp_ena_vsi(vsi))
5333 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5340 * ice_update_vsi_stats - Update VSI stats counters
5341 * @vsi: the VSI to be updated
5343 void ice_update_vsi_stats(struct ice_vsi *vsi)
5345 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5346 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5347 struct ice_pf *pf = vsi->back;
5349 if (test_bit(__ICE_DOWN, vsi->state) ||
5350 test_bit(__ICE_CFG_BUSY, pf->state))
5353 /* get stats as recorded by Tx/Rx rings */
5354 ice_update_vsi_ring_stats(vsi);
5356 /* get VSI stats as recorded by the hardware */
5357 ice_update_eth_stats(vsi);
5359 cur_ns->tx_errors = cur_es->tx_errors;
5360 cur_ns->rx_dropped = cur_es->rx_discards + vsi->rx_gro_dropped;
5361 cur_ns->tx_dropped = cur_es->tx_discards;
5362 cur_ns->multicast = cur_es->rx_multicast;
5364 /* update some more netdev stats if this is main VSI */
5365 if (vsi->type == ICE_VSI_PF) {
5366 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5367 cur_ns->rx_errors = pf->stats.crc_errors +
5368 pf->stats.illegal_bytes +
5369 pf->stats.rx_len_errors +
5370 pf->stats.rx_undersize +
5371 pf->hw_csum_rx_error +
5372 pf->stats.rx_jabber +
5373 pf->stats.rx_fragments +
5374 pf->stats.rx_oversize;
5375 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5376 /* record drops from the port level */
5377 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5382 * ice_update_pf_stats - Update PF port stats counters
5383 * @pf: PF whose stats needs to be updated
5385 void ice_update_pf_stats(struct ice_pf *pf)
5387 struct ice_hw_port_stats *prev_ps, *cur_ps;
5388 struct ice_hw *hw = &pf->hw;
5392 port = hw->port_info->lport;
5393 prev_ps = &pf->stats_prev;
5394 cur_ps = &pf->stats;
5396 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5397 &prev_ps->eth.rx_bytes,
5398 &cur_ps->eth.rx_bytes);
5400 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5401 &prev_ps->eth.rx_unicast,
5402 &cur_ps->eth.rx_unicast);
5404 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5405 &prev_ps->eth.rx_multicast,
5406 &cur_ps->eth.rx_multicast);
5408 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5409 &prev_ps->eth.rx_broadcast,
5410 &cur_ps->eth.rx_broadcast);
5412 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5413 &prev_ps->eth.rx_discards,
5414 &cur_ps->eth.rx_discards);
5416 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5417 &prev_ps->eth.tx_bytes,
5418 &cur_ps->eth.tx_bytes);
5420 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5421 &prev_ps->eth.tx_unicast,
5422 &cur_ps->eth.tx_unicast);
5424 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5425 &prev_ps->eth.tx_multicast,
5426 &cur_ps->eth.tx_multicast);
5428 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5429 &prev_ps->eth.tx_broadcast,
5430 &cur_ps->eth.tx_broadcast);
5432 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5433 &prev_ps->tx_dropped_link_down,
5434 &cur_ps->tx_dropped_link_down);
5436 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5437 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5439 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5440 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5442 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5443 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5445 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5446 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5448 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5449 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5451 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5452 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5454 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5455 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5457 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5458 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5460 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5461 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5463 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5464 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5466 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5467 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5469 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5470 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5472 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5473 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5475 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5476 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5478 fd_ctr_base = hw->fd_ctr_base;
5480 ice_stat_update40(hw,
5481 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5482 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5483 &cur_ps->fd_sb_match);
5484 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5485 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5487 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5488 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5490 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5491 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5493 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5494 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5496 ice_update_dcb_stats(pf);
5498 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5499 &prev_ps->crc_errors, &cur_ps->crc_errors);
5501 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5502 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5504 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5505 &prev_ps->mac_local_faults,
5506 &cur_ps->mac_local_faults);
5508 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5509 &prev_ps->mac_remote_faults,
5510 &cur_ps->mac_remote_faults);
5512 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5513 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5515 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5516 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5518 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5519 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5521 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5522 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5524 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5525 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5527 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5529 pf->stat_prev_loaded = true;
5533 * ice_get_stats64 - get statistics for network device structure
5534 * @netdev: network interface device structure
5535 * @stats: main device statistics structure
5538 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5540 struct ice_netdev_priv *np = netdev_priv(netdev);
5541 struct rtnl_link_stats64 *vsi_stats;
5542 struct ice_vsi *vsi = np->vsi;
5544 vsi_stats = &vsi->net_stats;
5546 if (!vsi->num_txq || !vsi->num_rxq)
5549 /* netdev packet/byte stats come from ring counter. These are obtained
5550 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5551 * But, only call the update routine and read the registers if VSI is
5554 if (!test_bit(__ICE_DOWN, vsi->state))
5555 ice_update_vsi_ring_stats(vsi);
5556 stats->tx_packets = vsi_stats->tx_packets;
5557 stats->tx_bytes = vsi_stats->tx_bytes;
5558 stats->rx_packets = vsi_stats->rx_packets;
5559 stats->rx_bytes = vsi_stats->rx_bytes;
5561 /* The rest of the stats can be read from the hardware but instead we
5562 * just return values that the watchdog task has already obtained from
5565 stats->multicast = vsi_stats->multicast;
5566 stats->tx_errors = vsi_stats->tx_errors;
5567 stats->tx_dropped = vsi_stats->tx_dropped;
5568 stats->rx_errors = vsi_stats->rx_errors;
5569 stats->rx_dropped = vsi_stats->rx_dropped;
5570 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5571 stats->rx_length_errors = vsi_stats->rx_length_errors;
5575 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5576 * @vsi: VSI having NAPI disabled
5578 static void ice_napi_disable_all(struct ice_vsi *vsi)
5585 ice_for_each_q_vector(vsi, q_idx) {
5586 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5588 if (q_vector->rx.ring || q_vector->tx.ring)
5589 napi_disable(&q_vector->napi);
5594 * ice_down - Shutdown the connection
5595 * @vsi: The VSI being stopped
5597 int ice_down(struct ice_vsi *vsi)
5599 int i, tx_err, rx_err, link_err = 0;
5601 /* Caller of this function is expected to set the
5602 * vsi->state __ICE_DOWN bit
5605 netif_carrier_off(vsi->netdev);
5606 netif_tx_disable(vsi->netdev);
5609 ice_vsi_dis_irq(vsi);
5611 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5613 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5614 vsi->vsi_num, tx_err);
5615 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5616 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5618 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5619 vsi->vsi_num, tx_err);
5622 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5624 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5625 vsi->vsi_num, rx_err);
5627 ice_napi_disable_all(vsi);
5629 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5630 link_err = ice_force_phys_link_state(vsi, false);
5632 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5633 vsi->vsi_num, link_err);
5636 ice_for_each_txq(vsi, i)
5637 ice_clean_tx_ring(vsi->tx_rings[i]);
5639 ice_for_each_rxq(vsi, i)
5640 ice_clean_rx_ring(vsi->rx_rings[i]);
5642 if (tx_err || rx_err || link_err) {
5643 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5644 vsi->vsi_num, vsi->vsw->sw_id);
5652 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5653 * @vsi: VSI having resources allocated
5655 * Return 0 on success, negative on failure
5657 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5661 if (!vsi->num_txq) {
5662 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5667 ice_for_each_txq(vsi, i) {
5668 struct ice_ring *ring = vsi->tx_rings[i];
5673 ring->netdev = vsi->netdev;
5674 err = ice_setup_tx_ring(ring);
5683 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5684 * @vsi: VSI having resources allocated
5686 * Return 0 on success, negative on failure
5688 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5692 if (!vsi->num_rxq) {
5693 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5698 ice_for_each_rxq(vsi, i) {
5699 struct ice_ring *ring = vsi->rx_rings[i];
5704 ring->netdev = vsi->netdev;
5705 err = ice_setup_rx_ring(ring);
5714 * ice_vsi_open_ctrl - open control VSI for use
5715 * @vsi: the VSI to open
5717 * Initialization of the Control VSI
5719 * Returns 0 on success, negative value on error
5721 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5723 char int_name[ICE_INT_NAME_STR_LEN];
5724 struct ice_pf *pf = vsi->back;
5728 dev = ice_pf_to_dev(pf);
5729 /* allocate descriptors */
5730 err = ice_vsi_setup_tx_rings(vsi);
5734 err = ice_vsi_setup_rx_rings(vsi);
5738 err = ice_vsi_cfg(vsi);
5742 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5743 dev_driver_string(dev), dev_name(dev));
5744 err = ice_vsi_req_irq_msix(vsi, int_name);
5748 ice_vsi_cfg_msix(vsi);
5750 err = ice_vsi_start_all_rx_rings(vsi);
5752 goto err_up_complete;
5754 clear_bit(__ICE_DOWN, vsi->state);
5755 ice_vsi_ena_irq(vsi);
5762 ice_vsi_free_rx_rings(vsi);
5764 ice_vsi_free_tx_rings(vsi);
5770 * ice_vsi_open - Called when a network interface is made active
5771 * @vsi: the VSI to open
5773 * Initialization of the VSI
5775 * Returns 0 on success, negative value on error
5777 static int ice_vsi_open(struct ice_vsi *vsi)
5779 char int_name[ICE_INT_NAME_STR_LEN];
5780 struct ice_pf *pf = vsi->back;
5783 /* allocate descriptors */
5784 err = ice_vsi_setup_tx_rings(vsi);
5788 err = ice_vsi_setup_rx_rings(vsi);
5792 err = ice_vsi_cfg(vsi);
5796 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5797 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
5798 err = ice_vsi_req_irq_msix(vsi, int_name);
5802 /* Notify the stack of the actual queue counts. */
5803 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
5807 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
5811 err = ice_up_complete(vsi);
5813 goto err_up_complete;
5820 ice_vsi_free_irq(vsi);
5822 ice_vsi_free_rx_rings(vsi);
5824 ice_vsi_free_tx_rings(vsi);
5830 * ice_vsi_release_all - Delete all VSIs
5831 * @pf: PF from which all VSIs are being removed
5833 static void ice_vsi_release_all(struct ice_pf *pf)
5840 ice_for_each_vsi(pf, i) {
5844 err = ice_vsi_release(pf->vsi[i]);
5846 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
5847 i, err, pf->vsi[i]->vsi_num);
5852 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
5853 * @pf: pointer to the PF instance
5854 * @type: VSI type to rebuild
5856 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
5858 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
5860 struct device *dev = ice_pf_to_dev(pf);
5861 enum ice_status status;
5864 ice_for_each_vsi(pf, i) {
5865 struct ice_vsi *vsi = pf->vsi[i];
5867 if (!vsi || vsi->type != type)
5870 /* rebuild the VSI */
5871 err = ice_vsi_rebuild(vsi, true);
5873 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
5874 err, vsi->idx, ice_vsi_type_str(type));
5878 /* replay filters for the VSI */
5879 status = ice_replay_vsi(&pf->hw, vsi->idx);
5881 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
5882 ice_stat_str(status), vsi->idx,
5883 ice_vsi_type_str(type));
5887 /* Re-map HW VSI number, using VSI handle that has been
5888 * previously validated in ice_replay_vsi() call above
5890 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
5892 /* enable the VSI */
5893 err = ice_ena_vsi(vsi, false);
5895 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
5896 err, vsi->idx, ice_vsi_type_str(type));
5900 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
5901 ice_vsi_type_str(type));
5908 * ice_update_pf_netdev_link - Update PF netdev link status
5909 * @pf: pointer to the PF instance
5911 static void ice_update_pf_netdev_link(struct ice_pf *pf)
5916 ice_for_each_vsi(pf, i) {
5917 struct ice_vsi *vsi = pf->vsi[i];
5919 if (!vsi || vsi->type != ICE_VSI_PF)
5922 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
5924 netif_carrier_on(pf->vsi[i]->netdev);
5925 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
5927 netif_carrier_off(pf->vsi[i]->netdev);
5928 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
5934 * ice_rebuild - rebuild after reset
5935 * @pf: PF to rebuild
5936 * @reset_type: type of reset
5938 * Do not rebuild VF VSI in this flow because that is already handled via
5939 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
5940 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
5941 * to reset/rebuild all the VF VSI twice.
5943 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
5945 struct device *dev = ice_pf_to_dev(pf);
5946 struct ice_hw *hw = &pf->hw;
5947 enum ice_status ret;
5950 if (test_bit(__ICE_DOWN, pf->state))
5951 goto clear_recovery;
5953 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
5955 ret = ice_init_all_ctrlq(hw);
5957 dev_err(dev, "control queues init failed %s\n",
5959 goto err_init_ctrlq;
5962 /* if DDP was previously loaded successfully */
5963 if (!ice_is_safe_mode(pf)) {
5964 /* reload the SW DB of filter tables */
5965 if (reset_type == ICE_RESET_PFR)
5966 ice_fill_blk_tbls(hw);
5968 /* Reload DDP Package after CORER/GLOBR reset */
5969 ice_load_pkg(NULL, pf);
5972 ret = ice_clear_pf_cfg(hw);
5974 dev_err(dev, "clear PF configuration failed %s\n",
5976 goto err_init_ctrlq;
5979 if (pf->first_sw->dflt_vsi_ena)
5980 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
5981 /* clear the default VSI configuration if it exists */
5982 pf->first_sw->dflt_vsi = NULL;
5983 pf->first_sw->dflt_vsi_ena = false;
5985 ice_clear_pxe_mode(hw);
5987 ret = ice_get_caps(hw);
5989 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
5990 goto err_init_ctrlq;
5993 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
5995 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
5996 goto err_init_ctrlq;
5999 err = ice_sched_init_port(hw->port_info);
6001 goto err_sched_init_port;
6003 /* start misc vector */
6004 err = ice_req_irq_msix_misc(pf);
6006 dev_err(dev, "misc vector setup failed: %d\n", err);
6007 goto err_sched_init_port;
6010 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6011 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6012 if (!rd32(hw, PFQF_FD_SIZE)) {
6013 u16 unused, guar, b_effort;
6015 guar = hw->func_caps.fd_fltr_guar;
6016 b_effort = hw->func_caps.fd_fltr_best_effort;
6018 /* force guaranteed filter pool for PF */
6019 ice_alloc_fd_guar_item(hw, &unused, guar);
6020 /* force shared filter pool for PF */
6021 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6025 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6026 ice_dcb_rebuild(pf);
6028 /* rebuild PF VSI */
6029 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6031 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6032 goto err_vsi_rebuild;
6035 /* If Flow Director is active */
6036 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6037 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6039 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6040 goto err_vsi_rebuild;
6043 /* replay HW Flow Director recipes */
6045 ice_fdir_replay_flows(hw);
6047 /* replay Flow Director filters */
6048 ice_fdir_replay_fltrs(pf);
6050 ice_rebuild_arfs(pf);
6053 ice_update_pf_netdev_link(pf);
6055 /* tell the firmware we are up */
6056 ret = ice_send_version(pf);
6058 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6060 goto err_vsi_rebuild;
6063 ice_replay_post(hw);
6065 /* if we get here, reset flow is successful */
6066 clear_bit(__ICE_RESET_FAILED, pf->state);
6070 err_sched_init_port:
6071 ice_sched_cleanup_all(hw);
6073 ice_shutdown_all_ctrlq(hw);
6074 set_bit(__ICE_RESET_FAILED, pf->state);
6076 /* set this bit in PF state to control service task scheduling */
6077 set_bit(__ICE_NEEDS_RESTART, pf->state);
6078 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6082 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6083 * @vsi: Pointer to VSI structure
6085 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6087 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6088 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6090 return ICE_RXBUF_3072;
6094 * ice_change_mtu - NDO callback to change the MTU
6095 * @netdev: network interface device structure
6096 * @new_mtu: new value for maximum frame size
6098 * Returns 0 on success, negative on failure
6100 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6102 struct ice_netdev_priv *np = netdev_priv(netdev);
6103 struct ice_vsi *vsi = np->vsi;
6104 struct ice_pf *pf = vsi->back;
6107 if (new_mtu == (int)netdev->mtu) {
6108 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6112 if (ice_is_xdp_ena_vsi(vsi)) {
6113 int frame_size = ice_max_xdp_frame_size(vsi);
6115 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6116 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6117 frame_size - ICE_ETH_PKT_HDR_PAD);
6122 if (new_mtu < (int)netdev->min_mtu) {
6123 netdev_err(netdev, "new MTU invalid. min_mtu is %d\n",
6126 } else if (new_mtu > (int)netdev->max_mtu) {
6127 netdev_err(netdev, "new MTU invalid. max_mtu is %d\n",
6131 /* if a reset is in progress, wait for some time for it to complete */
6133 if (ice_is_reset_in_progress(pf->state)) {
6135 usleep_range(1000, 2000);
6140 } while (count < 100);
6143 netdev_err(netdev, "can't change MTU. Device is busy\n");
6147 netdev->mtu = (unsigned int)new_mtu;
6149 /* if VSI is up, bring it down and then back up */
6150 if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
6153 err = ice_down(vsi);
6155 netdev_err(netdev, "change MTU if_up err %d\n", err);
6161 netdev_err(netdev, "change MTU if_up err %d\n", err);
6166 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6171 * ice_aq_str - convert AQ err code to a string
6172 * @aq_err: the AQ error code to convert
6174 const char *ice_aq_str(enum ice_aq_err aq_err)
6179 case ICE_AQ_RC_EPERM:
6180 return "ICE_AQ_RC_EPERM";
6181 case ICE_AQ_RC_ENOENT:
6182 return "ICE_AQ_RC_ENOENT";
6183 case ICE_AQ_RC_ENOMEM:
6184 return "ICE_AQ_RC_ENOMEM";
6185 case ICE_AQ_RC_EBUSY:
6186 return "ICE_AQ_RC_EBUSY";
6187 case ICE_AQ_RC_EEXIST:
6188 return "ICE_AQ_RC_EEXIST";
6189 case ICE_AQ_RC_EINVAL:
6190 return "ICE_AQ_RC_EINVAL";
6191 case ICE_AQ_RC_ENOSPC:
6192 return "ICE_AQ_RC_ENOSPC";
6193 case ICE_AQ_RC_ENOSYS:
6194 return "ICE_AQ_RC_ENOSYS";
6195 case ICE_AQ_RC_EMODE:
6196 return "ICE_AQ_RC_EMODE";
6197 case ICE_AQ_RC_ENOSEC:
6198 return "ICE_AQ_RC_ENOSEC";
6199 case ICE_AQ_RC_EBADSIG:
6200 return "ICE_AQ_RC_EBADSIG";
6201 case ICE_AQ_RC_ESVN:
6202 return "ICE_AQ_RC_ESVN";
6203 case ICE_AQ_RC_EBADMAN:
6204 return "ICE_AQ_RC_EBADMAN";
6205 case ICE_AQ_RC_EBADBUF:
6206 return "ICE_AQ_RC_EBADBUF";
6209 return "ICE_AQ_RC_UNKNOWN";
6213 * ice_stat_str - convert status err code to a string
6214 * @stat_err: the status error code to convert
6216 const char *ice_stat_str(enum ice_status stat_err)
6222 return "ICE_ERR_PARAM";
6223 case ICE_ERR_NOT_IMPL:
6224 return "ICE_ERR_NOT_IMPL";
6225 case ICE_ERR_NOT_READY:
6226 return "ICE_ERR_NOT_READY";
6227 case ICE_ERR_NOT_SUPPORTED:
6228 return "ICE_ERR_NOT_SUPPORTED";
6229 case ICE_ERR_BAD_PTR:
6230 return "ICE_ERR_BAD_PTR";
6231 case ICE_ERR_INVAL_SIZE:
6232 return "ICE_ERR_INVAL_SIZE";
6233 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6234 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6235 case ICE_ERR_RESET_FAILED:
6236 return "ICE_ERR_RESET_FAILED";
6237 case ICE_ERR_FW_API_VER:
6238 return "ICE_ERR_FW_API_VER";
6239 case ICE_ERR_NO_MEMORY:
6240 return "ICE_ERR_NO_MEMORY";
6242 return "ICE_ERR_CFG";
6243 case ICE_ERR_OUT_OF_RANGE:
6244 return "ICE_ERR_OUT_OF_RANGE";
6245 case ICE_ERR_ALREADY_EXISTS:
6246 return "ICE_ERR_ALREADY_EXISTS";
6247 case ICE_ERR_NVM_CHECKSUM:
6248 return "ICE_ERR_NVM_CHECKSUM";
6249 case ICE_ERR_BUF_TOO_SHORT:
6250 return "ICE_ERR_BUF_TOO_SHORT";
6251 case ICE_ERR_NVM_BLANK_MODE:
6252 return "ICE_ERR_NVM_BLANK_MODE";
6253 case ICE_ERR_IN_USE:
6254 return "ICE_ERR_IN_USE";
6255 case ICE_ERR_MAX_LIMIT:
6256 return "ICE_ERR_MAX_LIMIT";
6257 case ICE_ERR_RESET_ONGOING:
6258 return "ICE_ERR_RESET_ONGOING";
6259 case ICE_ERR_HW_TABLE:
6260 return "ICE_ERR_HW_TABLE";
6261 case ICE_ERR_DOES_NOT_EXIST:
6262 return "ICE_ERR_DOES_NOT_EXIST";
6263 case ICE_ERR_FW_DDP_MISMATCH:
6264 return "ICE_ERR_FW_DDP_MISMATCH";
6265 case ICE_ERR_AQ_ERROR:
6266 return "ICE_ERR_AQ_ERROR";
6267 case ICE_ERR_AQ_TIMEOUT:
6268 return "ICE_ERR_AQ_TIMEOUT";
6269 case ICE_ERR_AQ_FULL:
6270 return "ICE_ERR_AQ_FULL";
6271 case ICE_ERR_AQ_NO_WORK:
6272 return "ICE_ERR_AQ_NO_WORK";
6273 case ICE_ERR_AQ_EMPTY:
6274 return "ICE_ERR_AQ_EMPTY";
6275 case ICE_ERR_AQ_FW_CRITICAL:
6276 return "ICE_ERR_AQ_FW_CRITICAL";
6279 return "ICE_ERR_UNKNOWN";
6283 * ice_set_rss - Set RSS keys and lut
6284 * @vsi: Pointer to VSI structure
6285 * @seed: RSS hash seed
6286 * @lut: Lookup table
6287 * @lut_size: Lookup table size
6289 * Returns 0 on success, negative on failure
6291 int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6293 struct ice_pf *pf = vsi->back;
6294 struct ice_hw *hw = &pf->hw;
6295 enum ice_status status;
6298 dev = ice_pf_to_dev(pf);
6300 struct ice_aqc_get_set_rss_keys *buf =
6301 (struct ice_aqc_get_set_rss_keys *)seed;
6303 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
6306 dev_err(dev, "Cannot set RSS key, err %s aq_err %s\n",
6307 ice_stat_str(status),
6308 ice_aq_str(hw->adminq.sq_last_status));
6314 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6317 dev_err(dev, "Cannot set RSS lut, err %s aq_err %s\n",
6318 ice_stat_str(status),
6319 ice_aq_str(hw->adminq.sq_last_status));
6328 * ice_get_rss - Get RSS keys and lut
6329 * @vsi: Pointer to VSI structure
6330 * @seed: Buffer to store the keys
6331 * @lut: Buffer to store the lookup table entries
6332 * @lut_size: Size of buffer to store the lookup table entries
6334 * Returns 0 on success, negative on failure
6336 int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6338 struct ice_pf *pf = vsi->back;
6339 struct ice_hw *hw = &pf->hw;
6340 enum ice_status status;
6343 dev = ice_pf_to_dev(pf);
6345 struct ice_aqc_get_set_rss_keys *buf =
6346 (struct ice_aqc_get_set_rss_keys *)seed;
6348 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
6350 dev_err(dev, "Cannot get RSS key, err %s aq_err %s\n",
6351 ice_stat_str(status),
6352 ice_aq_str(hw->adminq.sq_last_status));
6358 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6361 dev_err(dev, "Cannot get RSS lut, err %s aq_err %s\n",
6362 ice_stat_str(status),
6363 ice_aq_str(hw->adminq.sq_last_status));
6372 * ice_bridge_getlink - Get the hardware bridge mode
6375 * @seq: RTNL message seq
6376 * @dev: the netdev being configured
6377 * @filter_mask: filter mask passed in
6378 * @nlflags: netlink flags passed in
6380 * Return the bridge mode (VEB/VEPA)
6383 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6384 struct net_device *dev, u32 filter_mask, int nlflags)
6386 struct ice_netdev_priv *np = netdev_priv(dev);
6387 struct ice_vsi *vsi = np->vsi;
6388 struct ice_pf *pf = vsi->back;
6391 bmode = pf->first_sw->bridge_mode;
6393 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6398 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6399 * @vsi: Pointer to VSI structure
6400 * @bmode: Hardware bridge mode (VEB/VEPA)
6402 * Returns 0 on success, negative on failure
6404 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6406 struct ice_aqc_vsi_props *vsi_props;
6407 struct ice_hw *hw = &vsi->back->hw;
6408 struct ice_vsi_ctx *ctxt;
6409 enum ice_status status;
6412 vsi_props = &vsi->info;
6414 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6418 ctxt->info = vsi->info;
6420 if (bmode == BRIDGE_MODE_VEB)
6421 /* change from VEPA to VEB mode */
6422 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6424 /* change from VEB to VEPA mode */
6425 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6426 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6428 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6430 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6431 bmode, ice_stat_str(status),
6432 ice_aq_str(hw->adminq.sq_last_status));
6436 /* Update sw flags for book keeping */
6437 vsi_props->sw_flags = ctxt->info.sw_flags;
6445 * ice_bridge_setlink - Set the hardware bridge mode
6446 * @dev: the netdev being configured
6447 * @nlh: RTNL message
6448 * @flags: bridge setlink flags
6449 * @extack: netlink extended ack
6451 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6452 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6453 * not already set for all VSIs connected to this switch. And also update the
6454 * unicast switch filter rules for the corresponding switch of the netdev.
6457 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6458 u16 __always_unused flags,
6459 struct netlink_ext_ack __always_unused *extack)
6461 struct ice_netdev_priv *np = netdev_priv(dev);
6462 struct ice_pf *pf = np->vsi->back;
6463 struct nlattr *attr, *br_spec;
6464 struct ice_hw *hw = &pf->hw;
6465 enum ice_status status;
6466 struct ice_sw *pf_sw;
6467 int rem, v, err = 0;
6469 pf_sw = pf->first_sw;
6470 /* find the attribute in the netlink message */
6471 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6473 nla_for_each_nested(attr, br_spec, rem) {
6476 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6478 mode = nla_get_u16(attr);
6479 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6481 /* Continue if bridge mode is not being flipped */
6482 if (mode == pf_sw->bridge_mode)
6484 /* Iterates through the PF VSI list and update the loopback
6487 ice_for_each_vsi(pf, v) {
6490 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6495 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6496 /* Update the unicast switch filter rules for the corresponding
6497 * switch of the netdev
6499 status = ice_update_sw_rule_bridge_mode(hw);
6501 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6502 mode, ice_stat_str(status),
6503 ice_aq_str(hw->adminq.sq_last_status));
6504 /* revert hw->evb_veb */
6505 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6509 pf_sw->bridge_mode = mode;
6516 * ice_tx_timeout - Respond to a Tx Hang
6517 * @netdev: network interface device structure
6518 * @txqueue: Tx queue
6520 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6522 struct ice_netdev_priv *np = netdev_priv(netdev);
6523 struct ice_ring *tx_ring = NULL;
6524 struct ice_vsi *vsi = np->vsi;
6525 struct ice_pf *pf = vsi->back;
6528 pf->tx_timeout_count++;
6530 /* Check if PFC is enabled for the TC to which the queue belongs
6531 * to. If yes then Tx timeout is not caused by a hung queue, no
6532 * need to reset and rebuild
6534 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6535 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6540 /* now that we have an index, find the tx_ring struct */
6541 for (i = 0; i < vsi->num_txq; i++)
6542 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6543 if (txqueue == vsi->tx_rings[i]->q_index) {
6544 tx_ring = vsi->tx_rings[i];
6548 /* Reset recovery level if enough time has elapsed after last timeout.
6549 * Also ensure no new reset action happens before next timeout period.
6551 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6552 pf->tx_timeout_recovery_level = 1;
6553 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6554 netdev->watchdog_timeo)))
6558 struct ice_hw *hw = &pf->hw;
6561 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6562 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6563 /* Read interrupt register */
6564 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6566 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6567 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6568 head, tx_ring->next_to_use, val);
6571 pf->tx_timeout_last_recovery = jiffies;
6572 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6573 pf->tx_timeout_recovery_level, txqueue);
6575 switch (pf->tx_timeout_recovery_level) {
6577 set_bit(__ICE_PFR_REQ, pf->state);
6580 set_bit(__ICE_CORER_REQ, pf->state);
6583 set_bit(__ICE_GLOBR_REQ, pf->state);
6586 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6587 set_bit(__ICE_DOWN, pf->state);
6588 set_bit(__ICE_NEEDS_RESTART, vsi->state);
6589 set_bit(__ICE_SERVICE_DIS, pf->state);
6593 ice_service_task_schedule(pf);
6594 pf->tx_timeout_recovery_level++;
6598 * ice_open - Called when a network interface becomes active
6599 * @netdev: network interface device structure
6601 * The open entry point is called when a network interface is made
6602 * active by the system (IFF_UP). At this point all resources needed
6603 * for transmit and receive operations are allocated, the interrupt
6604 * handler is registered with the OS, the netdev watchdog is enabled,
6605 * and the stack is notified that the interface is ready.
6607 * Returns 0 on success, negative value on failure
6609 int ice_open(struct net_device *netdev)
6611 struct ice_netdev_priv *np = netdev_priv(netdev);
6612 struct ice_vsi *vsi = np->vsi;
6613 struct ice_pf *pf = vsi->back;
6614 struct ice_port_info *pi;
6617 if (test_bit(__ICE_NEEDS_RESTART, pf->state)) {
6618 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6622 if (test_bit(__ICE_DOWN, pf->state)) {
6623 netdev_err(netdev, "device is not ready yet\n");
6627 netif_carrier_off(netdev);
6629 pi = vsi->port_info;
6630 err = ice_update_link_info(pi);
6632 netdev_err(netdev, "Failed to get link info, error %d\n",
6637 /* Set PHY if there is media, otherwise, turn off PHY */
6638 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6639 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6640 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state)) {
6641 err = ice_init_phy_user_cfg(pi);
6643 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6649 err = ice_configure_phy(vsi);
6651 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6656 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6657 err = ice_aq_set_link_restart_an(pi, false, NULL);
6659 netdev_err(netdev, "Failed to set PHY state, VSI %d error %d\n",
6665 err = ice_vsi_open(vsi);
6667 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6668 vsi->vsi_num, vsi->vsw->sw_id);
6670 /* Update existing tunnels information */
6671 udp_tunnel_get_rx_info(netdev);
6677 * ice_stop - Disables a network interface
6678 * @netdev: network interface device structure
6680 * The stop entry point is called when an interface is de-activated by the OS,
6681 * and the netdevice enters the DOWN state. The hardware is still under the
6682 * driver's control, but the netdev interface is disabled.
6684 * Returns success only - not allowed to fail
6686 int ice_stop(struct net_device *netdev)
6688 struct ice_netdev_priv *np = netdev_priv(netdev);
6689 struct ice_vsi *vsi = np->vsi;
6697 * ice_features_check - Validate encapsulated packet conforms to limits
6699 * @netdev: This port's netdev
6700 * @features: Offload features that the stack believes apply
6702 static netdev_features_t
6703 ice_features_check(struct sk_buff *skb,
6704 struct net_device __always_unused *netdev,
6705 netdev_features_t features)
6709 /* No point in doing any of this if neither checksum nor GSO are
6710 * being requested for this frame. We can rule out both by just
6711 * checking for CHECKSUM_PARTIAL
6713 if (skb->ip_summed != CHECKSUM_PARTIAL)
6716 /* We cannot support GSO if the MSS is going to be less than
6717 * 64 bytes. If it is then we need to drop support for GSO.
6719 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
6720 features &= ~NETIF_F_GSO_MASK;
6722 len = skb_network_header(skb) - skb->data;
6723 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
6724 goto out_rm_features;
6726 len = skb_transport_header(skb) - skb_network_header(skb);
6727 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6728 goto out_rm_features;
6730 if (skb->encapsulation) {
6731 len = skb_inner_network_header(skb) - skb_transport_header(skb);
6732 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
6733 goto out_rm_features;
6735 len = skb_inner_transport_header(skb) -
6736 skb_inner_network_header(skb);
6737 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6738 goto out_rm_features;
6743 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
6746 static const struct net_device_ops ice_netdev_safe_mode_ops = {
6747 .ndo_open = ice_open,
6748 .ndo_stop = ice_stop,
6749 .ndo_start_xmit = ice_start_xmit,
6750 .ndo_set_mac_address = ice_set_mac_address,
6751 .ndo_validate_addr = eth_validate_addr,
6752 .ndo_change_mtu = ice_change_mtu,
6753 .ndo_get_stats64 = ice_get_stats64,
6754 .ndo_tx_timeout = ice_tx_timeout,
6757 static const struct net_device_ops ice_netdev_ops = {
6758 .ndo_open = ice_open,
6759 .ndo_stop = ice_stop,
6760 .ndo_start_xmit = ice_start_xmit,
6761 .ndo_features_check = ice_features_check,
6762 .ndo_set_rx_mode = ice_set_rx_mode,
6763 .ndo_set_mac_address = ice_set_mac_address,
6764 .ndo_validate_addr = eth_validate_addr,
6765 .ndo_change_mtu = ice_change_mtu,
6766 .ndo_get_stats64 = ice_get_stats64,
6767 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
6768 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
6769 .ndo_set_vf_mac = ice_set_vf_mac,
6770 .ndo_get_vf_config = ice_get_vf_cfg,
6771 .ndo_set_vf_trust = ice_set_vf_trust,
6772 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
6773 .ndo_set_vf_link_state = ice_set_vf_link_state,
6774 .ndo_get_vf_stats = ice_get_vf_stats,
6775 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
6776 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
6777 .ndo_set_features = ice_set_features,
6778 .ndo_bridge_getlink = ice_bridge_getlink,
6779 .ndo_bridge_setlink = ice_bridge_setlink,
6780 .ndo_fdb_add = ice_fdb_add,
6781 .ndo_fdb_del = ice_fdb_del,
6782 #ifdef CONFIG_RFS_ACCEL
6783 .ndo_rx_flow_steer = ice_rx_flow_steer,
6785 .ndo_tx_timeout = ice_tx_timeout,
6787 .ndo_xdp_xmit = ice_xdp_xmit,
6788 .ndo_xsk_wakeup = ice_xsk_wakeup,
6789 .ndo_udp_tunnel_add = udp_tunnel_nic_add_port,
6790 .ndo_udp_tunnel_del = udp_tunnel_nic_del_port,