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)
227 if (vsi->num_vlan > 1) {
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) {
329 if (vsi->num_vlan > 1)
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) */
343 if (vsi->num_vlan > 1)
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 device *dev = ice_pf_to_dev(pf);
1060 struct ice_aq_task *task;
1061 unsigned long start;
1065 task = kzalloc(sizeof(*task), GFP_KERNEL);
1069 INIT_HLIST_NODE(&task->entry);
1070 task->opcode = opcode;
1071 task->event = event;
1072 task->state = ICE_AQ_TASK_WAITING;
1074 spin_lock_bh(&pf->aq_wait_lock);
1075 hlist_add_head(&task->entry, &pf->aq_wait_list);
1076 spin_unlock_bh(&pf->aq_wait_lock);
1080 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1082 switch (task->state) {
1083 case ICE_AQ_TASK_WAITING:
1084 err = ret < 0 ? ret : -ETIMEDOUT;
1086 case ICE_AQ_TASK_CANCELED:
1087 err = ret < 0 ? ret : -ECANCELED;
1089 case ICE_AQ_TASK_COMPLETE:
1090 err = ret < 0 ? ret : 0;
1093 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1098 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1099 jiffies_to_msecs(jiffies - start),
1100 jiffies_to_msecs(timeout),
1103 spin_lock_bh(&pf->aq_wait_lock);
1104 hlist_del(&task->entry);
1105 spin_unlock_bh(&pf->aq_wait_lock);
1112 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1113 * @pf: pointer to the PF private structure
1114 * @opcode: the opcode of the event
1115 * @event: the event to check
1117 * Loops over the current list of pending threads waiting for an AdminQ event.
1118 * For each matching task, copy the contents of the event into the task
1119 * structure and wake up the thread.
1121 * If multiple threads wait for the same opcode, they will all be woken up.
1123 * Note that event->msg_buf will only be duplicated if the event has a buffer
1124 * with enough space already allocated. Otherwise, only the descriptor and
1125 * message length will be copied.
1127 * Returns: true if an event was found, false otherwise
1129 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1130 struct ice_rq_event_info *event)
1132 struct ice_aq_task *task;
1135 spin_lock_bh(&pf->aq_wait_lock);
1136 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1137 if (task->state || task->opcode != opcode)
1140 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1141 task->event->msg_len = event->msg_len;
1143 /* Only copy the data buffer if a destination was set */
1144 if (task->event->msg_buf &&
1145 task->event->buf_len > event->buf_len) {
1146 memcpy(task->event->msg_buf, event->msg_buf,
1148 task->event->buf_len = event->buf_len;
1151 task->state = ICE_AQ_TASK_COMPLETE;
1154 spin_unlock_bh(&pf->aq_wait_lock);
1157 wake_up(&pf->aq_wait_queue);
1161 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1162 * @pf: the PF private structure
1164 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1165 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1167 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1169 struct ice_aq_task *task;
1171 spin_lock_bh(&pf->aq_wait_lock);
1172 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1173 task->state = ICE_AQ_TASK_CANCELED;
1174 spin_unlock_bh(&pf->aq_wait_lock);
1176 wake_up(&pf->aq_wait_queue);
1180 * __ice_clean_ctrlq - helper function to clean controlq rings
1181 * @pf: ptr to struct ice_pf
1182 * @q_type: specific Control queue type
1184 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1186 struct device *dev = ice_pf_to_dev(pf);
1187 struct ice_rq_event_info event;
1188 struct ice_hw *hw = &pf->hw;
1189 struct ice_ctl_q_info *cq;
1194 /* Do not clean control queue if/when PF reset fails */
1195 if (test_bit(__ICE_RESET_FAILED, pf->state))
1199 case ICE_CTL_Q_ADMIN:
1203 case ICE_CTL_Q_MAILBOX:
1208 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1212 /* check for error indications - PF_xx_AxQLEN register layout for
1213 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1215 val = rd32(hw, cq->rq.len);
1216 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1217 PF_FW_ARQLEN_ARQCRIT_M)) {
1219 if (val & PF_FW_ARQLEN_ARQVFE_M)
1220 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1222 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1223 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1226 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1227 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1229 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1230 PF_FW_ARQLEN_ARQCRIT_M);
1232 wr32(hw, cq->rq.len, val);
1235 val = rd32(hw, cq->sq.len);
1236 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1237 PF_FW_ATQLEN_ATQCRIT_M)) {
1239 if (val & PF_FW_ATQLEN_ATQVFE_M)
1240 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1242 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1243 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1246 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1247 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1249 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1250 PF_FW_ATQLEN_ATQCRIT_M);
1252 wr32(hw, cq->sq.len, val);
1255 event.buf_len = cq->rq_buf_size;
1256 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1261 enum ice_status ret;
1264 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1265 if (ret == ICE_ERR_AQ_NO_WORK)
1268 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1273 opcode = le16_to_cpu(event.desc.opcode);
1275 /* Notify any thread that might be waiting for this event */
1276 ice_aq_check_events(pf, opcode, &event);
1279 case ice_aqc_opc_get_link_status:
1280 if (ice_handle_link_event(pf, &event))
1281 dev_err(dev, "Could not handle link event\n");
1283 case ice_aqc_opc_event_lan_overflow:
1284 ice_vf_lan_overflow_event(pf, &event);
1286 case ice_mbx_opc_send_msg_to_pf:
1287 ice_vc_process_vf_msg(pf, &event);
1289 case ice_aqc_opc_fw_logging:
1290 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1292 case ice_aqc_opc_lldp_set_mib_change:
1293 ice_dcb_process_lldp_set_mib_change(pf, &event);
1296 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1300 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1302 kfree(event.msg_buf);
1304 return pending && (i == ICE_DFLT_IRQ_WORK);
1308 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1309 * @hw: pointer to hardware info
1310 * @cq: control queue information
1312 * returns true if there are pending messages in a queue, false if there aren't
1314 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1318 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1319 return cq->rq.next_to_clean != ntu;
1323 * ice_clean_adminq_subtask - clean the AdminQ rings
1324 * @pf: board private structure
1326 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1328 struct ice_hw *hw = &pf->hw;
1330 if (!test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
1333 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1336 clear_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
1338 /* There might be a situation where new messages arrive to a control
1339 * queue between processing the last message and clearing the
1340 * EVENT_PENDING bit. So before exiting, check queue head again (using
1341 * ice_ctrlq_pending) and process new messages if any.
1343 if (ice_ctrlq_pending(hw, &hw->adminq))
1344 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1350 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1351 * @pf: board private structure
1353 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1355 struct ice_hw *hw = &pf->hw;
1357 if (!test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1360 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1363 clear_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1365 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1366 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1372 * ice_service_task_schedule - schedule the service task to wake up
1373 * @pf: board private structure
1375 * If not already scheduled, this puts the task into the work queue.
1377 void ice_service_task_schedule(struct ice_pf *pf)
1379 if (!test_bit(__ICE_SERVICE_DIS, pf->state) &&
1380 !test_and_set_bit(__ICE_SERVICE_SCHED, pf->state) &&
1381 !test_bit(__ICE_NEEDS_RESTART, pf->state))
1382 queue_work(ice_wq, &pf->serv_task);
1386 * ice_service_task_complete - finish up the service task
1387 * @pf: board private structure
1389 static void ice_service_task_complete(struct ice_pf *pf)
1391 WARN_ON(!test_bit(__ICE_SERVICE_SCHED, pf->state));
1393 /* force memory (pf->state) to sync before next service task */
1394 smp_mb__before_atomic();
1395 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1399 * ice_service_task_stop - stop service task and cancel works
1400 * @pf: board private structure
1402 * Return 0 if the __ICE_SERVICE_DIS bit was not already set,
1405 static int ice_service_task_stop(struct ice_pf *pf)
1409 ret = test_and_set_bit(__ICE_SERVICE_DIS, pf->state);
1411 if (pf->serv_tmr.function)
1412 del_timer_sync(&pf->serv_tmr);
1413 if (pf->serv_task.func)
1414 cancel_work_sync(&pf->serv_task);
1416 clear_bit(__ICE_SERVICE_SCHED, pf->state);
1421 * ice_service_task_restart - restart service task and schedule works
1422 * @pf: board private structure
1424 * This function is needed for suspend and resume works (e.g WoL scenario)
1426 static void ice_service_task_restart(struct ice_pf *pf)
1428 clear_bit(__ICE_SERVICE_DIS, pf->state);
1429 ice_service_task_schedule(pf);
1433 * ice_service_timer - timer callback to schedule service task
1434 * @t: pointer to timer_list
1436 static void ice_service_timer(struct timer_list *t)
1438 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1440 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1441 ice_service_task_schedule(pf);
1445 * ice_handle_mdd_event - handle malicious driver detect event
1446 * @pf: pointer to the PF structure
1448 * Called from service task. OICR interrupt handler indicates MDD event.
1449 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1450 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1451 * disable the queue, the PF can be configured to reset the VF using ethtool
1452 * private flag mdd-auto-reset-vf.
1454 static void ice_handle_mdd_event(struct ice_pf *pf)
1456 struct device *dev = ice_pf_to_dev(pf);
1457 struct ice_hw *hw = &pf->hw;
1461 if (!test_and_clear_bit(__ICE_MDD_EVENT_PENDING, pf->state)) {
1462 /* Since the VF MDD event logging is rate limited, check if
1463 * there are pending MDD events.
1465 ice_print_vfs_mdd_events(pf);
1469 /* find what triggered an MDD event */
1470 reg = rd32(hw, GL_MDET_TX_PQM);
1471 if (reg & GL_MDET_TX_PQM_VALID_M) {
1472 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1473 GL_MDET_TX_PQM_PF_NUM_S;
1474 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1475 GL_MDET_TX_PQM_VF_NUM_S;
1476 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1477 GL_MDET_TX_PQM_MAL_TYPE_S;
1478 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1479 GL_MDET_TX_PQM_QNUM_S);
1481 if (netif_msg_tx_err(pf))
1482 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1483 event, queue, pf_num, vf_num);
1484 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1487 reg = rd32(hw, GL_MDET_TX_TCLAN);
1488 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1489 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1490 GL_MDET_TX_TCLAN_PF_NUM_S;
1491 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1492 GL_MDET_TX_TCLAN_VF_NUM_S;
1493 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1494 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1495 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1496 GL_MDET_TX_TCLAN_QNUM_S);
1498 if (netif_msg_tx_err(pf))
1499 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1500 event, queue, pf_num, vf_num);
1501 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1504 reg = rd32(hw, GL_MDET_RX);
1505 if (reg & GL_MDET_RX_VALID_M) {
1506 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1507 GL_MDET_RX_PF_NUM_S;
1508 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1509 GL_MDET_RX_VF_NUM_S;
1510 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1511 GL_MDET_RX_MAL_TYPE_S;
1512 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1515 if (netif_msg_rx_err(pf))
1516 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1517 event, queue, pf_num, vf_num);
1518 wr32(hw, GL_MDET_RX, 0xffffffff);
1521 /* check to see if this PF caused an MDD event */
1522 reg = rd32(hw, PF_MDET_TX_PQM);
1523 if (reg & PF_MDET_TX_PQM_VALID_M) {
1524 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1525 if (netif_msg_tx_err(pf))
1526 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1529 reg = rd32(hw, PF_MDET_TX_TCLAN);
1530 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1531 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1532 if (netif_msg_tx_err(pf))
1533 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1536 reg = rd32(hw, PF_MDET_RX);
1537 if (reg & PF_MDET_RX_VALID_M) {
1538 wr32(hw, PF_MDET_RX, 0xFFFF);
1539 if (netif_msg_rx_err(pf))
1540 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1543 /* Check to see if one of the VFs caused an MDD event, and then
1544 * increment counters and set print pending
1546 ice_for_each_vf(pf, i) {
1547 struct ice_vf *vf = &pf->vf[i];
1549 reg = rd32(hw, VP_MDET_TX_PQM(i));
1550 if (reg & VP_MDET_TX_PQM_VALID_M) {
1551 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1552 vf->mdd_tx_events.count++;
1553 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1554 if (netif_msg_tx_err(pf))
1555 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1559 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1560 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1561 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1562 vf->mdd_tx_events.count++;
1563 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1564 if (netif_msg_tx_err(pf))
1565 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1569 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1570 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1571 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1572 vf->mdd_tx_events.count++;
1573 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1574 if (netif_msg_tx_err(pf))
1575 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1579 reg = rd32(hw, VP_MDET_RX(i));
1580 if (reg & VP_MDET_RX_VALID_M) {
1581 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1582 vf->mdd_rx_events.count++;
1583 set_bit(__ICE_MDD_VF_PRINT_PENDING, pf->state);
1584 if (netif_msg_rx_err(pf))
1585 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1588 /* Since the queue is disabled on VF Rx MDD events, the
1589 * PF can be configured to reset the VF through ethtool
1590 * private flag mdd-auto-reset-vf.
1592 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1593 /* VF MDD event counters will be cleared by
1594 * reset, so print the event prior to reset.
1596 ice_print_vf_rx_mdd_event(vf);
1597 ice_reset_vf(&pf->vf[i], false);
1602 ice_print_vfs_mdd_events(pf);
1606 * ice_force_phys_link_state - Force the physical link state
1607 * @vsi: VSI to force the physical link state to up/down
1608 * @link_up: true/false indicates to set the physical link to up/down
1610 * Force the physical link state by getting the current PHY capabilities from
1611 * hardware and setting the PHY config based on the determined capabilities. If
1612 * link changes a link event will be triggered because both the Enable Automatic
1613 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1615 * Returns 0 on success, negative on failure
1617 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1619 struct ice_aqc_get_phy_caps_data *pcaps;
1620 struct ice_aqc_set_phy_cfg_data *cfg;
1621 struct ice_port_info *pi;
1625 if (!vsi || !vsi->port_info || !vsi->back)
1627 if (vsi->type != ICE_VSI_PF)
1630 dev = ice_pf_to_dev(vsi->back);
1632 pi = vsi->port_info;
1634 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1638 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
1641 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1642 vsi->vsi_num, retcode);
1647 /* No change in link */
1648 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1649 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1652 /* Use the current user PHY configuration. The current user PHY
1653 * configuration is initialized during probe from PHY capabilities
1654 * software mode, and updated on set PHY configuration.
1656 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1662 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1664 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1666 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1668 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1670 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1671 vsi->vsi_num, retcode);
1682 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1683 * @pi: port info structure
1685 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1687 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1689 struct ice_aqc_get_phy_caps_data *pcaps;
1690 struct ice_pf *pf = pi->hw->back;
1691 enum ice_status status;
1694 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1698 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_NVM_CAP, pcaps,
1702 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1707 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1708 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1716 * ice_init_link_dflt_override - Initialize link default override
1717 * @pi: port info structure
1719 * Initialize link default override and PHY total port shutdown during probe
1721 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1723 struct ice_link_default_override_tlv *ldo;
1724 struct ice_pf *pf = pi->hw->back;
1726 ldo = &pf->link_dflt_override;
1727 if (ice_get_link_default_override(ldo, pi))
1730 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1733 /* Enable Total Port Shutdown (override/replace link-down-on-close
1734 * ethtool private flag) for ports with Port Disable bit set.
1736 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1737 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1741 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1742 * @pi: port info structure
1744 * If default override is enabled, initialized the user PHY cfg speed and FEC
1745 * settings using the default override mask from the NVM.
1747 * The PHY should only be configured with the default override settings the
1748 * first time media is available. The __ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1749 * is used to indicate that the user PHY cfg default override is initialized
1750 * and the PHY has not been configured with the default override settings. The
1751 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1754 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1756 struct ice_link_default_override_tlv *ldo;
1757 struct ice_aqc_set_phy_cfg_data *cfg;
1758 struct ice_phy_info *phy = &pi->phy;
1759 struct ice_pf *pf = pi->hw->back;
1761 ldo = &pf->link_dflt_override;
1763 /* If link default override is enabled, use to mask NVM PHY capabilities
1764 * for speed and FEC default configuration.
1766 cfg = &phy->curr_user_phy_cfg;
1768 if (ldo->phy_type_low || ldo->phy_type_high) {
1769 cfg->phy_type_low = pf->nvm_phy_type_lo &
1770 cpu_to_le64(ldo->phy_type_low);
1771 cfg->phy_type_high = pf->nvm_phy_type_hi &
1772 cpu_to_le64(ldo->phy_type_high);
1774 cfg->link_fec_opt = ldo->fec_options;
1775 phy->curr_user_fec_req = ICE_FEC_AUTO;
1777 set_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1781 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1782 * @pi: port info structure
1784 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1785 * mode to default. The PHY defaults are from get PHY capabilities topology
1786 * with media so call when media is first available. An error is returned if
1787 * called when media is not available. The PHY initialization completed state is
1790 * These configurations are used when setting PHY
1791 * configuration. The user PHY configuration is updated on set PHY
1792 * configuration. Returns 0 on success, negative on failure
1794 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1796 struct ice_aqc_get_phy_caps_data *pcaps;
1797 struct ice_phy_info *phy = &pi->phy;
1798 struct ice_pf *pf = pi->hw->back;
1799 enum ice_status status;
1800 struct ice_vsi *vsi;
1803 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1806 vsi = ice_get_main_vsi(pf);
1810 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1814 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1817 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1822 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1824 /* check if lenient mode is supported and enabled */
1825 if (ice_fw_supports_link_override(&vsi->back->hw) &&
1826 !(pcaps->module_compliance_enforcement &
1827 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1828 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1830 /* if link default override is enabled, initialize user PHY
1831 * configuration with link default override values
1833 if (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN) {
1834 ice_init_phy_cfg_dflt_override(pi);
1839 /* if link default override is not enabled, initialize PHY using
1840 * topology with media
1842 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1843 pcaps->link_fec_options);
1844 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1847 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1848 set_bit(__ICE_PHY_INIT_COMPLETE, pf->state);
1855 * ice_configure_phy - configure PHY
1858 * Set the PHY configuration. If the current PHY configuration is the same as
1859 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1860 * configure the based get PHY capabilities for topology with media.
1862 static int ice_configure_phy(struct ice_vsi *vsi)
1864 struct device *dev = ice_pf_to_dev(vsi->back);
1865 struct ice_aqc_get_phy_caps_data *pcaps;
1866 struct ice_aqc_set_phy_cfg_data *cfg;
1867 struct ice_port_info *pi;
1868 enum ice_status status;
1871 pi = vsi->port_info;
1875 /* Ensure we have media as we cannot configure a medialess port */
1876 if (!(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1879 ice_print_topo_conflict(vsi);
1881 if (vsi->port_info->phy.link_info.topo_media_conflict ==
1882 ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1885 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
1886 return ice_force_phys_link_state(vsi, true);
1888 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1892 /* Get current PHY config */
1893 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_SW_CFG, pcaps,
1896 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1897 vsi->vsi_num, ice_stat_str(status));
1902 /* If PHY enable link is configured and configuration has not changed,
1903 * there's nothing to do
1905 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1906 ice_phy_caps_equals_cfg(pcaps, &pi->phy.curr_user_phy_cfg))
1909 /* Use PHY topology as baseline for configuration */
1910 memset(pcaps, 0, sizeof(*pcaps));
1911 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP, pcaps,
1914 dev_err(dev, "Failed to get PHY topology, VSI %d error %s\n",
1915 vsi->vsi_num, ice_stat_str(status));
1920 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
1926 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
1928 /* Speed - If default override pending, use curr_user_phy_cfg set in
1929 * ice_init_phy_user_cfg_ldo.
1931 if (test_and_clear_bit(__ICE_LINK_DEFAULT_OVERRIDE_PENDING,
1932 vsi->back->state)) {
1933 cfg->phy_type_low = pi->phy.curr_user_phy_cfg.phy_type_low;
1934 cfg->phy_type_high = pi->phy.curr_user_phy_cfg.phy_type_high;
1936 u64 phy_low = 0, phy_high = 0;
1938 ice_update_phy_type(&phy_low, &phy_high,
1939 pi->phy.curr_user_speed_req);
1940 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
1941 cfg->phy_type_high = pcaps->phy_type_high &
1942 cpu_to_le64(phy_high);
1945 /* Can't provide what was requested; use PHY capabilities */
1946 if (!cfg->phy_type_low && !cfg->phy_type_high) {
1947 cfg->phy_type_low = pcaps->phy_type_low;
1948 cfg->phy_type_high = pcaps->phy_type_high;
1952 ice_cfg_phy_fec(pi, cfg, pi->phy.curr_user_fec_req);
1954 /* Can't provide what was requested; use PHY capabilities */
1955 if (cfg->link_fec_opt !=
1956 (cfg->link_fec_opt & pcaps->link_fec_options)) {
1957 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
1958 cfg->link_fec_opt = pcaps->link_fec_options;
1961 /* Flow Control - always supported; no need to check against
1964 ice_cfg_phy_fc(pi, cfg, pi->phy.curr_user_fc_req);
1966 /* Enable link and link update */
1967 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
1969 status = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1971 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
1972 vsi->vsi_num, ice_stat_str(status));
1983 * ice_check_media_subtask - Check for media
1984 * @pf: pointer to PF struct
1986 * If media is available, then initialize PHY user configuration if it is not
1987 * been, and configure the PHY if the interface is up.
1989 static void ice_check_media_subtask(struct ice_pf *pf)
1991 struct ice_port_info *pi;
1992 struct ice_vsi *vsi;
1995 /* No need to check for media if it's already present */
1996 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
1999 vsi = ice_get_main_vsi(pf);
2003 /* Refresh link info and check if media is present */
2004 pi = vsi->port_info;
2005 err = ice_update_link_info(pi);
2009 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2010 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state))
2011 ice_init_phy_user_cfg(pi);
2013 /* PHY settings are reset on media insertion, reconfigure
2014 * PHY to preserve settings.
2016 if (test_bit(__ICE_DOWN, vsi->state) &&
2017 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2020 err = ice_configure_phy(vsi);
2022 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2024 /* A Link Status Event will be generated; the event handler
2025 * will complete bringing the interface up
2031 * ice_service_task - manage and run subtasks
2032 * @work: pointer to work_struct contained by the PF struct
2034 static void ice_service_task(struct work_struct *work)
2036 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2037 unsigned long start_time = jiffies;
2041 /* process reset requests first */
2042 ice_reset_subtask(pf);
2044 /* bail if a reset/recovery cycle is pending or rebuild failed */
2045 if (ice_is_reset_in_progress(pf->state) ||
2046 test_bit(__ICE_SUSPENDED, pf->state) ||
2047 test_bit(__ICE_NEEDS_RESTART, pf->state)) {
2048 ice_service_task_complete(pf);
2052 ice_clean_adminq_subtask(pf);
2053 ice_check_media_subtask(pf);
2054 ice_check_for_hang_subtask(pf);
2055 ice_sync_fltr_subtask(pf);
2056 ice_handle_mdd_event(pf);
2057 ice_watchdog_subtask(pf);
2059 if (ice_is_safe_mode(pf)) {
2060 ice_service_task_complete(pf);
2064 ice_process_vflr_event(pf);
2065 ice_clean_mailboxq_subtask(pf);
2066 ice_sync_arfs_fltrs(pf);
2067 /* Clear __ICE_SERVICE_SCHED flag to allow scheduling next event */
2068 ice_service_task_complete(pf);
2070 /* If the tasks have taken longer than one service timer period
2071 * or there is more work to be done, reset the service timer to
2072 * schedule the service task now.
2074 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2075 test_bit(__ICE_MDD_EVENT_PENDING, pf->state) ||
2076 test_bit(__ICE_VFLR_EVENT_PENDING, pf->state) ||
2077 test_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2078 test_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state))
2079 mod_timer(&pf->serv_tmr, jiffies);
2083 * ice_set_ctrlq_len - helper function to set controlq length
2084 * @hw: pointer to the HW instance
2086 static void ice_set_ctrlq_len(struct ice_hw *hw)
2088 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2089 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2090 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2091 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2092 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2093 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2094 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2095 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2099 * ice_schedule_reset - schedule a reset
2100 * @pf: board private structure
2101 * @reset: reset being requested
2103 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2105 struct device *dev = ice_pf_to_dev(pf);
2107 /* bail out if earlier reset has failed */
2108 if (test_bit(__ICE_RESET_FAILED, pf->state)) {
2109 dev_dbg(dev, "earlier reset has failed\n");
2112 /* bail if reset/recovery already in progress */
2113 if (ice_is_reset_in_progress(pf->state)) {
2114 dev_dbg(dev, "Reset already in progress\n");
2120 set_bit(__ICE_PFR_REQ, pf->state);
2122 case ICE_RESET_CORER:
2123 set_bit(__ICE_CORER_REQ, pf->state);
2125 case ICE_RESET_GLOBR:
2126 set_bit(__ICE_GLOBR_REQ, pf->state);
2132 ice_service_task_schedule(pf);
2137 * ice_irq_affinity_notify - Callback for affinity changes
2138 * @notify: context as to what irq was changed
2139 * @mask: the new affinity mask
2141 * This is a callback function used by the irq_set_affinity_notifier function
2142 * so that we may register to receive changes to the irq affinity masks.
2145 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2146 const cpumask_t *mask)
2148 struct ice_q_vector *q_vector =
2149 container_of(notify, struct ice_q_vector, affinity_notify);
2151 cpumask_copy(&q_vector->affinity_mask, mask);
2155 * ice_irq_affinity_release - Callback for affinity notifier release
2156 * @ref: internal core kernel usage
2158 * This is a callback function used by the irq_set_affinity_notifier function
2159 * to inform the current notification subscriber that they will no longer
2160 * receive notifications.
2162 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2165 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2166 * @vsi: the VSI being configured
2168 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2170 struct ice_hw *hw = &vsi->back->hw;
2173 ice_for_each_q_vector(vsi, i)
2174 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2181 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2182 * @vsi: the VSI being configured
2183 * @basename: name for the vector
2185 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2187 int q_vectors = vsi->num_q_vectors;
2188 struct ice_pf *pf = vsi->back;
2189 int base = vsi->base_vector;
2196 dev = ice_pf_to_dev(pf);
2197 for (vector = 0; vector < q_vectors; vector++) {
2198 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2200 irq_num = pf->msix_entries[base + vector].vector;
2202 if (q_vector->tx.ring && q_vector->rx.ring) {
2203 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2204 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2206 } else if (q_vector->rx.ring) {
2207 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2208 "%s-%s-%d", basename, "rx", rx_int_idx++);
2209 } else if (q_vector->tx.ring) {
2210 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2211 "%s-%s-%d", basename, "tx", tx_int_idx++);
2213 /* skip this unused q_vector */
2216 err = devm_request_irq(dev, irq_num, vsi->irq_handler, 0,
2217 q_vector->name, q_vector);
2219 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2224 /* register for affinity change notifications */
2225 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2226 struct irq_affinity_notify *affinity_notify;
2228 affinity_notify = &q_vector->affinity_notify;
2229 affinity_notify->notify = ice_irq_affinity_notify;
2230 affinity_notify->release = ice_irq_affinity_release;
2231 irq_set_affinity_notifier(irq_num, affinity_notify);
2234 /* assign the mask for this irq */
2235 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2238 vsi->irqs_ready = true;
2244 irq_num = pf->msix_entries[base + vector].vector;
2245 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2246 irq_set_affinity_notifier(irq_num, NULL);
2247 irq_set_affinity_hint(irq_num, NULL);
2248 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2254 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2255 * @vsi: VSI to setup Tx rings used by XDP
2257 * Return 0 on success and negative value on error
2259 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2261 struct device *dev = ice_pf_to_dev(vsi->back);
2264 for (i = 0; i < vsi->num_xdp_txq; i++) {
2265 u16 xdp_q_idx = vsi->alloc_txq + i;
2266 struct ice_ring *xdp_ring;
2268 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2271 goto free_xdp_rings;
2273 xdp_ring->q_index = xdp_q_idx;
2274 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2275 xdp_ring->ring_active = false;
2276 xdp_ring->vsi = vsi;
2277 xdp_ring->netdev = NULL;
2278 xdp_ring->dev = dev;
2279 xdp_ring->count = vsi->num_tx_desc;
2280 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2281 if (ice_setup_tx_ring(xdp_ring))
2282 goto free_xdp_rings;
2283 ice_set_ring_xdp(xdp_ring);
2284 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2291 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2292 ice_free_tx_ring(vsi->xdp_rings[i]);
2297 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2298 * @vsi: VSI to set the bpf prog on
2299 * @prog: the bpf prog pointer
2301 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2303 struct bpf_prog *old_prog;
2306 old_prog = xchg(&vsi->xdp_prog, prog);
2308 bpf_prog_put(old_prog);
2310 ice_for_each_rxq(vsi, i)
2311 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2315 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2316 * @vsi: VSI to bring up Tx rings used by XDP
2317 * @prog: bpf program that will be assigned to VSI
2319 * Return 0 on success and negative value on error
2321 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2323 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2324 int xdp_rings_rem = vsi->num_xdp_txq;
2325 struct ice_pf *pf = vsi->back;
2326 struct ice_qs_cfg xdp_qs_cfg = {
2327 .qs_mutex = &pf->avail_q_mutex,
2328 .pf_map = pf->avail_txqs,
2329 .pf_map_size = pf->max_pf_txqs,
2330 .q_count = vsi->num_xdp_txq,
2331 .scatter_count = ICE_MAX_SCATTER_TXQS,
2332 .vsi_map = vsi->txq_map,
2333 .vsi_map_offset = vsi->alloc_txq,
2334 .mapping_mode = ICE_VSI_MAP_CONTIG
2336 enum ice_status status;
2340 dev = ice_pf_to_dev(pf);
2341 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2342 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2343 if (!vsi->xdp_rings)
2346 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2347 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2350 if (ice_xdp_alloc_setup_rings(vsi))
2351 goto clear_xdp_rings;
2353 /* follow the logic from ice_vsi_map_rings_to_vectors */
2354 ice_for_each_q_vector(vsi, v_idx) {
2355 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2356 int xdp_rings_per_v, q_id, q_base;
2358 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2359 vsi->num_q_vectors - v_idx);
2360 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2362 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2363 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2365 xdp_ring->q_vector = q_vector;
2366 xdp_ring->next = q_vector->tx.ring;
2367 q_vector->tx.ring = xdp_ring;
2369 xdp_rings_rem -= xdp_rings_per_v;
2372 /* omit the scheduler update if in reset path; XDP queues will be
2373 * taken into account at the end of ice_vsi_rebuild, where
2374 * ice_cfg_vsi_lan is being called
2376 if (ice_is_reset_in_progress(pf->state))
2379 /* tell the Tx scheduler that right now we have
2382 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2383 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2385 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2388 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2389 ice_stat_str(status));
2390 goto clear_xdp_rings;
2392 ice_vsi_assign_bpf_prog(vsi, prog);
2396 for (i = 0; i < vsi->num_xdp_txq; i++)
2397 if (vsi->xdp_rings[i]) {
2398 kfree_rcu(vsi->xdp_rings[i], rcu);
2399 vsi->xdp_rings[i] = NULL;
2403 mutex_lock(&pf->avail_q_mutex);
2404 for (i = 0; i < vsi->num_xdp_txq; i++) {
2405 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2406 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2408 mutex_unlock(&pf->avail_q_mutex);
2410 devm_kfree(dev, vsi->xdp_rings);
2415 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2416 * @vsi: VSI to remove XDP rings
2418 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2421 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2423 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2424 struct ice_pf *pf = vsi->back;
2427 /* q_vectors are freed in reset path so there's no point in detaching
2428 * rings; in case of rebuild being triggered not from reset bits
2429 * in pf->state won't be set, so additionally check first q_vector
2432 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2435 ice_for_each_q_vector(vsi, v_idx) {
2436 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2437 struct ice_ring *ring;
2439 ice_for_each_ring(ring, q_vector->tx)
2440 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2443 /* restore the value of last node prior to XDP setup */
2444 q_vector->tx.ring = ring;
2448 mutex_lock(&pf->avail_q_mutex);
2449 for (i = 0; i < vsi->num_xdp_txq; i++) {
2450 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2451 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2453 mutex_unlock(&pf->avail_q_mutex);
2455 for (i = 0; i < vsi->num_xdp_txq; i++)
2456 if (vsi->xdp_rings[i]) {
2457 if (vsi->xdp_rings[i]->desc)
2458 ice_free_tx_ring(vsi->xdp_rings[i]);
2459 kfree_rcu(vsi->xdp_rings[i], rcu);
2460 vsi->xdp_rings[i] = NULL;
2463 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2464 vsi->xdp_rings = NULL;
2466 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2469 ice_vsi_assign_bpf_prog(vsi, NULL);
2471 /* notify Tx scheduler that we destroyed XDP queues and bring
2472 * back the old number of child nodes
2474 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2475 max_txqs[i] = vsi->num_txq;
2477 /* change number of XDP Tx queues to 0 */
2478 vsi->num_xdp_txq = 0;
2480 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2485 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2486 * @vsi: VSI to setup XDP for
2487 * @prog: XDP program
2488 * @extack: netlink extended ack
2491 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2492 struct netlink_ext_ack *extack)
2494 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2495 bool if_running = netif_running(vsi->netdev);
2496 int ret = 0, xdp_ring_err = 0;
2498 if (frame_size > vsi->rx_buf_len) {
2499 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2503 /* need to stop netdev while setting up the program for Rx rings */
2504 if (if_running && !test_and_set_bit(__ICE_DOWN, vsi->state)) {
2505 ret = ice_down(vsi);
2507 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2512 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2513 vsi->num_xdp_txq = vsi->alloc_rxq;
2514 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2516 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2517 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2518 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2520 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2522 ice_vsi_assign_bpf_prog(vsi, prog);
2528 if (!ret && prog && vsi->xsk_pools) {
2531 ice_for_each_rxq(vsi, i) {
2532 struct ice_ring *rx_ring = vsi->rx_rings[i];
2534 if (rx_ring->xsk_pool)
2535 napi_schedule(&rx_ring->q_vector->napi);
2539 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2543 * ice_xdp - implements XDP handler
2547 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2549 struct ice_netdev_priv *np = netdev_priv(dev);
2550 struct ice_vsi *vsi = np->vsi;
2552 if (vsi->type != ICE_VSI_PF) {
2553 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2557 switch (xdp->command) {
2558 case XDP_SETUP_PROG:
2559 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2560 case XDP_SETUP_XSK_POOL:
2561 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2569 * ice_ena_misc_vector - enable the non-queue interrupts
2570 * @pf: board private structure
2572 static void ice_ena_misc_vector(struct ice_pf *pf)
2574 struct ice_hw *hw = &pf->hw;
2577 /* Disable anti-spoof detection interrupt to prevent spurious event
2578 * interrupts during a function reset. Anti-spoof functionally is
2581 val = rd32(hw, GL_MDCK_TX_TDPU);
2582 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2583 wr32(hw, GL_MDCK_TX_TDPU, val);
2585 /* clear things first */
2586 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2587 rd32(hw, PFINT_OICR); /* read to clear */
2589 val = (PFINT_OICR_ECC_ERR_M |
2590 PFINT_OICR_MAL_DETECT_M |
2592 PFINT_OICR_PCI_EXCEPTION_M |
2594 PFINT_OICR_HMC_ERR_M |
2595 PFINT_OICR_PE_CRITERR_M);
2597 wr32(hw, PFINT_OICR_ENA, val);
2599 /* SW_ITR_IDX = 0, but don't change INTENA */
2600 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2601 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2605 * ice_misc_intr - misc interrupt handler
2606 * @irq: interrupt number
2607 * @data: pointer to a q_vector
2609 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2611 struct ice_pf *pf = (struct ice_pf *)data;
2612 struct ice_hw *hw = &pf->hw;
2613 irqreturn_t ret = IRQ_NONE;
2617 dev = ice_pf_to_dev(pf);
2618 set_bit(__ICE_ADMINQ_EVENT_PENDING, pf->state);
2619 set_bit(__ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2621 oicr = rd32(hw, PFINT_OICR);
2622 ena_mask = rd32(hw, PFINT_OICR_ENA);
2624 if (oicr & PFINT_OICR_SWINT_M) {
2625 ena_mask &= ~PFINT_OICR_SWINT_M;
2629 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2630 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2631 set_bit(__ICE_MDD_EVENT_PENDING, pf->state);
2633 if (oicr & PFINT_OICR_VFLR_M) {
2634 /* disable any further VFLR event notifications */
2635 if (test_bit(__ICE_VF_RESETS_DISABLED, pf->state)) {
2636 u32 reg = rd32(hw, PFINT_OICR_ENA);
2638 reg &= ~PFINT_OICR_VFLR_M;
2639 wr32(hw, PFINT_OICR_ENA, reg);
2641 ena_mask &= ~PFINT_OICR_VFLR_M;
2642 set_bit(__ICE_VFLR_EVENT_PENDING, pf->state);
2646 if (oicr & PFINT_OICR_GRST_M) {
2649 /* we have a reset warning */
2650 ena_mask &= ~PFINT_OICR_GRST_M;
2651 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2652 GLGEN_RSTAT_RESET_TYPE_S;
2654 if (reset == ICE_RESET_CORER)
2656 else if (reset == ICE_RESET_GLOBR)
2658 else if (reset == ICE_RESET_EMPR)
2661 dev_dbg(dev, "Invalid reset type %d\n", reset);
2663 /* If a reset cycle isn't already in progress, we set a bit in
2664 * pf->state so that the service task can start a reset/rebuild.
2665 * We also make note of which reset happened so that peer
2666 * devices/drivers can be informed.
2668 if (!test_and_set_bit(__ICE_RESET_OICR_RECV, pf->state)) {
2669 if (reset == ICE_RESET_CORER)
2670 set_bit(__ICE_CORER_RECV, pf->state);
2671 else if (reset == ICE_RESET_GLOBR)
2672 set_bit(__ICE_GLOBR_RECV, pf->state);
2674 set_bit(__ICE_EMPR_RECV, pf->state);
2676 /* There are couple of different bits at play here.
2677 * hw->reset_ongoing indicates whether the hardware is
2678 * in reset. This is set to true when a reset interrupt
2679 * is received and set back to false after the driver
2680 * has determined that the hardware is out of reset.
2682 * __ICE_RESET_OICR_RECV in pf->state indicates
2683 * that a post reset rebuild is required before the
2684 * driver is operational again. This is set above.
2686 * As this is the start of the reset/rebuild cycle, set
2687 * both to indicate that.
2689 hw->reset_ongoing = true;
2693 if (oicr & PFINT_OICR_HMC_ERR_M) {
2694 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
2695 dev_dbg(dev, "HMC Error interrupt - info 0x%x, data 0x%x\n",
2696 rd32(hw, PFHMC_ERRORINFO),
2697 rd32(hw, PFHMC_ERRORDATA));
2700 /* Report any remaining unexpected interrupts */
2703 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2704 /* If a critical error is pending there is no choice but to
2707 if (oicr & (PFINT_OICR_PE_CRITERR_M |
2708 PFINT_OICR_PCI_EXCEPTION_M |
2709 PFINT_OICR_ECC_ERR_M)) {
2710 set_bit(__ICE_PFR_REQ, pf->state);
2711 ice_service_task_schedule(pf);
2716 ice_service_task_schedule(pf);
2717 ice_irq_dynamic_ena(hw, NULL, NULL);
2723 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2724 * @hw: pointer to HW structure
2726 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2728 /* disable Admin queue Interrupt causes */
2729 wr32(hw, PFINT_FW_CTL,
2730 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2732 /* disable Mailbox queue Interrupt causes */
2733 wr32(hw, PFINT_MBX_CTL,
2734 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2736 /* disable Control queue Interrupt causes */
2737 wr32(hw, PFINT_OICR_CTL,
2738 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2744 * ice_free_irq_msix_misc - Unroll misc vector setup
2745 * @pf: board private structure
2747 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2749 struct ice_hw *hw = &pf->hw;
2751 ice_dis_ctrlq_interrupts(hw);
2753 /* disable OICR interrupt */
2754 wr32(hw, PFINT_OICR_ENA, 0);
2757 if (pf->msix_entries) {
2758 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2759 devm_free_irq(ice_pf_to_dev(pf),
2760 pf->msix_entries[pf->oicr_idx].vector, pf);
2763 pf->num_avail_sw_msix += 1;
2764 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2768 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2769 * @hw: pointer to HW structure
2770 * @reg_idx: HW vector index to associate the control queue interrupts with
2772 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2776 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2777 PFINT_OICR_CTL_CAUSE_ENA_M);
2778 wr32(hw, PFINT_OICR_CTL, val);
2780 /* enable Admin queue Interrupt causes */
2781 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2782 PFINT_FW_CTL_CAUSE_ENA_M);
2783 wr32(hw, PFINT_FW_CTL, val);
2785 /* enable Mailbox queue Interrupt causes */
2786 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2787 PFINT_MBX_CTL_CAUSE_ENA_M);
2788 wr32(hw, PFINT_MBX_CTL, val);
2794 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2795 * @pf: board private structure
2797 * This sets up the handler for MSIX 0, which is used to manage the
2798 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2799 * when in MSI or Legacy interrupt mode.
2801 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2803 struct device *dev = ice_pf_to_dev(pf);
2804 struct ice_hw *hw = &pf->hw;
2805 int oicr_idx, err = 0;
2807 if (!pf->int_name[0])
2808 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2809 dev_driver_string(dev), dev_name(dev));
2811 /* Do not request IRQ but do enable OICR interrupt since settings are
2812 * lost during reset. Note that this function is called only during
2813 * rebuild path and not while reset is in progress.
2815 if (ice_is_reset_in_progress(pf->state))
2818 /* reserve one vector in irq_tracker for misc interrupts */
2819 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2823 pf->num_avail_sw_msix -= 1;
2824 pf->oicr_idx = (u16)oicr_idx;
2826 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2827 ice_misc_intr, 0, pf->int_name, pf);
2829 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2831 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2832 pf->num_avail_sw_msix += 1;
2837 ice_ena_misc_vector(pf);
2839 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2840 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2841 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2844 ice_irq_dynamic_ena(hw, NULL, NULL);
2850 * ice_napi_add - register NAPI handler for the VSI
2851 * @vsi: VSI for which NAPI handler is to be registered
2853 * This function is only called in the driver's load path. Registering the NAPI
2854 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2855 * reset/rebuild, etc.)
2857 static void ice_napi_add(struct ice_vsi *vsi)
2864 ice_for_each_q_vector(vsi, v_idx)
2865 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2866 ice_napi_poll, NAPI_POLL_WEIGHT);
2870 * ice_set_ops - set netdev and ethtools ops for the given netdev
2871 * @netdev: netdev instance
2873 static void ice_set_ops(struct net_device *netdev)
2875 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2877 if (ice_is_safe_mode(pf)) {
2878 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2879 ice_set_ethtool_safe_mode_ops(netdev);
2883 netdev->netdev_ops = &ice_netdev_ops;
2884 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
2885 ice_set_ethtool_ops(netdev);
2889 * ice_set_netdev_features - set features for the given netdev
2890 * @netdev: netdev instance
2892 static void ice_set_netdev_features(struct net_device *netdev)
2894 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2895 netdev_features_t csumo_features;
2896 netdev_features_t vlano_features;
2897 netdev_features_t dflt_features;
2898 netdev_features_t tso_features;
2900 if (ice_is_safe_mode(pf)) {
2902 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2903 netdev->hw_features = netdev->features;
2907 dflt_features = NETIF_F_SG |
2912 csumo_features = NETIF_F_RXCSUM |
2917 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2918 NETIF_F_HW_VLAN_CTAG_TX |
2919 NETIF_F_HW_VLAN_CTAG_RX;
2921 tso_features = NETIF_F_TSO |
2925 NETIF_F_GSO_UDP_TUNNEL |
2926 NETIF_F_GSO_GRE_CSUM |
2927 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2928 NETIF_F_GSO_PARTIAL |
2929 NETIF_F_GSO_IPXIP4 |
2930 NETIF_F_GSO_IPXIP6 |
2933 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2934 NETIF_F_GSO_GRE_CSUM;
2935 /* set features that user can change */
2936 netdev->hw_features = dflt_features | csumo_features |
2937 vlano_features | tso_features;
2939 /* add support for HW_CSUM on packets with MPLS header */
2940 netdev->mpls_features = NETIF_F_HW_CSUM;
2942 /* enable features */
2943 netdev->features |= netdev->hw_features;
2944 /* encap and VLAN devices inherit default, csumo and tso features */
2945 netdev->hw_enc_features |= dflt_features | csumo_features |
2947 netdev->vlan_features |= dflt_features | csumo_features |
2952 * ice_cfg_netdev - Allocate, configure and register a netdev
2953 * @vsi: the VSI associated with the new netdev
2955 * Returns 0 on success, negative value on failure
2957 static int ice_cfg_netdev(struct ice_vsi *vsi)
2959 struct ice_pf *pf = vsi->back;
2960 struct ice_netdev_priv *np;
2961 struct net_device *netdev;
2962 u8 mac_addr[ETH_ALEN];
2965 err = ice_devlink_create_port(vsi);
2969 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
2973 goto err_destroy_devlink_port;
2976 vsi->netdev = netdev;
2977 np = netdev_priv(netdev);
2980 ice_set_netdev_features(netdev);
2982 ice_set_ops(netdev);
2984 if (vsi->type == ICE_VSI_PF) {
2985 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
2986 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
2987 ether_addr_copy(netdev->dev_addr, mac_addr);
2988 ether_addr_copy(netdev->perm_addr, mac_addr);
2991 netdev->priv_flags |= IFF_UNICAST_FLT;
2993 /* Setup netdev TC information */
2994 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
2996 /* setup watchdog timeout value to be 5 second */
2997 netdev->watchdog_timeo = 5 * HZ;
2999 netdev->min_mtu = ETH_MIN_MTU;
3000 netdev->max_mtu = ICE_MAX_MTU;
3002 err = register_netdev(vsi->netdev);
3004 goto err_free_netdev;
3006 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
3008 netif_carrier_off(vsi->netdev);
3010 /* make sure transmit queues start off as stopped */
3011 netif_tx_stop_all_queues(vsi->netdev);
3016 free_netdev(vsi->netdev);
3018 err_destroy_devlink_port:
3019 ice_devlink_destroy_port(vsi);
3024 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3025 * @lut: Lookup table
3026 * @rss_table_size: Lookup table size
3027 * @rss_size: Range of queue number for hashing
3029 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3033 for (i = 0; i < rss_table_size; i++)
3034 lut[i] = i % rss_size;
3038 * ice_pf_vsi_setup - Set up a PF VSI
3039 * @pf: board private structure
3040 * @pi: pointer to the port_info instance
3042 * Returns pointer to the successfully allocated VSI software struct
3043 * on success, otherwise returns NULL on failure.
3045 static struct ice_vsi *
3046 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3048 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3052 * ice_ctrl_vsi_setup - Set up a control VSI
3053 * @pf: board private structure
3054 * @pi: pointer to the port_info instance
3056 * Returns pointer to the successfully allocated VSI software struct
3057 * on success, otherwise returns NULL on failure.
3059 static struct ice_vsi *
3060 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3062 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3066 * ice_lb_vsi_setup - Set up a loopback VSI
3067 * @pf: board private structure
3068 * @pi: pointer to the port_info instance
3070 * Returns pointer to the successfully allocated VSI software struct
3071 * on success, otherwise returns NULL on failure.
3074 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3076 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3080 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3081 * @netdev: network interface to be adjusted
3082 * @proto: unused protocol
3083 * @vid: VLAN ID to be added
3085 * net_device_ops implementation for adding VLAN IDs
3088 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3091 struct ice_netdev_priv *np = netdev_priv(netdev);
3092 struct ice_vsi *vsi = np->vsi;
3095 if (vid >= VLAN_N_VID) {
3096 netdev_err(netdev, "VLAN id requested %d is out of range %d\n",
3104 /* VLAN 0 is added by default during load/reset */
3108 /* Enable VLAN pruning when a VLAN other than 0 is added */
3109 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3110 ret = ice_cfg_vlan_pruning(vsi, true, false);
3115 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3116 * packets aren't pruned by the device's internal switch on Rx
3118 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3120 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3126 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3127 * @netdev: network interface to be adjusted
3128 * @proto: unused protocol
3129 * @vid: VLAN ID to be removed
3131 * net_device_ops implementation for removing VLAN IDs
3134 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3137 struct ice_netdev_priv *np = netdev_priv(netdev);
3138 struct ice_vsi *vsi = np->vsi;
3144 /* don't allow removal of VLAN 0 */
3148 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3151 ret = ice_vsi_kill_vlan(vsi, vid);
3155 /* Disable pruning when VLAN 0 is the only VLAN rule */
3156 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3157 ret = ice_cfg_vlan_pruning(vsi, false, false);
3159 set_bit(ICE_VSI_FLAG_VLAN_FLTR_CHANGED, vsi->flags);
3164 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3165 * @pf: board private structure
3167 * Returns 0 on success, negative value on failure
3169 static int ice_setup_pf_sw(struct ice_pf *pf)
3171 struct ice_vsi *vsi;
3174 if (ice_is_reset_in_progress(pf->state))
3177 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3181 status = ice_cfg_netdev(vsi);
3184 goto unroll_vsi_setup;
3186 /* netdev has to be configured before setting frame size */
3187 ice_vsi_cfg_frame_size(vsi);
3189 /* Setup DCB netlink interface */
3190 ice_dcbnl_setup(vsi);
3192 /* registering the NAPI handler requires both the queues and
3193 * netdev to be created, which are done in ice_pf_vsi_setup()
3194 * and ice_cfg_netdev() respectively
3198 status = ice_set_cpu_rx_rmap(vsi);
3200 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3201 vsi->vsi_num, status);
3203 goto unroll_napi_add;
3205 status = ice_init_mac_fltr(pf);
3207 goto free_cpu_rx_map;
3212 ice_free_cpu_rx_rmap(vsi);
3218 if (vsi->netdev->reg_state == NETREG_REGISTERED)
3219 unregister_netdev(vsi->netdev);
3220 free_netdev(vsi->netdev);
3226 ice_vsi_release(vsi);
3231 * ice_get_avail_q_count - Get count of queues in use
3232 * @pf_qmap: bitmap to get queue use count from
3233 * @lock: pointer to a mutex that protects access to pf_qmap
3234 * @size: size of the bitmap
3237 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3243 for_each_clear_bit(bit, pf_qmap, size)
3251 * ice_get_avail_txq_count - Get count of Tx queues in use
3252 * @pf: pointer to an ice_pf instance
3254 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3256 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3261 * ice_get_avail_rxq_count - Get count of Rx queues in use
3262 * @pf: pointer to an ice_pf instance
3264 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3266 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3271 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3272 * @pf: board private structure to initialize
3274 static void ice_deinit_pf(struct ice_pf *pf)
3276 ice_service_task_stop(pf);
3277 mutex_destroy(&pf->sw_mutex);
3278 mutex_destroy(&pf->tc_mutex);
3279 mutex_destroy(&pf->avail_q_mutex);
3281 if (pf->avail_txqs) {
3282 bitmap_free(pf->avail_txqs);
3283 pf->avail_txqs = NULL;
3286 if (pf->avail_rxqs) {
3287 bitmap_free(pf->avail_rxqs);
3288 pf->avail_rxqs = NULL;
3293 * ice_set_pf_caps - set PFs capability flags
3294 * @pf: pointer to the PF instance
3296 static void ice_set_pf_caps(struct ice_pf *pf)
3298 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3300 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3301 if (func_caps->common_cap.dcb)
3302 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3303 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3304 if (func_caps->common_cap.sr_iov_1_1) {
3305 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3306 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3309 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3310 if (func_caps->common_cap.rss_table_size)
3311 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3313 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3314 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3317 /* ctrl_vsi_idx will be set to a valid value when flow director
3318 * is setup by ice_init_fdir
3320 pf->ctrl_vsi_idx = ICE_NO_VSI;
3321 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3322 /* force guaranteed filter pool for PF */
3323 ice_alloc_fd_guar_item(&pf->hw, &unused,
3324 func_caps->fd_fltr_guar);
3325 /* force shared filter pool for PF */
3326 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3327 func_caps->fd_fltr_best_effort);
3330 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3331 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3335 * ice_init_pf - Initialize general software structures (struct ice_pf)
3336 * @pf: board private structure to initialize
3338 static int ice_init_pf(struct ice_pf *pf)
3340 ice_set_pf_caps(pf);
3342 mutex_init(&pf->sw_mutex);
3343 mutex_init(&pf->tc_mutex);
3345 INIT_HLIST_HEAD(&pf->aq_wait_list);
3346 spin_lock_init(&pf->aq_wait_lock);
3347 init_waitqueue_head(&pf->aq_wait_queue);
3349 /* setup service timer and periodic service task */
3350 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3351 pf->serv_tmr_period = HZ;
3352 INIT_WORK(&pf->serv_task, ice_service_task);
3353 clear_bit(__ICE_SERVICE_SCHED, pf->state);
3355 mutex_init(&pf->avail_q_mutex);
3356 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3357 if (!pf->avail_txqs)
3360 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3361 if (!pf->avail_rxqs) {
3362 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3363 pf->avail_txqs = NULL;
3371 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3372 * @pf: board private structure
3374 * compute the number of MSIX vectors required (v_budget) and request from
3375 * the OS. Return the number of vectors reserved or negative on failure
3377 static int ice_ena_msix_range(struct ice_pf *pf)
3379 struct device *dev = ice_pf_to_dev(pf);
3380 int v_left, v_actual, v_budget = 0;
3383 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3385 /* reserve one vector for miscellaneous handler */
3387 if (v_left < needed)
3388 goto no_hw_vecs_left_err;
3392 /* reserve vectors for LAN traffic */
3393 needed = min_t(int, num_online_cpus(), v_left);
3394 if (v_left < needed)
3395 goto no_hw_vecs_left_err;
3396 pf->num_lan_msix = needed;
3400 /* reserve one vector for flow director */
3401 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3402 needed = ICE_FDIR_MSIX;
3403 if (v_left < needed)
3404 goto no_hw_vecs_left_err;
3409 pf->msix_entries = devm_kcalloc(dev, v_budget,
3410 sizeof(*pf->msix_entries), GFP_KERNEL);
3412 if (!pf->msix_entries) {
3417 for (i = 0; i < v_budget; i++)
3418 pf->msix_entries[i].entry = i;
3420 /* actually reserve the vectors */
3421 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3422 ICE_MIN_MSIX, v_budget);
3425 dev_err(dev, "unable to reserve MSI-X vectors\n");
3430 if (v_actual < v_budget) {
3431 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3432 v_budget, v_actual);
3433 /* 2 vectors each for LAN and RDMA (traffic + OICR), one for flow director */
3434 #define ICE_MIN_LAN_VECS 2
3435 #define ICE_MIN_RDMA_VECS 2
3436 #define ICE_MIN_VECS (ICE_MIN_LAN_VECS + ICE_MIN_RDMA_VECS + 1)
3438 if (v_actual < ICE_MIN_LAN_VECS) {
3439 /* error if we can't get minimum vectors */
3440 pci_disable_msix(pf->pdev);
3444 pf->num_lan_msix = ICE_MIN_LAN_VECS;
3451 devm_kfree(dev, pf->msix_entries);
3454 no_hw_vecs_left_err:
3455 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3459 pf->num_lan_msix = 0;
3464 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3465 * @pf: board private structure
3467 static void ice_dis_msix(struct ice_pf *pf)
3469 pci_disable_msix(pf->pdev);
3470 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3471 pf->msix_entries = NULL;
3475 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3476 * @pf: board private structure
3478 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3482 if (pf->irq_tracker) {
3483 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3484 pf->irq_tracker = NULL;
3489 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3490 * @pf: board private structure to initialize
3492 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3496 vectors = ice_ena_msix_range(pf);
3501 /* set up vector assignment tracking */
3503 devm_kzalloc(ice_pf_to_dev(pf), sizeof(*pf->irq_tracker) +
3504 (sizeof(u16) * vectors), GFP_KERNEL);
3505 if (!pf->irq_tracker) {
3510 /* populate SW interrupts pool with number of OS granted IRQs. */
3511 pf->num_avail_sw_msix = (u16)vectors;
3512 pf->irq_tracker->num_entries = (u16)vectors;
3513 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3519 * ice_is_wol_supported - get NVM state of WoL
3520 * @pf: board private structure
3522 * Check if WoL is supported based on the HW configuration.
3523 * Returns true if NVM supports and enables WoL for this port, false otherwise
3525 bool ice_is_wol_supported(struct ice_pf *pf)
3527 struct ice_hw *hw = &pf->hw;
3530 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3531 * word) indicates WoL is not supported on the corresponding PF ID.
3533 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3536 return !(BIT(hw->pf_id) & wol_ctrl);
3540 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3541 * @vsi: VSI being changed
3542 * @new_rx: new number of Rx queues
3543 * @new_tx: new number of Tx queues
3545 * Only change the number of queues if new_tx, or new_rx is non-0.
3547 * Returns 0 on success.
3549 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3551 struct ice_pf *pf = vsi->back;
3552 int err = 0, timeout = 50;
3554 if (!new_rx && !new_tx)
3557 while (test_and_set_bit(__ICE_CFG_BUSY, pf->state)) {
3561 usleep_range(1000, 2000);
3565 vsi->req_txq = (u16)new_tx;
3567 vsi->req_rxq = (u16)new_rx;
3569 /* set for the next time the netdev is started */
3570 if (!netif_running(vsi->netdev)) {
3571 ice_vsi_rebuild(vsi, false);
3572 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3577 ice_vsi_rebuild(vsi, false);
3578 ice_pf_dcb_recfg(pf);
3581 clear_bit(__ICE_CFG_BUSY, pf->state);
3586 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3587 * @pf: PF to configure
3589 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3590 * VSI can still Tx/Rx VLAN tagged packets.
3592 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3594 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3595 struct ice_vsi_ctx *ctxt;
3596 enum ice_status status;
3602 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3607 ctxt->info = vsi->info;
3609 ctxt->info.valid_sections =
3610 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3611 ICE_AQ_VSI_PROP_SECURITY_VALID |
3612 ICE_AQ_VSI_PROP_SW_VALID);
3614 /* disable VLAN anti-spoof */
3615 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3616 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3618 /* disable VLAN pruning and keep all other settings */
3619 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3621 /* allow all VLANs on Tx and don't strip on Rx */
3622 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3623 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3625 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3627 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3628 ice_stat_str(status),
3629 ice_aq_str(hw->adminq.sq_last_status));
3631 vsi->info.sec_flags = ctxt->info.sec_flags;
3632 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3633 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3640 * ice_log_pkg_init - log result of DDP package load
3641 * @hw: pointer to hardware info
3642 * @status: status of package load
3645 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3647 struct ice_pf *pf = (struct ice_pf *)hw->back;
3648 struct device *dev = ice_pf_to_dev(pf);
3652 /* The package download AdminQ command returned success because
3653 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3654 * already a package loaded on the device.
3656 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3657 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3658 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3659 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3660 !memcmp(hw->pkg_name, hw->active_pkg_name,
3661 sizeof(hw->pkg_name))) {
3662 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3663 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3664 hw->active_pkg_name,
3665 hw->active_pkg_ver.major,
3666 hw->active_pkg_ver.minor,
3667 hw->active_pkg_ver.update,
3668 hw->active_pkg_ver.draft);
3670 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3671 hw->active_pkg_name,
3672 hw->active_pkg_ver.major,
3673 hw->active_pkg_ver.minor,
3674 hw->active_pkg_ver.update,
3675 hw->active_pkg_ver.draft);
3676 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3677 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3678 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",
3679 hw->active_pkg_name,
3680 hw->active_pkg_ver.major,
3681 hw->active_pkg_ver.minor,
3682 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3683 *status = ICE_ERR_NOT_SUPPORTED;
3684 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3685 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3686 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",
3687 hw->active_pkg_name,
3688 hw->active_pkg_ver.major,
3689 hw->active_pkg_ver.minor,
3690 hw->active_pkg_ver.update,
3691 hw->active_pkg_ver.draft,
3698 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");
3699 *status = ICE_ERR_NOT_SUPPORTED;
3702 case ICE_ERR_FW_DDP_MISMATCH:
3703 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");
3705 case ICE_ERR_BUF_TOO_SHORT:
3707 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3709 case ICE_ERR_NOT_SUPPORTED:
3710 /* Package File version not supported */
3711 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3712 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3713 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3714 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3715 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3716 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3717 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3718 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",
3719 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3721 case ICE_ERR_AQ_ERROR:
3722 switch (hw->pkg_dwnld_status) {
3723 case ICE_AQ_RC_ENOSEC:
3724 case ICE_AQ_RC_EBADSIG:
3725 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");
3727 case ICE_AQ_RC_ESVN:
3728 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");
3730 case ICE_AQ_RC_EBADMAN:
3731 case ICE_AQ_RC_EBADBUF:
3732 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3733 /* poll for reset to complete */
3734 if (ice_check_reset(hw))
3735 dev_err(dev, "Error resetting device. Please reload the driver\n");
3742 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3749 * ice_load_pkg - load/reload the DDP Package file
3750 * @firmware: firmware structure when firmware requested or NULL for reload
3751 * @pf: pointer to the PF instance
3753 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3754 * initialize HW tables.
3757 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3759 enum ice_status status = ICE_ERR_PARAM;
3760 struct device *dev = ice_pf_to_dev(pf);
3761 struct ice_hw *hw = &pf->hw;
3763 /* Load DDP Package */
3764 if (firmware && !hw->pkg_copy) {
3765 status = ice_copy_and_init_pkg(hw, firmware->data,
3767 ice_log_pkg_init(hw, &status);
3768 } else if (!firmware && hw->pkg_copy) {
3769 /* Reload package during rebuild after CORER/GLOBR reset */
3770 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3771 ice_log_pkg_init(hw, &status);
3773 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3778 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3782 /* Successful download package is the precondition for advanced
3783 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3785 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3789 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3790 * @pf: pointer to the PF structure
3792 * There is no error returned here because the driver should be able to handle
3793 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3794 * specifically with Tx.
3796 static void ice_verify_cacheline_size(struct ice_pf *pf)
3798 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3799 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3800 ICE_CACHE_LINE_BYTES);
3804 * ice_send_version - update firmware with driver version
3807 * Returns ICE_SUCCESS on success, else error code
3809 static enum ice_status ice_send_version(struct ice_pf *pf)
3811 struct ice_driver_ver dv;
3813 dv.major_ver = 0xff;
3814 dv.minor_ver = 0xff;
3815 dv.build_ver = 0xff;
3816 dv.subbuild_ver = 0;
3817 strscpy((char *)dv.driver_string, UTS_RELEASE,
3818 sizeof(dv.driver_string));
3819 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3823 * ice_init_fdir - Initialize flow director VSI and configuration
3824 * @pf: pointer to the PF instance
3826 * returns 0 on success, negative on error
3828 static int ice_init_fdir(struct ice_pf *pf)
3830 struct device *dev = ice_pf_to_dev(pf);
3831 struct ice_vsi *ctrl_vsi;
3834 /* Side Band Flow Director needs to have a control VSI.
3835 * Allocate it and store it in the PF.
3837 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3839 dev_dbg(dev, "could not create control VSI\n");
3843 err = ice_vsi_open_ctrl(ctrl_vsi);
3845 dev_dbg(dev, "could not open control VSI\n");
3849 mutex_init(&pf->hw.fdir_fltr_lock);
3851 err = ice_fdir_create_dflt_rules(pf);
3858 ice_fdir_release_flows(&pf->hw);
3859 ice_vsi_close(ctrl_vsi);
3861 ice_vsi_release(ctrl_vsi);
3862 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3863 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3864 pf->ctrl_vsi_idx = ICE_NO_VSI;
3870 * ice_get_opt_fw_name - return optional firmware file name or NULL
3871 * @pf: pointer to the PF instance
3873 static char *ice_get_opt_fw_name(struct ice_pf *pf)
3875 /* Optional firmware name same as default with additional dash
3876 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3878 struct pci_dev *pdev = pf->pdev;
3879 char *opt_fw_filename;
3882 /* Determine the name of the optional file using the DSN (two
3883 * dwords following the start of the DSN Capability).
3885 dsn = pci_get_dsn(pdev);
3889 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3890 if (!opt_fw_filename)
3893 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3894 ICE_DDP_PKG_PATH, dsn);
3896 return opt_fw_filename;
3900 * ice_request_fw - Device initialization routine
3901 * @pf: pointer to the PF instance
3903 static void ice_request_fw(struct ice_pf *pf)
3905 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3906 const struct firmware *firmware = NULL;
3907 struct device *dev = ice_pf_to_dev(pf);
3910 /* optional device-specific DDP (if present) overrides the default DDP
3911 * package file. kernel logs a debug message if the file doesn't exist,
3912 * and warning messages for other errors.
3914 if (opt_fw_filename) {
3915 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3917 kfree(opt_fw_filename);
3921 /* request for firmware was successful. Download to device */
3922 ice_load_pkg(firmware, pf);
3923 kfree(opt_fw_filename);
3924 release_firmware(firmware);
3929 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3931 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3935 /* request for firmware was successful. Download to device */
3936 ice_load_pkg(firmware, pf);
3937 release_firmware(firmware);
3941 * ice_print_wake_reason - show the wake up cause in the log
3942 * @pf: pointer to the PF struct
3944 static void ice_print_wake_reason(struct ice_pf *pf)
3946 u32 wus = pf->wakeup_reason;
3947 const char *wake_str;
3949 /* if no wake event, nothing to print */
3953 if (wus & PFPM_WUS_LNKC_M)
3954 wake_str = "Link\n";
3955 else if (wus & PFPM_WUS_MAG_M)
3956 wake_str = "Magic Packet\n";
3957 else if (wus & PFPM_WUS_MNG_M)
3958 wake_str = "Management\n";
3959 else if (wus & PFPM_WUS_FW_RST_WK_M)
3960 wake_str = "Firmware Reset\n";
3962 wake_str = "Unknown\n";
3964 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
3968 * ice_probe - Device initialization routine
3969 * @pdev: PCI device information struct
3970 * @ent: entry in ice_pci_tbl
3972 * Returns 0 on success, negative on failure
3975 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
3977 struct device *dev = &pdev->dev;
3982 /* this driver uses devres, see
3983 * Documentation/driver-api/driver-model/devres.rst
3985 err = pcim_enable_device(pdev);
3989 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), pci_name(pdev));
3991 dev_err(dev, "BAR0 I/O map error %d\n", err);
3995 pf = ice_allocate_pf(dev);
3999 /* set up for high or low DMA */
4000 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4002 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4004 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4008 pci_enable_pcie_error_reporting(pdev);
4009 pci_set_master(pdev);
4012 pci_set_drvdata(pdev, pf);
4013 set_bit(__ICE_DOWN, pf->state);
4014 /* Disable service task until DOWN bit is cleared */
4015 set_bit(__ICE_SERVICE_DIS, pf->state);
4018 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4019 pci_save_state(pdev);
4022 hw->vendor_id = pdev->vendor;
4023 hw->device_id = pdev->device;
4024 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4025 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4026 hw->subsystem_device_id = pdev->subsystem_device;
4027 hw->bus.device = PCI_SLOT(pdev->devfn);
4028 hw->bus.func = PCI_FUNC(pdev->devfn);
4029 ice_set_ctrlq_len(hw);
4031 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4033 err = ice_devlink_register(pf);
4035 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4036 goto err_exit_unroll;
4039 #ifndef CONFIG_DYNAMIC_DEBUG
4041 hw->debug_mask = debug;
4044 err = ice_init_hw(hw);
4046 dev_err(dev, "ice_init_hw failed: %d\n", err);
4048 goto err_exit_unroll;
4053 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4054 * set in pf->state, which will cause ice_is_safe_mode to return
4057 if (ice_is_safe_mode(pf)) {
4058 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4059 /* we already got function/device capabilities but these don't
4060 * reflect what the driver needs to do in safe mode. Instead of
4061 * adding conditional logic everywhere to ignore these
4062 * device/function capabilities, override them.
4064 ice_set_safe_mode_caps(hw);
4067 err = ice_init_pf(pf);
4069 dev_err(dev, "ice_init_pf failed: %d\n", err);
4070 goto err_init_pf_unroll;
4073 ice_devlink_init_regions(pf);
4075 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4076 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4077 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4078 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4080 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4081 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4082 pf->hw.tnl.valid_count[TNL_VXLAN];
4083 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4084 UDP_TUNNEL_TYPE_VXLAN;
4087 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4088 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4089 pf->hw.tnl.valid_count[TNL_GENEVE];
4090 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4091 UDP_TUNNEL_TYPE_GENEVE;
4095 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4096 if (!pf->num_alloc_vsi) {
4098 goto err_init_pf_unroll;
4100 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4101 dev_warn(&pf->pdev->dev,
4102 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4103 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4104 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4107 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4111 goto err_init_pf_unroll;
4114 err = ice_init_interrupt_scheme(pf);
4116 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4118 goto err_init_vsi_unroll;
4121 /* In case of MSIX we are going to setup the misc vector right here
4122 * to handle admin queue events etc. In case of legacy and MSI
4123 * the misc functionality and queue processing is combined in
4124 * the same vector and that gets setup at open.
4126 err = ice_req_irq_msix_misc(pf);
4128 dev_err(dev, "setup of misc vector failed: %d\n", err);
4129 goto err_init_interrupt_unroll;
4132 /* create switch struct for the switch element created by FW on boot */
4133 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4134 if (!pf->first_sw) {
4136 goto err_msix_misc_unroll;
4140 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4142 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4144 pf->first_sw->pf = pf;
4146 /* record the sw_id available for later use */
4147 pf->first_sw->sw_id = hw->port_info->sw_id;
4149 err = ice_setup_pf_sw(pf);
4151 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4152 goto err_alloc_sw_unroll;
4155 clear_bit(__ICE_SERVICE_DIS, pf->state);
4157 /* tell the firmware we are up */
4158 err = ice_send_version(pf);
4160 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4162 goto err_send_version_unroll;
4165 /* since everything is good, start the service timer */
4166 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4168 err = ice_init_link_events(pf->hw.port_info);
4170 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4171 goto err_send_version_unroll;
4174 err = ice_init_nvm_phy_type(pf->hw.port_info);
4176 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4177 goto err_send_version_unroll;
4180 err = ice_update_link_info(pf->hw.port_info);
4182 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4183 goto err_send_version_unroll;
4186 ice_init_link_dflt_override(pf->hw.port_info);
4188 /* if media available, initialize PHY settings */
4189 if (pf->hw.port_info->phy.link_info.link_info &
4190 ICE_AQ_MEDIA_AVAILABLE) {
4191 err = ice_init_phy_user_cfg(pf->hw.port_info);
4193 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4194 goto err_send_version_unroll;
4197 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4198 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4201 ice_configure_phy(vsi);
4204 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4207 ice_verify_cacheline_size(pf);
4209 /* Save wakeup reason register for later use */
4210 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4212 /* check for a power management event */
4213 ice_print_wake_reason(pf);
4215 /* clear wake status, all bits */
4216 wr32(hw, PFPM_WUS, U32_MAX);
4218 /* Disable WoL at init, wait for user to enable */
4219 device_set_wakeup_enable(dev, false);
4221 if (ice_is_safe_mode(pf)) {
4222 ice_set_safe_mode_vlan_cfg(pf);
4226 /* initialize DDP driven features */
4228 /* Note: Flow director init failure is non-fatal to load */
4229 if (ice_init_fdir(pf))
4230 dev_err(dev, "could not initialize flow director\n");
4232 /* Note: DCB init failure is non-fatal to load */
4233 if (ice_init_pf_dcb(pf, false)) {
4234 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4235 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4237 ice_cfg_lldp_mib_change(&pf->hw, true);
4240 /* print PCI link speed and width */
4241 pcie_print_link_status(pf->pdev);
4244 /* ready to go, so clear down state bit */
4245 clear_bit(__ICE_DOWN, pf->state);
4248 err_send_version_unroll:
4249 ice_vsi_release_all(pf);
4250 err_alloc_sw_unroll:
4251 set_bit(__ICE_SERVICE_DIS, pf->state);
4252 set_bit(__ICE_DOWN, pf->state);
4253 devm_kfree(dev, pf->first_sw);
4254 err_msix_misc_unroll:
4255 ice_free_irq_msix_misc(pf);
4256 err_init_interrupt_unroll:
4257 ice_clear_interrupt_scheme(pf);
4258 err_init_vsi_unroll:
4259 devm_kfree(dev, pf->vsi);
4262 ice_devlink_destroy_regions(pf);
4265 ice_devlink_unregister(pf);
4266 pci_disable_pcie_error_reporting(pdev);
4267 pci_disable_device(pdev);
4272 * ice_set_wake - enable or disable Wake on LAN
4273 * @pf: pointer to the PF struct
4275 * Simple helper for WoL control
4277 static void ice_set_wake(struct ice_pf *pf)
4279 struct ice_hw *hw = &pf->hw;
4280 bool wol = pf->wol_ena;
4282 /* clear wake state, otherwise new wake events won't fire */
4283 wr32(hw, PFPM_WUS, U32_MAX);
4285 /* enable / disable APM wake up, no RMW needed */
4286 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4288 /* set magic packet filter enabled */
4289 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4293 * ice_setup_magic_mc_wake - setup device to wake on multicast magic packet
4294 * @pf: pointer to the PF struct
4296 * Issue firmware command to enable multicast magic wake, making
4297 * sure that any locally administered address (LAA) is used for
4298 * wake, and that PF reset doesn't undo the LAA.
4300 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4302 struct device *dev = ice_pf_to_dev(pf);
4303 struct ice_hw *hw = &pf->hw;
4304 enum ice_status status;
4305 u8 mac_addr[ETH_ALEN];
4306 struct ice_vsi *vsi;
4312 vsi = ice_get_main_vsi(pf);
4316 /* Get current MAC address in case it's an LAA */
4318 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4320 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4322 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4323 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4324 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4326 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4328 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4329 ice_stat_str(status),
4330 ice_aq_str(hw->adminq.sq_last_status));
4334 * ice_remove - Device removal routine
4335 * @pdev: PCI device information struct
4337 static void ice_remove(struct pci_dev *pdev)
4339 struct ice_pf *pf = pci_get_drvdata(pdev);
4345 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4346 if (!ice_is_reset_in_progress(pf->state))
4351 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4352 set_bit(__ICE_VF_RESETS_DISABLED, pf->state);
4356 set_bit(__ICE_DOWN, pf->state);
4357 ice_service_task_stop(pf);
4359 ice_aq_cancel_waiting_tasks(pf);
4361 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4362 if (!ice_is_safe_mode(pf))
4363 ice_remove_arfs(pf);
4364 ice_setup_mc_magic_wake(pf);
4365 ice_vsi_release_all(pf);
4367 ice_free_irq_msix_misc(pf);
4368 ice_for_each_vsi(pf, i) {
4371 ice_vsi_free_q_vectors(pf->vsi[i]);
4374 ice_devlink_destroy_regions(pf);
4375 ice_deinit_hw(&pf->hw);
4376 ice_devlink_unregister(pf);
4378 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4379 * do it via ice_schedule_reset() since there is no need to rebuild
4380 * and the service task is already stopped.
4382 ice_reset(&pf->hw, ICE_RESET_PFR);
4383 pci_wait_for_pending_transaction(pdev);
4384 ice_clear_interrupt_scheme(pf);
4385 pci_disable_pcie_error_reporting(pdev);
4386 pci_disable_device(pdev);
4390 * ice_shutdown - PCI callback for shutting down device
4391 * @pdev: PCI device information struct
4393 static void ice_shutdown(struct pci_dev *pdev)
4395 struct ice_pf *pf = pci_get_drvdata(pdev);
4399 if (system_state == SYSTEM_POWER_OFF) {
4400 pci_wake_from_d3(pdev, pf->wol_ena);
4401 pci_set_power_state(pdev, PCI_D3hot);
4407 * ice_prepare_for_shutdown - prep for PCI shutdown
4408 * @pf: board private structure
4410 * Inform or close all dependent features in prep for PCI device shutdown
4412 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4414 struct ice_hw *hw = &pf->hw;
4417 /* Notify VFs of impending reset */
4418 if (ice_check_sq_alive(hw, &hw->mailboxq))
4419 ice_vc_notify_reset(pf);
4421 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4423 /* disable the VSIs and their queues that are not already DOWN */
4424 ice_pf_dis_all_vsi(pf, false);
4426 ice_for_each_vsi(pf, v)
4428 pf->vsi[v]->vsi_num = 0;
4430 ice_shutdown_all_ctrlq(hw);
4434 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4435 * @pf: board private structure to reinitialize
4437 * This routine reinitialize interrupt scheme that was cleared during
4438 * power management suspend callback.
4440 * This should be called during resume routine to re-allocate the q_vectors
4441 * and reacquire interrupts.
4443 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4445 struct device *dev = ice_pf_to_dev(pf);
4448 /* Since we clear MSIX flag during suspend, we need to
4449 * set it back during resume...
4452 ret = ice_init_interrupt_scheme(pf);
4454 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4458 /* Remap vectors and rings, after successful re-init interrupts */
4459 ice_for_each_vsi(pf, v) {
4463 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4466 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4469 ret = ice_req_irq_msix_misc(pf);
4471 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4481 ice_vsi_free_q_vectors(pf->vsi[v]);
4488 * @dev: generic device information structure
4490 * Power Management callback to quiesce the device and prepare
4491 * for D3 transition.
4493 static int __maybe_unused ice_suspend(struct device *dev)
4495 struct pci_dev *pdev = to_pci_dev(dev);
4499 pf = pci_get_drvdata(pdev);
4501 if (!ice_pf_state_is_nominal(pf)) {
4502 dev_err(dev, "Device is not ready, no need to suspend it\n");
4506 /* Stop watchdog tasks until resume completion.
4507 * Even though it is most likely that the service task is
4508 * disabled if the device is suspended or down, the service task's
4509 * state is controlled by a different state bit, and we should
4510 * store and honor whatever state that bit is in at this point.
4512 disabled = ice_service_task_stop(pf);
4514 /* Already suspended?, then there is nothing to do */
4515 if (test_and_set_bit(__ICE_SUSPENDED, pf->state)) {
4517 ice_service_task_restart(pf);
4521 if (test_bit(__ICE_DOWN, pf->state) ||
4522 ice_is_reset_in_progress(pf->state)) {
4523 dev_err(dev, "can't suspend device in reset or already down\n");
4525 ice_service_task_restart(pf);
4529 ice_setup_mc_magic_wake(pf);
4531 ice_prepare_for_shutdown(pf);
4535 /* Free vectors, clear the interrupt scheme and release IRQs
4536 * for proper hibernation, especially with large number of CPUs.
4537 * Otherwise hibernation might fail when mapping all the vectors back
4540 ice_free_irq_msix_misc(pf);
4541 ice_for_each_vsi(pf, v) {
4544 ice_vsi_free_q_vectors(pf->vsi[v]);
4546 ice_clear_interrupt_scheme(pf);
4548 pci_save_state(pdev);
4549 pci_wake_from_d3(pdev, pf->wol_ena);
4550 pci_set_power_state(pdev, PCI_D3hot);
4555 * ice_resume - PM callback for waking up from D3
4556 * @dev: generic device information structure
4558 static int __maybe_unused ice_resume(struct device *dev)
4560 struct pci_dev *pdev = to_pci_dev(dev);
4561 enum ice_reset_req reset_type;
4566 pci_set_power_state(pdev, PCI_D0);
4567 pci_restore_state(pdev);
4568 pci_save_state(pdev);
4570 if (!pci_device_is_present(pdev))
4573 ret = pci_enable_device_mem(pdev);
4575 dev_err(dev, "Cannot enable device after suspend\n");
4579 pf = pci_get_drvdata(pdev);
4582 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4583 ice_print_wake_reason(pf);
4585 /* We cleared the interrupt scheme when we suspended, so we need to
4586 * restore it now to resume device functionality.
4588 ret = ice_reinit_interrupt_scheme(pf);
4590 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4592 clear_bit(__ICE_DOWN, pf->state);
4593 /* Now perform PF reset and rebuild */
4594 reset_type = ICE_RESET_PFR;
4595 /* re-enable service task for reset, but allow reset to schedule it */
4596 clear_bit(__ICE_SERVICE_DIS, pf->state);
4598 if (ice_schedule_reset(pf, reset_type))
4599 dev_err(dev, "Reset during resume failed.\n");
4601 clear_bit(__ICE_SUSPENDED, pf->state);
4602 ice_service_task_restart(pf);
4604 /* Restart the service task */
4605 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4609 #endif /* CONFIG_PM */
4612 * ice_pci_err_detected - warning that PCI error has been detected
4613 * @pdev: PCI device information struct
4614 * @err: the type of PCI error
4616 * Called to warn that something happened on the PCI bus and the error handling
4617 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4619 static pci_ers_result_t
4620 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4622 struct ice_pf *pf = pci_get_drvdata(pdev);
4625 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4627 return PCI_ERS_RESULT_DISCONNECT;
4630 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4631 ice_service_task_stop(pf);
4633 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4634 set_bit(__ICE_PFR_REQ, pf->state);
4635 ice_prepare_for_reset(pf);
4639 return PCI_ERS_RESULT_NEED_RESET;
4643 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4644 * @pdev: PCI device information struct
4646 * Called to determine if the driver can recover from the PCI slot reset by
4647 * using a register read to determine if the device is recoverable.
4649 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4651 struct ice_pf *pf = pci_get_drvdata(pdev);
4652 pci_ers_result_t result;
4656 err = pci_enable_device_mem(pdev);
4658 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4660 result = PCI_ERS_RESULT_DISCONNECT;
4662 pci_set_master(pdev);
4663 pci_restore_state(pdev);
4664 pci_save_state(pdev);
4665 pci_wake_from_d3(pdev, false);
4667 /* Check for life */
4668 reg = rd32(&pf->hw, GLGEN_RTRIG);
4670 result = PCI_ERS_RESULT_RECOVERED;
4672 result = PCI_ERS_RESULT_DISCONNECT;
4675 err = pci_aer_clear_nonfatal_status(pdev);
4677 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4679 /* non-fatal, continue */
4685 * ice_pci_err_resume - restart operations after PCI error recovery
4686 * @pdev: PCI device information struct
4688 * Called to allow the driver to bring things back up after PCI error and/or
4689 * reset recovery have finished
4691 static void ice_pci_err_resume(struct pci_dev *pdev)
4693 struct ice_pf *pf = pci_get_drvdata(pdev);
4696 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4701 if (test_bit(__ICE_SUSPENDED, pf->state)) {
4702 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4707 ice_restore_all_vfs_msi_state(pdev);
4709 ice_do_reset(pf, ICE_RESET_PFR);
4710 ice_service_task_restart(pf);
4711 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4715 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4716 * @pdev: PCI device information struct
4718 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4720 struct ice_pf *pf = pci_get_drvdata(pdev);
4722 if (!test_bit(__ICE_SUSPENDED, pf->state)) {
4723 ice_service_task_stop(pf);
4725 if (!test_bit(__ICE_PREPARED_FOR_RESET, pf->state)) {
4726 set_bit(__ICE_PFR_REQ, pf->state);
4727 ice_prepare_for_reset(pf);
4733 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4734 * @pdev: PCI device information struct
4736 static void ice_pci_err_reset_done(struct pci_dev *pdev)
4738 ice_pci_err_resume(pdev);
4741 /* ice_pci_tbl - PCI Device ID Table
4743 * Wildcard entries (PCI_ANY_ID) should come last
4744 * Last entry must be all 0s
4746 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4747 * Class, Class Mask, private data (not used) }
4749 static const struct pci_device_id ice_pci_tbl[] = {
4750 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4751 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4752 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4753 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4754 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4755 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4756 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4757 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4758 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4759 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4760 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4761 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4762 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4763 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4764 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4765 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4766 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4767 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4768 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4769 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4770 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4771 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4772 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4773 /* required last entry */
4776 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4778 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4780 static const struct pci_error_handlers ice_pci_err_handler = {
4781 .error_detected = ice_pci_err_detected,
4782 .slot_reset = ice_pci_err_slot_reset,
4783 .reset_prepare = ice_pci_err_reset_prepare,
4784 .reset_done = ice_pci_err_reset_done,
4785 .resume = ice_pci_err_resume
4788 static struct pci_driver ice_driver = {
4789 .name = KBUILD_MODNAME,
4790 .id_table = ice_pci_tbl,
4792 .remove = ice_remove,
4794 .driver.pm = &ice_pm_ops,
4795 #endif /* CONFIG_PM */
4796 .shutdown = ice_shutdown,
4797 .sriov_configure = ice_sriov_configure,
4798 .err_handler = &ice_pci_err_handler
4802 * ice_module_init - Driver registration routine
4804 * ice_module_init is the first routine called when the driver is
4805 * loaded. All it does is register with the PCI subsystem.
4807 static int __init ice_module_init(void)
4811 pr_info("%s\n", ice_driver_string);
4812 pr_info("%s\n", ice_copyright);
4814 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4816 pr_err("Failed to create workqueue\n");
4820 status = pci_register_driver(&ice_driver);
4822 pr_err("failed to register PCI driver, err %d\n", status);
4823 destroy_workqueue(ice_wq);
4828 module_init(ice_module_init);
4831 * ice_module_exit - Driver exit cleanup routine
4833 * ice_module_exit is called just before the driver is removed
4836 static void __exit ice_module_exit(void)
4838 pci_unregister_driver(&ice_driver);
4839 destroy_workqueue(ice_wq);
4840 pr_info("module unloaded\n");
4842 module_exit(ice_module_exit);
4845 * ice_set_mac_address - NDO callback to set MAC address
4846 * @netdev: network interface device structure
4847 * @pi: pointer to an address structure
4849 * Returns 0 on success, negative on failure
4851 static int ice_set_mac_address(struct net_device *netdev, void *pi)
4853 struct ice_netdev_priv *np = netdev_priv(netdev);
4854 struct ice_vsi *vsi = np->vsi;
4855 struct ice_pf *pf = vsi->back;
4856 struct ice_hw *hw = &pf->hw;
4857 struct sockaddr *addr = pi;
4858 enum ice_status status;
4863 mac = (u8 *)addr->sa_data;
4865 if (!is_valid_ether_addr(mac))
4866 return -EADDRNOTAVAIL;
4868 if (ether_addr_equal(netdev->dev_addr, mac)) {
4869 netdev_warn(netdev, "already using mac %pM\n", mac);
4873 if (test_bit(__ICE_DOWN, pf->state) ||
4874 ice_is_reset_in_progress(pf->state)) {
4875 netdev_err(netdev, "can't set mac %pM. device not ready\n",
4880 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
4881 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
4882 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
4883 err = -EADDRNOTAVAIL;
4884 goto err_update_filters;
4887 /* Add filter for new MAC. If filter exists, just return success */
4888 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
4889 if (status == ICE_ERR_ALREADY_EXISTS) {
4890 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
4894 /* error if the new filter addition failed */
4896 err = -EADDRNOTAVAIL;
4900 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
4905 /* change the netdev's MAC address */
4906 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4907 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
4910 /* write new MAC address to the firmware */
4911 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
4912 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
4914 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
4915 mac, ice_stat_str(status));
4921 * ice_set_rx_mode - NDO callback to set the netdev filters
4922 * @netdev: network interface device structure
4924 static void ice_set_rx_mode(struct net_device *netdev)
4926 struct ice_netdev_priv *np = netdev_priv(netdev);
4927 struct ice_vsi *vsi = np->vsi;
4932 /* Set the flags to synchronize filters
4933 * ndo_set_rx_mode may be triggered even without a change in netdev
4936 set_bit(ICE_VSI_FLAG_UMAC_FLTR_CHANGED, vsi->flags);
4937 set_bit(ICE_VSI_FLAG_MMAC_FLTR_CHANGED, vsi->flags);
4938 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
4940 /* schedule our worker thread which will take care of
4941 * applying the new filter changes
4943 ice_service_task_schedule(vsi->back);
4947 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
4948 * @netdev: network interface device structure
4949 * @queue_index: Queue ID
4950 * @maxrate: maximum bandwidth in Mbps
4953 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
4955 struct ice_netdev_priv *np = netdev_priv(netdev);
4956 struct ice_vsi *vsi = np->vsi;
4957 enum ice_status status;
4961 /* Validate maxrate requested is within permitted range */
4962 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
4963 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
4964 maxrate, queue_index);
4968 q_handle = vsi->tx_rings[queue_index]->q_handle;
4969 tc = ice_dcb_get_tc(vsi, queue_index);
4971 /* Set BW back to default, when user set maxrate to 0 */
4973 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
4974 q_handle, ICE_MAX_BW);
4976 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
4977 q_handle, ICE_MAX_BW, maxrate * 1000);
4979 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
4980 ice_stat_str(status));
4988 * ice_fdb_add - add an entry to the hardware database
4989 * @ndm: the input from the stack
4990 * @tb: pointer to array of nladdr (unused)
4991 * @dev: the net device pointer
4992 * @addr: the MAC address entry being added
4994 * @flags: instructions from stack about fdb operation
4995 * @extack: netlink extended ack
4998 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
4999 struct net_device *dev, const unsigned char *addr, u16 vid,
5000 u16 flags, struct netlink_ext_ack __always_unused *extack)
5005 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5008 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5009 netdev_err(dev, "FDB only supports static addresses\n");
5013 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5014 err = dev_uc_add_excl(dev, addr);
5015 else if (is_multicast_ether_addr(addr))
5016 err = dev_mc_add_excl(dev, addr);
5020 /* Only return duplicate errors if NLM_F_EXCL is set */
5021 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5028 * ice_fdb_del - delete an entry from the hardware database
5029 * @ndm: the input from the stack
5030 * @tb: pointer to array of nladdr (unused)
5031 * @dev: the net device pointer
5032 * @addr: the MAC address entry being added
5036 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5037 struct net_device *dev, const unsigned char *addr,
5038 __always_unused u16 vid)
5042 if (ndm->ndm_state & NUD_PERMANENT) {
5043 netdev_err(dev, "FDB only supports static addresses\n");
5047 if (is_unicast_ether_addr(addr))
5048 err = dev_uc_del(dev, addr);
5049 else if (is_multicast_ether_addr(addr))
5050 err = dev_mc_del(dev, addr);
5058 * ice_set_features - set the netdev feature flags
5059 * @netdev: ptr to the netdev being adjusted
5060 * @features: the feature set that the stack is suggesting
5063 ice_set_features(struct net_device *netdev, netdev_features_t features)
5065 struct ice_netdev_priv *np = netdev_priv(netdev);
5066 struct ice_vsi *vsi = np->vsi;
5067 struct ice_pf *pf = vsi->back;
5070 /* Don't set any netdev advanced features with device in Safe Mode */
5071 if (ice_is_safe_mode(vsi->back)) {
5072 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5076 /* Do not change setting during reset */
5077 if (ice_is_reset_in_progress(pf->state)) {
5078 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5082 /* Multiple features can be changed in one call so keep features in
5083 * separate if/else statements to guarantee each feature is checked
5085 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5086 ret = ice_vsi_manage_rss_lut(vsi, true);
5087 else if (!(features & NETIF_F_RXHASH) &&
5088 netdev->features & NETIF_F_RXHASH)
5089 ret = ice_vsi_manage_rss_lut(vsi, false);
5091 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5092 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5093 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5094 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5095 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5096 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5098 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5099 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5100 ret = ice_vsi_manage_vlan_insertion(vsi);
5101 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5102 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5103 ret = ice_vsi_manage_vlan_insertion(vsi);
5105 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5106 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5107 ret = ice_cfg_vlan_pruning(vsi, true, false);
5108 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5109 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5110 ret = ice_cfg_vlan_pruning(vsi, false, false);
5112 if ((features & NETIF_F_NTUPLE) &&
5113 !(netdev->features & NETIF_F_NTUPLE)) {
5114 ice_vsi_manage_fdir(vsi, true);
5116 } else if (!(features & NETIF_F_NTUPLE) &&
5117 (netdev->features & NETIF_F_NTUPLE)) {
5118 ice_vsi_manage_fdir(vsi, false);
5119 ice_clear_arfs(vsi);
5126 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5127 * @vsi: VSI to setup VLAN properties for
5129 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5133 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5134 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5135 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5136 ret = ice_vsi_manage_vlan_insertion(vsi);
5142 * ice_vsi_cfg - Setup the VSI
5143 * @vsi: the VSI being configured
5145 * Return 0 on success and negative value on error
5147 int ice_vsi_cfg(struct ice_vsi *vsi)
5152 ice_set_rx_mode(vsi->netdev);
5154 err = ice_vsi_vlan_setup(vsi);
5159 ice_vsi_cfg_dcb_rings(vsi);
5161 err = ice_vsi_cfg_lan_txqs(vsi);
5162 if (!err && ice_is_xdp_ena_vsi(vsi))
5163 err = ice_vsi_cfg_xdp_txqs(vsi);
5165 err = ice_vsi_cfg_rxqs(vsi);
5171 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5172 * @vsi: the VSI being configured
5174 static void ice_napi_enable_all(struct ice_vsi *vsi)
5181 ice_for_each_q_vector(vsi, q_idx) {
5182 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5184 if (q_vector->rx.ring || q_vector->tx.ring)
5185 napi_enable(&q_vector->napi);
5190 * ice_up_complete - Finish the last steps of bringing up a connection
5191 * @vsi: The VSI being configured
5193 * Return 0 on success and negative value on error
5195 static int ice_up_complete(struct ice_vsi *vsi)
5197 struct ice_pf *pf = vsi->back;
5200 ice_vsi_cfg_msix(vsi);
5202 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5203 * Tx queue group list was configured and the context bits were
5204 * programmed using ice_vsi_cfg_txqs
5206 err = ice_vsi_start_all_rx_rings(vsi);
5210 clear_bit(__ICE_DOWN, vsi->state);
5211 ice_napi_enable_all(vsi);
5212 ice_vsi_ena_irq(vsi);
5214 if (vsi->port_info &&
5215 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5217 ice_print_link_msg(vsi, true);
5218 netif_tx_start_all_queues(vsi->netdev);
5219 netif_carrier_on(vsi->netdev);
5222 ice_service_task_schedule(pf);
5228 * ice_up - Bring the connection back up after being down
5229 * @vsi: VSI being configured
5231 int ice_up(struct ice_vsi *vsi)
5235 err = ice_vsi_cfg(vsi);
5237 err = ice_up_complete(vsi);
5243 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5244 * @ring: Tx or Rx ring to read stats from
5245 * @pkts: packets stats counter
5246 * @bytes: bytes stats counter
5248 * This function fetches stats from the ring considering the atomic operations
5249 * that needs to be performed to read u64 values in 32 bit machine.
5252 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5261 start = u64_stats_fetch_begin_irq(&ring->syncp);
5262 *pkts = ring->stats.pkts;
5263 *bytes = ring->stats.bytes;
5264 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5268 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5269 * @vsi: the VSI to be updated
5270 * @rings: rings to work on
5271 * @count: number of rings
5274 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5277 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5280 for (i = 0; i < count; i++) {
5281 struct ice_ring *ring;
5284 ring = READ_ONCE(rings[i]);
5285 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5286 vsi_stats->tx_packets += pkts;
5287 vsi_stats->tx_bytes += bytes;
5288 vsi->tx_restart += ring->tx_stats.restart_q;
5289 vsi->tx_busy += ring->tx_stats.tx_busy;
5290 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5295 * ice_update_vsi_ring_stats - Update VSI stats counters
5296 * @vsi: the VSI to be updated
5298 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5300 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5301 struct ice_ring *ring;
5305 /* reset netdev stats */
5306 vsi_stats->tx_packets = 0;
5307 vsi_stats->tx_bytes = 0;
5308 vsi_stats->rx_packets = 0;
5309 vsi_stats->rx_bytes = 0;
5311 /* reset non-netdev (extended) stats */
5312 vsi->tx_restart = 0;
5314 vsi->tx_linearize = 0;
5315 vsi->rx_buf_failed = 0;
5316 vsi->rx_page_failed = 0;
5317 vsi->rx_gro_dropped = 0;
5321 /* update Tx rings counters */
5322 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5324 /* update Rx rings counters */
5325 ice_for_each_rxq(vsi, i) {
5326 ring = READ_ONCE(vsi->rx_rings[i]);
5327 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5328 vsi_stats->rx_packets += pkts;
5329 vsi_stats->rx_bytes += bytes;
5330 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5331 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5332 vsi->rx_gro_dropped += ring->rx_stats.gro_dropped;
5335 /* update XDP Tx rings counters */
5336 if (ice_is_xdp_ena_vsi(vsi))
5337 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5344 * ice_update_vsi_stats - Update VSI stats counters
5345 * @vsi: the VSI to be updated
5347 void ice_update_vsi_stats(struct ice_vsi *vsi)
5349 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5350 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5351 struct ice_pf *pf = vsi->back;
5353 if (test_bit(__ICE_DOWN, vsi->state) ||
5354 test_bit(__ICE_CFG_BUSY, pf->state))
5357 /* get stats as recorded by Tx/Rx rings */
5358 ice_update_vsi_ring_stats(vsi);
5360 /* get VSI stats as recorded by the hardware */
5361 ice_update_eth_stats(vsi);
5363 cur_ns->tx_errors = cur_es->tx_errors;
5364 cur_ns->rx_dropped = cur_es->rx_discards + vsi->rx_gro_dropped;
5365 cur_ns->tx_dropped = cur_es->tx_discards;
5366 cur_ns->multicast = cur_es->rx_multicast;
5368 /* update some more netdev stats if this is main VSI */
5369 if (vsi->type == ICE_VSI_PF) {
5370 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5371 cur_ns->rx_errors = pf->stats.crc_errors +
5372 pf->stats.illegal_bytes +
5373 pf->stats.rx_len_errors +
5374 pf->stats.rx_undersize +
5375 pf->hw_csum_rx_error +
5376 pf->stats.rx_jabber +
5377 pf->stats.rx_fragments +
5378 pf->stats.rx_oversize;
5379 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5380 /* record drops from the port level */
5381 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5386 * ice_update_pf_stats - Update PF port stats counters
5387 * @pf: PF whose stats needs to be updated
5389 void ice_update_pf_stats(struct ice_pf *pf)
5391 struct ice_hw_port_stats *prev_ps, *cur_ps;
5392 struct ice_hw *hw = &pf->hw;
5396 port = hw->port_info->lport;
5397 prev_ps = &pf->stats_prev;
5398 cur_ps = &pf->stats;
5400 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5401 &prev_ps->eth.rx_bytes,
5402 &cur_ps->eth.rx_bytes);
5404 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5405 &prev_ps->eth.rx_unicast,
5406 &cur_ps->eth.rx_unicast);
5408 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5409 &prev_ps->eth.rx_multicast,
5410 &cur_ps->eth.rx_multicast);
5412 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5413 &prev_ps->eth.rx_broadcast,
5414 &cur_ps->eth.rx_broadcast);
5416 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5417 &prev_ps->eth.rx_discards,
5418 &cur_ps->eth.rx_discards);
5420 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5421 &prev_ps->eth.tx_bytes,
5422 &cur_ps->eth.tx_bytes);
5424 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5425 &prev_ps->eth.tx_unicast,
5426 &cur_ps->eth.tx_unicast);
5428 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5429 &prev_ps->eth.tx_multicast,
5430 &cur_ps->eth.tx_multicast);
5432 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5433 &prev_ps->eth.tx_broadcast,
5434 &cur_ps->eth.tx_broadcast);
5436 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5437 &prev_ps->tx_dropped_link_down,
5438 &cur_ps->tx_dropped_link_down);
5440 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5441 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5443 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5444 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5446 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5447 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5449 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5450 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5452 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5453 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5455 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5456 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5458 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5459 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5461 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5462 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5464 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5465 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5467 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5468 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5470 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5471 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5473 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5474 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5476 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5477 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5479 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5480 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5482 fd_ctr_base = hw->fd_ctr_base;
5484 ice_stat_update40(hw,
5485 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5486 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5487 &cur_ps->fd_sb_match);
5488 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5489 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5491 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5492 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5494 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5495 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5497 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5498 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5500 ice_update_dcb_stats(pf);
5502 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5503 &prev_ps->crc_errors, &cur_ps->crc_errors);
5505 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5506 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5508 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5509 &prev_ps->mac_local_faults,
5510 &cur_ps->mac_local_faults);
5512 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5513 &prev_ps->mac_remote_faults,
5514 &cur_ps->mac_remote_faults);
5516 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5517 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5519 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5520 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5522 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5523 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5525 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5526 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5528 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5529 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5531 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5533 pf->stat_prev_loaded = true;
5537 * ice_get_stats64 - get statistics for network device structure
5538 * @netdev: network interface device structure
5539 * @stats: main device statistics structure
5542 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5544 struct ice_netdev_priv *np = netdev_priv(netdev);
5545 struct rtnl_link_stats64 *vsi_stats;
5546 struct ice_vsi *vsi = np->vsi;
5548 vsi_stats = &vsi->net_stats;
5550 if (!vsi->num_txq || !vsi->num_rxq)
5553 /* netdev packet/byte stats come from ring counter. These are obtained
5554 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5555 * But, only call the update routine and read the registers if VSI is
5558 if (!test_bit(__ICE_DOWN, vsi->state))
5559 ice_update_vsi_ring_stats(vsi);
5560 stats->tx_packets = vsi_stats->tx_packets;
5561 stats->tx_bytes = vsi_stats->tx_bytes;
5562 stats->rx_packets = vsi_stats->rx_packets;
5563 stats->rx_bytes = vsi_stats->rx_bytes;
5565 /* The rest of the stats can be read from the hardware but instead we
5566 * just return values that the watchdog task has already obtained from
5569 stats->multicast = vsi_stats->multicast;
5570 stats->tx_errors = vsi_stats->tx_errors;
5571 stats->tx_dropped = vsi_stats->tx_dropped;
5572 stats->rx_errors = vsi_stats->rx_errors;
5573 stats->rx_dropped = vsi_stats->rx_dropped;
5574 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5575 stats->rx_length_errors = vsi_stats->rx_length_errors;
5579 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5580 * @vsi: VSI having NAPI disabled
5582 static void ice_napi_disable_all(struct ice_vsi *vsi)
5589 ice_for_each_q_vector(vsi, q_idx) {
5590 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5592 if (q_vector->rx.ring || q_vector->tx.ring)
5593 napi_disable(&q_vector->napi);
5598 * ice_down - Shutdown the connection
5599 * @vsi: The VSI being stopped
5601 int ice_down(struct ice_vsi *vsi)
5603 int i, tx_err, rx_err, link_err = 0;
5605 /* Caller of this function is expected to set the
5606 * vsi->state __ICE_DOWN bit
5609 netif_carrier_off(vsi->netdev);
5610 netif_tx_disable(vsi->netdev);
5613 ice_vsi_dis_irq(vsi);
5615 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5617 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5618 vsi->vsi_num, tx_err);
5619 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5620 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5622 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5623 vsi->vsi_num, tx_err);
5626 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5628 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5629 vsi->vsi_num, rx_err);
5631 ice_napi_disable_all(vsi);
5633 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5634 link_err = ice_force_phys_link_state(vsi, false);
5636 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5637 vsi->vsi_num, link_err);
5640 ice_for_each_txq(vsi, i)
5641 ice_clean_tx_ring(vsi->tx_rings[i]);
5643 ice_for_each_rxq(vsi, i)
5644 ice_clean_rx_ring(vsi->rx_rings[i]);
5646 if (tx_err || rx_err || link_err) {
5647 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5648 vsi->vsi_num, vsi->vsw->sw_id);
5656 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5657 * @vsi: VSI having resources allocated
5659 * Return 0 on success, negative on failure
5661 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5665 if (!vsi->num_txq) {
5666 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5671 ice_for_each_txq(vsi, i) {
5672 struct ice_ring *ring = vsi->tx_rings[i];
5677 ring->netdev = vsi->netdev;
5678 err = ice_setup_tx_ring(ring);
5687 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5688 * @vsi: VSI having resources allocated
5690 * Return 0 on success, negative on failure
5692 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5696 if (!vsi->num_rxq) {
5697 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5702 ice_for_each_rxq(vsi, i) {
5703 struct ice_ring *ring = vsi->rx_rings[i];
5708 ring->netdev = vsi->netdev;
5709 err = ice_setup_rx_ring(ring);
5718 * ice_vsi_open_ctrl - open control VSI for use
5719 * @vsi: the VSI to open
5721 * Initialization of the Control VSI
5723 * Returns 0 on success, negative value on error
5725 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5727 char int_name[ICE_INT_NAME_STR_LEN];
5728 struct ice_pf *pf = vsi->back;
5732 dev = ice_pf_to_dev(pf);
5733 /* allocate descriptors */
5734 err = ice_vsi_setup_tx_rings(vsi);
5738 err = ice_vsi_setup_rx_rings(vsi);
5742 err = ice_vsi_cfg(vsi);
5746 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5747 dev_driver_string(dev), dev_name(dev));
5748 err = ice_vsi_req_irq_msix(vsi, int_name);
5752 ice_vsi_cfg_msix(vsi);
5754 err = ice_vsi_start_all_rx_rings(vsi);
5756 goto err_up_complete;
5758 clear_bit(__ICE_DOWN, vsi->state);
5759 ice_vsi_ena_irq(vsi);
5766 ice_vsi_free_rx_rings(vsi);
5768 ice_vsi_free_tx_rings(vsi);
5774 * ice_vsi_open - Called when a network interface is made active
5775 * @vsi: the VSI to open
5777 * Initialization of the VSI
5779 * Returns 0 on success, negative value on error
5781 static int ice_vsi_open(struct ice_vsi *vsi)
5783 char int_name[ICE_INT_NAME_STR_LEN];
5784 struct ice_pf *pf = vsi->back;
5787 /* allocate descriptors */
5788 err = ice_vsi_setup_tx_rings(vsi);
5792 err = ice_vsi_setup_rx_rings(vsi);
5796 err = ice_vsi_cfg(vsi);
5800 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5801 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
5802 err = ice_vsi_req_irq_msix(vsi, int_name);
5806 /* Notify the stack of the actual queue counts. */
5807 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
5811 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
5815 err = ice_up_complete(vsi);
5817 goto err_up_complete;
5824 ice_vsi_free_irq(vsi);
5826 ice_vsi_free_rx_rings(vsi);
5828 ice_vsi_free_tx_rings(vsi);
5834 * ice_vsi_release_all - Delete all VSIs
5835 * @pf: PF from which all VSIs are being removed
5837 static void ice_vsi_release_all(struct ice_pf *pf)
5844 ice_for_each_vsi(pf, i) {
5848 err = ice_vsi_release(pf->vsi[i]);
5850 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
5851 i, err, pf->vsi[i]->vsi_num);
5856 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
5857 * @pf: pointer to the PF instance
5858 * @type: VSI type to rebuild
5860 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
5862 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
5864 struct device *dev = ice_pf_to_dev(pf);
5865 enum ice_status status;
5868 ice_for_each_vsi(pf, i) {
5869 struct ice_vsi *vsi = pf->vsi[i];
5871 if (!vsi || vsi->type != type)
5874 /* rebuild the VSI */
5875 err = ice_vsi_rebuild(vsi, true);
5877 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
5878 err, vsi->idx, ice_vsi_type_str(type));
5882 /* replay filters for the VSI */
5883 status = ice_replay_vsi(&pf->hw, vsi->idx);
5885 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
5886 ice_stat_str(status), vsi->idx,
5887 ice_vsi_type_str(type));
5891 /* Re-map HW VSI number, using VSI handle that has been
5892 * previously validated in ice_replay_vsi() call above
5894 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
5896 /* enable the VSI */
5897 err = ice_ena_vsi(vsi, false);
5899 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
5900 err, vsi->idx, ice_vsi_type_str(type));
5904 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
5905 ice_vsi_type_str(type));
5912 * ice_update_pf_netdev_link - Update PF netdev link status
5913 * @pf: pointer to the PF instance
5915 static void ice_update_pf_netdev_link(struct ice_pf *pf)
5920 ice_for_each_vsi(pf, i) {
5921 struct ice_vsi *vsi = pf->vsi[i];
5923 if (!vsi || vsi->type != ICE_VSI_PF)
5926 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
5928 netif_carrier_on(pf->vsi[i]->netdev);
5929 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
5931 netif_carrier_off(pf->vsi[i]->netdev);
5932 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
5938 * ice_rebuild - rebuild after reset
5939 * @pf: PF to rebuild
5940 * @reset_type: type of reset
5942 * Do not rebuild VF VSI in this flow because that is already handled via
5943 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
5944 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
5945 * to reset/rebuild all the VF VSI twice.
5947 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
5949 struct device *dev = ice_pf_to_dev(pf);
5950 struct ice_hw *hw = &pf->hw;
5951 enum ice_status ret;
5954 if (test_bit(__ICE_DOWN, pf->state))
5955 goto clear_recovery;
5957 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
5959 ret = ice_init_all_ctrlq(hw);
5961 dev_err(dev, "control queues init failed %s\n",
5963 goto err_init_ctrlq;
5966 /* if DDP was previously loaded successfully */
5967 if (!ice_is_safe_mode(pf)) {
5968 /* reload the SW DB of filter tables */
5969 if (reset_type == ICE_RESET_PFR)
5970 ice_fill_blk_tbls(hw);
5972 /* Reload DDP Package after CORER/GLOBR reset */
5973 ice_load_pkg(NULL, pf);
5976 ret = ice_clear_pf_cfg(hw);
5978 dev_err(dev, "clear PF configuration failed %s\n",
5980 goto err_init_ctrlq;
5983 if (pf->first_sw->dflt_vsi_ena)
5984 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
5985 /* clear the default VSI configuration if it exists */
5986 pf->first_sw->dflt_vsi = NULL;
5987 pf->first_sw->dflt_vsi_ena = false;
5989 ice_clear_pxe_mode(hw);
5991 ret = ice_get_caps(hw);
5993 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
5994 goto err_init_ctrlq;
5997 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
5999 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6000 goto err_init_ctrlq;
6003 err = ice_sched_init_port(hw->port_info);
6005 goto err_sched_init_port;
6007 /* start misc vector */
6008 err = ice_req_irq_msix_misc(pf);
6010 dev_err(dev, "misc vector setup failed: %d\n", err);
6011 goto err_sched_init_port;
6014 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6015 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6016 if (!rd32(hw, PFQF_FD_SIZE)) {
6017 u16 unused, guar, b_effort;
6019 guar = hw->func_caps.fd_fltr_guar;
6020 b_effort = hw->func_caps.fd_fltr_best_effort;
6022 /* force guaranteed filter pool for PF */
6023 ice_alloc_fd_guar_item(hw, &unused, guar);
6024 /* force shared filter pool for PF */
6025 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6029 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6030 ice_dcb_rebuild(pf);
6032 /* rebuild PF VSI */
6033 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6035 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6036 goto err_vsi_rebuild;
6039 /* If Flow Director is active */
6040 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6041 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6043 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6044 goto err_vsi_rebuild;
6047 /* replay HW Flow Director recipes */
6049 ice_fdir_replay_flows(hw);
6051 /* replay Flow Director filters */
6052 ice_fdir_replay_fltrs(pf);
6054 ice_rebuild_arfs(pf);
6057 ice_update_pf_netdev_link(pf);
6059 /* tell the firmware we are up */
6060 ret = ice_send_version(pf);
6062 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6064 goto err_vsi_rebuild;
6067 ice_replay_post(hw);
6069 /* if we get here, reset flow is successful */
6070 clear_bit(__ICE_RESET_FAILED, pf->state);
6074 err_sched_init_port:
6075 ice_sched_cleanup_all(hw);
6077 ice_shutdown_all_ctrlq(hw);
6078 set_bit(__ICE_RESET_FAILED, pf->state);
6080 /* set this bit in PF state to control service task scheduling */
6081 set_bit(__ICE_NEEDS_RESTART, pf->state);
6082 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6086 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6087 * @vsi: Pointer to VSI structure
6089 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6091 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6092 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6094 return ICE_RXBUF_3072;
6098 * ice_change_mtu - NDO callback to change the MTU
6099 * @netdev: network interface device structure
6100 * @new_mtu: new value for maximum frame size
6102 * Returns 0 on success, negative on failure
6104 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6106 struct ice_netdev_priv *np = netdev_priv(netdev);
6107 struct ice_vsi *vsi = np->vsi;
6108 struct ice_pf *pf = vsi->back;
6111 if (new_mtu == (int)netdev->mtu) {
6112 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6116 if (ice_is_xdp_ena_vsi(vsi)) {
6117 int frame_size = ice_max_xdp_frame_size(vsi);
6119 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6120 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6121 frame_size - ICE_ETH_PKT_HDR_PAD);
6126 if (new_mtu < (int)netdev->min_mtu) {
6127 netdev_err(netdev, "new MTU invalid. min_mtu is %d\n",
6130 } else if (new_mtu > (int)netdev->max_mtu) {
6131 netdev_err(netdev, "new MTU invalid. max_mtu is %d\n",
6135 /* if a reset is in progress, wait for some time for it to complete */
6137 if (ice_is_reset_in_progress(pf->state)) {
6139 usleep_range(1000, 2000);
6144 } while (count < 100);
6147 netdev_err(netdev, "can't change MTU. Device is busy\n");
6151 netdev->mtu = (unsigned int)new_mtu;
6153 /* if VSI is up, bring it down and then back up */
6154 if (!test_and_set_bit(__ICE_DOWN, vsi->state)) {
6157 err = ice_down(vsi);
6159 netdev_err(netdev, "change MTU if_up err %d\n", err);
6165 netdev_err(netdev, "change MTU if_up err %d\n", err);
6170 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6175 * ice_aq_str - convert AQ err code to a string
6176 * @aq_err: the AQ error code to convert
6178 const char *ice_aq_str(enum ice_aq_err aq_err)
6183 case ICE_AQ_RC_EPERM:
6184 return "ICE_AQ_RC_EPERM";
6185 case ICE_AQ_RC_ENOENT:
6186 return "ICE_AQ_RC_ENOENT";
6187 case ICE_AQ_RC_ENOMEM:
6188 return "ICE_AQ_RC_ENOMEM";
6189 case ICE_AQ_RC_EBUSY:
6190 return "ICE_AQ_RC_EBUSY";
6191 case ICE_AQ_RC_EEXIST:
6192 return "ICE_AQ_RC_EEXIST";
6193 case ICE_AQ_RC_EINVAL:
6194 return "ICE_AQ_RC_EINVAL";
6195 case ICE_AQ_RC_ENOSPC:
6196 return "ICE_AQ_RC_ENOSPC";
6197 case ICE_AQ_RC_ENOSYS:
6198 return "ICE_AQ_RC_ENOSYS";
6199 case ICE_AQ_RC_EMODE:
6200 return "ICE_AQ_RC_EMODE";
6201 case ICE_AQ_RC_ENOSEC:
6202 return "ICE_AQ_RC_ENOSEC";
6203 case ICE_AQ_RC_EBADSIG:
6204 return "ICE_AQ_RC_EBADSIG";
6205 case ICE_AQ_RC_ESVN:
6206 return "ICE_AQ_RC_ESVN";
6207 case ICE_AQ_RC_EBADMAN:
6208 return "ICE_AQ_RC_EBADMAN";
6209 case ICE_AQ_RC_EBADBUF:
6210 return "ICE_AQ_RC_EBADBUF";
6213 return "ICE_AQ_RC_UNKNOWN";
6217 * ice_stat_str - convert status err code to a string
6218 * @stat_err: the status error code to convert
6220 const char *ice_stat_str(enum ice_status stat_err)
6226 return "ICE_ERR_PARAM";
6227 case ICE_ERR_NOT_IMPL:
6228 return "ICE_ERR_NOT_IMPL";
6229 case ICE_ERR_NOT_READY:
6230 return "ICE_ERR_NOT_READY";
6231 case ICE_ERR_NOT_SUPPORTED:
6232 return "ICE_ERR_NOT_SUPPORTED";
6233 case ICE_ERR_BAD_PTR:
6234 return "ICE_ERR_BAD_PTR";
6235 case ICE_ERR_INVAL_SIZE:
6236 return "ICE_ERR_INVAL_SIZE";
6237 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6238 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6239 case ICE_ERR_RESET_FAILED:
6240 return "ICE_ERR_RESET_FAILED";
6241 case ICE_ERR_FW_API_VER:
6242 return "ICE_ERR_FW_API_VER";
6243 case ICE_ERR_NO_MEMORY:
6244 return "ICE_ERR_NO_MEMORY";
6246 return "ICE_ERR_CFG";
6247 case ICE_ERR_OUT_OF_RANGE:
6248 return "ICE_ERR_OUT_OF_RANGE";
6249 case ICE_ERR_ALREADY_EXISTS:
6250 return "ICE_ERR_ALREADY_EXISTS";
6251 case ICE_ERR_NVM_CHECKSUM:
6252 return "ICE_ERR_NVM_CHECKSUM";
6253 case ICE_ERR_BUF_TOO_SHORT:
6254 return "ICE_ERR_BUF_TOO_SHORT";
6255 case ICE_ERR_NVM_BLANK_MODE:
6256 return "ICE_ERR_NVM_BLANK_MODE";
6257 case ICE_ERR_IN_USE:
6258 return "ICE_ERR_IN_USE";
6259 case ICE_ERR_MAX_LIMIT:
6260 return "ICE_ERR_MAX_LIMIT";
6261 case ICE_ERR_RESET_ONGOING:
6262 return "ICE_ERR_RESET_ONGOING";
6263 case ICE_ERR_HW_TABLE:
6264 return "ICE_ERR_HW_TABLE";
6265 case ICE_ERR_DOES_NOT_EXIST:
6266 return "ICE_ERR_DOES_NOT_EXIST";
6267 case ICE_ERR_FW_DDP_MISMATCH:
6268 return "ICE_ERR_FW_DDP_MISMATCH";
6269 case ICE_ERR_AQ_ERROR:
6270 return "ICE_ERR_AQ_ERROR";
6271 case ICE_ERR_AQ_TIMEOUT:
6272 return "ICE_ERR_AQ_TIMEOUT";
6273 case ICE_ERR_AQ_FULL:
6274 return "ICE_ERR_AQ_FULL";
6275 case ICE_ERR_AQ_NO_WORK:
6276 return "ICE_ERR_AQ_NO_WORK";
6277 case ICE_ERR_AQ_EMPTY:
6278 return "ICE_ERR_AQ_EMPTY";
6279 case ICE_ERR_AQ_FW_CRITICAL:
6280 return "ICE_ERR_AQ_FW_CRITICAL";
6283 return "ICE_ERR_UNKNOWN";
6287 * ice_set_rss - Set RSS keys and lut
6288 * @vsi: Pointer to VSI structure
6289 * @seed: RSS hash seed
6290 * @lut: Lookup table
6291 * @lut_size: Lookup table size
6293 * Returns 0 on success, negative on failure
6295 int ice_set_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6297 struct ice_pf *pf = vsi->back;
6298 struct ice_hw *hw = &pf->hw;
6299 enum ice_status status;
6302 dev = ice_pf_to_dev(pf);
6304 struct ice_aqc_get_set_rss_keys *buf =
6305 (struct ice_aqc_get_set_rss_keys *)seed;
6307 status = ice_aq_set_rss_key(hw, vsi->idx, buf);
6310 dev_err(dev, "Cannot set RSS key, err %s aq_err %s\n",
6311 ice_stat_str(status),
6312 ice_aq_str(hw->adminq.sq_last_status));
6318 status = ice_aq_set_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6321 dev_err(dev, "Cannot set RSS lut, err %s aq_err %s\n",
6322 ice_stat_str(status),
6323 ice_aq_str(hw->adminq.sq_last_status));
6332 * ice_get_rss - Get RSS keys and lut
6333 * @vsi: Pointer to VSI structure
6334 * @seed: Buffer to store the keys
6335 * @lut: Buffer to store the lookup table entries
6336 * @lut_size: Size of buffer to store the lookup table entries
6338 * Returns 0 on success, negative on failure
6340 int ice_get_rss(struct ice_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
6342 struct ice_pf *pf = vsi->back;
6343 struct ice_hw *hw = &pf->hw;
6344 enum ice_status status;
6347 dev = ice_pf_to_dev(pf);
6349 struct ice_aqc_get_set_rss_keys *buf =
6350 (struct ice_aqc_get_set_rss_keys *)seed;
6352 status = ice_aq_get_rss_key(hw, vsi->idx, buf);
6354 dev_err(dev, "Cannot get RSS key, err %s aq_err %s\n",
6355 ice_stat_str(status),
6356 ice_aq_str(hw->adminq.sq_last_status));
6362 status = ice_aq_get_rss_lut(hw, vsi->idx, vsi->rss_lut_type,
6365 dev_err(dev, "Cannot get RSS lut, err %s aq_err %s\n",
6366 ice_stat_str(status),
6367 ice_aq_str(hw->adminq.sq_last_status));
6376 * ice_bridge_getlink - Get the hardware bridge mode
6379 * @seq: RTNL message seq
6380 * @dev: the netdev being configured
6381 * @filter_mask: filter mask passed in
6382 * @nlflags: netlink flags passed in
6384 * Return the bridge mode (VEB/VEPA)
6387 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6388 struct net_device *dev, u32 filter_mask, int nlflags)
6390 struct ice_netdev_priv *np = netdev_priv(dev);
6391 struct ice_vsi *vsi = np->vsi;
6392 struct ice_pf *pf = vsi->back;
6395 bmode = pf->first_sw->bridge_mode;
6397 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6402 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6403 * @vsi: Pointer to VSI structure
6404 * @bmode: Hardware bridge mode (VEB/VEPA)
6406 * Returns 0 on success, negative on failure
6408 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6410 struct ice_aqc_vsi_props *vsi_props;
6411 struct ice_hw *hw = &vsi->back->hw;
6412 struct ice_vsi_ctx *ctxt;
6413 enum ice_status status;
6416 vsi_props = &vsi->info;
6418 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6422 ctxt->info = vsi->info;
6424 if (bmode == BRIDGE_MODE_VEB)
6425 /* change from VEPA to VEB mode */
6426 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6428 /* change from VEB to VEPA mode */
6429 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6430 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6432 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6434 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6435 bmode, ice_stat_str(status),
6436 ice_aq_str(hw->adminq.sq_last_status));
6440 /* Update sw flags for book keeping */
6441 vsi_props->sw_flags = ctxt->info.sw_flags;
6449 * ice_bridge_setlink - Set the hardware bridge mode
6450 * @dev: the netdev being configured
6451 * @nlh: RTNL message
6452 * @flags: bridge setlink flags
6453 * @extack: netlink extended ack
6455 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6456 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6457 * not already set for all VSIs connected to this switch. And also update the
6458 * unicast switch filter rules for the corresponding switch of the netdev.
6461 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6462 u16 __always_unused flags,
6463 struct netlink_ext_ack __always_unused *extack)
6465 struct ice_netdev_priv *np = netdev_priv(dev);
6466 struct ice_pf *pf = np->vsi->back;
6467 struct nlattr *attr, *br_spec;
6468 struct ice_hw *hw = &pf->hw;
6469 enum ice_status status;
6470 struct ice_sw *pf_sw;
6471 int rem, v, err = 0;
6473 pf_sw = pf->first_sw;
6474 /* find the attribute in the netlink message */
6475 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6477 nla_for_each_nested(attr, br_spec, rem) {
6480 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6482 mode = nla_get_u16(attr);
6483 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6485 /* Continue if bridge mode is not being flipped */
6486 if (mode == pf_sw->bridge_mode)
6488 /* Iterates through the PF VSI list and update the loopback
6491 ice_for_each_vsi(pf, v) {
6494 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6499 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6500 /* Update the unicast switch filter rules for the corresponding
6501 * switch of the netdev
6503 status = ice_update_sw_rule_bridge_mode(hw);
6505 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6506 mode, ice_stat_str(status),
6507 ice_aq_str(hw->adminq.sq_last_status));
6508 /* revert hw->evb_veb */
6509 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6513 pf_sw->bridge_mode = mode;
6520 * ice_tx_timeout - Respond to a Tx Hang
6521 * @netdev: network interface device structure
6522 * @txqueue: Tx queue
6524 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6526 struct ice_netdev_priv *np = netdev_priv(netdev);
6527 struct ice_ring *tx_ring = NULL;
6528 struct ice_vsi *vsi = np->vsi;
6529 struct ice_pf *pf = vsi->back;
6532 pf->tx_timeout_count++;
6534 /* Check if PFC is enabled for the TC to which the queue belongs
6535 * to. If yes then Tx timeout is not caused by a hung queue, no
6536 * need to reset and rebuild
6538 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6539 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6544 /* now that we have an index, find the tx_ring struct */
6545 for (i = 0; i < vsi->num_txq; i++)
6546 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6547 if (txqueue == vsi->tx_rings[i]->q_index) {
6548 tx_ring = vsi->tx_rings[i];
6552 /* Reset recovery level if enough time has elapsed after last timeout.
6553 * Also ensure no new reset action happens before next timeout period.
6555 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6556 pf->tx_timeout_recovery_level = 1;
6557 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6558 netdev->watchdog_timeo)))
6562 struct ice_hw *hw = &pf->hw;
6565 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6566 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6567 /* Read interrupt register */
6568 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6570 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6571 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6572 head, tx_ring->next_to_use, val);
6575 pf->tx_timeout_last_recovery = jiffies;
6576 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6577 pf->tx_timeout_recovery_level, txqueue);
6579 switch (pf->tx_timeout_recovery_level) {
6581 set_bit(__ICE_PFR_REQ, pf->state);
6584 set_bit(__ICE_CORER_REQ, pf->state);
6587 set_bit(__ICE_GLOBR_REQ, pf->state);
6590 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6591 set_bit(__ICE_DOWN, pf->state);
6592 set_bit(__ICE_NEEDS_RESTART, vsi->state);
6593 set_bit(__ICE_SERVICE_DIS, pf->state);
6597 ice_service_task_schedule(pf);
6598 pf->tx_timeout_recovery_level++;
6602 * ice_open - Called when a network interface becomes active
6603 * @netdev: network interface device structure
6605 * The open entry point is called when a network interface is made
6606 * active by the system (IFF_UP). At this point all resources needed
6607 * for transmit and receive operations are allocated, the interrupt
6608 * handler is registered with the OS, the netdev watchdog is enabled,
6609 * and the stack is notified that the interface is ready.
6611 * Returns 0 on success, negative value on failure
6613 int ice_open(struct net_device *netdev)
6615 struct ice_netdev_priv *np = netdev_priv(netdev);
6616 struct ice_vsi *vsi = np->vsi;
6617 struct ice_pf *pf = vsi->back;
6618 struct ice_port_info *pi;
6621 if (test_bit(__ICE_NEEDS_RESTART, pf->state)) {
6622 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6626 if (test_bit(__ICE_DOWN, pf->state)) {
6627 netdev_err(netdev, "device is not ready yet\n");
6631 netif_carrier_off(netdev);
6633 pi = vsi->port_info;
6634 err = ice_update_link_info(pi);
6636 netdev_err(netdev, "Failed to get link info, error %d\n",
6641 /* Set PHY if there is media, otherwise, turn off PHY */
6642 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6643 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6644 if (!test_bit(__ICE_PHY_INIT_COMPLETE, pf->state)) {
6645 err = ice_init_phy_user_cfg(pi);
6647 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6653 err = ice_configure_phy(vsi);
6655 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6660 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6661 err = ice_aq_set_link_restart_an(pi, false, NULL);
6663 netdev_err(netdev, "Failed to set PHY state, VSI %d error %d\n",
6669 err = ice_vsi_open(vsi);
6671 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6672 vsi->vsi_num, vsi->vsw->sw_id);
6674 /* Update existing tunnels information */
6675 udp_tunnel_get_rx_info(netdev);
6681 * ice_stop - Disables a network interface
6682 * @netdev: network interface device structure
6684 * The stop entry point is called when an interface is de-activated by the OS,
6685 * and the netdevice enters the DOWN state. The hardware is still under the
6686 * driver's control, but the netdev interface is disabled.
6688 * Returns success only - not allowed to fail
6690 int ice_stop(struct net_device *netdev)
6692 struct ice_netdev_priv *np = netdev_priv(netdev);
6693 struct ice_vsi *vsi = np->vsi;
6701 * ice_features_check - Validate encapsulated packet conforms to limits
6703 * @netdev: This port's netdev
6704 * @features: Offload features that the stack believes apply
6706 static netdev_features_t
6707 ice_features_check(struct sk_buff *skb,
6708 struct net_device __always_unused *netdev,
6709 netdev_features_t features)
6713 /* No point in doing any of this if neither checksum nor GSO are
6714 * being requested for this frame. We can rule out both by just
6715 * checking for CHECKSUM_PARTIAL
6717 if (skb->ip_summed != CHECKSUM_PARTIAL)
6720 /* We cannot support GSO if the MSS is going to be less than
6721 * 64 bytes. If it is then we need to drop support for GSO.
6723 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
6724 features &= ~NETIF_F_GSO_MASK;
6726 len = skb_network_header(skb) - skb->data;
6727 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
6728 goto out_rm_features;
6730 len = skb_transport_header(skb) - skb_network_header(skb);
6731 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6732 goto out_rm_features;
6734 if (skb->encapsulation) {
6735 len = skb_inner_network_header(skb) - skb_transport_header(skb);
6736 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
6737 goto out_rm_features;
6739 len = skb_inner_transport_header(skb) -
6740 skb_inner_network_header(skb);
6741 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6742 goto out_rm_features;
6747 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
6750 static const struct net_device_ops ice_netdev_safe_mode_ops = {
6751 .ndo_open = ice_open,
6752 .ndo_stop = ice_stop,
6753 .ndo_start_xmit = ice_start_xmit,
6754 .ndo_set_mac_address = ice_set_mac_address,
6755 .ndo_validate_addr = eth_validate_addr,
6756 .ndo_change_mtu = ice_change_mtu,
6757 .ndo_get_stats64 = ice_get_stats64,
6758 .ndo_tx_timeout = ice_tx_timeout,
6761 static const struct net_device_ops ice_netdev_ops = {
6762 .ndo_open = ice_open,
6763 .ndo_stop = ice_stop,
6764 .ndo_start_xmit = ice_start_xmit,
6765 .ndo_features_check = ice_features_check,
6766 .ndo_set_rx_mode = ice_set_rx_mode,
6767 .ndo_set_mac_address = ice_set_mac_address,
6768 .ndo_validate_addr = eth_validate_addr,
6769 .ndo_change_mtu = ice_change_mtu,
6770 .ndo_get_stats64 = ice_get_stats64,
6771 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
6772 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
6773 .ndo_set_vf_mac = ice_set_vf_mac,
6774 .ndo_get_vf_config = ice_get_vf_cfg,
6775 .ndo_set_vf_trust = ice_set_vf_trust,
6776 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
6777 .ndo_set_vf_link_state = ice_set_vf_link_state,
6778 .ndo_get_vf_stats = ice_get_vf_stats,
6779 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
6780 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
6781 .ndo_set_features = ice_set_features,
6782 .ndo_bridge_getlink = ice_bridge_getlink,
6783 .ndo_bridge_setlink = ice_bridge_setlink,
6784 .ndo_fdb_add = ice_fdb_add,
6785 .ndo_fdb_del = ice_fdb_del,
6786 #ifdef CONFIG_RFS_ACCEL
6787 .ndo_rx_flow_steer = ice_rx_flow_steer,
6789 .ndo_tx_timeout = ice_tx_timeout,
6791 .ndo_xdp_xmit = ice_xdp_xmit,
6792 .ndo_xsk_wakeup = ice_xsk_wakeup,
6793 .ndo_udp_tunnel_add = udp_tunnel_nic_add_port,
6794 .ndo_udp_tunnel_del = udp_tunnel_nic_del_port,