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"
25 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
26 MODULE_DESCRIPTION(DRV_SUMMARY);
27 MODULE_LICENSE("GPL v2");
28 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
30 static int debug = -1;
31 module_param(debug, int, 0644);
32 #ifndef CONFIG_DYNAMIC_DEBUG
33 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
35 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
36 #endif /* !CONFIG_DYNAMIC_DEBUG */
38 static struct workqueue_struct *ice_wq;
39 static const struct net_device_ops ice_netdev_safe_mode_ops;
40 static const struct net_device_ops ice_netdev_ops;
41 static int ice_vsi_open(struct ice_vsi *vsi);
43 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
45 static void ice_vsi_release_all(struct ice_pf *pf);
47 bool netif_is_ice(struct net_device *dev)
49 return dev && (dev->netdev_ops == &ice_netdev_ops);
53 * ice_get_tx_pending - returns number of Tx descriptors not processed
54 * @ring: the ring of descriptors
56 static u16 ice_get_tx_pending(struct ice_ring *ring)
60 head = ring->next_to_clean;
61 tail = ring->next_to_use;
64 return (head < tail) ?
65 tail - head : (tail + ring->count - head);
70 * ice_check_for_hang_subtask - check for and recover hung queues
71 * @pf: pointer to PF struct
73 static void ice_check_for_hang_subtask(struct ice_pf *pf)
75 struct ice_vsi *vsi = NULL;
81 ice_for_each_vsi(pf, v)
82 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
87 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
90 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
95 for (i = 0; i < vsi->num_txq; i++) {
96 struct ice_ring *tx_ring = vsi->tx_rings[i];
98 if (tx_ring && tx_ring->desc) {
99 /* If packet counter has not changed the queue is
100 * likely stalled, so force an interrupt for this
103 * prev_pkt would be negative if there was no
106 packets = tx_ring->stats.pkts & INT_MAX;
107 if (tx_ring->tx_stats.prev_pkt == packets) {
108 /* Trigger sw interrupt to revive the queue */
109 ice_trigger_sw_intr(hw, tx_ring->q_vector);
113 /* Memory barrier between read of packet count and call
114 * to ice_get_tx_pending()
117 tx_ring->tx_stats.prev_pkt =
118 ice_get_tx_pending(tx_ring) ? packets : -1;
124 * ice_init_mac_fltr - Set initial MAC filters
125 * @pf: board private structure
127 * Set initial set of MAC filters for PF VSI; configure filters for permanent
128 * address and broadcast address. If an error is encountered, netdevice will be
131 static int ice_init_mac_fltr(struct ice_pf *pf)
133 enum ice_status status;
137 vsi = ice_get_main_vsi(pf);
141 perm_addr = vsi->port_info->mac.perm_addr;
142 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
150 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
151 * @netdev: the net device on which the sync is happening
152 * @addr: MAC address to sync
154 * This is a callback function which is called by the in kernel device sync
155 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
156 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
157 * MAC filters from the hardware.
159 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
161 struct ice_netdev_priv *np = netdev_priv(netdev);
162 struct ice_vsi *vsi = np->vsi;
164 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
172 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
173 * @netdev: the net device on which the unsync is happening
174 * @addr: MAC address to unsync
176 * This is a callback function which is called by the in kernel device unsync
177 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
178 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
179 * delete the MAC filters from the hardware.
181 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
183 struct ice_netdev_priv *np = netdev_priv(netdev);
184 struct ice_vsi *vsi = np->vsi;
186 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
194 * ice_vsi_fltr_changed - check if filter state changed
195 * @vsi: VSI to be checked
197 * returns true if filter state has changed, false otherwise.
199 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
201 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
202 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
203 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
207 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
208 * @vsi: the VSI being configured
209 * @promisc_m: mask of promiscuous config bits
210 * @set_promisc: enable or disable promisc flag request
213 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
215 struct ice_hw *hw = &vsi->back->hw;
216 enum ice_status status = 0;
218 if (vsi->type != ICE_VSI_PF)
221 if (vsi->num_vlan > 1) {
222 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
226 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
229 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
240 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
241 * @vsi: ptr to the VSI
243 * Push any outstanding VSI filter changes through the AdminQ.
245 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
247 struct device *dev = ice_pf_to_dev(vsi->back);
248 struct net_device *netdev = vsi->netdev;
249 bool promisc_forced_on = false;
250 struct ice_pf *pf = vsi->back;
251 struct ice_hw *hw = &pf->hw;
252 enum ice_status status = 0;
253 u32 changed_flags = 0;
260 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
261 usleep_range(1000, 2000);
263 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
264 vsi->current_netdev_flags = vsi->netdev->flags;
266 INIT_LIST_HEAD(&vsi->tmp_sync_list);
267 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
269 if (ice_vsi_fltr_changed(vsi)) {
270 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
271 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
272 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
274 /* grab the netdev's addr_list_lock */
275 netif_addr_lock_bh(netdev);
276 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
277 ice_add_mac_to_unsync_list);
278 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
279 ice_add_mac_to_unsync_list);
280 /* our temp lists are populated. release lock */
281 netif_addr_unlock_bh(netdev);
284 /* Remove MAC addresses in the unsync list */
285 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
286 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
288 netdev_err(netdev, "Failed to delete MAC filters\n");
289 /* if we failed because of alloc failures, just bail */
290 if (status == ICE_ERR_NO_MEMORY) {
296 /* Add MAC addresses in the sync list */
297 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
298 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
299 /* If filter is added successfully or already exists, do not go into
300 * 'if' condition and report it as error. Instead continue processing
301 * rest of the function.
303 if (status && status != ICE_ERR_ALREADY_EXISTS) {
304 netdev_err(netdev, "Failed to add MAC filters\n");
305 /* If there is no more space for new umac filters, VSI
306 * should go into promiscuous mode. There should be some
307 * space reserved for promiscuous filters.
309 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
310 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
312 promisc_forced_on = true;
313 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
320 /* check for changes in promiscuous modes */
321 if (changed_flags & IFF_ALLMULTI) {
322 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
323 if (vsi->num_vlan > 1)
324 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
326 promisc_m = ICE_MCAST_PROMISC_BITS;
328 err = ice_cfg_promisc(vsi, promisc_m, true);
330 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
332 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
336 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
337 if (vsi->num_vlan > 1)
338 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
340 promisc_m = ICE_MCAST_PROMISC_BITS;
342 err = ice_cfg_promisc(vsi, promisc_m, false);
344 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
346 vsi->current_netdev_flags |= IFF_ALLMULTI;
352 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
353 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
354 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
355 if (vsi->current_netdev_flags & IFF_PROMISC) {
356 /* Apply Rx filter rule to get traffic from wire */
357 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
358 err = ice_set_dflt_vsi(pf->first_sw, vsi);
359 if (err && err != -EEXIST) {
360 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
362 vsi->current_netdev_flags &=
366 ice_cfg_vlan_pruning(vsi, false, false);
369 /* Clear Rx filter to remove traffic from wire */
370 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
371 err = ice_clear_dflt_vsi(pf->first_sw);
373 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
375 vsi->current_netdev_flags |=
379 if (vsi->num_vlan > 1)
380 ice_cfg_vlan_pruning(vsi, true, false);
387 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
390 /* if something went wrong then set the changed flag so we try again */
391 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
392 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
394 clear_bit(ICE_CFG_BUSY, vsi->state);
399 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
400 * @pf: board private structure
402 static void ice_sync_fltr_subtask(struct ice_pf *pf)
406 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
409 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
411 ice_for_each_vsi(pf, v)
412 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
413 ice_vsi_sync_fltr(pf->vsi[v])) {
414 /* come back and try again later */
415 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
421 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
423 * @locked: is the rtnl_lock already held
425 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
430 ice_for_each_vsi(pf, v)
432 ice_dis_vsi(pf->vsi[v], locked);
434 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
435 pf->pf_agg_node[node].num_vsis = 0;
437 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
438 pf->vf_agg_node[node].num_vsis = 0;
442 * ice_prepare_for_reset - prep for the core to reset
443 * @pf: board private structure
445 * Inform or close all dependent features in prep for reset.
448 ice_prepare_for_reset(struct ice_pf *pf)
450 struct ice_hw *hw = &pf->hw;
453 /* already prepared for reset */
454 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
457 /* Notify VFs of impending reset */
458 if (ice_check_sq_alive(hw, &hw->mailboxq))
459 ice_vc_notify_reset(pf);
461 /* Disable VFs until reset is completed */
462 ice_for_each_vf(pf, i)
463 ice_set_vf_state_qs_dis(&pf->vf[i]);
465 /* clear SW filtering DB */
466 ice_clear_hw_tbls(hw);
467 /* disable the VSIs and their queues that are not already DOWN */
468 ice_pf_dis_all_vsi(pf, false);
471 ice_sched_clear_port(hw->port_info);
473 ice_shutdown_all_ctrlq(hw);
475 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
479 * ice_do_reset - Initiate one of many types of resets
480 * @pf: board private structure
481 * @reset_type: reset type requested
482 * before this function was called.
484 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
486 struct device *dev = ice_pf_to_dev(pf);
487 struct ice_hw *hw = &pf->hw;
489 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
491 ice_prepare_for_reset(pf);
493 /* trigger the reset */
494 if (ice_reset(hw, reset_type)) {
495 dev_err(dev, "reset %d failed\n", reset_type);
496 set_bit(ICE_RESET_FAILED, pf->state);
497 clear_bit(ICE_RESET_OICR_RECV, pf->state);
498 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
499 clear_bit(ICE_PFR_REQ, pf->state);
500 clear_bit(ICE_CORER_REQ, pf->state);
501 clear_bit(ICE_GLOBR_REQ, pf->state);
505 /* PFR is a bit of a special case because it doesn't result in an OICR
506 * interrupt. So for PFR, rebuild after the reset and clear the reset-
507 * associated state bits.
509 if (reset_type == ICE_RESET_PFR) {
511 ice_rebuild(pf, reset_type);
512 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
513 clear_bit(ICE_PFR_REQ, pf->state);
514 ice_reset_all_vfs(pf, true);
519 * ice_reset_subtask - Set up for resetting the device and driver
520 * @pf: board private structure
522 static void ice_reset_subtask(struct ice_pf *pf)
524 enum ice_reset_req reset_type = ICE_RESET_INVAL;
526 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
527 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
528 * of reset is pending and sets bits in pf->state indicating the reset
529 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
530 * prepare for pending reset if not already (for PF software-initiated
531 * global resets the software should already be prepared for it as
532 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
533 * by firmware or software on other PFs, that bit is not set so prepare
534 * for the reset now), poll for reset done, rebuild and return.
536 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
537 /* Perform the largest reset requested */
538 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
539 reset_type = ICE_RESET_CORER;
540 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
541 reset_type = ICE_RESET_GLOBR;
542 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
543 reset_type = ICE_RESET_EMPR;
544 /* return if no valid reset type requested */
545 if (reset_type == ICE_RESET_INVAL)
547 ice_prepare_for_reset(pf);
549 /* make sure we are ready to rebuild */
550 if (ice_check_reset(&pf->hw)) {
551 set_bit(ICE_RESET_FAILED, pf->state);
553 /* done with reset. start rebuild */
554 pf->hw.reset_ongoing = false;
555 ice_rebuild(pf, reset_type);
556 /* clear bit to resume normal operations, but
557 * ICE_NEEDS_RESTART bit is set in case rebuild failed
559 clear_bit(ICE_RESET_OICR_RECV, pf->state);
560 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
561 clear_bit(ICE_PFR_REQ, pf->state);
562 clear_bit(ICE_CORER_REQ, pf->state);
563 clear_bit(ICE_GLOBR_REQ, pf->state);
564 ice_reset_all_vfs(pf, true);
570 /* No pending resets to finish processing. Check for new resets */
571 if (test_bit(ICE_PFR_REQ, pf->state))
572 reset_type = ICE_RESET_PFR;
573 if (test_bit(ICE_CORER_REQ, pf->state))
574 reset_type = ICE_RESET_CORER;
575 if (test_bit(ICE_GLOBR_REQ, pf->state))
576 reset_type = ICE_RESET_GLOBR;
577 /* If no valid reset type requested just return */
578 if (reset_type == ICE_RESET_INVAL)
581 /* reset if not already down or busy */
582 if (!test_bit(ICE_DOWN, pf->state) &&
583 !test_bit(ICE_CFG_BUSY, pf->state)) {
584 ice_do_reset(pf, reset_type);
589 * ice_print_topo_conflict - print topology conflict message
590 * @vsi: the VSI whose topology status is being checked
592 static void ice_print_topo_conflict(struct ice_vsi *vsi)
594 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
595 case ICE_AQ_LINK_TOPO_CONFLICT:
596 case ICE_AQ_LINK_MEDIA_CONFLICT:
597 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
598 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
599 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
600 netdev_info(vsi->netdev, "Potential misconfiguration of the Ethernet port detected. If it was not intended, please use the Intel (R) Ethernet Port Configuration Tool to address the issue.\n");
602 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
603 netdev_info(vsi->netdev, "Rx/Tx is disabled on this device because an unsupported module type was detected. Refer to the Intel(R) Ethernet Adapters and Devices User Guide for a list of supported modules.\n");
611 * ice_print_link_msg - print link up or down message
612 * @vsi: the VSI whose link status is being queried
613 * @isup: boolean for if the link is now up or down
615 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
617 struct ice_aqc_get_phy_caps_data *caps;
618 const char *an_advertised;
619 enum ice_status status;
629 if (vsi->current_isup == isup)
632 vsi->current_isup = isup;
635 netdev_info(vsi->netdev, "NIC Link is Down\n");
639 switch (vsi->port_info->phy.link_info.link_speed) {
640 case ICE_AQ_LINK_SPEED_100GB:
643 case ICE_AQ_LINK_SPEED_50GB:
646 case ICE_AQ_LINK_SPEED_40GB:
649 case ICE_AQ_LINK_SPEED_25GB:
652 case ICE_AQ_LINK_SPEED_20GB:
655 case ICE_AQ_LINK_SPEED_10GB:
658 case ICE_AQ_LINK_SPEED_5GB:
661 case ICE_AQ_LINK_SPEED_2500MB:
664 case ICE_AQ_LINK_SPEED_1000MB:
667 case ICE_AQ_LINK_SPEED_100MB:
675 switch (vsi->port_info->fc.current_mode) {
679 case ICE_FC_TX_PAUSE:
682 case ICE_FC_RX_PAUSE:
693 /* Get FEC mode based on negotiated link info */
694 switch (vsi->port_info->phy.link_info.fec_info) {
695 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
696 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
699 case ICE_AQ_LINK_25G_KR_FEC_EN:
700 fec = "FC-FEC/BASE-R";
707 /* check if autoneg completed, might be false due to not supported */
708 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
713 /* Get FEC mode requested based on PHY caps last SW configuration */
714 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
717 an_advertised = "Unknown";
721 status = ice_aq_get_phy_caps(vsi->port_info, false,
722 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
724 netdev_info(vsi->netdev, "Get phy capability failed.\n");
726 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
728 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
729 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
731 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
732 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
733 fec_req = "FC-FEC/BASE-R";
740 netdev_info(vsi->netdev, "NIC Link is up %sbps Full Duplex, Requested FEC: %s, Negotiated FEC: %s, Autoneg Advertised: %s, Autoneg Negotiated: %s, Flow Control: %s\n",
741 speed, fec_req, fec, an_advertised, an, fc);
742 ice_print_topo_conflict(vsi);
746 * ice_vsi_link_event - update the VSI's netdev
747 * @vsi: the VSI on which the link event occurred
748 * @link_up: whether or not the VSI needs to be set up or down
750 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
755 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
758 if (vsi->type == ICE_VSI_PF) {
759 if (link_up == netif_carrier_ok(vsi->netdev))
763 netif_carrier_on(vsi->netdev);
764 netif_tx_wake_all_queues(vsi->netdev);
766 netif_carrier_off(vsi->netdev);
767 netif_tx_stop_all_queues(vsi->netdev);
773 * ice_set_dflt_mib - send a default config MIB to the FW
774 * @pf: private PF struct
776 * This function sends a default configuration MIB to the FW.
778 * If this function errors out at any point, the driver is still able to
779 * function. The main impact is that LFC may not operate as expected.
780 * Therefore an error state in this function should be treated with a DBG
781 * message and continue on with driver rebuild/reenable.
783 static void ice_set_dflt_mib(struct ice_pf *pf)
785 struct device *dev = ice_pf_to_dev(pf);
786 u8 mib_type, *buf, *lldpmib = NULL;
787 u16 len, typelen, offset = 0;
788 struct ice_lldp_org_tlv *tlv;
789 struct ice_hw *hw = &pf->hw;
792 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
793 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
795 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
800 /* Add ETS CFG TLV */
801 tlv = (struct ice_lldp_org_tlv *)lldpmib;
802 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
803 ICE_IEEE_ETS_TLV_LEN);
804 tlv->typelen = htons(typelen);
805 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
806 ICE_IEEE_SUBTYPE_ETS_CFG);
807 tlv->ouisubtype = htonl(ouisubtype);
812 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
813 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
814 * Octets 13 - 20 are TSA values - leave as zeros
817 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
819 tlv = (struct ice_lldp_org_tlv *)
820 ((char *)tlv + sizeof(tlv->typelen) + len);
822 /* Add ETS REC TLV */
824 tlv->typelen = htons(typelen);
826 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
827 ICE_IEEE_SUBTYPE_ETS_REC);
828 tlv->ouisubtype = htonl(ouisubtype);
830 /* First octet of buf is reserved
831 * Octets 1 - 4 map UP to TC - all UPs map to zero
832 * Octets 5 - 12 are BW values - set TC 0 to 100%.
833 * Octets 13 - 20 are TSA value - leave as zeros
837 tlv = (struct ice_lldp_org_tlv *)
838 ((char *)tlv + sizeof(tlv->typelen) + len);
840 /* Add PFC CFG TLV */
841 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
842 ICE_IEEE_PFC_TLV_LEN);
843 tlv->typelen = htons(typelen);
845 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
846 ICE_IEEE_SUBTYPE_PFC_CFG);
847 tlv->ouisubtype = htonl(ouisubtype);
849 /* Octet 1 left as all zeros - PFC disabled */
851 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
854 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
855 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
861 * ice_link_event - process the link event
862 * @pf: PF that the link event is associated with
863 * @pi: port_info for the port that the link event is associated with
864 * @link_up: true if the physical link is up and false if it is down
865 * @link_speed: current link speed received from the link event
867 * Returns 0 on success and negative on failure
870 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
873 struct device *dev = ice_pf_to_dev(pf);
874 struct ice_phy_info *phy_info;
875 enum ice_status status;
881 phy_info->link_info_old = phy_info->link_info;
883 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
884 old_link_speed = phy_info->link_info_old.link_speed;
886 /* update the link info structures and re-enable link events,
887 * don't bail on failure due to other book keeping needed
889 status = ice_update_link_info(pi);
891 dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
892 pi->lport, ice_stat_str(status),
893 ice_aq_str(pi->hw->adminq.sq_last_status));
895 /* Check if the link state is up after updating link info, and treat
896 * this event as an UP event since the link is actually UP now.
898 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
901 vsi = ice_get_main_vsi(pf);
902 if (!vsi || !vsi->port_info)
905 /* turn off PHY if media was removed */
906 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
907 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
908 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
909 ice_set_link(vsi, false);
912 /* if the old link up/down and speed is the same as the new */
913 if (link_up == old_link && link_speed == old_link_speed)
916 if (ice_is_dcb_active(pf)) {
917 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
921 ice_set_dflt_mib(pf);
923 ice_vsi_link_event(vsi, link_up);
924 ice_print_link_msg(vsi, link_up);
926 ice_vc_notify_link_state(pf);
932 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
933 * @pf: board private structure
935 static void ice_watchdog_subtask(struct ice_pf *pf)
939 /* if interface is down do nothing */
940 if (test_bit(ICE_DOWN, pf->state) ||
941 test_bit(ICE_CFG_BUSY, pf->state))
944 /* make sure we don't do these things too often */
945 if (time_before(jiffies,
946 pf->serv_tmr_prev + pf->serv_tmr_period))
949 pf->serv_tmr_prev = jiffies;
951 /* Update the stats for active netdevs so the network stack
952 * can look at updated numbers whenever it cares to
954 ice_update_pf_stats(pf);
955 ice_for_each_vsi(pf, i)
956 if (pf->vsi[i] && pf->vsi[i]->netdev)
957 ice_update_vsi_stats(pf->vsi[i]);
961 * ice_init_link_events - enable/initialize link events
962 * @pi: pointer to the port_info instance
964 * Returns -EIO on failure, 0 on success
966 static int ice_init_link_events(struct ice_port_info *pi)
970 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
971 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
973 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
974 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
979 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
980 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
989 * ice_handle_link_event - handle link event via ARQ
990 * @pf: PF that the link event is associated with
991 * @event: event structure containing link status info
994 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
996 struct ice_aqc_get_link_status_data *link_data;
997 struct ice_port_info *port_info;
1000 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1001 port_info = pf->hw.port_info;
1005 status = ice_link_event(pf, port_info,
1006 !!(link_data->link_info & ICE_AQ_LINK_UP),
1007 le16_to_cpu(link_data->link_speed));
1009 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1015 enum ice_aq_task_state {
1016 ICE_AQ_TASK_WAITING = 0,
1017 ICE_AQ_TASK_COMPLETE,
1018 ICE_AQ_TASK_CANCELED,
1021 struct ice_aq_task {
1022 struct hlist_node entry;
1025 struct ice_rq_event_info *event;
1026 enum ice_aq_task_state state;
1030 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1031 * @pf: pointer to the PF private structure
1032 * @opcode: the opcode to wait for
1033 * @timeout: how long to wait, in jiffies
1034 * @event: storage for the event info
1036 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1037 * current thread will be put to sleep until the specified event occurs or
1038 * until the given timeout is reached.
1040 * To obtain only the descriptor contents, pass an event without an allocated
1041 * msg_buf. If the complete data buffer is desired, allocate the
1042 * event->msg_buf with enough space ahead of time.
1044 * Returns: zero on success, or a negative error code on failure.
1046 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1047 struct ice_rq_event_info *event)
1049 struct device *dev = ice_pf_to_dev(pf);
1050 struct ice_aq_task *task;
1051 unsigned long start;
1055 task = kzalloc(sizeof(*task), GFP_KERNEL);
1059 INIT_HLIST_NODE(&task->entry);
1060 task->opcode = opcode;
1061 task->event = event;
1062 task->state = ICE_AQ_TASK_WAITING;
1064 spin_lock_bh(&pf->aq_wait_lock);
1065 hlist_add_head(&task->entry, &pf->aq_wait_list);
1066 spin_unlock_bh(&pf->aq_wait_lock);
1070 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1072 switch (task->state) {
1073 case ICE_AQ_TASK_WAITING:
1074 err = ret < 0 ? ret : -ETIMEDOUT;
1076 case ICE_AQ_TASK_CANCELED:
1077 err = ret < 0 ? ret : -ECANCELED;
1079 case ICE_AQ_TASK_COMPLETE:
1080 err = ret < 0 ? ret : 0;
1083 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1088 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1089 jiffies_to_msecs(jiffies - start),
1090 jiffies_to_msecs(timeout),
1093 spin_lock_bh(&pf->aq_wait_lock);
1094 hlist_del(&task->entry);
1095 spin_unlock_bh(&pf->aq_wait_lock);
1102 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1103 * @pf: pointer to the PF private structure
1104 * @opcode: the opcode of the event
1105 * @event: the event to check
1107 * Loops over the current list of pending threads waiting for an AdminQ event.
1108 * For each matching task, copy the contents of the event into the task
1109 * structure and wake up the thread.
1111 * If multiple threads wait for the same opcode, they will all be woken up.
1113 * Note that event->msg_buf will only be duplicated if the event has a buffer
1114 * with enough space already allocated. Otherwise, only the descriptor and
1115 * message length will be copied.
1117 * Returns: true if an event was found, false otherwise
1119 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1120 struct ice_rq_event_info *event)
1122 struct ice_aq_task *task;
1125 spin_lock_bh(&pf->aq_wait_lock);
1126 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1127 if (task->state || task->opcode != opcode)
1130 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1131 task->event->msg_len = event->msg_len;
1133 /* Only copy the data buffer if a destination was set */
1134 if (task->event->msg_buf &&
1135 task->event->buf_len > event->buf_len) {
1136 memcpy(task->event->msg_buf, event->msg_buf,
1138 task->event->buf_len = event->buf_len;
1141 task->state = ICE_AQ_TASK_COMPLETE;
1144 spin_unlock_bh(&pf->aq_wait_lock);
1147 wake_up(&pf->aq_wait_queue);
1151 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1152 * @pf: the PF private structure
1154 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1155 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1157 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1159 struct ice_aq_task *task;
1161 spin_lock_bh(&pf->aq_wait_lock);
1162 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1163 task->state = ICE_AQ_TASK_CANCELED;
1164 spin_unlock_bh(&pf->aq_wait_lock);
1166 wake_up(&pf->aq_wait_queue);
1170 * __ice_clean_ctrlq - helper function to clean controlq rings
1171 * @pf: ptr to struct ice_pf
1172 * @q_type: specific Control queue type
1174 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1176 struct device *dev = ice_pf_to_dev(pf);
1177 struct ice_rq_event_info event;
1178 struct ice_hw *hw = &pf->hw;
1179 struct ice_ctl_q_info *cq;
1184 /* Do not clean control queue if/when PF reset fails */
1185 if (test_bit(ICE_RESET_FAILED, pf->state))
1189 case ICE_CTL_Q_ADMIN:
1193 case ICE_CTL_Q_MAILBOX:
1196 /* we are going to try to detect a malicious VF, so set the
1197 * state to begin detection
1199 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1202 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1206 /* check for error indications - PF_xx_AxQLEN register layout for
1207 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1209 val = rd32(hw, cq->rq.len);
1210 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1211 PF_FW_ARQLEN_ARQCRIT_M)) {
1213 if (val & PF_FW_ARQLEN_ARQVFE_M)
1214 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1216 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1217 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1220 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1221 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1223 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1224 PF_FW_ARQLEN_ARQCRIT_M);
1226 wr32(hw, cq->rq.len, val);
1229 val = rd32(hw, cq->sq.len);
1230 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1231 PF_FW_ATQLEN_ATQCRIT_M)) {
1233 if (val & PF_FW_ATQLEN_ATQVFE_M)
1234 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1236 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1237 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1240 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1241 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1243 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1244 PF_FW_ATQLEN_ATQCRIT_M);
1246 wr32(hw, cq->sq.len, val);
1249 event.buf_len = cq->rq_buf_size;
1250 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1255 enum ice_status ret;
1258 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1259 if (ret == ICE_ERR_AQ_NO_WORK)
1262 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1267 opcode = le16_to_cpu(event.desc.opcode);
1269 /* Notify any thread that might be waiting for this event */
1270 ice_aq_check_events(pf, opcode, &event);
1273 case ice_aqc_opc_get_link_status:
1274 if (ice_handle_link_event(pf, &event))
1275 dev_err(dev, "Could not handle link event\n");
1277 case ice_aqc_opc_event_lan_overflow:
1278 ice_vf_lan_overflow_event(pf, &event);
1280 case ice_mbx_opc_send_msg_to_pf:
1281 if (!ice_is_malicious_vf(pf, &event, i, pending))
1282 ice_vc_process_vf_msg(pf, &event);
1284 case ice_aqc_opc_fw_logging:
1285 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1287 case ice_aqc_opc_lldp_set_mib_change:
1288 ice_dcb_process_lldp_set_mib_change(pf, &event);
1291 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1295 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1297 kfree(event.msg_buf);
1299 return pending && (i == ICE_DFLT_IRQ_WORK);
1303 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1304 * @hw: pointer to hardware info
1305 * @cq: control queue information
1307 * returns true if there are pending messages in a queue, false if there aren't
1309 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1313 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1314 return cq->rq.next_to_clean != ntu;
1318 * ice_clean_adminq_subtask - clean the AdminQ rings
1319 * @pf: board private structure
1321 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1323 struct ice_hw *hw = &pf->hw;
1325 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1328 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1331 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1333 /* There might be a situation where new messages arrive to a control
1334 * queue between processing the last message and clearing the
1335 * EVENT_PENDING bit. So before exiting, check queue head again (using
1336 * ice_ctrlq_pending) and process new messages if any.
1338 if (ice_ctrlq_pending(hw, &hw->adminq))
1339 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1345 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1346 * @pf: board private structure
1348 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1350 struct ice_hw *hw = &pf->hw;
1352 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1355 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1358 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1360 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1361 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1367 * ice_service_task_schedule - schedule the service task to wake up
1368 * @pf: board private structure
1370 * If not already scheduled, this puts the task into the work queue.
1372 void ice_service_task_schedule(struct ice_pf *pf)
1374 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1375 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1376 !test_bit(ICE_NEEDS_RESTART, pf->state))
1377 queue_work(ice_wq, &pf->serv_task);
1381 * ice_service_task_complete - finish up the service task
1382 * @pf: board private structure
1384 static void ice_service_task_complete(struct ice_pf *pf)
1386 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1388 /* force memory (pf->state) to sync before next service task */
1389 smp_mb__before_atomic();
1390 clear_bit(ICE_SERVICE_SCHED, pf->state);
1394 * ice_service_task_stop - stop service task and cancel works
1395 * @pf: board private structure
1397 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1400 static int ice_service_task_stop(struct ice_pf *pf)
1404 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1406 if (pf->serv_tmr.function)
1407 del_timer_sync(&pf->serv_tmr);
1408 if (pf->serv_task.func)
1409 cancel_work_sync(&pf->serv_task);
1411 clear_bit(ICE_SERVICE_SCHED, pf->state);
1416 * ice_service_task_restart - restart service task and schedule works
1417 * @pf: board private structure
1419 * This function is needed for suspend and resume works (e.g WoL scenario)
1421 static void ice_service_task_restart(struct ice_pf *pf)
1423 clear_bit(ICE_SERVICE_DIS, pf->state);
1424 ice_service_task_schedule(pf);
1428 * ice_service_timer - timer callback to schedule service task
1429 * @t: pointer to timer_list
1431 static void ice_service_timer(struct timer_list *t)
1433 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1435 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1436 ice_service_task_schedule(pf);
1440 * ice_handle_mdd_event - handle malicious driver detect event
1441 * @pf: pointer to the PF structure
1443 * Called from service task. OICR interrupt handler indicates MDD event.
1444 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1445 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1446 * disable the queue, the PF can be configured to reset the VF using ethtool
1447 * private flag mdd-auto-reset-vf.
1449 static void ice_handle_mdd_event(struct ice_pf *pf)
1451 struct device *dev = ice_pf_to_dev(pf);
1452 struct ice_hw *hw = &pf->hw;
1456 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1457 /* Since the VF MDD event logging is rate limited, check if
1458 * there are pending MDD events.
1460 ice_print_vfs_mdd_events(pf);
1464 /* find what triggered an MDD event */
1465 reg = rd32(hw, GL_MDET_TX_PQM);
1466 if (reg & GL_MDET_TX_PQM_VALID_M) {
1467 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1468 GL_MDET_TX_PQM_PF_NUM_S;
1469 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1470 GL_MDET_TX_PQM_VF_NUM_S;
1471 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1472 GL_MDET_TX_PQM_MAL_TYPE_S;
1473 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1474 GL_MDET_TX_PQM_QNUM_S);
1476 if (netif_msg_tx_err(pf))
1477 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1478 event, queue, pf_num, vf_num);
1479 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1482 reg = rd32(hw, GL_MDET_TX_TCLAN);
1483 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1484 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1485 GL_MDET_TX_TCLAN_PF_NUM_S;
1486 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1487 GL_MDET_TX_TCLAN_VF_NUM_S;
1488 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1489 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1490 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1491 GL_MDET_TX_TCLAN_QNUM_S);
1493 if (netif_msg_tx_err(pf))
1494 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1495 event, queue, pf_num, vf_num);
1496 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1499 reg = rd32(hw, GL_MDET_RX);
1500 if (reg & GL_MDET_RX_VALID_M) {
1501 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1502 GL_MDET_RX_PF_NUM_S;
1503 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1504 GL_MDET_RX_VF_NUM_S;
1505 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1506 GL_MDET_RX_MAL_TYPE_S;
1507 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1510 if (netif_msg_rx_err(pf))
1511 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1512 event, queue, pf_num, vf_num);
1513 wr32(hw, GL_MDET_RX, 0xffffffff);
1516 /* check to see if this PF caused an MDD event */
1517 reg = rd32(hw, PF_MDET_TX_PQM);
1518 if (reg & PF_MDET_TX_PQM_VALID_M) {
1519 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1520 if (netif_msg_tx_err(pf))
1521 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1524 reg = rd32(hw, PF_MDET_TX_TCLAN);
1525 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1526 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1527 if (netif_msg_tx_err(pf))
1528 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1531 reg = rd32(hw, PF_MDET_RX);
1532 if (reg & PF_MDET_RX_VALID_M) {
1533 wr32(hw, PF_MDET_RX, 0xFFFF);
1534 if (netif_msg_rx_err(pf))
1535 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1538 /* Check to see if one of the VFs caused an MDD event, and then
1539 * increment counters and set print pending
1541 ice_for_each_vf(pf, i) {
1542 struct ice_vf *vf = &pf->vf[i];
1544 reg = rd32(hw, VP_MDET_TX_PQM(i));
1545 if (reg & VP_MDET_TX_PQM_VALID_M) {
1546 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1547 vf->mdd_tx_events.count++;
1548 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1549 if (netif_msg_tx_err(pf))
1550 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1554 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1555 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1556 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1557 vf->mdd_tx_events.count++;
1558 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1559 if (netif_msg_tx_err(pf))
1560 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1564 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1565 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1566 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1567 vf->mdd_tx_events.count++;
1568 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1569 if (netif_msg_tx_err(pf))
1570 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1574 reg = rd32(hw, VP_MDET_RX(i));
1575 if (reg & VP_MDET_RX_VALID_M) {
1576 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1577 vf->mdd_rx_events.count++;
1578 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1579 if (netif_msg_rx_err(pf))
1580 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1583 /* Since the queue is disabled on VF Rx MDD events, the
1584 * PF can be configured to reset the VF through ethtool
1585 * private flag mdd-auto-reset-vf.
1587 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1588 /* VF MDD event counters will be cleared by
1589 * reset, so print the event prior to reset.
1591 ice_print_vf_rx_mdd_event(vf);
1592 ice_reset_vf(&pf->vf[i], false);
1597 ice_print_vfs_mdd_events(pf);
1601 * ice_force_phys_link_state - Force the physical link state
1602 * @vsi: VSI to force the physical link state to up/down
1603 * @link_up: true/false indicates to set the physical link to up/down
1605 * Force the physical link state by getting the current PHY capabilities from
1606 * hardware and setting the PHY config based on the determined capabilities. If
1607 * link changes a link event will be triggered because both the Enable Automatic
1608 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1610 * Returns 0 on success, negative on failure
1612 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1614 struct ice_aqc_get_phy_caps_data *pcaps;
1615 struct ice_aqc_set_phy_cfg_data *cfg;
1616 struct ice_port_info *pi;
1620 if (!vsi || !vsi->port_info || !vsi->back)
1622 if (vsi->type != ICE_VSI_PF)
1625 dev = ice_pf_to_dev(vsi->back);
1627 pi = vsi->port_info;
1629 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1633 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1636 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1637 vsi->vsi_num, retcode);
1642 /* No change in link */
1643 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1644 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1647 /* Use the current user PHY configuration. The current user PHY
1648 * configuration is initialized during probe from PHY capabilities
1649 * software mode, and updated on set PHY configuration.
1651 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1657 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1659 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1661 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1663 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1665 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1666 vsi->vsi_num, retcode);
1677 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1678 * @pi: port info structure
1680 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1682 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1684 struct ice_aqc_get_phy_caps_data *pcaps;
1685 struct ice_pf *pf = pi->hw->back;
1686 enum ice_status status;
1689 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1693 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1697 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1702 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1703 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1711 * ice_init_link_dflt_override - Initialize link default override
1712 * @pi: port info structure
1714 * Initialize link default override and PHY total port shutdown during probe
1716 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1718 struct ice_link_default_override_tlv *ldo;
1719 struct ice_pf *pf = pi->hw->back;
1721 ldo = &pf->link_dflt_override;
1722 if (ice_get_link_default_override(ldo, pi))
1725 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1728 /* Enable Total Port Shutdown (override/replace link-down-on-close
1729 * ethtool private flag) for ports with Port Disable bit set.
1731 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1732 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1736 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1737 * @pi: port info structure
1739 * If default override is enabled, initialize the user PHY cfg speed and FEC
1740 * settings using the default override mask from the NVM.
1742 * The PHY should only be configured with the default override settings the
1743 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1744 * is used to indicate that the user PHY cfg default override is initialized
1745 * and the PHY has not been configured with the default override settings. The
1746 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1749 * This function should be called only if the FW doesn't support default
1750 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1752 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1754 struct ice_link_default_override_tlv *ldo;
1755 struct ice_aqc_set_phy_cfg_data *cfg;
1756 struct ice_phy_info *phy = &pi->phy;
1757 struct ice_pf *pf = pi->hw->back;
1759 ldo = &pf->link_dflt_override;
1761 /* If link default override is enabled, use to mask NVM PHY capabilities
1762 * for speed and FEC default configuration.
1764 cfg = &phy->curr_user_phy_cfg;
1766 if (ldo->phy_type_low || ldo->phy_type_high) {
1767 cfg->phy_type_low = pf->nvm_phy_type_lo &
1768 cpu_to_le64(ldo->phy_type_low);
1769 cfg->phy_type_high = pf->nvm_phy_type_hi &
1770 cpu_to_le64(ldo->phy_type_high);
1772 cfg->link_fec_opt = ldo->fec_options;
1773 phy->curr_user_fec_req = ICE_FEC_AUTO;
1775 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1779 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1780 * @pi: port info structure
1782 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1783 * mode to default. The PHY defaults are from get PHY capabilities topology
1784 * with media so call when media is first available. An error is returned if
1785 * called when media is not available. The PHY initialization completed state is
1788 * These configurations are used when setting PHY
1789 * configuration. The user PHY configuration is updated on set PHY
1790 * configuration. Returns 0 on success, negative on failure
1792 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1794 struct ice_aqc_get_phy_caps_data *pcaps;
1795 struct ice_phy_info *phy = &pi->phy;
1796 struct ice_pf *pf = pi->hw->back;
1797 enum ice_status status;
1800 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1803 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1807 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1808 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1811 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1814 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1819 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1821 /* check if lenient mode is supported and enabled */
1822 if (ice_fw_supports_link_override(pi->hw) &&
1823 !(pcaps->module_compliance_enforcement &
1824 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1825 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1827 /* if the FW supports default PHY configuration mode, then the driver
1828 * does not have to apply link override settings. If not,
1829 * initialize user PHY configuration with link override values
1831 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1832 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1833 ice_init_phy_cfg_dflt_override(pi);
1838 /* if link default override is not enabled, set user flow control and
1839 * FEC settings based on what get_phy_caps returned
1841 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1842 pcaps->link_fec_options);
1843 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1846 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1847 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
1854 * ice_configure_phy - configure PHY
1857 * Set the PHY configuration. If the current PHY configuration is the same as
1858 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1859 * configure the based get PHY capabilities for topology with media.
1861 static int ice_configure_phy(struct ice_vsi *vsi)
1863 struct device *dev = ice_pf_to_dev(vsi->back);
1864 struct ice_port_info *pi = vsi->port_info;
1865 struct ice_aqc_get_phy_caps_data *pcaps;
1866 struct ice_aqc_set_phy_cfg_data *cfg;
1867 struct ice_phy_info *phy = &pi->phy;
1868 struct ice_pf *pf = vsi->back;
1869 enum ice_status status;
1872 /* Ensure we have media as we cannot configure a medialess port */
1873 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1876 ice_print_topo_conflict(vsi);
1878 if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1881 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1882 return ice_force_phys_link_state(vsi, true);
1884 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1888 /* Get current PHY config */
1889 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1892 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1893 vsi->vsi_num, ice_stat_str(status));
1898 /* If PHY enable link is configured and configuration has not changed,
1899 * there's nothing to do
1901 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1902 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1905 /* Use PHY topology as baseline for configuration */
1906 memset(pcaps, 0, sizeof(*pcaps));
1907 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1908 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1911 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1914 dev_err(dev, "Failed to get PHY caps, 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 = phy->curr_user_phy_cfg.phy_type_low;
1934 cfg->phy_type_high = 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, 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, 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(&pf->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_VSI_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 ice_flush_fdir_ctx(pf);
2069 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2070 ice_service_task_complete(pf);
2072 /* If the tasks have taken longer than one service timer period
2073 * or there is more work to be done, reset the service timer to
2074 * schedule the service task now.
2076 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2077 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2078 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2079 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2080 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2081 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2082 mod_timer(&pf->serv_tmr, jiffies);
2086 * ice_set_ctrlq_len - helper function to set controlq length
2087 * @hw: pointer to the HW instance
2089 static void ice_set_ctrlq_len(struct ice_hw *hw)
2091 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2092 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2093 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2094 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2095 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2096 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2097 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2098 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2102 * ice_schedule_reset - schedule a reset
2103 * @pf: board private structure
2104 * @reset: reset being requested
2106 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2108 struct device *dev = ice_pf_to_dev(pf);
2110 /* bail out if earlier reset has failed */
2111 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2112 dev_dbg(dev, "earlier reset has failed\n");
2115 /* bail if reset/recovery already in progress */
2116 if (ice_is_reset_in_progress(pf->state)) {
2117 dev_dbg(dev, "Reset already in progress\n");
2123 set_bit(ICE_PFR_REQ, pf->state);
2125 case ICE_RESET_CORER:
2126 set_bit(ICE_CORER_REQ, pf->state);
2128 case ICE_RESET_GLOBR:
2129 set_bit(ICE_GLOBR_REQ, pf->state);
2135 ice_service_task_schedule(pf);
2140 * ice_irq_affinity_notify - Callback for affinity changes
2141 * @notify: context as to what irq was changed
2142 * @mask: the new affinity mask
2144 * This is a callback function used by the irq_set_affinity_notifier function
2145 * so that we may register to receive changes to the irq affinity masks.
2148 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2149 const cpumask_t *mask)
2151 struct ice_q_vector *q_vector =
2152 container_of(notify, struct ice_q_vector, affinity_notify);
2154 cpumask_copy(&q_vector->affinity_mask, mask);
2158 * ice_irq_affinity_release - Callback for affinity notifier release
2159 * @ref: internal core kernel usage
2161 * This is a callback function used by the irq_set_affinity_notifier function
2162 * to inform the current notification subscriber that they will no longer
2163 * receive notifications.
2165 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2168 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2169 * @vsi: the VSI being configured
2171 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2173 struct ice_hw *hw = &vsi->back->hw;
2176 ice_for_each_q_vector(vsi, i)
2177 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2184 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2185 * @vsi: the VSI being configured
2186 * @basename: name for the vector
2188 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2190 int q_vectors = vsi->num_q_vectors;
2191 struct ice_pf *pf = vsi->back;
2192 int base = vsi->base_vector;
2199 dev = ice_pf_to_dev(pf);
2200 for (vector = 0; vector < q_vectors; vector++) {
2201 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2203 irq_num = pf->msix_entries[base + vector].vector;
2205 if (q_vector->tx.ring && q_vector->rx.ring) {
2206 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2207 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2209 } else if (q_vector->rx.ring) {
2210 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2211 "%s-%s-%d", basename, "rx", rx_int_idx++);
2212 } else if (q_vector->tx.ring) {
2213 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2214 "%s-%s-%d", basename, "tx", tx_int_idx++);
2216 /* skip this unused q_vector */
2219 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2220 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2221 IRQF_SHARED, q_vector->name,
2224 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2225 0, q_vector->name, q_vector);
2227 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2232 /* register for affinity change notifications */
2233 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2234 struct irq_affinity_notify *affinity_notify;
2236 affinity_notify = &q_vector->affinity_notify;
2237 affinity_notify->notify = ice_irq_affinity_notify;
2238 affinity_notify->release = ice_irq_affinity_release;
2239 irq_set_affinity_notifier(irq_num, affinity_notify);
2242 /* assign the mask for this irq */
2243 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2246 vsi->irqs_ready = true;
2252 irq_num = pf->msix_entries[base + vector].vector;
2253 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2254 irq_set_affinity_notifier(irq_num, NULL);
2255 irq_set_affinity_hint(irq_num, NULL);
2256 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2262 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2263 * @vsi: VSI to setup Tx rings used by XDP
2265 * Return 0 on success and negative value on error
2267 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2269 struct device *dev = ice_pf_to_dev(vsi->back);
2272 for (i = 0; i < vsi->num_xdp_txq; i++) {
2273 u16 xdp_q_idx = vsi->alloc_txq + i;
2274 struct ice_ring *xdp_ring;
2276 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2279 goto free_xdp_rings;
2281 xdp_ring->q_index = xdp_q_idx;
2282 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2283 xdp_ring->ring_active = false;
2284 xdp_ring->vsi = vsi;
2285 xdp_ring->netdev = NULL;
2286 xdp_ring->dev = dev;
2287 xdp_ring->count = vsi->num_tx_desc;
2288 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2289 if (ice_setup_tx_ring(xdp_ring))
2290 goto free_xdp_rings;
2291 ice_set_ring_xdp(xdp_ring);
2292 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2299 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2300 ice_free_tx_ring(vsi->xdp_rings[i]);
2305 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2306 * @vsi: VSI to set the bpf prog on
2307 * @prog: the bpf prog pointer
2309 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2311 struct bpf_prog *old_prog;
2314 old_prog = xchg(&vsi->xdp_prog, prog);
2316 bpf_prog_put(old_prog);
2318 ice_for_each_rxq(vsi, i)
2319 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2323 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2324 * @vsi: VSI to bring up Tx rings used by XDP
2325 * @prog: bpf program that will be assigned to VSI
2327 * Return 0 on success and negative value on error
2329 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2331 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2332 int xdp_rings_rem = vsi->num_xdp_txq;
2333 struct ice_pf *pf = vsi->back;
2334 struct ice_qs_cfg xdp_qs_cfg = {
2335 .qs_mutex = &pf->avail_q_mutex,
2336 .pf_map = pf->avail_txqs,
2337 .pf_map_size = pf->max_pf_txqs,
2338 .q_count = vsi->num_xdp_txq,
2339 .scatter_count = ICE_MAX_SCATTER_TXQS,
2340 .vsi_map = vsi->txq_map,
2341 .vsi_map_offset = vsi->alloc_txq,
2342 .mapping_mode = ICE_VSI_MAP_CONTIG
2344 enum ice_status status;
2348 dev = ice_pf_to_dev(pf);
2349 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2350 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2351 if (!vsi->xdp_rings)
2354 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2355 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2358 if (ice_xdp_alloc_setup_rings(vsi))
2359 goto clear_xdp_rings;
2361 /* follow the logic from ice_vsi_map_rings_to_vectors */
2362 ice_for_each_q_vector(vsi, v_idx) {
2363 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2364 int xdp_rings_per_v, q_id, q_base;
2366 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2367 vsi->num_q_vectors - v_idx);
2368 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2370 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2371 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2373 xdp_ring->q_vector = q_vector;
2374 xdp_ring->next = q_vector->tx.ring;
2375 q_vector->tx.ring = xdp_ring;
2377 xdp_rings_rem -= xdp_rings_per_v;
2380 /* omit the scheduler update if in reset path; XDP queues will be
2381 * taken into account at the end of ice_vsi_rebuild, where
2382 * ice_cfg_vsi_lan is being called
2384 if (ice_is_reset_in_progress(pf->state))
2387 /* tell the Tx scheduler that right now we have
2390 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2391 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2393 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2396 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2397 ice_stat_str(status));
2398 goto clear_xdp_rings;
2400 ice_vsi_assign_bpf_prog(vsi, prog);
2404 for (i = 0; i < vsi->num_xdp_txq; i++)
2405 if (vsi->xdp_rings[i]) {
2406 kfree_rcu(vsi->xdp_rings[i], rcu);
2407 vsi->xdp_rings[i] = NULL;
2411 mutex_lock(&pf->avail_q_mutex);
2412 for (i = 0; i < vsi->num_xdp_txq; i++) {
2413 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2414 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2416 mutex_unlock(&pf->avail_q_mutex);
2418 devm_kfree(dev, vsi->xdp_rings);
2423 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2424 * @vsi: VSI to remove XDP rings
2426 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2429 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2431 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2432 struct ice_pf *pf = vsi->back;
2435 /* q_vectors are freed in reset path so there's no point in detaching
2436 * rings; in case of rebuild being triggered not from reset bits
2437 * in pf->state won't be set, so additionally check first q_vector
2440 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2443 ice_for_each_q_vector(vsi, v_idx) {
2444 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2445 struct ice_ring *ring;
2447 ice_for_each_ring(ring, q_vector->tx)
2448 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2451 /* restore the value of last node prior to XDP setup */
2452 q_vector->tx.ring = ring;
2456 mutex_lock(&pf->avail_q_mutex);
2457 for (i = 0; i < vsi->num_xdp_txq; i++) {
2458 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2459 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2461 mutex_unlock(&pf->avail_q_mutex);
2463 for (i = 0; i < vsi->num_xdp_txq; i++)
2464 if (vsi->xdp_rings[i]) {
2465 if (vsi->xdp_rings[i]->desc)
2466 ice_free_tx_ring(vsi->xdp_rings[i]);
2467 kfree_rcu(vsi->xdp_rings[i], rcu);
2468 vsi->xdp_rings[i] = NULL;
2471 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2472 vsi->xdp_rings = NULL;
2474 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2477 ice_vsi_assign_bpf_prog(vsi, NULL);
2479 /* notify Tx scheduler that we destroyed XDP queues and bring
2480 * back the old number of child nodes
2482 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2483 max_txqs[i] = vsi->num_txq;
2485 /* change number of XDP Tx queues to 0 */
2486 vsi->num_xdp_txq = 0;
2488 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2493 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2494 * @vsi: VSI to schedule napi on
2496 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2500 ice_for_each_rxq(vsi, i) {
2501 struct ice_ring *rx_ring = vsi->rx_rings[i];
2503 if (rx_ring->xsk_pool)
2504 napi_schedule(&rx_ring->q_vector->napi);
2509 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2510 * @vsi: VSI to setup XDP for
2511 * @prog: XDP program
2512 * @extack: netlink extended ack
2515 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2516 struct netlink_ext_ack *extack)
2518 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2519 bool if_running = netif_running(vsi->netdev);
2520 int ret = 0, xdp_ring_err = 0;
2522 if (frame_size > vsi->rx_buf_len) {
2523 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2527 /* need to stop netdev while setting up the program for Rx rings */
2528 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2529 ret = ice_down(vsi);
2531 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2536 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2537 vsi->num_xdp_txq = vsi->alloc_rxq;
2538 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2540 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2541 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2542 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2544 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2546 ice_vsi_assign_bpf_prog(vsi, prog);
2553 ice_vsi_rx_napi_schedule(vsi);
2555 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2559 * ice_xdp - implements XDP handler
2563 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2565 struct ice_netdev_priv *np = netdev_priv(dev);
2566 struct ice_vsi *vsi = np->vsi;
2568 if (vsi->type != ICE_VSI_PF) {
2569 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2573 switch (xdp->command) {
2574 case XDP_SETUP_PROG:
2575 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2576 case XDP_SETUP_XSK_POOL:
2577 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2585 * ice_ena_misc_vector - enable the non-queue interrupts
2586 * @pf: board private structure
2588 static void ice_ena_misc_vector(struct ice_pf *pf)
2590 struct ice_hw *hw = &pf->hw;
2593 /* Disable anti-spoof detection interrupt to prevent spurious event
2594 * interrupts during a function reset. Anti-spoof functionally is
2597 val = rd32(hw, GL_MDCK_TX_TDPU);
2598 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2599 wr32(hw, GL_MDCK_TX_TDPU, val);
2601 /* clear things first */
2602 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2603 rd32(hw, PFINT_OICR); /* read to clear */
2605 val = (PFINT_OICR_ECC_ERR_M |
2606 PFINT_OICR_MAL_DETECT_M |
2608 PFINT_OICR_PCI_EXCEPTION_M |
2610 PFINT_OICR_HMC_ERR_M |
2611 PFINT_OICR_PE_CRITERR_M);
2613 wr32(hw, PFINT_OICR_ENA, val);
2615 /* SW_ITR_IDX = 0, but don't change INTENA */
2616 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2617 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2621 * ice_misc_intr - misc interrupt handler
2622 * @irq: interrupt number
2623 * @data: pointer to a q_vector
2625 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2627 struct ice_pf *pf = (struct ice_pf *)data;
2628 struct ice_hw *hw = &pf->hw;
2629 irqreturn_t ret = IRQ_NONE;
2633 dev = ice_pf_to_dev(pf);
2634 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2635 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2637 oicr = rd32(hw, PFINT_OICR);
2638 ena_mask = rd32(hw, PFINT_OICR_ENA);
2640 if (oicr & PFINT_OICR_SWINT_M) {
2641 ena_mask &= ~PFINT_OICR_SWINT_M;
2645 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2646 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2647 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2649 if (oicr & PFINT_OICR_VFLR_M) {
2650 /* disable any further VFLR event notifications */
2651 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2652 u32 reg = rd32(hw, PFINT_OICR_ENA);
2654 reg &= ~PFINT_OICR_VFLR_M;
2655 wr32(hw, PFINT_OICR_ENA, reg);
2657 ena_mask &= ~PFINT_OICR_VFLR_M;
2658 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2662 if (oicr & PFINT_OICR_GRST_M) {
2665 /* we have a reset warning */
2666 ena_mask &= ~PFINT_OICR_GRST_M;
2667 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2668 GLGEN_RSTAT_RESET_TYPE_S;
2670 if (reset == ICE_RESET_CORER)
2672 else if (reset == ICE_RESET_GLOBR)
2674 else if (reset == ICE_RESET_EMPR)
2677 dev_dbg(dev, "Invalid reset type %d\n", reset);
2679 /* If a reset cycle isn't already in progress, we set a bit in
2680 * pf->state so that the service task can start a reset/rebuild.
2681 * We also make note of which reset happened so that peer
2682 * devices/drivers can be informed.
2684 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
2685 if (reset == ICE_RESET_CORER)
2686 set_bit(ICE_CORER_RECV, pf->state);
2687 else if (reset == ICE_RESET_GLOBR)
2688 set_bit(ICE_GLOBR_RECV, pf->state);
2690 set_bit(ICE_EMPR_RECV, pf->state);
2692 /* There are couple of different bits at play here.
2693 * hw->reset_ongoing indicates whether the hardware is
2694 * in reset. This is set to true when a reset interrupt
2695 * is received and set back to false after the driver
2696 * has determined that the hardware is out of reset.
2698 * ICE_RESET_OICR_RECV in pf->state indicates
2699 * that a post reset rebuild is required before the
2700 * driver is operational again. This is set above.
2702 * As this is the start of the reset/rebuild cycle, set
2703 * both to indicate that.
2705 hw->reset_ongoing = true;
2709 if (oicr & PFINT_OICR_HMC_ERR_M) {
2710 ena_mask &= ~PFINT_OICR_HMC_ERR_M;
2711 dev_dbg(dev, "HMC Error interrupt - info 0x%x, data 0x%x\n",
2712 rd32(hw, PFHMC_ERRORINFO),
2713 rd32(hw, PFHMC_ERRORDATA));
2716 /* Report any remaining unexpected interrupts */
2719 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2720 /* If a critical error is pending there is no choice but to
2723 if (oicr & (PFINT_OICR_PE_CRITERR_M |
2724 PFINT_OICR_PCI_EXCEPTION_M |
2725 PFINT_OICR_ECC_ERR_M)) {
2726 set_bit(ICE_PFR_REQ, pf->state);
2727 ice_service_task_schedule(pf);
2732 ice_service_task_schedule(pf);
2733 ice_irq_dynamic_ena(hw, NULL, NULL);
2739 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2740 * @hw: pointer to HW structure
2742 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2744 /* disable Admin queue Interrupt causes */
2745 wr32(hw, PFINT_FW_CTL,
2746 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2748 /* disable Mailbox queue Interrupt causes */
2749 wr32(hw, PFINT_MBX_CTL,
2750 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2752 /* disable Control queue Interrupt causes */
2753 wr32(hw, PFINT_OICR_CTL,
2754 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2760 * ice_free_irq_msix_misc - Unroll misc vector setup
2761 * @pf: board private structure
2763 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2765 struct ice_hw *hw = &pf->hw;
2767 ice_dis_ctrlq_interrupts(hw);
2769 /* disable OICR interrupt */
2770 wr32(hw, PFINT_OICR_ENA, 0);
2773 if (pf->msix_entries) {
2774 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2775 devm_free_irq(ice_pf_to_dev(pf),
2776 pf->msix_entries[pf->oicr_idx].vector, pf);
2779 pf->num_avail_sw_msix += 1;
2780 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2784 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2785 * @hw: pointer to HW structure
2786 * @reg_idx: HW vector index to associate the control queue interrupts with
2788 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2792 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2793 PFINT_OICR_CTL_CAUSE_ENA_M);
2794 wr32(hw, PFINT_OICR_CTL, val);
2796 /* enable Admin queue Interrupt causes */
2797 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2798 PFINT_FW_CTL_CAUSE_ENA_M);
2799 wr32(hw, PFINT_FW_CTL, val);
2801 /* enable Mailbox queue Interrupt causes */
2802 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2803 PFINT_MBX_CTL_CAUSE_ENA_M);
2804 wr32(hw, PFINT_MBX_CTL, val);
2810 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2811 * @pf: board private structure
2813 * This sets up the handler for MSIX 0, which is used to manage the
2814 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2815 * when in MSI or Legacy interrupt mode.
2817 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2819 struct device *dev = ice_pf_to_dev(pf);
2820 struct ice_hw *hw = &pf->hw;
2821 int oicr_idx, err = 0;
2823 if (!pf->int_name[0])
2824 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2825 dev_driver_string(dev), dev_name(dev));
2827 /* Do not request IRQ but do enable OICR interrupt since settings are
2828 * lost during reset. Note that this function is called only during
2829 * rebuild path and not while reset is in progress.
2831 if (ice_is_reset_in_progress(pf->state))
2834 /* reserve one vector in irq_tracker for misc interrupts */
2835 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2839 pf->num_avail_sw_msix -= 1;
2840 pf->oicr_idx = (u16)oicr_idx;
2842 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2843 ice_misc_intr, 0, pf->int_name, pf);
2845 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2847 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2848 pf->num_avail_sw_msix += 1;
2853 ice_ena_misc_vector(pf);
2855 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
2856 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
2857 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
2860 ice_irq_dynamic_ena(hw, NULL, NULL);
2866 * ice_napi_add - register NAPI handler for the VSI
2867 * @vsi: VSI for which NAPI handler is to be registered
2869 * This function is only called in the driver's load path. Registering the NAPI
2870 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
2871 * reset/rebuild, etc.)
2873 static void ice_napi_add(struct ice_vsi *vsi)
2880 ice_for_each_q_vector(vsi, v_idx)
2881 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
2882 ice_napi_poll, NAPI_POLL_WEIGHT);
2886 * ice_set_ops - set netdev and ethtools ops for the given netdev
2887 * @netdev: netdev instance
2889 static void ice_set_ops(struct net_device *netdev)
2891 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2893 if (ice_is_safe_mode(pf)) {
2894 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
2895 ice_set_ethtool_safe_mode_ops(netdev);
2899 netdev->netdev_ops = &ice_netdev_ops;
2900 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
2901 ice_set_ethtool_ops(netdev);
2905 * ice_set_netdev_features - set features for the given netdev
2906 * @netdev: netdev instance
2908 static void ice_set_netdev_features(struct net_device *netdev)
2910 struct ice_pf *pf = ice_netdev_to_pf(netdev);
2911 netdev_features_t csumo_features;
2912 netdev_features_t vlano_features;
2913 netdev_features_t dflt_features;
2914 netdev_features_t tso_features;
2916 if (ice_is_safe_mode(pf)) {
2918 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
2919 netdev->hw_features = netdev->features;
2923 dflt_features = NETIF_F_SG |
2928 csumo_features = NETIF_F_RXCSUM |
2933 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
2934 NETIF_F_HW_VLAN_CTAG_TX |
2935 NETIF_F_HW_VLAN_CTAG_RX;
2937 tso_features = NETIF_F_TSO |
2941 NETIF_F_GSO_UDP_TUNNEL |
2942 NETIF_F_GSO_GRE_CSUM |
2943 NETIF_F_GSO_UDP_TUNNEL_CSUM |
2944 NETIF_F_GSO_PARTIAL |
2945 NETIF_F_GSO_IPXIP4 |
2946 NETIF_F_GSO_IPXIP6 |
2949 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
2950 NETIF_F_GSO_GRE_CSUM;
2951 /* set features that user can change */
2952 netdev->hw_features = dflt_features | csumo_features |
2953 vlano_features | tso_features;
2955 /* add support for HW_CSUM on packets with MPLS header */
2956 netdev->mpls_features = NETIF_F_HW_CSUM;
2958 /* enable features */
2959 netdev->features |= netdev->hw_features;
2960 /* encap and VLAN devices inherit default, csumo and tso features */
2961 netdev->hw_enc_features |= dflt_features | csumo_features |
2963 netdev->vlan_features |= dflt_features | csumo_features |
2968 * ice_cfg_netdev - Allocate, configure and register a netdev
2969 * @vsi: the VSI associated with the new netdev
2971 * Returns 0 on success, negative value on failure
2973 static int ice_cfg_netdev(struct ice_vsi *vsi)
2975 struct ice_pf *pf = vsi->back;
2976 struct ice_netdev_priv *np;
2977 struct net_device *netdev;
2978 u8 mac_addr[ETH_ALEN];
2980 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
2985 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
2986 vsi->netdev = netdev;
2987 np = netdev_priv(netdev);
2990 ice_set_netdev_features(netdev);
2992 ice_set_ops(netdev);
2994 if (vsi->type == ICE_VSI_PF) {
2995 SET_NETDEV_DEV(netdev, ice_pf_to_dev(pf));
2996 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
2997 ether_addr_copy(netdev->dev_addr, mac_addr);
2998 ether_addr_copy(netdev->perm_addr, mac_addr);
3001 netdev->priv_flags |= IFF_UNICAST_FLT;
3003 /* Setup netdev TC information */
3004 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3006 /* setup watchdog timeout value to be 5 second */
3007 netdev->watchdog_timeo = 5 * HZ;
3009 netdev->min_mtu = ETH_MIN_MTU;
3010 netdev->max_mtu = ICE_MAX_MTU;
3016 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3017 * @lut: Lookup table
3018 * @rss_table_size: Lookup table size
3019 * @rss_size: Range of queue number for hashing
3021 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3025 for (i = 0; i < rss_table_size; i++)
3026 lut[i] = i % rss_size;
3030 * ice_pf_vsi_setup - Set up a PF VSI
3031 * @pf: board private structure
3032 * @pi: pointer to the port_info instance
3034 * Returns pointer to the successfully allocated VSI software struct
3035 * on success, otherwise returns NULL on failure.
3037 static struct ice_vsi *
3038 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3040 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3044 * ice_ctrl_vsi_setup - Set up a control VSI
3045 * @pf: board private structure
3046 * @pi: pointer to the port_info instance
3048 * Returns pointer to the successfully allocated VSI software struct
3049 * on success, otherwise returns NULL on failure.
3051 static struct ice_vsi *
3052 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3054 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3058 * ice_lb_vsi_setup - Set up a loopback VSI
3059 * @pf: board private structure
3060 * @pi: pointer to the port_info instance
3062 * Returns pointer to the successfully allocated VSI software struct
3063 * on success, otherwise returns NULL on failure.
3066 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3068 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3072 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3073 * @netdev: network interface to be adjusted
3074 * @proto: unused protocol
3075 * @vid: VLAN ID to be added
3077 * net_device_ops implementation for adding VLAN IDs
3080 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3083 struct ice_netdev_priv *np = netdev_priv(netdev);
3084 struct ice_vsi *vsi = np->vsi;
3087 /* VLAN 0 is added by default during load/reset */
3091 /* Enable VLAN pruning when a VLAN other than 0 is added */
3092 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3093 ret = ice_cfg_vlan_pruning(vsi, true, false);
3098 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3099 * packets aren't pruned by the device's internal switch on Rx
3101 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3103 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3109 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3110 * @netdev: network interface to be adjusted
3111 * @proto: unused protocol
3112 * @vid: VLAN ID to be removed
3114 * net_device_ops implementation for removing VLAN IDs
3117 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3120 struct ice_netdev_priv *np = netdev_priv(netdev);
3121 struct ice_vsi *vsi = np->vsi;
3124 /* don't allow removal of VLAN 0 */
3128 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3131 ret = ice_vsi_kill_vlan(vsi, vid);
3135 /* Disable pruning when VLAN 0 is the only VLAN rule */
3136 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3137 ret = ice_cfg_vlan_pruning(vsi, false, false);
3139 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3144 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3145 * @pf: board private structure
3147 * Returns 0 on success, negative value on failure
3149 static int ice_setup_pf_sw(struct ice_pf *pf)
3151 struct ice_vsi *vsi;
3154 if (ice_is_reset_in_progress(pf->state))
3157 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3161 status = ice_cfg_netdev(vsi);
3164 goto unroll_vsi_setup;
3166 /* netdev has to be configured before setting frame size */
3167 ice_vsi_cfg_frame_size(vsi);
3169 /* Setup DCB netlink interface */
3170 ice_dcbnl_setup(vsi);
3172 /* registering the NAPI handler requires both the queues and
3173 * netdev to be created, which are done in ice_pf_vsi_setup()
3174 * and ice_cfg_netdev() respectively
3178 status = ice_set_cpu_rx_rmap(vsi);
3180 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3181 vsi->vsi_num, status);
3183 goto unroll_napi_add;
3185 status = ice_init_mac_fltr(pf);
3187 goto free_cpu_rx_map;
3192 ice_free_cpu_rx_rmap(vsi);
3198 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3199 free_netdev(vsi->netdev);
3205 ice_vsi_release(vsi);
3210 * ice_get_avail_q_count - Get count of queues in use
3211 * @pf_qmap: bitmap to get queue use count from
3212 * @lock: pointer to a mutex that protects access to pf_qmap
3213 * @size: size of the bitmap
3216 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3222 for_each_clear_bit(bit, pf_qmap, size)
3230 * ice_get_avail_txq_count - Get count of Tx queues in use
3231 * @pf: pointer to an ice_pf instance
3233 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3235 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3240 * ice_get_avail_rxq_count - Get count of Rx queues in use
3241 * @pf: pointer to an ice_pf instance
3243 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3245 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3250 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3251 * @pf: board private structure to initialize
3253 static void ice_deinit_pf(struct ice_pf *pf)
3255 ice_service_task_stop(pf);
3256 mutex_destroy(&pf->sw_mutex);
3257 mutex_destroy(&pf->tc_mutex);
3258 mutex_destroy(&pf->avail_q_mutex);
3260 if (pf->avail_txqs) {
3261 bitmap_free(pf->avail_txqs);
3262 pf->avail_txqs = NULL;
3265 if (pf->avail_rxqs) {
3266 bitmap_free(pf->avail_rxqs);
3267 pf->avail_rxqs = NULL;
3272 * ice_set_pf_caps - set PFs capability flags
3273 * @pf: pointer to the PF instance
3275 static void ice_set_pf_caps(struct ice_pf *pf)
3277 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3279 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3280 if (func_caps->common_cap.dcb)
3281 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3282 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3283 if (func_caps->common_cap.sr_iov_1_1) {
3284 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3285 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3288 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3289 if (func_caps->common_cap.rss_table_size)
3290 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3292 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3293 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3296 /* ctrl_vsi_idx will be set to a valid value when flow director
3297 * is setup by ice_init_fdir
3299 pf->ctrl_vsi_idx = ICE_NO_VSI;
3300 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3301 /* force guaranteed filter pool for PF */
3302 ice_alloc_fd_guar_item(&pf->hw, &unused,
3303 func_caps->fd_fltr_guar);
3304 /* force shared filter pool for PF */
3305 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3306 func_caps->fd_fltr_best_effort);
3309 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3310 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3314 * ice_init_pf - Initialize general software structures (struct ice_pf)
3315 * @pf: board private structure to initialize
3317 static int ice_init_pf(struct ice_pf *pf)
3319 ice_set_pf_caps(pf);
3321 mutex_init(&pf->sw_mutex);
3322 mutex_init(&pf->tc_mutex);
3324 INIT_HLIST_HEAD(&pf->aq_wait_list);
3325 spin_lock_init(&pf->aq_wait_lock);
3326 init_waitqueue_head(&pf->aq_wait_queue);
3328 /* setup service timer and periodic service task */
3329 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3330 pf->serv_tmr_period = HZ;
3331 INIT_WORK(&pf->serv_task, ice_service_task);
3332 clear_bit(ICE_SERVICE_SCHED, pf->state);
3334 mutex_init(&pf->avail_q_mutex);
3335 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3336 if (!pf->avail_txqs)
3339 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3340 if (!pf->avail_rxqs) {
3341 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3342 pf->avail_txqs = NULL;
3350 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3351 * @pf: board private structure
3353 * compute the number of MSIX vectors required (v_budget) and request from
3354 * the OS. Return the number of vectors reserved or negative on failure
3356 static int ice_ena_msix_range(struct ice_pf *pf)
3358 int v_left, v_actual, v_other, v_budget = 0;
3359 struct device *dev = ice_pf_to_dev(pf);
3362 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3364 /* reserve for LAN miscellaneous handler */
3365 needed = ICE_MIN_LAN_OICR_MSIX;
3366 if (v_left < needed)
3367 goto no_hw_vecs_left_err;
3371 /* reserve for flow director */
3372 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3373 needed = ICE_FDIR_MSIX;
3374 if (v_left < needed)
3375 goto no_hw_vecs_left_err;
3380 /* total used for non-traffic vectors */
3383 /* reserve vectors for LAN traffic */
3384 needed = min_t(int, num_online_cpus(), v_left);
3385 if (v_left < needed)
3386 goto no_hw_vecs_left_err;
3387 pf->num_lan_msix = needed;
3391 pf->msix_entries = devm_kcalloc(dev, v_budget,
3392 sizeof(*pf->msix_entries), GFP_KERNEL);
3393 if (!pf->msix_entries) {
3398 for (i = 0; i < v_budget; i++)
3399 pf->msix_entries[i].entry = i;
3401 /* actually reserve the vectors */
3402 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3403 ICE_MIN_MSIX, v_budget);
3405 dev_err(dev, "unable to reserve MSI-X vectors\n");
3410 if (v_actual < v_budget) {
3411 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3412 v_budget, v_actual);
3414 if (v_actual < ICE_MIN_MSIX) {
3415 /* error if we can't get minimum vectors */
3416 pci_disable_msix(pf->pdev);
3420 int v_traffic = v_actual - v_other;
3422 if (v_actual == ICE_MIN_MSIX ||
3423 v_traffic < ICE_MIN_LAN_TXRX_MSIX)
3424 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3426 pf->num_lan_msix = v_traffic;
3428 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3436 devm_kfree(dev, pf->msix_entries);
3439 no_hw_vecs_left_err:
3440 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3444 pf->num_lan_msix = 0;
3449 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3450 * @pf: board private structure
3452 static void ice_dis_msix(struct ice_pf *pf)
3454 pci_disable_msix(pf->pdev);
3455 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3456 pf->msix_entries = NULL;
3460 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3461 * @pf: board private structure
3463 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3467 if (pf->irq_tracker) {
3468 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3469 pf->irq_tracker = NULL;
3474 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3475 * @pf: board private structure to initialize
3477 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3481 vectors = ice_ena_msix_range(pf);
3486 /* set up vector assignment tracking */
3487 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3488 struct_size(pf->irq_tracker, list, vectors),
3490 if (!pf->irq_tracker) {
3495 /* populate SW interrupts pool with number of OS granted IRQs. */
3496 pf->num_avail_sw_msix = (u16)vectors;
3497 pf->irq_tracker->num_entries = (u16)vectors;
3498 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3504 * ice_is_wol_supported - check if WoL is supported
3505 * @hw: pointer to hardware info
3507 * Check if WoL is supported based on the HW configuration.
3508 * Returns true if NVM supports and enables WoL for this port, false otherwise
3510 bool ice_is_wol_supported(struct ice_hw *hw)
3514 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3515 * word) indicates WoL is not supported on the corresponding PF ID.
3517 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3520 return !(BIT(hw->port_info->lport) & wol_ctrl);
3524 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3525 * @vsi: VSI being changed
3526 * @new_rx: new number of Rx queues
3527 * @new_tx: new number of Tx queues
3529 * Only change the number of queues if new_tx, or new_rx is non-0.
3531 * Returns 0 on success.
3533 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3535 struct ice_pf *pf = vsi->back;
3536 int err = 0, timeout = 50;
3538 if (!new_rx && !new_tx)
3541 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3545 usleep_range(1000, 2000);
3549 vsi->req_txq = (u16)new_tx;
3551 vsi->req_rxq = (u16)new_rx;
3553 /* set for the next time the netdev is started */
3554 if (!netif_running(vsi->netdev)) {
3555 ice_vsi_rebuild(vsi, false);
3556 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3561 ice_vsi_rebuild(vsi, false);
3562 ice_pf_dcb_recfg(pf);
3565 clear_bit(ICE_CFG_BUSY, pf->state);
3570 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3571 * @pf: PF to configure
3573 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3574 * VSI can still Tx/Rx VLAN tagged packets.
3576 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3578 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3579 struct ice_vsi_ctx *ctxt;
3580 enum ice_status status;
3586 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3591 ctxt->info = vsi->info;
3593 ctxt->info.valid_sections =
3594 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3595 ICE_AQ_VSI_PROP_SECURITY_VALID |
3596 ICE_AQ_VSI_PROP_SW_VALID);
3598 /* disable VLAN anti-spoof */
3599 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3600 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3602 /* disable VLAN pruning and keep all other settings */
3603 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3605 /* allow all VLANs on Tx and don't strip on Rx */
3606 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3607 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3609 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3611 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3612 ice_stat_str(status),
3613 ice_aq_str(hw->adminq.sq_last_status));
3615 vsi->info.sec_flags = ctxt->info.sec_flags;
3616 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3617 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3624 * ice_log_pkg_init - log result of DDP package load
3625 * @hw: pointer to hardware info
3626 * @status: status of package load
3629 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3631 struct ice_pf *pf = (struct ice_pf *)hw->back;
3632 struct device *dev = ice_pf_to_dev(pf);
3636 /* The package download AdminQ command returned success because
3637 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3638 * already a package loaded on the device.
3640 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3641 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3642 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3643 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3644 !memcmp(hw->pkg_name, hw->active_pkg_name,
3645 sizeof(hw->pkg_name))) {
3646 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3647 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3648 hw->active_pkg_name,
3649 hw->active_pkg_ver.major,
3650 hw->active_pkg_ver.minor,
3651 hw->active_pkg_ver.update,
3652 hw->active_pkg_ver.draft);
3654 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3655 hw->active_pkg_name,
3656 hw->active_pkg_ver.major,
3657 hw->active_pkg_ver.minor,
3658 hw->active_pkg_ver.update,
3659 hw->active_pkg_ver.draft);
3660 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3661 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3662 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",
3663 hw->active_pkg_name,
3664 hw->active_pkg_ver.major,
3665 hw->active_pkg_ver.minor,
3666 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3667 *status = ICE_ERR_NOT_SUPPORTED;
3668 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3669 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3670 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",
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,
3682 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");
3683 *status = ICE_ERR_NOT_SUPPORTED;
3686 case ICE_ERR_FW_DDP_MISMATCH:
3687 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");
3689 case ICE_ERR_BUF_TOO_SHORT:
3691 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3693 case ICE_ERR_NOT_SUPPORTED:
3694 /* Package File version not supported */
3695 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3696 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3697 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3698 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3699 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3700 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3701 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3702 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",
3703 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3705 case ICE_ERR_AQ_ERROR:
3706 switch (hw->pkg_dwnld_status) {
3707 case ICE_AQ_RC_ENOSEC:
3708 case ICE_AQ_RC_EBADSIG:
3709 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");
3711 case ICE_AQ_RC_ESVN:
3712 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");
3714 case ICE_AQ_RC_EBADMAN:
3715 case ICE_AQ_RC_EBADBUF:
3716 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3717 /* poll for reset to complete */
3718 if (ice_check_reset(hw))
3719 dev_err(dev, "Error resetting device. Please reload the driver\n");
3726 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3733 * ice_load_pkg - load/reload the DDP Package file
3734 * @firmware: firmware structure when firmware requested or NULL for reload
3735 * @pf: pointer to the PF instance
3737 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3738 * initialize HW tables.
3741 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3743 enum ice_status status = ICE_ERR_PARAM;
3744 struct device *dev = ice_pf_to_dev(pf);
3745 struct ice_hw *hw = &pf->hw;
3747 /* Load DDP Package */
3748 if (firmware && !hw->pkg_copy) {
3749 status = ice_copy_and_init_pkg(hw, firmware->data,
3751 ice_log_pkg_init(hw, &status);
3752 } else if (!firmware && hw->pkg_copy) {
3753 /* Reload package during rebuild after CORER/GLOBR reset */
3754 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3755 ice_log_pkg_init(hw, &status);
3757 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3762 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3766 /* Successful download package is the precondition for advanced
3767 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3769 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3773 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3774 * @pf: pointer to the PF structure
3776 * There is no error returned here because the driver should be able to handle
3777 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3778 * specifically with Tx.
3780 static void ice_verify_cacheline_size(struct ice_pf *pf)
3782 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3783 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3784 ICE_CACHE_LINE_BYTES);
3788 * ice_send_version - update firmware with driver version
3791 * Returns ICE_SUCCESS on success, else error code
3793 static enum ice_status ice_send_version(struct ice_pf *pf)
3795 struct ice_driver_ver dv;
3797 dv.major_ver = 0xff;
3798 dv.minor_ver = 0xff;
3799 dv.build_ver = 0xff;
3800 dv.subbuild_ver = 0;
3801 strscpy((char *)dv.driver_string, UTS_RELEASE,
3802 sizeof(dv.driver_string));
3803 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
3807 * ice_init_fdir - Initialize flow director VSI and configuration
3808 * @pf: pointer to the PF instance
3810 * returns 0 on success, negative on error
3812 static int ice_init_fdir(struct ice_pf *pf)
3814 struct device *dev = ice_pf_to_dev(pf);
3815 struct ice_vsi *ctrl_vsi;
3818 /* Side Band Flow Director needs to have a control VSI.
3819 * Allocate it and store it in the PF.
3821 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
3823 dev_dbg(dev, "could not create control VSI\n");
3827 err = ice_vsi_open_ctrl(ctrl_vsi);
3829 dev_dbg(dev, "could not open control VSI\n");
3833 mutex_init(&pf->hw.fdir_fltr_lock);
3835 err = ice_fdir_create_dflt_rules(pf);
3842 ice_fdir_release_flows(&pf->hw);
3843 ice_vsi_close(ctrl_vsi);
3845 ice_vsi_release(ctrl_vsi);
3846 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
3847 pf->vsi[pf->ctrl_vsi_idx] = NULL;
3848 pf->ctrl_vsi_idx = ICE_NO_VSI;
3854 * ice_get_opt_fw_name - return optional firmware file name or NULL
3855 * @pf: pointer to the PF instance
3857 static char *ice_get_opt_fw_name(struct ice_pf *pf)
3859 /* Optional firmware name same as default with additional dash
3860 * followed by a EUI-64 identifier (PCIe Device Serial Number)
3862 struct pci_dev *pdev = pf->pdev;
3863 char *opt_fw_filename;
3866 /* Determine the name of the optional file using the DSN (two
3867 * dwords following the start of the DSN Capability).
3869 dsn = pci_get_dsn(pdev);
3873 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
3874 if (!opt_fw_filename)
3877 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
3878 ICE_DDP_PKG_PATH, dsn);
3880 return opt_fw_filename;
3884 * ice_request_fw - Device initialization routine
3885 * @pf: pointer to the PF instance
3887 static void ice_request_fw(struct ice_pf *pf)
3889 char *opt_fw_filename = ice_get_opt_fw_name(pf);
3890 const struct firmware *firmware = NULL;
3891 struct device *dev = ice_pf_to_dev(pf);
3894 /* optional device-specific DDP (if present) overrides the default DDP
3895 * package file. kernel logs a debug message if the file doesn't exist,
3896 * and warning messages for other errors.
3898 if (opt_fw_filename) {
3899 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
3901 kfree(opt_fw_filename);
3905 /* request for firmware was successful. Download to device */
3906 ice_load_pkg(firmware, pf);
3907 kfree(opt_fw_filename);
3908 release_firmware(firmware);
3913 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
3915 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
3919 /* request for firmware was successful. Download to device */
3920 ice_load_pkg(firmware, pf);
3921 release_firmware(firmware);
3925 * ice_print_wake_reason - show the wake up cause in the log
3926 * @pf: pointer to the PF struct
3928 static void ice_print_wake_reason(struct ice_pf *pf)
3930 u32 wus = pf->wakeup_reason;
3931 const char *wake_str;
3933 /* if no wake event, nothing to print */
3937 if (wus & PFPM_WUS_LNKC_M)
3938 wake_str = "Link\n";
3939 else if (wus & PFPM_WUS_MAG_M)
3940 wake_str = "Magic Packet\n";
3941 else if (wus & PFPM_WUS_MNG_M)
3942 wake_str = "Management\n";
3943 else if (wus & PFPM_WUS_FW_RST_WK_M)
3944 wake_str = "Firmware Reset\n";
3946 wake_str = "Unknown\n";
3948 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
3952 * ice_register_netdev - register netdev and devlink port
3953 * @pf: pointer to the PF struct
3955 static int ice_register_netdev(struct ice_pf *pf)
3957 struct ice_vsi *vsi;
3960 vsi = ice_get_main_vsi(pf);
3961 if (!vsi || !vsi->netdev)
3964 err = register_netdev(vsi->netdev);
3966 goto err_register_netdev;
3968 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
3969 netif_carrier_off(vsi->netdev);
3970 netif_tx_stop_all_queues(vsi->netdev);
3971 err = ice_devlink_create_port(vsi);
3973 goto err_devlink_create;
3975 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
3979 unregister_netdev(vsi->netdev);
3980 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
3981 err_register_netdev:
3982 free_netdev(vsi->netdev);
3984 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3989 * ice_probe - Device initialization routine
3990 * @pdev: PCI device information struct
3991 * @ent: entry in ice_pci_tbl
3993 * Returns 0 on success, negative on failure
3996 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
3998 struct device *dev = &pdev->dev;
4003 /* this driver uses devres, see
4004 * Documentation/driver-api/driver-model/devres.rst
4006 err = pcim_enable_device(pdev);
4010 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4012 dev_err(dev, "BAR0 I/O map error %d\n", err);
4016 pf = ice_allocate_pf(dev);
4020 /* set up for high or low DMA */
4021 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4023 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4025 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4029 pci_enable_pcie_error_reporting(pdev);
4030 pci_set_master(pdev);
4033 pci_set_drvdata(pdev, pf);
4034 set_bit(ICE_DOWN, pf->state);
4035 /* Disable service task until DOWN bit is cleared */
4036 set_bit(ICE_SERVICE_DIS, pf->state);
4039 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4040 pci_save_state(pdev);
4043 hw->vendor_id = pdev->vendor;
4044 hw->device_id = pdev->device;
4045 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4046 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4047 hw->subsystem_device_id = pdev->subsystem_device;
4048 hw->bus.device = PCI_SLOT(pdev->devfn);
4049 hw->bus.func = PCI_FUNC(pdev->devfn);
4050 ice_set_ctrlq_len(hw);
4052 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4054 err = ice_devlink_register(pf);
4056 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4057 goto err_exit_unroll;
4060 #ifndef CONFIG_DYNAMIC_DEBUG
4062 hw->debug_mask = debug;
4065 err = ice_init_hw(hw);
4067 dev_err(dev, "ice_init_hw failed: %d\n", err);
4069 goto err_exit_unroll;
4074 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4075 * set in pf->state, which will cause ice_is_safe_mode to return
4078 if (ice_is_safe_mode(pf)) {
4079 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4080 /* we already got function/device capabilities but these don't
4081 * reflect what the driver needs to do in safe mode. Instead of
4082 * adding conditional logic everywhere to ignore these
4083 * device/function capabilities, override them.
4085 ice_set_safe_mode_caps(hw);
4088 err = ice_init_pf(pf);
4090 dev_err(dev, "ice_init_pf failed: %d\n", err);
4091 goto err_init_pf_unroll;
4094 ice_devlink_init_regions(pf);
4096 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4097 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4098 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4099 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4101 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4102 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4103 pf->hw.tnl.valid_count[TNL_VXLAN];
4104 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4105 UDP_TUNNEL_TYPE_VXLAN;
4108 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4109 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4110 pf->hw.tnl.valid_count[TNL_GENEVE];
4111 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4112 UDP_TUNNEL_TYPE_GENEVE;
4116 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4117 if (!pf->num_alloc_vsi) {
4119 goto err_init_pf_unroll;
4121 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4122 dev_warn(&pf->pdev->dev,
4123 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4124 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4125 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4128 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4132 goto err_init_pf_unroll;
4135 err = ice_init_interrupt_scheme(pf);
4137 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4139 goto err_init_vsi_unroll;
4142 /* In case of MSIX we are going to setup the misc vector right here
4143 * to handle admin queue events etc. In case of legacy and MSI
4144 * the misc functionality and queue processing is combined in
4145 * the same vector and that gets setup at open.
4147 err = ice_req_irq_msix_misc(pf);
4149 dev_err(dev, "setup of misc vector failed: %d\n", err);
4150 goto err_init_interrupt_unroll;
4153 /* create switch struct for the switch element created by FW on boot */
4154 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4155 if (!pf->first_sw) {
4157 goto err_msix_misc_unroll;
4161 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4163 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4165 pf->first_sw->pf = pf;
4167 /* record the sw_id available for later use */
4168 pf->first_sw->sw_id = hw->port_info->sw_id;
4170 err = ice_setup_pf_sw(pf);
4172 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4173 goto err_alloc_sw_unroll;
4176 clear_bit(ICE_SERVICE_DIS, pf->state);
4178 /* tell the firmware we are up */
4179 err = ice_send_version(pf);
4181 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4183 goto err_send_version_unroll;
4186 /* since everything is good, start the service timer */
4187 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4189 err = ice_init_link_events(pf->hw.port_info);
4191 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4192 goto err_send_version_unroll;
4195 /* not a fatal error if this fails */
4196 err = ice_init_nvm_phy_type(pf->hw.port_info);
4198 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4200 /* not a fatal error if this fails */
4201 err = ice_update_link_info(pf->hw.port_info);
4203 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4205 ice_init_link_dflt_override(pf->hw.port_info);
4207 /* if media available, initialize PHY settings */
4208 if (pf->hw.port_info->phy.link_info.link_info &
4209 ICE_AQ_MEDIA_AVAILABLE) {
4210 /* not a fatal error if this fails */
4211 err = ice_init_phy_user_cfg(pf->hw.port_info);
4213 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4215 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4216 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4219 ice_configure_phy(vsi);
4222 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4225 ice_verify_cacheline_size(pf);
4227 /* Save wakeup reason register for later use */
4228 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4230 /* check for a power management event */
4231 ice_print_wake_reason(pf);
4233 /* clear wake status, all bits */
4234 wr32(hw, PFPM_WUS, U32_MAX);
4236 /* Disable WoL at init, wait for user to enable */
4237 device_set_wakeup_enable(dev, false);
4239 if (ice_is_safe_mode(pf)) {
4240 ice_set_safe_mode_vlan_cfg(pf);
4244 /* initialize DDP driven features */
4246 /* Note: Flow director init failure is non-fatal to load */
4247 if (ice_init_fdir(pf))
4248 dev_err(dev, "could not initialize flow director\n");
4250 /* Note: DCB init failure is non-fatal to load */
4251 if (ice_init_pf_dcb(pf, false)) {
4252 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4253 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4255 ice_cfg_lldp_mib_change(&pf->hw, true);
4258 if (ice_init_lag(pf))
4259 dev_warn(dev, "Failed to init link aggregation support\n");
4261 /* print PCI link speed and width */
4262 pcie_print_link_status(pf->pdev);
4265 err = ice_register_netdev(pf);
4267 goto err_netdev_reg;
4269 /* ready to go, so clear down state bit */
4270 clear_bit(ICE_DOWN, pf->state);
4274 err_send_version_unroll:
4275 ice_vsi_release_all(pf);
4276 err_alloc_sw_unroll:
4277 set_bit(ICE_SERVICE_DIS, pf->state);
4278 set_bit(ICE_DOWN, pf->state);
4279 devm_kfree(dev, pf->first_sw);
4280 err_msix_misc_unroll:
4281 ice_free_irq_msix_misc(pf);
4282 err_init_interrupt_unroll:
4283 ice_clear_interrupt_scheme(pf);
4284 err_init_vsi_unroll:
4285 devm_kfree(dev, pf->vsi);
4288 ice_devlink_destroy_regions(pf);
4291 ice_devlink_unregister(pf);
4292 pci_disable_pcie_error_reporting(pdev);
4293 pci_disable_device(pdev);
4298 * ice_set_wake - enable or disable Wake on LAN
4299 * @pf: pointer to the PF struct
4301 * Simple helper for WoL control
4303 static void ice_set_wake(struct ice_pf *pf)
4305 struct ice_hw *hw = &pf->hw;
4306 bool wol = pf->wol_ena;
4308 /* clear wake state, otherwise new wake events won't fire */
4309 wr32(hw, PFPM_WUS, U32_MAX);
4311 /* enable / disable APM wake up, no RMW needed */
4312 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4314 /* set magic packet filter enabled */
4315 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4319 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4320 * @pf: pointer to the PF struct
4322 * Issue firmware command to enable multicast magic wake, making
4323 * sure that any locally administered address (LAA) is used for
4324 * wake, and that PF reset doesn't undo the LAA.
4326 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4328 struct device *dev = ice_pf_to_dev(pf);
4329 struct ice_hw *hw = &pf->hw;
4330 enum ice_status status;
4331 u8 mac_addr[ETH_ALEN];
4332 struct ice_vsi *vsi;
4338 vsi = ice_get_main_vsi(pf);
4342 /* Get current MAC address in case it's an LAA */
4344 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4346 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4348 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4349 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4350 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4352 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4354 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4355 ice_stat_str(status),
4356 ice_aq_str(hw->adminq.sq_last_status));
4360 * ice_remove - Device removal routine
4361 * @pdev: PCI device information struct
4363 static void ice_remove(struct pci_dev *pdev)
4365 struct ice_pf *pf = pci_get_drvdata(pdev);
4371 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4372 if (!ice_is_reset_in_progress(pf->state))
4377 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4378 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4382 set_bit(ICE_DOWN, pf->state);
4383 ice_service_task_stop(pf);
4385 ice_aq_cancel_waiting_tasks(pf);
4387 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4389 if (!ice_is_safe_mode(pf))
4390 ice_remove_arfs(pf);
4391 ice_setup_mc_magic_wake(pf);
4392 ice_vsi_release_all(pf);
4394 ice_free_irq_msix_misc(pf);
4395 ice_for_each_vsi(pf, i) {
4398 ice_vsi_free_q_vectors(pf->vsi[i]);
4401 ice_devlink_destroy_regions(pf);
4402 ice_deinit_hw(&pf->hw);
4403 ice_devlink_unregister(pf);
4405 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4406 * do it via ice_schedule_reset() since there is no need to rebuild
4407 * and the service task is already stopped.
4409 ice_reset(&pf->hw, ICE_RESET_PFR);
4410 pci_wait_for_pending_transaction(pdev);
4411 ice_clear_interrupt_scheme(pf);
4412 pci_disable_pcie_error_reporting(pdev);
4413 pci_disable_device(pdev);
4417 * ice_shutdown - PCI callback for shutting down device
4418 * @pdev: PCI device information struct
4420 static void ice_shutdown(struct pci_dev *pdev)
4422 struct ice_pf *pf = pci_get_drvdata(pdev);
4426 if (system_state == SYSTEM_POWER_OFF) {
4427 pci_wake_from_d3(pdev, pf->wol_ena);
4428 pci_set_power_state(pdev, PCI_D3hot);
4434 * ice_prepare_for_shutdown - prep for PCI shutdown
4435 * @pf: board private structure
4437 * Inform or close all dependent features in prep for PCI device shutdown
4439 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4441 struct ice_hw *hw = &pf->hw;
4444 /* Notify VFs of impending reset */
4445 if (ice_check_sq_alive(hw, &hw->mailboxq))
4446 ice_vc_notify_reset(pf);
4448 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4450 /* disable the VSIs and their queues that are not already DOWN */
4451 ice_pf_dis_all_vsi(pf, false);
4453 ice_for_each_vsi(pf, v)
4455 pf->vsi[v]->vsi_num = 0;
4457 ice_shutdown_all_ctrlq(hw);
4461 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4462 * @pf: board private structure to reinitialize
4464 * This routine reinitialize interrupt scheme that was cleared during
4465 * power management suspend callback.
4467 * This should be called during resume routine to re-allocate the q_vectors
4468 * and reacquire interrupts.
4470 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4472 struct device *dev = ice_pf_to_dev(pf);
4475 /* Since we clear MSIX flag during suspend, we need to
4476 * set it back during resume...
4479 ret = ice_init_interrupt_scheme(pf);
4481 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4485 /* Remap vectors and rings, after successful re-init interrupts */
4486 ice_for_each_vsi(pf, v) {
4490 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4493 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4496 ret = ice_req_irq_msix_misc(pf);
4498 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4508 ice_vsi_free_q_vectors(pf->vsi[v]);
4515 * @dev: generic device information structure
4517 * Power Management callback to quiesce the device and prepare
4518 * for D3 transition.
4520 static int __maybe_unused ice_suspend(struct device *dev)
4522 struct pci_dev *pdev = to_pci_dev(dev);
4526 pf = pci_get_drvdata(pdev);
4528 if (!ice_pf_state_is_nominal(pf)) {
4529 dev_err(dev, "Device is not ready, no need to suspend it\n");
4533 /* Stop watchdog tasks until resume completion.
4534 * Even though it is most likely that the service task is
4535 * disabled if the device is suspended or down, the service task's
4536 * state is controlled by a different state bit, and we should
4537 * store and honor whatever state that bit is in at this point.
4539 disabled = ice_service_task_stop(pf);
4541 /* Already suspended?, then there is nothing to do */
4542 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
4544 ice_service_task_restart(pf);
4548 if (test_bit(ICE_DOWN, pf->state) ||
4549 ice_is_reset_in_progress(pf->state)) {
4550 dev_err(dev, "can't suspend device in reset or already down\n");
4552 ice_service_task_restart(pf);
4556 ice_setup_mc_magic_wake(pf);
4558 ice_prepare_for_shutdown(pf);
4562 /* Free vectors, clear the interrupt scheme and release IRQs
4563 * for proper hibernation, especially with large number of CPUs.
4564 * Otherwise hibernation might fail when mapping all the vectors back
4567 ice_free_irq_msix_misc(pf);
4568 ice_for_each_vsi(pf, v) {
4571 ice_vsi_free_q_vectors(pf->vsi[v]);
4573 ice_free_cpu_rx_rmap(ice_get_main_vsi(pf));
4574 ice_clear_interrupt_scheme(pf);
4576 pci_save_state(pdev);
4577 pci_wake_from_d3(pdev, pf->wol_ena);
4578 pci_set_power_state(pdev, PCI_D3hot);
4583 * ice_resume - PM callback for waking up from D3
4584 * @dev: generic device information structure
4586 static int __maybe_unused ice_resume(struct device *dev)
4588 struct pci_dev *pdev = to_pci_dev(dev);
4589 enum ice_reset_req reset_type;
4594 pci_set_power_state(pdev, PCI_D0);
4595 pci_restore_state(pdev);
4596 pci_save_state(pdev);
4598 if (!pci_device_is_present(pdev))
4601 ret = pci_enable_device_mem(pdev);
4603 dev_err(dev, "Cannot enable device after suspend\n");
4607 pf = pci_get_drvdata(pdev);
4610 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4611 ice_print_wake_reason(pf);
4613 /* We cleared the interrupt scheme when we suspended, so we need to
4614 * restore it now to resume device functionality.
4616 ret = ice_reinit_interrupt_scheme(pf);
4618 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4620 clear_bit(ICE_DOWN, pf->state);
4621 /* Now perform PF reset and rebuild */
4622 reset_type = ICE_RESET_PFR;
4623 /* re-enable service task for reset, but allow reset to schedule it */
4624 clear_bit(ICE_SERVICE_DIS, pf->state);
4626 if (ice_schedule_reset(pf, reset_type))
4627 dev_err(dev, "Reset during resume failed.\n");
4629 clear_bit(ICE_SUSPENDED, pf->state);
4630 ice_service_task_restart(pf);
4632 /* Restart the service task */
4633 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4637 #endif /* CONFIG_PM */
4640 * ice_pci_err_detected - warning that PCI error has been detected
4641 * @pdev: PCI device information struct
4642 * @err: the type of PCI error
4644 * Called to warn that something happened on the PCI bus and the error handling
4645 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4647 static pci_ers_result_t
4648 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4650 struct ice_pf *pf = pci_get_drvdata(pdev);
4653 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4655 return PCI_ERS_RESULT_DISCONNECT;
4658 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4659 ice_service_task_stop(pf);
4661 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4662 set_bit(ICE_PFR_REQ, pf->state);
4663 ice_prepare_for_reset(pf);
4667 return PCI_ERS_RESULT_NEED_RESET;
4671 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4672 * @pdev: PCI device information struct
4674 * Called to determine if the driver can recover from the PCI slot reset by
4675 * using a register read to determine if the device is recoverable.
4677 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4679 struct ice_pf *pf = pci_get_drvdata(pdev);
4680 pci_ers_result_t result;
4684 err = pci_enable_device_mem(pdev);
4686 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4688 result = PCI_ERS_RESULT_DISCONNECT;
4690 pci_set_master(pdev);
4691 pci_restore_state(pdev);
4692 pci_save_state(pdev);
4693 pci_wake_from_d3(pdev, false);
4695 /* Check for life */
4696 reg = rd32(&pf->hw, GLGEN_RTRIG);
4698 result = PCI_ERS_RESULT_RECOVERED;
4700 result = PCI_ERS_RESULT_DISCONNECT;
4703 err = pci_aer_clear_nonfatal_status(pdev);
4705 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4707 /* non-fatal, continue */
4713 * ice_pci_err_resume - restart operations after PCI error recovery
4714 * @pdev: PCI device information struct
4716 * Called to allow the driver to bring things back up after PCI error and/or
4717 * reset recovery have finished
4719 static void ice_pci_err_resume(struct pci_dev *pdev)
4721 struct ice_pf *pf = pci_get_drvdata(pdev);
4724 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4729 if (test_bit(ICE_SUSPENDED, pf->state)) {
4730 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4735 ice_restore_all_vfs_msi_state(pdev);
4737 ice_do_reset(pf, ICE_RESET_PFR);
4738 ice_service_task_restart(pf);
4739 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4743 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4744 * @pdev: PCI device information struct
4746 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4748 struct ice_pf *pf = pci_get_drvdata(pdev);
4750 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4751 ice_service_task_stop(pf);
4753 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4754 set_bit(ICE_PFR_REQ, pf->state);
4755 ice_prepare_for_reset(pf);
4761 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
4762 * @pdev: PCI device information struct
4764 static void ice_pci_err_reset_done(struct pci_dev *pdev)
4766 ice_pci_err_resume(pdev);
4769 /* ice_pci_tbl - PCI Device ID Table
4771 * Wildcard entries (PCI_ANY_ID) should come last
4772 * Last entry must be all 0s
4774 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
4775 * Class, Class Mask, private data (not used) }
4777 static const struct pci_device_id ice_pci_tbl[] = {
4778 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
4779 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
4780 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
4781 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
4782 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
4783 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
4784 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
4785 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
4786 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
4787 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
4788 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
4789 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
4790 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
4791 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
4792 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
4793 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
4794 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
4795 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
4796 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
4797 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
4798 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
4799 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
4800 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
4801 /* required last entry */
4804 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
4806 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
4808 static const struct pci_error_handlers ice_pci_err_handler = {
4809 .error_detected = ice_pci_err_detected,
4810 .slot_reset = ice_pci_err_slot_reset,
4811 .reset_prepare = ice_pci_err_reset_prepare,
4812 .reset_done = ice_pci_err_reset_done,
4813 .resume = ice_pci_err_resume
4816 static struct pci_driver ice_driver = {
4817 .name = KBUILD_MODNAME,
4818 .id_table = ice_pci_tbl,
4820 .remove = ice_remove,
4822 .driver.pm = &ice_pm_ops,
4823 #endif /* CONFIG_PM */
4824 .shutdown = ice_shutdown,
4825 .sriov_configure = ice_sriov_configure,
4826 .err_handler = &ice_pci_err_handler
4830 * ice_module_init - Driver registration routine
4832 * ice_module_init is the first routine called when the driver is
4833 * loaded. All it does is register with the PCI subsystem.
4835 static int __init ice_module_init(void)
4839 pr_info("%s\n", ice_driver_string);
4840 pr_info("%s\n", ice_copyright);
4842 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
4844 pr_err("Failed to create workqueue\n");
4848 status = pci_register_driver(&ice_driver);
4850 pr_err("failed to register PCI driver, err %d\n", status);
4851 destroy_workqueue(ice_wq);
4856 module_init(ice_module_init);
4859 * ice_module_exit - Driver exit cleanup routine
4861 * ice_module_exit is called just before the driver is removed
4864 static void __exit ice_module_exit(void)
4866 pci_unregister_driver(&ice_driver);
4867 destroy_workqueue(ice_wq);
4868 pr_info("module unloaded\n");
4870 module_exit(ice_module_exit);
4873 * ice_set_mac_address - NDO callback to set MAC address
4874 * @netdev: network interface device structure
4875 * @pi: pointer to an address structure
4877 * Returns 0 on success, negative on failure
4879 static int ice_set_mac_address(struct net_device *netdev, void *pi)
4881 struct ice_netdev_priv *np = netdev_priv(netdev);
4882 struct ice_vsi *vsi = np->vsi;
4883 struct ice_pf *pf = vsi->back;
4884 struct ice_hw *hw = &pf->hw;
4885 struct sockaddr *addr = pi;
4886 enum ice_status status;
4891 mac = (u8 *)addr->sa_data;
4893 if (!is_valid_ether_addr(mac))
4894 return -EADDRNOTAVAIL;
4896 if (ether_addr_equal(netdev->dev_addr, mac)) {
4897 netdev_warn(netdev, "already using mac %pM\n", mac);
4901 if (test_bit(ICE_DOWN, pf->state) ||
4902 ice_is_reset_in_progress(pf->state)) {
4903 netdev_err(netdev, "can't set mac %pM. device not ready\n",
4908 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
4909 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
4910 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
4911 err = -EADDRNOTAVAIL;
4912 goto err_update_filters;
4915 /* Add filter for new MAC. If filter exists, return success */
4916 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
4917 if (status == ICE_ERR_ALREADY_EXISTS) {
4918 /* Although this MAC filter is already present in hardware it's
4919 * possible in some cases (e.g. bonding) that dev_addr was
4920 * modified outside of the driver and needs to be restored back
4923 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4924 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
4928 /* error if the new filter addition failed */
4930 err = -EADDRNOTAVAIL;
4934 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
4939 /* change the netdev's MAC address */
4940 memcpy(netdev->dev_addr, mac, netdev->addr_len);
4941 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
4944 /* write new MAC address to the firmware */
4945 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
4946 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
4948 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
4949 mac, ice_stat_str(status));
4955 * ice_set_rx_mode - NDO callback to set the netdev filters
4956 * @netdev: network interface device structure
4958 static void ice_set_rx_mode(struct net_device *netdev)
4960 struct ice_netdev_priv *np = netdev_priv(netdev);
4961 struct ice_vsi *vsi = np->vsi;
4966 /* Set the flags to synchronize filters
4967 * ndo_set_rx_mode may be triggered even without a change in netdev
4970 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
4971 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
4972 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
4974 /* schedule our worker thread which will take care of
4975 * applying the new filter changes
4977 ice_service_task_schedule(vsi->back);
4981 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
4982 * @netdev: network interface device structure
4983 * @queue_index: Queue ID
4984 * @maxrate: maximum bandwidth in Mbps
4987 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
4989 struct ice_netdev_priv *np = netdev_priv(netdev);
4990 struct ice_vsi *vsi = np->vsi;
4991 enum ice_status status;
4995 /* Validate maxrate requested is within permitted range */
4996 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
4997 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
4998 maxrate, queue_index);
5002 q_handle = vsi->tx_rings[queue_index]->q_handle;
5003 tc = ice_dcb_get_tc(vsi, queue_index);
5005 /* Set BW back to default, when user set maxrate to 0 */
5007 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5008 q_handle, ICE_MAX_BW);
5010 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5011 q_handle, ICE_MAX_BW, maxrate * 1000);
5013 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
5014 ice_stat_str(status));
5022 * ice_fdb_add - add an entry to the hardware database
5023 * @ndm: the input from the stack
5024 * @tb: pointer to array of nladdr (unused)
5025 * @dev: the net device pointer
5026 * @addr: the MAC address entry being added
5028 * @flags: instructions from stack about fdb operation
5029 * @extack: netlink extended ack
5032 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5033 struct net_device *dev, const unsigned char *addr, u16 vid,
5034 u16 flags, struct netlink_ext_ack __always_unused *extack)
5039 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5042 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5043 netdev_err(dev, "FDB only supports static addresses\n");
5047 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5048 err = dev_uc_add_excl(dev, addr);
5049 else if (is_multicast_ether_addr(addr))
5050 err = dev_mc_add_excl(dev, addr);
5054 /* Only return duplicate errors if NLM_F_EXCL is set */
5055 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5062 * ice_fdb_del - delete an entry from the hardware database
5063 * @ndm: the input from the stack
5064 * @tb: pointer to array of nladdr (unused)
5065 * @dev: the net device pointer
5066 * @addr: the MAC address entry being added
5070 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5071 struct net_device *dev, const unsigned char *addr,
5072 __always_unused u16 vid)
5076 if (ndm->ndm_state & NUD_PERMANENT) {
5077 netdev_err(dev, "FDB only supports static addresses\n");
5081 if (is_unicast_ether_addr(addr))
5082 err = dev_uc_del(dev, addr);
5083 else if (is_multicast_ether_addr(addr))
5084 err = dev_mc_del(dev, addr);
5092 * ice_set_features - set the netdev feature flags
5093 * @netdev: ptr to the netdev being adjusted
5094 * @features: the feature set that the stack is suggesting
5097 ice_set_features(struct net_device *netdev, netdev_features_t features)
5099 struct ice_netdev_priv *np = netdev_priv(netdev);
5100 struct ice_vsi *vsi = np->vsi;
5101 struct ice_pf *pf = vsi->back;
5104 /* Don't set any netdev advanced features with device in Safe Mode */
5105 if (ice_is_safe_mode(vsi->back)) {
5106 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5110 /* Do not change setting during reset */
5111 if (ice_is_reset_in_progress(pf->state)) {
5112 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5116 /* Multiple features can be changed in one call so keep features in
5117 * separate if/else statements to guarantee each feature is checked
5119 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5120 ice_vsi_manage_rss_lut(vsi, true);
5121 else if (!(features & NETIF_F_RXHASH) &&
5122 netdev->features & NETIF_F_RXHASH)
5123 ice_vsi_manage_rss_lut(vsi, false);
5125 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5126 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5127 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5128 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5129 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5130 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5132 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5133 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5134 ret = ice_vsi_manage_vlan_insertion(vsi);
5135 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5136 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5137 ret = ice_vsi_manage_vlan_insertion(vsi);
5139 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5140 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5141 ret = ice_cfg_vlan_pruning(vsi, true, false);
5142 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5143 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5144 ret = ice_cfg_vlan_pruning(vsi, false, false);
5146 if ((features & NETIF_F_NTUPLE) &&
5147 !(netdev->features & NETIF_F_NTUPLE)) {
5148 ice_vsi_manage_fdir(vsi, true);
5150 } else if (!(features & NETIF_F_NTUPLE) &&
5151 (netdev->features & NETIF_F_NTUPLE)) {
5152 ice_vsi_manage_fdir(vsi, false);
5153 ice_clear_arfs(vsi);
5160 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5161 * @vsi: VSI to setup VLAN properties for
5163 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5167 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5168 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5169 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5170 ret = ice_vsi_manage_vlan_insertion(vsi);
5176 * ice_vsi_cfg - Setup the VSI
5177 * @vsi: the VSI being configured
5179 * Return 0 on success and negative value on error
5181 int ice_vsi_cfg(struct ice_vsi *vsi)
5186 ice_set_rx_mode(vsi->netdev);
5188 err = ice_vsi_vlan_setup(vsi);
5193 ice_vsi_cfg_dcb_rings(vsi);
5195 err = ice_vsi_cfg_lan_txqs(vsi);
5196 if (!err && ice_is_xdp_ena_vsi(vsi))
5197 err = ice_vsi_cfg_xdp_txqs(vsi);
5199 err = ice_vsi_cfg_rxqs(vsi);
5204 /* THEORY OF MODERATION:
5205 * The below code creates custom DIM profiles for use by this driver, because
5206 * the ice driver hardware works differently than the hardware that DIMLIB was
5207 * originally made for. ice hardware doesn't have packet count limits that
5208 * can trigger an interrupt, but it *does* have interrupt rate limit support,
5209 * and this code adds that capability to be used by the driver when it's using
5210 * DIMLIB. The DIMLIB code was always designed to be a suggestion to the driver
5211 * for how to "respond" to traffic and interrupts, so this driver uses a
5212 * slightly different set of moderation parameters to get best performance.
5215 /* the throttle rate for interrupts, basically worst case delay before
5216 * an initial interrupt fires, value is stored in microseconds.
5219 /* the rate limit for interrupts, which can cap a delay from a small
5220 * ITR at a certain amount of interrupts per second. f.e. a 2us ITR
5221 * could yield as much as 500,000 interrupts per second, but with a
5222 * 10us rate limit, it limits to 100,000 interrupts per second. Value
5223 * is stored in microseconds.
5228 /* Make a different profile for Rx that doesn't allow quite so aggressive
5229 * moderation at the high end (it maxes out at 128us or about 8k interrupts a
5230 * second. The INTRL/rate parameters here are only useful to cap small ITR
5231 * values, which is why for larger ITR's - like 128, which can only generate
5232 * 8k interrupts per second, there is no point to rate limit and the values
5233 * are set to zero. The rate limit values do affect latency, and so must
5234 * be reasonably small so to not impact latency sensitive tests.
5236 static const struct ice_dim rx_profile[] = {
5244 /* The transmit profile, which has the same sorts of values
5245 * as the previous struct
5247 static const struct ice_dim tx_profile[] = {
5255 static void ice_tx_dim_work(struct work_struct *work)
5257 struct ice_ring_container *rc;
5258 struct ice_q_vector *q_vector;
5262 dim = container_of(work, struct dim, work);
5263 rc = container_of(dim, struct ice_ring_container, dim);
5264 q_vector = container_of(rc, struct ice_q_vector, tx);
5266 if (dim->profile_ix >= ARRAY_SIZE(tx_profile))
5267 dim->profile_ix = ARRAY_SIZE(tx_profile) - 1;
5269 /* look up the values in our local table */
5270 itr = tx_profile[dim->profile_ix].itr;
5271 intrl = tx_profile[dim->profile_ix].intrl;
5273 ice_write_itr(rc, itr);
5274 ice_write_intrl(q_vector, intrl);
5276 dim->state = DIM_START_MEASURE;
5279 static void ice_rx_dim_work(struct work_struct *work)
5281 struct ice_ring_container *rc;
5282 struct ice_q_vector *q_vector;
5286 dim = container_of(work, struct dim, work);
5287 rc = container_of(dim, struct ice_ring_container, dim);
5288 q_vector = container_of(rc, struct ice_q_vector, rx);
5290 if (dim->profile_ix >= ARRAY_SIZE(rx_profile))
5291 dim->profile_ix = ARRAY_SIZE(rx_profile) - 1;
5293 /* look up the values in our local table */
5294 itr = rx_profile[dim->profile_ix].itr;
5295 intrl = rx_profile[dim->profile_ix].intrl;
5297 ice_write_itr(rc, itr);
5298 ice_write_intrl(q_vector, intrl);
5300 dim->state = DIM_START_MEASURE;
5304 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5305 * @vsi: the VSI being configured
5307 static void ice_napi_enable_all(struct ice_vsi *vsi)
5314 ice_for_each_q_vector(vsi, q_idx) {
5315 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5317 INIT_WORK(&q_vector->tx.dim.work, ice_tx_dim_work);
5318 q_vector->tx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5320 INIT_WORK(&q_vector->rx.dim.work, ice_rx_dim_work);
5321 q_vector->rx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5323 if (q_vector->rx.ring || q_vector->tx.ring)
5324 napi_enable(&q_vector->napi);
5329 * ice_up_complete - Finish the last steps of bringing up a connection
5330 * @vsi: The VSI being configured
5332 * Return 0 on success and negative value on error
5334 static int ice_up_complete(struct ice_vsi *vsi)
5336 struct ice_pf *pf = vsi->back;
5339 ice_vsi_cfg_msix(vsi);
5341 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5342 * Tx queue group list was configured and the context bits were
5343 * programmed using ice_vsi_cfg_txqs
5345 err = ice_vsi_start_all_rx_rings(vsi);
5349 clear_bit(ICE_VSI_DOWN, vsi->state);
5350 ice_napi_enable_all(vsi);
5351 ice_vsi_ena_irq(vsi);
5353 if (vsi->port_info &&
5354 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5356 ice_print_link_msg(vsi, true);
5357 netif_tx_start_all_queues(vsi->netdev);
5358 netif_carrier_on(vsi->netdev);
5361 ice_service_task_schedule(pf);
5367 * ice_up - Bring the connection back up after being down
5368 * @vsi: VSI being configured
5370 int ice_up(struct ice_vsi *vsi)
5374 err = ice_vsi_cfg(vsi);
5376 err = ice_up_complete(vsi);
5382 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5383 * @ring: Tx or Rx ring to read stats from
5384 * @pkts: packets stats counter
5385 * @bytes: bytes stats counter
5387 * This function fetches stats from the ring considering the atomic operations
5388 * that needs to be performed to read u64 values in 32 bit machine.
5391 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5400 start = u64_stats_fetch_begin_irq(&ring->syncp);
5401 *pkts = ring->stats.pkts;
5402 *bytes = ring->stats.bytes;
5403 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5407 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5408 * @vsi: the VSI to be updated
5409 * @rings: rings to work on
5410 * @count: number of rings
5413 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5416 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5419 for (i = 0; i < count; i++) {
5420 struct ice_ring *ring;
5423 ring = READ_ONCE(rings[i]);
5424 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5425 vsi_stats->tx_packets += pkts;
5426 vsi_stats->tx_bytes += bytes;
5427 vsi->tx_restart += ring->tx_stats.restart_q;
5428 vsi->tx_busy += ring->tx_stats.tx_busy;
5429 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5434 * ice_update_vsi_ring_stats - Update VSI stats counters
5435 * @vsi: the VSI to be updated
5437 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5439 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5440 struct ice_ring *ring;
5444 /* reset netdev stats */
5445 vsi_stats->tx_packets = 0;
5446 vsi_stats->tx_bytes = 0;
5447 vsi_stats->rx_packets = 0;
5448 vsi_stats->rx_bytes = 0;
5450 /* reset non-netdev (extended) stats */
5451 vsi->tx_restart = 0;
5453 vsi->tx_linearize = 0;
5454 vsi->rx_buf_failed = 0;
5455 vsi->rx_page_failed = 0;
5459 /* update Tx rings counters */
5460 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5462 /* update Rx rings counters */
5463 ice_for_each_rxq(vsi, i) {
5464 ring = READ_ONCE(vsi->rx_rings[i]);
5465 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5466 vsi_stats->rx_packets += pkts;
5467 vsi_stats->rx_bytes += bytes;
5468 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5469 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5472 /* update XDP Tx rings counters */
5473 if (ice_is_xdp_ena_vsi(vsi))
5474 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5481 * ice_update_vsi_stats - Update VSI stats counters
5482 * @vsi: the VSI to be updated
5484 void ice_update_vsi_stats(struct ice_vsi *vsi)
5486 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5487 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5488 struct ice_pf *pf = vsi->back;
5490 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
5491 test_bit(ICE_CFG_BUSY, pf->state))
5494 /* get stats as recorded by Tx/Rx rings */
5495 ice_update_vsi_ring_stats(vsi);
5497 /* get VSI stats as recorded by the hardware */
5498 ice_update_eth_stats(vsi);
5500 cur_ns->tx_errors = cur_es->tx_errors;
5501 cur_ns->rx_dropped = cur_es->rx_discards;
5502 cur_ns->tx_dropped = cur_es->tx_discards;
5503 cur_ns->multicast = cur_es->rx_multicast;
5505 /* update some more netdev stats if this is main VSI */
5506 if (vsi->type == ICE_VSI_PF) {
5507 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5508 cur_ns->rx_errors = pf->stats.crc_errors +
5509 pf->stats.illegal_bytes +
5510 pf->stats.rx_len_errors +
5511 pf->stats.rx_undersize +
5512 pf->hw_csum_rx_error +
5513 pf->stats.rx_jabber +
5514 pf->stats.rx_fragments +
5515 pf->stats.rx_oversize;
5516 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5517 /* record drops from the port level */
5518 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5523 * ice_update_pf_stats - Update PF port stats counters
5524 * @pf: PF whose stats needs to be updated
5526 void ice_update_pf_stats(struct ice_pf *pf)
5528 struct ice_hw_port_stats *prev_ps, *cur_ps;
5529 struct ice_hw *hw = &pf->hw;
5533 port = hw->port_info->lport;
5534 prev_ps = &pf->stats_prev;
5535 cur_ps = &pf->stats;
5537 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5538 &prev_ps->eth.rx_bytes,
5539 &cur_ps->eth.rx_bytes);
5541 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5542 &prev_ps->eth.rx_unicast,
5543 &cur_ps->eth.rx_unicast);
5545 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5546 &prev_ps->eth.rx_multicast,
5547 &cur_ps->eth.rx_multicast);
5549 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5550 &prev_ps->eth.rx_broadcast,
5551 &cur_ps->eth.rx_broadcast);
5553 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5554 &prev_ps->eth.rx_discards,
5555 &cur_ps->eth.rx_discards);
5557 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5558 &prev_ps->eth.tx_bytes,
5559 &cur_ps->eth.tx_bytes);
5561 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5562 &prev_ps->eth.tx_unicast,
5563 &cur_ps->eth.tx_unicast);
5565 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5566 &prev_ps->eth.tx_multicast,
5567 &cur_ps->eth.tx_multicast);
5569 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5570 &prev_ps->eth.tx_broadcast,
5571 &cur_ps->eth.tx_broadcast);
5573 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5574 &prev_ps->tx_dropped_link_down,
5575 &cur_ps->tx_dropped_link_down);
5577 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5578 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5580 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5581 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5583 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5584 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5586 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5587 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5589 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5590 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5592 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5593 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5595 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5596 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5598 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5599 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5601 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5602 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5604 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5605 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5607 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5608 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5610 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5611 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5613 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5614 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5616 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5617 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5619 fd_ctr_base = hw->fd_ctr_base;
5621 ice_stat_update40(hw,
5622 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5623 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5624 &cur_ps->fd_sb_match);
5625 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5626 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5628 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5629 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5631 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5632 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5634 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5635 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5637 ice_update_dcb_stats(pf);
5639 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5640 &prev_ps->crc_errors, &cur_ps->crc_errors);
5642 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5643 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5645 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5646 &prev_ps->mac_local_faults,
5647 &cur_ps->mac_local_faults);
5649 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5650 &prev_ps->mac_remote_faults,
5651 &cur_ps->mac_remote_faults);
5653 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5654 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5656 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5657 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5659 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5660 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5662 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5663 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5665 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5666 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5668 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5670 pf->stat_prev_loaded = true;
5674 * ice_get_stats64 - get statistics for network device structure
5675 * @netdev: network interface device structure
5676 * @stats: main device statistics structure
5679 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5681 struct ice_netdev_priv *np = netdev_priv(netdev);
5682 struct rtnl_link_stats64 *vsi_stats;
5683 struct ice_vsi *vsi = np->vsi;
5685 vsi_stats = &vsi->net_stats;
5687 if (!vsi->num_txq || !vsi->num_rxq)
5690 /* netdev packet/byte stats come from ring counter. These are obtained
5691 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5692 * But, only call the update routine and read the registers if VSI is
5695 if (!test_bit(ICE_VSI_DOWN, vsi->state))
5696 ice_update_vsi_ring_stats(vsi);
5697 stats->tx_packets = vsi_stats->tx_packets;
5698 stats->tx_bytes = vsi_stats->tx_bytes;
5699 stats->rx_packets = vsi_stats->rx_packets;
5700 stats->rx_bytes = vsi_stats->rx_bytes;
5702 /* The rest of the stats can be read from the hardware but instead we
5703 * just return values that the watchdog task has already obtained from
5706 stats->multicast = vsi_stats->multicast;
5707 stats->tx_errors = vsi_stats->tx_errors;
5708 stats->tx_dropped = vsi_stats->tx_dropped;
5709 stats->rx_errors = vsi_stats->rx_errors;
5710 stats->rx_dropped = vsi_stats->rx_dropped;
5711 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5712 stats->rx_length_errors = vsi_stats->rx_length_errors;
5716 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5717 * @vsi: VSI having NAPI disabled
5719 static void ice_napi_disable_all(struct ice_vsi *vsi)
5726 ice_for_each_q_vector(vsi, q_idx) {
5727 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5729 if (q_vector->rx.ring || q_vector->tx.ring)
5730 napi_disable(&q_vector->napi);
5732 cancel_work_sync(&q_vector->tx.dim.work);
5733 cancel_work_sync(&q_vector->rx.dim.work);
5738 * ice_down - Shutdown the connection
5739 * @vsi: The VSI being stopped
5741 int ice_down(struct ice_vsi *vsi)
5743 int i, tx_err, rx_err, link_err = 0;
5745 /* Caller of this function is expected to set the
5746 * vsi->state ICE_DOWN bit
5749 netif_carrier_off(vsi->netdev);
5750 netif_tx_disable(vsi->netdev);
5753 ice_vsi_dis_irq(vsi);
5755 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5757 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5758 vsi->vsi_num, tx_err);
5759 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5760 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
5762 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
5763 vsi->vsi_num, tx_err);
5766 rx_err = ice_vsi_stop_all_rx_rings(vsi);
5768 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
5769 vsi->vsi_num, rx_err);
5771 ice_napi_disable_all(vsi);
5773 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
5774 link_err = ice_force_phys_link_state(vsi, false);
5776 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
5777 vsi->vsi_num, link_err);
5780 ice_for_each_txq(vsi, i)
5781 ice_clean_tx_ring(vsi->tx_rings[i]);
5783 ice_for_each_rxq(vsi, i)
5784 ice_clean_rx_ring(vsi->rx_rings[i]);
5786 if (tx_err || rx_err || link_err) {
5787 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
5788 vsi->vsi_num, vsi->vsw->sw_id);
5796 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
5797 * @vsi: VSI having resources allocated
5799 * Return 0 on success, negative on failure
5801 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
5805 if (!vsi->num_txq) {
5806 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
5811 ice_for_each_txq(vsi, i) {
5812 struct ice_ring *ring = vsi->tx_rings[i];
5817 ring->netdev = vsi->netdev;
5818 err = ice_setup_tx_ring(ring);
5827 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
5828 * @vsi: VSI having resources allocated
5830 * Return 0 on success, negative on failure
5832 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
5836 if (!vsi->num_rxq) {
5837 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
5842 ice_for_each_rxq(vsi, i) {
5843 struct ice_ring *ring = vsi->rx_rings[i];
5848 ring->netdev = vsi->netdev;
5849 err = ice_setup_rx_ring(ring);
5858 * ice_vsi_open_ctrl - open control VSI for use
5859 * @vsi: the VSI to open
5861 * Initialization of the Control VSI
5863 * Returns 0 on success, negative value on error
5865 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
5867 char int_name[ICE_INT_NAME_STR_LEN];
5868 struct ice_pf *pf = vsi->back;
5872 dev = ice_pf_to_dev(pf);
5873 /* allocate descriptors */
5874 err = ice_vsi_setup_tx_rings(vsi);
5878 err = ice_vsi_setup_rx_rings(vsi);
5882 err = ice_vsi_cfg(vsi);
5886 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
5887 dev_driver_string(dev), dev_name(dev));
5888 err = ice_vsi_req_irq_msix(vsi, int_name);
5892 ice_vsi_cfg_msix(vsi);
5894 err = ice_vsi_start_all_rx_rings(vsi);
5896 goto err_up_complete;
5898 clear_bit(ICE_VSI_DOWN, vsi->state);
5899 ice_vsi_ena_irq(vsi);
5906 ice_vsi_free_rx_rings(vsi);
5908 ice_vsi_free_tx_rings(vsi);
5914 * ice_vsi_open - Called when a network interface is made active
5915 * @vsi: the VSI to open
5917 * Initialization of the VSI
5919 * Returns 0 on success, negative value on error
5921 static int ice_vsi_open(struct ice_vsi *vsi)
5923 char int_name[ICE_INT_NAME_STR_LEN];
5924 struct ice_pf *pf = vsi->back;
5927 /* allocate descriptors */
5928 err = ice_vsi_setup_tx_rings(vsi);
5932 err = ice_vsi_setup_rx_rings(vsi);
5936 err = ice_vsi_cfg(vsi);
5940 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5941 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
5942 err = ice_vsi_req_irq_msix(vsi, int_name);
5946 /* Notify the stack of the actual queue counts. */
5947 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
5951 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
5955 err = ice_up_complete(vsi);
5957 goto err_up_complete;
5964 ice_vsi_free_irq(vsi);
5966 ice_vsi_free_rx_rings(vsi);
5968 ice_vsi_free_tx_rings(vsi);
5974 * ice_vsi_release_all - Delete all VSIs
5975 * @pf: PF from which all VSIs are being removed
5977 static void ice_vsi_release_all(struct ice_pf *pf)
5984 ice_for_each_vsi(pf, i) {
5988 err = ice_vsi_release(pf->vsi[i]);
5990 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
5991 i, err, pf->vsi[i]->vsi_num);
5996 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
5997 * @pf: pointer to the PF instance
5998 * @type: VSI type to rebuild
6000 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6002 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6004 struct device *dev = ice_pf_to_dev(pf);
6005 enum ice_status status;
6008 ice_for_each_vsi(pf, i) {
6009 struct ice_vsi *vsi = pf->vsi[i];
6011 if (!vsi || vsi->type != type)
6014 /* rebuild the VSI */
6015 err = ice_vsi_rebuild(vsi, true);
6017 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6018 err, vsi->idx, ice_vsi_type_str(type));
6022 /* replay filters for the VSI */
6023 status = ice_replay_vsi(&pf->hw, vsi->idx);
6025 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
6026 ice_stat_str(status), vsi->idx,
6027 ice_vsi_type_str(type));
6031 /* Re-map HW VSI number, using VSI handle that has been
6032 * previously validated in ice_replay_vsi() call above
6034 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6036 /* enable the VSI */
6037 err = ice_ena_vsi(vsi, false);
6039 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6040 err, vsi->idx, ice_vsi_type_str(type));
6044 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6045 ice_vsi_type_str(type));
6052 * ice_update_pf_netdev_link - Update PF netdev link status
6053 * @pf: pointer to the PF instance
6055 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6060 ice_for_each_vsi(pf, i) {
6061 struct ice_vsi *vsi = pf->vsi[i];
6063 if (!vsi || vsi->type != ICE_VSI_PF)
6066 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6068 netif_carrier_on(pf->vsi[i]->netdev);
6069 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6071 netif_carrier_off(pf->vsi[i]->netdev);
6072 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6078 * ice_rebuild - rebuild after reset
6079 * @pf: PF to rebuild
6080 * @reset_type: type of reset
6082 * Do not rebuild VF VSI in this flow because that is already handled via
6083 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6084 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6085 * to reset/rebuild all the VF VSI twice.
6087 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6089 struct device *dev = ice_pf_to_dev(pf);
6090 struct ice_hw *hw = &pf->hw;
6091 enum ice_status ret;
6094 if (test_bit(ICE_DOWN, pf->state))
6095 goto clear_recovery;
6097 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6099 ret = ice_init_all_ctrlq(hw);
6101 dev_err(dev, "control queues init failed %s\n",
6103 goto err_init_ctrlq;
6106 /* if DDP was previously loaded successfully */
6107 if (!ice_is_safe_mode(pf)) {
6108 /* reload the SW DB of filter tables */
6109 if (reset_type == ICE_RESET_PFR)
6110 ice_fill_blk_tbls(hw);
6112 /* Reload DDP Package after CORER/GLOBR reset */
6113 ice_load_pkg(NULL, pf);
6116 ret = ice_clear_pf_cfg(hw);
6118 dev_err(dev, "clear PF configuration failed %s\n",
6120 goto err_init_ctrlq;
6123 if (pf->first_sw->dflt_vsi_ena)
6124 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6125 /* clear the default VSI configuration if it exists */
6126 pf->first_sw->dflt_vsi = NULL;
6127 pf->first_sw->dflt_vsi_ena = false;
6129 ice_clear_pxe_mode(hw);
6131 ret = ice_get_caps(hw);
6133 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
6134 goto err_init_ctrlq;
6137 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6139 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6140 goto err_init_ctrlq;
6143 err = ice_sched_init_port(hw->port_info);
6145 goto err_sched_init_port;
6147 /* start misc vector */
6148 err = ice_req_irq_msix_misc(pf);
6150 dev_err(dev, "misc vector setup failed: %d\n", err);
6151 goto err_sched_init_port;
6154 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6155 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6156 if (!rd32(hw, PFQF_FD_SIZE)) {
6157 u16 unused, guar, b_effort;
6159 guar = hw->func_caps.fd_fltr_guar;
6160 b_effort = hw->func_caps.fd_fltr_best_effort;
6162 /* force guaranteed filter pool for PF */
6163 ice_alloc_fd_guar_item(hw, &unused, guar);
6164 /* force shared filter pool for PF */
6165 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6169 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6170 ice_dcb_rebuild(pf);
6172 /* rebuild PF VSI */
6173 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6175 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6176 goto err_vsi_rebuild;
6179 /* If Flow Director is active */
6180 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6181 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6183 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6184 goto err_vsi_rebuild;
6187 /* replay HW Flow Director recipes */
6189 ice_fdir_replay_flows(hw);
6191 /* replay Flow Director filters */
6192 ice_fdir_replay_fltrs(pf);
6194 ice_rebuild_arfs(pf);
6197 ice_update_pf_netdev_link(pf);
6199 /* tell the firmware we are up */
6200 ret = ice_send_version(pf);
6202 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6204 goto err_vsi_rebuild;
6207 ice_replay_post(hw);
6209 /* if we get here, reset flow is successful */
6210 clear_bit(ICE_RESET_FAILED, pf->state);
6214 err_sched_init_port:
6215 ice_sched_cleanup_all(hw);
6217 ice_shutdown_all_ctrlq(hw);
6218 set_bit(ICE_RESET_FAILED, pf->state);
6220 /* set this bit in PF state to control service task scheduling */
6221 set_bit(ICE_NEEDS_RESTART, pf->state);
6222 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6226 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6227 * @vsi: Pointer to VSI structure
6229 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6231 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6232 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6234 return ICE_RXBUF_3072;
6238 * ice_change_mtu - NDO callback to change the MTU
6239 * @netdev: network interface device structure
6240 * @new_mtu: new value for maximum frame size
6242 * Returns 0 on success, negative on failure
6244 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6246 struct ice_netdev_priv *np = netdev_priv(netdev);
6247 struct ice_vsi *vsi = np->vsi;
6248 struct ice_pf *pf = vsi->back;
6251 if (new_mtu == (int)netdev->mtu) {
6252 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6256 if (ice_is_xdp_ena_vsi(vsi)) {
6257 int frame_size = ice_max_xdp_frame_size(vsi);
6259 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6260 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6261 frame_size - ICE_ETH_PKT_HDR_PAD);
6266 /* if a reset is in progress, wait for some time for it to complete */
6268 if (ice_is_reset_in_progress(pf->state)) {
6270 usleep_range(1000, 2000);
6275 } while (count < 100);
6278 netdev_err(netdev, "can't change MTU. Device is busy\n");
6282 netdev->mtu = (unsigned int)new_mtu;
6284 /* if VSI is up, bring it down and then back up */
6285 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6288 err = ice_down(vsi);
6290 netdev_err(netdev, "change MTU if_down err %d\n", err);
6296 netdev_err(netdev, "change MTU if_up err %d\n", err);
6301 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6306 * ice_aq_str - convert AQ err code to a string
6307 * @aq_err: the AQ error code to convert
6309 const char *ice_aq_str(enum ice_aq_err aq_err)
6314 case ICE_AQ_RC_EPERM:
6315 return "ICE_AQ_RC_EPERM";
6316 case ICE_AQ_RC_ENOENT:
6317 return "ICE_AQ_RC_ENOENT";
6318 case ICE_AQ_RC_ENOMEM:
6319 return "ICE_AQ_RC_ENOMEM";
6320 case ICE_AQ_RC_EBUSY:
6321 return "ICE_AQ_RC_EBUSY";
6322 case ICE_AQ_RC_EEXIST:
6323 return "ICE_AQ_RC_EEXIST";
6324 case ICE_AQ_RC_EINVAL:
6325 return "ICE_AQ_RC_EINVAL";
6326 case ICE_AQ_RC_ENOSPC:
6327 return "ICE_AQ_RC_ENOSPC";
6328 case ICE_AQ_RC_ENOSYS:
6329 return "ICE_AQ_RC_ENOSYS";
6330 case ICE_AQ_RC_EMODE:
6331 return "ICE_AQ_RC_EMODE";
6332 case ICE_AQ_RC_ENOSEC:
6333 return "ICE_AQ_RC_ENOSEC";
6334 case ICE_AQ_RC_EBADSIG:
6335 return "ICE_AQ_RC_EBADSIG";
6336 case ICE_AQ_RC_ESVN:
6337 return "ICE_AQ_RC_ESVN";
6338 case ICE_AQ_RC_EBADMAN:
6339 return "ICE_AQ_RC_EBADMAN";
6340 case ICE_AQ_RC_EBADBUF:
6341 return "ICE_AQ_RC_EBADBUF";
6344 return "ICE_AQ_RC_UNKNOWN";
6348 * ice_stat_str - convert status err code to a string
6349 * @stat_err: the status error code to convert
6351 const char *ice_stat_str(enum ice_status stat_err)
6357 return "ICE_ERR_PARAM";
6358 case ICE_ERR_NOT_IMPL:
6359 return "ICE_ERR_NOT_IMPL";
6360 case ICE_ERR_NOT_READY:
6361 return "ICE_ERR_NOT_READY";
6362 case ICE_ERR_NOT_SUPPORTED:
6363 return "ICE_ERR_NOT_SUPPORTED";
6364 case ICE_ERR_BAD_PTR:
6365 return "ICE_ERR_BAD_PTR";
6366 case ICE_ERR_INVAL_SIZE:
6367 return "ICE_ERR_INVAL_SIZE";
6368 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6369 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6370 case ICE_ERR_RESET_FAILED:
6371 return "ICE_ERR_RESET_FAILED";
6372 case ICE_ERR_FW_API_VER:
6373 return "ICE_ERR_FW_API_VER";
6374 case ICE_ERR_NO_MEMORY:
6375 return "ICE_ERR_NO_MEMORY";
6377 return "ICE_ERR_CFG";
6378 case ICE_ERR_OUT_OF_RANGE:
6379 return "ICE_ERR_OUT_OF_RANGE";
6380 case ICE_ERR_ALREADY_EXISTS:
6381 return "ICE_ERR_ALREADY_EXISTS";
6383 return "ICE_ERR_NVM";
6384 case ICE_ERR_NVM_CHECKSUM:
6385 return "ICE_ERR_NVM_CHECKSUM";
6386 case ICE_ERR_BUF_TOO_SHORT:
6387 return "ICE_ERR_BUF_TOO_SHORT";
6388 case ICE_ERR_NVM_BLANK_MODE:
6389 return "ICE_ERR_NVM_BLANK_MODE";
6390 case ICE_ERR_IN_USE:
6391 return "ICE_ERR_IN_USE";
6392 case ICE_ERR_MAX_LIMIT:
6393 return "ICE_ERR_MAX_LIMIT";
6394 case ICE_ERR_RESET_ONGOING:
6395 return "ICE_ERR_RESET_ONGOING";
6396 case ICE_ERR_HW_TABLE:
6397 return "ICE_ERR_HW_TABLE";
6398 case ICE_ERR_DOES_NOT_EXIST:
6399 return "ICE_ERR_DOES_NOT_EXIST";
6400 case ICE_ERR_FW_DDP_MISMATCH:
6401 return "ICE_ERR_FW_DDP_MISMATCH";
6402 case ICE_ERR_AQ_ERROR:
6403 return "ICE_ERR_AQ_ERROR";
6404 case ICE_ERR_AQ_TIMEOUT:
6405 return "ICE_ERR_AQ_TIMEOUT";
6406 case ICE_ERR_AQ_FULL:
6407 return "ICE_ERR_AQ_FULL";
6408 case ICE_ERR_AQ_NO_WORK:
6409 return "ICE_ERR_AQ_NO_WORK";
6410 case ICE_ERR_AQ_EMPTY:
6411 return "ICE_ERR_AQ_EMPTY";
6412 case ICE_ERR_AQ_FW_CRITICAL:
6413 return "ICE_ERR_AQ_FW_CRITICAL";
6416 return "ICE_ERR_UNKNOWN";
6420 * ice_set_rss_lut - Set RSS LUT
6421 * @vsi: Pointer to VSI structure
6422 * @lut: Lookup table
6423 * @lut_size: Lookup table size
6425 * Returns 0 on success, negative on failure
6427 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6429 struct ice_aq_get_set_rss_lut_params params = {};
6430 struct ice_hw *hw = &vsi->back->hw;
6431 enum ice_status status;
6436 params.vsi_handle = vsi->idx;
6437 params.lut_size = lut_size;
6438 params.lut_type = vsi->rss_lut_type;
6441 status = ice_aq_set_rss_lut(hw, ¶ms);
6443 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %s aq_err %s\n",
6444 ice_stat_str(status),
6445 ice_aq_str(hw->adminq.sq_last_status));
6453 * ice_set_rss_key - Set RSS key
6454 * @vsi: Pointer to the VSI structure
6455 * @seed: RSS hash seed
6457 * Returns 0 on success, negative on failure
6459 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
6461 struct ice_hw *hw = &vsi->back->hw;
6462 enum ice_status status;
6467 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6469 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %s aq_err %s\n",
6470 ice_stat_str(status),
6471 ice_aq_str(hw->adminq.sq_last_status));
6479 * ice_get_rss_lut - Get RSS LUT
6480 * @vsi: Pointer to VSI structure
6481 * @lut: Buffer to store the lookup table entries
6482 * @lut_size: Size of buffer to store the lookup table entries
6484 * Returns 0 on success, negative on failure
6486 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6488 struct ice_aq_get_set_rss_lut_params params = {};
6489 struct ice_hw *hw = &vsi->back->hw;
6490 enum ice_status status;
6495 params.vsi_handle = vsi->idx;
6496 params.lut_size = lut_size;
6497 params.lut_type = vsi->rss_lut_type;
6500 status = ice_aq_get_rss_lut(hw, ¶ms);
6502 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %s aq_err %s\n",
6503 ice_stat_str(status),
6504 ice_aq_str(hw->adminq.sq_last_status));
6512 * ice_get_rss_key - Get RSS key
6513 * @vsi: Pointer to VSI structure
6514 * @seed: Buffer to store the key in
6516 * Returns 0 on success, negative on failure
6518 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
6520 struct ice_hw *hw = &vsi->back->hw;
6521 enum ice_status status;
6526 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6528 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %s aq_err %s\n",
6529 ice_stat_str(status),
6530 ice_aq_str(hw->adminq.sq_last_status));
6538 * ice_bridge_getlink - Get the hardware bridge mode
6541 * @seq: RTNL message seq
6542 * @dev: the netdev being configured
6543 * @filter_mask: filter mask passed in
6544 * @nlflags: netlink flags passed in
6546 * Return the bridge mode (VEB/VEPA)
6549 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6550 struct net_device *dev, u32 filter_mask, int nlflags)
6552 struct ice_netdev_priv *np = netdev_priv(dev);
6553 struct ice_vsi *vsi = np->vsi;
6554 struct ice_pf *pf = vsi->back;
6557 bmode = pf->first_sw->bridge_mode;
6559 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6564 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6565 * @vsi: Pointer to VSI structure
6566 * @bmode: Hardware bridge mode (VEB/VEPA)
6568 * Returns 0 on success, negative on failure
6570 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6572 struct ice_aqc_vsi_props *vsi_props;
6573 struct ice_hw *hw = &vsi->back->hw;
6574 struct ice_vsi_ctx *ctxt;
6575 enum ice_status status;
6578 vsi_props = &vsi->info;
6580 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6584 ctxt->info = vsi->info;
6586 if (bmode == BRIDGE_MODE_VEB)
6587 /* change from VEPA to VEB mode */
6588 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6590 /* change from VEB to VEPA mode */
6591 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6592 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6594 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6596 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6597 bmode, ice_stat_str(status),
6598 ice_aq_str(hw->adminq.sq_last_status));
6602 /* Update sw flags for book keeping */
6603 vsi_props->sw_flags = ctxt->info.sw_flags;
6611 * ice_bridge_setlink - Set the hardware bridge mode
6612 * @dev: the netdev being configured
6613 * @nlh: RTNL message
6614 * @flags: bridge setlink flags
6615 * @extack: netlink extended ack
6617 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6618 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6619 * not already set for all VSIs connected to this switch. And also update the
6620 * unicast switch filter rules for the corresponding switch of the netdev.
6623 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6624 u16 __always_unused flags,
6625 struct netlink_ext_ack __always_unused *extack)
6627 struct ice_netdev_priv *np = netdev_priv(dev);
6628 struct ice_pf *pf = np->vsi->back;
6629 struct nlattr *attr, *br_spec;
6630 struct ice_hw *hw = &pf->hw;
6631 enum ice_status status;
6632 struct ice_sw *pf_sw;
6633 int rem, v, err = 0;
6635 pf_sw = pf->first_sw;
6636 /* find the attribute in the netlink message */
6637 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6639 nla_for_each_nested(attr, br_spec, rem) {
6642 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6644 mode = nla_get_u16(attr);
6645 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6647 /* Continue if bridge mode is not being flipped */
6648 if (mode == pf_sw->bridge_mode)
6650 /* Iterates through the PF VSI list and update the loopback
6653 ice_for_each_vsi(pf, v) {
6656 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6661 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6662 /* Update the unicast switch filter rules for the corresponding
6663 * switch of the netdev
6665 status = ice_update_sw_rule_bridge_mode(hw);
6667 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6668 mode, ice_stat_str(status),
6669 ice_aq_str(hw->adminq.sq_last_status));
6670 /* revert hw->evb_veb */
6671 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6675 pf_sw->bridge_mode = mode;
6682 * ice_tx_timeout - Respond to a Tx Hang
6683 * @netdev: network interface device structure
6684 * @txqueue: Tx queue
6686 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6688 struct ice_netdev_priv *np = netdev_priv(netdev);
6689 struct ice_ring *tx_ring = NULL;
6690 struct ice_vsi *vsi = np->vsi;
6691 struct ice_pf *pf = vsi->back;
6694 pf->tx_timeout_count++;
6696 /* Check if PFC is enabled for the TC to which the queue belongs
6697 * to. If yes then Tx timeout is not caused by a hung queue, no
6698 * need to reset and rebuild
6700 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6701 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6706 /* now that we have an index, find the tx_ring struct */
6707 for (i = 0; i < vsi->num_txq; i++)
6708 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6709 if (txqueue == vsi->tx_rings[i]->q_index) {
6710 tx_ring = vsi->tx_rings[i];
6714 /* Reset recovery level if enough time has elapsed after last timeout.
6715 * Also ensure no new reset action happens before next timeout period.
6717 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
6718 pf->tx_timeout_recovery_level = 1;
6719 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
6720 netdev->watchdog_timeo)))
6724 struct ice_hw *hw = &pf->hw;
6727 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
6728 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
6729 /* Read interrupt register */
6730 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
6732 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
6733 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
6734 head, tx_ring->next_to_use, val);
6737 pf->tx_timeout_last_recovery = jiffies;
6738 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
6739 pf->tx_timeout_recovery_level, txqueue);
6741 switch (pf->tx_timeout_recovery_level) {
6743 set_bit(ICE_PFR_REQ, pf->state);
6746 set_bit(ICE_CORER_REQ, pf->state);
6749 set_bit(ICE_GLOBR_REQ, pf->state);
6752 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
6753 set_bit(ICE_DOWN, pf->state);
6754 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
6755 set_bit(ICE_SERVICE_DIS, pf->state);
6759 ice_service_task_schedule(pf);
6760 pf->tx_timeout_recovery_level++;
6764 * ice_open - Called when a network interface becomes active
6765 * @netdev: network interface device structure
6767 * The open entry point is called when a network interface is made
6768 * active by the system (IFF_UP). At this point all resources needed
6769 * for transmit and receive operations are allocated, the interrupt
6770 * handler is registered with the OS, the netdev watchdog is enabled,
6771 * and the stack is notified that the interface is ready.
6773 * Returns 0 on success, negative value on failure
6775 int ice_open(struct net_device *netdev)
6777 struct ice_netdev_priv *np = netdev_priv(netdev);
6778 struct ice_pf *pf = np->vsi->back;
6780 if (ice_is_reset_in_progress(pf->state)) {
6781 netdev_err(netdev, "can't open net device while reset is in progress");
6785 return ice_open_internal(netdev);
6789 * ice_open_internal - Called when a network interface becomes active
6790 * @netdev: network interface device structure
6792 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
6795 * Returns 0 on success, negative value on failure
6797 int ice_open_internal(struct net_device *netdev)
6799 struct ice_netdev_priv *np = netdev_priv(netdev);
6800 struct ice_vsi *vsi = np->vsi;
6801 struct ice_pf *pf = vsi->back;
6802 struct ice_port_info *pi;
6803 enum ice_status status;
6806 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
6807 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
6811 netif_carrier_off(netdev);
6813 pi = vsi->port_info;
6814 status = ice_update_link_info(pi);
6816 netdev_err(netdev, "Failed to get link info, error %s\n",
6817 ice_stat_str(status));
6821 /* Set PHY if there is media, otherwise, turn off PHY */
6822 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
6823 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6824 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
6825 err = ice_init_phy_user_cfg(pi);
6827 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
6833 err = ice_configure_phy(vsi);
6835 netdev_err(netdev, "Failed to set physical link up, error %d\n",
6840 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
6841 ice_set_link(vsi, false);
6844 err = ice_vsi_open(vsi);
6846 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
6847 vsi->vsi_num, vsi->vsw->sw_id);
6849 /* Update existing tunnels information */
6850 udp_tunnel_get_rx_info(netdev);
6856 * ice_stop - Disables a network interface
6857 * @netdev: network interface device structure
6859 * The stop entry point is called when an interface is de-activated by the OS,
6860 * and the netdevice enters the DOWN state. The hardware is still under the
6861 * driver's control, but the netdev interface is disabled.
6863 * Returns success only - not allowed to fail
6865 int ice_stop(struct net_device *netdev)
6867 struct ice_netdev_priv *np = netdev_priv(netdev);
6868 struct ice_vsi *vsi = np->vsi;
6869 struct ice_pf *pf = vsi->back;
6871 if (ice_is_reset_in_progress(pf->state)) {
6872 netdev_err(netdev, "can't stop net device while reset is in progress");
6882 * ice_features_check - Validate encapsulated packet conforms to limits
6884 * @netdev: This port's netdev
6885 * @features: Offload features that the stack believes apply
6887 static netdev_features_t
6888 ice_features_check(struct sk_buff *skb,
6889 struct net_device __always_unused *netdev,
6890 netdev_features_t features)
6894 /* No point in doing any of this if neither checksum nor GSO are
6895 * being requested for this frame. We can rule out both by just
6896 * checking for CHECKSUM_PARTIAL
6898 if (skb->ip_summed != CHECKSUM_PARTIAL)
6901 /* We cannot support GSO if the MSS is going to be less than
6902 * 64 bytes. If it is then we need to drop support for GSO.
6904 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
6905 features &= ~NETIF_F_GSO_MASK;
6907 len = skb_network_header(skb) - skb->data;
6908 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
6909 goto out_rm_features;
6911 len = skb_transport_header(skb) - skb_network_header(skb);
6912 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6913 goto out_rm_features;
6915 if (skb->encapsulation) {
6916 len = skb_inner_network_header(skb) - skb_transport_header(skb);
6917 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
6918 goto out_rm_features;
6920 len = skb_inner_transport_header(skb) -
6921 skb_inner_network_header(skb);
6922 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
6923 goto out_rm_features;
6928 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
6931 static const struct net_device_ops ice_netdev_safe_mode_ops = {
6932 .ndo_open = ice_open,
6933 .ndo_stop = ice_stop,
6934 .ndo_start_xmit = ice_start_xmit,
6935 .ndo_set_mac_address = ice_set_mac_address,
6936 .ndo_validate_addr = eth_validate_addr,
6937 .ndo_change_mtu = ice_change_mtu,
6938 .ndo_get_stats64 = ice_get_stats64,
6939 .ndo_tx_timeout = ice_tx_timeout,
6942 static const struct net_device_ops ice_netdev_ops = {
6943 .ndo_open = ice_open,
6944 .ndo_stop = ice_stop,
6945 .ndo_start_xmit = ice_start_xmit,
6946 .ndo_features_check = ice_features_check,
6947 .ndo_set_rx_mode = ice_set_rx_mode,
6948 .ndo_set_mac_address = ice_set_mac_address,
6949 .ndo_validate_addr = eth_validate_addr,
6950 .ndo_change_mtu = ice_change_mtu,
6951 .ndo_get_stats64 = ice_get_stats64,
6952 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
6953 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
6954 .ndo_set_vf_mac = ice_set_vf_mac,
6955 .ndo_get_vf_config = ice_get_vf_cfg,
6956 .ndo_set_vf_trust = ice_set_vf_trust,
6957 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
6958 .ndo_set_vf_link_state = ice_set_vf_link_state,
6959 .ndo_get_vf_stats = ice_get_vf_stats,
6960 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
6961 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
6962 .ndo_set_features = ice_set_features,
6963 .ndo_bridge_getlink = ice_bridge_getlink,
6964 .ndo_bridge_setlink = ice_bridge_setlink,
6965 .ndo_fdb_add = ice_fdb_add,
6966 .ndo_fdb_del = ice_fdb_del,
6967 #ifdef CONFIG_RFS_ACCEL
6968 .ndo_rx_flow_steer = ice_rx_flow_steer,
6970 .ndo_tx_timeout = ice_tx_timeout,
6972 .ndo_xdp_xmit = ice_xdp_xmit,
6973 .ndo_xsk_wakeup = ice_xsk_wakeup,