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"
16 /* Including ice_trace.h with CREATE_TRACE_POINTS defined will generate the
17 * ice tracepoint functions. This must be done exactly once across the
20 #define CREATE_TRACE_POINTS
21 #include "ice_trace.h"
23 #define DRV_SUMMARY "Intel(R) Ethernet Connection E800 Series Linux Driver"
24 static const char ice_driver_string[] = DRV_SUMMARY;
25 static const char ice_copyright[] = "Copyright (c) 2018, Intel Corporation.";
27 /* DDP Package file located in firmware search paths (e.g. /lib/firmware/) */
28 #define ICE_DDP_PKG_PATH "intel/ice/ddp/"
29 #define ICE_DDP_PKG_FILE ICE_DDP_PKG_PATH "ice.pkg"
31 MODULE_AUTHOR("Intel Corporation, <linux.nics@intel.com>");
32 MODULE_DESCRIPTION(DRV_SUMMARY);
33 MODULE_LICENSE("GPL v2");
34 MODULE_FIRMWARE(ICE_DDP_PKG_FILE);
36 static int debug = -1;
37 module_param(debug, int, 0644);
38 #ifndef CONFIG_DYNAMIC_DEBUG
39 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all), hw debug_mask (0x8XXXXXXX)");
41 MODULE_PARM_DESC(debug, "netif level (0=none,...,16=all)");
42 #endif /* !CONFIG_DYNAMIC_DEBUG */
44 static DEFINE_IDA(ice_aux_ida);
46 static struct workqueue_struct *ice_wq;
47 static const struct net_device_ops ice_netdev_safe_mode_ops;
48 static const struct net_device_ops ice_netdev_ops;
49 static int ice_vsi_open(struct ice_vsi *vsi);
51 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type);
53 static void ice_vsi_release_all(struct ice_pf *pf);
55 bool netif_is_ice(struct net_device *dev)
57 return dev && (dev->netdev_ops == &ice_netdev_ops);
61 * ice_get_tx_pending - returns number of Tx descriptors not processed
62 * @ring: the ring of descriptors
64 static u16 ice_get_tx_pending(struct ice_ring *ring)
68 head = ring->next_to_clean;
69 tail = ring->next_to_use;
72 return (head < tail) ?
73 tail - head : (tail + ring->count - head);
78 * ice_check_for_hang_subtask - check for and recover hung queues
79 * @pf: pointer to PF struct
81 static void ice_check_for_hang_subtask(struct ice_pf *pf)
83 struct ice_vsi *vsi = NULL;
89 ice_for_each_vsi(pf, v)
90 if (pf->vsi[v] && pf->vsi[v]->type == ICE_VSI_PF) {
95 if (!vsi || test_bit(ICE_VSI_DOWN, vsi->state))
98 if (!(vsi->netdev && netif_carrier_ok(vsi->netdev)))
103 for (i = 0; i < vsi->num_txq; i++) {
104 struct ice_ring *tx_ring = vsi->tx_rings[i];
106 if (tx_ring && tx_ring->desc) {
107 /* If packet counter has not changed the queue is
108 * likely stalled, so force an interrupt for this
111 * prev_pkt would be negative if there was no
114 packets = tx_ring->stats.pkts & INT_MAX;
115 if (tx_ring->tx_stats.prev_pkt == packets) {
116 /* Trigger sw interrupt to revive the queue */
117 ice_trigger_sw_intr(hw, tx_ring->q_vector);
121 /* Memory barrier between read of packet count and call
122 * to ice_get_tx_pending()
125 tx_ring->tx_stats.prev_pkt =
126 ice_get_tx_pending(tx_ring) ? packets : -1;
132 * ice_init_mac_fltr - Set initial MAC filters
133 * @pf: board private structure
135 * Set initial set of MAC filters for PF VSI; configure filters for permanent
136 * address and broadcast address. If an error is encountered, netdevice will be
139 static int ice_init_mac_fltr(struct ice_pf *pf)
141 enum ice_status status;
145 vsi = ice_get_main_vsi(pf);
149 perm_addr = vsi->port_info->mac.perm_addr;
150 status = ice_fltr_add_mac_and_broadcast(vsi, perm_addr, ICE_FWD_TO_VSI);
158 * ice_add_mac_to_sync_list - creates list of MAC addresses to be synced
159 * @netdev: the net device on which the sync is happening
160 * @addr: MAC address to sync
162 * This is a callback function which is called by the in kernel device sync
163 * functions (like __dev_uc_sync, __dev_mc_sync, etc). This function only
164 * populates the tmp_sync_list, which is later used by ice_add_mac to add the
165 * MAC filters from the hardware.
167 static int ice_add_mac_to_sync_list(struct net_device *netdev, const u8 *addr)
169 struct ice_netdev_priv *np = netdev_priv(netdev);
170 struct ice_vsi *vsi = np->vsi;
172 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_sync_list, addr,
180 * ice_add_mac_to_unsync_list - creates list of MAC addresses to be unsynced
181 * @netdev: the net device on which the unsync is happening
182 * @addr: MAC address to unsync
184 * This is a callback function which is called by the in kernel device unsync
185 * functions (like __dev_uc_unsync, __dev_mc_unsync, etc). This function only
186 * populates the tmp_unsync_list, which is later used by ice_remove_mac to
187 * delete the MAC filters from the hardware.
189 static int ice_add_mac_to_unsync_list(struct net_device *netdev, const u8 *addr)
191 struct ice_netdev_priv *np = netdev_priv(netdev);
192 struct ice_vsi *vsi = np->vsi;
194 /* Under some circumstances, we might receive a request to delete our
195 * own device address from our uc list. Because we store the device
196 * address in the VSI's MAC filter list, we need to ignore such
197 * requests and not delete our device address from this list.
199 if (ether_addr_equal(addr, netdev->dev_addr))
202 if (ice_fltr_add_mac_to_list(vsi, &vsi->tmp_unsync_list, addr,
210 * ice_vsi_fltr_changed - check if filter state changed
211 * @vsi: VSI to be checked
213 * returns true if filter state has changed, false otherwise.
215 static bool ice_vsi_fltr_changed(struct ice_vsi *vsi)
217 return test_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state) ||
218 test_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state) ||
219 test_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
223 * ice_cfg_promisc - Enable or disable promiscuous mode for a given PF
224 * @vsi: the VSI being configured
225 * @promisc_m: mask of promiscuous config bits
226 * @set_promisc: enable or disable promisc flag request
229 static int ice_cfg_promisc(struct ice_vsi *vsi, u8 promisc_m, bool set_promisc)
231 struct ice_hw *hw = &vsi->back->hw;
232 enum ice_status status = 0;
234 if (vsi->type != ICE_VSI_PF)
237 if (vsi->num_vlan > 1) {
238 status = ice_set_vlan_vsi_promisc(hw, vsi->idx, promisc_m,
242 status = ice_set_vsi_promisc(hw, vsi->idx, promisc_m,
245 status = ice_clear_vsi_promisc(hw, vsi->idx, promisc_m,
256 * ice_vsi_sync_fltr - Update the VSI filter list to the HW
257 * @vsi: ptr to the VSI
259 * Push any outstanding VSI filter changes through the AdminQ.
261 static int ice_vsi_sync_fltr(struct ice_vsi *vsi)
263 struct device *dev = ice_pf_to_dev(vsi->back);
264 struct net_device *netdev = vsi->netdev;
265 bool promisc_forced_on = false;
266 struct ice_pf *pf = vsi->back;
267 struct ice_hw *hw = &pf->hw;
268 enum ice_status status = 0;
269 u32 changed_flags = 0;
276 while (test_and_set_bit(ICE_CFG_BUSY, vsi->state))
277 usleep_range(1000, 2000);
279 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
280 vsi->current_netdev_flags = vsi->netdev->flags;
282 INIT_LIST_HEAD(&vsi->tmp_sync_list);
283 INIT_LIST_HEAD(&vsi->tmp_unsync_list);
285 if (ice_vsi_fltr_changed(vsi)) {
286 clear_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
287 clear_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
288 clear_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
290 /* grab the netdev's addr_list_lock */
291 netif_addr_lock_bh(netdev);
292 __dev_uc_sync(netdev, ice_add_mac_to_sync_list,
293 ice_add_mac_to_unsync_list);
294 __dev_mc_sync(netdev, ice_add_mac_to_sync_list,
295 ice_add_mac_to_unsync_list);
296 /* our temp lists are populated. release lock */
297 netif_addr_unlock_bh(netdev);
300 /* Remove MAC addresses in the unsync list */
301 status = ice_fltr_remove_mac_list(vsi, &vsi->tmp_unsync_list);
302 ice_fltr_free_list(dev, &vsi->tmp_unsync_list);
304 netdev_err(netdev, "Failed to delete MAC filters\n");
305 /* if we failed because of alloc failures, just bail */
306 if (status == ICE_ERR_NO_MEMORY) {
312 /* Add MAC addresses in the sync list */
313 status = ice_fltr_add_mac_list(vsi, &vsi->tmp_sync_list);
314 ice_fltr_free_list(dev, &vsi->tmp_sync_list);
315 /* If filter is added successfully or already exists, do not go into
316 * 'if' condition and report it as error. Instead continue processing
317 * rest of the function.
319 if (status && status != ICE_ERR_ALREADY_EXISTS) {
320 netdev_err(netdev, "Failed to add MAC filters\n");
321 /* If there is no more space for new umac filters, VSI
322 * should go into promiscuous mode. There should be some
323 * space reserved for promiscuous filters.
325 if (hw->adminq.sq_last_status == ICE_AQ_RC_ENOSPC &&
326 !test_and_set_bit(ICE_FLTR_OVERFLOW_PROMISC,
328 promisc_forced_on = true;
329 netdev_warn(netdev, "Reached MAC filter limit, forcing promisc mode on VSI %d\n",
336 /* check for changes in promiscuous modes */
337 if (changed_flags & IFF_ALLMULTI) {
338 if (vsi->current_netdev_flags & IFF_ALLMULTI) {
339 if (vsi->num_vlan > 1)
340 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
342 promisc_m = ICE_MCAST_PROMISC_BITS;
344 err = ice_cfg_promisc(vsi, promisc_m, true);
346 netdev_err(netdev, "Error setting Multicast promiscuous mode on VSI %i\n",
348 vsi->current_netdev_flags &= ~IFF_ALLMULTI;
352 /* !(vsi->current_netdev_flags & IFF_ALLMULTI) */
353 if (vsi->num_vlan > 1)
354 promisc_m = ICE_MCAST_VLAN_PROMISC_BITS;
356 promisc_m = ICE_MCAST_PROMISC_BITS;
358 err = ice_cfg_promisc(vsi, promisc_m, false);
360 netdev_err(netdev, "Error clearing Multicast promiscuous mode on VSI %i\n",
362 vsi->current_netdev_flags |= IFF_ALLMULTI;
368 if (((changed_flags & IFF_PROMISC) || promisc_forced_on) ||
369 test_bit(ICE_VSI_PROMISC_CHANGED, vsi->state)) {
370 clear_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
371 if (vsi->current_netdev_flags & IFF_PROMISC) {
372 /* Apply Rx filter rule to get traffic from wire */
373 if (!ice_is_dflt_vsi_in_use(pf->first_sw)) {
374 err = ice_set_dflt_vsi(pf->first_sw, vsi);
375 if (err && err != -EEXIST) {
376 netdev_err(netdev, "Error %d setting default VSI %i Rx rule\n",
378 vsi->current_netdev_flags &=
382 ice_cfg_vlan_pruning(vsi, false, false);
385 /* Clear Rx filter to remove traffic from wire */
386 if (ice_is_vsi_dflt_vsi(pf->first_sw, vsi)) {
387 err = ice_clear_dflt_vsi(pf->first_sw);
389 netdev_err(netdev, "Error %d clearing default VSI %i Rx rule\n",
391 vsi->current_netdev_flags |=
395 if (vsi->num_vlan > 1)
396 ice_cfg_vlan_pruning(vsi, true, false);
403 set_bit(ICE_VSI_PROMISC_CHANGED, vsi->state);
406 /* if something went wrong then set the changed flag so we try again */
407 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
408 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
410 clear_bit(ICE_CFG_BUSY, vsi->state);
415 * ice_sync_fltr_subtask - Sync the VSI filter list with HW
416 * @pf: board private structure
418 static void ice_sync_fltr_subtask(struct ice_pf *pf)
422 if (!pf || !(test_bit(ICE_FLAG_FLTR_SYNC, pf->flags)))
425 clear_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
427 ice_for_each_vsi(pf, v)
428 if (pf->vsi[v] && ice_vsi_fltr_changed(pf->vsi[v]) &&
429 ice_vsi_sync_fltr(pf->vsi[v])) {
430 /* come back and try again later */
431 set_bit(ICE_FLAG_FLTR_SYNC, pf->flags);
437 * ice_pf_dis_all_vsi - Pause all VSIs on a PF
439 * @locked: is the rtnl_lock already held
441 static void ice_pf_dis_all_vsi(struct ice_pf *pf, bool locked)
446 ice_for_each_vsi(pf, v)
448 ice_dis_vsi(pf->vsi[v], locked);
450 for (node = 0; node < ICE_MAX_PF_AGG_NODES; node++)
451 pf->pf_agg_node[node].num_vsis = 0;
453 for (node = 0; node < ICE_MAX_VF_AGG_NODES; node++)
454 pf->vf_agg_node[node].num_vsis = 0;
458 * ice_prepare_for_reset - prep for the core to reset
459 * @pf: board private structure
461 * Inform or close all dependent features in prep for reset.
464 ice_prepare_for_reset(struct ice_pf *pf)
466 struct ice_hw *hw = &pf->hw;
469 /* already prepared for reset */
470 if (test_bit(ICE_PREPARED_FOR_RESET, pf->state))
473 ice_unplug_aux_dev(pf);
475 /* Notify VFs of impending reset */
476 if (ice_check_sq_alive(hw, &hw->mailboxq))
477 ice_vc_notify_reset(pf);
479 /* Disable VFs until reset is completed */
480 ice_for_each_vf(pf, i)
481 ice_set_vf_state_qs_dis(&pf->vf[i]);
483 /* clear SW filtering DB */
484 ice_clear_hw_tbls(hw);
485 /* disable the VSIs and their queues that are not already DOWN */
486 ice_pf_dis_all_vsi(pf, false);
488 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
492 ice_sched_clear_port(hw->port_info);
494 ice_shutdown_all_ctrlq(hw);
496 set_bit(ICE_PREPARED_FOR_RESET, pf->state);
500 * ice_do_reset - Initiate one of many types of resets
501 * @pf: board private structure
502 * @reset_type: reset type requested
503 * before this function was called.
505 static void ice_do_reset(struct ice_pf *pf, enum ice_reset_req reset_type)
507 struct device *dev = ice_pf_to_dev(pf);
508 struct ice_hw *hw = &pf->hw;
510 dev_dbg(dev, "reset_type 0x%x requested\n", reset_type);
512 ice_prepare_for_reset(pf);
514 /* trigger the reset */
515 if (ice_reset(hw, reset_type)) {
516 dev_err(dev, "reset %d failed\n", reset_type);
517 set_bit(ICE_RESET_FAILED, pf->state);
518 clear_bit(ICE_RESET_OICR_RECV, pf->state);
519 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
520 clear_bit(ICE_PFR_REQ, pf->state);
521 clear_bit(ICE_CORER_REQ, pf->state);
522 clear_bit(ICE_GLOBR_REQ, pf->state);
523 wake_up(&pf->reset_wait_queue);
527 /* PFR is a bit of a special case because it doesn't result in an OICR
528 * interrupt. So for PFR, rebuild after the reset and clear the reset-
529 * associated state bits.
531 if (reset_type == ICE_RESET_PFR) {
533 ice_rebuild(pf, reset_type);
534 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
535 clear_bit(ICE_PFR_REQ, pf->state);
536 wake_up(&pf->reset_wait_queue);
537 ice_reset_all_vfs(pf, true);
542 * ice_reset_subtask - Set up for resetting the device and driver
543 * @pf: board private structure
545 static void ice_reset_subtask(struct ice_pf *pf)
547 enum ice_reset_req reset_type = ICE_RESET_INVAL;
549 /* When a CORER/GLOBR/EMPR is about to happen, the hardware triggers an
550 * OICR interrupt. The OICR handler (ice_misc_intr) determines what type
551 * of reset is pending and sets bits in pf->state indicating the reset
552 * type and ICE_RESET_OICR_RECV. So, if the latter bit is set
553 * prepare for pending reset if not already (for PF software-initiated
554 * global resets the software should already be prepared for it as
555 * indicated by ICE_PREPARED_FOR_RESET; for global resets initiated
556 * by firmware or software on other PFs, that bit is not set so prepare
557 * for the reset now), poll for reset done, rebuild and return.
559 if (test_bit(ICE_RESET_OICR_RECV, pf->state)) {
560 /* Perform the largest reset requested */
561 if (test_and_clear_bit(ICE_CORER_RECV, pf->state))
562 reset_type = ICE_RESET_CORER;
563 if (test_and_clear_bit(ICE_GLOBR_RECV, pf->state))
564 reset_type = ICE_RESET_GLOBR;
565 if (test_and_clear_bit(ICE_EMPR_RECV, pf->state))
566 reset_type = ICE_RESET_EMPR;
567 /* return if no valid reset type requested */
568 if (reset_type == ICE_RESET_INVAL)
570 ice_prepare_for_reset(pf);
572 /* make sure we are ready to rebuild */
573 if (ice_check_reset(&pf->hw)) {
574 set_bit(ICE_RESET_FAILED, pf->state);
576 /* done with reset. start rebuild */
577 pf->hw.reset_ongoing = false;
578 ice_rebuild(pf, reset_type);
579 /* clear bit to resume normal operations, but
580 * ICE_NEEDS_RESTART bit is set in case rebuild failed
582 clear_bit(ICE_RESET_OICR_RECV, pf->state);
583 clear_bit(ICE_PREPARED_FOR_RESET, pf->state);
584 clear_bit(ICE_PFR_REQ, pf->state);
585 clear_bit(ICE_CORER_REQ, pf->state);
586 clear_bit(ICE_GLOBR_REQ, pf->state);
587 wake_up(&pf->reset_wait_queue);
588 ice_reset_all_vfs(pf, true);
594 /* No pending resets to finish processing. Check for new resets */
595 if (test_bit(ICE_PFR_REQ, pf->state))
596 reset_type = ICE_RESET_PFR;
597 if (test_bit(ICE_CORER_REQ, pf->state))
598 reset_type = ICE_RESET_CORER;
599 if (test_bit(ICE_GLOBR_REQ, pf->state))
600 reset_type = ICE_RESET_GLOBR;
601 /* If no valid reset type requested just return */
602 if (reset_type == ICE_RESET_INVAL)
605 /* reset if not already down or busy */
606 if (!test_bit(ICE_DOWN, pf->state) &&
607 !test_bit(ICE_CFG_BUSY, pf->state)) {
608 ice_do_reset(pf, reset_type);
613 * ice_print_topo_conflict - print topology conflict message
614 * @vsi: the VSI whose topology status is being checked
616 static void ice_print_topo_conflict(struct ice_vsi *vsi)
618 switch (vsi->port_info->phy.link_info.topo_media_conflict) {
619 case ICE_AQ_LINK_TOPO_CONFLICT:
620 case ICE_AQ_LINK_MEDIA_CONFLICT:
621 case ICE_AQ_LINK_TOPO_UNREACH_PRT:
622 case ICE_AQ_LINK_TOPO_UNDRUTIL_PRT:
623 case ICE_AQ_LINK_TOPO_UNDRUTIL_MEDIA:
624 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");
626 case ICE_AQ_LINK_TOPO_UNSUPP_MEDIA:
627 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");
635 * ice_print_link_msg - print link up or down message
636 * @vsi: the VSI whose link status is being queried
637 * @isup: boolean for if the link is now up or down
639 void ice_print_link_msg(struct ice_vsi *vsi, bool isup)
641 struct ice_aqc_get_phy_caps_data *caps;
642 const char *an_advertised;
643 enum ice_status status;
653 if (vsi->current_isup == isup)
656 vsi->current_isup = isup;
659 netdev_info(vsi->netdev, "NIC Link is Down\n");
663 switch (vsi->port_info->phy.link_info.link_speed) {
664 case ICE_AQ_LINK_SPEED_100GB:
667 case ICE_AQ_LINK_SPEED_50GB:
670 case ICE_AQ_LINK_SPEED_40GB:
673 case ICE_AQ_LINK_SPEED_25GB:
676 case ICE_AQ_LINK_SPEED_20GB:
679 case ICE_AQ_LINK_SPEED_10GB:
682 case ICE_AQ_LINK_SPEED_5GB:
685 case ICE_AQ_LINK_SPEED_2500MB:
688 case ICE_AQ_LINK_SPEED_1000MB:
691 case ICE_AQ_LINK_SPEED_100MB:
699 switch (vsi->port_info->fc.current_mode) {
703 case ICE_FC_TX_PAUSE:
706 case ICE_FC_RX_PAUSE:
717 /* Get FEC mode based on negotiated link info */
718 switch (vsi->port_info->phy.link_info.fec_info) {
719 case ICE_AQ_LINK_25G_RS_528_FEC_EN:
720 case ICE_AQ_LINK_25G_RS_544_FEC_EN:
723 case ICE_AQ_LINK_25G_KR_FEC_EN:
724 fec = "FC-FEC/BASE-R";
731 /* check if autoneg completed, might be false due to not supported */
732 if (vsi->port_info->phy.link_info.an_info & ICE_AQ_AN_COMPLETED)
737 /* Get FEC mode requested based on PHY caps last SW configuration */
738 caps = kzalloc(sizeof(*caps), GFP_KERNEL);
741 an_advertised = "Unknown";
745 status = ice_aq_get_phy_caps(vsi->port_info, false,
746 ICE_AQC_REPORT_ACTIVE_CFG, caps, NULL);
748 netdev_info(vsi->netdev, "Get phy capability failed.\n");
750 an_advertised = ice_is_phy_caps_an_enabled(caps) ? "On" : "Off";
752 if (caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_528_REQ ||
753 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_RS_544_REQ)
755 else if (caps->link_fec_options & ICE_AQC_PHY_FEC_10G_KR_40G_KR4_REQ ||
756 caps->link_fec_options & ICE_AQC_PHY_FEC_25G_KR_REQ)
757 fec_req = "FC-FEC/BASE-R";
764 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",
765 speed, fec_req, fec, an_advertised, an, fc);
766 ice_print_topo_conflict(vsi);
770 * ice_vsi_link_event - update the VSI's netdev
771 * @vsi: the VSI on which the link event occurred
772 * @link_up: whether or not the VSI needs to be set up or down
774 static void ice_vsi_link_event(struct ice_vsi *vsi, bool link_up)
779 if (test_bit(ICE_VSI_DOWN, vsi->state) || !vsi->netdev)
782 if (vsi->type == ICE_VSI_PF) {
783 if (link_up == netif_carrier_ok(vsi->netdev))
787 netif_carrier_on(vsi->netdev);
788 netif_tx_wake_all_queues(vsi->netdev);
790 netif_carrier_off(vsi->netdev);
791 netif_tx_stop_all_queues(vsi->netdev);
797 * ice_set_dflt_mib - send a default config MIB to the FW
798 * @pf: private PF struct
800 * This function sends a default configuration MIB to the FW.
802 * If this function errors out at any point, the driver is still able to
803 * function. The main impact is that LFC may not operate as expected.
804 * Therefore an error state in this function should be treated with a DBG
805 * message and continue on with driver rebuild/reenable.
807 static void ice_set_dflt_mib(struct ice_pf *pf)
809 struct device *dev = ice_pf_to_dev(pf);
810 u8 mib_type, *buf, *lldpmib = NULL;
811 u16 len, typelen, offset = 0;
812 struct ice_lldp_org_tlv *tlv;
813 struct ice_hw *hw = &pf->hw;
816 mib_type = SET_LOCAL_MIB_TYPE_LOCAL_MIB;
817 lldpmib = kzalloc(ICE_LLDPDU_SIZE, GFP_KERNEL);
819 dev_dbg(dev, "%s Failed to allocate MIB memory\n",
824 /* Add ETS CFG TLV */
825 tlv = (struct ice_lldp_org_tlv *)lldpmib;
826 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
827 ICE_IEEE_ETS_TLV_LEN);
828 tlv->typelen = htons(typelen);
829 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
830 ICE_IEEE_SUBTYPE_ETS_CFG);
831 tlv->ouisubtype = htonl(ouisubtype);
836 /* ETS CFG all UPs map to TC 0. Next 4 (1 - 4) Octets = 0.
837 * Octets 5 - 12 are BW values, set octet 5 to 100% BW.
838 * Octets 13 - 20 are TSA values - leave as zeros
841 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
843 tlv = (struct ice_lldp_org_tlv *)
844 ((char *)tlv + sizeof(tlv->typelen) + len);
846 /* Add ETS REC TLV */
848 tlv->typelen = htons(typelen);
850 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
851 ICE_IEEE_SUBTYPE_ETS_REC);
852 tlv->ouisubtype = htonl(ouisubtype);
854 /* First octet of buf is reserved
855 * Octets 1 - 4 map UP to TC - all UPs map to zero
856 * Octets 5 - 12 are BW values - set TC 0 to 100%.
857 * Octets 13 - 20 are TSA value - leave as zeros
861 tlv = (struct ice_lldp_org_tlv *)
862 ((char *)tlv + sizeof(tlv->typelen) + len);
864 /* Add PFC CFG TLV */
865 typelen = ((ICE_TLV_TYPE_ORG << ICE_LLDP_TLV_TYPE_S) |
866 ICE_IEEE_PFC_TLV_LEN);
867 tlv->typelen = htons(typelen);
869 ouisubtype = ((ICE_IEEE_8021QAZ_OUI << ICE_LLDP_TLV_OUI_S) |
870 ICE_IEEE_SUBTYPE_PFC_CFG);
871 tlv->ouisubtype = htonl(ouisubtype);
873 /* Octet 1 left as all zeros - PFC disabled */
875 len = (typelen & ICE_LLDP_TLV_LEN_M) >> ICE_LLDP_TLV_LEN_S;
878 if (ice_aq_set_lldp_mib(hw, mib_type, (void *)lldpmib, offset, NULL))
879 dev_dbg(dev, "%s Failed to set default LLDP MIB\n", __func__);
885 * ice_check_module_power
886 * @pf: pointer to PF struct
887 * @link_cfg_err: bitmap from the link info structure
889 * check module power level returned by a previous call to aq_get_link_info
890 * and print error messages if module power level is not supported
892 static void ice_check_module_power(struct ice_pf *pf, u8 link_cfg_err)
894 /* if module power level is supported, clear the flag */
895 if (!(link_cfg_err & (ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT |
896 ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED))) {
897 clear_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
901 /* if ICE_FLAG_MOD_POWER_UNSUPPORTED was previously set and the
902 * above block didn't clear this bit, there's nothing to do
904 if (test_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags))
907 if (link_cfg_err & ICE_AQ_LINK_INVAL_MAX_POWER_LIMIT) {
908 dev_err(ice_pf_to_dev(pf), "The installed module is incompatible with the device's NVM image. Cannot start link\n");
909 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
910 } else if (link_cfg_err & ICE_AQ_LINK_MODULE_POWER_UNSUPPORTED) {
911 dev_err(ice_pf_to_dev(pf), "The module's power requirements exceed the device's power supply. Cannot start link\n");
912 set_bit(ICE_FLAG_MOD_POWER_UNSUPPORTED, pf->flags);
917 * ice_link_event - process the link event
918 * @pf: PF that the link event is associated with
919 * @pi: port_info for the port that the link event is associated with
920 * @link_up: true if the physical link is up and false if it is down
921 * @link_speed: current link speed received from the link event
923 * Returns 0 on success and negative on failure
926 ice_link_event(struct ice_pf *pf, struct ice_port_info *pi, bool link_up,
929 struct device *dev = ice_pf_to_dev(pf);
930 struct ice_phy_info *phy_info;
931 enum ice_status status;
937 phy_info->link_info_old = phy_info->link_info;
939 old_link = !!(phy_info->link_info_old.link_info & ICE_AQ_LINK_UP);
940 old_link_speed = phy_info->link_info_old.link_speed;
942 /* update the link info structures and re-enable link events,
943 * don't bail on failure due to other book keeping needed
945 status = ice_update_link_info(pi);
947 dev_dbg(dev, "Failed to update link status on port %d, err %s aq_err %s\n",
948 pi->lport, ice_stat_str(status),
949 ice_aq_str(pi->hw->adminq.sq_last_status));
951 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
953 /* Check if the link state is up after updating link info, and treat
954 * this event as an UP event since the link is actually UP now.
956 if (phy_info->link_info.link_info & ICE_AQ_LINK_UP)
959 vsi = ice_get_main_vsi(pf);
960 if (!vsi || !vsi->port_info)
963 /* turn off PHY if media was removed */
964 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags) &&
965 !(pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE)) {
966 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
967 ice_set_link(vsi, false);
970 /* if the old link up/down and speed is the same as the new */
971 if (link_up == old_link && link_speed == old_link_speed)
974 if (ice_is_dcb_active(pf)) {
975 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
979 ice_set_dflt_mib(pf);
981 ice_vsi_link_event(vsi, link_up);
982 ice_print_link_msg(vsi, link_up);
984 ice_vc_notify_link_state(pf);
990 * ice_watchdog_subtask - periodic tasks not using event driven scheduling
991 * @pf: board private structure
993 static void ice_watchdog_subtask(struct ice_pf *pf)
997 /* if interface is down do nothing */
998 if (test_bit(ICE_DOWN, pf->state) ||
999 test_bit(ICE_CFG_BUSY, pf->state))
1002 /* make sure we don't do these things too often */
1003 if (time_before(jiffies,
1004 pf->serv_tmr_prev + pf->serv_tmr_period))
1007 pf->serv_tmr_prev = jiffies;
1009 /* Update the stats for active netdevs so the network stack
1010 * can look at updated numbers whenever it cares to
1012 ice_update_pf_stats(pf);
1013 ice_for_each_vsi(pf, i)
1014 if (pf->vsi[i] && pf->vsi[i]->netdev)
1015 ice_update_vsi_stats(pf->vsi[i]);
1019 * ice_init_link_events - enable/initialize link events
1020 * @pi: pointer to the port_info instance
1022 * Returns -EIO on failure, 0 on success
1024 static int ice_init_link_events(struct ice_port_info *pi)
1028 mask = ~((u16)(ICE_AQ_LINK_EVENT_UPDOWN | ICE_AQ_LINK_EVENT_MEDIA_NA |
1029 ICE_AQ_LINK_EVENT_MODULE_QUAL_FAIL));
1031 if (ice_aq_set_event_mask(pi->hw, pi->lport, mask, NULL)) {
1032 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to set link event mask for port %d\n",
1037 if (ice_aq_get_link_info(pi, true, NULL, NULL)) {
1038 dev_dbg(ice_hw_to_dev(pi->hw), "Failed to enable link events for port %d\n",
1047 * ice_handle_link_event - handle link event via ARQ
1048 * @pf: PF that the link event is associated with
1049 * @event: event structure containing link status info
1052 ice_handle_link_event(struct ice_pf *pf, struct ice_rq_event_info *event)
1054 struct ice_aqc_get_link_status_data *link_data;
1055 struct ice_port_info *port_info;
1058 link_data = (struct ice_aqc_get_link_status_data *)event->msg_buf;
1059 port_info = pf->hw.port_info;
1063 status = ice_link_event(pf, port_info,
1064 !!(link_data->link_info & ICE_AQ_LINK_UP),
1065 le16_to_cpu(link_data->link_speed));
1067 dev_dbg(ice_pf_to_dev(pf), "Could not process link event, error %d\n",
1073 enum ice_aq_task_state {
1074 ICE_AQ_TASK_WAITING = 0,
1075 ICE_AQ_TASK_COMPLETE,
1076 ICE_AQ_TASK_CANCELED,
1079 struct ice_aq_task {
1080 struct hlist_node entry;
1083 struct ice_rq_event_info *event;
1084 enum ice_aq_task_state state;
1088 * ice_aq_wait_for_event - Wait for an AdminQ event from firmware
1089 * @pf: pointer to the PF private structure
1090 * @opcode: the opcode to wait for
1091 * @timeout: how long to wait, in jiffies
1092 * @event: storage for the event info
1094 * Waits for a specific AdminQ completion event on the ARQ for a given PF. The
1095 * current thread will be put to sleep until the specified event occurs or
1096 * until the given timeout is reached.
1098 * To obtain only the descriptor contents, pass an event without an allocated
1099 * msg_buf. If the complete data buffer is desired, allocate the
1100 * event->msg_buf with enough space ahead of time.
1102 * Returns: zero on success, or a negative error code on failure.
1104 int ice_aq_wait_for_event(struct ice_pf *pf, u16 opcode, unsigned long timeout,
1105 struct ice_rq_event_info *event)
1107 struct device *dev = ice_pf_to_dev(pf);
1108 struct ice_aq_task *task;
1109 unsigned long start;
1113 task = kzalloc(sizeof(*task), GFP_KERNEL);
1117 INIT_HLIST_NODE(&task->entry);
1118 task->opcode = opcode;
1119 task->event = event;
1120 task->state = ICE_AQ_TASK_WAITING;
1122 spin_lock_bh(&pf->aq_wait_lock);
1123 hlist_add_head(&task->entry, &pf->aq_wait_list);
1124 spin_unlock_bh(&pf->aq_wait_lock);
1128 ret = wait_event_interruptible_timeout(pf->aq_wait_queue, task->state,
1130 switch (task->state) {
1131 case ICE_AQ_TASK_WAITING:
1132 err = ret < 0 ? ret : -ETIMEDOUT;
1134 case ICE_AQ_TASK_CANCELED:
1135 err = ret < 0 ? ret : -ECANCELED;
1137 case ICE_AQ_TASK_COMPLETE:
1138 err = ret < 0 ? ret : 0;
1141 WARN(1, "Unexpected AdminQ wait task state %u", task->state);
1146 dev_dbg(dev, "Waited %u msecs (max %u msecs) for firmware response to op 0x%04x\n",
1147 jiffies_to_msecs(jiffies - start),
1148 jiffies_to_msecs(timeout),
1151 spin_lock_bh(&pf->aq_wait_lock);
1152 hlist_del(&task->entry);
1153 spin_unlock_bh(&pf->aq_wait_lock);
1160 * ice_aq_check_events - Check if any thread is waiting for an AdminQ event
1161 * @pf: pointer to the PF private structure
1162 * @opcode: the opcode of the event
1163 * @event: the event to check
1165 * Loops over the current list of pending threads waiting for an AdminQ event.
1166 * For each matching task, copy the contents of the event into the task
1167 * structure and wake up the thread.
1169 * If multiple threads wait for the same opcode, they will all be woken up.
1171 * Note that event->msg_buf will only be duplicated if the event has a buffer
1172 * with enough space already allocated. Otherwise, only the descriptor and
1173 * message length will be copied.
1175 * Returns: true if an event was found, false otherwise
1177 static void ice_aq_check_events(struct ice_pf *pf, u16 opcode,
1178 struct ice_rq_event_info *event)
1180 struct ice_aq_task *task;
1183 spin_lock_bh(&pf->aq_wait_lock);
1184 hlist_for_each_entry(task, &pf->aq_wait_list, entry) {
1185 if (task->state || task->opcode != opcode)
1188 memcpy(&task->event->desc, &event->desc, sizeof(event->desc));
1189 task->event->msg_len = event->msg_len;
1191 /* Only copy the data buffer if a destination was set */
1192 if (task->event->msg_buf &&
1193 task->event->buf_len > event->buf_len) {
1194 memcpy(task->event->msg_buf, event->msg_buf,
1196 task->event->buf_len = event->buf_len;
1199 task->state = ICE_AQ_TASK_COMPLETE;
1202 spin_unlock_bh(&pf->aq_wait_lock);
1205 wake_up(&pf->aq_wait_queue);
1209 * ice_aq_cancel_waiting_tasks - Immediately cancel all waiting tasks
1210 * @pf: the PF private structure
1212 * Set all waiting tasks to ICE_AQ_TASK_CANCELED, and wake up their threads.
1213 * This will then cause ice_aq_wait_for_event to exit with -ECANCELED.
1215 static void ice_aq_cancel_waiting_tasks(struct ice_pf *pf)
1217 struct ice_aq_task *task;
1219 spin_lock_bh(&pf->aq_wait_lock);
1220 hlist_for_each_entry(task, &pf->aq_wait_list, entry)
1221 task->state = ICE_AQ_TASK_CANCELED;
1222 spin_unlock_bh(&pf->aq_wait_lock);
1224 wake_up(&pf->aq_wait_queue);
1228 * __ice_clean_ctrlq - helper function to clean controlq rings
1229 * @pf: ptr to struct ice_pf
1230 * @q_type: specific Control queue type
1232 static int __ice_clean_ctrlq(struct ice_pf *pf, enum ice_ctl_q q_type)
1234 struct device *dev = ice_pf_to_dev(pf);
1235 struct ice_rq_event_info event;
1236 struct ice_hw *hw = &pf->hw;
1237 struct ice_ctl_q_info *cq;
1242 /* Do not clean control queue if/when PF reset fails */
1243 if (test_bit(ICE_RESET_FAILED, pf->state))
1247 case ICE_CTL_Q_ADMIN:
1255 case ICE_CTL_Q_MAILBOX:
1258 /* we are going to try to detect a malicious VF, so set the
1259 * state to begin detection
1261 hw->mbx_snapshot.mbx_buf.state = ICE_MAL_VF_DETECT_STATE_NEW_SNAPSHOT;
1264 dev_warn(dev, "Unknown control queue type 0x%x\n", q_type);
1268 /* check for error indications - PF_xx_AxQLEN register layout for
1269 * FW/MBX/SB are identical so just use defines for PF_FW_AxQLEN.
1271 val = rd32(hw, cq->rq.len);
1272 if (val & (PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1273 PF_FW_ARQLEN_ARQCRIT_M)) {
1275 if (val & PF_FW_ARQLEN_ARQVFE_M)
1276 dev_dbg(dev, "%s Receive Queue VF Error detected\n",
1278 if (val & PF_FW_ARQLEN_ARQOVFL_M) {
1279 dev_dbg(dev, "%s Receive Queue Overflow Error detected\n",
1282 if (val & PF_FW_ARQLEN_ARQCRIT_M)
1283 dev_dbg(dev, "%s Receive Queue Critical Error detected\n",
1285 val &= ~(PF_FW_ARQLEN_ARQVFE_M | PF_FW_ARQLEN_ARQOVFL_M |
1286 PF_FW_ARQLEN_ARQCRIT_M);
1288 wr32(hw, cq->rq.len, val);
1291 val = rd32(hw, cq->sq.len);
1292 if (val & (PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1293 PF_FW_ATQLEN_ATQCRIT_M)) {
1295 if (val & PF_FW_ATQLEN_ATQVFE_M)
1296 dev_dbg(dev, "%s Send Queue VF Error detected\n",
1298 if (val & PF_FW_ATQLEN_ATQOVFL_M) {
1299 dev_dbg(dev, "%s Send Queue Overflow Error detected\n",
1302 if (val & PF_FW_ATQLEN_ATQCRIT_M)
1303 dev_dbg(dev, "%s Send Queue Critical Error detected\n",
1305 val &= ~(PF_FW_ATQLEN_ATQVFE_M | PF_FW_ATQLEN_ATQOVFL_M |
1306 PF_FW_ATQLEN_ATQCRIT_M);
1308 wr32(hw, cq->sq.len, val);
1311 event.buf_len = cq->rq_buf_size;
1312 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
1317 enum ice_status ret;
1320 ret = ice_clean_rq_elem(hw, cq, &event, &pending);
1321 if (ret == ICE_ERR_AQ_NO_WORK)
1324 dev_err(dev, "%s Receive Queue event error %s\n", qtype,
1329 opcode = le16_to_cpu(event.desc.opcode);
1331 /* Notify any thread that might be waiting for this event */
1332 ice_aq_check_events(pf, opcode, &event);
1335 case ice_aqc_opc_get_link_status:
1336 if (ice_handle_link_event(pf, &event))
1337 dev_err(dev, "Could not handle link event\n");
1339 case ice_aqc_opc_event_lan_overflow:
1340 ice_vf_lan_overflow_event(pf, &event);
1342 case ice_mbx_opc_send_msg_to_pf:
1343 if (!ice_is_malicious_vf(pf, &event, i, pending))
1344 ice_vc_process_vf_msg(pf, &event);
1346 case ice_aqc_opc_fw_logging:
1347 ice_output_fw_log(hw, &event.desc, event.msg_buf);
1349 case ice_aqc_opc_lldp_set_mib_change:
1350 ice_dcb_process_lldp_set_mib_change(pf, &event);
1353 dev_dbg(dev, "%s Receive Queue unknown event 0x%04x ignored\n",
1357 } while (pending && (i++ < ICE_DFLT_IRQ_WORK));
1359 kfree(event.msg_buf);
1361 return pending && (i == ICE_DFLT_IRQ_WORK);
1365 * ice_ctrlq_pending - check if there is a difference between ntc and ntu
1366 * @hw: pointer to hardware info
1367 * @cq: control queue information
1369 * returns true if there are pending messages in a queue, false if there aren't
1371 static bool ice_ctrlq_pending(struct ice_hw *hw, struct ice_ctl_q_info *cq)
1375 ntu = (u16)(rd32(hw, cq->rq.head) & cq->rq.head_mask);
1376 return cq->rq.next_to_clean != ntu;
1380 * ice_clean_adminq_subtask - clean the AdminQ rings
1381 * @pf: board private structure
1383 static void ice_clean_adminq_subtask(struct ice_pf *pf)
1385 struct ice_hw *hw = &pf->hw;
1387 if (!test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
1390 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN))
1393 clear_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
1395 /* There might be a situation where new messages arrive to a control
1396 * queue between processing the last message and clearing the
1397 * EVENT_PENDING bit. So before exiting, check queue head again (using
1398 * ice_ctrlq_pending) and process new messages if any.
1400 if (ice_ctrlq_pending(hw, &hw->adminq))
1401 __ice_clean_ctrlq(pf, ICE_CTL_Q_ADMIN);
1407 * ice_clean_mailboxq_subtask - clean the MailboxQ rings
1408 * @pf: board private structure
1410 static void ice_clean_mailboxq_subtask(struct ice_pf *pf)
1412 struct ice_hw *hw = &pf->hw;
1414 if (!test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state))
1417 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX))
1420 clear_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
1422 if (ice_ctrlq_pending(hw, &hw->mailboxq))
1423 __ice_clean_ctrlq(pf, ICE_CTL_Q_MAILBOX);
1429 * ice_clean_sbq_subtask - clean the Sideband Queue rings
1430 * @pf: board private structure
1432 static void ice_clean_sbq_subtask(struct ice_pf *pf)
1434 struct ice_hw *hw = &pf->hw;
1436 /* Nothing to do here if sideband queue is not supported */
1437 if (!ice_is_sbq_supported(hw)) {
1438 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1442 if (!test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state))
1445 if (__ice_clean_ctrlq(pf, ICE_CTL_Q_SB))
1448 clear_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
1450 if (ice_ctrlq_pending(hw, &hw->sbq))
1451 __ice_clean_ctrlq(pf, ICE_CTL_Q_SB);
1457 * ice_service_task_schedule - schedule the service task to wake up
1458 * @pf: board private structure
1460 * If not already scheduled, this puts the task into the work queue.
1462 void ice_service_task_schedule(struct ice_pf *pf)
1464 if (!test_bit(ICE_SERVICE_DIS, pf->state) &&
1465 !test_and_set_bit(ICE_SERVICE_SCHED, pf->state) &&
1466 !test_bit(ICE_NEEDS_RESTART, pf->state))
1467 queue_work(ice_wq, &pf->serv_task);
1471 * ice_service_task_complete - finish up the service task
1472 * @pf: board private structure
1474 static void ice_service_task_complete(struct ice_pf *pf)
1476 WARN_ON(!test_bit(ICE_SERVICE_SCHED, pf->state));
1478 /* force memory (pf->state) to sync before next service task */
1479 smp_mb__before_atomic();
1480 clear_bit(ICE_SERVICE_SCHED, pf->state);
1484 * ice_service_task_stop - stop service task and cancel works
1485 * @pf: board private structure
1487 * Return 0 if the ICE_SERVICE_DIS bit was not already set,
1490 static int ice_service_task_stop(struct ice_pf *pf)
1494 ret = test_and_set_bit(ICE_SERVICE_DIS, pf->state);
1496 if (pf->serv_tmr.function)
1497 del_timer_sync(&pf->serv_tmr);
1498 if (pf->serv_task.func)
1499 cancel_work_sync(&pf->serv_task);
1501 clear_bit(ICE_SERVICE_SCHED, pf->state);
1506 * ice_service_task_restart - restart service task and schedule works
1507 * @pf: board private structure
1509 * This function is needed for suspend and resume works (e.g WoL scenario)
1511 static void ice_service_task_restart(struct ice_pf *pf)
1513 clear_bit(ICE_SERVICE_DIS, pf->state);
1514 ice_service_task_schedule(pf);
1518 * ice_service_timer - timer callback to schedule service task
1519 * @t: pointer to timer_list
1521 static void ice_service_timer(struct timer_list *t)
1523 struct ice_pf *pf = from_timer(pf, t, serv_tmr);
1525 mod_timer(&pf->serv_tmr, round_jiffies(pf->serv_tmr_period + jiffies));
1526 ice_service_task_schedule(pf);
1530 * ice_handle_mdd_event - handle malicious driver detect event
1531 * @pf: pointer to the PF structure
1533 * Called from service task. OICR interrupt handler indicates MDD event.
1534 * VF MDD logging is guarded by net_ratelimit. Additional PF and VF log
1535 * messages are wrapped by netif_msg_[rx|tx]_err. Since VF Rx MDD events
1536 * disable the queue, the PF can be configured to reset the VF using ethtool
1537 * private flag mdd-auto-reset-vf.
1539 static void ice_handle_mdd_event(struct ice_pf *pf)
1541 struct device *dev = ice_pf_to_dev(pf);
1542 struct ice_hw *hw = &pf->hw;
1546 if (!test_and_clear_bit(ICE_MDD_EVENT_PENDING, pf->state)) {
1547 /* Since the VF MDD event logging is rate limited, check if
1548 * there are pending MDD events.
1550 ice_print_vfs_mdd_events(pf);
1554 /* find what triggered an MDD event */
1555 reg = rd32(hw, GL_MDET_TX_PQM);
1556 if (reg & GL_MDET_TX_PQM_VALID_M) {
1557 u8 pf_num = (reg & GL_MDET_TX_PQM_PF_NUM_M) >>
1558 GL_MDET_TX_PQM_PF_NUM_S;
1559 u16 vf_num = (reg & GL_MDET_TX_PQM_VF_NUM_M) >>
1560 GL_MDET_TX_PQM_VF_NUM_S;
1561 u8 event = (reg & GL_MDET_TX_PQM_MAL_TYPE_M) >>
1562 GL_MDET_TX_PQM_MAL_TYPE_S;
1563 u16 queue = ((reg & GL_MDET_TX_PQM_QNUM_M) >>
1564 GL_MDET_TX_PQM_QNUM_S);
1566 if (netif_msg_tx_err(pf))
1567 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1568 event, queue, pf_num, vf_num);
1569 wr32(hw, GL_MDET_TX_PQM, 0xffffffff);
1572 reg = rd32(hw, GL_MDET_TX_TCLAN);
1573 if (reg & GL_MDET_TX_TCLAN_VALID_M) {
1574 u8 pf_num = (reg & GL_MDET_TX_TCLAN_PF_NUM_M) >>
1575 GL_MDET_TX_TCLAN_PF_NUM_S;
1576 u16 vf_num = (reg & GL_MDET_TX_TCLAN_VF_NUM_M) >>
1577 GL_MDET_TX_TCLAN_VF_NUM_S;
1578 u8 event = (reg & GL_MDET_TX_TCLAN_MAL_TYPE_M) >>
1579 GL_MDET_TX_TCLAN_MAL_TYPE_S;
1580 u16 queue = ((reg & GL_MDET_TX_TCLAN_QNUM_M) >>
1581 GL_MDET_TX_TCLAN_QNUM_S);
1583 if (netif_msg_tx_err(pf))
1584 dev_info(dev, "Malicious Driver Detection event %d on TX queue %d PF# %d VF# %d\n",
1585 event, queue, pf_num, vf_num);
1586 wr32(hw, GL_MDET_TX_TCLAN, 0xffffffff);
1589 reg = rd32(hw, GL_MDET_RX);
1590 if (reg & GL_MDET_RX_VALID_M) {
1591 u8 pf_num = (reg & GL_MDET_RX_PF_NUM_M) >>
1592 GL_MDET_RX_PF_NUM_S;
1593 u16 vf_num = (reg & GL_MDET_RX_VF_NUM_M) >>
1594 GL_MDET_RX_VF_NUM_S;
1595 u8 event = (reg & GL_MDET_RX_MAL_TYPE_M) >>
1596 GL_MDET_RX_MAL_TYPE_S;
1597 u16 queue = ((reg & GL_MDET_RX_QNUM_M) >>
1600 if (netif_msg_rx_err(pf))
1601 dev_info(dev, "Malicious Driver Detection event %d on RX queue %d PF# %d VF# %d\n",
1602 event, queue, pf_num, vf_num);
1603 wr32(hw, GL_MDET_RX, 0xffffffff);
1606 /* check to see if this PF caused an MDD event */
1607 reg = rd32(hw, PF_MDET_TX_PQM);
1608 if (reg & PF_MDET_TX_PQM_VALID_M) {
1609 wr32(hw, PF_MDET_TX_PQM, 0xFFFF);
1610 if (netif_msg_tx_err(pf))
1611 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on PF\n");
1614 reg = rd32(hw, PF_MDET_TX_TCLAN);
1615 if (reg & PF_MDET_TX_TCLAN_VALID_M) {
1616 wr32(hw, PF_MDET_TX_TCLAN, 0xFFFF);
1617 if (netif_msg_tx_err(pf))
1618 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on PF\n");
1621 reg = rd32(hw, PF_MDET_RX);
1622 if (reg & PF_MDET_RX_VALID_M) {
1623 wr32(hw, PF_MDET_RX, 0xFFFF);
1624 if (netif_msg_rx_err(pf))
1625 dev_info(dev, "Malicious Driver Detection event RX detected on PF\n");
1628 /* Check to see if one of the VFs caused an MDD event, and then
1629 * increment counters and set print pending
1631 ice_for_each_vf(pf, i) {
1632 struct ice_vf *vf = &pf->vf[i];
1634 reg = rd32(hw, VP_MDET_TX_PQM(i));
1635 if (reg & VP_MDET_TX_PQM_VALID_M) {
1636 wr32(hw, VP_MDET_TX_PQM(i), 0xFFFF);
1637 vf->mdd_tx_events.count++;
1638 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1639 if (netif_msg_tx_err(pf))
1640 dev_info(dev, "Malicious Driver Detection event TX_PQM detected on VF %d\n",
1644 reg = rd32(hw, VP_MDET_TX_TCLAN(i));
1645 if (reg & VP_MDET_TX_TCLAN_VALID_M) {
1646 wr32(hw, VP_MDET_TX_TCLAN(i), 0xFFFF);
1647 vf->mdd_tx_events.count++;
1648 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1649 if (netif_msg_tx_err(pf))
1650 dev_info(dev, "Malicious Driver Detection event TX_TCLAN detected on VF %d\n",
1654 reg = rd32(hw, VP_MDET_TX_TDPU(i));
1655 if (reg & VP_MDET_TX_TDPU_VALID_M) {
1656 wr32(hw, VP_MDET_TX_TDPU(i), 0xFFFF);
1657 vf->mdd_tx_events.count++;
1658 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1659 if (netif_msg_tx_err(pf))
1660 dev_info(dev, "Malicious Driver Detection event TX_TDPU detected on VF %d\n",
1664 reg = rd32(hw, VP_MDET_RX(i));
1665 if (reg & VP_MDET_RX_VALID_M) {
1666 wr32(hw, VP_MDET_RX(i), 0xFFFF);
1667 vf->mdd_rx_events.count++;
1668 set_bit(ICE_MDD_VF_PRINT_PENDING, pf->state);
1669 if (netif_msg_rx_err(pf))
1670 dev_info(dev, "Malicious Driver Detection event RX detected on VF %d\n",
1673 /* Since the queue is disabled on VF Rx MDD events, the
1674 * PF can be configured to reset the VF through ethtool
1675 * private flag mdd-auto-reset-vf.
1677 if (test_bit(ICE_FLAG_MDD_AUTO_RESET_VF, pf->flags)) {
1678 /* VF MDD event counters will be cleared by
1679 * reset, so print the event prior to reset.
1681 ice_print_vf_rx_mdd_event(vf);
1682 ice_reset_vf(&pf->vf[i], false);
1687 ice_print_vfs_mdd_events(pf);
1691 * ice_force_phys_link_state - Force the physical link state
1692 * @vsi: VSI to force the physical link state to up/down
1693 * @link_up: true/false indicates to set the physical link to up/down
1695 * Force the physical link state by getting the current PHY capabilities from
1696 * hardware and setting the PHY config based on the determined capabilities. If
1697 * link changes a link event will be triggered because both the Enable Automatic
1698 * Link Update and LESM Enable bits are set when setting the PHY capabilities.
1700 * Returns 0 on success, negative on failure
1702 static int ice_force_phys_link_state(struct ice_vsi *vsi, bool link_up)
1704 struct ice_aqc_get_phy_caps_data *pcaps;
1705 struct ice_aqc_set_phy_cfg_data *cfg;
1706 struct ice_port_info *pi;
1710 if (!vsi || !vsi->port_info || !vsi->back)
1712 if (vsi->type != ICE_VSI_PF)
1715 dev = ice_pf_to_dev(vsi->back);
1717 pi = vsi->port_info;
1719 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1723 retcode = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1726 dev_err(dev, "Failed to get phy capabilities, VSI %d error %d\n",
1727 vsi->vsi_num, retcode);
1732 /* No change in link */
1733 if (link_up == !!(pcaps->caps & ICE_AQC_PHY_EN_LINK) &&
1734 link_up == !!(pi->phy.link_info.link_info & ICE_AQ_LINK_UP))
1737 /* Use the current user PHY configuration. The current user PHY
1738 * configuration is initialized during probe from PHY capabilities
1739 * software mode, and updated on set PHY configuration.
1741 cfg = kmemdup(&pi->phy.curr_user_phy_cfg, sizeof(*cfg), GFP_KERNEL);
1747 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT;
1749 cfg->caps |= ICE_AQ_PHY_ENA_LINK;
1751 cfg->caps &= ~ICE_AQ_PHY_ENA_LINK;
1753 retcode = ice_aq_set_phy_cfg(&vsi->back->hw, pi, cfg, NULL);
1755 dev_err(dev, "Failed to set phy config, VSI %d error %d\n",
1756 vsi->vsi_num, retcode);
1767 * ice_init_nvm_phy_type - Initialize the NVM PHY type
1768 * @pi: port info structure
1770 * Initialize nvm_phy_type_[low|high] for link lenient mode support
1772 static int ice_init_nvm_phy_type(struct ice_port_info *pi)
1774 struct ice_aqc_get_phy_caps_data *pcaps;
1775 struct ice_pf *pf = pi->hw->back;
1776 enum ice_status status;
1779 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1783 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_NO_MEDIA, pcaps,
1787 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1792 pf->nvm_phy_type_hi = pcaps->phy_type_high;
1793 pf->nvm_phy_type_lo = pcaps->phy_type_low;
1801 * ice_init_link_dflt_override - Initialize link default override
1802 * @pi: port info structure
1804 * Initialize link default override and PHY total port shutdown during probe
1806 static void ice_init_link_dflt_override(struct ice_port_info *pi)
1808 struct ice_link_default_override_tlv *ldo;
1809 struct ice_pf *pf = pi->hw->back;
1811 ldo = &pf->link_dflt_override;
1812 if (ice_get_link_default_override(ldo, pi))
1815 if (!(ldo->options & ICE_LINK_OVERRIDE_PORT_DIS))
1818 /* Enable Total Port Shutdown (override/replace link-down-on-close
1819 * ethtool private flag) for ports with Port Disable bit set.
1821 set_bit(ICE_FLAG_TOTAL_PORT_SHUTDOWN_ENA, pf->flags);
1822 set_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags);
1826 * ice_init_phy_cfg_dflt_override - Initialize PHY cfg default override settings
1827 * @pi: port info structure
1829 * If default override is enabled, initialize the user PHY cfg speed and FEC
1830 * settings using the default override mask from the NVM.
1832 * The PHY should only be configured with the default override settings the
1833 * first time media is available. The ICE_LINK_DEFAULT_OVERRIDE_PENDING state
1834 * is used to indicate that the user PHY cfg default override is initialized
1835 * and the PHY has not been configured with the default override settings. The
1836 * state is set here, and cleared in ice_configure_phy the first time the PHY is
1839 * This function should be called only if the FW doesn't support default
1840 * configuration mode, as reported by ice_fw_supports_report_dflt_cfg.
1842 static void ice_init_phy_cfg_dflt_override(struct ice_port_info *pi)
1844 struct ice_link_default_override_tlv *ldo;
1845 struct ice_aqc_set_phy_cfg_data *cfg;
1846 struct ice_phy_info *phy = &pi->phy;
1847 struct ice_pf *pf = pi->hw->back;
1849 ldo = &pf->link_dflt_override;
1851 /* If link default override is enabled, use to mask NVM PHY capabilities
1852 * for speed and FEC default configuration.
1854 cfg = &phy->curr_user_phy_cfg;
1856 if (ldo->phy_type_low || ldo->phy_type_high) {
1857 cfg->phy_type_low = pf->nvm_phy_type_lo &
1858 cpu_to_le64(ldo->phy_type_low);
1859 cfg->phy_type_high = pf->nvm_phy_type_hi &
1860 cpu_to_le64(ldo->phy_type_high);
1862 cfg->link_fec_opt = ldo->fec_options;
1863 phy->curr_user_fec_req = ICE_FEC_AUTO;
1865 set_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING, pf->state);
1869 * ice_init_phy_user_cfg - Initialize the PHY user configuration
1870 * @pi: port info structure
1872 * Initialize the current user PHY configuration, speed, FEC, and FC requested
1873 * mode to default. The PHY defaults are from get PHY capabilities topology
1874 * with media so call when media is first available. An error is returned if
1875 * called when media is not available. The PHY initialization completed state is
1878 * These configurations are used when setting PHY
1879 * configuration. The user PHY configuration is updated on set PHY
1880 * configuration. Returns 0 on success, negative on failure
1882 static int ice_init_phy_user_cfg(struct ice_port_info *pi)
1884 struct ice_aqc_get_phy_caps_data *pcaps;
1885 struct ice_phy_info *phy = &pi->phy;
1886 struct ice_pf *pf = pi->hw->back;
1887 enum ice_status status;
1890 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1893 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1897 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1898 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
1901 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
1904 dev_err(ice_pf_to_dev(pf), "Get PHY capability failed.\n");
1909 ice_copy_phy_caps_to_cfg(pi, pcaps, &pi->phy.curr_user_phy_cfg);
1911 /* check if lenient mode is supported and enabled */
1912 if (ice_fw_supports_link_override(pi->hw) &&
1913 !(pcaps->module_compliance_enforcement &
1914 ICE_AQC_MOD_ENFORCE_STRICT_MODE)) {
1915 set_bit(ICE_FLAG_LINK_LENIENT_MODE_ENA, pf->flags);
1917 /* if the FW supports default PHY configuration mode, then the driver
1918 * does not have to apply link override settings. If not,
1919 * initialize user PHY configuration with link override values
1921 if (!ice_fw_supports_report_dflt_cfg(pi->hw) &&
1922 (pf->link_dflt_override.options & ICE_LINK_OVERRIDE_EN)) {
1923 ice_init_phy_cfg_dflt_override(pi);
1928 /* if link default override is not enabled, set user flow control and
1929 * FEC settings based on what get_phy_caps returned
1931 phy->curr_user_fec_req = ice_caps_to_fec_mode(pcaps->caps,
1932 pcaps->link_fec_options);
1933 phy->curr_user_fc_req = ice_caps_to_fc_mode(pcaps->caps);
1936 phy->curr_user_speed_req = ICE_AQ_LINK_SPEED_M;
1937 set_bit(ICE_PHY_INIT_COMPLETE, pf->state);
1944 * ice_configure_phy - configure PHY
1947 * Set the PHY configuration. If the current PHY configuration is the same as
1948 * the curr_user_phy_cfg, then do nothing to avoid link flap. Otherwise
1949 * configure the based get PHY capabilities for topology with media.
1951 static int ice_configure_phy(struct ice_vsi *vsi)
1953 struct device *dev = ice_pf_to_dev(vsi->back);
1954 struct ice_port_info *pi = vsi->port_info;
1955 struct ice_aqc_get_phy_caps_data *pcaps;
1956 struct ice_aqc_set_phy_cfg_data *cfg;
1957 struct ice_phy_info *phy = &pi->phy;
1958 struct ice_pf *pf = vsi->back;
1959 enum ice_status status;
1962 /* Ensure we have media as we cannot configure a medialess port */
1963 if (!(phy->link_info.link_info & ICE_AQ_MEDIA_AVAILABLE))
1966 ice_print_topo_conflict(vsi);
1968 if (phy->link_info.topo_media_conflict == ICE_AQ_LINK_TOPO_UNSUPP_MEDIA)
1971 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags))
1972 return ice_force_phys_link_state(vsi, true);
1974 pcaps = kzalloc(sizeof(*pcaps), GFP_KERNEL);
1978 /* Get current PHY config */
1979 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_ACTIVE_CFG, pcaps,
1982 dev_err(dev, "Failed to get PHY configuration, VSI %d error %s\n",
1983 vsi->vsi_num, ice_stat_str(status));
1988 /* If PHY enable link is configured and configuration has not changed,
1989 * there's nothing to do
1991 if (pcaps->caps & ICE_AQC_PHY_EN_LINK &&
1992 ice_phy_caps_equals_cfg(pcaps, &phy->curr_user_phy_cfg))
1995 /* Use PHY topology as baseline for configuration */
1996 memset(pcaps, 0, sizeof(*pcaps));
1997 if (ice_fw_supports_report_dflt_cfg(pi->hw))
1998 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_DFLT_CFG,
2001 status = ice_aq_get_phy_caps(pi, false, ICE_AQC_REPORT_TOPO_CAP_MEDIA,
2004 dev_err(dev, "Failed to get PHY caps, VSI %d error %s\n",
2005 vsi->vsi_num, ice_stat_str(status));
2010 cfg = kzalloc(sizeof(*cfg), GFP_KERNEL);
2016 ice_copy_phy_caps_to_cfg(pi, pcaps, cfg);
2018 /* Speed - If default override pending, use curr_user_phy_cfg set in
2019 * ice_init_phy_user_cfg_ldo.
2021 if (test_and_clear_bit(ICE_LINK_DEFAULT_OVERRIDE_PENDING,
2022 vsi->back->state)) {
2023 cfg->phy_type_low = phy->curr_user_phy_cfg.phy_type_low;
2024 cfg->phy_type_high = phy->curr_user_phy_cfg.phy_type_high;
2026 u64 phy_low = 0, phy_high = 0;
2028 ice_update_phy_type(&phy_low, &phy_high,
2029 pi->phy.curr_user_speed_req);
2030 cfg->phy_type_low = pcaps->phy_type_low & cpu_to_le64(phy_low);
2031 cfg->phy_type_high = pcaps->phy_type_high &
2032 cpu_to_le64(phy_high);
2035 /* Can't provide what was requested; use PHY capabilities */
2036 if (!cfg->phy_type_low && !cfg->phy_type_high) {
2037 cfg->phy_type_low = pcaps->phy_type_low;
2038 cfg->phy_type_high = pcaps->phy_type_high;
2042 ice_cfg_phy_fec(pi, cfg, phy->curr_user_fec_req);
2044 /* Can't provide what was requested; use PHY capabilities */
2045 if (cfg->link_fec_opt !=
2046 (cfg->link_fec_opt & pcaps->link_fec_options)) {
2047 cfg->caps |= pcaps->caps & ICE_AQC_PHY_EN_AUTO_FEC;
2048 cfg->link_fec_opt = pcaps->link_fec_options;
2051 /* Flow Control - always supported; no need to check against
2054 ice_cfg_phy_fc(pi, cfg, phy->curr_user_fc_req);
2056 /* Enable link and link update */
2057 cfg->caps |= ICE_AQ_PHY_ENA_AUTO_LINK_UPDT | ICE_AQ_PHY_ENA_LINK;
2059 status = ice_aq_set_phy_cfg(&pf->hw, pi, cfg, NULL);
2061 dev_err(dev, "Failed to set phy config, VSI %d error %s\n",
2062 vsi->vsi_num, ice_stat_str(status));
2073 * ice_check_media_subtask - Check for media
2074 * @pf: pointer to PF struct
2076 * If media is available, then initialize PHY user configuration if it is not
2077 * been, and configure the PHY if the interface is up.
2079 static void ice_check_media_subtask(struct ice_pf *pf)
2081 struct ice_port_info *pi;
2082 struct ice_vsi *vsi;
2085 /* No need to check for media if it's already present */
2086 if (!test_bit(ICE_FLAG_NO_MEDIA, pf->flags))
2089 vsi = ice_get_main_vsi(pf);
2093 /* Refresh link info and check if media is present */
2094 pi = vsi->port_info;
2095 err = ice_update_link_info(pi);
2099 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
2101 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
2102 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state))
2103 ice_init_phy_user_cfg(pi);
2105 /* PHY settings are reset on media insertion, reconfigure
2106 * PHY to preserve settings.
2108 if (test_bit(ICE_VSI_DOWN, vsi->state) &&
2109 test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags))
2112 err = ice_configure_phy(vsi);
2114 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
2116 /* A Link Status Event will be generated; the event handler
2117 * will complete bringing the interface up
2123 * ice_service_task - manage and run subtasks
2124 * @work: pointer to work_struct contained by the PF struct
2126 static void ice_service_task(struct work_struct *work)
2128 struct ice_pf *pf = container_of(work, struct ice_pf, serv_task);
2129 unsigned long start_time = jiffies;
2133 /* process reset requests first */
2134 ice_reset_subtask(pf);
2136 /* bail if a reset/recovery cycle is pending or rebuild failed */
2137 if (ice_is_reset_in_progress(pf->state) ||
2138 test_bit(ICE_SUSPENDED, pf->state) ||
2139 test_bit(ICE_NEEDS_RESTART, pf->state)) {
2140 ice_service_task_complete(pf);
2144 ice_clean_adminq_subtask(pf);
2145 ice_check_media_subtask(pf);
2146 ice_check_for_hang_subtask(pf);
2147 ice_sync_fltr_subtask(pf);
2148 ice_handle_mdd_event(pf);
2149 ice_watchdog_subtask(pf);
2151 if (ice_is_safe_mode(pf)) {
2152 ice_service_task_complete(pf);
2156 ice_process_vflr_event(pf);
2157 ice_clean_mailboxq_subtask(pf);
2158 ice_clean_sbq_subtask(pf);
2159 ice_sync_arfs_fltrs(pf);
2160 ice_flush_fdir_ctx(pf);
2162 /* Clear ICE_SERVICE_SCHED flag to allow scheduling next event */
2163 ice_service_task_complete(pf);
2165 /* If the tasks have taken longer than one service timer period
2166 * or there is more work to be done, reset the service timer to
2167 * schedule the service task now.
2169 if (time_after(jiffies, (start_time + pf->serv_tmr_period)) ||
2170 test_bit(ICE_MDD_EVENT_PENDING, pf->state) ||
2171 test_bit(ICE_VFLR_EVENT_PENDING, pf->state) ||
2172 test_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state) ||
2173 test_bit(ICE_FD_VF_FLUSH_CTX, pf->state) ||
2174 test_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state) ||
2175 test_bit(ICE_ADMINQ_EVENT_PENDING, pf->state))
2176 mod_timer(&pf->serv_tmr, jiffies);
2180 * ice_set_ctrlq_len - helper function to set controlq length
2181 * @hw: pointer to the HW instance
2183 static void ice_set_ctrlq_len(struct ice_hw *hw)
2185 hw->adminq.num_rq_entries = ICE_AQ_LEN;
2186 hw->adminq.num_sq_entries = ICE_AQ_LEN;
2187 hw->adminq.rq_buf_size = ICE_AQ_MAX_BUF_LEN;
2188 hw->adminq.sq_buf_size = ICE_AQ_MAX_BUF_LEN;
2189 hw->mailboxq.num_rq_entries = PF_MBX_ARQLEN_ARQLEN_M;
2190 hw->mailboxq.num_sq_entries = ICE_MBXSQ_LEN;
2191 hw->mailboxq.rq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2192 hw->mailboxq.sq_buf_size = ICE_MBXQ_MAX_BUF_LEN;
2193 hw->sbq.num_rq_entries = ICE_SBQ_LEN;
2194 hw->sbq.num_sq_entries = ICE_SBQ_LEN;
2195 hw->sbq.rq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2196 hw->sbq.sq_buf_size = ICE_SBQ_MAX_BUF_LEN;
2200 * ice_schedule_reset - schedule a reset
2201 * @pf: board private structure
2202 * @reset: reset being requested
2204 int ice_schedule_reset(struct ice_pf *pf, enum ice_reset_req reset)
2206 struct device *dev = ice_pf_to_dev(pf);
2208 /* bail out if earlier reset has failed */
2209 if (test_bit(ICE_RESET_FAILED, pf->state)) {
2210 dev_dbg(dev, "earlier reset has failed\n");
2213 /* bail if reset/recovery already in progress */
2214 if (ice_is_reset_in_progress(pf->state)) {
2215 dev_dbg(dev, "Reset already in progress\n");
2219 ice_unplug_aux_dev(pf);
2223 set_bit(ICE_PFR_REQ, pf->state);
2225 case ICE_RESET_CORER:
2226 set_bit(ICE_CORER_REQ, pf->state);
2228 case ICE_RESET_GLOBR:
2229 set_bit(ICE_GLOBR_REQ, pf->state);
2235 ice_service_task_schedule(pf);
2240 * ice_irq_affinity_notify - Callback for affinity changes
2241 * @notify: context as to what irq was changed
2242 * @mask: the new affinity mask
2244 * This is a callback function used by the irq_set_affinity_notifier function
2245 * so that we may register to receive changes to the irq affinity masks.
2248 ice_irq_affinity_notify(struct irq_affinity_notify *notify,
2249 const cpumask_t *mask)
2251 struct ice_q_vector *q_vector =
2252 container_of(notify, struct ice_q_vector, affinity_notify);
2254 cpumask_copy(&q_vector->affinity_mask, mask);
2258 * ice_irq_affinity_release - Callback for affinity notifier release
2259 * @ref: internal core kernel usage
2261 * This is a callback function used by the irq_set_affinity_notifier function
2262 * to inform the current notification subscriber that they will no longer
2263 * receive notifications.
2265 static void ice_irq_affinity_release(struct kref __always_unused *ref) {}
2268 * ice_vsi_ena_irq - Enable IRQ for the given VSI
2269 * @vsi: the VSI being configured
2271 static int ice_vsi_ena_irq(struct ice_vsi *vsi)
2273 struct ice_hw *hw = &vsi->back->hw;
2276 ice_for_each_q_vector(vsi, i)
2277 ice_irq_dynamic_ena(hw, vsi, vsi->q_vectors[i]);
2284 * ice_vsi_req_irq_msix - get MSI-X vectors from the OS for the VSI
2285 * @vsi: the VSI being configured
2286 * @basename: name for the vector
2288 static int ice_vsi_req_irq_msix(struct ice_vsi *vsi, char *basename)
2290 int q_vectors = vsi->num_q_vectors;
2291 struct ice_pf *pf = vsi->back;
2292 int base = vsi->base_vector;
2299 dev = ice_pf_to_dev(pf);
2300 for (vector = 0; vector < q_vectors; vector++) {
2301 struct ice_q_vector *q_vector = vsi->q_vectors[vector];
2303 irq_num = pf->msix_entries[base + vector].vector;
2305 if (q_vector->tx.ring && q_vector->rx.ring) {
2306 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2307 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2309 } else if (q_vector->rx.ring) {
2310 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2311 "%s-%s-%d", basename, "rx", rx_int_idx++);
2312 } else if (q_vector->tx.ring) {
2313 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2314 "%s-%s-%d", basename, "tx", tx_int_idx++);
2316 /* skip this unused q_vector */
2319 if (vsi->type == ICE_VSI_CTRL && vsi->vf_id != ICE_INVAL_VFID)
2320 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2321 IRQF_SHARED, q_vector->name,
2324 err = devm_request_irq(dev, irq_num, vsi->irq_handler,
2325 0, q_vector->name, q_vector);
2327 netdev_err(vsi->netdev, "MSIX request_irq failed, error: %d\n",
2332 /* register for affinity change notifications */
2333 if (!IS_ENABLED(CONFIG_RFS_ACCEL)) {
2334 struct irq_affinity_notify *affinity_notify;
2336 affinity_notify = &q_vector->affinity_notify;
2337 affinity_notify->notify = ice_irq_affinity_notify;
2338 affinity_notify->release = ice_irq_affinity_release;
2339 irq_set_affinity_notifier(irq_num, affinity_notify);
2342 /* assign the mask for this irq */
2343 irq_set_affinity_hint(irq_num, &q_vector->affinity_mask);
2346 vsi->irqs_ready = true;
2352 irq_num = pf->msix_entries[base + vector].vector;
2353 if (!IS_ENABLED(CONFIG_RFS_ACCEL))
2354 irq_set_affinity_notifier(irq_num, NULL);
2355 irq_set_affinity_hint(irq_num, NULL);
2356 devm_free_irq(dev, irq_num, &vsi->q_vectors[vector]);
2362 * ice_xdp_alloc_setup_rings - Allocate and setup Tx rings for XDP
2363 * @vsi: VSI to setup Tx rings used by XDP
2365 * Return 0 on success and negative value on error
2367 static int ice_xdp_alloc_setup_rings(struct ice_vsi *vsi)
2369 struct device *dev = ice_pf_to_dev(vsi->back);
2372 for (i = 0; i < vsi->num_xdp_txq; i++) {
2373 u16 xdp_q_idx = vsi->alloc_txq + i;
2374 struct ice_ring *xdp_ring;
2376 xdp_ring = kzalloc(sizeof(*xdp_ring), GFP_KERNEL);
2379 goto free_xdp_rings;
2381 xdp_ring->q_index = xdp_q_idx;
2382 xdp_ring->reg_idx = vsi->txq_map[xdp_q_idx];
2383 xdp_ring->ring_active = false;
2384 xdp_ring->vsi = vsi;
2385 xdp_ring->netdev = NULL;
2386 xdp_ring->dev = dev;
2387 xdp_ring->count = vsi->num_tx_desc;
2388 WRITE_ONCE(vsi->xdp_rings[i], xdp_ring);
2389 if (ice_setup_tx_ring(xdp_ring))
2390 goto free_xdp_rings;
2391 ice_set_ring_xdp(xdp_ring);
2392 xdp_ring->xsk_pool = ice_xsk_pool(xdp_ring);
2399 if (vsi->xdp_rings[i] && vsi->xdp_rings[i]->desc)
2400 ice_free_tx_ring(vsi->xdp_rings[i]);
2405 * ice_vsi_assign_bpf_prog - set or clear bpf prog pointer on VSI
2406 * @vsi: VSI to set the bpf prog on
2407 * @prog: the bpf prog pointer
2409 static void ice_vsi_assign_bpf_prog(struct ice_vsi *vsi, struct bpf_prog *prog)
2411 struct bpf_prog *old_prog;
2414 old_prog = xchg(&vsi->xdp_prog, prog);
2416 bpf_prog_put(old_prog);
2418 ice_for_each_rxq(vsi, i)
2419 WRITE_ONCE(vsi->rx_rings[i]->xdp_prog, vsi->xdp_prog);
2423 * ice_prepare_xdp_rings - Allocate, configure and setup Tx rings for XDP
2424 * @vsi: VSI to bring up Tx rings used by XDP
2425 * @prog: bpf program that will be assigned to VSI
2427 * Return 0 on success and negative value on error
2429 int ice_prepare_xdp_rings(struct ice_vsi *vsi, struct bpf_prog *prog)
2431 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2432 int xdp_rings_rem = vsi->num_xdp_txq;
2433 struct ice_pf *pf = vsi->back;
2434 struct ice_qs_cfg xdp_qs_cfg = {
2435 .qs_mutex = &pf->avail_q_mutex,
2436 .pf_map = pf->avail_txqs,
2437 .pf_map_size = pf->max_pf_txqs,
2438 .q_count = vsi->num_xdp_txq,
2439 .scatter_count = ICE_MAX_SCATTER_TXQS,
2440 .vsi_map = vsi->txq_map,
2441 .vsi_map_offset = vsi->alloc_txq,
2442 .mapping_mode = ICE_VSI_MAP_CONTIG
2444 enum ice_status status;
2448 dev = ice_pf_to_dev(pf);
2449 vsi->xdp_rings = devm_kcalloc(dev, vsi->num_xdp_txq,
2450 sizeof(*vsi->xdp_rings), GFP_KERNEL);
2451 if (!vsi->xdp_rings)
2454 vsi->xdp_mapping_mode = xdp_qs_cfg.mapping_mode;
2455 if (__ice_vsi_get_qs(&xdp_qs_cfg))
2458 if (ice_xdp_alloc_setup_rings(vsi))
2459 goto clear_xdp_rings;
2461 /* follow the logic from ice_vsi_map_rings_to_vectors */
2462 ice_for_each_q_vector(vsi, v_idx) {
2463 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2464 int xdp_rings_per_v, q_id, q_base;
2466 xdp_rings_per_v = DIV_ROUND_UP(xdp_rings_rem,
2467 vsi->num_q_vectors - v_idx);
2468 q_base = vsi->num_xdp_txq - xdp_rings_rem;
2470 for (q_id = q_base; q_id < (q_base + xdp_rings_per_v); q_id++) {
2471 struct ice_ring *xdp_ring = vsi->xdp_rings[q_id];
2473 xdp_ring->q_vector = q_vector;
2474 xdp_ring->next = q_vector->tx.ring;
2475 q_vector->tx.ring = xdp_ring;
2477 xdp_rings_rem -= xdp_rings_per_v;
2480 /* omit the scheduler update if in reset path; XDP queues will be
2481 * taken into account at the end of ice_vsi_rebuild, where
2482 * ice_cfg_vsi_lan is being called
2484 if (ice_is_reset_in_progress(pf->state))
2487 /* tell the Tx scheduler that right now we have
2490 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2491 max_txqs[i] = vsi->num_txq + vsi->num_xdp_txq;
2493 status = ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2496 dev_err(dev, "Failed VSI LAN queue config for XDP, error: %s\n",
2497 ice_stat_str(status));
2498 goto clear_xdp_rings;
2500 ice_vsi_assign_bpf_prog(vsi, prog);
2504 for (i = 0; i < vsi->num_xdp_txq; i++)
2505 if (vsi->xdp_rings[i]) {
2506 kfree_rcu(vsi->xdp_rings[i], rcu);
2507 vsi->xdp_rings[i] = NULL;
2511 mutex_lock(&pf->avail_q_mutex);
2512 for (i = 0; i < vsi->num_xdp_txq; i++) {
2513 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2514 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2516 mutex_unlock(&pf->avail_q_mutex);
2518 devm_kfree(dev, vsi->xdp_rings);
2523 * ice_destroy_xdp_rings - undo the configuration made by ice_prepare_xdp_rings
2524 * @vsi: VSI to remove XDP rings
2526 * Detach XDP rings from irq vectors, clean up the PF bitmap and free
2529 int ice_destroy_xdp_rings(struct ice_vsi *vsi)
2531 u16 max_txqs[ICE_MAX_TRAFFIC_CLASS] = { 0 };
2532 struct ice_pf *pf = vsi->back;
2535 /* q_vectors are freed in reset path so there's no point in detaching
2536 * rings; in case of rebuild being triggered not from reset bits
2537 * in pf->state won't be set, so additionally check first q_vector
2540 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2543 ice_for_each_q_vector(vsi, v_idx) {
2544 struct ice_q_vector *q_vector = vsi->q_vectors[v_idx];
2545 struct ice_ring *ring;
2547 ice_for_each_ring(ring, q_vector->tx)
2548 if (!ring->tx_buf || !ice_ring_is_xdp(ring))
2551 /* restore the value of last node prior to XDP setup */
2552 q_vector->tx.ring = ring;
2556 mutex_lock(&pf->avail_q_mutex);
2557 for (i = 0; i < vsi->num_xdp_txq; i++) {
2558 clear_bit(vsi->txq_map[i + vsi->alloc_txq], pf->avail_txqs);
2559 vsi->txq_map[i + vsi->alloc_txq] = ICE_INVAL_Q_INDEX;
2561 mutex_unlock(&pf->avail_q_mutex);
2563 for (i = 0; i < vsi->num_xdp_txq; i++)
2564 if (vsi->xdp_rings[i]) {
2565 if (vsi->xdp_rings[i]->desc)
2566 ice_free_tx_ring(vsi->xdp_rings[i]);
2567 kfree_rcu(vsi->xdp_rings[i], rcu);
2568 vsi->xdp_rings[i] = NULL;
2571 devm_kfree(ice_pf_to_dev(pf), vsi->xdp_rings);
2572 vsi->xdp_rings = NULL;
2574 if (ice_is_reset_in_progress(pf->state) || !vsi->q_vectors[0])
2577 ice_vsi_assign_bpf_prog(vsi, NULL);
2579 /* notify Tx scheduler that we destroyed XDP queues and bring
2580 * back the old number of child nodes
2582 for (i = 0; i < vsi->tc_cfg.numtc; i++)
2583 max_txqs[i] = vsi->num_txq;
2585 /* change number of XDP Tx queues to 0 */
2586 vsi->num_xdp_txq = 0;
2588 return ice_cfg_vsi_lan(vsi->port_info, vsi->idx, vsi->tc_cfg.ena_tc,
2593 * ice_vsi_rx_napi_schedule - Schedule napi on RX queues from VSI
2594 * @vsi: VSI to schedule napi on
2596 static void ice_vsi_rx_napi_schedule(struct ice_vsi *vsi)
2600 ice_for_each_rxq(vsi, i) {
2601 struct ice_ring *rx_ring = vsi->rx_rings[i];
2603 if (rx_ring->xsk_pool)
2604 napi_schedule(&rx_ring->q_vector->napi);
2609 * ice_xdp_setup_prog - Add or remove XDP eBPF program
2610 * @vsi: VSI to setup XDP for
2611 * @prog: XDP program
2612 * @extack: netlink extended ack
2615 ice_xdp_setup_prog(struct ice_vsi *vsi, struct bpf_prog *prog,
2616 struct netlink_ext_ack *extack)
2618 int frame_size = vsi->netdev->mtu + ICE_ETH_PKT_HDR_PAD;
2619 bool if_running = netif_running(vsi->netdev);
2620 int ret = 0, xdp_ring_err = 0;
2622 if (frame_size > vsi->rx_buf_len) {
2623 NL_SET_ERR_MSG_MOD(extack, "MTU too large for loading XDP");
2627 /* need to stop netdev while setting up the program for Rx rings */
2628 if (if_running && !test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
2629 ret = ice_down(vsi);
2631 NL_SET_ERR_MSG_MOD(extack, "Preparing device for XDP attach failed");
2636 if (!ice_is_xdp_ena_vsi(vsi) && prog) {
2637 vsi->num_xdp_txq = vsi->alloc_rxq;
2638 xdp_ring_err = ice_prepare_xdp_rings(vsi, prog);
2640 NL_SET_ERR_MSG_MOD(extack, "Setting up XDP Tx resources failed");
2641 } else if (ice_is_xdp_ena_vsi(vsi) && !prog) {
2642 xdp_ring_err = ice_destroy_xdp_rings(vsi);
2644 NL_SET_ERR_MSG_MOD(extack, "Freeing XDP Tx resources failed");
2646 ice_vsi_assign_bpf_prog(vsi, prog);
2653 ice_vsi_rx_napi_schedule(vsi);
2655 return (ret || xdp_ring_err) ? -ENOMEM : 0;
2659 * ice_xdp_safe_mode - XDP handler for safe mode
2663 static int ice_xdp_safe_mode(struct net_device __always_unused *dev,
2664 struct netdev_bpf *xdp)
2666 NL_SET_ERR_MSG_MOD(xdp->extack,
2667 "Please provide working DDP firmware package in order to use XDP\n"
2668 "Refer to Documentation/networking/device_drivers/ethernet/intel/ice.rst");
2673 * ice_xdp - implements XDP handler
2677 static int ice_xdp(struct net_device *dev, struct netdev_bpf *xdp)
2679 struct ice_netdev_priv *np = netdev_priv(dev);
2680 struct ice_vsi *vsi = np->vsi;
2682 if (vsi->type != ICE_VSI_PF) {
2683 NL_SET_ERR_MSG_MOD(xdp->extack, "XDP can be loaded only on PF VSI");
2687 switch (xdp->command) {
2688 case XDP_SETUP_PROG:
2689 return ice_xdp_setup_prog(vsi, xdp->prog, xdp->extack);
2690 case XDP_SETUP_XSK_POOL:
2691 return ice_xsk_pool_setup(vsi, xdp->xsk.pool,
2699 * ice_ena_misc_vector - enable the non-queue interrupts
2700 * @pf: board private structure
2702 static void ice_ena_misc_vector(struct ice_pf *pf)
2704 struct ice_hw *hw = &pf->hw;
2707 /* Disable anti-spoof detection interrupt to prevent spurious event
2708 * interrupts during a function reset. Anti-spoof functionally is
2711 val = rd32(hw, GL_MDCK_TX_TDPU);
2712 val |= GL_MDCK_TX_TDPU_RCU_ANTISPOOF_ITR_DIS_M;
2713 wr32(hw, GL_MDCK_TX_TDPU, val);
2715 /* clear things first */
2716 wr32(hw, PFINT_OICR_ENA, 0); /* disable all */
2717 rd32(hw, PFINT_OICR); /* read to clear */
2719 val = (PFINT_OICR_ECC_ERR_M |
2720 PFINT_OICR_MAL_DETECT_M |
2722 PFINT_OICR_PCI_EXCEPTION_M |
2724 PFINT_OICR_HMC_ERR_M |
2725 PFINT_OICR_PE_PUSH_M |
2726 PFINT_OICR_PE_CRITERR_M);
2728 wr32(hw, PFINT_OICR_ENA, val);
2730 /* SW_ITR_IDX = 0, but don't change INTENA */
2731 wr32(hw, GLINT_DYN_CTL(pf->oicr_idx),
2732 GLINT_DYN_CTL_SW_ITR_INDX_M | GLINT_DYN_CTL_INTENA_MSK_M);
2736 * ice_misc_intr - misc interrupt handler
2737 * @irq: interrupt number
2738 * @data: pointer to a q_vector
2740 static irqreturn_t ice_misc_intr(int __always_unused irq, void *data)
2742 struct ice_pf *pf = (struct ice_pf *)data;
2743 struct ice_hw *hw = &pf->hw;
2744 irqreturn_t ret = IRQ_NONE;
2748 dev = ice_pf_to_dev(pf);
2749 set_bit(ICE_ADMINQ_EVENT_PENDING, pf->state);
2750 set_bit(ICE_MAILBOXQ_EVENT_PENDING, pf->state);
2751 set_bit(ICE_SIDEBANDQ_EVENT_PENDING, pf->state);
2753 oicr = rd32(hw, PFINT_OICR);
2754 ena_mask = rd32(hw, PFINT_OICR_ENA);
2756 if (oicr & PFINT_OICR_SWINT_M) {
2757 ena_mask &= ~PFINT_OICR_SWINT_M;
2761 if (oicr & PFINT_OICR_MAL_DETECT_M) {
2762 ena_mask &= ~PFINT_OICR_MAL_DETECT_M;
2763 set_bit(ICE_MDD_EVENT_PENDING, pf->state);
2765 if (oicr & PFINT_OICR_VFLR_M) {
2766 /* disable any further VFLR event notifications */
2767 if (test_bit(ICE_VF_RESETS_DISABLED, pf->state)) {
2768 u32 reg = rd32(hw, PFINT_OICR_ENA);
2770 reg &= ~PFINT_OICR_VFLR_M;
2771 wr32(hw, PFINT_OICR_ENA, reg);
2773 ena_mask &= ~PFINT_OICR_VFLR_M;
2774 set_bit(ICE_VFLR_EVENT_PENDING, pf->state);
2778 if (oicr & PFINT_OICR_GRST_M) {
2781 /* we have a reset warning */
2782 ena_mask &= ~PFINT_OICR_GRST_M;
2783 reset = (rd32(hw, GLGEN_RSTAT) & GLGEN_RSTAT_RESET_TYPE_M) >>
2784 GLGEN_RSTAT_RESET_TYPE_S;
2786 if (reset == ICE_RESET_CORER)
2788 else if (reset == ICE_RESET_GLOBR)
2790 else if (reset == ICE_RESET_EMPR)
2793 dev_dbg(dev, "Invalid reset type %d\n", reset);
2795 /* If a reset cycle isn't already in progress, we set a bit in
2796 * pf->state so that the service task can start a reset/rebuild.
2798 if (!test_and_set_bit(ICE_RESET_OICR_RECV, pf->state)) {
2799 if (reset == ICE_RESET_CORER)
2800 set_bit(ICE_CORER_RECV, pf->state);
2801 else if (reset == ICE_RESET_GLOBR)
2802 set_bit(ICE_GLOBR_RECV, pf->state);
2804 set_bit(ICE_EMPR_RECV, pf->state);
2806 /* There are couple of different bits at play here.
2807 * hw->reset_ongoing indicates whether the hardware is
2808 * in reset. This is set to true when a reset interrupt
2809 * is received and set back to false after the driver
2810 * has determined that the hardware is out of reset.
2812 * ICE_RESET_OICR_RECV in pf->state indicates
2813 * that a post reset rebuild is required before the
2814 * driver is operational again. This is set above.
2816 * As this is the start of the reset/rebuild cycle, set
2817 * both to indicate that.
2819 hw->reset_ongoing = true;
2823 if (oicr & PFINT_OICR_TSYN_TX_M) {
2824 ena_mask &= ~PFINT_OICR_TSYN_TX_M;
2825 ice_ptp_process_ts(pf);
2828 if (oicr & PFINT_OICR_TSYN_EVNT_M) {
2829 u8 tmr_idx = hw->func_caps.ts_func_info.tmr_index_owned;
2830 u32 gltsyn_stat = rd32(hw, GLTSYN_STAT(tmr_idx));
2832 /* Save EVENTs from GTSYN register */
2833 pf->ptp.ext_ts_irq |= gltsyn_stat & (GLTSYN_STAT_EVENT0_M |
2834 GLTSYN_STAT_EVENT1_M |
2835 GLTSYN_STAT_EVENT2_M);
2836 ena_mask &= ~PFINT_OICR_TSYN_EVNT_M;
2837 kthread_queue_work(pf->ptp.kworker, &pf->ptp.extts_work);
2840 #define ICE_AUX_CRIT_ERR (PFINT_OICR_PE_CRITERR_M | PFINT_OICR_HMC_ERR_M | PFINT_OICR_PE_PUSH_M)
2841 if (oicr & ICE_AUX_CRIT_ERR) {
2842 struct iidc_event *event;
2844 ena_mask &= ~ICE_AUX_CRIT_ERR;
2845 event = kzalloc(sizeof(*event), GFP_KERNEL);
2847 set_bit(IIDC_EVENT_CRIT_ERR, event->type);
2848 /* report the entire OICR value to AUX driver */
2850 ice_send_event_to_aux(pf, event);
2855 /* Report any remaining unexpected interrupts */
2858 dev_dbg(dev, "unhandled interrupt oicr=0x%08x\n", oicr);
2859 /* If a critical error is pending there is no choice but to
2862 if (oicr & (PFINT_OICR_PCI_EXCEPTION_M |
2863 PFINT_OICR_ECC_ERR_M)) {
2864 set_bit(ICE_PFR_REQ, pf->state);
2865 ice_service_task_schedule(pf);
2870 ice_service_task_schedule(pf);
2871 ice_irq_dynamic_ena(hw, NULL, NULL);
2877 * ice_dis_ctrlq_interrupts - disable control queue interrupts
2878 * @hw: pointer to HW structure
2880 static void ice_dis_ctrlq_interrupts(struct ice_hw *hw)
2882 /* disable Admin queue Interrupt causes */
2883 wr32(hw, PFINT_FW_CTL,
2884 rd32(hw, PFINT_FW_CTL) & ~PFINT_FW_CTL_CAUSE_ENA_M);
2886 /* disable Mailbox queue Interrupt causes */
2887 wr32(hw, PFINT_MBX_CTL,
2888 rd32(hw, PFINT_MBX_CTL) & ~PFINT_MBX_CTL_CAUSE_ENA_M);
2890 wr32(hw, PFINT_SB_CTL,
2891 rd32(hw, PFINT_SB_CTL) & ~PFINT_SB_CTL_CAUSE_ENA_M);
2893 /* disable Control queue Interrupt causes */
2894 wr32(hw, PFINT_OICR_CTL,
2895 rd32(hw, PFINT_OICR_CTL) & ~PFINT_OICR_CTL_CAUSE_ENA_M);
2901 * ice_free_irq_msix_misc - Unroll misc vector setup
2902 * @pf: board private structure
2904 static void ice_free_irq_msix_misc(struct ice_pf *pf)
2906 struct ice_hw *hw = &pf->hw;
2908 ice_dis_ctrlq_interrupts(hw);
2910 /* disable OICR interrupt */
2911 wr32(hw, PFINT_OICR_ENA, 0);
2914 if (pf->msix_entries) {
2915 synchronize_irq(pf->msix_entries[pf->oicr_idx].vector);
2916 devm_free_irq(ice_pf_to_dev(pf),
2917 pf->msix_entries[pf->oicr_idx].vector, pf);
2920 pf->num_avail_sw_msix += 1;
2921 ice_free_res(pf->irq_tracker, pf->oicr_idx, ICE_RES_MISC_VEC_ID);
2925 * ice_ena_ctrlq_interrupts - enable control queue interrupts
2926 * @hw: pointer to HW structure
2927 * @reg_idx: HW vector index to associate the control queue interrupts with
2929 static void ice_ena_ctrlq_interrupts(struct ice_hw *hw, u16 reg_idx)
2933 val = ((reg_idx & PFINT_OICR_CTL_MSIX_INDX_M) |
2934 PFINT_OICR_CTL_CAUSE_ENA_M);
2935 wr32(hw, PFINT_OICR_CTL, val);
2937 /* enable Admin queue Interrupt causes */
2938 val = ((reg_idx & PFINT_FW_CTL_MSIX_INDX_M) |
2939 PFINT_FW_CTL_CAUSE_ENA_M);
2940 wr32(hw, PFINT_FW_CTL, val);
2942 /* enable Mailbox queue Interrupt causes */
2943 val = ((reg_idx & PFINT_MBX_CTL_MSIX_INDX_M) |
2944 PFINT_MBX_CTL_CAUSE_ENA_M);
2945 wr32(hw, PFINT_MBX_CTL, val);
2947 /* This enables Sideband queue Interrupt causes */
2948 val = ((reg_idx & PFINT_SB_CTL_MSIX_INDX_M) |
2949 PFINT_SB_CTL_CAUSE_ENA_M);
2950 wr32(hw, PFINT_SB_CTL, val);
2956 * ice_req_irq_msix_misc - Setup the misc vector to handle non queue events
2957 * @pf: board private structure
2959 * This sets up the handler for MSIX 0, which is used to manage the
2960 * non-queue interrupts, e.g. AdminQ and errors. This is not used
2961 * when in MSI or Legacy interrupt mode.
2963 static int ice_req_irq_msix_misc(struct ice_pf *pf)
2965 struct device *dev = ice_pf_to_dev(pf);
2966 struct ice_hw *hw = &pf->hw;
2967 int oicr_idx, err = 0;
2969 if (!pf->int_name[0])
2970 snprintf(pf->int_name, sizeof(pf->int_name) - 1, "%s-%s:misc",
2971 dev_driver_string(dev), dev_name(dev));
2973 /* Do not request IRQ but do enable OICR interrupt since settings are
2974 * lost during reset. Note that this function is called only during
2975 * rebuild path and not while reset is in progress.
2977 if (ice_is_reset_in_progress(pf->state))
2980 /* reserve one vector in irq_tracker for misc interrupts */
2981 oicr_idx = ice_get_res(pf, pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2985 pf->num_avail_sw_msix -= 1;
2986 pf->oicr_idx = (u16)oicr_idx;
2988 err = devm_request_irq(dev, pf->msix_entries[pf->oicr_idx].vector,
2989 ice_misc_intr, 0, pf->int_name, pf);
2991 dev_err(dev, "devm_request_irq for %s failed: %d\n",
2993 ice_free_res(pf->irq_tracker, 1, ICE_RES_MISC_VEC_ID);
2994 pf->num_avail_sw_msix += 1;
2999 ice_ena_misc_vector(pf);
3001 ice_ena_ctrlq_interrupts(hw, pf->oicr_idx);
3002 wr32(hw, GLINT_ITR(ICE_RX_ITR, pf->oicr_idx),
3003 ITR_REG_ALIGN(ICE_ITR_8K) >> ICE_ITR_GRAN_S);
3006 ice_irq_dynamic_ena(hw, NULL, NULL);
3012 * ice_napi_add - register NAPI handler for the VSI
3013 * @vsi: VSI for which NAPI handler is to be registered
3015 * This function is only called in the driver's load path. Registering the NAPI
3016 * handler is done in ice_vsi_alloc_q_vector() for all other cases (i.e. resume,
3017 * reset/rebuild, etc.)
3019 static void ice_napi_add(struct ice_vsi *vsi)
3026 ice_for_each_q_vector(vsi, v_idx)
3027 netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx]->napi,
3028 ice_napi_poll, NAPI_POLL_WEIGHT);
3032 * ice_set_ops - set netdev and ethtools ops for the given netdev
3033 * @netdev: netdev instance
3035 static void ice_set_ops(struct net_device *netdev)
3037 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3039 if (ice_is_safe_mode(pf)) {
3040 netdev->netdev_ops = &ice_netdev_safe_mode_ops;
3041 ice_set_ethtool_safe_mode_ops(netdev);
3045 netdev->netdev_ops = &ice_netdev_ops;
3046 netdev->udp_tunnel_nic_info = &pf->hw.udp_tunnel_nic;
3047 ice_set_ethtool_ops(netdev);
3051 * ice_set_netdev_features - set features for the given netdev
3052 * @netdev: netdev instance
3054 static void ice_set_netdev_features(struct net_device *netdev)
3056 struct ice_pf *pf = ice_netdev_to_pf(netdev);
3057 netdev_features_t csumo_features;
3058 netdev_features_t vlano_features;
3059 netdev_features_t dflt_features;
3060 netdev_features_t tso_features;
3062 if (ice_is_safe_mode(pf)) {
3064 netdev->features = NETIF_F_SG | NETIF_F_HIGHDMA;
3065 netdev->hw_features = netdev->features;
3069 dflt_features = NETIF_F_SG |
3074 csumo_features = NETIF_F_RXCSUM |
3079 vlano_features = NETIF_F_HW_VLAN_CTAG_FILTER |
3080 NETIF_F_HW_VLAN_CTAG_TX |
3081 NETIF_F_HW_VLAN_CTAG_RX;
3083 tso_features = NETIF_F_TSO |
3087 NETIF_F_GSO_UDP_TUNNEL |
3088 NETIF_F_GSO_GRE_CSUM |
3089 NETIF_F_GSO_UDP_TUNNEL_CSUM |
3090 NETIF_F_GSO_PARTIAL |
3091 NETIF_F_GSO_IPXIP4 |
3092 NETIF_F_GSO_IPXIP6 |
3095 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM |
3096 NETIF_F_GSO_GRE_CSUM;
3097 /* set features that user can change */
3098 netdev->hw_features = dflt_features | csumo_features |
3099 vlano_features | tso_features;
3101 /* add support for HW_CSUM on packets with MPLS header */
3102 netdev->mpls_features = NETIF_F_HW_CSUM;
3104 /* enable features */
3105 netdev->features |= netdev->hw_features;
3106 /* encap and VLAN devices inherit default, csumo and tso features */
3107 netdev->hw_enc_features |= dflt_features | csumo_features |
3109 netdev->vlan_features |= dflt_features | csumo_features |
3114 * ice_cfg_netdev - Allocate, configure and register a netdev
3115 * @vsi: the VSI associated with the new netdev
3117 * Returns 0 on success, negative value on failure
3119 static int ice_cfg_netdev(struct ice_vsi *vsi)
3121 struct ice_netdev_priv *np;
3122 struct net_device *netdev;
3123 u8 mac_addr[ETH_ALEN];
3125 netdev = alloc_etherdev_mqs(sizeof(*np), vsi->alloc_txq,
3130 set_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3131 vsi->netdev = netdev;
3132 np = netdev_priv(netdev);
3135 ice_set_netdev_features(netdev);
3137 ice_set_ops(netdev);
3139 if (vsi->type == ICE_VSI_PF) {
3140 SET_NETDEV_DEV(netdev, ice_pf_to_dev(vsi->back));
3141 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
3142 ether_addr_copy(netdev->dev_addr, mac_addr);
3143 ether_addr_copy(netdev->perm_addr, mac_addr);
3146 netdev->priv_flags |= IFF_UNICAST_FLT;
3148 /* Setup netdev TC information */
3149 ice_vsi_cfg_netdev_tc(vsi, vsi->tc_cfg.ena_tc);
3151 /* setup watchdog timeout value to be 5 second */
3152 netdev->watchdog_timeo = 5 * HZ;
3154 netdev->min_mtu = ETH_MIN_MTU;
3155 netdev->max_mtu = ICE_MAX_MTU;
3161 * ice_fill_rss_lut - Fill the RSS lookup table with default values
3162 * @lut: Lookup table
3163 * @rss_table_size: Lookup table size
3164 * @rss_size: Range of queue number for hashing
3166 void ice_fill_rss_lut(u8 *lut, u16 rss_table_size, u16 rss_size)
3170 for (i = 0; i < rss_table_size; i++)
3171 lut[i] = i % rss_size;
3175 * ice_pf_vsi_setup - Set up a PF VSI
3176 * @pf: board private structure
3177 * @pi: pointer to the port_info instance
3179 * Returns pointer to the successfully allocated VSI software struct
3180 * on success, otherwise returns NULL on failure.
3182 static struct ice_vsi *
3183 ice_pf_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3185 return ice_vsi_setup(pf, pi, ICE_VSI_PF, ICE_INVAL_VFID);
3189 * ice_ctrl_vsi_setup - Set up a control VSI
3190 * @pf: board private structure
3191 * @pi: pointer to the port_info instance
3193 * Returns pointer to the successfully allocated VSI software struct
3194 * on success, otherwise returns NULL on failure.
3196 static struct ice_vsi *
3197 ice_ctrl_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3199 return ice_vsi_setup(pf, pi, ICE_VSI_CTRL, ICE_INVAL_VFID);
3203 * ice_lb_vsi_setup - Set up a loopback VSI
3204 * @pf: board private structure
3205 * @pi: pointer to the port_info instance
3207 * Returns pointer to the successfully allocated VSI software struct
3208 * on success, otherwise returns NULL on failure.
3211 ice_lb_vsi_setup(struct ice_pf *pf, struct ice_port_info *pi)
3213 return ice_vsi_setup(pf, pi, ICE_VSI_LB, ICE_INVAL_VFID);
3217 * ice_vlan_rx_add_vid - Add a VLAN ID filter to HW offload
3218 * @netdev: network interface to be adjusted
3219 * @proto: unused protocol
3220 * @vid: VLAN ID to be added
3222 * net_device_ops implementation for adding VLAN IDs
3225 ice_vlan_rx_add_vid(struct net_device *netdev, __always_unused __be16 proto,
3228 struct ice_netdev_priv *np = netdev_priv(netdev);
3229 struct ice_vsi *vsi = np->vsi;
3232 /* VLAN 0 is added by default during load/reset */
3236 /* Enable VLAN pruning when a VLAN other than 0 is added */
3237 if (!ice_vsi_is_vlan_pruning_ena(vsi)) {
3238 ret = ice_cfg_vlan_pruning(vsi, true, false);
3243 /* Add a switch rule for this VLAN ID so its corresponding VLAN tagged
3244 * packets aren't pruned by the device's internal switch on Rx
3246 ret = ice_vsi_add_vlan(vsi, vid, ICE_FWD_TO_VSI);
3248 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3254 * ice_vlan_rx_kill_vid - Remove a VLAN ID filter from HW offload
3255 * @netdev: network interface to be adjusted
3256 * @proto: unused protocol
3257 * @vid: VLAN ID to be removed
3259 * net_device_ops implementation for removing VLAN IDs
3262 ice_vlan_rx_kill_vid(struct net_device *netdev, __always_unused __be16 proto,
3265 struct ice_netdev_priv *np = netdev_priv(netdev);
3266 struct ice_vsi *vsi = np->vsi;
3269 /* don't allow removal of VLAN 0 */
3273 /* Make sure ice_vsi_kill_vlan is successful before updating VLAN
3276 ret = ice_vsi_kill_vlan(vsi, vid);
3280 /* Disable pruning when VLAN 0 is the only VLAN rule */
3281 if (vsi->num_vlan == 1 && ice_vsi_is_vlan_pruning_ena(vsi))
3282 ret = ice_cfg_vlan_pruning(vsi, false, false);
3284 set_bit(ICE_VSI_VLAN_FLTR_CHANGED, vsi->state);
3289 * ice_setup_pf_sw - Setup the HW switch on startup or after reset
3290 * @pf: board private structure
3292 * Returns 0 on success, negative value on failure
3294 static int ice_setup_pf_sw(struct ice_pf *pf)
3296 struct ice_vsi *vsi;
3299 if (ice_is_reset_in_progress(pf->state))
3302 vsi = ice_pf_vsi_setup(pf, pf->hw.port_info);
3306 status = ice_cfg_netdev(vsi);
3309 goto unroll_vsi_setup;
3311 /* netdev has to be configured before setting frame size */
3312 ice_vsi_cfg_frame_size(vsi);
3314 /* Setup DCB netlink interface */
3315 ice_dcbnl_setup(vsi);
3317 /* registering the NAPI handler requires both the queues and
3318 * netdev to be created, which are done in ice_pf_vsi_setup()
3319 * and ice_cfg_netdev() respectively
3323 status = ice_set_cpu_rx_rmap(vsi);
3325 dev_err(ice_pf_to_dev(pf), "Failed to set CPU Rx map VSI %d error %d\n",
3326 vsi->vsi_num, status);
3328 goto unroll_napi_add;
3330 status = ice_init_mac_fltr(pf);
3332 goto free_cpu_rx_map;
3337 ice_free_cpu_rx_rmap(vsi);
3343 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
3344 free_netdev(vsi->netdev);
3350 ice_vsi_release(vsi);
3355 * ice_get_avail_q_count - Get count of queues in use
3356 * @pf_qmap: bitmap to get queue use count from
3357 * @lock: pointer to a mutex that protects access to pf_qmap
3358 * @size: size of the bitmap
3361 ice_get_avail_q_count(unsigned long *pf_qmap, struct mutex *lock, u16 size)
3367 for_each_clear_bit(bit, pf_qmap, size)
3375 * ice_get_avail_txq_count - Get count of Tx queues in use
3376 * @pf: pointer to an ice_pf instance
3378 u16 ice_get_avail_txq_count(struct ice_pf *pf)
3380 return ice_get_avail_q_count(pf->avail_txqs, &pf->avail_q_mutex,
3385 * ice_get_avail_rxq_count - Get count of Rx queues in use
3386 * @pf: pointer to an ice_pf instance
3388 u16 ice_get_avail_rxq_count(struct ice_pf *pf)
3390 return ice_get_avail_q_count(pf->avail_rxqs, &pf->avail_q_mutex,
3395 * ice_deinit_pf - Unrolls initialziations done by ice_init_pf
3396 * @pf: board private structure to initialize
3398 static void ice_deinit_pf(struct ice_pf *pf)
3400 ice_service_task_stop(pf);
3401 mutex_destroy(&pf->sw_mutex);
3402 mutex_destroy(&pf->tc_mutex);
3403 mutex_destroy(&pf->avail_q_mutex);
3405 if (pf->avail_txqs) {
3406 bitmap_free(pf->avail_txqs);
3407 pf->avail_txqs = NULL;
3410 if (pf->avail_rxqs) {
3411 bitmap_free(pf->avail_rxqs);
3412 pf->avail_rxqs = NULL;
3416 ptp_clock_unregister(pf->ptp.clock);
3420 * ice_set_pf_caps - set PFs capability flags
3421 * @pf: pointer to the PF instance
3423 static void ice_set_pf_caps(struct ice_pf *pf)
3425 struct ice_hw_func_caps *func_caps = &pf->hw.func_caps;
3427 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3428 clear_bit(ICE_FLAG_AUX_ENA, pf->flags);
3429 if (func_caps->common_cap.rdma) {
3430 set_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3431 set_bit(ICE_FLAG_AUX_ENA, pf->flags);
3433 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3434 if (func_caps->common_cap.dcb)
3435 set_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
3436 clear_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3437 if (func_caps->common_cap.sr_iov_1_1) {
3438 set_bit(ICE_FLAG_SRIOV_CAPABLE, pf->flags);
3439 pf->num_vfs_supported = min_t(int, func_caps->num_allocd_vfs,
3442 clear_bit(ICE_FLAG_RSS_ENA, pf->flags);
3443 if (func_caps->common_cap.rss_table_size)
3444 set_bit(ICE_FLAG_RSS_ENA, pf->flags);
3446 clear_bit(ICE_FLAG_FD_ENA, pf->flags);
3447 if (func_caps->fd_fltr_guar > 0 || func_caps->fd_fltr_best_effort > 0) {
3450 /* ctrl_vsi_idx will be set to a valid value when flow director
3451 * is setup by ice_init_fdir
3453 pf->ctrl_vsi_idx = ICE_NO_VSI;
3454 set_bit(ICE_FLAG_FD_ENA, pf->flags);
3455 /* force guaranteed filter pool for PF */
3456 ice_alloc_fd_guar_item(&pf->hw, &unused,
3457 func_caps->fd_fltr_guar);
3458 /* force shared filter pool for PF */
3459 ice_alloc_fd_shrd_item(&pf->hw, &unused,
3460 func_caps->fd_fltr_best_effort);
3463 clear_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3464 if (func_caps->common_cap.ieee_1588)
3465 set_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags);
3467 pf->max_pf_txqs = func_caps->common_cap.num_txq;
3468 pf->max_pf_rxqs = func_caps->common_cap.num_rxq;
3472 * ice_init_pf - Initialize general software structures (struct ice_pf)
3473 * @pf: board private structure to initialize
3475 static int ice_init_pf(struct ice_pf *pf)
3477 ice_set_pf_caps(pf);
3479 mutex_init(&pf->sw_mutex);
3480 mutex_init(&pf->tc_mutex);
3482 INIT_HLIST_HEAD(&pf->aq_wait_list);
3483 spin_lock_init(&pf->aq_wait_lock);
3484 init_waitqueue_head(&pf->aq_wait_queue);
3486 init_waitqueue_head(&pf->reset_wait_queue);
3488 /* setup service timer and periodic service task */
3489 timer_setup(&pf->serv_tmr, ice_service_timer, 0);
3490 pf->serv_tmr_period = HZ;
3491 INIT_WORK(&pf->serv_task, ice_service_task);
3492 clear_bit(ICE_SERVICE_SCHED, pf->state);
3494 mutex_init(&pf->avail_q_mutex);
3495 pf->avail_txqs = bitmap_zalloc(pf->max_pf_txqs, GFP_KERNEL);
3496 if (!pf->avail_txqs)
3499 pf->avail_rxqs = bitmap_zalloc(pf->max_pf_rxqs, GFP_KERNEL);
3500 if (!pf->avail_rxqs) {
3501 devm_kfree(ice_pf_to_dev(pf), pf->avail_txqs);
3502 pf->avail_txqs = NULL;
3510 * ice_ena_msix_range - Request a range of MSIX vectors from the OS
3511 * @pf: board private structure
3513 * compute the number of MSIX vectors required (v_budget) and request from
3514 * the OS. Return the number of vectors reserved or negative on failure
3516 static int ice_ena_msix_range(struct ice_pf *pf)
3518 int num_cpus, v_left, v_actual, v_other, v_budget = 0;
3519 struct device *dev = ice_pf_to_dev(pf);
3522 v_left = pf->hw.func_caps.common_cap.num_msix_vectors;
3523 num_cpus = num_online_cpus();
3525 /* reserve for LAN miscellaneous handler */
3526 needed = ICE_MIN_LAN_OICR_MSIX;
3527 if (v_left < needed)
3528 goto no_hw_vecs_left_err;
3532 /* reserve for flow director */
3533 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
3534 needed = ICE_FDIR_MSIX;
3535 if (v_left < needed)
3536 goto no_hw_vecs_left_err;
3541 /* total used for non-traffic vectors */
3544 /* reserve vectors for LAN traffic */
3546 if (v_left < needed)
3547 goto no_hw_vecs_left_err;
3548 pf->num_lan_msix = needed;
3552 /* reserve vectors for RDMA auxiliary driver */
3553 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3554 needed = num_cpus + ICE_RDMA_NUM_AEQ_MSIX;
3555 if (v_left < needed)
3556 goto no_hw_vecs_left_err;
3557 pf->num_rdma_msix = needed;
3562 pf->msix_entries = devm_kcalloc(dev, v_budget,
3563 sizeof(*pf->msix_entries), GFP_KERNEL);
3564 if (!pf->msix_entries) {
3569 for (i = 0; i < v_budget; i++)
3570 pf->msix_entries[i].entry = i;
3572 /* actually reserve the vectors */
3573 v_actual = pci_enable_msix_range(pf->pdev, pf->msix_entries,
3574 ICE_MIN_MSIX, v_budget);
3576 dev_err(dev, "unable to reserve MSI-X vectors\n");
3581 if (v_actual < v_budget) {
3582 dev_warn(dev, "not enough OS MSI-X vectors. requested = %d, obtained = %d\n",
3583 v_budget, v_actual);
3585 if (v_actual < ICE_MIN_MSIX) {
3586 /* error if we can't get minimum vectors */
3587 pci_disable_msix(pf->pdev);
3591 int v_remain = v_actual - v_other;
3592 int v_rdma = 0, v_min_rdma = 0;
3594 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags)) {
3595 /* Need at least 1 interrupt in addition to
3598 v_rdma = ICE_RDMA_NUM_AEQ_MSIX + 1;
3599 v_min_rdma = ICE_MIN_RDMA_MSIX;
3602 if (v_actual == ICE_MIN_MSIX ||
3603 v_remain < ICE_MIN_LAN_TXRX_MSIX + v_min_rdma) {
3604 dev_warn(dev, "Not enough MSI-X vectors to support RDMA.\n");
3605 clear_bit(ICE_FLAG_RDMA_ENA, pf->flags);
3607 pf->num_rdma_msix = 0;
3608 pf->num_lan_msix = ICE_MIN_LAN_TXRX_MSIX;
3609 } else if ((v_remain < ICE_MIN_LAN_TXRX_MSIX + v_rdma) ||
3610 (v_remain - v_rdma < v_rdma)) {
3611 /* Support minimum RDMA and give remaining
3612 * vectors to LAN MSIX
3614 pf->num_rdma_msix = v_min_rdma;
3615 pf->num_lan_msix = v_remain - v_min_rdma;
3617 /* Split remaining MSIX with RDMA after
3618 * accounting for AEQ MSIX
3620 pf->num_rdma_msix = (v_remain - ICE_RDMA_NUM_AEQ_MSIX) / 2 +
3621 ICE_RDMA_NUM_AEQ_MSIX;
3622 pf->num_lan_msix = v_remain - pf->num_rdma_msix;
3625 dev_notice(dev, "Enabled %d MSI-X vectors for LAN traffic.\n",
3628 if (test_bit(ICE_FLAG_RDMA_ENA, pf->flags))
3629 dev_notice(dev, "Enabled %d MSI-X vectors for RDMA.\n",
3637 devm_kfree(dev, pf->msix_entries);
3640 no_hw_vecs_left_err:
3641 dev_err(dev, "not enough device MSI-X vectors. requested = %d, available = %d\n",
3645 pf->num_rdma_msix = 0;
3646 pf->num_lan_msix = 0;
3651 * ice_dis_msix - Disable MSI-X interrupt setup in OS
3652 * @pf: board private structure
3654 static void ice_dis_msix(struct ice_pf *pf)
3656 pci_disable_msix(pf->pdev);
3657 devm_kfree(ice_pf_to_dev(pf), pf->msix_entries);
3658 pf->msix_entries = NULL;
3662 * ice_clear_interrupt_scheme - Undo things done by ice_init_interrupt_scheme
3663 * @pf: board private structure
3665 static void ice_clear_interrupt_scheme(struct ice_pf *pf)
3669 if (pf->irq_tracker) {
3670 devm_kfree(ice_pf_to_dev(pf), pf->irq_tracker);
3671 pf->irq_tracker = NULL;
3676 * ice_init_interrupt_scheme - Determine proper interrupt scheme
3677 * @pf: board private structure to initialize
3679 static int ice_init_interrupt_scheme(struct ice_pf *pf)
3683 vectors = ice_ena_msix_range(pf);
3688 /* set up vector assignment tracking */
3689 pf->irq_tracker = devm_kzalloc(ice_pf_to_dev(pf),
3690 struct_size(pf->irq_tracker, list, vectors),
3692 if (!pf->irq_tracker) {
3697 /* populate SW interrupts pool with number of OS granted IRQs. */
3698 pf->num_avail_sw_msix = (u16)vectors;
3699 pf->irq_tracker->num_entries = (u16)vectors;
3700 pf->irq_tracker->end = pf->irq_tracker->num_entries;
3706 * ice_is_wol_supported - check if WoL is supported
3707 * @hw: pointer to hardware info
3709 * Check if WoL is supported based on the HW configuration.
3710 * Returns true if NVM supports and enables WoL for this port, false otherwise
3712 bool ice_is_wol_supported(struct ice_hw *hw)
3716 /* A bit set to 1 in the NVM Software Reserved Word 2 (WoL control
3717 * word) indicates WoL is not supported on the corresponding PF ID.
3719 if (ice_read_sr_word(hw, ICE_SR_NVM_WOL_CFG, &wol_ctrl))
3722 return !(BIT(hw->port_info->lport) & wol_ctrl);
3726 * ice_vsi_recfg_qs - Change the number of queues on a VSI
3727 * @vsi: VSI being changed
3728 * @new_rx: new number of Rx queues
3729 * @new_tx: new number of Tx queues
3731 * Only change the number of queues if new_tx, or new_rx is non-0.
3733 * Returns 0 on success.
3735 int ice_vsi_recfg_qs(struct ice_vsi *vsi, int new_rx, int new_tx)
3737 struct ice_pf *pf = vsi->back;
3738 int err = 0, timeout = 50;
3740 if (!new_rx && !new_tx)
3743 while (test_and_set_bit(ICE_CFG_BUSY, pf->state)) {
3747 usleep_range(1000, 2000);
3751 vsi->req_txq = (u16)new_tx;
3753 vsi->req_rxq = (u16)new_rx;
3755 /* set for the next time the netdev is started */
3756 if (!netif_running(vsi->netdev)) {
3757 ice_vsi_rebuild(vsi, false);
3758 dev_dbg(ice_pf_to_dev(pf), "Link is down, queue count change happens when link is brought up\n");
3763 ice_vsi_rebuild(vsi, false);
3764 ice_pf_dcb_recfg(pf);
3767 clear_bit(ICE_CFG_BUSY, pf->state);
3772 * ice_set_safe_mode_vlan_cfg - configure PF VSI to allow all VLANs in safe mode
3773 * @pf: PF to configure
3775 * No VLAN offloads/filtering are advertised in safe mode so make sure the PF
3776 * VSI can still Tx/Rx VLAN tagged packets.
3778 static void ice_set_safe_mode_vlan_cfg(struct ice_pf *pf)
3780 struct ice_vsi *vsi = ice_get_main_vsi(pf);
3781 struct ice_vsi_ctx *ctxt;
3782 enum ice_status status;
3788 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
3793 ctxt->info = vsi->info;
3795 ctxt->info.valid_sections =
3796 cpu_to_le16(ICE_AQ_VSI_PROP_VLAN_VALID |
3797 ICE_AQ_VSI_PROP_SECURITY_VALID |
3798 ICE_AQ_VSI_PROP_SW_VALID);
3800 /* disable VLAN anti-spoof */
3801 ctxt->info.sec_flags &= ~(ICE_AQ_VSI_SEC_TX_VLAN_PRUNE_ENA <<
3802 ICE_AQ_VSI_SEC_TX_PRUNE_ENA_S);
3804 /* disable VLAN pruning and keep all other settings */
3805 ctxt->info.sw_flags2 &= ~ICE_AQ_VSI_SW_FLAG_RX_VLAN_PRUNE_ENA;
3807 /* allow all VLANs on Tx and don't strip on Rx */
3808 ctxt->info.vlan_flags = ICE_AQ_VSI_VLAN_MODE_ALL |
3809 ICE_AQ_VSI_VLAN_EMOD_NOTHING;
3811 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
3813 dev_err(ice_pf_to_dev(vsi->back), "Failed to update VSI for safe mode VLANs, err %s aq_err %s\n",
3814 ice_stat_str(status),
3815 ice_aq_str(hw->adminq.sq_last_status));
3817 vsi->info.sec_flags = ctxt->info.sec_flags;
3818 vsi->info.sw_flags2 = ctxt->info.sw_flags2;
3819 vsi->info.vlan_flags = ctxt->info.vlan_flags;
3826 * ice_log_pkg_init - log result of DDP package load
3827 * @hw: pointer to hardware info
3828 * @status: status of package load
3831 ice_log_pkg_init(struct ice_hw *hw, enum ice_status *status)
3833 struct ice_pf *pf = (struct ice_pf *)hw->back;
3834 struct device *dev = ice_pf_to_dev(pf);
3838 /* The package download AdminQ command returned success because
3839 * this download succeeded or ICE_ERR_AQ_NO_WORK since there is
3840 * already a package loaded on the device.
3842 if (hw->pkg_ver.major == hw->active_pkg_ver.major &&
3843 hw->pkg_ver.minor == hw->active_pkg_ver.minor &&
3844 hw->pkg_ver.update == hw->active_pkg_ver.update &&
3845 hw->pkg_ver.draft == hw->active_pkg_ver.draft &&
3846 !memcmp(hw->pkg_name, hw->active_pkg_name,
3847 sizeof(hw->pkg_name))) {
3848 if (hw->pkg_dwnld_status == ICE_AQ_RC_EEXIST)
3849 dev_info(dev, "DDP package already present on device: %s version %d.%d.%d.%d\n",
3850 hw->active_pkg_name,
3851 hw->active_pkg_ver.major,
3852 hw->active_pkg_ver.minor,
3853 hw->active_pkg_ver.update,
3854 hw->active_pkg_ver.draft);
3856 dev_info(dev, "The DDP package was successfully loaded: %s version %d.%d.%d.%d\n",
3857 hw->active_pkg_name,
3858 hw->active_pkg_ver.major,
3859 hw->active_pkg_ver.minor,
3860 hw->active_pkg_ver.update,
3861 hw->active_pkg_ver.draft);
3862 } else if (hw->active_pkg_ver.major != ICE_PKG_SUPP_VER_MAJ ||
3863 hw->active_pkg_ver.minor != ICE_PKG_SUPP_VER_MNR) {
3864 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",
3865 hw->active_pkg_name,
3866 hw->active_pkg_ver.major,
3867 hw->active_pkg_ver.minor,
3868 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3869 *status = ICE_ERR_NOT_SUPPORTED;
3870 } else if (hw->active_pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3871 hw->active_pkg_ver.minor == ICE_PKG_SUPP_VER_MNR) {
3872 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",
3873 hw->active_pkg_name,
3874 hw->active_pkg_ver.major,
3875 hw->active_pkg_ver.minor,
3876 hw->active_pkg_ver.update,
3877 hw->active_pkg_ver.draft,
3884 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");
3885 *status = ICE_ERR_NOT_SUPPORTED;
3888 case ICE_ERR_FW_DDP_MISMATCH:
3889 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");
3891 case ICE_ERR_BUF_TOO_SHORT:
3893 dev_err(dev, "The DDP package file is invalid. Entering Safe Mode.\n");
3895 case ICE_ERR_NOT_SUPPORTED:
3896 /* Package File version not supported */
3897 if (hw->pkg_ver.major > ICE_PKG_SUPP_VER_MAJ ||
3898 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3899 hw->pkg_ver.minor > ICE_PKG_SUPP_VER_MNR))
3900 dev_err(dev, "The DDP package file version is higher than the driver supports. Please use an updated driver. Entering Safe Mode.\n");
3901 else if (hw->pkg_ver.major < ICE_PKG_SUPP_VER_MAJ ||
3902 (hw->pkg_ver.major == ICE_PKG_SUPP_VER_MAJ &&
3903 hw->pkg_ver.minor < ICE_PKG_SUPP_VER_MNR))
3904 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",
3905 ICE_PKG_SUPP_VER_MAJ, ICE_PKG_SUPP_VER_MNR);
3907 case ICE_ERR_AQ_ERROR:
3908 switch (hw->pkg_dwnld_status) {
3909 case ICE_AQ_RC_ENOSEC:
3910 case ICE_AQ_RC_EBADSIG:
3911 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");
3913 case ICE_AQ_RC_ESVN:
3914 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");
3916 case ICE_AQ_RC_EBADMAN:
3917 case ICE_AQ_RC_EBADBUF:
3918 dev_err(dev, "An error occurred on the device while loading the DDP package. The device will be reset.\n");
3919 /* poll for reset to complete */
3920 if (ice_check_reset(hw))
3921 dev_err(dev, "Error resetting device. Please reload the driver\n");
3928 dev_err(dev, "An unknown error (%d) occurred when loading the DDP package. Entering Safe Mode.\n",
3935 * ice_load_pkg - load/reload the DDP Package file
3936 * @firmware: firmware structure when firmware requested or NULL for reload
3937 * @pf: pointer to the PF instance
3939 * Called on probe and post CORER/GLOBR rebuild to load DDP Package and
3940 * initialize HW tables.
3943 ice_load_pkg(const struct firmware *firmware, struct ice_pf *pf)
3945 enum ice_status status = ICE_ERR_PARAM;
3946 struct device *dev = ice_pf_to_dev(pf);
3947 struct ice_hw *hw = &pf->hw;
3949 /* Load DDP Package */
3950 if (firmware && !hw->pkg_copy) {
3951 status = ice_copy_and_init_pkg(hw, firmware->data,
3953 ice_log_pkg_init(hw, &status);
3954 } else if (!firmware && hw->pkg_copy) {
3955 /* Reload package during rebuild after CORER/GLOBR reset */
3956 status = ice_init_pkg(hw, hw->pkg_copy, hw->pkg_size);
3957 ice_log_pkg_init(hw, &status);
3959 dev_err(dev, "The DDP package file failed to load. Entering Safe Mode.\n");
3964 clear_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3968 /* Successful download package is the precondition for advanced
3969 * features, hence setting the ICE_FLAG_ADV_FEATURES flag
3971 set_bit(ICE_FLAG_ADV_FEATURES, pf->flags);
3975 * ice_verify_cacheline_size - verify driver's assumption of 64 Byte cache lines
3976 * @pf: pointer to the PF structure
3978 * There is no error returned here because the driver should be able to handle
3979 * 128 Byte cache lines, so we only print a warning in case issues are seen,
3980 * specifically with Tx.
3982 static void ice_verify_cacheline_size(struct ice_pf *pf)
3984 if (rd32(&pf->hw, GLPCI_CNF2) & GLPCI_CNF2_CACHELINE_SIZE_M)
3985 dev_warn(ice_pf_to_dev(pf), "%d Byte cache line assumption is invalid, driver may have Tx timeouts!\n",
3986 ICE_CACHE_LINE_BYTES);
3990 * ice_send_version - update firmware with driver version
3993 * Returns ICE_SUCCESS on success, else error code
3995 static enum ice_status ice_send_version(struct ice_pf *pf)
3997 struct ice_driver_ver dv;
3999 dv.major_ver = 0xff;
4000 dv.minor_ver = 0xff;
4001 dv.build_ver = 0xff;
4002 dv.subbuild_ver = 0;
4003 strscpy((char *)dv.driver_string, UTS_RELEASE,
4004 sizeof(dv.driver_string));
4005 return ice_aq_send_driver_ver(&pf->hw, &dv, NULL);
4009 * ice_init_fdir - Initialize flow director VSI and configuration
4010 * @pf: pointer to the PF instance
4012 * returns 0 on success, negative on error
4014 static int ice_init_fdir(struct ice_pf *pf)
4016 struct device *dev = ice_pf_to_dev(pf);
4017 struct ice_vsi *ctrl_vsi;
4020 /* Side Band Flow Director needs to have a control VSI.
4021 * Allocate it and store it in the PF.
4023 ctrl_vsi = ice_ctrl_vsi_setup(pf, pf->hw.port_info);
4025 dev_dbg(dev, "could not create control VSI\n");
4029 err = ice_vsi_open_ctrl(ctrl_vsi);
4031 dev_dbg(dev, "could not open control VSI\n");
4035 mutex_init(&pf->hw.fdir_fltr_lock);
4037 err = ice_fdir_create_dflt_rules(pf);
4044 ice_fdir_release_flows(&pf->hw);
4045 ice_vsi_close(ctrl_vsi);
4047 ice_vsi_release(ctrl_vsi);
4048 if (pf->ctrl_vsi_idx != ICE_NO_VSI) {
4049 pf->vsi[pf->ctrl_vsi_idx] = NULL;
4050 pf->ctrl_vsi_idx = ICE_NO_VSI;
4056 * ice_get_opt_fw_name - return optional firmware file name or NULL
4057 * @pf: pointer to the PF instance
4059 static char *ice_get_opt_fw_name(struct ice_pf *pf)
4061 /* Optional firmware name same as default with additional dash
4062 * followed by a EUI-64 identifier (PCIe Device Serial Number)
4064 struct pci_dev *pdev = pf->pdev;
4065 char *opt_fw_filename;
4068 /* Determine the name of the optional file using the DSN (two
4069 * dwords following the start of the DSN Capability).
4071 dsn = pci_get_dsn(pdev);
4075 opt_fw_filename = kzalloc(NAME_MAX, GFP_KERNEL);
4076 if (!opt_fw_filename)
4079 snprintf(opt_fw_filename, NAME_MAX, "%sice-%016llx.pkg",
4080 ICE_DDP_PKG_PATH, dsn);
4082 return opt_fw_filename;
4086 * ice_request_fw - Device initialization routine
4087 * @pf: pointer to the PF instance
4089 static void ice_request_fw(struct ice_pf *pf)
4091 char *opt_fw_filename = ice_get_opt_fw_name(pf);
4092 const struct firmware *firmware = NULL;
4093 struct device *dev = ice_pf_to_dev(pf);
4096 /* optional device-specific DDP (if present) overrides the default DDP
4097 * package file. kernel logs a debug message if the file doesn't exist,
4098 * and warning messages for other errors.
4100 if (opt_fw_filename) {
4101 err = firmware_request_nowarn(&firmware, opt_fw_filename, dev);
4103 kfree(opt_fw_filename);
4107 /* request for firmware was successful. Download to device */
4108 ice_load_pkg(firmware, pf);
4109 kfree(opt_fw_filename);
4110 release_firmware(firmware);
4115 err = request_firmware(&firmware, ICE_DDP_PKG_FILE, dev);
4117 dev_err(dev, "The DDP package file was not found or could not be read. Entering Safe Mode\n");
4121 /* request for firmware was successful. Download to device */
4122 ice_load_pkg(firmware, pf);
4123 release_firmware(firmware);
4127 * ice_print_wake_reason - show the wake up cause in the log
4128 * @pf: pointer to the PF struct
4130 static void ice_print_wake_reason(struct ice_pf *pf)
4132 u32 wus = pf->wakeup_reason;
4133 const char *wake_str;
4135 /* if no wake event, nothing to print */
4139 if (wus & PFPM_WUS_LNKC_M)
4140 wake_str = "Link\n";
4141 else if (wus & PFPM_WUS_MAG_M)
4142 wake_str = "Magic Packet\n";
4143 else if (wus & PFPM_WUS_MNG_M)
4144 wake_str = "Management\n";
4145 else if (wus & PFPM_WUS_FW_RST_WK_M)
4146 wake_str = "Firmware Reset\n";
4148 wake_str = "Unknown\n";
4150 dev_info(ice_pf_to_dev(pf), "Wake reason: %s", wake_str);
4154 * ice_register_netdev - register netdev and devlink port
4155 * @pf: pointer to the PF struct
4157 static int ice_register_netdev(struct ice_pf *pf)
4159 struct ice_vsi *vsi;
4162 vsi = ice_get_main_vsi(pf);
4163 if (!vsi || !vsi->netdev)
4166 err = register_netdev(vsi->netdev);
4168 goto err_register_netdev;
4170 set_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4171 netif_carrier_off(vsi->netdev);
4172 netif_tx_stop_all_queues(vsi->netdev);
4173 err = ice_devlink_create_port(vsi);
4175 goto err_devlink_create;
4177 devlink_port_type_eth_set(&vsi->devlink_port, vsi->netdev);
4181 unregister_netdev(vsi->netdev);
4182 clear_bit(ICE_VSI_NETDEV_REGISTERED, vsi->state);
4183 err_register_netdev:
4184 free_netdev(vsi->netdev);
4186 clear_bit(ICE_VSI_NETDEV_ALLOCD, vsi->state);
4191 * ice_probe - Device initialization routine
4192 * @pdev: PCI device information struct
4193 * @ent: entry in ice_pci_tbl
4195 * Returns 0 on success, negative on failure
4198 ice_probe(struct pci_dev *pdev, const struct pci_device_id __always_unused *ent)
4200 struct device *dev = &pdev->dev;
4205 if (pdev->is_virtfn) {
4206 dev_err(dev, "can't probe a virtual function\n");
4210 /* this driver uses devres, see
4211 * Documentation/driver-api/driver-model/devres.rst
4213 err = pcim_enable_device(pdev);
4217 err = pcim_iomap_regions(pdev, BIT(ICE_BAR0), dev_driver_string(dev));
4219 dev_err(dev, "BAR0 I/O map error %d\n", err);
4223 pf = ice_allocate_pf(dev);
4227 /* set up for high or low DMA */
4228 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(64));
4230 err = dma_set_mask_and_coherent(dev, DMA_BIT_MASK(32));
4232 dev_err(dev, "DMA configuration failed: 0x%x\n", err);
4236 pci_enable_pcie_error_reporting(pdev);
4237 pci_set_master(pdev);
4240 pci_set_drvdata(pdev, pf);
4241 set_bit(ICE_DOWN, pf->state);
4242 /* Disable service task until DOWN bit is cleared */
4243 set_bit(ICE_SERVICE_DIS, pf->state);
4246 hw->hw_addr = pcim_iomap_table(pdev)[ICE_BAR0];
4247 pci_save_state(pdev);
4250 hw->vendor_id = pdev->vendor;
4251 hw->device_id = pdev->device;
4252 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
4253 hw->subsystem_vendor_id = pdev->subsystem_vendor;
4254 hw->subsystem_device_id = pdev->subsystem_device;
4255 hw->bus.device = PCI_SLOT(pdev->devfn);
4256 hw->bus.func = PCI_FUNC(pdev->devfn);
4257 ice_set_ctrlq_len(hw);
4259 pf->msg_enable = netif_msg_init(debug, ICE_DFLT_NETIF_M);
4261 err = ice_devlink_register(pf);
4263 dev_err(dev, "ice_devlink_register failed: %d\n", err);
4264 goto err_exit_unroll;
4267 #ifndef CONFIG_DYNAMIC_DEBUG
4269 hw->debug_mask = debug;
4272 err = ice_init_hw(hw);
4274 dev_err(dev, "ice_init_hw failed: %d\n", err);
4276 goto err_exit_unroll;
4281 /* if ice_request_fw fails, ICE_FLAG_ADV_FEATURES bit won't be
4282 * set in pf->state, which will cause ice_is_safe_mode to return
4285 if (ice_is_safe_mode(pf)) {
4286 dev_err(dev, "Package download failed. Advanced features disabled - Device now in Safe Mode\n");
4287 /* we already got function/device capabilities but these don't
4288 * reflect what the driver needs to do in safe mode. Instead of
4289 * adding conditional logic everywhere to ignore these
4290 * device/function capabilities, override them.
4292 ice_set_safe_mode_caps(hw);
4295 err = ice_init_pf(pf);
4297 dev_err(dev, "ice_init_pf failed: %d\n", err);
4298 goto err_init_pf_unroll;
4301 ice_devlink_init_regions(pf);
4303 pf->hw.udp_tunnel_nic.set_port = ice_udp_tunnel_set_port;
4304 pf->hw.udp_tunnel_nic.unset_port = ice_udp_tunnel_unset_port;
4305 pf->hw.udp_tunnel_nic.flags = UDP_TUNNEL_NIC_INFO_MAY_SLEEP;
4306 pf->hw.udp_tunnel_nic.shared = &pf->hw.udp_tunnel_shared;
4308 if (pf->hw.tnl.valid_count[TNL_VXLAN]) {
4309 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4310 pf->hw.tnl.valid_count[TNL_VXLAN];
4311 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4312 UDP_TUNNEL_TYPE_VXLAN;
4315 if (pf->hw.tnl.valid_count[TNL_GENEVE]) {
4316 pf->hw.udp_tunnel_nic.tables[i].n_entries =
4317 pf->hw.tnl.valid_count[TNL_GENEVE];
4318 pf->hw.udp_tunnel_nic.tables[i].tunnel_types =
4319 UDP_TUNNEL_TYPE_GENEVE;
4323 pf->num_alloc_vsi = hw->func_caps.guar_num_vsi;
4324 if (!pf->num_alloc_vsi) {
4326 goto err_init_pf_unroll;
4328 if (pf->num_alloc_vsi > UDP_TUNNEL_NIC_MAX_SHARING_DEVICES) {
4329 dev_warn(&pf->pdev->dev,
4330 "limiting the VSI count due to UDP tunnel limitation %d > %d\n",
4331 pf->num_alloc_vsi, UDP_TUNNEL_NIC_MAX_SHARING_DEVICES);
4332 pf->num_alloc_vsi = UDP_TUNNEL_NIC_MAX_SHARING_DEVICES;
4335 pf->vsi = devm_kcalloc(dev, pf->num_alloc_vsi, sizeof(*pf->vsi),
4339 goto err_init_pf_unroll;
4342 err = ice_init_interrupt_scheme(pf);
4344 dev_err(dev, "ice_init_interrupt_scheme failed: %d\n", err);
4346 goto err_init_vsi_unroll;
4349 /* In case of MSIX we are going to setup the misc vector right here
4350 * to handle admin queue events etc. In case of legacy and MSI
4351 * the misc functionality and queue processing is combined in
4352 * the same vector and that gets setup at open.
4354 err = ice_req_irq_msix_misc(pf);
4356 dev_err(dev, "setup of misc vector failed: %d\n", err);
4357 goto err_init_interrupt_unroll;
4360 /* create switch struct for the switch element created by FW on boot */
4361 pf->first_sw = devm_kzalloc(dev, sizeof(*pf->first_sw), GFP_KERNEL);
4362 if (!pf->first_sw) {
4364 goto err_msix_misc_unroll;
4368 pf->first_sw->bridge_mode = BRIDGE_MODE_VEB;
4370 pf->first_sw->bridge_mode = BRIDGE_MODE_VEPA;
4372 pf->first_sw->pf = pf;
4374 /* record the sw_id available for later use */
4375 pf->first_sw->sw_id = hw->port_info->sw_id;
4377 err = ice_setup_pf_sw(pf);
4379 dev_err(dev, "probe failed due to setup PF switch: %d\n", err);
4380 goto err_alloc_sw_unroll;
4383 clear_bit(ICE_SERVICE_DIS, pf->state);
4385 /* tell the firmware we are up */
4386 err = ice_send_version(pf);
4388 dev_err(dev, "probe failed sending driver version %s. error: %d\n",
4390 goto err_send_version_unroll;
4393 /* since everything is good, start the service timer */
4394 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4396 err = ice_init_link_events(pf->hw.port_info);
4398 dev_err(dev, "ice_init_link_events failed: %d\n", err);
4399 goto err_send_version_unroll;
4402 /* not a fatal error if this fails */
4403 err = ice_init_nvm_phy_type(pf->hw.port_info);
4405 dev_err(dev, "ice_init_nvm_phy_type failed: %d\n", err);
4407 /* not a fatal error if this fails */
4408 err = ice_update_link_info(pf->hw.port_info);
4410 dev_err(dev, "ice_update_link_info failed: %d\n", err);
4412 ice_init_link_dflt_override(pf->hw.port_info);
4414 ice_check_module_power(pf, pf->hw.port_info->phy.link_info.link_cfg_err);
4416 /* if media available, initialize PHY settings */
4417 if (pf->hw.port_info->phy.link_info.link_info &
4418 ICE_AQ_MEDIA_AVAILABLE) {
4419 /* not a fatal error if this fails */
4420 err = ice_init_phy_user_cfg(pf->hw.port_info);
4422 dev_err(dev, "ice_init_phy_user_cfg failed: %d\n", err);
4424 if (!test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, pf->flags)) {
4425 struct ice_vsi *vsi = ice_get_main_vsi(pf);
4428 ice_configure_phy(vsi);
4431 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
4434 ice_verify_cacheline_size(pf);
4436 /* Save wakeup reason register for later use */
4437 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4439 /* check for a power management event */
4440 ice_print_wake_reason(pf);
4442 /* clear wake status, all bits */
4443 wr32(hw, PFPM_WUS, U32_MAX);
4445 /* Disable WoL at init, wait for user to enable */
4446 device_set_wakeup_enable(dev, false);
4448 if (ice_is_safe_mode(pf)) {
4449 ice_set_safe_mode_vlan_cfg(pf);
4453 /* initialize DDP driven features */
4454 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4457 /* Note: Flow director init failure is non-fatal to load */
4458 if (ice_init_fdir(pf))
4459 dev_err(dev, "could not initialize flow director\n");
4461 /* Note: DCB init failure is non-fatal to load */
4462 if (ice_init_pf_dcb(pf, false)) {
4463 clear_bit(ICE_FLAG_DCB_CAPABLE, pf->flags);
4464 clear_bit(ICE_FLAG_DCB_ENA, pf->flags);
4466 ice_cfg_lldp_mib_change(&pf->hw, true);
4469 if (ice_init_lag(pf))
4470 dev_warn(dev, "Failed to init link aggregation support\n");
4472 /* print PCI link speed and width */
4473 pcie_print_link_status(pf->pdev);
4476 err = ice_register_netdev(pf);
4478 goto err_netdev_reg;
4480 /* ready to go, so clear down state bit */
4481 clear_bit(ICE_DOWN, pf->state);
4482 if (ice_is_aux_ena(pf)) {
4483 pf->aux_idx = ida_alloc(&ice_aux_ida, GFP_KERNEL);
4484 if (pf->aux_idx < 0) {
4485 dev_err(dev, "Failed to allocate device ID for AUX driver\n");
4487 goto err_netdev_reg;
4490 err = ice_init_rdma(pf);
4492 dev_err(dev, "Failed to initialize RDMA: %d\n", err);
4494 goto err_init_aux_unroll;
4497 dev_warn(dev, "RDMA is not supported on this device\n");
4502 err_init_aux_unroll:
4504 ida_free(&ice_aux_ida, pf->aux_idx);
4506 err_send_version_unroll:
4507 ice_vsi_release_all(pf);
4508 err_alloc_sw_unroll:
4509 set_bit(ICE_SERVICE_DIS, pf->state);
4510 set_bit(ICE_DOWN, pf->state);
4511 devm_kfree(dev, pf->first_sw);
4512 err_msix_misc_unroll:
4513 ice_free_irq_msix_misc(pf);
4514 err_init_interrupt_unroll:
4515 ice_clear_interrupt_scheme(pf);
4516 err_init_vsi_unroll:
4517 devm_kfree(dev, pf->vsi);
4520 ice_devlink_destroy_regions(pf);
4523 ice_devlink_unregister(pf);
4524 pci_disable_pcie_error_reporting(pdev);
4525 pci_disable_device(pdev);
4530 * ice_set_wake - enable or disable Wake on LAN
4531 * @pf: pointer to the PF struct
4533 * Simple helper for WoL control
4535 static void ice_set_wake(struct ice_pf *pf)
4537 struct ice_hw *hw = &pf->hw;
4538 bool wol = pf->wol_ena;
4540 /* clear wake state, otherwise new wake events won't fire */
4541 wr32(hw, PFPM_WUS, U32_MAX);
4543 /* enable / disable APM wake up, no RMW needed */
4544 wr32(hw, PFPM_APM, wol ? PFPM_APM_APME_M : 0);
4546 /* set magic packet filter enabled */
4547 wr32(hw, PFPM_WUFC, wol ? PFPM_WUFC_MAG_M : 0);
4551 * ice_setup_mc_magic_wake - setup device to wake on multicast magic packet
4552 * @pf: pointer to the PF struct
4554 * Issue firmware command to enable multicast magic wake, making
4555 * sure that any locally administered address (LAA) is used for
4556 * wake, and that PF reset doesn't undo the LAA.
4558 static void ice_setup_mc_magic_wake(struct ice_pf *pf)
4560 struct device *dev = ice_pf_to_dev(pf);
4561 struct ice_hw *hw = &pf->hw;
4562 enum ice_status status;
4563 u8 mac_addr[ETH_ALEN];
4564 struct ice_vsi *vsi;
4570 vsi = ice_get_main_vsi(pf);
4574 /* Get current MAC address in case it's an LAA */
4576 ether_addr_copy(mac_addr, vsi->netdev->dev_addr);
4578 ether_addr_copy(mac_addr, vsi->port_info->mac.perm_addr);
4580 flags = ICE_AQC_MAN_MAC_WR_MC_MAG_EN |
4581 ICE_AQC_MAN_MAC_UPDATE_LAA_WOL |
4582 ICE_AQC_MAN_MAC_WR_WOL_LAA_PFR_KEEP;
4584 status = ice_aq_manage_mac_write(hw, mac_addr, flags, NULL);
4586 dev_err(dev, "Failed to enable Multicast Magic Packet wake, err %s aq_err %s\n",
4587 ice_stat_str(status),
4588 ice_aq_str(hw->adminq.sq_last_status));
4592 * ice_remove - Device removal routine
4593 * @pdev: PCI device information struct
4595 static void ice_remove(struct pci_dev *pdev)
4597 struct ice_pf *pf = pci_get_drvdata(pdev);
4603 for (i = 0; i < ICE_MAX_RESET_WAIT; i++) {
4604 if (!ice_is_reset_in_progress(pf->state))
4609 if (test_bit(ICE_FLAG_SRIOV_ENA, pf->flags)) {
4610 set_bit(ICE_VF_RESETS_DISABLED, pf->state);
4614 ice_service_task_stop(pf);
4616 ice_aq_cancel_waiting_tasks(pf);
4617 ice_unplug_aux_dev(pf);
4618 ida_free(&ice_aux_ida, pf->aux_idx);
4619 set_bit(ICE_DOWN, pf->state);
4621 mutex_destroy(&(&pf->hw)->fdir_fltr_lock);
4623 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
4624 ice_ptp_release(pf);
4625 if (!ice_is_safe_mode(pf))
4626 ice_remove_arfs(pf);
4627 ice_setup_mc_magic_wake(pf);
4628 ice_vsi_release_all(pf);
4630 ice_free_irq_msix_misc(pf);
4631 ice_for_each_vsi(pf, i) {
4634 ice_vsi_free_q_vectors(pf->vsi[i]);
4637 ice_devlink_destroy_regions(pf);
4638 ice_deinit_hw(&pf->hw);
4639 ice_devlink_unregister(pf);
4641 /* Issue a PFR as part of the prescribed driver unload flow. Do not
4642 * do it via ice_schedule_reset() since there is no need to rebuild
4643 * and the service task is already stopped.
4645 ice_reset(&pf->hw, ICE_RESET_PFR);
4646 pci_wait_for_pending_transaction(pdev);
4647 ice_clear_interrupt_scheme(pf);
4648 pci_disable_pcie_error_reporting(pdev);
4649 pci_disable_device(pdev);
4653 * ice_shutdown - PCI callback for shutting down device
4654 * @pdev: PCI device information struct
4656 static void ice_shutdown(struct pci_dev *pdev)
4658 struct ice_pf *pf = pci_get_drvdata(pdev);
4662 if (system_state == SYSTEM_POWER_OFF) {
4663 pci_wake_from_d3(pdev, pf->wol_ena);
4664 pci_set_power_state(pdev, PCI_D3hot);
4670 * ice_prepare_for_shutdown - prep for PCI shutdown
4671 * @pf: board private structure
4673 * Inform or close all dependent features in prep for PCI device shutdown
4675 static void ice_prepare_for_shutdown(struct ice_pf *pf)
4677 struct ice_hw *hw = &pf->hw;
4680 /* Notify VFs of impending reset */
4681 if (ice_check_sq_alive(hw, &hw->mailboxq))
4682 ice_vc_notify_reset(pf);
4684 dev_dbg(ice_pf_to_dev(pf), "Tearing down internal switch for shutdown\n");
4686 /* disable the VSIs and their queues that are not already DOWN */
4687 ice_pf_dis_all_vsi(pf, false);
4689 ice_for_each_vsi(pf, v)
4691 pf->vsi[v]->vsi_num = 0;
4693 ice_shutdown_all_ctrlq(hw);
4697 * ice_reinit_interrupt_scheme - Reinitialize interrupt scheme
4698 * @pf: board private structure to reinitialize
4700 * This routine reinitialize interrupt scheme that was cleared during
4701 * power management suspend callback.
4703 * This should be called during resume routine to re-allocate the q_vectors
4704 * and reacquire interrupts.
4706 static int ice_reinit_interrupt_scheme(struct ice_pf *pf)
4708 struct device *dev = ice_pf_to_dev(pf);
4711 /* Since we clear MSIX flag during suspend, we need to
4712 * set it back during resume...
4715 ret = ice_init_interrupt_scheme(pf);
4717 dev_err(dev, "Failed to re-initialize interrupt %d\n", ret);
4721 /* Remap vectors and rings, after successful re-init interrupts */
4722 ice_for_each_vsi(pf, v) {
4726 ret = ice_vsi_alloc_q_vectors(pf->vsi[v]);
4729 ice_vsi_map_rings_to_vectors(pf->vsi[v]);
4732 ret = ice_req_irq_msix_misc(pf);
4734 dev_err(dev, "Setting up misc vector failed after device suspend %d\n",
4744 ice_vsi_free_q_vectors(pf->vsi[v]);
4751 * @dev: generic device information structure
4753 * Power Management callback to quiesce the device and prepare
4754 * for D3 transition.
4756 static int __maybe_unused ice_suspend(struct device *dev)
4758 struct pci_dev *pdev = to_pci_dev(dev);
4762 pf = pci_get_drvdata(pdev);
4764 if (!ice_pf_state_is_nominal(pf)) {
4765 dev_err(dev, "Device is not ready, no need to suspend it\n");
4769 /* Stop watchdog tasks until resume completion.
4770 * Even though it is most likely that the service task is
4771 * disabled if the device is suspended or down, the service task's
4772 * state is controlled by a different state bit, and we should
4773 * store and honor whatever state that bit is in at this point.
4775 disabled = ice_service_task_stop(pf);
4777 ice_unplug_aux_dev(pf);
4779 /* Already suspended?, then there is nothing to do */
4780 if (test_and_set_bit(ICE_SUSPENDED, pf->state)) {
4782 ice_service_task_restart(pf);
4786 if (test_bit(ICE_DOWN, pf->state) ||
4787 ice_is_reset_in_progress(pf->state)) {
4788 dev_err(dev, "can't suspend device in reset or already down\n");
4790 ice_service_task_restart(pf);
4794 ice_setup_mc_magic_wake(pf);
4796 ice_prepare_for_shutdown(pf);
4800 /* Free vectors, clear the interrupt scheme and release IRQs
4801 * for proper hibernation, especially with large number of CPUs.
4802 * Otherwise hibernation might fail when mapping all the vectors back
4805 ice_free_irq_msix_misc(pf);
4806 ice_for_each_vsi(pf, v) {
4809 ice_vsi_free_q_vectors(pf->vsi[v]);
4811 ice_free_cpu_rx_rmap(ice_get_main_vsi(pf));
4812 ice_clear_interrupt_scheme(pf);
4814 pci_save_state(pdev);
4815 pci_wake_from_d3(pdev, pf->wol_ena);
4816 pci_set_power_state(pdev, PCI_D3hot);
4821 * ice_resume - PM callback for waking up from D3
4822 * @dev: generic device information structure
4824 static int __maybe_unused ice_resume(struct device *dev)
4826 struct pci_dev *pdev = to_pci_dev(dev);
4827 enum ice_reset_req reset_type;
4832 pci_set_power_state(pdev, PCI_D0);
4833 pci_restore_state(pdev);
4834 pci_save_state(pdev);
4836 if (!pci_device_is_present(pdev))
4839 ret = pci_enable_device_mem(pdev);
4841 dev_err(dev, "Cannot enable device after suspend\n");
4845 pf = pci_get_drvdata(pdev);
4848 pf->wakeup_reason = rd32(hw, PFPM_WUS);
4849 ice_print_wake_reason(pf);
4851 /* We cleared the interrupt scheme when we suspended, so we need to
4852 * restore it now to resume device functionality.
4854 ret = ice_reinit_interrupt_scheme(pf);
4856 dev_err(dev, "Cannot restore interrupt scheme: %d\n", ret);
4858 clear_bit(ICE_DOWN, pf->state);
4859 /* Now perform PF reset and rebuild */
4860 reset_type = ICE_RESET_PFR;
4861 /* re-enable service task for reset, but allow reset to schedule it */
4862 clear_bit(ICE_SERVICE_DIS, pf->state);
4864 if (ice_schedule_reset(pf, reset_type))
4865 dev_err(dev, "Reset during resume failed.\n");
4867 clear_bit(ICE_SUSPENDED, pf->state);
4868 ice_service_task_restart(pf);
4870 /* Restart the service task */
4871 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4875 #endif /* CONFIG_PM */
4878 * ice_pci_err_detected - warning that PCI error has been detected
4879 * @pdev: PCI device information struct
4880 * @err: the type of PCI error
4882 * Called to warn that something happened on the PCI bus and the error handling
4883 * is in progress. Allows the driver to gracefully prepare/handle PCI errors.
4885 static pci_ers_result_t
4886 ice_pci_err_detected(struct pci_dev *pdev, pci_channel_state_t err)
4888 struct ice_pf *pf = pci_get_drvdata(pdev);
4891 dev_err(&pdev->dev, "%s: unrecoverable device error %d\n",
4893 return PCI_ERS_RESULT_DISCONNECT;
4896 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4897 ice_service_task_stop(pf);
4899 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4900 set_bit(ICE_PFR_REQ, pf->state);
4901 ice_prepare_for_reset(pf);
4905 return PCI_ERS_RESULT_NEED_RESET;
4909 * ice_pci_err_slot_reset - a PCI slot reset has just happened
4910 * @pdev: PCI device information struct
4912 * Called to determine if the driver can recover from the PCI slot reset by
4913 * using a register read to determine if the device is recoverable.
4915 static pci_ers_result_t ice_pci_err_slot_reset(struct pci_dev *pdev)
4917 struct ice_pf *pf = pci_get_drvdata(pdev);
4918 pci_ers_result_t result;
4922 err = pci_enable_device_mem(pdev);
4924 dev_err(&pdev->dev, "Cannot re-enable PCI device after reset, error %d\n",
4926 result = PCI_ERS_RESULT_DISCONNECT;
4928 pci_set_master(pdev);
4929 pci_restore_state(pdev);
4930 pci_save_state(pdev);
4931 pci_wake_from_d3(pdev, false);
4933 /* Check for life */
4934 reg = rd32(&pf->hw, GLGEN_RTRIG);
4936 result = PCI_ERS_RESULT_RECOVERED;
4938 result = PCI_ERS_RESULT_DISCONNECT;
4941 err = pci_aer_clear_nonfatal_status(pdev);
4943 dev_dbg(&pdev->dev, "pci_aer_clear_nonfatal_status() failed, error %d\n",
4945 /* non-fatal, continue */
4951 * ice_pci_err_resume - restart operations after PCI error recovery
4952 * @pdev: PCI device information struct
4954 * Called to allow the driver to bring things back up after PCI error and/or
4955 * reset recovery have finished
4957 static void ice_pci_err_resume(struct pci_dev *pdev)
4959 struct ice_pf *pf = pci_get_drvdata(pdev);
4962 dev_err(&pdev->dev, "%s failed, device is unrecoverable\n",
4967 if (test_bit(ICE_SUSPENDED, pf->state)) {
4968 dev_dbg(&pdev->dev, "%s failed to resume normal operations!\n",
4973 ice_restore_all_vfs_msi_state(pdev);
4975 ice_do_reset(pf, ICE_RESET_PFR);
4976 ice_service_task_restart(pf);
4977 mod_timer(&pf->serv_tmr, round_jiffies(jiffies + pf->serv_tmr_period));
4981 * ice_pci_err_reset_prepare - prepare device driver for PCI reset
4982 * @pdev: PCI device information struct
4984 static void ice_pci_err_reset_prepare(struct pci_dev *pdev)
4986 struct ice_pf *pf = pci_get_drvdata(pdev);
4988 if (!test_bit(ICE_SUSPENDED, pf->state)) {
4989 ice_service_task_stop(pf);
4991 if (!test_bit(ICE_PREPARED_FOR_RESET, pf->state)) {
4992 set_bit(ICE_PFR_REQ, pf->state);
4993 ice_prepare_for_reset(pf);
4999 * ice_pci_err_reset_done - PCI reset done, device driver reset can begin
5000 * @pdev: PCI device information struct
5002 static void ice_pci_err_reset_done(struct pci_dev *pdev)
5004 ice_pci_err_resume(pdev);
5007 /* ice_pci_tbl - PCI Device ID Table
5009 * Wildcard entries (PCI_ANY_ID) should come last
5010 * Last entry must be all 0s
5012 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
5013 * Class, Class Mask, private data (not used) }
5015 static const struct pci_device_id ice_pci_tbl[] = {
5016 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_BACKPLANE), 0 },
5017 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_QSFP), 0 },
5018 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810C_SFP), 0 },
5019 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E810_XXV_SFP), 0 },
5020 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_BACKPLANE), 0 },
5021 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_QSFP), 0 },
5022 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SFP), 0 },
5023 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_10G_BASE_T), 0 },
5024 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823C_SGMII), 0 },
5025 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_BACKPLANE), 0 },
5026 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_QSFP), 0 },
5027 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SFP), 0 },
5028 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_10G_BASE_T), 0 },
5029 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822C_SGMII), 0 },
5030 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_BACKPLANE), 0 },
5031 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SFP), 0 },
5032 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_10G_BASE_T), 0 },
5033 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E822L_SGMII), 0 },
5034 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_BACKPLANE), 0 },
5035 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_SFP), 0 },
5036 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_10G_BASE_T), 0 },
5037 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_1GBE), 0 },
5038 { PCI_VDEVICE(INTEL, ICE_DEV_ID_E823L_QSFP), 0 },
5039 /* required last entry */
5042 MODULE_DEVICE_TABLE(pci, ice_pci_tbl);
5044 static __maybe_unused SIMPLE_DEV_PM_OPS(ice_pm_ops, ice_suspend, ice_resume);
5046 static const struct pci_error_handlers ice_pci_err_handler = {
5047 .error_detected = ice_pci_err_detected,
5048 .slot_reset = ice_pci_err_slot_reset,
5049 .reset_prepare = ice_pci_err_reset_prepare,
5050 .reset_done = ice_pci_err_reset_done,
5051 .resume = ice_pci_err_resume
5054 static struct pci_driver ice_driver = {
5055 .name = KBUILD_MODNAME,
5056 .id_table = ice_pci_tbl,
5058 .remove = ice_remove,
5060 .driver.pm = &ice_pm_ops,
5061 #endif /* CONFIG_PM */
5062 .shutdown = ice_shutdown,
5063 .sriov_configure = ice_sriov_configure,
5064 .err_handler = &ice_pci_err_handler
5068 * ice_module_init - Driver registration routine
5070 * ice_module_init is the first routine called when the driver is
5071 * loaded. All it does is register with the PCI subsystem.
5073 static int __init ice_module_init(void)
5077 pr_info("%s\n", ice_driver_string);
5078 pr_info("%s\n", ice_copyright);
5080 ice_wq = alloc_workqueue("%s", WQ_MEM_RECLAIM, 0, KBUILD_MODNAME);
5082 pr_err("Failed to create workqueue\n");
5086 status = pci_register_driver(&ice_driver);
5088 pr_err("failed to register PCI driver, err %d\n", status);
5089 destroy_workqueue(ice_wq);
5094 module_init(ice_module_init);
5097 * ice_module_exit - Driver exit cleanup routine
5099 * ice_module_exit is called just before the driver is removed
5102 static void __exit ice_module_exit(void)
5104 pci_unregister_driver(&ice_driver);
5105 destroy_workqueue(ice_wq);
5106 pr_info("module unloaded\n");
5108 module_exit(ice_module_exit);
5111 * ice_set_mac_address - NDO callback to set MAC address
5112 * @netdev: network interface device structure
5113 * @pi: pointer to an address structure
5115 * Returns 0 on success, negative on failure
5117 static int ice_set_mac_address(struct net_device *netdev, void *pi)
5119 struct ice_netdev_priv *np = netdev_priv(netdev);
5120 struct ice_vsi *vsi = np->vsi;
5121 struct ice_pf *pf = vsi->back;
5122 struct ice_hw *hw = &pf->hw;
5123 struct sockaddr *addr = pi;
5124 enum ice_status status;
5129 mac = (u8 *)addr->sa_data;
5131 if (!is_valid_ether_addr(mac))
5132 return -EADDRNOTAVAIL;
5134 if (ether_addr_equal(netdev->dev_addr, mac)) {
5135 netdev_dbg(netdev, "already using mac %pM\n", mac);
5139 if (test_bit(ICE_DOWN, pf->state) ||
5140 ice_is_reset_in_progress(pf->state)) {
5141 netdev_err(netdev, "can't set mac %pM. device not ready\n",
5146 netif_addr_lock_bh(netdev);
5147 /* Clean up old MAC filter. Not an error if old filter doesn't exist */
5148 status = ice_fltr_remove_mac(vsi, netdev->dev_addr, ICE_FWD_TO_VSI);
5149 if (status && status != ICE_ERR_DOES_NOT_EXIST) {
5150 err = -EADDRNOTAVAIL;
5151 goto err_update_filters;
5154 /* Add filter for new MAC. If filter exists, return success */
5155 status = ice_fltr_add_mac(vsi, mac, ICE_FWD_TO_VSI);
5156 if (status == ICE_ERR_ALREADY_EXISTS)
5157 /* Although this MAC filter is already present in hardware it's
5158 * possible in some cases (e.g. bonding) that dev_addr was
5159 * modified outside of the driver and needs to be restored back
5162 netdev_dbg(netdev, "filter for MAC %pM already exists\n", mac);
5164 /* error if the new filter addition failed */
5165 err = -EADDRNOTAVAIL;
5169 netdev_err(netdev, "can't set MAC %pM. filter update failed\n",
5171 netif_addr_unlock_bh(netdev);
5175 /* change the netdev's MAC address */
5176 memcpy(netdev->dev_addr, mac, netdev->addr_len);
5177 netif_addr_unlock_bh(netdev);
5178 netdev_dbg(vsi->netdev, "updated MAC address to %pM\n",
5181 /* write new MAC address to the firmware */
5182 flags = ICE_AQC_MAN_MAC_UPDATE_LAA_WOL;
5183 status = ice_aq_manage_mac_write(hw, mac, flags, NULL);
5185 netdev_err(netdev, "can't set MAC %pM. write to firmware failed error %s\n",
5186 mac, ice_stat_str(status));
5192 * ice_set_rx_mode - NDO callback to set the netdev filters
5193 * @netdev: network interface device structure
5195 static void ice_set_rx_mode(struct net_device *netdev)
5197 struct ice_netdev_priv *np = netdev_priv(netdev);
5198 struct ice_vsi *vsi = np->vsi;
5203 /* Set the flags to synchronize filters
5204 * ndo_set_rx_mode may be triggered even without a change in netdev
5207 set_bit(ICE_VSI_UMAC_FLTR_CHANGED, vsi->state);
5208 set_bit(ICE_VSI_MMAC_FLTR_CHANGED, vsi->state);
5209 set_bit(ICE_FLAG_FLTR_SYNC, vsi->back->flags);
5211 /* schedule our worker thread which will take care of
5212 * applying the new filter changes
5214 ice_service_task_schedule(vsi->back);
5218 * ice_set_tx_maxrate - NDO callback to set the maximum per-queue bitrate
5219 * @netdev: network interface device structure
5220 * @queue_index: Queue ID
5221 * @maxrate: maximum bandwidth in Mbps
5224 ice_set_tx_maxrate(struct net_device *netdev, int queue_index, u32 maxrate)
5226 struct ice_netdev_priv *np = netdev_priv(netdev);
5227 struct ice_vsi *vsi = np->vsi;
5228 enum ice_status status;
5232 /* Validate maxrate requested is within permitted range */
5233 if (maxrate && (maxrate > (ICE_SCHED_MAX_BW / 1000))) {
5234 netdev_err(netdev, "Invalid max rate %d specified for the queue %d\n",
5235 maxrate, queue_index);
5239 q_handle = vsi->tx_rings[queue_index]->q_handle;
5240 tc = ice_dcb_get_tc(vsi, queue_index);
5242 /* Set BW back to default, when user set maxrate to 0 */
5244 status = ice_cfg_q_bw_dflt_lmt(vsi->port_info, vsi->idx, tc,
5245 q_handle, ICE_MAX_BW);
5247 status = ice_cfg_q_bw_lmt(vsi->port_info, vsi->idx, tc,
5248 q_handle, ICE_MAX_BW, maxrate * 1000);
5250 netdev_err(netdev, "Unable to set Tx max rate, error %s\n",
5251 ice_stat_str(status));
5259 * ice_fdb_add - add an entry to the hardware database
5260 * @ndm: the input from the stack
5261 * @tb: pointer to array of nladdr (unused)
5262 * @dev: the net device pointer
5263 * @addr: the MAC address entry being added
5265 * @flags: instructions from stack about fdb operation
5266 * @extack: netlink extended ack
5269 ice_fdb_add(struct ndmsg *ndm, struct nlattr __always_unused *tb[],
5270 struct net_device *dev, const unsigned char *addr, u16 vid,
5271 u16 flags, struct netlink_ext_ack __always_unused *extack)
5276 netdev_err(dev, "VLANs aren't supported yet for dev_uc|mc_add()\n");
5279 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
5280 netdev_err(dev, "FDB only supports static addresses\n");
5284 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
5285 err = dev_uc_add_excl(dev, addr);
5286 else if (is_multicast_ether_addr(addr))
5287 err = dev_mc_add_excl(dev, addr);
5291 /* Only return duplicate errors if NLM_F_EXCL is set */
5292 if (err == -EEXIST && !(flags & NLM_F_EXCL))
5299 * ice_fdb_del - delete an entry from the hardware database
5300 * @ndm: the input from the stack
5301 * @tb: pointer to array of nladdr (unused)
5302 * @dev: the net device pointer
5303 * @addr: the MAC address entry being added
5307 ice_fdb_del(struct ndmsg *ndm, __always_unused struct nlattr *tb[],
5308 struct net_device *dev, const unsigned char *addr,
5309 __always_unused u16 vid)
5313 if (ndm->ndm_state & NUD_PERMANENT) {
5314 netdev_err(dev, "FDB only supports static addresses\n");
5318 if (is_unicast_ether_addr(addr))
5319 err = dev_uc_del(dev, addr);
5320 else if (is_multicast_ether_addr(addr))
5321 err = dev_mc_del(dev, addr);
5329 * ice_set_features - set the netdev feature flags
5330 * @netdev: ptr to the netdev being adjusted
5331 * @features: the feature set that the stack is suggesting
5334 ice_set_features(struct net_device *netdev, netdev_features_t features)
5336 struct ice_netdev_priv *np = netdev_priv(netdev);
5337 struct ice_vsi *vsi = np->vsi;
5338 struct ice_pf *pf = vsi->back;
5341 /* Don't set any netdev advanced features with device in Safe Mode */
5342 if (ice_is_safe_mode(vsi->back)) {
5343 dev_err(ice_pf_to_dev(vsi->back), "Device is in Safe Mode - not enabling advanced netdev features\n");
5347 /* Do not change setting during reset */
5348 if (ice_is_reset_in_progress(pf->state)) {
5349 dev_err(ice_pf_to_dev(vsi->back), "Device is resetting, changing advanced netdev features temporarily unavailable.\n");
5353 /* Multiple features can be changed in one call so keep features in
5354 * separate if/else statements to guarantee each feature is checked
5356 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
5357 ice_vsi_manage_rss_lut(vsi, true);
5358 else if (!(features & NETIF_F_RXHASH) &&
5359 netdev->features & NETIF_F_RXHASH)
5360 ice_vsi_manage_rss_lut(vsi, false);
5362 if ((features & NETIF_F_HW_VLAN_CTAG_RX) &&
5363 !(netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5364 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5365 else if (!(features & NETIF_F_HW_VLAN_CTAG_RX) &&
5366 (netdev->features & NETIF_F_HW_VLAN_CTAG_RX))
5367 ret = ice_vsi_manage_vlan_stripping(vsi, false);
5369 if ((features & NETIF_F_HW_VLAN_CTAG_TX) &&
5370 !(netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5371 ret = ice_vsi_manage_vlan_insertion(vsi);
5372 else if (!(features & NETIF_F_HW_VLAN_CTAG_TX) &&
5373 (netdev->features & NETIF_F_HW_VLAN_CTAG_TX))
5374 ret = ice_vsi_manage_vlan_insertion(vsi);
5376 if ((features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5377 !(netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5378 ret = ice_cfg_vlan_pruning(vsi, true, false);
5379 else if (!(features & NETIF_F_HW_VLAN_CTAG_FILTER) &&
5380 (netdev->features & NETIF_F_HW_VLAN_CTAG_FILTER))
5381 ret = ice_cfg_vlan_pruning(vsi, false, false);
5383 if ((features & NETIF_F_NTUPLE) &&
5384 !(netdev->features & NETIF_F_NTUPLE)) {
5385 ice_vsi_manage_fdir(vsi, true);
5387 } else if (!(features & NETIF_F_NTUPLE) &&
5388 (netdev->features & NETIF_F_NTUPLE)) {
5389 ice_vsi_manage_fdir(vsi, false);
5390 ice_clear_arfs(vsi);
5397 * ice_vsi_vlan_setup - Setup VLAN offload properties on a VSI
5398 * @vsi: VSI to setup VLAN properties for
5400 static int ice_vsi_vlan_setup(struct ice_vsi *vsi)
5404 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
5405 ret = ice_vsi_manage_vlan_stripping(vsi, true);
5406 if (vsi->netdev->features & NETIF_F_HW_VLAN_CTAG_TX)
5407 ret = ice_vsi_manage_vlan_insertion(vsi);
5413 * ice_vsi_cfg - Setup the VSI
5414 * @vsi: the VSI being configured
5416 * Return 0 on success and negative value on error
5418 int ice_vsi_cfg(struct ice_vsi *vsi)
5423 ice_set_rx_mode(vsi->netdev);
5425 err = ice_vsi_vlan_setup(vsi);
5430 ice_vsi_cfg_dcb_rings(vsi);
5432 err = ice_vsi_cfg_lan_txqs(vsi);
5433 if (!err && ice_is_xdp_ena_vsi(vsi))
5434 err = ice_vsi_cfg_xdp_txqs(vsi);
5436 err = ice_vsi_cfg_rxqs(vsi);
5441 /* THEORY OF MODERATION:
5442 * The below code creates custom DIM profiles for use by this driver, because
5443 * the ice driver hardware works differently than the hardware that DIMLIB was
5444 * originally made for. ice hardware doesn't have packet count limits that
5445 * can trigger an interrupt, but it *does* have interrupt rate limit support,
5446 * and this code adds that capability to be used by the driver when it's using
5447 * DIMLIB. The DIMLIB code was always designed to be a suggestion to the driver
5448 * for how to "respond" to traffic and interrupts, so this driver uses a
5449 * slightly different set of moderation parameters to get best performance.
5452 /* the throttle rate for interrupts, basically worst case delay before
5453 * an initial interrupt fires, value is stored in microseconds.
5456 /* the rate limit for interrupts, which can cap a delay from a small
5457 * ITR at a certain amount of interrupts per second. f.e. a 2us ITR
5458 * could yield as much as 500,000 interrupts per second, but with a
5459 * 10us rate limit, it limits to 100,000 interrupts per second. Value
5460 * is stored in microseconds.
5465 /* Make a different profile for Rx that doesn't allow quite so aggressive
5466 * moderation at the high end (it maxes out at 128us or about 8k interrupts a
5467 * second. The INTRL/rate parameters here are only useful to cap small ITR
5468 * values, which is why for larger ITR's - like 128, which can only generate
5469 * 8k interrupts per second, there is no point to rate limit and the values
5470 * are set to zero. The rate limit values do affect latency, and so must
5471 * be reasonably small so to not impact latency sensitive tests.
5473 static const struct ice_dim rx_profile[] = {
5481 /* The transmit profile, which has the same sorts of values
5482 * as the previous struct
5484 static const struct ice_dim tx_profile[] = {
5492 static void ice_tx_dim_work(struct work_struct *work)
5494 struct ice_ring_container *rc;
5495 struct ice_q_vector *q_vector;
5499 dim = container_of(work, struct dim, work);
5500 rc = container_of(dim, struct ice_ring_container, dim);
5501 q_vector = container_of(rc, struct ice_q_vector, tx);
5503 if (dim->profile_ix >= ARRAY_SIZE(tx_profile))
5504 dim->profile_ix = ARRAY_SIZE(tx_profile) - 1;
5506 /* look up the values in our local table */
5507 itr = tx_profile[dim->profile_ix].itr;
5508 intrl = tx_profile[dim->profile_ix].intrl;
5510 ice_trace(tx_dim_work, q_vector, dim);
5511 ice_write_itr(rc, itr);
5512 ice_write_intrl(q_vector, intrl);
5514 dim->state = DIM_START_MEASURE;
5517 static void ice_rx_dim_work(struct work_struct *work)
5519 struct ice_ring_container *rc;
5520 struct ice_q_vector *q_vector;
5524 dim = container_of(work, struct dim, work);
5525 rc = container_of(dim, struct ice_ring_container, dim);
5526 q_vector = container_of(rc, struct ice_q_vector, rx);
5528 if (dim->profile_ix >= ARRAY_SIZE(rx_profile))
5529 dim->profile_ix = ARRAY_SIZE(rx_profile) - 1;
5531 /* look up the values in our local table */
5532 itr = rx_profile[dim->profile_ix].itr;
5533 intrl = rx_profile[dim->profile_ix].intrl;
5535 ice_trace(rx_dim_work, q_vector, dim);
5536 ice_write_itr(rc, itr);
5537 ice_write_intrl(q_vector, intrl);
5539 dim->state = DIM_START_MEASURE;
5543 * ice_napi_enable_all - Enable NAPI for all q_vectors in the VSI
5544 * @vsi: the VSI being configured
5546 static void ice_napi_enable_all(struct ice_vsi *vsi)
5553 ice_for_each_q_vector(vsi, q_idx) {
5554 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5556 INIT_WORK(&q_vector->tx.dim.work, ice_tx_dim_work);
5557 q_vector->tx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5559 INIT_WORK(&q_vector->rx.dim.work, ice_rx_dim_work);
5560 q_vector->rx.dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
5562 if (q_vector->rx.ring || q_vector->tx.ring)
5563 napi_enable(&q_vector->napi);
5568 * ice_up_complete - Finish the last steps of bringing up a connection
5569 * @vsi: The VSI being configured
5571 * Return 0 on success and negative value on error
5573 static int ice_up_complete(struct ice_vsi *vsi)
5575 struct ice_pf *pf = vsi->back;
5578 ice_vsi_cfg_msix(vsi);
5580 /* Enable only Rx rings, Tx rings were enabled by the FW when the
5581 * Tx queue group list was configured and the context bits were
5582 * programmed using ice_vsi_cfg_txqs
5584 err = ice_vsi_start_all_rx_rings(vsi);
5588 clear_bit(ICE_VSI_DOWN, vsi->state);
5589 ice_napi_enable_all(vsi);
5590 ice_vsi_ena_irq(vsi);
5592 if (vsi->port_info &&
5593 (vsi->port_info->phy.link_info.link_info & ICE_AQ_LINK_UP) &&
5595 ice_print_link_msg(vsi, true);
5596 netif_tx_start_all_queues(vsi->netdev);
5597 netif_carrier_on(vsi->netdev);
5600 ice_service_task_schedule(pf);
5606 * ice_up - Bring the connection back up after being down
5607 * @vsi: VSI being configured
5609 int ice_up(struct ice_vsi *vsi)
5613 err = ice_vsi_cfg(vsi);
5615 err = ice_up_complete(vsi);
5621 * ice_fetch_u64_stats_per_ring - get packets and bytes stats per ring
5622 * @ring: Tx or Rx ring to read stats from
5623 * @pkts: packets stats counter
5624 * @bytes: bytes stats counter
5626 * This function fetches stats from the ring considering the atomic operations
5627 * that needs to be performed to read u64 values in 32 bit machine.
5630 ice_fetch_u64_stats_per_ring(struct ice_ring *ring, u64 *pkts, u64 *bytes)
5639 start = u64_stats_fetch_begin_irq(&ring->syncp);
5640 *pkts = ring->stats.pkts;
5641 *bytes = ring->stats.bytes;
5642 } while (u64_stats_fetch_retry_irq(&ring->syncp, start));
5646 * ice_update_vsi_tx_ring_stats - Update VSI Tx ring stats counters
5647 * @vsi: the VSI to be updated
5648 * @rings: rings to work on
5649 * @count: number of rings
5652 ice_update_vsi_tx_ring_stats(struct ice_vsi *vsi, struct ice_ring **rings,
5655 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5658 for (i = 0; i < count; i++) {
5659 struct ice_ring *ring;
5662 ring = READ_ONCE(rings[i]);
5663 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5664 vsi_stats->tx_packets += pkts;
5665 vsi_stats->tx_bytes += bytes;
5666 vsi->tx_restart += ring->tx_stats.restart_q;
5667 vsi->tx_busy += ring->tx_stats.tx_busy;
5668 vsi->tx_linearize += ring->tx_stats.tx_linearize;
5673 * ice_update_vsi_ring_stats - Update VSI stats counters
5674 * @vsi: the VSI to be updated
5676 static void ice_update_vsi_ring_stats(struct ice_vsi *vsi)
5678 struct rtnl_link_stats64 *vsi_stats = &vsi->net_stats;
5682 /* reset netdev stats */
5683 vsi_stats->tx_packets = 0;
5684 vsi_stats->tx_bytes = 0;
5685 vsi_stats->rx_packets = 0;
5686 vsi_stats->rx_bytes = 0;
5688 /* reset non-netdev (extended) stats */
5689 vsi->tx_restart = 0;
5691 vsi->tx_linearize = 0;
5692 vsi->rx_buf_failed = 0;
5693 vsi->rx_page_failed = 0;
5697 /* update Tx rings counters */
5698 ice_update_vsi_tx_ring_stats(vsi, vsi->tx_rings, vsi->num_txq);
5700 /* update Rx rings counters */
5701 ice_for_each_rxq(vsi, i) {
5702 struct ice_ring *ring = READ_ONCE(vsi->rx_rings[i]);
5704 ice_fetch_u64_stats_per_ring(ring, &pkts, &bytes);
5705 vsi_stats->rx_packets += pkts;
5706 vsi_stats->rx_bytes += bytes;
5707 vsi->rx_buf_failed += ring->rx_stats.alloc_buf_failed;
5708 vsi->rx_page_failed += ring->rx_stats.alloc_page_failed;
5711 /* update XDP Tx rings counters */
5712 if (ice_is_xdp_ena_vsi(vsi))
5713 ice_update_vsi_tx_ring_stats(vsi, vsi->xdp_rings,
5720 * ice_update_vsi_stats - Update VSI stats counters
5721 * @vsi: the VSI to be updated
5723 void ice_update_vsi_stats(struct ice_vsi *vsi)
5725 struct rtnl_link_stats64 *cur_ns = &vsi->net_stats;
5726 struct ice_eth_stats *cur_es = &vsi->eth_stats;
5727 struct ice_pf *pf = vsi->back;
5729 if (test_bit(ICE_VSI_DOWN, vsi->state) ||
5730 test_bit(ICE_CFG_BUSY, pf->state))
5733 /* get stats as recorded by Tx/Rx rings */
5734 ice_update_vsi_ring_stats(vsi);
5736 /* get VSI stats as recorded by the hardware */
5737 ice_update_eth_stats(vsi);
5739 cur_ns->tx_errors = cur_es->tx_errors;
5740 cur_ns->rx_dropped = cur_es->rx_discards;
5741 cur_ns->tx_dropped = cur_es->tx_discards;
5742 cur_ns->multicast = cur_es->rx_multicast;
5744 /* update some more netdev stats if this is main VSI */
5745 if (vsi->type == ICE_VSI_PF) {
5746 cur_ns->rx_crc_errors = pf->stats.crc_errors;
5747 cur_ns->rx_errors = pf->stats.crc_errors +
5748 pf->stats.illegal_bytes +
5749 pf->stats.rx_len_errors +
5750 pf->stats.rx_undersize +
5751 pf->hw_csum_rx_error +
5752 pf->stats.rx_jabber +
5753 pf->stats.rx_fragments +
5754 pf->stats.rx_oversize;
5755 cur_ns->rx_length_errors = pf->stats.rx_len_errors;
5756 /* record drops from the port level */
5757 cur_ns->rx_missed_errors = pf->stats.eth.rx_discards;
5762 * ice_update_pf_stats - Update PF port stats counters
5763 * @pf: PF whose stats needs to be updated
5765 void ice_update_pf_stats(struct ice_pf *pf)
5767 struct ice_hw_port_stats *prev_ps, *cur_ps;
5768 struct ice_hw *hw = &pf->hw;
5772 port = hw->port_info->lport;
5773 prev_ps = &pf->stats_prev;
5774 cur_ps = &pf->stats;
5776 ice_stat_update40(hw, GLPRT_GORCL(port), pf->stat_prev_loaded,
5777 &prev_ps->eth.rx_bytes,
5778 &cur_ps->eth.rx_bytes);
5780 ice_stat_update40(hw, GLPRT_UPRCL(port), pf->stat_prev_loaded,
5781 &prev_ps->eth.rx_unicast,
5782 &cur_ps->eth.rx_unicast);
5784 ice_stat_update40(hw, GLPRT_MPRCL(port), pf->stat_prev_loaded,
5785 &prev_ps->eth.rx_multicast,
5786 &cur_ps->eth.rx_multicast);
5788 ice_stat_update40(hw, GLPRT_BPRCL(port), pf->stat_prev_loaded,
5789 &prev_ps->eth.rx_broadcast,
5790 &cur_ps->eth.rx_broadcast);
5792 ice_stat_update32(hw, PRTRPB_RDPC, pf->stat_prev_loaded,
5793 &prev_ps->eth.rx_discards,
5794 &cur_ps->eth.rx_discards);
5796 ice_stat_update40(hw, GLPRT_GOTCL(port), pf->stat_prev_loaded,
5797 &prev_ps->eth.tx_bytes,
5798 &cur_ps->eth.tx_bytes);
5800 ice_stat_update40(hw, GLPRT_UPTCL(port), pf->stat_prev_loaded,
5801 &prev_ps->eth.tx_unicast,
5802 &cur_ps->eth.tx_unicast);
5804 ice_stat_update40(hw, GLPRT_MPTCL(port), pf->stat_prev_loaded,
5805 &prev_ps->eth.tx_multicast,
5806 &cur_ps->eth.tx_multicast);
5808 ice_stat_update40(hw, GLPRT_BPTCL(port), pf->stat_prev_loaded,
5809 &prev_ps->eth.tx_broadcast,
5810 &cur_ps->eth.tx_broadcast);
5812 ice_stat_update32(hw, GLPRT_TDOLD(port), pf->stat_prev_loaded,
5813 &prev_ps->tx_dropped_link_down,
5814 &cur_ps->tx_dropped_link_down);
5816 ice_stat_update40(hw, GLPRT_PRC64L(port), pf->stat_prev_loaded,
5817 &prev_ps->rx_size_64, &cur_ps->rx_size_64);
5819 ice_stat_update40(hw, GLPRT_PRC127L(port), pf->stat_prev_loaded,
5820 &prev_ps->rx_size_127, &cur_ps->rx_size_127);
5822 ice_stat_update40(hw, GLPRT_PRC255L(port), pf->stat_prev_loaded,
5823 &prev_ps->rx_size_255, &cur_ps->rx_size_255);
5825 ice_stat_update40(hw, GLPRT_PRC511L(port), pf->stat_prev_loaded,
5826 &prev_ps->rx_size_511, &cur_ps->rx_size_511);
5828 ice_stat_update40(hw, GLPRT_PRC1023L(port), pf->stat_prev_loaded,
5829 &prev_ps->rx_size_1023, &cur_ps->rx_size_1023);
5831 ice_stat_update40(hw, GLPRT_PRC1522L(port), pf->stat_prev_loaded,
5832 &prev_ps->rx_size_1522, &cur_ps->rx_size_1522);
5834 ice_stat_update40(hw, GLPRT_PRC9522L(port), pf->stat_prev_loaded,
5835 &prev_ps->rx_size_big, &cur_ps->rx_size_big);
5837 ice_stat_update40(hw, GLPRT_PTC64L(port), pf->stat_prev_loaded,
5838 &prev_ps->tx_size_64, &cur_ps->tx_size_64);
5840 ice_stat_update40(hw, GLPRT_PTC127L(port), pf->stat_prev_loaded,
5841 &prev_ps->tx_size_127, &cur_ps->tx_size_127);
5843 ice_stat_update40(hw, GLPRT_PTC255L(port), pf->stat_prev_loaded,
5844 &prev_ps->tx_size_255, &cur_ps->tx_size_255);
5846 ice_stat_update40(hw, GLPRT_PTC511L(port), pf->stat_prev_loaded,
5847 &prev_ps->tx_size_511, &cur_ps->tx_size_511);
5849 ice_stat_update40(hw, GLPRT_PTC1023L(port), pf->stat_prev_loaded,
5850 &prev_ps->tx_size_1023, &cur_ps->tx_size_1023);
5852 ice_stat_update40(hw, GLPRT_PTC1522L(port), pf->stat_prev_loaded,
5853 &prev_ps->tx_size_1522, &cur_ps->tx_size_1522);
5855 ice_stat_update40(hw, GLPRT_PTC9522L(port), pf->stat_prev_loaded,
5856 &prev_ps->tx_size_big, &cur_ps->tx_size_big);
5858 fd_ctr_base = hw->fd_ctr_base;
5860 ice_stat_update40(hw,
5861 GLSTAT_FD_CNT0L(ICE_FD_SB_STAT_IDX(fd_ctr_base)),
5862 pf->stat_prev_loaded, &prev_ps->fd_sb_match,
5863 &cur_ps->fd_sb_match);
5864 ice_stat_update32(hw, GLPRT_LXONRXC(port), pf->stat_prev_loaded,
5865 &prev_ps->link_xon_rx, &cur_ps->link_xon_rx);
5867 ice_stat_update32(hw, GLPRT_LXOFFRXC(port), pf->stat_prev_loaded,
5868 &prev_ps->link_xoff_rx, &cur_ps->link_xoff_rx);
5870 ice_stat_update32(hw, GLPRT_LXONTXC(port), pf->stat_prev_loaded,
5871 &prev_ps->link_xon_tx, &cur_ps->link_xon_tx);
5873 ice_stat_update32(hw, GLPRT_LXOFFTXC(port), pf->stat_prev_loaded,
5874 &prev_ps->link_xoff_tx, &cur_ps->link_xoff_tx);
5876 ice_update_dcb_stats(pf);
5878 ice_stat_update32(hw, GLPRT_CRCERRS(port), pf->stat_prev_loaded,
5879 &prev_ps->crc_errors, &cur_ps->crc_errors);
5881 ice_stat_update32(hw, GLPRT_ILLERRC(port), pf->stat_prev_loaded,
5882 &prev_ps->illegal_bytes, &cur_ps->illegal_bytes);
5884 ice_stat_update32(hw, GLPRT_MLFC(port), pf->stat_prev_loaded,
5885 &prev_ps->mac_local_faults,
5886 &cur_ps->mac_local_faults);
5888 ice_stat_update32(hw, GLPRT_MRFC(port), pf->stat_prev_loaded,
5889 &prev_ps->mac_remote_faults,
5890 &cur_ps->mac_remote_faults);
5892 ice_stat_update32(hw, GLPRT_RLEC(port), pf->stat_prev_loaded,
5893 &prev_ps->rx_len_errors, &cur_ps->rx_len_errors);
5895 ice_stat_update32(hw, GLPRT_RUC(port), pf->stat_prev_loaded,
5896 &prev_ps->rx_undersize, &cur_ps->rx_undersize);
5898 ice_stat_update32(hw, GLPRT_RFC(port), pf->stat_prev_loaded,
5899 &prev_ps->rx_fragments, &cur_ps->rx_fragments);
5901 ice_stat_update32(hw, GLPRT_ROC(port), pf->stat_prev_loaded,
5902 &prev_ps->rx_oversize, &cur_ps->rx_oversize);
5904 ice_stat_update32(hw, GLPRT_RJC(port), pf->stat_prev_loaded,
5905 &prev_ps->rx_jabber, &cur_ps->rx_jabber);
5907 cur_ps->fd_sb_status = test_bit(ICE_FLAG_FD_ENA, pf->flags) ? 1 : 0;
5909 pf->stat_prev_loaded = true;
5913 * ice_get_stats64 - get statistics for network device structure
5914 * @netdev: network interface device structure
5915 * @stats: main device statistics structure
5918 void ice_get_stats64(struct net_device *netdev, struct rtnl_link_stats64 *stats)
5920 struct ice_netdev_priv *np = netdev_priv(netdev);
5921 struct rtnl_link_stats64 *vsi_stats;
5922 struct ice_vsi *vsi = np->vsi;
5924 vsi_stats = &vsi->net_stats;
5926 if (!vsi->num_txq || !vsi->num_rxq)
5929 /* netdev packet/byte stats come from ring counter. These are obtained
5930 * by summing up ring counters (done by ice_update_vsi_ring_stats).
5931 * But, only call the update routine and read the registers if VSI is
5934 if (!test_bit(ICE_VSI_DOWN, vsi->state))
5935 ice_update_vsi_ring_stats(vsi);
5936 stats->tx_packets = vsi_stats->tx_packets;
5937 stats->tx_bytes = vsi_stats->tx_bytes;
5938 stats->rx_packets = vsi_stats->rx_packets;
5939 stats->rx_bytes = vsi_stats->rx_bytes;
5941 /* The rest of the stats can be read from the hardware but instead we
5942 * just return values that the watchdog task has already obtained from
5945 stats->multicast = vsi_stats->multicast;
5946 stats->tx_errors = vsi_stats->tx_errors;
5947 stats->tx_dropped = vsi_stats->tx_dropped;
5948 stats->rx_errors = vsi_stats->rx_errors;
5949 stats->rx_dropped = vsi_stats->rx_dropped;
5950 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
5951 stats->rx_length_errors = vsi_stats->rx_length_errors;
5955 * ice_napi_disable_all - Disable NAPI for all q_vectors in the VSI
5956 * @vsi: VSI having NAPI disabled
5958 static void ice_napi_disable_all(struct ice_vsi *vsi)
5965 ice_for_each_q_vector(vsi, q_idx) {
5966 struct ice_q_vector *q_vector = vsi->q_vectors[q_idx];
5968 if (q_vector->rx.ring || q_vector->tx.ring)
5969 napi_disable(&q_vector->napi);
5971 cancel_work_sync(&q_vector->tx.dim.work);
5972 cancel_work_sync(&q_vector->rx.dim.work);
5977 * ice_down - Shutdown the connection
5978 * @vsi: The VSI being stopped
5980 int ice_down(struct ice_vsi *vsi)
5982 int i, tx_err, rx_err, link_err = 0;
5984 /* Caller of this function is expected to set the
5985 * vsi->state ICE_DOWN bit
5988 netif_carrier_off(vsi->netdev);
5989 netif_tx_disable(vsi->netdev);
5992 ice_vsi_dis_irq(vsi);
5994 tx_err = ice_vsi_stop_lan_tx_rings(vsi, ICE_NO_RESET, 0);
5996 netdev_err(vsi->netdev, "Failed stop Tx rings, VSI %d error %d\n",
5997 vsi->vsi_num, tx_err);
5998 if (!tx_err && ice_is_xdp_ena_vsi(vsi)) {
5999 tx_err = ice_vsi_stop_xdp_tx_rings(vsi);
6001 netdev_err(vsi->netdev, "Failed stop XDP rings, VSI %d error %d\n",
6002 vsi->vsi_num, tx_err);
6005 rx_err = ice_vsi_stop_all_rx_rings(vsi);
6007 netdev_err(vsi->netdev, "Failed stop Rx rings, VSI %d error %d\n",
6008 vsi->vsi_num, rx_err);
6010 ice_napi_disable_all(vsi);
6012 if (test_bit(ICE_FLAG_LINK_DOWN_ON_CLOSE_ENA, vsi->back->flags)) {
6013 link_err = ice_force_phys_link_state(vsi, false);
6015 netdev_err(vsi->netdev, "Failed to set physical link down, VSI %d error %d\n",
6016 vsi->vsi_num, link_err);
6019 ice_for_each_txq(vsi, i)
6020 ice_clean_tx_ring(vsi->tx_rings[i]);
6022 ice_for_each_rxq(vsi, i)
6023 ice_clean_rx_ring(vsi->rx_rings[i]);
6025 if (tx_err || rx_err || link_err) {
6026 netdev_err(vsi->netdev, "Failed to close VSI 0x%04X on switch 0x%04X\n",
6027 vsi->vsi_num, vsi->vsw->sw_id);
6035 * ice_vsi_setup_tx_rings - Allocate VSI Tx queue resources
6036 * @vsi: VSI having resources allocated
6038 * Return 0 on success, negative on failure
6040 int ice_vsi_setup_tx_rings(struct ice_vsi *vsi)
6044 if (!vsi->num_txq) {
6045 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Tx queues\n",
6050 ice_for_each_txq(vsi, i) {
6051 struct ice_ring *ring = vsi->tx_rings[i];
6056 ring->netdev = vsi->netdev;
6057 err = ice_setup_tx_ring(ring);
6066 * ice_vsi_setup_rx_rings - Allocate VSI Rx queue resources
6067 * @vsi: VSI having resources allocated
6069 * Return 0 on success, negative on failure
6071 int ice_vsi_setup_rx_rings(struct ice_vsi *vsi)
6075 if (!vsi->num_rxq) {
6076 dev_err(ice_pf_to_dev(vsi->back), "VSI %d has 0 Rx queues\n",
6081 ice_for_each_rxq(vsi, i) {
6082 struct ice_ring *ring = vsi->rx_rings[i];
6087 ring->netdev = vsi->netdev;
6088 err = ice_setup_rx_ring(ring);
6097 * ice_vsi_open_ctrl - open control VSI for use
6098 * @vsi: the VSI to open
6100 * Initialization of the Control VSI
6102 * Returns 0 on success, negative value on error
6104 int ice_vsi_open_ctrl(struct ice_vsi *vsi)
6106 char int_name[ICE_INT_NAME_STR_LEN];
6107 struct ice_pf *pf = vsi->back;
6111 dev = ice_pf_to_dev(pf);
6112 /* allocate descriptors */
6113 err = ice_vsi_setup_tx_rings(vsi);
6117 err = ice_vsi_setup_rx_rings(vsi);
6121 err = ice_vsi_cfg(vsi);
6125 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:ctrl",
6126 dev_driver_string(dev), dev_name(dev));
6127 err = ice_vsi_req_irq_msix(vsi, int_name);
6131 ice_vsi_cfg_msix(vsi);
6133 err = ice_vsi_start_all_rx_rings(vsi);
6135 goto err_up_complete;
6137 clear_bit(ICE_VSI_DOWN, vsi->state);
6138 ice_vsi_ena_irq(vsi);
6145 ice_vsi_free_rx_rings(vsi);
6147 ice_vsi_free_tx_rings(vsi);
6153 * ice_vsi_open - Called when a network interface is made active
6154 * @vsi: the VSI to open
6156 * Initialization of the VSI
6158 * Returns 0 on success, negative value on error
6160 static int ice_vsi_open(struct ice_vsi *vsi)
6162 char int_name[ICE_INT_NAME_STR_LEN];
6163 struct ice_pf *pf = vsi->back;
6166 /* allocate descriptors */
6167 err = ice_vsi_setup_tx_rings(vsi);
6171 err = ice_vsi_setup_rx_rings(vsi);
6175 err = ice_vsi_cfg(vsi);
6179 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
6180 dev_driver_string(ice_pf_to_dev(pf)), vsi->netdev->name);
6181 err = ice_vsi_req_irq_msix(vsi, int_name);
6185 /* Notify the stack of the actual queue counts. */
6186 err = netif_set_real_num_tx_queues(vsi->netdev, vsi->num_txq);
6190 err = netif_set_real_num_rx_queues(vsi->netdev, vsi->num_rxq);
6194 err = ice_up_complete(vsi);
6196 goto err_up_complete;
6203 ice_vsi_free_irq(vsi);
6205 ice_vsi_free_rx_rings(vsi);
6207 ice_vsi_free_tx_rings(vsi);
6213 * ice_vsi_release_all - Delete all VSIs
6214 * @pf: PF from which all VSIs are being removed
6216 static void ice_vsi_release_all(struct ice_pf *pf)
6223 ice_for_each_vsi(pf, i) {
6227 err = ice_vsi_release(pf->vsi[i]);
6229 dev_dbg(ice_pf_to_dev(pf), "Failed to release pf->vsi[%d], err %d, vsi_num = %d\n",
6230 i, err, pf->vsi[i]->vsi_num);
6235 * ice_vsi_rebuild_by_type - Rebuild VSI of a given type
6236 * @pf: pointer to the PF instance
6237 * @type: VSI type to rebuild
6239 * Iterates through the pf->vsi array and rebuilds VSIs of the requested type
6241 static int ice_vsi_rebuild_by_type(struct ice_pf *pf, enum ice_vsi_type type)
6243 struct device *dev = ice_pf_to_dev(pf);
6244 enum ice_status status;
6247 ice_for_each_vsi(pf, i) {
6248 struct ice_vsi *vsi = pf->vsi[i];
6250 if (!vsi || vsi->type != type)
6253 /* rebuild the VSI */
6254 err = ice_vsi_rebuild(vsi, true);
6256 dev_err(dev, "rebuild VSI failed, err %d, VSI index %d, type %s\n",
6257 err, vsi->idx, ice_vsi_type_str(type));
6261 /* replay filters for the VSI */
6262 status = ice_replay_vsi(&pf->hw, vsi->idx);
6264 dev_err(dev, "replay VSI failed, status %s, VSI index %d, type %s\n",
6265 ice_stat_str(status), vsi->idx,
6266 ice_vsi_type_str(type));
6270 /* Re-map HW VSI number, using VSI handle that has been
6271 * previously validated in ice_replay_vsi() call above
6273 vsi->vsi_num = ice_get_hw_vsi_num(&pf->hw, vsi->idx);
6275 /* enable the VSI */
6276 err = ice_ena_vsi(vsi, false);
6278 dev_err(dev, "enable VSI failed, err %d, VSI index %d, type %s\n",
6279 err, vsi->idx, ice_vsi_type_str(type));
6283 dev_info(dev, "VSI rebuilt. VSI index %d, type %s\n", vsi->idx,
6284 ice_vsi_type_str(type));
6291 * ice_update_pf_netdev_link - Update PF netdev link status
6292 * @pf: pointer to the PF instance
6294 static void ice_update_pf_netdev_link(struct ice_pf *pf)
6299 ice_for_each_vsi(pf, i) {
6300 struct ice_vsi *vsi = pf->vsi[i];
6302 if (!vsi || vsi->type != ICE_VSI_PF)
6305 ice_get_link_status(pf->vsi[i]->port_info, &link_up);
6307 netif_carrier_on(pf->vsi[i]->netdev);
6308 netif_tx_wake_all_queues(pf->vsi[i]->netdev);
6310 netif_carrier_off(pf->vsi[i]->netdev);
6311 netif_tx_stop_all_queues(pf->vsi[i]->netdev);
6317 * ice_rebuild - rebuild after reset
6318 * @pf: PF to rebuild
6319 * @reset_type: type of reset
6321 * Do not rebuild VF VSI in this flow because that is already handled via
6322 * ice_reset_all_vfs(). This is because requirements for resetting a VF after a
6323 * PFR/CORER/GLOBER/etc. are different than the normal flow. Also, we don't want
6324 * to reset/rebuild all the VF VSI twice.
6326 static void ice_rebuild(struct ice_pf *pf, enum ice_reset_req reset_type)
6328 struct device *dev = ice_pf_to_dev(pf);
6329 struct ice_hw *hw = &pf->hw;
6330 enum ice_status ret;
6333 if (test_bit(ICE_DOWN, pf->state))
6334 goto clear_recovery;
6336 dev_dbg(dev, "rebuilding PF after reset_type=%d\n", reset_type);
6338 ret = ice_init_all_ctrlq(hw);
6340 dev_err(dev, "control queues init failed %s\n",
6342 goto err_init_ctrlq;
6345 /* if DDP was previously loaded successfully */
6346 if (!ice_is_safe_mode(pf)) {
6347 /* reload the SW DB of filter tables */
6348 if (reset_type == ICE_RESET_PFR)
6349 ice_fill_blk_tbls(hw);
6351 /* Reload DDP Package after CORER/GLOBR reset */
6352 ice_load_pkg(NULL, pf);
6355 ret = ice_clear_pf_cfg(hw);
6357 dev_err(dev, "clear PF configuration failed %s\n",
6359 goto err_init_ctrlq;
6362 if (pf->first_sw->dflt_vsi_ena)
6363 dev_info(dev, "Clearing default VSI, re-enable after reset completes\n");
6364 /* clear the default VSI configuration if it exists */
6365 pf->first_sw->dflt_vsi = NULL;
6366 pf->first_sw->dflt_vsi_ena = false;
6368 ice_clear_pxe_mode(hw);
6370 ret = ice_init_nvm(hw);
6372 dev_err(dev, "ice_init_nvm failed %s\n", ice_stat_str(ret));
6373 goto err_init_ctrlq;
6376 ret = ice_get_caps(hw);
6378 dev_err(dev, "ice_get_caps failed %s\n", ice_stat_str(ret));
6379 goto err_init_ctrlq;
6382 ret = ice_aq_set_mac_cfg(hw, ICE_AQ_SET_MAC_FRAME_SIZE_MAX, NULL);
6384 dev_err(dev, "set_mac_cfg failed %s\n", ice_stat_str(ret));
6385 goto err_init_ctrlq;
6388 err = ice_sched_init_port(hw->port_info);
6390 goto err_sched_init_port;
6392 /* start misc vector */
6393 err = ice_req_irq_msix_misc(pf);
6395 dev_err(dev, "misc vector setup failed: %d\n", err);
6396 goto err_sched_init_port;
6399 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6400 wr32(hw, PFQF_FD_ENA, PFQF_FD_ENA_FD_ENA_M);
6401 if (!rd32(hw, PFQF_FD_SIZE)) {
6402 u16 unused, guar, b_effort;
6404 guar = hw->func_caps.fd_fltr_guar;
6405 b_effort = hw->func_caps.fd_fltr_best_effort;
6407 /* force guaranteed filter pool for PF */
6408 ice_alloc_fd_guar_item(hw, &unused, guar);
6409 /* force shared filter pool for PF */
6410 ice_alloc_fd_shrd_item(hw, &unused, b_effort);
6414 if (test_bit(ICE_FLAG_DCB_ENA, pf->flags))
6415 ice_dcb_rebuild(pf);
6417 /* If the PF previously had enabled PTP, PTP init needs to happen before
6418 * the VSI rebuild. If not, this causes the PTP link status events to
6421 if (test_bit(ICE_FLAG_PTP_SUPPORTED, pf->flags))
6424 /* rebuild PF VSI */
6425 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_PF);
6427 dev_err(dev, "PF VSI rebuild failed: %d\n", err);
6428 goto err_vsi_rebuild;
6431 /* If Flow Director is active */
6432 if (test_bit(ICE_FLAG_FD_ENA, pf->flags)) {
6433 err = ice_vsi_rebuild_by_type(pf, ICE_VSI_CTRL);
6435 dev_err(dev, "control VSI rebuild failed: %d\n", err);
6436 goto err_vsi_rebuild;
6439 /* replay HW Flow Director recipes */
6441 ice_fdir_replay_flows(hw);
6443 /* replay Flow Director filters */
6444 ice_fdir_replay_fltrs(pf);
6446 ice_rebuild_arfs(pf);
6449 ice_update_pf_netdev_link(pf);
6451 /* tell the firmware we are up */
6452 ret = ice_send_version(pf);
6454 dev_err(dev, "Rebuild failed due to error sending driver version: %s\n",
6456 goto err_vsi_rebuild;
6459 ice_replay_post(hw);
6461 /* if we get here, reset flow is successful */
6462 clear_bit(ICE_RESET_FAILED, pf->state);
6464 ice_plug_aux_dev(pf);
6468 err_sched_init_port:
6469 ice_sched_cleanup_all(hw);
6471 ice_shutdown_all_ctrlq(hw);
6472 set_bit(ICE_RESET_FAILED, pf->state);
6474 /* set this bit in PF state to control service task scheduling */
6475 set_bit(ICE_NEEDS_RESTART, pf->state);
6476 dev_err(dev, "Rebuild failed, unload and reload driver\n");
6480 * ice_max_xdp_frame_size - returns the maximum allowed frame size for XDP
6481 * @vsi: Pointer to VSI structure
6483 static int ice_max_xdp_frame_size(struct ice_vsi *vsi)
6485 if (PAGE_SIZE >= 8192 || test_bit(ICE_FLAG_LEGACY_RX, vsi->back->flags))
6486 return ICE_RXBUF_2048 - XDP_PACKET_HEADROOM;
6488 return ICE_RXBUF_3072;
6492 * ice_change_mtu - NDO callback to change the MTU
6493 * @netdev: network interface device structure
6494 * @new_mtu: new value for maximum frame size
6496 * Returns 0 on success, negative on failure
6498 static int ice_change_mtu(struct net_device *netdev, int new_mtu)
6500 struct ice_netdev_priv *np = netdev_priv(netdev);
6501 struct ice_vsi *vsi = np->vsi;
6502 struct ice_pf *pf = vsi->back;
6503 struct iidc_event *event;
6507 if (new_mtu == (int)netdev->mtu) {
6508 netdev_warn(netdev, "MTU is already %u\n", netdev->mtu);
6512 if (ice_is_xdp_ena_vsi(vsi)) {
6513 int frame_size = ice_max_xdp_frame_size(vsi);
6515 if (new_mtu + ICE_ETH_PKT_HDR_PAD > frame_size) {
6516 netdev_err(netdev, "max MTU for XDP usage is %d\n",
6517 frame_size - ICE_ETH_PKT_HDR_PAD);
6522 /* if a reset is in progress, wait for some time for it to complete */
6524 if (ice_is_reset_in_progress(pf->state)) {
6526 usleep_range(1000, 2000);
6531 } while (count < 100);
6534 netdev_err(netdev, "can't change MTU. Device is busy\n");
6538 event = kzalloc(sizeof(*event), GFP_KERNEL);
6542 set_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6543 ice_send_event_to_aux(pf, event);
6544 clear_bit(IIDC_EVENT_BEFORE_MTU_CHANGE, event->type);
6546 netdev->mtu = (unsigned int)new_mtu;
6548 /* if VSI is up, bring it down and then back up */
6549 if (!test_and_set_bit(ICE_VSI_DOWN, vsi->state)) {
6550 err = ice_down(vsi);
6552 netdev_err(netdev, "change MTU if_down err %d\n", err);
6558 netdev_err(netdev, "change MTU if_up err %d\n", err);
6563 netdev_dbg(netdev, "changed MTU to %d\n", new_mtu);
6565 set_bit(IIDC_EVENT_AFTER_MTU_CHANGE, event->type);
6566 ice_send_event_to_aux(pf, event);
6573 * ice_do_ioctl - Access the hwtstamp interface
6574 * @netdev: network interface device structure
6575 * @ifr: interface request data
6576 * @cmd: ioctl command
6578 static int ice_do_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6580 struct ice_netdev_priv *np = netdev_priv(netdev);
6581 struct ice_pf *pf = np->vsi->back;
6585 return ice_ptp_get_ts_config(pf, ifr);
6587 return ice_ptp_set_ts_config(pf, ifr);
6594 * ice_aq_str - convert AQ err code to a string
6595 * @aq_err: the AQ error code to convert
6597 const char *ice_aq_str(enum ice_aq_err aq_err)
6602 case ICE_AQ_RC_EPERM:
6603 return "ICE_AQ_RC_EPERM";
6604 case ICE_AQ_RC_ENOENT:
6605 return "ICE_AQ_RC_ENOENT";
6606 case ICE_AQ_RC_ENOMEM:
6607 return "ICE_AQ_RC_ENOMEM";
6608 case ICE_AQ_RC_EBUSY:
6609 return "ICE_AQ_RC_EBUSY";
6610 case ICE_AQ_RC_EEXIST:
6611 return "ICE_AQ_RC_EEXIST";
6612 case ICE_AQ_RC_EINVAL:
6613 return "ICE_AQ_RC_EINVAL";
6614 case ICE_AQ_RC_ENOSPC:
6615 return "ICE_AQ_RC_ENOSPC";
6616 case ICE_AQ_RC_ENOSYS:
6617 return "ICE_AQ_RC_ENOSYS";
6618 case ICE_AQ_RC_EMODE:
6619 return "ICE_AQ_RC_EMODE";
6620 case ICE_AQ_RC_ENOSEC:
6621 return "ICE_AQ_RC_ENOSEC";
6622 case ICE_AQ_RC_EBADSIG:
6623 return "ICE_AQ_RC_EBADSIG";
6624 case ICE_AQ_RC_ESVN:
6625 return "ICE_AQ_RC_ESVN";
6626 case ICE_AQ_RC_EBADMAN:
6627 return "ICE_AQ_RC_EBADMAN";
6628 case ICE_AQ_RC_EBADBUF:
6629 return "ICE_AQ_RC_EBADBUF";
6632 return "ICE_AQ_RC_UNKNOWN";
6636 * ice_stat_str - convert status err code to a string
6637 * @stat_err: the status error code to convert
6639 const char *ice_stat_str(enum ice_status stat_err)
6645 return "ICE_ERR_PARAM";
6646 case ICE_ERR_NOT_IMPL:
6647 return "ICE_ERR_NOT_IMPL";
6648 case ICE_ERR_NOT_READY:
6649 return "ICE_ERR_NOT_READY";
6650 case ICE_ERR_NOT_SUPPORTED:
6651 return "ICE_ERR_NOT_SUPPORTED";
6652 case ICE_ERR_BAD_PTR:
6653 return "ICE_ERR_BAD_PTR";
6654 case ICE_ERR_INVAL_SIZE:
6655 return "ICE_ERR_INVAL_SIZE";
6656 case ICE_ERR_DEVICE_NOT_SUPPORTED:
6657 return "ICE_ERR_DEVICE_NOT_SUPPORTED";
6658 case ICE_ERR_RESET_FAILED:
6659 return "ICE_ERR_RESET_FAILED";
6660 case ICE_ERR_FW_API_VER:
6661 return "ICE_ERR_FW_API_VER";
6662 case ICE_ERR_NO_MEMORY:
6663 return "ICE_ERR_NO_MEMORY";
6665 return "ICE_ERR_CFG";
6666 case ICE_ERR_OUT_OF_RANGE:
6667 return "ICE_ERR_OUT_OF_RANGE";
6668 case ICE_ERR_ALREADY_EXISTS:
6669 return "ICE_ERR_ALREADY_EXISTS";
6671 return "ICE_ERR_NVM";
6672 case ICE_ERR_NVM_CHECKSUM:
6673 return "ICE_ERR_NVM_CHECKSUM";
6674 case ICE_ERR_BUF_TOO_SHORT:
6675 return "ICE_ERR_BUF_TOO_SHORT";
6676 case ICE_ERR_NVM_BLANK_MODE:
6677 return "ICE_ERR_NVM_BLANK_MODE";
6678 case ICE_ERR_IN_USE:
6679 return "ICE_ERR_IN_USE";
6680 case ICE_ERR_MAX_LIMIT:
6681 return "ICE_ERR_MAX_LIMIT";
6682 case ICE_ERR_RESET_ONGOING:
6683 return "ICE_ERR_RESET_ONGOING";
6684 case ICE_ERR_HW_TABLE:
6685 return "ICE_ERR_HW_TABLE";
6686 case ICE_ERR_DOES_NOT_EXIST:
6687 return "ICE_ERR_DOES_NOT_EXIST";
6688 case ICE_ERR_FW_DDP_MISMATCH:
6689 return "ICE_ERR_FW_DDP_MISMATCH";
6690 case ICE_ERR_AQ_ERROR:
6691 return "ICE_ERR_AQ_ERROR";
6692 case ICE_ERR_AQ_TIMEOUT:
6693 return "ICE_ERR_AQ_TIMEOUT";
6694 case ICE_ERR_AQ_FULL:
6695 return "ICE_ERR_AQ_FULL";
6696 case ICE_ERR_AQ_NO_WORK:
6697 return "ICE_ERR_AQ_NO_WORK";
6698 case ICE_ERR_AQ_EMPTY:
6699 return "ICE_ERR_AQ_EMPTY";
6700 case ICE_ERR_AQ_FW_CRITICAL:
6701 return "ICE_ERR_AQ_FW_CRITICAL";
6704 return "ICE_ERR_UNKNOWN";
6708 * ice_set_rss_lut - Set RSS LUT
6709 * @vsi: Pointer to VSI structure
6710 * @lut: Lookup table
6711 * @lut_size: Lookup table size
6713 * Returns 0 on success, negative on failure
6715 int ice_set_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6717 struct ice_aq_get_set_rss_lut_params params = {};
6718 struct ice_hw *hw = &vsi->back->hw;
6719 enum ice_status status;
6724 params.vsi_handle = vsi->idx;
6725 params.lut_size = lut_size;
6726 params.lut_type = vsi->rss_lut_type;
6729 status = ice_aq_set_rss_lut(hw, ¶ms);
6731 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS lut, err %s aq_err %s\n",
6732 ice_stat_str(status),
6733 ice_aq_str(hw->adminq.sq_last_status));
6741 * ice_set_rss_key - Set RSS key
6742 * @vsi: Pointer to the VSI structure
6743 * @seed: RSS hash seed
6745 * Returns 0 on success, negative on failure
6747 int ice_set_rss_key(struct ice_vsi *vsi, u8 *seed)
6749 struct ice_hw *hw = &vsi->back->hw;
6750 enum ice_status status;
6755 status = ice_aq_set_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6757 dev_err(ice_pf_to_dev(vsi->back), "Cannot set RSS key, err %s aq_err %s\n",
6758 ice_stat_str(status),
6759 ice_aq_str(hw->adminq.sq_last_status));
6767 * ice_get_rss_lut - Get RSS LUT
6768 * @vsi: Pointer to VSI structure
6769 * @lut: Buffer to store the lookup table entries
6770 * @lut_size: Size of buffer to store the lookup table entries
6772 * Returns 0 on success, negative on failure
6774 int ice_get_rss_lut(struct ice_vsi *vsi, u8 *lut, u16 lut_size)
6776 struct ice_aq_get_set_rss_lut_params params = {};
6777 struct ice_hw *hw = &vsi->back->hw;
6778 enum ice_status status;
6783 params.vsi_handle = vsi->idx;
6784 params.lut_size = lut_size;
6785 params.lut_type = vsi->rss_lut_type;
6788 status = ice_aq_get_rss_lut(hw, ¶ms);
6790 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS lut, err %s aq_err %s\n",
6791 ice_stat_str(status),
6792 ice_aq_str(hw->adminq.sq_last_status));
6800 * ice_get_rss_key - Get RSS key
6801 * @vsi: Pointer to VSI structure
6802 * @seed: Buffer to store the key in
6804 * Returns 0 on success, negative on failure
6806 int ice_get_rss_key(struct ice_vsi *vsi, u8 *seed)
6808 struct ice_hw *hw = &vsi->back->hw;
6809 enum ice_status status;
6814 status = ice_aq_get_rss_key(hw, vsi->idx, (struct ice_aqc_get_set_rss_keys *)seed);
6816 dev_err(ice_pf_to_dev(vsi->back), "Cannot get RSS key, err %s aq_err %s\n",
6817 ice_stat_str(status),
6818 ice_aq_str(hw->adminq.sq_last_status));
6826 * ice_bridge_getlink - Get the hardware bridge mode
6829 * @seq: RTNL message seq
6830 * @dev: the netdev being configured
6831 * @filter_mask: filter mask passed in
6832 * @nlflags: netlink flags passed in
6834 * Return the bridge mode (VEB/VEPA)
6837 ice_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
6838 struct net_device *dev, u32 filter_mask, int nlflags)
6840 struct ice_netdev_priv *np = netdev_priv(dev);
6841 struct ice_vsi *vsi = np->vsi;
6842 struct ice_pf *pf = vsi->back;
6845 bmode = pf->first_sw->bridge_mode;
6847 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, bmode, 0, 0, nlflags,
6852 * ice_vsi_update_bridge_mode - Update VSI for switching bridge mode (VEB/VEPA)
6853 * @vsi: Pointer to VSI structure
6854 * @bmode: Hardware bridge mode (VEB/VEPA)
6856 * Returns 0 on success, negative on failure
6858 static int ice_vsi_update_bridge_mode(struct ice_vsi *vsi, u16 bmode)
6860 struct ice_aqc_vsi_props *vsi_props;
6861 struct ice_hw *hw = &vsi->back->hw;
6862 struct ice_vsi_ctx *ctxt;
6863 enum ice_status status;
6866 vsi_props = &vsi->info;
6868 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL);
6872 ctxt->info = vsi->info;
6874 if (bmode == BRIDGE_MODE_VEB)
6875 /* change from VEPA to VEB mode */
6876 ctxt->info.sw_flags |= ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6878 /* change from VEB to VEPA mode */
6879 ctxt->info.sw_flags &= ~ICE_AQ_VSI_SW_FLAG_ALLOW_LB;
6880 ctxt->info.valid_sections = cpu_to_le16(ICE_AQ_VSI_PROP_SW_VALID);
6882 status = ice_update_vsi(hw, vsi->idx, ctxt, NULL);
6884 dev_err(ice_pf_to_dev(vsi->back), "update VSI for bridge mode failed, bmode = %d err %s aq_err %s\n",
6885 bmode, ice_stat_str(status),
6886 ice_aq_str(hw->adminq.sq_last_status));
6890 /* Update sw flags for book keeping */
6891 vsi_props->sw_flags = ctxt->info.sw_flags;
6899 * ice_bridge_setlink - Set the hardware bridge mode
6900 * @dev: the netdev being configured
6901 * @nlh: RTNL message
6902 * @flags: bridge setlink flags
6903 * @extack: netlink extended ack
6905 * Sets the bridge mode (VEB/VEPA) of the switch to which the netdev (VSI) is
6906 * hooked up to. Iterates through the PF VSI list and sets the loopback mode (if
6907 * not already set for all VSIs connected to this switch. And also update the
6908 * unicast switch filter rules for the corresponding switch of the netdev.
6911 ice_bridge_setlink(struct net_device *dev, struct nlmsghdr *nlh,
6912 u16 __always_unused flags,
6913 struct netlink_ext_ack __always_unused *extack)
6915 struct ice_netdev_priv *np = netdev_priv(dev);
6916 struct ice_pf *pf = np->vsi->back;
6917 struct nlattr *attr, *br_spec;
6918 struct ice_hw *hw = &pf->hw;
6919 enum ice_status status;
6920 struct ice_sw *pf_sw;
6921 int rem, v, err = 0;
6923 pf_sw = pf->first_sw;
6924 /* find the attribute in the netlink message */
6925 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
6927 nla_for_each_nested(attr, br_spec, rem) {
6930 if (nla_type(attr) != IFLA_BRIDGE_MODE)
6932 mode = nla_get_u16(attr);
6933 if (mode != BRIDGE_MODE_VEPA && mode != BRIDGE_MODE_VEB)
6935 /* Continue if bridge mode is not being flipped */
6936 if (mode == pf_sw->bridge_mode)
6938 /* Iterates through the PF VSI list and update the loopback
6941 ice_for_each_vsi(pf, v) {
6944 err = ice_vsi_update_bridge_mode(pf->vsi[v], mode);
6949 hw->evb_veb = (mode == BRIDGE_MODE_VEB);
6950 /* Update the unicast switch filter rules for the corresponding
6951 * switch of the netdev
6953 status = ice_update_sw_rule_bridge_mode(hw);
6955 netdev_err(dev, "switch rule update failed, mode = %d err %s aq_err %s\n",
6956 mode, ice_stat_str(status),
6957 ice_aq_str(hw->adminq.sq_last_status));
6958 /* revert hw->evb_veb */
6959 hw->evb_veb = (pf_sw->bridge_mode == BRIDGE_MODE_VEB);
6963 pf_sw->bridge_mode = mode;
6970 * ice_tx_timeout - Respond to a Tx Hang
6971 * @netdev: network interface device structure
6972 * @txqueue: Tx queue
6974 static void ice_tx_timeout(struct net_device *netdev, unsigned int txqueue)
6976 struct ice_netdev_priv *np = netdev_priv(netdev);
6977 struct ice_ring *tx_ring = NULL;
6978 struct ice_vsi *vsi = np->vsi;
6979 struct ice_pf *pf = vsi->back;
6982 pf->tx_timeout_count++;
6984 /* Check if PFC is enabled for the TC to which the queue belongs
6985 * to. If yes then Tx timeout is not caused by a hung queue, no
6986 * need to reset and rebuild
6988 if (ice_is_pfc_causing_hung_q(pf, txqueue)) {
6989 dev_info(ice_pf_to_dev(pf), "Fake Tx hang detected on queue %u, timeout caused by PFC storm\n",
6994 /* now that we have an index, find the tx_ring struct */
6995 for (i = 0; i < vsi->num_txq; i++)
6996 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
6997 if (txqueue == vsi->tx_rings[i]->q_index) {
6998 tx_ring = vsi->tx_rings[i];
7002 /* Reset recovery level if enough time has elapsed after last timeout.
7003 * Also ensure no new reset action happens before next timeout period.
7005 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ * 20)))
7006 pf->tx_timeout_recovery_level = 1;
7007 else if (time_before(jiffies, (pf->tx_timeout_last_recovery +
7008 netdev->watchdog_timeo)))
7012 struct ice_hw *hw = &pf->hw;
7015 head = (rd32(hw, QTX_COMM_HEAD(vsi->txq_map[txqueue])) &
7016 QTX_COMM_HEAD_HEAD_M) >> QTX_COMM_HEAD_HEAD_S;
7017 /* Read interrupt register */
7018 val = rd32(hw, GLINT_DYN_CTL(tx_ring->q_vector->reg_idx));
7020 netdev_info(netdev, "tx_timeout: VSI_num: %d, Q %u, NTC: 0x%x, HW_HEAD: 0x%x, NTU: 0x%x, INT: 0x%x\n",
7021 vsi->vsi_num, txqueue, tx_ring->next_to_clean,
7022 head, tx_ring->next_to_use, val);
7025 pf->tx_timeout_last_recovery = jiffies;
7026 netdev_info(netdev, "tx_timeout recovery level %d, txqueue %u\n",
7027 pf->tx_timeout_recovery_level, txqueue);
7029 switch (pf->tx_timeout_recovery_level) {
7031 set_bit(ICE_PFR_REQ, pf->state);
7034 set_bit(ICE_CORER_REQ, pf->state);
7037 set_bit(ICE_GLOBR_REQ, pf->state);
7040 netdev_err(netdev, "tx_timeout recovery unsuccessful, device is in unrecoverable state.\n");
7041 set_bit(ICE_DOWN, pf->state);
7042 set_bit(ICE_VSI_NEEDS_RESTART, vsi->state);
7043 set_bit(ICE_SERVICE_DIS, pf->state);
7047 ice_service_task_schedule(pf);
7048 pf->tx_timeout_recovery_level++;
7052 * ice_open - Called when a network interface becomes active
7053 * @netdev: network interface device structure
7055 * The open entry point is called when a network interface is made
7056 * active by the system (IFF_UP). At this point all resources needed
7057 * for transmit and receive operations are allocated, the interrupt
7058 * handler is registered with the OS, the netdev watchdog is enabled,
7059 * and the stack is notified that the interface is ready.
7061 * Returns 0 on success, negative value on failure
7063 int ice_open(struct net_device *netdev)
7065 struct ice_netdev_priv *np = netdev_priv(netdev);
7066 struct ice_pf *pf = np->vsi->back;
7068 if (ice_is_reset_in_progress(pf->state)) {
7069 netdev_err(netdev, "can't open net device while reset is in progress");
7073 return ice_open_internal(netdev);
7077 * ice_open_internal - Called when a network interface becomes active
7078 * @netdev: network interface device structure
7080 * Internal ice_open implementation. Should not be used directly except for ice_open and reset
7083 * Returns 0 on success, negative value on failure
7085 int ice_open_internal(struct net_device *netdev)
7087 struct ice_netdev_priv *np = netdev_priv(netdev);
7088 struct ice_vsi *vsi = np->vsi;
7089 struct ice_pf *pf = vsi->back;
7090 struct ice_port_info *pi;
7091 enum ice_status status;
7094 if (test_bit(ICE_NEEDS_RESTART, pf->state)) {
7095 netdev_err(netdev, "driver needs to be unloaded and reloaded\n");
7099 netif_carrier_off(netdev);
7101 pi = vsi->port_info;
7102 status = ice_update_link_info(pi);
7104 netdev_err(netdev, "Failed to get link info, error %s\n",
7105 ice_stat_str(status));
7109 ice_check_module_power(pf, pi->phy.link_info.link_cfg_err);
7111 /* Set PHY if there is media, otherwise, turn off PHY */
7112 if (pi->phy.link_info.link_info & ICE_AQ_MEDIA_AVAILABLE) {
7113 clear_bit(ICE_FLAG_NO_MEDIA, pf->flags);
7114 if (!test_bit(ICE_PHY_INIT_COMPLETE, pf->state)) {
7115 err = ice_init_phy_user_cfg(pi);
7117 netdev_err(netdev, "Failed to initialize PHY settings, error %d\n",
7123 err = ice_configure_phy(vsi);
7125 netdev_err(netdev, "Failed to set physical link up, error %d\n",
7130 set_bit(ICE_FLAG_NO_MEDIA, pf->flags);
7131 ice_set_link(vsi, false);
7134 err = ice_vsi_open(vsi);
7136 netdev_err(netdev, "Failed to open VSI 0x%04X on switch 0x%04X\n",
7137 vsi->vsi_num, vsi->vsw->sw_id);
7139 /* Update existing tunnels information */
7140 udp_tunnel_get_rx_info(netdev);
7146 * ice_stop - Disables a network interface
7147 * @netdev: network interface device structure
7149 * The stop entry point is called when an interface is de-activated by the OS,
7150 * and the netdevice enters the DOWN state. The hardware is still under the
7151 * driver's control, but the netdev interface is disabled.
7153 * Returns success only - not allowed to fail
7155 int ice_stop(struct net_device *netdev)
7157 struct ice_netdev_priv *np = netdev_priv(netdev);
7158 struct ice_vsi *vsi = np->vsi;
7159 struct ice_pf *pf = vsi->back;
7161 if (ice_is_reset_in_progress(pf->state)) {
7162 netdev_err(netdev, "can't stop net device while reset is in progress");
7172 * ice_features_check - Validate encapsulated packet conforms to limits
7174 * @netdev: This port's netdev
7175 * @features: Offload features that the stack believes apply
7177 static netdev_features_t
7178 ice_features_check(struct sk_buff *skb,
7179 struct net_device __always_unused *netdev,
7180 netdev_features_t features)
7184 /* No point in doing any of this if neither checksum nor GSO are
7185 * being requested for this frame. We can rule out both by just
7186 * checking for CHECKSUM_PARTIAL
7188 if (skb->ip_summed != CHECKSUM_PARTIAL)
7191 /* We cannot support GSO if the MSS is going to be less than
7192 * 64 bytes. If it is then we need to drop support for GSO.
7194 if (skb_is_gso(skb) && (skb_shinfo(skb)->gso_size < 64))
7195 features &= ~NETIF_F_GSO_MASK;
7197 len = skb_network_header(skb) - skb->data;
7198 if (len > ICE_TXD_MACLEN_MAX || len & 0x1)
7199 goto out_rm_features;
7201 len = skb_transport_header(skb) - skb_network_header(skb);
7202 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7203 goto out_rm_features;
7205 if (skb->encapsulation) {
7206 len = skb_inner_network_header(skb) - skb_transport_header(skb);
7207 if (len > ICE_TXD_L4LEN_MAX || len & 0x1)
7208 goto out_rm_features;
7210 len = skb_inner_transport_header(skb) -
7211 skb_inner_network_header(skb);
7212 if (len > ICE_TXD_IPLEN_MAX || len & 0x1)
7213 goto out_rm_features;
7218 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
7221 static const struct net_device_ops ice_netdev_safe_mode_ops = {
7222 .ndo_open = ice_open,
7223 .ndo_stop = ice_stop,
7224 .ndo_start_xmit = ice_start_xmit,
7225 .ndo_set_mac_address = ice_set_mac_address,
7226 .ndo_validate_addr = eth_validate_addr,
7227 .ndo_change_mtu = ice_change_mtu,
7228 .ndo_get_stats64 = ice_get_stats64,
7229 .ndo_tx_timeout = ice_tx_timeout,
7230 .ndo_bpf = ice_xdp_safe_mode,
7233 static const struct net_device_ops ice_netdev_ops = {
7234 .ndo_open = ice_open,
7235 .ndo_stop = ice_stop,
7236 .ndo_start_xmit = ice_start_xmit,
7237 .ndo_features_check = ice_features_check,
7238 .ndo_set_rx_mode = ice_set_rx_mode,
7239 .ndo_set_mac_address = ice_set_mac_address,
7240 .ndo_validate_addr = eth_validate_addr,
7241 .ndo_change_mtu = ice_change_mtu,
7242 .ndo_get_stats64 = ice_get_stats64,
7243 .ndo_set_tx_maxrate = ice_set_tx_maxrate,
7244 .ndo_do_ioctl = ice_do_ioctl,
7245 .ndo_set_vf_spoofchk = ice_set_vf_spoofchk,
7246 .ndo_set_vf_mac = ice_set_vf_mac,
7247 .ndo_get_vf_config = ice_get_vf_cfg,
7248 .ndo_set_vf_trust = ice_set_vf_trust,
7249 .ndo_set_vf_vlan = ice_set_vf_port_vlan,
7250 .ndo_set_vf_link_state = ice_set_vf_link_state,
7251 .ndo_get_vf_stats = ice_get_vf_stats,
7252 .ndo_vlan_rx_add_vid = ice_vlan_rx_add_vid,
7253 .ndo_vlan_rx_kill_vid = ice_vlan_rx_kill_vid,
7254 .ndo_set_features = ice_set_features,
7255 .ndo_bridge_getlink = ice_bridge_getlink,
7256 .ndo_bridge_setlink = ice_bridge_setlink,
7257 .ndo_fdb_add = ice_fdb_add,
7258 .ndo_fdb_del = ice_fdb_del,
7259 #ifdef CONFIG_RFS_ACCEL
7260 .ndo_rx_flow_steer = ice_rx_flow_steer,
7262 .ndo_tx_timeout = ice_tx_timeout,
7264 .ndo_xdp_xmit = ice_xdp_xmit,
7265 .ndo_xsk_wakeup = ice_xsk_wakeup,