1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 - 2016 Intel Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * You should have received a copy of the GNU General Public License along
16 * with this program. If not, see <http://www.gnu.org/licenses/>.
18 * The full GNU General Public License is included in this distribution in
19 * the file called "COPYING".
21 * Contact Information:
22 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 ******************************************************************************/
27 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
33 #include "i40e_diag.h"
34 #include <net/udp_tunnel.h>
36 const char i40e_driver_name[] = "i40e";
37 static const char i40e_driver_string[] =
38 "Intel(R) Ethernet Connection XL710 Network Driver";
42 #define DRV_VERSION_MAJOR 1
43 #define DRV_VERSION_MINOR 6
44 #define DRV_VERSION_BUILD 11
45 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
46 __stringify(DRV_VERSION_MINOR) "." \
47 __stringify(DRV_VERSION_BUILD) DRV_KERN
48 const char i40e_driver_version_str[] = DRV_VERSION;
49 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
51 /* a bit of forward declarations */
52 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
53 static void i40e_handle_reset_warning(struct i40e_pf *pf);
54 static int i40e_add_vsi(struct i40e_vsi *vsi);
55 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
56 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
57 static int i40e_setup_misc_vector(struct i40e_pf *pf);
58 static void i40e_determine_queue_usage(struct i40e_pf *pf);
59 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
60 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
61 u16 rss_table_size, u16 rss_size);
62 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
63 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
65 /* i40e_pci_tbl - PCI Device ID Table
67 * Last entry must be all 0s
69 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
70 * Class, Class Mask, private data (not used) }
72 static const struct pci_device_id i40e_pci_tbl[] = {
73 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
74 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
75 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
76 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
77 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
78 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
79 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
80 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
81 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
82 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
83 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
84 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
85 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
86 {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
87 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
88 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
89 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
90 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
91 /* required last entry */
94 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
96 #define I40E_MAX_VF_COUNT 128
97 static int debug = -1;
98 module_param(debug, int, 0);
99 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
101 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
102 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
103 MODULE_LICENSE("GPL");
104 MODULE_VERSION(DRV_VERSION);
106 static struct workqueue_struct *i40e_wq;
109 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
110 * @hw: pointer to the HW structure
111 * @mem: ptr to mem struct to fill out
112 * @size: size of memory requested
113 * @alignment: what to align the allocation to
115 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
116 u64 size, u32 alignment)
118 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
120 mem->size = ALIGN(size, alignment);
121 mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
122 &mem->pa, GFP_KERNEL);
130 * i40e_free_dma_mem_d - OS specific memory free for shared code
131 * @hw: pointer to the HW structure
132 * @mem: ptr to mem struct to free
134 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
136 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
138 dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
147 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
148 * @hw: pointer to the HW structure
149 * @mem: ptr to mem struct to fill out
150 * @size: size of memory requested
152 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
156 mem->va = kzalloc(size, GFP_KERNEL);
165 * i40e_free_virt_mem_d - OS specific memory free for shared code
166 * @hw: pointer to the HW structure
167 * @mem: ptr to mem struct to free
169 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
171 /* it's ok to kfree a NULL pointer */
180 * i40e_get_lump - find a lump of free generic resource
181 * @pf: board private structure
182 * @pile: the pile of resource to search
183 * @needed: the number of items needed
184 * @id: an owner id to stick on the items assigned
186 * Returns the base item index of the lump, or negative for error
188 * The search_hint trick and lack of advanced fit-finding only work
189 * because we're highly likely to have all the same size lump requests.
190 * Linear search time and any fragmentation should be minimal.
192 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
198 if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
199 dev_info(&pf->pdev->dev,
200 "param err: pile=%p needed=%d id=0x%04x\n",
205 /* start the linear search with an imperfect hint */
206 i = pile->search_hint;
207 while (i < pile->num_entries) {
208 /* skip already allocated entries */
209 if (pile->list[i] & I40E_PILE_VALID_BIT) {
214 /* do we have enough in this lump? */
215 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
216 if (pile->list[i+j] & I40E_PILE_VALID_BIT)
221 /* there was enough, so assign it to the requestor */
222 for (j = 0; j < needed; j++)
223 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
225 pile->search_hint = i + j;
229 /* not enough, so skip over it and continue looking */
237 * i40e_put_lump - return a lump of generic resource
238 * @pile: the pile of resource to search
239 * @index: the base item index
240 * @id: the owner id of the items assigned
242 * Returns the count of items in the lump
244 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
246 int valid_id = (id | I40E_PILE_VALID_BIT);
250 if (!pile || index >= pile->num_entries)
254 i < pile->num_entries && pile->list[i] == valid_id;
260 if (count && index < pile->search_hint)
261 pile->search_hint = index;
267 * i40e_find_vsi_from_id - searches for the vsi with the given id
268 * @pf - the pf structure to search for the vsi
269 * @id - id of the vsi it is searching for
271 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
275 for (i = 0; i < pf->num_alloc_vsi; i++)
276 if (pf->vsi[i] && (pf->vsi[i]->id == id))
283 * i40e_service_event_schedule - Schedule the service task to wake up
284 * @pf: board private structure
286 * If not already scheduled, this puts the task into the work queue
288 void i40e_service_event_schedule(struct i40e_pf *pf)
290 if (!test_bit(__I40E_DOWN, &pf->state) &&
291 !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
292 !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
293 queue_work(i40e_wq, &pf->service_task);
297 * i40e_tx_timeout - Respond to a Tx Hang
298 * @netdev: network interface device structure
300 * If any port has noticed a Tx timeout, it is likely that the whole
301 * device is munged, not just the one netdev port, so go for the full
305 void i40e_tx_timeout(struct net_device *netdev)
307 static void i40e_tx_timeout(struct net_device *netdev)
310 struct i40e_netdev_priv *np = netdev_priv(netdev);
311 struct i40e_vsi *vsi = np->vsi;
312 struct i40e_pf *pf = vsi->back;
313 struct i40e_ring *tx_ring = NULL;
314 unsigned int i, hung_queue = 0;
317 pf->tx_timeout_count++;
319 /* find the stopped queue the same way the stack does */
320 for (i = 0; i < netdev->num_tx_queues; i++) {
321 struct netdev_queue *q;
322 unsigned long trans_start;
324 q = netdev_get_tx_queue(netdev, i);
325 trans_start = q->trans_start;
326 if (netif_xmit_stopped(q) &&
328 (trans_start + netdev->watchdog_timeo))) {
334 if (i == netdev->num_tx_queues) {
335 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
337 /* now that we have an index, find the tx_ring struct */
338 for (i = 0; i < vsi->num_queue_pairs; i++) {
339 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
341 vsi->tx_rings[i]->queue_index) {
342 tx_ring = vsi->tx_rings[i];
349 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
350 pf->tx_timeout_recovery_level = 1; /* reset after some time */
351 else if (time_before(jiffies,
352 (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
353 return; /* don't do any new action before the next timeout */
356 head = i40e_get_head(tx_ring);
357 /* Read interrupt register */
358 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
360 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
361 tx_ring->vsi->base_vector - 1));
363 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
365 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
366 vsi->seid, hung_queue, tx_ring->next_to_clean,
367 head, tx_ring->next_to_use,
368 readl(tx_ring->tail), val);
371 pf->tx_timeout_last_recovery = jiffies;
372 netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
373 pf->tx_timeout_recovery_level, hung_queue);
375 switch (pf->tx_timeout_recovery_level) {
377 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
380 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
383 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
386 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
390 i40e_service_event_schedule(pf);
391 pf->tx_timeout_recovery_level++;
395 * i40e_get_vsi_stats_struct - Get System Network Statistics
396 * @vsi: the VSI we care about
398 * Returns the address of the device statistics structure.
399 * The statistics are actually updated from the service task.
401 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
403 return &vsi->net_stats;
407 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
408 * @netdev: network interface device structure
410 * Returns the address of the device statistics structure.
411 * The statistics are actually updated from the service task.
414 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
415 struct net_device *netdev,
416 struct rtnl_link_stats64 *stats)
418 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
419 struct net_device *netdev,
420 struct rtnl_link_stats64 *stats)
423 struct i40e_netdev_priv *np = netdev_priv(netdev);
424 struct i40e_ring *tx_ring, *rx_ring;
425 struct i40e_vsi *vsi = np->vsi;
426 struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
429 if (test_bit(__I40E_DOWN, &vsi->state))
436 for (i = 0; i < vsi->num_queue_pairs; i++) {
440 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
445 start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
446 packets = tx_ring->stats.packets;
447 bytes = tx_ring->stats.bytes;
448 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
450 stats->tx_packets += packets;
451 stats->tx_bytes += bytes;
452 rx_ring = &tx_ring[1];
455 start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
456 packets = rx_ring->stats.packets;
457 bytes = rx_ring->stats.bytes;
458 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
460 stats->rx_packets += packets;
461 stats->rx_bytes += bytes;
465 /* following stats updated by i40e_watchdog_subtask() */
466 stats->multicast = vsi_stats->multicast;
467 stats->tx_errors = vsi_stats->tx_errors;
468 stats->tx_dropped = vsi_stats->tx_dropped;
469 stats->rx_errors = vsi_stats->rx_errors;
470 stats->rx_dropped = vsi_stats->rx_dropped;
471 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
472 stats->rx_length_errors = vsi_stats->rx_length_errors;
478 * i40e_vsi_reset_stats - Resets all stats of the given vsi
479 * @vsi: the VSI to have its stats reset
481 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
483 struct rtnl_link_stats64 *ns;
489 ns = i40e_get_vsi_stats_struct(vsi);
490 memset(ns, 0, sizeof(*ns));
491 memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
492 memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
493 memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
494 if (vsi->rx_rings && vsi->rx_rings[0]) {
495 for (i = 0; i < vsi->num_queue_pairs; i++) {
496 memset(&vsi->rx_rings[i]->stats, 0,
497 sizeof(vsi->rx_rings[i]->stats));
498 memset(&vsi->rx_rings[i]->rx_stats, 0,
499 sizeof(vsi->rx_rings[i]->rx_stats));
500 memset(&vsi->tx_rings[i]->stats, 0,
501 sizeof(vsi->tx_rings[i]->stats));
502 memset(&vsi->tx_rings[i]->tx_stats, 0,
503 sizeof(vsi->tx_rings[i]->tx_stats));
506 vsi->stat_offsets_loaded = false;
510 * i40e_pf_reset_stats - Reset all of the stats for the given PF
511 * @pf: the PF to be reset
513 void i40e_pf_reset_stats(struct i40e_pf *pf)
517 memset(&pf->stats, 0, sizeof(pf->stats));
518 memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
519 pf->stat_offsets_loaded = false;
521 for (i = 0; i < I40E_MAX_VEB; i++) {
523 memset(&pf->veb[i]->stats, 0,
524 sizeof(pf->veb[i]->stats));
525 memset(&pf->veb[i]->stats_offsets, 0,
526 sizeof(pf->veb[i]->stats_offsets));
527 pf->veb[i]->stat_offsets_loaded = false;
533 * i40e_stat_update48 - read and update a 48 bit stat from the chip
534 * @hw: ptr to the hardware info
535 * @hireg: the high 32 bit reg to read
536 * @loreg: the low 32 bit reg to read
537 * @offset_loaded: has the initial offset been loaded yet
538 * @offset: ptr to current offset value
539 * @stat: ptr to the stat
541 * Since the device stats are not reset at PFReset, they likely will not
542 * be zeroed when the driver starts. We'll save the first values read
543 * and use them as offsets to be subtracted from the raw values in order
544 * to report stats that count from zero. In the process, we also manage
545 * the potential roll-over.
547 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
548 bool offset_loaded, u64 *offset, u64 *stat)
552 if (hw->device_id == I40E_DEV_ID_QEMU) {
553 new_data = rd32(hw, loreg);
554 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
556 new_data = rd64(hw, loreg);
560 if (likely(new_data >= *offset))
561 *stat = new_data - *offset;
563 *stat = (new_data + BIT_ULL(48)) - *offset;
564 *stat &= 0xFFFFFFFFFFFFULL;
568 * i40e_stat_update32 - read and update a 32 bit stat from the chip
569 * @hw: ptr to the hardware info
570 * @reg: the hw reg to read
571 * @offset_loaded: has the initial offset been loaded yet
572 * @offset: ptr to current offset value
573 * @stat: ptr to the stat
575 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
576 bool offset_loaded, u64 *offset, u64 *stat)
580 new_data = rd32(hw, reg);
583 if (likely(new_data >= *offset))
584 *stat = (u32)(new_data - *offset);
586 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
590 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
591 * @vsi: the VSI to be updated
593 void i40e_update_eth_stats(struct i40e_vsi *vsi)
595 int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
596 struct i40e_pf *pf = vsi->back;
597 struct i40e_hw *hw = &pf->hw;
598 struct i40e_eth_stats *oes;
599 struct i40e_eth_stats *es; /* device's eth stats */
601 es = &vsi->eth_stats;
602 oes = &vsi->eth_stats_offsets;
604 /* Gather up the stats that the hw collects */
605 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
606 vsi->stat_offsets_loaded,
607 &oes->tx_errors, &es->tx_errors);
608 i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
609 vsi->stat_offsets_loaded,
610 &oes->rx_discards, &es->rx_discards);
611 i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
612 vsi->stat_offsets_loaded,
613 &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
614 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
615 vsi->stat_offsets_loaded,
616 &oes->tx_errors, &es->tx_errors);
618 i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
619 I40E_GLV_GORCL(stat_idx),
620 vsi->stat_offsets_loaded,
621 &oes->rx_bytes, &es->rx_bytes);
622 i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
623 I40E_GLV_UPRCL(stat_idx),
624 vsi->stat_offsets_loaded,
625 &oes->rx_unicast, &es->rx_unicast);
626 i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
627 I40E_GLV_MPRCL(stat_idx),
628 vsi->stat_offsets_loaded,
629 &oes->rx_multicast, &es->rx_multicast);
630 i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
631 I40E_GLV_BPRCL(stat_idx),
632 vsi->stat_offsets_loaded,
633 &oes->rx_broadcast, &es->rx_broadcast);
635 i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
636 I40E_GLV_GOTCL(stat_idx),
637 vsi->stat_offsets_loaded,
638 &oes->tx_bytes, &es->tx_bytes);
639 i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
640 I40E_GLV_UPTCL(stat_idx),
641 vsi->stat_offsets_loaded,
642 &oes->tx_unicast, &es->tx_unicast);
643 i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
644 I40E_GLV_MPTCL(stat_idx),
645 vsi->stat_offsets_loaded,
646 &oes->tx_multicast, &es->tx_multicast);
647 i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
648 I40E_GLV_BPTCL(stat_idx),
649 vsi->stat_offsets_loaded,
650 &oes->tx_broadcast, &es->tx_broadcast);
651 vsi->stat_offsets_loaded = true;
655 * i40e_update_veb_stats - Update Switch component statistics
656 * @veb: the VEB being updated
658 static void i40e_update_veb_stats(struct i40e_veb *veb)
660 struct i40e_pf *pf = veb->pf;
661 struct i40e_hw *hw = &pf->hw;
662 struct i40e_eth_stats *oes;
663 struct i40e_eth_stats *es; /* device's eth stats */
664 struct i40e_veb_tc_stats *veb_oes;
665 struct i40e_veb_tc_stats *veb_es;
668 idx = veb->stats_idx;
670 oes = &veb->stats_offsets;
671 veb_es = &veb->tc_stats;
672 veb_oes = &veb->tc_stats_offsets;
674 /* Gather up the stats that the hw collects */
675 i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
676 veb->stat_offsets_loaded,
677 &oes->tx_discards, &es->tx_discards);
678 if (hw->revision_id > 0)
679 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
680 veb->stat_offsets_loaded,
681 &oes->rx_unknown_protocol,
682 &es->rx_unknown_protocol);
683 i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
684 veb->stat_offsets_loaded,
685 &oes->rx_bytes, &es->rx_bytes);
686 i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
687 veb->stat_offsets_loaded,
688 &oes->rx_unicast, &es->rx_unicast);
689 i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
690 veb->stat_offsets_loaded,
691 &oes->rx_multicast, &es->rx_multicast);
692 i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
693 veb->stat_offsets_loaded,
694 &oes->rx_broadcast, &es->rx_broadcast);
696 i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
697 veb->stat_offsets_loaded,
698 &oes->tx_bytes, &es->tx_bytes);
699 i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
700 veb->stat_offsets_loaded,
701 &oes->tx_unicast, &es->tx_unicast);
702 i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
703 veb->stat_offsets_loaded,
704 &oes->tx_multicast, &es->tx_multicast);
705 i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
706 veb->stat_offsets_loaded,
707 &oes->tx_broadcast, &es->tx_broadcast);
708 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
709 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
710 I40E_GLVEBTC_RPCL(i, idx),
711 veb->stat_offsets_loaded,
712 &veb_oes->tc_rx_packets[i],
713 &veb_es->tc_rx_packets[i]);
714 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
715 I40E_GLVEBTC_RBCL(i, idx),
716 veb->stat_offsets_loaded,
717 &veb_oes->tc_rx_bytes[i],
718 &veb_es->tc_rx_bytes[i]);
719 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
720 I40E_GLVEBTC_TPCL(i, idx),
721 veb->stat_offsets_loaded,
722 &veb_oes->tc_tx_packets[i],
723 &veb_es->tc_tx_packets[i]);
724 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
725 I40E_GLVEBTC_TBCL(i, idx),
726 veb->stat_offsets_loaded,
727 &veb_oes->tc_tx_bytes[i],
728 &veb_es->tc_tx_bytes[i]);
730 veb->stat_offsets_loaded = true;
735 * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
736 * @vsi: the VSI that is capable of doing FCoE
738 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
740 struct i40e_pf *pf = vsi->back;
741 struct i40e_hw *hw = &pf->hw;
742 struct i40e_fcoe_stats *ofs;
743 struct i40e_fcoe_stats *fs; /* device's eth stats */
746 if (vsi->type != I40E_VSI_FCOE)
749 idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
750 fs = &vsi->fcoe_stats;
751 ofs = &vsi->fcoe_stats_offsets;
753 i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
754 vsi->fcoe_stat_offsets_loaded,
755 &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
756 i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
757 vsi->fcoe_stat_offsets_loaded,
758 &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
759 i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
760 vsi->fcoe_stat_offsets_loaded,
761 &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
762 i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
763 vsi->fcoe_stat_offsets_loaded,
764 &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
765 i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
766 vsi->fcoe_stat_offsets_loaded,
767 &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
768 i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
769 vsi->fcoe_stat_offsets_loaded,
770 &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
771 i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
772 vsi->fcoe_stat_offsets_loaded,
773 &ofs->fcoe_last_error, &fs->fcoe_last_error);
774 i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
775 vsi->fcoe_stat_offsets_loaded,
776 &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
778 vsi->fcoe_stat_offsets_loaded = true;
783 * i40e_update_vsi_stats - Update the vsi statistics counters.
784 * @vsi: the VSI to be updated
786 * There are a few instances where we store the same stat in a
787 * couple of different structs. This is partly because we have
788 * the netdev stats that need to be filled out, which is slightly
789 * different from the "eth_stats" defined by the chip and used in
790 * VF communications. We sort it out here.
792 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
794 struct i40e_pf *pf = vsi->back;
795 struct rtnl_link_stats64 *ons;
796 struct rtnl_link_stats64 *ns; /* netdev stats */
797 struct i40e_eth_stats *oes;
798 struct i40e_eth_stats *es; /* device's eth stats */
799 u32 tx_restart, tx_busy;
800 u64 tx_lost_interrupt;
811 if (test_bit(__I40E_DOWN, &vsi->state) ||
812 test_bit(__I40E_CONFIG_BUSY, &pf->state))
815 ns = i40e_get_vsi_stats_struct(vsi);
816 ons = &vsi->net_stats_offsets;
817 es = &vsi->eth_stats;
818 oes = &vsi->eth_stats_offsets;
820 /* Gather up the netdev and vsi stats that the driver collects
821 * on the fly during packet processing
825 tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
826 tx_lost_interrupt = 0;
830 for (q = 0; q < vsi->num_queue_pairs; q++) {
832 p = ACCESS_ONCE(vsi->tx_rings[q]);
835 start = u64_stats_fetch_begin_irq(&p->syncp);
836 packets = p->stats.packets;
837 bytes = p->stats.bytes;
838 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
841 tx_restart += p->tx_stats.restart_queue;
842 tx_busy += p->tx_stats.tx_busy;
843 tx_linearize += p->tx_stats.tx_linearize;
844 tx_force_wb += p->tx_stats.tx_force_wb;
845 tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
847 /* Rx queue is part of the same block as Tx queue */
850 start = u64_stats_fetch_begin_irq(&p->syncp);
851 packets = p->stats.packets;
852 bytes = p->stats.bytes;
853 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
856 rx_buf += p->rx_stats.alloc_buff_failed;
857 rx_page += p->rx_stats.alloc_page_failed;
860 vsi->tx_restart = tx_restart;
861 vsi->tx_busy = tx_busy;
862 vsi->tx_linearize = tx_linearize;
863 vsi->tx_force_wb = tx_force_wb;
864 vsi->tx_lost_interrupt = tx_lost_interrupt;
865 vsi->rx_page_failed = rx_page;
866 vsi->rx_buf_failed = rx_buf;
868 ns->rx_packets = rx_p;
870 ns->tx_packets = tx_p;
873 /* update netdev stats from eth stats */
874 i40e_update_eth_stats(vsi);
875 ons->tx_errors = oes->tx_errors;
876 ns->tx_errors = es->tx_errors;
877 ons->multicast = oes->rx_multicast;
878 ns->multicast = es->rx_multicast;
879 ons->rx_dropped = oes->rx_discards;
880 ns->rx_dropped = es->rx_discards;
881 ons->tx_dropped = oes->tx_discards;
882 ns->tx_dropped = es->tx_discards;
884 /* pull in a couple PF stats if this is the main vsi */
885 if (vsi == pf->vsi[pf->lan_vsi]) {
886 ns->rx_crc_errors = pf->stats.crc_errors;
887 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
888 ns->rx_length_errors = pf->stats.rx_length_errors;
893 * i40e_update_pf_stats - Update the PF statistics counters.
894 * @pf: the PF to be updated
896 static void i40e_update_pf_stats(struct i40e_pf *pf)
898 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
899 struct i40e_hw_port_stats *nsd = &pf->stats;
900 struct i40e_hw *hw = &pf->hw;
904 i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
905 I40E_GLPRT_GORCL(hw->port),
906 pf->stat_offsets_loaded,
907 &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
908 i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
909 I40E_GLPRT_GOTCL(hw->port),
910 pf->stat_offsets_loaded,
911 &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
912 i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
913 pf->stat_offsets_loaded,
914 &osd->eth.rx_discards,
915 &nsd->eth.rx_discards);
916 i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
917 I40E_GLPRT_UPRCL(hw->port),
918 pf->stat_offsets_loaded,
919 &osd->eth.rx_unicast,
920 &nsd->eth.rx_unicast);
921 i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
922 I40E_GLPRT_MPRCL(hw->port),
923 pf->stat_offsets_loaded,
924 &osd->eth.rx_multicast,
925 &nsd->eth.rx_multicast);
926 i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
927 I40E_GLPRT_BPRCL(hw->port),
928 pf->stat_offsets_loaded,
929 &osd->eth.rx_broadcast,
930 &nsd->eth.rx_broadcast);
931 i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
932 I40E_GLPRT_UPTCL(hw->port),
933 pf->stat_offsets_loaded,
934 &osd->eth.tx_unicast,
935 &nsd->eth.tx_unicast);
936 i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
937 I40E_GLPRT_MPTCL(hw->port),
938 pf->stat_offsets_loaded,
939 &osd->eth.tx_multicast,
940 &nsd->eth.tx_multicast);
941 i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
942 I40E_GLPRT_BPTCL(hw->port),
943 pf->stat_offsets_loaded,
944 &osd->eth.tx_broadcast,
945 &nsd->eth.tx_broadcast);
947 i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
948 pf->stat_offsets_loaded,
949 &osd->tx_dropped_link_down,
950 &nsd->tx_dropped_link_down);
952 i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
953 pf->stat_offsets_loaded,
954 &osd->crc_errors, &nsd->crc_errors);
956 i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
957 pf->stat_offsets_loaded,
958 &osd->illegal_bytes, &nsd->illegal_bytes);
960 i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
961 pf->stat_offsets_loaded,
962 &osd->mac_local_faults,
963 &nsd->mac_local_faults);
964 i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
965 pf->stat_offsets_loaded,
966 &osd->mac_remote_faults,
967 &nsd->mac_remote_faults);
969 i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
970 pf->stat_offsets_loaded,
971 &osd->rx_length_errors,
972 &nsd->rx_length_errors);
974 i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
975 pf->stat_offsets_loaded,
976 &osd->link_xon_rx, &nsd->link_xon_rx);
977 i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
978 pf->stat_offsets_loaded,
979 &osd->link_xon_tx, &nsd->link_xon_tx);
980 i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
981 pf->stat_offsets_loaded,
982 &osd->link_xoff_rx, &nsd->link_xoff_rx);
983 i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
984 pf->stat_offsets_loaded,
985 &osd->link_xoff_tx, &nsd->link_xoff_tx);
987 for (i = 0; i < 8; i++) {
988 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
989 pf->stat_offsets_loaded,
990 &osd->priority_xoff_rx[i],
991 &nsd->priority_xoff_rx[i]);
992 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
993 pf->stat_offsets_loaded,
994 &osd->priority_xon_rx[i],
995 &nsd->priority_xon_rx[i]);
996 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
997 pf->stat_offsets_loaded,
998 &osd->priority_xon_tx[i],
999 &nsd->priority_xon_tx[i]);
1000 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1001 pf->stat_offsets_loaded,
1002 &osd->priority_xoff_tx[i],
1003 &nsd->priority_xoff_tx[i]);
1004 i40e_stat_update32(hw,
1005 I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1006 pf->stat_offsets_loaded,
1007 &osd->priority_xon_2_xoff[i],
1008 &nsd->priority_xon_2_xoff[i]);
1011 i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1012 I40E_GLPRT_PRC64L(hw->port),
1013 pf->stat_offsets_loaded,
1014 &osd->rx_size_64, &nsd->rx_size_64);
1015 i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1016 I40E_GLPRT_PRC127L(hw->port),
1017 pf->stat_offsets_loaded,
1018 &osd->rx_size_127, &nsd->rx_size_127);
1019 i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1020 I40E_GLPRT_PRC255L(hw->port),
1021 pf->stat_offsets_loaded,
1022 &osd->rx_size_255, &nsd->rx_size_255);
1023 i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1024 I40E_GLPRT_PRC511L(hw->port),
1025 pf->stat_offsets_loaded,
1026 &osd->rx_size_511, &nsd->rx_size_511);
1027 i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1028 I40E_GLPRT_PRC1023L(hw->port),
1029 pf->stat_offsets_loaded,
1030 &osd->rx_size_1023, &nsd->rx_size_1023);
1031 i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1032 I40E_GLPRT_PRC1522L(hw->port),
1033 pf->stat_offsets_loaded,
1034 &osd->rx_size_1522, &nsd->rx_size_1522);
1035 i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1036 I40E_GLPRT_PRC9522L(hw->port),
1037 pf->stat_offsets_loaded,
1038 &osd->rx_size_big, &nsd->rx_size_big);
1040 i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1041 I40E_GLPRT_PTC64L(hw->port),
1042 pf->stat_offsets_loaded,
1043 &osd->tx_size_64, &nsd->tx_size_64);
1044 i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1045 I40E_GLPRT_PTC127L(hw->port),
1046 pf->stat_offsets_loaded,
1047 &osd->tx_size_127, &nsd->tx_size_127);
1048 i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1049 I40E_GLPRT_PTC255L(hw->port),
1050 pf->stat_offsets_loaded,
1051 &osd->tx_size_255, &nsd->tx_size_255);
1052 i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1053 I40E_GLPRT_PTC511L(hw->port),
1054 pf->stat_offsets_loaded,
1055 &osd->tx_size_511, &nsd->tx_size_511);
1056 i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1057 I40E_GLPRT_PTC1023L(hw->port),
1058 pf->stat_offsets_loaded,
1059 &osd->tx_size_1023, &nsd->tx_size_1023);
1060 i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1061 I40E_GLPRT_PTC1522L(hw->port),
1062 pf->stat_offsets_loaded,
1063 &osd->tx_size_1522, &nsd->tx_size_1522);
1064 i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1065 I40E_GLPRT_PTC9522L(hw->port),
1066 pf->stat_offsets_loaded,
1067 &osd->tx_size_big, &nsd->tx_size_big);
1069 i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1070 pf->stat_offsets_loaded,
1071 &osd->rx_undersize, &nsd->rx_undersize);
1072 i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1073 pf->stat_offsets_loaded,
1074 &osd->rx_fragments, &nsd->rx_fragments);
1075 i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1076 pf->stat_offsets_loaded,
1077 &osd->rx_oversize, &nsd->rx_oversize);
1078 i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1079 pf->stat_offsets_loaded,
1080 &osd->rx_jabber, &nsd->rx_jabber);
1083 i40e_stat_update32(hw,
1084 I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1085 pf->stat_offsets_loaded,
1086 &osd->fd_atr_match, &nsd->fd_atr_match);
1087 i40e_stat_update32(hw,
1088 I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1089 pf->stat_offsets_loaded,
1090 &osd->fd_sb_match, &nsd->fd_sb_match);
1091 i40e_stat_update32(hw,
1092 I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1093 pf->stat_offsets_loaded,
1094 &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1096 val = rd32(hw, I40E_PRTPM_EEE_STAT);
1097 nsd->tx_lpi_status =
1098 (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1099 I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1100 nsd->rx_lpi_status =
1101 (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1102 I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1103 i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1104 pf->stat_offsets_loaded,
1105 &osd->tx_lpi_count, &nsd->tx_lpi_count);
1106 i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1107 pf->stat_offsets_loaded,
1108 &osd->rx_lpi_count, &nsd->rx_lpi_count);
1110 if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1111 !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1112 nsd->fd_sb_status = true;
1114 nsd->fd_sb_status = false;
1116 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1117 !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1118 nsd->fd_atr_status = true;
1120 nsd->fd_atr_status = false;
1122 pf->stat_offsets_loaded = true;
1126 * i40e_update_stats - Update the various statistics counters.
1127 * @vsi: the VSI to be updated
1129 * Update the various stats for this VSI and its related entities.
1131 void i40e_update_stats(struct i40e_vsi *vsi)
1133 struct i40e_pf *pf = vsi->back;
1135 if (vsi == pf->vsi[pf->lan_vsi])
1136 i40e_update_pf_stats(pf);
1138 i40e_update_vsi_stats(vsi);
1140 i40e_update_fcoe_stats(vsi);
1145 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1146 * @vsi: the VSI to be searched
1147 * @macaddr: the MAC address
1149 * @is_vf: make sure its a VF filter, else doesn't matter
1150 * @is_netdev: make sure its a netdev filter, else doesn't matter
1152 * Returns ptr to the filter object or NULL
1154 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1155 u8 *macaddr, s16 vlan,
1156 bool is_vf, bool is_netdev)
1158 struct i40e_mac_filter *f;
1160 if (!vsi || !macaddr)
1163 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1164 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1165 (vlan == f->vlan) &&
1166 (!is_vf || f->is_vf) &&
1167 (!is_netdev || f->is_netdev))
1174 * i40e_find_mac - Find a mac addr in the macvlan filters list
1175 * @vsi: the VSI to be searched
1176 * @macaddr: the MAC address we are searching for
1177 * @is_vf: make sure its a VF filter, else doesn't matter
1178 * @is_netdev: make sure its a netdev filter, else doesn't matter
1180 * Returns the first filter with the provided MAC address or NULL if
1181 * MAC address was not found
1183 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1184 bool is_vf, bool is_netdev)
1186 struct i40e_mac_filter *f;
1188 if (!vsi || !macaddr)
1191 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1192 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1193 (!is_vf || f->is_vf) &&
1194 (!is_netdev || f->is_netdev))
1201 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1202 * @vsi: the VSI to be searched
1204 * Returns true if VSI is in vlan mode or false otherwise
1206 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1208 struct i40e_mac_filter *f;
1210 /* Only -1 for all the filters denotes not in vlan mode
1211 * so we have to go through all the list in order to make sure
1213 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1214 if (f->vlan >= 0 || vsi->info.pvid)
1222 * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1223 * @vsi: the VSI to be searched
1224 * @macaddr: the mac address to be filtered
1225 * @is_vf: true if it is a VF
1226 * @is_netdev: true if it is a netdev
1228 * Goes through all the macvlan filters and adds a
1229 * macvlan filter for each unique vlan that already exists
1231 * Returns first filter found on success, else NULL
1233 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1234 bool is_vf, bool is_netdev)
1236 struct i40e_mac_filter *f;
1238 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1240 f->vlan = le16_to_cpu(vsi->info.pvid);
1241 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1242 is_vf, is_netdev)) {
1243 if (!i40e_add_filter(vsi, macaddr, f->vlan,
1249 return list_first_entry_or_null(&vsi->mac_filter_list,
1250 struct i40e_mac_filter, list);
1254 * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1255 * @vsi: the VSI to be searched
1256 * @macaddr: the mac address to be removed
1257 * @is_vf: true if it is a VF
1258 * @is_netdev: true if it is a netdev
1260 * Removes a given MAC address from a VSI, regardless of VLAN
1262 * Returns 0 for success, or error
1264 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1265 bool is_vf, bool is_netdev)
1267 struct i40e_mac_filter *f = NULL;
1270 WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1271 "Missing mac_filter_list_lock\n");
1272 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1273 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1274 (is_vf == f->is_vf) &&
1275 (is_netdev == f->is_netdev)) {
1278 if (f->counter == 0)
1279 f->state = I40E_FILTER_REMOVE;
1283 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1284 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1291 * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1292 * @vsi: the PF Main VSI - inappropriate for any other VSI
1293 * @macaddr: the MAC address
1295 * Remove whatever filter the firmware set up so the driver can manage
1296 * its own filtering intelligently.
1298 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1300 struct i40e_aqc_remove_macvlan_element_data element;
1301 struct i40e_pf *pf = vsi->back;
1303 /* Only appropriate for the PF main VSI */
1304 if (vsi->type != I40E_VSI_MAIN)
1307 memset(&element, 0, sizeof(element));
1308 ether_addr_copy(element.mac_addr, macaddr);
1309 element.vlan_tag = 0;
1310 /* Ignore error returns, some firmware does it this way... */
1311 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1312 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1314 memset(&element, 0, sizeof(element));
1315 ether_addr_copy(element.mac_addr, macaddr);
1316 element.vlan_tag = 0;
1317 /* ...and some firmware does it this way. */
1318 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1319 I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1320 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1324 * i40e_add_filter - Add a mac/vlan filter to the VSI
1325 * @vsi: the VSI to be searched
1326 * @macaddr: the MAC address
1328 * @is_vf: make sure its a VF filter, else doesn't matter
1329 * @is_netdev: make sure its a netdev filter, else doesn't matter
1331 * Returns ptr to the filter object or NULL when no memory available.
1333 * NOTE: This function is expected to be called with mac_filter_list_lock
1336 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1337 u8 *macaddr, s16 vlan,
1338 bool is_vf, bool is_netdev)
1340 struct i40e_mac_filter *f;
1341 int changed = false;
1343 if (!vsi || !macaddr)
1346 /* Do not allow broadcast filter to be added since broadcast filter
1347 * is added as part of add VSI for any newly created VSI except
1350 if (is_broadcast_ether_addr(macaddr))
1353 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1355 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1357 goto add_filter_out;
1359 ether_addr_copy(f->macaddr, macaddr);
1361 /* If we're in overflow promisc mode, set the state directly
1362 * to failed, so we don't bother to try sending the filter
1365 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))
1366 f->state = I40E_FILTER_FAILED;
1368 f->state = I40E_FILTER_NEW;
1370 INIT_LIST_HEAD(&f->list);
1371 list_add_tail(&f->list, &vsi->mac_filter_list);
1374 /* increment counter and add a new flag if needed */
1380 } else if (is_netdev) {
1381 if (!f->is_netdev) {
1382 f->is_netdev = true;
1390 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1391 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1399 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1400 * @vsi: the VSI to be searched
1401 * @macaddr: the MAC address
1403 * @is_vf: make sure it's a VF filter, else doesn't matter
1404 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1406 * NOTE: This function is expected to be called with mac_filter_list_lock
1408 * ANOTHER NOTE: This function MUST be called from within the context of
1409 * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1410 * instead of list_for_each_entry().
1412 void i40e_del_filter(struct i40e_vsi *vsi,
1413 u8 *macaddr, s16 vlan,
1414 bool is_vf, bool is_netdev)
1416 struct i40e_mac_filter *f;
1418 if (!vsi || !macaddr)
1421 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1422 if (!f || f->counter == 0)
1430 } else if (is_netdev) {
1432 f->is_netdev = false;
1436 /* make sure we don't remove a filter in use by VF or netdev */
1439 min_f += (f->is_vf ? 1 : 0);
1440 min_f += (f->is_netdev ? 1 : 0);
1442 if (f->counter > min_f)
1446 /* counter == 0 tells sync_filters_subtask to
1447 * remove the filter from the firmware's list
1449 if (f->counter == 0) {
1450 if ((f->state == I40E_FILTER_FAILED) ||
1451 (f->state == I40E_FILTER_NEW)) {
1452 /* this one never got added by the FW. Just remove it,
1453 * no need to sync anything.
1458 f->state = I40E_FILTER_REMOVE;
1459 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1460 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1466 * i40e_set_mac - NDO callback to set mac address
1467 * @netdev: network interface device structure
1468 * @p: pointer to an address structure
1470 * Returns 0 on success, negative on failure
1473 int i40e_set_mac(struct net_device *netdev, void *p)
1475 static int i40e_set_mac(struct net_device *netdev, void *p)
1478 struct i40e_netdev_priv *np = netdev_priv(netdev);
1479 struct i40e_vsi *vsi = np->vsi;
1480 struct i40e_pf *pf = vsi->back;
1481 struct i40e_hw *hw = &pf->hw;
1482 struct sockaddr *addr = p;
1484 if (!is_valid_ether_addr(addr->sa_data))
1485 return -EADDRNOTAVAIL;
1487 if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1488 netdev_info(netdev, "already using mac address %pM\n",
1493 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1494 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1495 return -EADDRNOTAVAIL;
1497 if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1498 netdev_info(netdev, "returning to hw mac address %pM\n",
1501 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1503 spin_lock_bh(&vsi->mac_filter_list_lock);
1504 i40e_del_mac_all_vlan(vsi, netdev->dev_addr, false, true);
1505 i40e_put_mac_in_vlan(vsi, addr->sa_data, false, true);
1506 spin_unlock_bh(&vsi->mac_filter_list_lock);
1507 ether_addr_copy(netdev->dev_addr, addr->sa_data);
1508 if (vsi->type == I40E_VSI_MAIN) {
1511 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1512 I40E_AQC_WRITE_TYPE_LAA_WOL,
1513 addr->sa_data, NULL);
1515 netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1516 i40e_stat_str(hw, ret),
1517 i40e_aq_str(hw, hw->aq.asq_last_status));
1520 /* schedule our worker thread which will take care of
1521 * applying the new filter changes
1523 i40e_service_event_schedule(vsi->back);
1528 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1529 * @vsi: the VSI being setup
1530 * @ctxt: VSI context structure
1531 * @enabled_tc: Enabled TCs bitmap
1532 * @is_add: True if called before Add VSI
1534 * Setup VSI queue mapping for enabled traffic classes.
1537 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1538 struct i40e_vsi_context *ctxt,
1542 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1543 struct i40e_vsi_context *ctxt,
1548 struct i40e_pf *pf = vsi->back;
1558 sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1561 if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1562 /* Find numtc from enabled TC bitmap */
1563 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1564 if (enabled_tc & BIT(i)) /* TC is enabled */
1568 dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1572 /* At least TC0 is enabled in case of non-DCB case */
1576 vsi->tc_config.numtc = numtc;
1577 vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1578 /* Number of queues per enabled TC */
1579 qcount = vsi->alloc_queue_pairs;
1581 num_tc_qps = qcount / numtc;
1582 num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1584 /* Setup queue offset/count for all TCs for given VSI */
1585 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1586 /* See if the given TC is enabled for the given VSI */
1587 if (vsi->tc_config.enabled_tc & BIT(i)) {
1591 switch (vsi->type) {
1593 qcount = min_t(int, pf->alloc_rss_size,
1598 qcount = num_tc_qps;
1602 case I40E_VSI_SRIOV:
1603 case I40E_VSI_VMDQ2:
1605 qcount = num_tc_qps;
1609 vsi->tc_config.tc_info[i].qoffset = offset;
1610 vsi->tc_config.tc_info[i].qcount = qcount;
1612 /* find the next higher power-of-2 of num queue pairs */
1615 while (num_qps && (BIT_ULL(pow) < qcount)) {
1620 vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1622 (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1623 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1627 /* TC is not enabled so set the offset to
1628 * default queue and allocate one queue
1631 vsi->tc_config.tc_info[i].qoffset = 0;
1632 vsi->tc_config.tc_info[i].qcount = 1;
1633 vsi->tc_config.tc_info[i].netdev_tc = 0;
1637 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1640 /* Set actual Tx/Rx queue pairs */
1641 vsi->num_queue_pairs = offset;
1642 if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1643 if (vsi->req_queue_pairs > 0)
1644 vsi->num_queue_pairs = vsi->req_queue_pairs;
1645 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1646 vsi->num_queue_pairs = pf->num_lan_msix;
1649 /* Scheduler section valid can only be set for ADD VSI */
1651 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1653 ctxt->info.up_enable_bits = enabled_tc;
1655 if (vsi->type == I40E_VSI_SRIOV) {
1656 ctxt->info.mapping_flags |=
1657 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1658 for (i = 0; i < vsi->num_queue_pairs; i++)
1659 ctxt->info.queue_mapping[i] =
1660 cpu_to_le16(vsi->base_queue + i);
1662 ctxt->info.mapping_flags |=
1663 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1664 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1666 ctxt->info.valid_sections |= cpu_to_le16(sections);
1670 * i40e_set_rx_mode - NDO callback to set the netdev filters
1671 * @netdev: network interface device structure
1674 void i40e_set_rx_mode(struct net_device *netdev)
1676 static void i40e_set_rx_mode(struct net_device *netdev)
1679 struct i40e_netdev_priv *np = netdev_priv(netdev);
1680 struct i40e_mac_filter *f, *ftmp;
1681 struct i40e_vsi *vsi = np->vsi;
1682 struct netdev_hw_addr *uca;
1683 struct netdev_hw_addr *mca;
1684 struct netdev_hw_addr *ha;
1686 spin_lock_bh(&vsi->mac_filter_list_lock);
1688 /* add addr if not already in the filter list */
1689 netdev_for_each_uc_addr(uca, netdev) {
1690 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1691 if (i40e_is_vsi_in_vlan(vsi))
1692 i40e_put_mac_in_vlan(vsi, uca->addr,
1695 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1700 netdev_for_each_mc_addr(mca, netdev) {
1701 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1702 if (i40e_is_vsi_in_vlan(vsi))
1703 i40e_put_mac_in_vlan(vsi, mca->addr,
1706 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1711 /* remove filter if not in netdev list */
1712 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1717 netdev_for_each_mc_addr(mca, netdev)
1718 if (ether_addr_equal(mca->addr, f->macaddr))
1719 goto bottom_of_search_loop;
1721 netdev_for_each_uc_addr(uca, netdev)
1722 if (ether_addr_equal(uca->addr, f->macaddr))
1723 goto bottom_of_search_loop;
1725 for_each_dev_addr(netdev, ha)
1726 if (ether_addr_equal(ha->addr, f->macaddr))
1727 goto bottom_of_search_loop;
1729 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1730 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1732 bottom_of_search_loop:
1735 spin_unlock_bh(&vsi->mac_filter_list_lock);
1737 /* check for other flag changes */
1738 if (vsi->current_netdev_flags != vsi->netdev->flags) {
1739 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1740 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1743 /* schedule our worker thread which will take care of
1744 * applying the new filter changes
1746 i40e_service_event_schedule(vsi->back);
1750 * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1751 * @vsi: pointer to vsi struct
1752 * @from: Pointer to list which contains MAC filter entries - changes to
1753 * those entries needs to be undone.
1755 * MAC filter entries from list were slated to be removed from device.
1757 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1758 struct list_head *from)
1760 struct i40e_mac_filter *f, *ftmp;
1762 list_for_each_entry_safe(f, ftmp, from, list) {
1763 /* Move the element back into MAC filter list*/
1764 list_move_tail(&f->list, &vsi->mac_filter_list);
1769 * i40e_update_filter_state - Update filter state based on return data
1771 * @count: Number of filters added
1772 * @add_list: return data from fw
1773 * @head: pointer to first filter in current batch
1774 * @aq_err: status from fw
1776 * MAC filter entries from list were slated to be added to device. Returns
1777 * number of successful filters. Note that 0 does NOT mean success!
1780 i40e_update_filter_state(int count,
1781 struct i40e_aqc_add_macvlan_element_data *add_list,
1782 struct i40e_mac_filter *add_head, int aq_err)
1790 /* Everything's good, mark all filters active. */
1791 for (i = 0; i < count ; i++) {
1792 add_head->state = I40E_FILTER_ACTIVE;
1793 add_head = list_next_entry(add_head, list);
1795 } else if (aq_err == I40E_AQ_RC_ENOSPC) {
1796 /* Device ran out of filter space. Check the return value
1797 * for each filter to see which ones are active.
1799 for (i = 0; i < count ; i++) {
1800 if (add_list[i].match_method ==
1801 I40E_AQC_MM_ERR_NO_RES) {
1802 add_head->state = I40E_FILTER_FAILED;
1804 add_head->state = I40E_FILTER_ACTIVE;
1807 add_head = list_next_entry(add_head, list);
1810 /* Some other horrible thing happened, fail all filters */
1812 for (i = 0; i < count ; i++) {
1813 add_head->state = I40E_FILTER_FAILED;
1814 add_head = list_next_entry(add_head, list);
1821 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1822 * @vsi: ptr to the VSI
1824 * Push any outstanding VSI filter changes through the AdminQ.
1826 * Returns 0 or error value
1828 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1830 struct i40e_mac_filter *f, *ftmp, *add_head = NULL;
1831 struct list_head tmp_add_list, tmp_del_list;
1832 struct i40e_hw *hw = &vsi->back->hw;
1833 bool promisc_changed = false;
1834 char vsi_name[16] = "PF";
1835 int filter_list_len = 0;
1836 u32 changed_flags = 0;
1837 i40e_status aq_ret = 0;
1847 /* empty array typed pointers, kcalloc later */
1848 struct i40e_aqc_add_macvlan_element_data *add_list;
1849 struct i40e_aqc_remove_macvlan_element_data *del_list;
1851 while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1852 usleep_range(1000, 2000);
1856 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1857 vsi->current_netdev_flags = vsi->netdev->flags;
1860 INIT_LIST_HEAD(&tmp_add_list);
1861 INIT_LIST_HEAD(&tmp_del_list);
1863 if (vsi->type == I40E_VSI_SRIOV)
1864 snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
1865 else if (vsi->type != I40E_VSI_MAIN)
1866 snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
1868 if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1869 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1871 spin_lock_bh(&vsi->mac_filter_list_lock);
1872 /* Create a list of filters to delete. */
1873 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1874 if (f->state == I40E_FILTER_REMOVE) {
1875 WARN_ON(f->counter != 0);
1876 /* Move the element into temporary del_list */
1877 list_move_tail(&f->list, &tmp_del_list);
1878 vsi->active_filters--;
1880 if (f->state == I40E_FILTER_NEW) {
1881 WARN_ON(f->counter == 0);
1882 /* Move the element into temporary add_list */
1883 list_move_tail(&f->list, &tmp_add_list);
1886 spin_unlock_bh(&vsi->mac_filter_list_lock);
1889 /* Now process 'del_list' outside the lock */
1890 if (!list_empty(&tmp_del_list)) {
1891 filter_list_len = hw->aq.asq_buf_size /
1892 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1893 list_size = filter_list_len *
1894 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1895 del_list = kzalloc(list_size, GFP_ATOMIC);
1897 /* Undo VSI's MAC filter entry element updates */
1898 spin_lock_bh(&vsi->mac_filter_list_lock);
1899 i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1900 spin_unlock_bh(&vsi->mac_filter_list_lock);
1905 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1908 /* add to delete list */
1909 ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1910 if (f->vlan == I40E_VLAN_ANY) {
1911 del_list[num_del].vlan_tag = 0;
1912 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1914 del_list[num_del].vlan_tag =
1915 cpu_to_le16((u16)(f->vlan));
1918 cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1919 del_list[num_del].flags = cmd_flags;
1922 /* flush a full buffer */
1923 if (num_del == filter_list_len) {
1924 aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid,
1927 aq_err = hw->aq.asq_last_status;
1929 memset(del_list, 0, list_size);
1931 /* Explicitly ignore and do not report when
1932 * firmware returns ENOENT.
1934 if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1936 dev_info(&pf->pdev->dev,
1937 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1939 i40e_stat_str(hw, aq_ret),
1940 i40e_aq_str(hw, aq_err));
1943 /* Release memory for MAC filter entries which were
1944 * synced up with HW.
1951 aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, del_list,
1953 aq_err = hw->aq.asq_last_status;
1956 /* Explicitly ignore and do not report when firmware
1959 if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1961 dev_info(&pf->pdev->dev,
1962 "ignoring delete macvlan error on %s, err %s aq_err %s\n",
1964 i40e_stat_str(hw, aq_ret),
1965 i40e_aq_str(hw, aq_err));
1973 if (!list_empty(&tmp_add_list)) {
1974 /* Do all the adds now. */
1975 filter_list_len = hw->aq.asq_buf_size /
1976 sizeof(struct i40e_aqc_add_macvlan_element_data);
1977 list_size = filter_list_len *
1978 sizeof(struct i40e_aqc_add_macvlan_element_data);
1979 add_list = kzalloc(list_size, GFP_ATOMIC);
1985 list_for_each_entry(f, &tmp_add_list, list) {
1986 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1988 f->state = I40E_FILTER_FAILED;
1991 /* add to add array */
1995 ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
1996 if (f->vlan == I40E_VLAN_ANY) {
1997 add_list[num_add].vlan_tag = 0;
1998 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2000 add_list[num_add].vlan_tag =
2001 cpu_to_le16((u16)(f->vlan));
2003 add_list[num_add].queue_number = 0;
2004 cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2005 add_list[num_add].flags = cpu_to_le16(cmd_flags);
2008 /* flush a full buffer */
2009 if (num_add == filter_list_len) {
2010 aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2013 aq_err = hw->aq.asq_last_status;
2014 fcnt = i40e_update_filter_state(num_add,
2018 vsi->active_filters += fcnt;
2020 if (fcnt != num_add) {
2021 promisc_changed = true;
2022 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2024 vsi->promisc_threshold =
2025 (vsi->active_filters * 3) / 4;
2026 dev_warn(&pf->pdev->dev,
2027 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2028 i40e_aq_str(hw, aq_err),
2031 memset(add_list, 0, list_size);
2036 aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2037 add_list, num_add, NULL);
2038 aq_err = hw->aq.asq_last_status;
2039 fcnt = i40e_update_filter_state(num_add, add_list,
2041 vsi->active_filters += fcnt;
2042 if (fcnt != num_add) {
2043 promisc_changed = true;
2044 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2046 vsi->promisc_threshold =
2047 (vsi->active_filters * 3) / 4;
2048 dev_warn(&pf->pdev->dev,
2049 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2050 i40e_aq_str(hw, aq_err), vsi_name);
2053 /* Now move all of the filters from the temp add list back to
2056 spin_lock_bh(&vsi->mac_filter_list_lock);
2057 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2058 list_move_tail(&f->list, &vsi->mac_filter_list);
2060 spin_unlock_bh(&vsi->mac_filter_list_lock);
2065 /* Check to see if we can drop out of overflow promiscuous mode. */
2066 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state) &&
2067 (vsi->active_filters < vsi->promisc_threshold)) {
2068 int failed_count = 0;
2069 /* See if we have any failed filters. We can't drop out of
2070 * promiscuous until these have all been deleted.
2072 spin_lock_bh(&vsi->mac_filter_list_lock);
2073 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2074 if (f->state == I40E_FILTER_FAILED)
2077 spin_unlock_bh(&vsi->mac_filter_list_lock);
2078 if (!failed_count) {
2079 dev_info(&pf->pdev->dev,
2080 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2082 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2083 promisc_changed = true;
2084 vsi->promisc_threshold = 0;
2088 /* if the VF is not trusted do not do promisc */
2089 if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2090 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2094 /* check for changes in promiscuous modes */
2095 if (changed_flags & IFF_ALLMULTI) {
2096 bool cur_multipromisc;
2098 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2099 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2104 retval = i40e_aq_rc_to_posix(aq_ret,
2105 hw->aq.asq_last_status);
2106 dev_info(&pf->pdev->dev,
2107 "set multi promisc failed on %s, err %s aq_err %s\n",
2109 i40e_stat_str(hw, aq_ret),
2110 i40e_aq_str(hw, hw->aq.asq_last_status));
2113 if ((changed_flags & IFF_PROMISC) ||
2115 test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))) {
2118 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2119 test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2121 if ((vsi->type == I40E_VSI_MAIN) &&
2122 (pf->lan_veb != I40E_NO_VEB) &&
2123 !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2124 /* set defport ON for Main VSI instead of true promisc
2125 * this way we will get all unicast/multicast and VLAN
2126 * promisc behavior but will not get VF or VMDq traffic
2127 * replicated on the Main VSI.
2129 if (pf->cur_promisc != cur_promisc) {
2130 pf->cur_promisc = cur_promisc;
2133 i40e_aq_set_default_vsi(hw,
2138 i40e_aq_clear_default_vsi(hw,
2142 retval = i40e_aq_rc_to_posix(aq_ret,
2143 hw->aq.asq_last_status);
2144 dev_info(&pf->pdev->dev,
2145 "Set default VSI failed on %s, err %s, aq_err %s\n",
2147 i40e_stat_str(hw, aq_ret),
2149 hw->aq.asq_last_status));
2153 aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2160 i40e_aq_rc_to_posix(aq_ret,
2161 hw->aq.asq_last_status);
2162 dev_info(&pf->pdev->dev,
2163 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2165 i40e_stat_str(hw, aq_ret),
2167 hw->aq.asq_last_status));
2169 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2175 i40e_aq_rc_to_posix(aq_ret,
2176 hw->aq.asq_last_status);
2177 dev_info(&pf->pdev->dev,
2178 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2180 i40e_stat_str(hw, aq_ret),
2182 hw->aq.asq_last_status));
2185 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2189 retval = i40e_aq_rc_to_posix(aq_ret,
2190 pf->hw.aq.asq_last_status);
2191 dev_info(&pf->pdev->dev,
2192 "set brdcast promisc failed, err %s, aq_err %s\n",
2193 i40e_stat_str(hw, aq_ret),
2195 hw->aq.asq_last_status));
2199 /* if something went wrong then set the changed flag so we try again */
2201 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2203 clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2208 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2209 * @pf: board private structure
2211 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2215 if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2217 pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2219 for (v = 0; v < pf->num_alloc_vsi; v++) {
2221 (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2222 int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2225 /* come back and try again later */
2226 pf->flags |= I40E_FLAG_FILTER_SYNC;
2234 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2235 * @netdev: network interface device structure
2236 * @new_mtu: new value for maximum frame size
2238 * Returns 0 on success, negative on failure
2240 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2242 struct i40e_netdev_priv *np = netdev_priv(netdev);
2243 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2244 struct i40e_vsi *vsi = np->vsi;
2246 /* MTU < 68 is an error and causes problems on some kernels */
2247 if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2250 netdev_info(netdev, "changing MTU from %d to %d\n",
2251 netdev->mtu, new_mtu);
2252 netdev->mtu = new_mtu;
2253 if (netif_running(netdev))
2254 i40e_vsi_reinit_locked(vsi);
2255 i40e_notify_client_of_l2_param_changes(vsi);
2260 * i40e_ioctl - Access the hwtstamp interface
2261 * @netdev: network interface device structure
2262 * @ifr: interface request data
2263 * @cmd: ioctl command
2265 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2267 struct i40e_netdev_priv *np = netdev_priv(netdev);
2268 struct i40e_pf *pf = np->vsi->back;
2272 return i40e_ptp_get_ts_config(pf, ifr);
2274 return i40e_ptp_set_ts_config(pf, ifr);
2281 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2282 * @vsi: the vsi being adjusted
2284 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2286 struct i40e_vsi_context ctxt;
2289 if ((vsi->info.valid_sections &
2290 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2291 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2292 return; /* already enabled */
2294 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2295 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2296 I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2298 ctxt.seid = vsi->seid;
2299 ctxt.info = vsi->info;
2300 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2302 dev_info(&vsi->back->pdev->dev,
2303 "update vlan stripping failed, err %s aq_err %s\n",
2304 i40e_stat_str(&vsi->back->hw, ret),
2305 i40e_aq_str(&vsi->back->hw,
2306 vsi->back->hw.aq.asq_last_status));
2311 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2312 * @vsi: the vsi being adjusted
2314 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2316 struct i40e_vsi_context ctxt;
2319 if ((vsi->info.valid_sections &
2320 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2321 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2322 I40E_AQ_VSI_PVLAN_EMOD_MASK))
2323 return; /* already disabled */
2325 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2326 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2327 I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2329 ctxt.seid = vsi->seid;
2330 ctxt.info = vsi->info;
2331 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2333 dev_info(&vsi->back->pdev->dev,
2334 "update vlan stripping failed, err %s aq_err %s\n",
2335 i40e_stat_str(&vsi->back->hw, ret),
2336 i40e_aq_str(&vsi->back->hw,
2337 vsi->back->hw.aq.asq_last_status));
2342 * i40e_vlan_rx_register - Setup or shutdown vlan offload
2343 * @netdev: network interface to be adjusted
2344 * @features: netdev features to test if VLAN offload is enabled or not
2346 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2348 struct i40e_netdev_priv *np = netdev_priv(netdev);
2349 struct i40e_vsi *vsi = np->vsi;
2351 if (features & NETIF_F_HW_VLAN_CTAG_RX)
2352 i40e_vlan_stripping_enable(vsi);
2354 i40e_vlan_stripping_disable(vsi);
2358 * i40e_vsi_add_vlan - Add vsi membership for given vlan
2359 * @vsi: the vsi being configured
2360 * @vid: vlan id to be added (0 = untagged only , -1 = any)
2362 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2364 struct i40e_mac_filter *f, *ftmp, *add_f;
2365 bool is_netdev, is_vf;
2367 is_vf = (vsi->type == I40E_VSI_SRIOV);
2368 is_netdev = !!(vsi->netdev);
2370 /* Locked once because all functions invoked below iterates list*/
2371 spin_lock_bh(&vsi->mac_filter_list_lock);
2374 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2377 dev_info(&vsi->back->pdev->dev,
2378 "Could not add vlan filter %d for %pM\n",
2379 vid, vsi->netdev->dev_addr);
2380 spin_unlock_bh(&vsi->mac_filter_list_lock);
2385 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2386 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2388 dev_info(&vsi->back->pdev->dev,
2389 "Could not add vlan filter %d for %pM\n",
2391 spin_unlock_bh(&vsi->mac_filter_list_lock);
2396 /* Now if we add a vlan tag, make sure to check if it is the first
2397 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2398 * with 0, so we now accept untagged and specified tagged traffic
2399 * (and not all tags along with untagged)
2402 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2404 is_vf, is_netdev)) {
2405 i40e_del_filter(vsi, vsi->netdev->dev_addr,
2406 I40E_VLAN_ANY, is_vf, is_netdev);
2407 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2410 dev_info(&vsi->back->pdev->dev,
2411 "Could not add filter 0 for %pM\n",
2412 vsi->netdev->dev_addr);
2413 spin_unlock_bh(&vsi->mac_filter_list_lock);
2419 /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2420 if (vid > 0 && !vsi->info.pvid) {
2421 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2422 if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2425 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2427 add_f = i40e_add_filter(vsi, f->macaddr,
2428 0, is_vf, is_netdev);
2430 dev_info(&vsi->back->pdev->dev,
2431 "Could not add filter 0 for %pM\n",
2433 spin_unlock_bh(&vsi->mac_filter_list_lock);
2439 spin_unlock_bh(&vsi->mac_filter_list_lock);
2441 /* schedule our worker thread which will take care of
2442 * applying the new filter changes
2444 i40e_service_event_schedule(vsi->back);
2449 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2450 * @vsi: the vsi being configured
2451 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2453 * Return: 0 on success or negative otherwise
2455 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2457 struct net_device *netdev = vsi->netdev;
2458 struct i40e_mac_filter *f, *ftmp, *add_f;
2459 bool is_vf, is_netdev;
2460 int filter_count = 0;
2462 is_vf = (vsi->type == I40E_VSI_SRIOV);
2463 is_netdev = !!(netdev);
2465 /* Locked once because all functions invoked below iterates list */
2466 spin_lock_bh(&vsi->mac_filter_list_lock);
2469 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2471 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
2472 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2474 /* go through all the filters for this VSI and if there is only
2475 * vid == 0 it means there are no other filters, so vid 0 must
2476 * be replaced with -1. This signifies that we should from now
2477 * on accept any traffic (with any tag present, or untagged)
2479 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2482 ether_addr_equal(netdev->dev_addr, f->macaddr))
2490 if (!filter_count && is_netdev) {
2491 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2492 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2495 dev_info(&vsi->back->pdev->dev,
2496 "Could not add filter %d for %pM\n",
2497 I40E_VLAN_ANY, netdev->dev_addr);
2498 spin_unlock_bh(&vsi->mac_filter_list_lock);
2503 if (!filter_count) {
2504 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2505 i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2506 add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2509 dev_info(&vsi->back->pdev->dev,
2510 "Could not add filter %d for %pM\n",
2511 I40E_VLAN_ANY, f->macaddr);
2512 spin_unlock_bh(&vsi->mac_filter_list_lock);
2518 spin_unlock_bh(&vsi->mac_filter_list_lock);
2520 /* schedule our worker thread which will take care of
2521 * applying the new filter changes
2523 i40e_service_event_schedule(vsi->back);
2528 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2529 * @netdev: network interface to be adjusted
2530 * @vid: vlan id to be added
2532 * net_device_ops implementation for adding vlan ids
2535 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2536 __always_unused __be16 proto, u16 vid)
2538 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2539 __always_unused __be16 proto, u16 vid)
2542 struct i40e_netdev_priv *np = netdev_priv(netdev);
2543 struct i40e_vsi *vsi = np->vsi;
2549 /* If the network stack called us with vid = 0 then
2550 * it is asking to receive priority tagged packets with
2551 * vlan id 0. Our HW receives them by default when configured
2552 * to receive untagged packets so there is no need to add an
2553 * extra filter for vlan 0 tagged packets.
2556 ret = i40e_vsi_add_vlan(vsi, vid);
2558 if (!ret && (vid < VLAN_N_VID))
2559 set_bit(vid, vsi->active_vlans);
2565 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2566 * @netdev: network interface to be adjusted
2567 * @vid: vlan id to be removed
2569 * net_device_ops implementation for removing vlan ids
2572 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2573 __always_unused __be16 proto, u16 vid)
2575 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2576 __always_unused __be16 proto, u16 vid)
2579 struct i40e_netdev_priv *np = netdev_priv(netdev);
2580 struct i40e_vsi *vsi = np->vsi;
2582 /* return code is ignored as there is nothing a user
2583 * can do about failure to remove and a log message was
2584 * already printed from the other function
2586 i40e_vsi_kill_vlan(vsi, vid);
2588 clear_bit(vid, vsi->active_vlans);
2594 * i40e_macaddr_init - explicitly write the mac address filters
2596 * @vsi: pointer to the vsi
2597 * @macaddr: the MAC address
2599 * This is needed when the macaddr has been obtained by other
2600 * means than the default, e.g., from Open Firmware or IDPROM.
2601 * Returns 0 on success, negative on failure
2603 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2606 struct i40e_aqc_add_macvlan_element_data element;
2608 ret = i40e_aq_mac_address_write(&vsi->back->hw,
2609 I40E_AQC_WRITE_TYPE_LAA_WOL,
2612 dev_info(&vsi->back->pdev->dev,
2613 "Addr change for VSI failed: %d\n", ret);
2614 return -EADDRNOTAVAIL;
2617 memset(&element, 0, sizeof(element));
2618 ether_addr_copy(element.mac_addr, macaddr);
2619 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2620 ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2622 dev_info(&vsi->back->pdev->dev,
2623 "add filter failed err %s aq_err %s\n",
2624 i40e_stat_str(&vsi->back->hw, ret),
2625 i40e_aq_str(&vsi->back->hw,
2626 vsi->back->hw.aq.asq_last_status));
2632 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2633 * @vsi: the vsi being brought back up
2635 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2642 i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2644 for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2645 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2650 * i40e_vsi_add_pvid - Add pvid for the VSI
2651 * @vsi: the vsi being adjusted
2652 * @vid: the vlan id to set as a PVID
2654 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2656 struct i40e_vsi_context ctxt;
2659 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2660 vsi->info.pvid = cpu_to_le16(vid);
2661 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2662 I40E_AQ_VSI_PVLAN_INSERT_PVID |
2663 I40E_AQ_VSI_PVLAN_EMOD_STR;
2665 ctxt.seid = vsi->seid;
2666 ctxt.info = vsi->info;
2667 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2669 dev_info(&vsi->back->pdev->dev,
2670 "add pvid failed, err %s aq_err %s\n",
2671 i40e_stat_str(&vsi->back->hw, ret),
2672 i40e_aq_str(&vsi->back->hw,
2673 vsi->back->hw.aq.asq_last_status));
2681 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2682 * @vsi: the vsi being adjusted
2684 * Just use the vlan_rx_register() service to put it back to normal
2686 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2688 i40e_vlan_stripping_disable(vsi);
2694 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2695 * @vsi: ptr to the VSI
2697 * If this function returns with an error, then it's possible one or
2698 * more of the rings is populated (while the rest are not). It is the
2699 * callers duty to clean those orphaned rings.
2701 * Return 0 on success, negative on failure
2703 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2707 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2708 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2714 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2715 * @vsi: ptr to the VSI
2717 * Free VSI's transmit software resources
2719 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2726 for (i = 0; i < vsi->num_queue_pairs; i++)
2727 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2728 i40e_free_tx_resources(vsi->tx_rings[i]);
2732 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2733 * @vsi: ptr to the VSI
2735 * If this function returns with an error, then it's possible one or
2736 * more of the rings is populated (while the rest are not). It is the
2737 * callers duty to clean those orphaned rings.
2739 * Return 0 on success, negative on failure
2741 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2745 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2746 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2748 i40e_fcoe_setup_ddp_resources(vsi);
2754 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2755 * @vsi: ptr to the VSI
2757 * Free all receive software resources
2759 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2766 for (i = 0; i < vsi->num_queue_pairs; i++)
2767 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2768 i40e_free_rx_resources(vsi->rx_rings[i]);
2770 i40e_fcoe_free_ddp_resources(vsi);
2775 * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2776 * @ring: The Tx ring to configure
2778 * This enables/disables XPS for a given Tx descriptor ring
2779 * based on the TCs enabled for the VSI that ring belongs to.
2781 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2783 struct i40e_vsi *vsi = ring->vsi;
2786 if (!ring->q_vector || !ring->netdev)
2789 /* Single TC mode enable XPS */
2790 if (vsi->tc_config.numtc <= 1) {
2791 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2792 netif_set_xps_queue(ring->netdev,
2793 &ring->q_vector->affinity_mask,
2795 } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2796 /* Disable XPS to allow selection based on TC */
2797 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2798 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2799 free_cpumask_var(mask);
2802 /* schedule our worker thread which will take care of
2803 * applying the new filter changes
2805 i40e_service_event_schedule(vsi->back);
2809 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2810 * @ring: The Tx ring to configure
2812 * Configure the Tx descriptor ring in the HMC context.
2814 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2816 struct i40e_vsi *vsi = ring->vsi;
2817 u16 pf_q = vsi->base_queue + ring->queue_index;
2818 struct i40e_hw *hw = &vsi->back->hw;
2819 struct i40e_hmc_obj_txq tx_ctx;
2820 i40e_status err = 0;
2823 /* some ATR related tx ring init */
2824 if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2825 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2826 ring->atr_count = 0;
2828 ring->atr_sample_rate = 0;
2832 i40e_config_xps_tx_ring(ring);
2834 /* clear the context structure first */
2835 memset(&tx_ctx, 0, sizeof(tx_ctx));
2837 tx_ctx.new_context = 1;
2838 tx_ctx.base = (ring->dma / 128);
2839 tx_ctx.qlen = ring->count;
2840 tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2841 I40E_FLAG_FD_ATR_ENABLED));
2843 tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2845 tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2846 /* FDIR VSI tx ring can still use RS bit and writebacks */
2847 if (vsi->type != I40E_VSI_FDIR)
2848 tx_ctx.head_wb_ena = 1;
2849 tx_ctx.head_wb_addr = ring->dma +
2850 (ring->count * sizeof(struct i40e_tx_desc));
2852 /* As part of VSI creation/update, FW allocates certain
2853 * Tx arbitration queue sets for each TC enabled for
2854 * the VSI. The FW returns the handles to these queue
2855 * sets as part of the response buffer to Add VSI,
2856 * Update VSI, etc. AQ commands. It is expected that
2857 * these queue set handles be associated with the Tx
2858 * queues by the driver as part of the TX queue context
2859 * initialization. This has to be done regardless of
2860 * DCB as by default everything is mapped to TC0.
2862 tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2863 tx_ctx.rdylist_act = 0;
2865 /* clear the context in the HMC */
2866 err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2868 dev_info(&vsi->back->pdev->dev,
2869 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2870 ring->queue_index, pf_q, err);
2874 /* set the context in the HMC */
2875 err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2877 dev_info(&vsi->back->pdev->dev,
2878 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2879 ring->queue_index, pf_q, err);
2883 /* Now associate this queue with this PCI function */
2884 if (vsi->type == I40E_VSI_VMDQ2) {
2885 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2886 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2887 I40E_QTX_CTL_VFVM_INDX_MASK;
2889 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2892 qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2893 I40E_QTX_CTL_PF_INDX_MASK);
2894 wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2897 /* cache tail off for easier writes later */
2898 ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2904 * i40e_configure_rx_ring - Configure a receive ring context
2905 * @ring: The Rx ring to configure
2907 * Configure the Rx descriptor ring in the HMC context.
2909 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2911 struct i40e_vsi *vsi = ring->vsi;
2912 u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2913 u16 pf_q = vsi->base_queue + ring->queue_index;
2914 struct i40e_hw *hw = &vsi->back->hw;
2915 struct i40e_hmc_obj_rxq rx_ctx;
2916 i40e_status err = 0;
2920 /* clear the context structure first */
2921 memset(&rx_ctx, 0, sizeof(rx_ctx));
2923 ring->rx_buf_len = vsi->rx_buf_len;
2925 rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2927 rx_ctx.base = (ring->dma / 128);
2928 rx_ctx.qlen = ring->count;
2930 /* use 32 byte descriptors */
2933 /* descriptor type is always zero
2936 rx_ctx.hsplit_0 = 0;
2938 rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
2939 if (hw->revision_id == 0)
2940 rx_ctx.lrxqthresh = 0;
2942 rx_ctx.lrxqthresh = 2;
2943 rx_ctx.crcstrip = 1;
2945 /* this controls whether VLAN is stripped from inner headers */
2948 rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2950 /* set the prefena field to 1 because the manual says to */
2953 /* clear the context in the HMC */
2954 err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2956 dev_info(&vsi->back->pdev->dev,
2957 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2958 ring->queue_index, pf_q, err);
2962 /* set the context in the HMC */
2963 err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2965 dev_info(&vsi->back->pdev->dev,
2966 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2967 ring->queue_index, pf_q, err);
2971 /* cache tail for quicker writes, and clear the reg before use */
2972 ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2973 writel(0, ring->tail);
2975 i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2981 * i40e_vsi_configure_tx - Configure the VSI for Tx
2982 * @vsi: VSI structure describing this set of rings and resources
2984 * Configure the Tx VSI for operation.
2986 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2991 for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2992 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2998 * i40e_vsi_configure_rx - Configure the VSI for Rx
2999 * @vsi: the VSI being configured
3001 * Configure the Rx VSI for operation.
3003 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3008 if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
3009 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
3010 + ETH_FCS_LEN + VLAN_HLEN;
3012 vsi->max_frame = I40E_RXBUFFER_2048;
3014 vsi->rx_buf_len = I40E_RXBUFFER_2048;
3017 /* setup rx buffer for FCoE */
3018 if ((vsi->type == I40E_VSI_FCOE) &&
3019 (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
3020 vsi->rx_buf_len = I40E_RXBUFFER_3072;
3021 vsi->max_frame = I40E_RXBUFFER_3072;
3024 #endif /* I40E_FCOE */
3025 /* round up for the chip's needs */
3026 vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
3027 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3029 /* set up individual rings */
3030 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3031 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3037 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3038 * @vsi: ptr to the VSI
3040 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3042 struct i40e_ring *tx_ring, *rx_ring;
3043 u16 qoffset, qcount;
3046 if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3047 /* Reset the TC information */
3048 for (i = 0; i < vsi->num_queue_pairs; i++) {
3049 rx_ring = vsi->rx_rings[i];
3050 tx_ring = vsi->tx_rings[i];
3051 rx_ring->dcb_tc = 0;
3052 tx_ring->dcb_tc = 0;
3056 for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3057 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3060 qoffset = vsi->tc_config.tc_info[n].qoffset;
3061 qcount = vsi->tc_config.tc_info[n].qcount;
3062 for (i = qoffset; i < (qoffset + qcount); i++) {
3063 rx_ring = vsi->rx_rings[i];
3064 tx_ring = vsi->tx_rings[i];
3065 rx_ring->dcb_tc = n;
3066 tx_ring->dcb_tc = n;
3072 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3073 * @vsi: ptr to the VSI
3075 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3077 struct i40e_pf *pf = vsi->back;
3081 i40e_set_rx_mode(vsi->netdev);
3083 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3084 err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3086 dev_warn(&pf->pdev->dev,
3087 "could not set up macaddr; err %d\n", err);
3093 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3094 * @vsi: Pointer to the targeted VSI
3096 * This function replays the hlist on the hw where all the SB Flow Director
3097 * filters were saved.
3099 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3101 struct i40e_fdir_filter *filter;
3102 struct i40e_pf *pf = vsi->back;
3103 struct hlist_node *node;
3105 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3108 hlist_for_each_entry_safe(filter, node,
3109 &pf->fdir_filter_list, fdir_node) {
3110 i40e_add_del_fdir(vsi, filter, true);
3115 * i40e_vsi_configure - Set up the VSI for action
3116 * @vsi: the VSI being configured
3118 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3122 i40e_set_vsi_rx_mode(vsi);
3123 i40e_restore_vlan(vsi);
3124 i40e_vsi_config_dcb_rings(vsi);
3125 err = i40e_vsi_configure_tx(vsi);
3127 err = i40e_vsi_configure_rx(vsi);
3133 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3134 * @vsi: the VSI being configured
3136 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3138 struct i40e_pf *pf = vsi->back;
3139 struct i40e_hw *hw = &pf->hw;
3144 /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3145 * and PFINT_LNKLSTn registers, e.g.:
3146 * PFINT_ITRn[0..n-1] gets msix-1..msix-n (qpair interrupts)
3148 qp = vsi->base_queue;
3149 vector = vsi->base_vector;
3150 for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3151 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3153 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3154 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3155 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3156 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3158 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3159 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3160 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3162 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3163 INTRL_USEC_TO_REG(vsi->int_rate_limit));
3165 /* Linked list for the queuepairs assigned to this vector */
3166 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3167 for (q = 0; q < q_vector->num_ringpairs; q++) {
3170 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3171 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3172 (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3173 (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3175 << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3177 wr32(hw, I40E_QINT_RQCTL(qp), val);
3179 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3180 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3181 (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3182 ((qp+1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3184 << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3186 /* Terminate the linked list */
3187 if (q == (q_vector->num_ringpairs - 1))
3188 val |= (I40E_QUEUE_END_OF_LIST
3189 << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3191 wr32(hw, I40E_QINT_TQCTL(qp), val);
3200 * i40e_enable_misc_int_causes - enable the non-queue interrupts
3201 * @hw: ptr to the hardware info
3203 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3205 struct i40e_hw *hw = &pf->hw;
3208 /* clear things first */
3209 wr32(hw, I40E_PFINT_ICR0_ENA, 0); /* disable all */
3210 rd32(hw, I40E_PFINT_ICR0); /* read to clear */
3212 val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK |
3213 I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK |
3214 I40E_PFINT_ICR0_ENA_GRST_MASK |
3215 I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3216 I40E_PFINT_ICR0_ENA_GPIO_MASK |
3217 I40E_PFINT_ICR0_ENA_HMC_ERR_MASK |
3218 I40E_PFINT_ICR0_ENA_VFLR_MASK |
3219 I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3221 if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3222 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3224 if (pf->flags & I40E_FLAG_PTP)
3225 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3227 wr32(hw, I40E_PFINT_ICR0_ENA, val);
3229 /* SW_ITR_IDX = 0, but don't change INTENA */
3230 wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3231 I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3233 /* OTHER_ITR_IDX = 0 */
3234 wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3238 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3239 * @vsi: the VSI being configured
3241 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3243 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3244 struct i40e_pf *pf = vsi->back;
3245 struct i40e_hw *hw = &pf->hw;
3248 /* set the ITR configuration */
3249 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3250 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3251 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3252 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3253 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3254 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3255 wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3257 i40e_enable_misc_int_causes(pf);
3259 /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3260 wr32(hw, I40E_PFINT_LNKLST0, 0);
3262 /* Associate the queue pair to the vector and enable the queue int */
3263 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3264 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3265 (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3267 wr32(hw, I40E_QINT_RQCTL(0), val);
3269 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3270 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3271 (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3273 wr32(hw, I40E_QINT_TQCTL(0), val);
3278 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3279 * @pf: board private structure
3281 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3283 struct i40e_hw *hw = &pf->hw;
3285 wr32(hw, I40E_PFINT_DYN_CTL0,
3286 I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3291 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3292 * @pf: board private structure
3293 * @clearpba: true when all pending interrupt events should be cleared
3295 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3297 struct i40e_hw *hw = &pf->hw;
3300 val = I40E_PFINT_DYN_CTL0_INTENA_MASK |
3301 (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3302 (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3304 wr32(hw, I40E_PFINT_DYN_CTL0, val);
3309 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3310 * @irq: interrupt number
3311 * @data: pointer to a q_vector
3313 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3315 struct i40e_q_vector *q_vector = data;
3317 if (!q_vector->tx.ring && !q_vector->rx.ring)
3320 napi_schedule_irqoff(&q_vector->napi);
3326 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3327 * @vsi: the VSI being configured
3328 * @basename: name for the vector
3330 * Allocates MSI-X vectors and requests interrupts from the kernel.
3332 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3334 int q_vectors = vsi->num_q_vectors;
3335 struct i40e_pf *pf = vsi->back;
3336 int base = vsi->base_vector;
3341 for (vector = 0; vector < q_vectors; vector++) {
3342 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3344 if (q_vector->tx.ring && q_vector->rx.ring) {
3345 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3346 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3348 } else if (q_vector->rx.ring) {
3349 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3350 "%s-%s-%d", basename, "rx", rx_int_idx++);
3351 } else if (q_vector->tx.ring) {
3352 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3353 "%s-%s-%d", basename, "tx", tx_int_idx++);
3355 /* skip this unused q_vector */
3358 err = request_irq(pf->msix_entries[base + vector].vector,
3364 dev_info(&pf->pdev->dev,
3365 "MSIX request_irq failed, error: %d\n", err);
3366 goto free_queue_irqs;
3368 /* assign the mask for this irq */
3369 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3370 &q_vector->affinity_mask);
3373 vsi->irqs_ready = true;
3379 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3381 free_irq(pf->msix_entries[base + vector].vector,
3382 &(vsi->q_vectors[vector]));
3388 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3389 * @vsi: the VSI being un-configured
3391 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3393 struct i40e_pf *pf = vsi->back;
3394 struct i40e_hw *hw = &pf->hw;
3395 int base = vsi->base_vector;
3398 for (i = 0; i < vsi->num_queue_pairs; i++) {
3399 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3400 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3403 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3404 for (i = vsi->base_vector;
3405 i < (vsi->num_q_vectors + vsi->base_vector); i++)
3406 wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3409 for (i = 0; i < vsi->num_q_vectors; i++)
3410 synchronize_irq(pf->msix_entries[i + base].vector);
3412 /* Legacy and MSI mode - this stops all interrupt handling */
3413 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3414 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3416 synchronize_irq(pf->pdev->irq);
3421 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3422 * @vsi: the VSI being configured
3424 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3426 struct i40e_pf *pf = vsi->back;
3429 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3430 for (i = 0; i < vsi->num_q_vectors; i++)
3431 i40e_irq_dynamic_enable(vsi, i);
3433 i40e_irq_dynamic_enable_icr0(pf, true);
3436 i40e_flush(&pf->hw);
3441 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3442 * @pf: board private structure
3444 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3447 wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3448 i40e_flush(&pf->hw);
3452 * i40e_intr - MSI/Legacy and non-queue interrupt handler
3453 * @irq: interrupt number
3454 * @data: pointer to a q_vector
3456 * This is the handler used for all MSI/Legacy interrupts, and deals
3457 * with both queue and non-queue interrupts. This is also used in
3458 * MSIX mode to handle the non-queue interrupts.
3460 static irqreturn_t i40e_intr(int irq, void *data)
3462 struct i40e_pf *pf = (struct i40e_pf *)data;
3463 struct i40e_hw *hw = &pf->hw;
3464 irqreturn_t ret = IRQ_NONE;
3465 u32 icr0, icr0_remaining;
3468 icr0 = rd32(hw, I40E_PFINT_ICR0);
3469 ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3471 /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3472 if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3475 /* if interrupt but no bits showing, must be SWINT */
3476 if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3477 (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3480 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3481 (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3482 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3483 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3484 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3487 /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3488 if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3489 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3490 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3492 /* We do not have a way to disarm Queue causes while leaving
3493 * interrupt enabled for all other causes, ideally
3494 * interrupt should be disabled while we are in NAPI but
3495 * this is not a performance path and napi_schedule()
3496 * can deal with rescheduling.
3498 if (!test_bit(__I40E_DOWN, &pf->state))
3499 napi_schedule_irqoff(&q_vector->napi);
3502 if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3503 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3504 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3505 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3508 if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3509 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3510 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3513 if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3514 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3515 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3518 if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3519 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3520 set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3521 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3522 val = rd32(hw, I40E_GLGEN_RSTAT);
3523 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3524 >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3525 if (val == I40E_RESET_CORER) {
3527 } else if (val == I40E_RESET_GLOBR) {
3529 } else if (val == I40E_RESET_EMPR) {
3531 set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3535 if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3536 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3537 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3538 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3539 rd32(hw, I40E_PFHMC_ERRORINFO),
3540 rd32(hw, I40E_PFHMC_ERRORDATA));
3543 if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3544 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3546 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3547 icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3548 i40e_ptp_tx_hwtstamp(pf);
3552 /* If a critical error is pending we have no choice but to reset the
3554 * Report and mask out any remaining unexpected interrupts.
3556 icr0_remaining = icr0 & ena_mask;
3557 if (icr0_remaining) {
3558 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3560 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3561 (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3562 (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3563 dev_info(&pf->pdev->dev, "device will be reset\n");
3564 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3565 i40e_service_event_schedule(pf);
3567 ena_mask &= ~icr0_remaining;
3572 /* re-enable interrupt causes */
3573 wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3574 if (!test_bit(__I40E_DOWN, &pf->state)) {
3575 i40e_service_event_schedule(pf);
3576 i40e_irq_dynamic_enable_icr0(pf, false);
3583 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3584 * @tx_ring: tx ring to clean
3585 * @budget: how many cleans we're allowed
3587 * Returns true if there's any budget left (e.g. the clean is finished)
3589 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3591 struct i40e_vsi *vsi = tx_ring->vsi;
3592 u16 i = tx_ring->next_to_clean;
3593 struct i40e_tx_buffer *tx_buf;
3594 struct i40e_tx_desc *tx_desc;
3596 tx_buf = &tx_ring->tx_bi[i];
3597 tx_desc = I40E_TX_DESC(tx_ring, i);
3598 i -= tx_ring->count;
3601 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3603 /* if next_to_watch is not set then there is no work pending */
3607 /* prevent any other reads prior to eop_desc */
3608 read_barrier_depends();
3610 /* if the descriptor isn't done, no work yet to do */
3611 if (!(eop_desc->cmd_type_offset_bsz &
3612 cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3615 /* clear next_to_watch to prevent false hangs */
3616 tx_buf->next_to_watch = NULL;
3618 tx_desc->buffer_addr = 0;
3619 tx_desc->cmd_type_offset_bsz = 0;
3620 /* move past filter desc */
3625 i -= tx_ring->count;
3626 tx_buf = tx_ring->tx_bi;
3627 tx_desc = I40E_TX_DESC(tx_ring, 0);
3629 /* unmap skb header data */
3630 dma_unmap_single(tx_ring->dev,
3631 dma_unmap_addr(tx_buf, dma),
3632 dma_unmap_len(tx_buf, len),
3634 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3635 kfree(tx_buf->raw_buf);
3637 tx_buf->raw_buf = NULL;
3638 tx_buf->tx_flags = 0;
3639 tx_buf->next_to_watch = NULL;
3640 dma_unmap_len_set(tx_buf, len, 0);
3641 tx_desc->buffer_addr = 0;
3642 tx_desc->cmd_type_offset_bsz = 0;
3644 /* move us past the eop_desc for start of next FD desc */
3649 i -= tx_ring->count;
3650 tx_buf = tx_ring->tx_bi;
3651 tx_desc = I40E_TX_DESC(tx_ring, 0);
3654 /* update budget accounting */
3656 } while (likely(budget));
3658 i += tx_ring->count;
3659 tx_ring->next_to_clean = i;
3661 if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3662 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3668 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3669 * @irq: interrupt number
3670 * @data: pointer to a q_vector
3672 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3674 struct i40e_q_vector *q_vector = data;
3675 struct i40e_vsi *vsi;
3677 if (!q_vector->tx.ring)
3680 vsi = q_vector->tx.ring->vsi;
3681 i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3687 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3688 * @vsi: the VSI being configured
3689 * @v_idx: vector index
3690 * @qp_idx: queue pair index
3692 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3694 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3695 struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3696 struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3698 tx_ring->q_vector = q_vector;
3699 tx_ring->next = q_vector->tx.ring;
3700 q_vector->tx.ring = tx_ring;
3701 q_vector->tx.count++;
3703 rx_ring->q_vector = q_vector;
3704 rx_ring->next = q_vector->rx.ring;
3705 q_vector->rx.ring = rx_ring;
3706 q_vector->rx.count++;
3710 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3711 * @vsi: the VSI being configured
3713 * This function maps descriptor rings to the queue-specific vectors
3714 * we were allotted through the MSI-X enabling code. Ideally, we'd have
3715 * one vector per queue pair, but on a constrained vector budget, we
3716 * group the queue pairs as "efficiently" as possible.
3718 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3720 int qp_remaining = vsi->num_queue_pairs;
3721 int q_vectors = vsi->num_q_vectors;
3726 /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3727 * group them so there are multiple queues per vector.
3728 * It is also important to go through all the vectors available to be
3729 * sure that if we don't use all the vectors, that the remaining vectors
3730 * are cleared. This is especially important when decreasing the
3731 * number of queues in use.
3733 for (; v_start < q_vectors; v_start++) {
3734 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3736 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3738 q_vector->num_ringpairs = num_ringpairs;
3740 q_vector->rx.count = 0;
3741 q_vector->tx.count = 0;
3742 q_vector->rx.ring = NULL;
3743 q_vector->tx.ring = NULL;
3745 while (num_ringpairs--) {
3746 i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3754 * i40e_vsi_request_irq - Request IRQ from the OS
3755 * @vsi: the VSI being configured
3756 * @basename: name for the vector
3758 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3760 struct i40e_pf *pf = vsi->back;
3763 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3764 err = i40e_vsi_request_irq_msix(vsi, basename);
3765 else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3766 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3769 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3773 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3778 #ifdef CONFIG_NET_POLL_CONTROLLER
3780 * i40e_netpoll - A Polling 'interrupt' handler
3781 * @netdev: network interface device structure
3783 * This is used by netconsole to send skbs without having to re-enable
3784 * interrupts. It's not called while the normal interrupt routine is executing.
3787 void i40e_netpoll(struct net_device *netdev)
3789 static void i40e_netpoll(struct net_device *netdev)
3792 struct i40e_netdev_priv *np = netdev_priv(netdev);
3793 struct i40e_vsi *vsi = np->vsi;
3794 struct i40e_pf *pf = vsi->back;
3797 /* if interface is down do nothing */
3798 if (test_bit(__I40E_DOWN, &vsi->state))
3801 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3802 for (i = 0; i < vsi->num_q_vectors; i++)
3803 i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3805 i40e_intr(pf->pdev->irq, netdev);
3811 * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3812 * @pf: the PF being configured
3813 * @pf_q: the PF queue
3814 * @enable: enable or disable state of the queue
3816 * This routine will wait for the given Tx queue of the PF to reach the
3817 * enabled or disabled state.
3818 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3819 * multiple retries; else will return 0 in case of success.
3821 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3826 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3827 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3828 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3831 usleep_range(10, 20);
3833 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3840 * i40e_vsi_control_tx - Start or stop a VSI's rings
3841 * @vsi: the VSI being configured
3842 * @enable: start or stop the rings
3844 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3846 struct i40e_pf *pf = vsi->back;
3847 struct i40e_hw *hw = &pf->hw;
3848 int i, j, pf_q, ret = 0;
3851 pf_q = vsi->base_queue;
3852 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3854 /* warn the TX unit of coming changes */
3855 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3857 usleep_range(10, 20);
3859 for (j = 0; j < 50; j++) {
3860 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3861 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3862 ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3864 usleep_range(1000, 2000);
3866 /* Skip if the queue is already in the requested state */
3867 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3870 /* turn on/off the queue */
3872 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3873 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3875 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3878 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3879 /* No waiting for the Tx queue to disable */
3880 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3883 /* wait for the change to finish */
3884 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3886 dev_info(&pf->pdev->dev,
3887 "VSI seid %d Tx ring %d %sable timeout\n",
3888 vsi->seid, pf_q, (enable ? "en" : "dis"));
3893 if (hw->revision_id == 0)
3899 * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3900 * @pf: the PF being configured
3901 * @pf_q: the PF queue
3902 * @enable: enable or disable state of the queue
3904 * This routine will wait for the given Rx queue of the PF to reach the
3905 * enabled or disabled state.
3906 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3907 * multiple retries; else will return 0 in case of success.
3909 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3914 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3915 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3916 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3919 usleep_range(10, 20);
3921 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3928 * i40e_vsi_control_rx - Start or stop a VSI's rings
3929 * @vsi: the VSI being configured
3930 * @enable: start or stop the rings
3932 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3934 struct i40e_pf *pf = vsi->back;
3935 struct i40e_hw *hw = &pf->hw;
3936 int i, j, pf_q, ret = 0;
3939 pf_q = vsi->base_queue;
3940 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3941 for (j = 0; j < 50; j++) {
3942 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3943 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3944 ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3946 usleep_range(1000, 2000);
3949 /* Skip if the queue is already in the requested state */
3950 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3953 /* turn on/off the queue */
3955 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3957 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3958 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3959 /* No waiting for the Tx queue to disable */
3960 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3963 /* wait for the change to finish */
3964 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3966 dev_info(&pf->pdev->dev,
3967 "VSI seid %d Rx ring %d %sable timeout\n",
3968 vsi->seid, pf_q, (enable ? "en" : "dis"));
3977 * i40e_vsi_control_rings - Start or stop a VSI's rings
3978 * @vsi: the VSI being configured
3979 * @enable: start or stop the rings
3981 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3985 /* do rx first for enable and last for disable */
3987 ret = i40e_vsi_control_rx(vsi, request);
3990 ret = i40e_vsi_control_tx(vsi, request);
3992 /* Ignore return value, we need to shutdown whatever we can */
3993 i40e_vsi_control_tx(vsi, request);
3994 i40e_vsi_control_rx(vsi, request);
4001 * i40e_vsi_free_irq - Free the irq association with the OS
4002 * @vsi: the VSI being configured
4004 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4006 struct i40e_pf *pf = vsi->back;
4007 struct i40e_hw *hw = &pf->hw;
4008 int base = vsi->base_vector;
4012 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4013 if (!vsi->q_vectors)
4016 if (!vsi->irqs_ready)
4019 vsi->irqs_ready = false;
4020 for (i = 0; i < vsi->num_q_vectors; i++) {
4021 u16 vector = i + base;
4023 /* free only the irqs that were actually requested */
4024 if (!vsi->q_vectors[i] ||
4025 !vsi->q_vectors[i]->num_ringpairs)
4028 /* clear the affinity_mask in the IRQ descriptor */
4029 irq_set_affinity_hint(pf->msix_entries[vector].vector,
4031 synchronize_irq(pf->msix_entries[vector].vector);
4032 free_irq(pf->msix_entries[vector].vector,
4035 /* Tear down the interrupt queue link list
4037 * We know that they come in pairs and always
4038 * the Rx first, then the Tx. To clear the
4039 * link list, stick the EOL value into the
4040 * next_q field of the registers.
4042 val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4043 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4044 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4045 val |= I40E_QUEUE_END_OF_LIST
4046 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4047 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4049 while (qp != I40E_QUEUE_END_OF_LIST) {
4052 val = rd32(hw, I40E_QINT_RQCTL(qp));
4054 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4055 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4056 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4057 I40E_QINT_RQCTL_INTEVENT_MASK);
4059 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4060 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4062 wr32(hw, I40E_QINT_RQCTL(qp), val);
4064 val = rd32(hw, I40E_QINT_TQCTL(qp));
4066 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4067 >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4069 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4070 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4071 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4072 I40E_QINT_TQCTL_INTEVENT_MASK);
4074 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4075 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4077 wr32(hw, I40E_QINT_TQCTL(qp), val);
4082 free_irq(pf->pdev->irq, pf);
4084 val = rd32(hw, I40E_PFINT_LNKLST0);
4085 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4086 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4087 val |= I40E_QUEUE_END_OF_LIST
4088 << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4089 wr32(hw, I40E_PFINT_LNKLST0, val);
4091 val = rd32(hw, I40E_QINT_RQCTL(qp));
4092 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4093 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4094 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4095 I40E_QINT_RQCTL_INTEVENT_MASK);
4097 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4098 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4100 wr32(hw, I40E_QINT_RQCTL(qp), val);
4102 val = rd32(hw, I40E_QINT_TQCTL(qp));
4104 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4105 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4106 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4107 I40E_QINT_TQCTL_INTEVENT_MASK);
4109 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4110 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4112 wr32(hw, I40E_QINT_TQCTL(qp), val);
4117 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4118 * @vsi: the VSI being configured
4119 * @v_idx: Index of vector to be freed
4121 * This function frees the memory allocated to the q_vector. In addition if
4122 * NAPI is enabled it will delete any references to the NAPI struct prior
4123 * to freeing the q_vector.
4125 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4127 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4128 struct i40e_ring *ring;
4133 /* disassociate q_vector from rings */
4134 i40e_for_each_ring(ring, q_vector->tx)
4135 ring->q_vector = NULL;
4137 i40e_for_each_ring(ring, q_vector->rx)
4138 ring->q_vector = NULL;
4140 /* only VSI w/ an associated netdev is set up w/ NAPI */
4142 netif_napi_del(&q_vector->napi);
4144 vsi->q_vectors[v_idx] = NULL;
4146 kfree_rcu(q_vector, rcu);
4150 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4151 * @vsi: the VSI being un-configured
4153 * This frees the memory allocated to the q_vectors and
4154 * deletes references to the NAPI struct.
4156 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4160 for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4161 i40e_free_q_vector(vsi, v_idx);
4165 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4166 * @pf: board private structure
4168 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4170 /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4171 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4172 pci_disable_msix(pf->pdev);
4173 kfree(pf->msix_entries);
4174 pf->msix_entries = NULL;
4175 kfree(pf->irq_pile);
4176 pf->irq_pile = NULL;
4177 } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4178 pci_disable_msi(pf->pdev);
4180 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4184 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4185 * @pf: board private structure
4187 * We go through and clear interrupt specific resources and reset the structure
4188 * to pre-load conditions
4190 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4194 i40e_stop_misc_vector(pf);
4195 if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4196 synchronize_irq(pf->msix_entries[0].vector);
4197 free_irq(pf->msix_entries[0].vector, pf);
4200 i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4201 I40E_IWARP_IRQ_PILE_ID);
4203 i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4204 for (i = 0; i < pf->num_alloc_vsi; i++)
4206 i40e_vsi_free_q_vectors(pf->vsi[i]);
4207 i40e_reset_interrupt_capability(pf);
4211 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4212 * @vsi: the VSI being configured
4214 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4221 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4222 napi_enable(&vsi->q_vectors[q_idx]->napi);
4226 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4227 * @vsi: the VSI being configured
4229 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4236 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4237 napi_disable(&vsi->q_vectors[q_idx]->napi);
4241 * i40e_vsi_close - Shut down a VSI
4242 * @vsi: the vsi to be quelled
4244 static void i40e_vsi_close(struct i40e_vsi *vsi)
4248 if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4250 i40e_vsi_free_irq(vsi);
4251 i40e_vsi_free_tx_resources(vsi);
4252 i40e_vsi_free_rx_resources(vsi);
4253 vsi->current_netdev_flags = 0;
4254 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4256 i40e_notify_client_of_netdev_close(vsi, reset);
4260 * i40e_quiesce_vsi - Pause a given VSI
4261 * @vsi: the VSI being paused
4263 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4265 if (test_bit(__I40E_DOWN, &vsi->state))
4268 /* No need to disable FCoE VSI when Tx suspended */
4269 if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4270 vsi->type == I40E_VSI_FCOE) {
4271 dev_dbg(&vsi->back->pdev->dev,
4272 "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4276 set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4277 if (vsi->netdev && netif_running(vsi->netdev))
4278 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4280 i40e_vsi_close(vsi);
4284 * i40e_unquiesce_vsi - Resume a given VSI
4285 * @vsi: the VSI being resumed
4287 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4289 if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4292 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4293 if (vsi->netdev && netif_running(vsi->netdev))
4294 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4296 i40e_vsi_open(vsi); /* this clears the DOWN bit */
4300 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4303 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4307 for (v = 0; v < pf->num_alloc_vsi; v++) {
4309 i40e_quiesce_vsi(pf->vsi[v]);
4314 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4317 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4321 for (v = 0; v < pf->num_alloc_vsi; v++) {
4323 i40e_unquiesce_vsi(pf->vsi[v]);
4327 #ifdef CONFIG_I40E_DCB
4329 * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4330 * @vsi: the VSI being configured
4332 * This function waits for the given VSI's queues to be disabled.
4334 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4336 struct i40e_pf *pf = vsi->back;
4339 pf_q = vsi->base_queue;
4340 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4341 /* Check and wait for the disable status of the queue */
4342 ret = i40e_pf_txq_wait(pf, pf_q, false);
4344 dev_info(&pf->pdev->dev,
4345 "VSI seid %d Tx ring %d disable timeout\n",
4351 pf_q = vsi->base_queue;
4352 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4353 /* Check and wait for the disable status of the queue */
4354 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4356 dev_info(&pf->pdev->dev,
4357 "VSI seid %d Rx ring %d disable timeout\n",
4367 * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4370 * This function waits for the queues to be in disabled state for all the
4371 * VSIs that are managed by this PF.
4373 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4377 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4378 /* No need to wait for FCoE VSI queues */
4379 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4380 ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4392 * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4393 * @q_idx: TX queue number
4394 * @vsi: Pointer to VSI struct
4396 * This function checks specified queue for given VSI. Detects hung condition.
4397 * Sets hung bit since it is two step process. Before next run of service task
4398 * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4399 * hung condition remain unchanged and during subsequent run, this function
4400 * issues SW interrupt to recover from hung condition.
4402 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4404 struct i40e_ring *tx_ring = NULL;
4406 u32 head, val, tx_pending_hw;
4411 /* now that we have an index, find the tx_ring struct */
4412 for (i = 0; i < vsi->num_queue_pairs; i++) {
4413 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4414 if (q_idx == vsi->tx_rings[i]->queue_index) {
4415 tx_ring = vsi->tx_rings[i];
4424 /* Read interrupt register */
4425 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4427 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4428 tx_ring->vsi->base_vector - 1));
4430 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4432 head = i40e_get_head(tx_ring);
4434 tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4436 /* HW is done executing descriptors, updated HEAD write back,
4437 * but SW hasn't processed those descriptors. If interrupt is
4438 * not generated from this point ON, it could result into
4439 * dev_watchdog detecting timeout on those netdev_queue,
4440 * hence proactively trigger SW interrupt.
4442 if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4443 /* NAPI Poll didn't run and clear since it was set */
4444 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4445 &tx_ring->q_vector->hung_detected)) {
4446 netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4447 vsi->seid, q_idx, tx_pending_hw,
4448 tx_ring->next_to_clean, head,
4449 tx_ring->next_to_use,
4450 readl(tx_ring->tail));
4451 netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4452 vsi->seid, q_idx, val);
4453 i40e_force_wb(vsi, tx_ring->q_vector);
4455 /* First Chance - detected possible hung */
4456 set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4457 &tx_ring->q_vector->hung_detected);
4461 /* This is the case where we have interrupts missing,
4462 * so the tx_pending in HW will most likely be 0, but we
4463 * will have tx_pending in SW since the WB happened but the
4464 * interrupt got lost.
4466 if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4467 (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4468 if (napi_reschedule(&tx_ring->q_vector->napi))
4469 tx_ring->tx_stats.tx_lost_interrupt++;
4474 * i40e_detect_recover_hung - Function to detect and recover hung_queues
4475 * @pf: pointer to PF struct
4477 * LAN VSI has netdev and netdev has TX queues. This function is to check
4478 * each of those TX queues if they are hung, trigger recovery by issuing
4481 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4483 struct net_device *netdev;
4484 struct i40e_vsi *vsi;
4487 /* Only for LAN VSI */
4488 vsi = pf->vsi[pf->lan_vsi];
4493 /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4494 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4495 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4498 /* Make sure type is MAIN VSI */
4499 if (vsi->type != I40E_VSI_MAIN)
4502 netdev = vsi->netdev;
4506 /* Bail out if netif_carrier is not OK */
4507 if (!netif_carrier_ok(netdev))
4510 /* Go thru' TX queues for netdev */
4511 for (i = 0; i < netdev->num_tx_queues; i++) {
4512 struct netdev_queue *q;
4514 q = netdev_get_tx_queue(netdev, i);
4516 i40e_detect_recover_hung_queue(i, vsi);
4521 * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4522 * @pf: pointer to PF
4524 * Get TC map for ISCSI PF type that will include iSCSI TC
4527 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4529 struct i40e_dcb_app_priority_table app;
4530 struct i40e_hw *hw = &pf->hw;
4531 u8 enabled_tc = 1; /* TC0 is always enabled */
4533 /* Get the iSCSI APP TLV */
4534 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4536 for (i = 0; i < dcbcfg->numapps; i++) {
4537 app = dcbcfg->app[i];
4538 if (app.selector == I40E_APP_SEL_TCPIP &&
4539 app.protocolid == I40E_APP_PROTOID_ISCSI) {
4540 tc = dcbcfg->etscfg.prioritytable[app.priority];
4541 enabled_tc |= BIT(tc);
4550 * i40e_dcb_get_num_tc - Get the number of TCs from DCBx config
4551 * @dcbcfg: the corresponding DCBx configuration structure
4553 * Return the number of TCs from given DCBx configuration
4555 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4557 int i, tc_unused = 0;
4561 /* Scan the ETS Config Priority Table to find
4562 * traffic class enabled for a given priority
4563 * and create a bitmask of enabled TCs
4565 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4566 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4568 /* Now scan the bitmask to check for
4569 * contiguous TCs starting with TC0
4571 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4572 if (num_tc & BIT(i)) {
4576 pr_err("Non-contiguous TC - Disabling DCB\n");
4584 /* There is always at least TC0 */
4592 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4593 * @dcbcfg: the corresponding DCBx configuration structure
4595 * Query the current DCB configuration and return the number of
4596 * traffic classes enabled from the given DCBX config
4598 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4600 u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4604 for (i = 0; i < num_tc; i++)
4605 enabled_tc |= BIT(i);
4611 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4612 * @pf: PF being queried
4614 * Return number of traffic classes enabled for the given PF
4616 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4618 struct i40e_hw *hw = &pf->hw;
4621 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4623 /* If DCB is not enabled then always in single TC */
4624 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4627 /* SFP mode will be enabled for all TCs on port */
4628 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4629 return i40e_dcb_get_num_tc(dcbcfg);
4631 /* MFP mode return count of enabled TCs for this PF */
4632 if (pf->hw.func_caps.iscsi)
4633 enabled_tc = i40e_get_iscsi_tc_map(pf);
4635 return 1; /* Only TC0 */
4637 /* At least have TC0 */
4638 enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4639 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4640 if (enabled_tc & BIT(i))
4647 * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4648 * @pf: PF being queried
4650 * Return a bitmap for first enabled traffic class for this PF.
4652 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4654 u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4658 return 0x1; /* TC0 */
4660 /* Find the first enabled TC */
4661 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4662 if (enabled_tc & BIT(i))
4670 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4671 * @pf: PF being queried
4673 * Return a bitmap for enabled traffic classes for this PF.
4675 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4677 /* If DCB is not enabled for this PF then just return default TC */
4678 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4679 return i40e_pf_get_default_tc(pf);
4681 /* SFP mode we want PF to be enabled for all TCs */
4682 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4683 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4685 /* MFP enabled and iSCSI PF type */
4686 if (pf->hw.func_caps.iscsi)
4687 return i40e_get_iscsi_tc_map(pf);
4689 return i40e_pf_get_default_tc(pf);
4693 * i40e_vsi_get_bw_info - Query VSI BW Information
4694 * @vsi: the VSI being queried
4696 * Returns 0 on success, negative value on failure
4698 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4700 struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4701 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4702 struct i40e_pf *pf = vsi->back;
4703 struct i40e_hw *hw = &pf->hw;
4708 /* Get the VSI level BW configuration */
4709 ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4711 dev_info(&pf->pdev->dev,
4712 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4713 i40e_stat_str(&pf->hw, ret),
4714 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4718 /* Get the VSI level BW configuration per TC */
4719 ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4722 dev_info(&pf->pdev->dev,
4723 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4724 i40e_stat_str(&pf->hw, ret),
4725 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4729 if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4730 dev_info(&pf->pdev->dev,
4731 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4732 bw_config.tc_valid_bits,
4733 bw_ets_config.tc_valid_bits);
4734 /* Still continuing */
4737 vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4738 vsi->bw_max_quanta = bw_config.max_bw;
4739 tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4740 (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4741 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4742 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4743 vsi->bw_ets_limit_credits[i] =
4744 le16_to_cpu(bw_ets_config.credits[i]);
4745 /* 3 bits out of 4 for each TC */
4746 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4753 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4754 * @vsi: the VSI being configured
4755 * @enabled_tc: TC bitmap
4756 * @bw_credits: BW shared credits per TC
4758 * Returns 0 on success, negative value on failure
4760 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4763 struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4767 bw_data.tc_valid_bits = enabled_tc;
4768 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4769 bw_data.tc_bw_credits[i] = bw_share[i];
4771 ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4774 dev_info(&vsi->back->pdev->dev,
4775 "AQ command Config VSI BW allocation per TC failed = %d\n",
4776 vsi->back->hw.aq.asq_last_status);
4780 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4781 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4787 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4788 * @vsi: the VSI being configured
4789 * @enabled_tc: TC map to be enabled
4792 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4794 struct net_device *netdev = vsi->netdev;
4795 struct i40e_pf *pf = vsi->back;
4796 struct i40e_hw *hw = &pf->hw;
4799 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4805 netdev_reset_tc(netdev);
4809 /* Set up actual enabled TCs on the VSI */
4810 if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4813 /* set per TC queues for the VSI */
4814 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4815 /* Only set TC queues for enabled tcs
4817 * e.g. For a VSI that has TC0 and TC3 enabled the
4818 * enabled_tc bitmap would be 0x00001001; the driver
4819 * will set the numtc for netdev as 2 that will be
4820 * referenced by the netdev layer as TC 0 and 1.
4822 if (vsi->tc_config.enabled_tc & BIT(i))
4823 netdev_set_tc_queue(netdev,
4824 vsi->tc_config.tc_info[i].netdev_tc,
4825 vsi->tc_config.tc_info[i].qcount,
4826 vsi->tc_config.tc_info[i].qoffset);
4829 /* Assign UP2TC map for the VSI */
4830 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4831 /* Get the actual TC# for the UP */
4832 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4833 /* Get the mapped netdev TC# for the UP */
4834 netdev_tc = vsi->tc_config.tc_info[ets_tc].netdev_tc;
4835 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4840 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4841 * @vsi: the VSI being configured
4842 * @ctxt: the ctxt buffer returned from AQ VSI update param command
4844 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4845 struct i40e_vsi_context *ctxt)
4847 /* copy just the sections touched not the entire info
4848 * since not all sections are valid as returned by
4851 vsi->info.mapping_flags = ctxt->info.mapping_flags;
4852 memcpy(&vsi->info.queue_mapping,
4853 &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4854 memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4855 sizeof(vsi->info.tc_mapping));
4859 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4860 * @vsi: VSI to be configured
4861 * @enabled_tc: TC bitmap
4863 * This configures a particular VSI for TCs that are mapped to the
4864 * given TC bitmap. It uses default bandwidth share for TCs across
4865 * VSIs to configure TC for a particular VSI.
4868 * It is expected that the VSI queues have been quisced before calling
4871 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4873 u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4874 struct i40e_vsi_context ctxt;
4878 /* Check if enabled_tc is same as existing or new TCs */
4879 if (vsi->tc_config.enabled_tc == enabled_tc)
4882 /* Enable ETS TCs with equal BW Share for now across all VSIs */
4883 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4884 if (enabled_tc & BIT(i))
4888 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4890 dev_info(&vsi->back->pdev->dev,
4891 "Failed configuring TC map %d for VSI %d\n",
4892 enabled_tc, vsi->seid);
4896 /* Update Queue Pairs Mapping for currently enabled UPs */
4897 ctxt.seid = vsi->seid;
4898 ctxt.pf_num = vsi->back->hw.pf_id;
4900 ctxt.uplink_seid = vsi->uplink_seid;
4901 ctxt.info = vsi->info;
4902 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4904 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4905 ctxt.info.valid_sections |=
4906 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4907 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4910 /* Update the VSI after updating the VSI queue-mapping information */
4911 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4913 dev_info(&vsi->back->pdev->dev,
4914 "Update vsi tc config failed, err %s aq_err %s\n",
4915 i40e_stat_str(&vsi->back->hw, ret),
4916 i40e_aq_str(&vsi->back->hw,
4917 vsi->back->hw.aq.asq_last_status));
4920 /* update the local VSI info with updated queue map */
4921 i40e_vsi_update_queue_map(vsi, &ctxt);
4922 vsi->info.valid_sections = 0;
4924 /* Update current VSI BW information */
4925 ret = i40e_vsi_get_bw_info(vsi);
4927 dev_info(&vsi->back->pdev->dev,
4928 "Failed updating vsi bw info, err %s aq_err %s\n",
4929 i40e_stat_str(&vsi->back->hw, ret),
4930 i40e_aq_str(&vsi->back->hw,
4931 vsi->back->hw.aq.asq_last_status));
4935 /* Update the netdev TC setup */
4936 i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4942 * i40e_veb_config_tc - Configure TCs for given VEB
4944 * @enabled_tc: TC bitmap
4946 * Configures given TC bitmap for VEB (switching) element
4948 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4950 struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4951 struct i40e_pf *pf = veb->pf;
4955 /* No TCs or already enabled TCs just return */
4956 if (!enabled_tc || veb->enabled_tc == enabled_tc)
4959 bw_data.tc_valid_bits = enabled_tc;
4960 /* bw_data.absolute_credits is not set (relative) */
4962 /* Enable ETS TCs with equal BW Share for now */
4963 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4964 if (enabled_tc & BIT(i))
4965 bw_data.tc_bw_share_credits[i] = 1;
4968 ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4971 dev_info(&pf->pdev->dev,
4972 "VEB bw config failed, err %s aq_err %s\n",
4973 i40e_stat_str(&pf->hw, ret),
4974 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4978 /* Update the BW information */
4979 ret = i40e_veb_get_bw_info(veb);
4981 dev_info(&pf->pdev->dev,
4982 "Failed getting veb bw config, err %s aq_err %s\n",
4983 i40e_stat_str(&pf->hw, ret),
4984 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4991 #ifdef CONFIG_I40E_DCB
4993 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4996 * Reconfigure VEB/VSIs on a given PF; it is assumed that
4997 * the caller would've quiesce all the VSIs before calling
5000 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5006 /* Enable the TCs available on PF to all VEBs */
5007 tc_map = i40e_pf_get_tc_map(pf);
5008 for (v = 0; v < I40E_MAX_VEB; v++) {
5011 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5013 dev_info(&pf->pdev->dev,
5014 "Failed configuring TC for VEB seid=%d\n",
5016 /* Will try to configure as many components */
5020 /* Update each VSI */
5021 for (v = 0; v < pf->num_alloc_vsi; v++) {
5025 /* - Enable all TCs for the LAN VSI
5027 * - For FCoE VSI only enable the TC configured
5028 * as per the APP TLV
5030 * - For all others keep them at TC0 for now
5032 if (v == pf->lan_vsi)
5033 tc_map = i40e_pf_get_tc_map(pf);
5035 tc_map = i40e_pf_get_default_tc(pf);
5037 if (pf->vsi[v]->type == I40E_VSI_FCOE)
5038 tc_map = i40e_get_fcoe_tc_map(pf);
5039 #endif /* #ifdef I40E_FCOE */
5041 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5043 dev_info(&pf->pdev->dev,
5044 "Failed configuring TC for VSI seid=%d\n",
5046 /* Will try to configure as many components */
5048 /* Re-configure VSI vectors based on updated TC map */
5049 i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5050 if (pf->vsi[v]->netdev)
5051 i40e_dcbnl_set_all(pf->vsi[v]);
5057 * i40e_resume_port_tx - Resume port Tx
5060 * Resume a port's Tx and issue a PF reset in case of failure to
5063 static int i40e_resume_port_tx(struct i40e_pf *pf)
5065 struct i40e_hw *hw = &pf->hw;
5068 ret = i40e_aq_resume_port_tx(hw, NULL);
5070 dev_info(&pf->pdev->dev,
5071 "Resume Port Tx failed, err %s aq_err %s\n",
5072 i40e_stat_str(&pf->hw, ret),
5073 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5074 /* Schedule PF reset to recover */
5075 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5076 i40e_service_event_schedule(pf);
5083 * i40e_init_pf_dcb - Initialize DCB configuration
5084 * @pf: PF being configured
5086 * Query the current DCB configuration and cache it
5087 * in the hardware structure
5089 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5091 struct i40e_hw *hw = &pf->hw;
5094 /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5095 if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5098 /* Get the initial DCB configuration */
5099 err = i40e_init_dcb(hw);
5101 /* Device/Function is not DCBX capable */
5102 if ((!hw->func_caps.dcb) ||
5103 (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5104 dev_info(&pf->pdev->dev,
5105 "DCBX offload is not supported or is disabled for this PF.\n");
5107 if (pf->flags & I40E_FLAG_MFP_ENABLED)
5111 /* When status is not DISABLED then DCBX in FW */
5112 pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5113 DCB_CAP_DCBX_VER_IEEE;
5115 pf->flags |= I40E_FLAG_DCB_CAPABLE;
5116 /* Enable DCB tagging only when more than one TC
5117 * or explicitly disable if only one TC
5119 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5120 pf->flags |= I40E_FLAG_DCB_ENABLED;
5122 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5123 dev_dbg(&pf->pdev->dev,
5124 "DCBX offload is supported for this PF.\n");
5127 dev_info(&pf->pdev->dev,
5128 "Query for DCB configuration failed, err %s aq_err %s\n",
5129 i40e_stat_str(&pf->hw, err),
5130 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5136 #endif /* CONFIG_I40E_DCB */
5137 #define SPEED_SIZE 14
5140 * i40e_print_link_message - print link up or down
5141 * @vsi: the VSI for which link needs a message
5143 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5145 char *speed = "Unknown";
5146 char *fc = "Unknown";
5148 if (vsi->current_isup == isup)
5150 vsi->current_isup = isup;
5152 netdev_info(vsi->netdev, "NIC Link is Down\n");
5156 /* Warn user if link speed on NPAR enabled partition is not at
5159 if (vsi->back->hw.func_caps.npar_enable &&
5160 (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5161 vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5162 netdev_warn(vsi->netdev,
5163 "The partition detected link speed that is less than 10Gbps\n");
5165 switch (vsi->back->hw.phy.link_info.link_speed) {
5166 case I40E_LINK_SPEED_40GB:
5169 case I40E_LINK_SPEED_20GB:
5172 case I40E_LINK_SPEED_10GB:
5175 case I40E_LINK_SPEED_1GB:
5178 case I40E_LINK_SPEED_100MB:
5185 switch (vsi->back->hw.fc.current_mode) {
5189 case I40E_FC_TX_PAUSE:
5192 case I40E_FC_RX_PAUSE:
5200 netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5205 * i40e_up_complete - Finish the last steps of bringing up a connection
5206 * @vsi: the VSI being configured
5208 static int i40e_up_complete(struct i40e_vsi *vsi)
5210 struct i40e_pf *pf = vsi->back;
5213 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5214 i40e_vsi_configure_msix(vsi);
5216 i40e_configure_msi_and_legacy(vsi);
5219 err = i40e_vsi_control_rings(vsi, true);
5223 clear_bit(__I40E_DOWN, &vsi->state);
5224 i40e_napi_enable_all(vsi);
5225 i40e_vsi_enable_irq(vsi);
5227 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5229 i40e_print_link_message(vsi, true);
5230 netif_tx_start_all_queues(vsi->netdev);
5231 netif_carrier_on(vsi->netdev);
5232 } else if (vsi->netdev) {
5233 i40e_print_link_message(vsi, false);
5234 /* need to check for qualified module here*/
5235 if ((pf->hw.phy.link_info.link_info &
5236 I40E_AQ_MEDIA_AVAILABLE) &&
5237 (!(pf->hw.phy.link_info.an_info &
5238 I40E_AQ_QUALIFIED_MODULE)))
5239 netdev_err(vsi->netdev,
5240 "the driver failed to link because an unqualified module was detected.");
5243 /* replay FDIR SB filters */
5244 if (vsi->type == I40E_VSI_FDIR) {
5245 /* reset fd counters */
5246 pf->fd_add_err = pf->fd_atr_cnt = 0;
5247 if (pf->fd_tcp_rule > 0) {
5248 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5249 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5250 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5251 pf->fd_tcp_rule = 0;
5253 i40e_fdir_filter_restore(vsi);
5256 /* On the next run of the service_task, notify any clients of the new
5259 pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5260 i40e_service_event_schedule(pf);
5266 * i40e_vsi_reinit_locked - Reset the VSI
5267 * @vsi: the VSI being configured
5269 * Rebuild the ring structs after some configuration
5270 * has changed, e.g. MTU size.
5272 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5274 struct i40e_pf *pf = vsi->back;
5276 WARN_ON(in_interrupt());
5277 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5278 usleep_range(1000, 2000);
5282 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5286 * i40e_up - Bring the connection back up after being down
5287 * @vsi: the VSI being configured
5289 int i40e_up(struct i40e_vsi *vsi)
5293 err = i40e_vsi_configure(vsi);
5295 err = i40e_up_complete(vsi);
5301 * i40e_down - Shutdown the connection processing
5302 * @vsi: the VSI being stopped
5304 void i40e_down(struct i40e_vsi *vsi)
5308 /* It is assumed that the caller of this function
5309 * sets the vsi->state __I40E_DOWN bit.
5312 netif_carrier_off(vsi->netdev);
5313 netif_tx_disable(vsi->netdev);
5315 i40e_vsi_disable_irq(vsi);
5316 i40e_vsi_control_rings(vsi, false);
5317 i40e_napi_disable_all(vsi);
5319 for (i = 0; i < vsi->num_queue_pairs; i++) {
5320 i40e_clean_tx_ring(vsi->tx_rings[i]);
5321 i40e_clean_rx_ring(vsi->rx_rings[i]);
5324 i40e_notify_client_of_netdev_close(vsi, false);
5329 * i40e_setup_tc - configure multiple traffic classes
5330 * @netdev: net device to configure
5331 * @tc: number of traffic classes to enable
5333 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5335 struct i40e_netdev_priv *np = netdev_priv(netdev);
5336 struct i40e_vsi *vsi = np->vsi;
5337 struct i40e_pf *pf = vsi->back;
5342 /* Check if DCB enabled to continue */
5343 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5344 netdev_info(netdev, "DCB is not enabled for adapter\n");
5348 /* Check if MFP enabled */
5349 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5350 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5354 /* Check whether tc count is within enabled limit */
5355 if (tc > i40e_pf_get_num_tc(pf)) {
5356 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5360 /* Generate TC map for number of tc requested */
5361 for (i = 0; i < tc; i++)
5362 enabled_tc |= BIT(i);
5364 /* Requesting same TC configuration as already enabled */
5365 if (enabled_tc == vsi->tc_config.enabled_tc)
5368 /* Quiesce VSI queues */
5369 i40e_quiesce_vsi(vsi);
5371 /* Configure VSI for enabled TCs */
5372 ret = i40e_vsi_config_tc(vsi, enabled_tc);
5374 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5380 i40e_unquiesce_vsi(vsi);
5387 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5388 struct tc_to_netdev *tc)
5390 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5391 struct tc_to_netdev *tc)
5394 if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5396 return i40e_setup_tc(netdev, tc->tc);
5400 * i40e_open - Called when a network interface is made active
5401 * @netdev: network interface device structure
5403 * The open entry point is called when a network interface is made
5404 * active by the system (IFF_UP). At this point all resources needed
5405 * for transmit and receive operations are allocated, the interrupt
5406 * handler is registered with the OS, the netdev watchdog subtask is
5407 * enabled, and the stack is notified that the interface is ready.
5409 * Returns 0 on success, negative value on failure
5411 int i40e_open(struct net_device *netdev)
5413 struct i40e_netdev_priv *np = netdev_priv(netdev);
5414 struct i40e_vsi *vsi = np->vsi;
5415 struct i40e_pf *pf = vsi->back;
5418 /* disallow open during test or if eeprom is broken */
5419 if (test_bit(__I40E_TESTING, &pf->state) ||
5420 test_bit(__I40E_BAD_EEPROM, &pf->state))
5423 netif_carrier_off(netdev);
5425 err = i40e_vsi_open(vsi);
5429 /* configure global TSO hardware offload settings */
5430 wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5431 TCP_FLAG_FIN) >> 16);
5432 wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5434 TCP_FLAG_CWR) >> 16);
5435 wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5437 udp_tunnel_get_rx_info(netdev);
5444 * @vsi: the VSI to open
5446 * Finish initialization of the VSI.
5448 * Returns 0 on success, negative value on failure
5450 int i40e_vsi_open(struct i40e_vsi *vsi)
5452 struct i40e_pf *pf = vsi->back;
5453 char int_name[I40E_INT_NAME_STR_LEN];
5456 /* allocate descriptors */
5457 err = i40e_vsi_setup_tx_resources(vsi);
5460 err = i40e_vsi_setup_rx_resources(vsi);
5464 err = i40e_vsi_configure(vsi);
5469 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5470 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5471 err = i40e_vsi_request_irq(vsi, int_name);
5475 /* Notify the stack of the actual queue counts. */
5476 err = netif_set_real_num_tx_queues(vsi->netdev,
5477 vsi->num_queue_pairs);
5479 goto err_set_queues;
5481 err = netif_set_real_num_rx_queues(vsi->netdev,
5482 vsi->num_queue_pairs);
5484 goto err_set_queues;
5486 } else if (vsi->type == I40E_VSI_FDIR) {
5487 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5488 dev_driver_string(&pf->pdev->dev),
5489 dev_name(&pf->pdev->dev));
5490 err = i40e_vsi_request_irq(vsi, int_name);
5497 err = i40e_up_complete(vsi);
5499 goto err_up_complete;
5506 i40e_vsi_free_irq(vsi);
5508 i40e_vsi_free_rx_resources(vsi);
5510 i40e_vsi_free_tx_resources(vsi);
5511 if (vsi == pf->vsi[pf->lan_vsi])
5512 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5518 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5519 * @pf: Pointer to PF
5521 * This function destroys the hlist where all the Flow Director
5522 * filters were saved.
5524 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5526 struct i40e_fdir_filter *filter;
5527 struct hlist_node *node2;
5529 hlist_for_each_entry_safe(filter, node2,
5530 &pf->fdir_filter_list, fdir_node) {
5531 hlist_del(&filter->fdir_node);
5534 pf->fdir_pf_active_filters = 0;
5538 * i40e_close - Disables a network interface
5539 * @netdev: network interface device structure
5541 * The close entry point is called when an interface is de-activated
5542 * by the OS. The hardware is still under the driver's control, but
5543 * this netdev interface is disabled.
5545 * Returns 0, this is not allowed to fail
5547 int i40e_close(struct net_device *netdev)
5549 struct i40e_netdev_priv *np = netdev_priv(netdev);
5550 struct i40e_vsi *vsi = np->vsi;
5552 i40e_vsi_close(vsi);
5558 * i40e_do_reset - Start a PF or Core Reset sequence
5559 * @pf: board private structure
5560 * @reset_flags: which reset is requested
5562 * The essential difference in resets is that the PF Reset
5563 * doesn't clear the packet buffers, doesn't reset the PE
5564 * firmware, and doesn't bother the other PFs on the chip.
5566 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5570 WARN_ON(in_interrupt());
5573 /* do the biggest reset indicated */
5574 if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5576 /* Request a Global Reset
5578 * This will start the chip's countdown to the actual full
5579 * chip reset event, and a warning interrupt to be sent
5580 * to all PFs, including the requestor. Our handler
5581 * for the warning interrupt will deal with the shutdown
5582 * and recovery of the switch setup.
5584 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5585 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5586 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5587 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5589 } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5591 /* Request a Core Reset
5593 * Same as Global Reset, except does *not* include the MAC/PHY
5595 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5596 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5597 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5598 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5599 i40e_flush(&pf->hw);
5601 } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5603 /* Request a PF Reset
5605 * Resets only the PF-specific registers
5607 * This goes directly to the tear-down and rebuild of
5608 * the switch, since we need to do all the recovery as
5609 * for the Core Reset.
5611 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5612 i40e_handle_reset_warning(pf);
5614 } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5617 /* Find the VSI(s) that requested a re-init */
5618 dev_info(&pf->pdev->dev,
5619 "VSI reinit requested\n");
5620 for (v = 0; v < pf->num_alloc_vsi; v++) {
5621 struct i40e_vsi *vsi = pf->vsi[v];
5624 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5625 i40e_vsi_reinit_locked(pf->vsi[v]);
5626 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5629 } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5632 /* Find the VSI(s) that needs to be brought down */
5633 dev_info(&pf->pdev->dev, "VSI down requested\n");
5634 for (v = 0; v < pf->num_alloc_vsi; v++) {
5635 struct i40e_vsi *vsi = pf->vsi[v];
5638 test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5639 set_bit(__I40E_DOWN, &vsi->state);
5641 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5645 dev_info(&pf->pdev->dev,
5646 "bad reset request 0x%08x\n", reset_flags);
5650 #ifdef CONFIG_I40E_DCB
5652 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5653 * @pf: board private structure
5654 * @old_cfg: current DCB config
5655 * @new_cfg: new DCB config
5657 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5658 struct i40e_dcbx_config *old_cfg,
5659 struct i40e_dcbx_config *new_cfg)
5661 bool need_reconfig = false;
5663 /* Check if ETS configuration has changed */
5664 if (memcmp(&new_cfg->etscfg,
5666 sizeof(new_cfg->etscfg))) {
5667 /* If Priority Table has changed reconfig is needed */
5668 if (memcmp(&new_cfg->etscfg.prioritytable,
5669 &old_cfg->etscfg.prioritytable,
5670 sizeof(new_cfg->etscfg.prioritytable))) {
5671 need_reconfig = true;
5672 dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5675 if (memcmp(&new_cfg->etscfg.tcbwtable,
5676 &old_cfg->etscfg.tcbwtable,
5677 sizeof(new_cfg->etscfg.tcbwtable)))
5678 dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5680 if (memcmp(&new_cfg->etscfg.tsatable,
5681 &old_cfg->etscfg.tsatable,
5682 sizeof(new_cfg->etscfg.tsatable)))
5683 dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5686 /* Check if PFC configuration has changed */
5687 if (memcmp(&new_cfg->pfc,
5689 sizeof(new_cfg->pfc))) {
5690 need_reconfig = true;
5691 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5694 /* Check if APP Table has changed */
5695 if (memcmp(&new_cfg->app,
5697 sizeof(new_cfg->app))) {
5698 need_reconfig = true;
5699 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5702 dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5703 return need_reconfig;
5707 * i40e_handle_lldp_event - Handle LLDP Change MIB event
5708 * @pf: board private structure
5709 * @e: event info posted on ARQ
5711 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5712 struct i40e_arq_event_info *e)
5714 struct i40e_aqc_lldp_get_mib *mib =
5715 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5716 struct i40e_hw *hw = &pf->hw;
5717 struct i40e_dcbx_config tmp_dcbx_cfg;
5718 bool need_reconfig = false;
5722 /* Not DCB capable or capability disabled */
5723 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5726 /* Ignore if event is not for Nearest Bridge */
5727 type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5728 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5729 dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5730 if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5733 /* Check MIB Type and return if event for Remote MIB update */
5734 type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5735 dev_dbg(&pf->pdev->dev,
5736 "LLDP event mib type %s\n", type ? "remote" : "local");
5737 if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5738 /* Update the remote cached instance and return */
5739 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5740 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5741 &hw->remote_dcbx_config);
5745 /* Store the old configuration */
5746 tmp_dcbx_cfg = hw->local_dcbx_config;
5748 /* Reset the old DCBx configuration data */
5749 memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5750 /* Get updated DCBX data from firmware */
5751 ret = i40e_get_dcb_config(&pf->hw);
5753 dev_info(&pf->pdev->dev,
5754 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5755 i40e_stat_str(&pf->hw, ret),
5756 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5760 /* No change detected in DCBX configs */
5761 if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5762 sizeof(tmp_dcbx_cfg))) {
5763 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5767 need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5768 &hw->local_dcbx_config);
5770 i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5775 /* Enable DCB tagging only when more than one TC */
5776 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5777 pf->flags |= I40E_FLAG_DCB_ENABLED;
5779 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5781 set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5782 /* Reconfiguration needed quiesce all VSIs */
5783 i40e_pf_quiesce_all_vsi(pf);
5785 /* Changes in configuration update VEB/VSI */
5786 i40e_dcb_reconfigure(pf);
5788 ret = i40e_resume_port_tx(pf);
5790 clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5791 /* In case of error no point in resuming VSIs */
5795 /* Wait for the PF's queues to be disabled */
5796 ret = i40e_pf_wait_queues_disabled(pf);
5798 /* Schedule PF reset to recover */
5799 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5800 i40e_service_event_schedule(pf);
5802 i40e_pf_unquiesce_all_vsi(pf);
5803 /* Notify the client for the DCB changes */
5804 i40e_notify_client_of_l2_param_changes(pf->vsi[pf->lan_vsi]);
5810 #endif /* CONFIG_I40E_DCB */
5813 * i40e_do_reset_safe - Protected reset path for userland calls.
5814 * @pf: board private structure
5815 * @reset_flags: which reset is requested
5818 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5821 i40e_do_reset(pf, reset_flags);
5826 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5827 * @pf: board private structure
5828 * @e: event info posted on ARQ
5830 * Handler for LAN Queue Overflow Event generated by the firmware for PF
5833 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5834 struct i40e_arq_event_info *e)
5836 struct i40e_aqc_lan_overflow *data =
5837 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5838 u32 queue = le32_to_cpu(data->prtdcb_rupto);
5839 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5840 struct i40e_hw *hw = &pf->hw;
5844 dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5847 /* Queue belongs to VF, find the VF and issue VF reset */
5848 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5849 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5850 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5851 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5852 vf_id -= hw->func_caps.vf_base_id;
5853 vf = &pf->vf[vf_id];
5854 i40e_vc_notify_vf_reset(vf);
5855 /* Allow VF to process pending reset notification */
5857 i40e_reset_vf(vf, false);
5862 * i40e_service_event_complete - Finish up the service event
5863 * @pf: board private structure
5865 static void i40e_service_event_complete(struct i40e_pf *pf)
5867 WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5869 /* flush memory to make sure state is correct before next watchog */
5870 smp_mb__before_atomic();
5871 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5875 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5876 * @pf: board private structure
5878 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5882 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5883 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5888 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5889 * @pf: board private structure
5891 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5895 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5896 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5897 ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5898 I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5903 * i40e_get_global_fd_count - Get total FD filters programmed on device
5904 * @pf: board private structure
5906 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5910 val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5911 fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5912 ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5913 I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5918 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5919 * @pf: board private structure
5921 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5923 struct i40e_fdir_filter *filter;
5924 u32 fcnt_prog, fcnt_avail;
5925 struct hlist_node *node;
5927 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5930 /* Check if, FD SB or ATR was auto disabled and if there is enough room
5933 fcnt_prog = i40e_get_global_fd_count(pf);
5934 fcnt_avail = pf->fdir_pf_filter_count;
5935 if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5936 (pf->fd_add_err == 0) ||
5937 (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5938 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5939 (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5940 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5941 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5942 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5945 /* Wait for some more space to be available to turn on ATR */
5946 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5947 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5948 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5949 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5950 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5951 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5955 /* if hw had a problem adding a filter, delete it */
5956 if (pf->fd_inv > 0) {
5957 hlist_for_each_entry_safe(filter, node,
5958 &pf->fdir_filter_list, fdir_node) {
5959 if (filter->fd_id == pf->fd_inv) {
5960 hlist_del(&filter->fdir_node);
5962 pf->fdir_pf_active_filters--;
5968 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5969 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5971 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5972 * @pf: board private structure
5974 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5976 unsigned long min_flush_time;
5977 int flush_wait_retry = 50;
5978 bool disable_atr = false;
5982 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5985 if (!time_after(jiffies, pf->fd_flush_timestamp +
5986 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5989 /* If the flush is happening too quick and we have mostly SB rules we
5990 * should not re-enable ATR for some time.
5992 min_flush_time = pf->fd_flush_timestamp +
5993 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5994 fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5996 if (!(time_after(jiffies, min_flush_time)) &&
5997 (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5998 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5999 dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
6003 pf->fd_flush_timestamp = jiffies;
6004 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6005 /* flush all filters */
6006 wr32(&pf->hw, I40E_PFQF_CTL_1,
6007 I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6008 i40e_flush(&pf->hw);
6012 /* Check FD flush status every 5-6msec */
6013 usleep_range(5000, 6000);
6014 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6015 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6017 } while (flush_wait_retry--);
6018 if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6019 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6021 /* replay sideband filters */
6022 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6024 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
6025 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6026 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6027 dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6032 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6033 * @pf: board private structure
6035 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6037 return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6040 /* We can see up to 256 filter programming desc in transit if the filters are
6041 * being applied really fast; before we see the first
6042 * filter miss error on Rx queue 0. Accumulating enough error messages before
6043 * reacting will make sure we don't cause flush too often.
6045 #define I40E_MAX_FD_PROGRAM_ERROR 256
6048 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6049 * @pf: board private structure
6051 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6054 /* if interface is down do nothing */
6055 if (test_bit(__I40E_DOWN, &pf->state))
6058 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
6061 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6062 i40e_fdir_flush_and_replay(pf);
6064 i40e_fdir_check_and_reenable(pf);
6069 * i40e_vsi_link_event - notify VSI of a link event
6070 * @vsi: vsi to be notified
6071 * @link_up: link up or down
6073 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6075 if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6078 switch (vsi->type) {
6083 if (!vsi->netdev || !vsi->netdev_registered)
6087 netif_carrier_on(vsi->netdev);
6088 netif_tx_wake_all_queues(vsi->netdev);
6090 netif_carrier_off(vsi->netdev);
6091 netif_tx_stop_all_queues(vsi->netdev);
6095 case I40E_VSI_SRIOV:
6096 case I40E_VSI_VMDQ2:
6098 case I40E_VSI_IWARP:
6099 case I40E_VSI_MIRROR:
6101 /* there is no notification for other VSIs */
6107 * i40e_veb_link_event - notify elements on the veb of a link event
6108 * @veb: veb to be notified
6109 * @link_up: link up or down
6111 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6116 if (!veb || !veb->pf)
6120 /* depth first... */
6121 for (i = 0; i < I40E_MAX_VEB; i++)
6122 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6123 i40e_veb_link_event(pf->veb[i], link_up);
6125 /* ... now the local VSIs */
6126 for (i = 0; i < pf->num_alloc_vsi; i++)
6127 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6128 i40e_vsi_link_event(pf->vsi[i], link_up);
6132 * i40e_link_event - Update netif_carrier status
6133 * @pf: board private structure
6135 static void i40e_link_event(struct i40e_pf *pf)
6137 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6138 u8 new_link_speed, old_link_speed;
6140 bool new_link, old_link;
6142 /* save off old link status information */
6143 pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6145 /* set this to force the get_link_status call to refresh state */
6146 pf->hw.phy.get_link_info = true;
6148 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6150 status = i40e_get_link_status(&pf->hw, &new_link);
6152 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6157 old_link_speed = pf->hw.phy.link_info_old.link_speed;
6158 new_link_speed = pf->hw.phy.link_info.link_speed;
6160 if (new_link == old_link &&
6161 new_link_speed == old_link_speed &&
6162 (test_bit(__I40E_DOWN, &vsi->state) ||
6163 new_link == netif_carrier_ok(vsi->netdev)))
6166 if (!test_bit(__I40E_DOWN, &vsi->state))
6167 i40e_print_link_message(vsi, new_link);
6169 /* Notify the base of the switch tree connected to
6170 * the link. Floating VEBs are not notified.
6172 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6173 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6175 i40e_vsi_link_event(vsi, new_link);
6178 i40e_vc_notify_link_state(pf);
6180 if (pf->flags & I40E_FLAG_PTP)
6181 i40e_ptp_set_increment(pf);
6185 * i40e_watchdog_subtask - periodic checks not using event driven response
6186 * @pf: board private structure
6188 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6192 /* if interface is down do nothing */
6193 if (test_bit(__I40E_DOWN, &pf->state) ||
6194 test_bit(__I40E_CONFIG_BUSY, &pf->state))
6197 /* make sure we don't do these things too often */
6198 if (time_before(jiffies, (pf->service_timer_previous +
6199 pf->service_timer_period)))
6201 pf->service_timer_previous = jiffies;
6203 if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6204 i40e_link_event(pf);
6206 /* Update the stats for active netdevs so the network stack
6207 * can look at updated numbers whenever it cares to
6209 for (i = 0; i < pf->num_alloc_vsi; i++)
6210 if (pf->vsi[i] && pf->vsi[i]->netdev)
6211 i40e_update_stats(pf->vsi[i]);
6213 if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6214 /* Update the stats for the active switching components */
6215 for (i = 0; i < I40E_MAX_VEB; i++)
6217 i40e_update_veb_stats(pf->veb[i]);
6220 i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6224 * i40e_reset_subtask - Set up for resetting the device and driver
6225 * @pf: board private structure
6227 static void i40e_reset_subtask(struct i40e_pf *pf)
6229 u32 reset_flags = 0;
6232 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6233 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6234 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6236 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6237 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6238 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6240 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6241 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6242 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6244 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6245 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6246 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6248 if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6249 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6250 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6253 /* If there's a recovery already waiting, it takes
6254 * precedence before starting a new reset sequence.
6256 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6257 i40e_handle_reset_warning(pf);
6261 /* If we're already down or resetting, just bail */
6263 !test_bit(__I40E_DOWN, &pf->state) &&
6264 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6265 i40e_do_reset(pf, reset_flags);
6272 * i40e_handle_link_event - Handle link event
6273 * @pf: board private structure
6274 * @e: event info posted on ARQ
6276 static void i40e_handle_link_event(struct i40e_pf *pf,
6277 struct i40e_arq_event_info *e)
6279 struct i40e_aqc_get_link_status *status =
6280 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6282 /* Do a new status request to re-enable LSE reporting
6283 * and load new status information into the hw struct
6284 * This completely ignores any state information
6285 * in the ARQ event info, instead choosing to always
6286 * issue the AQ update link status command.
6288 i40e_link_event(pf);
6290 /* check for unqualified module, if link is down */
6291 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6292 (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6293 (!(status->link_info & I40E_AQ_LINK_UP)))
6294 dev_err(&pf->pdev->dev,
6295 "The driver failed to link because an unqualified module was detected.\n");
6299 * i40e_clean_adminq_subtask - Clean the AdminQ rings
6300 * @pf: board private structure
6302 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6304 struct i40e_arq_event_info event;
6305 struct i40e_hw *hw = &pf->hw;
6312 /* Do not run clean AQ when PF reset fails */
6313 if (test_bit(__I40E_RESET_FAILED, &pf->state))
6316 /* check for error indications */
6317 val = rd32(&pf->hw, pf->hw.aq.arq.len);
6319 if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6320 if (hw->debug_mask & I40E_DEBUG_AQ)
6321 dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6322 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6324 if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6325 if (hw->debug_mask & I40E_DEBUG_AQ)
6326 dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6327 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6328 pf->arq_overflows++;
6330 if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6331 if (hw->debug_mask & I40E_DEBUG_AQ)
6332 dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6333 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6336 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6338 val = rd32(&pf->hw, pf->hw.aq.asq.len);
6340 if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6341 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6342 dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6343 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6345 if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6346 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6347 dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6348 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6350 if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6351 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6352 dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6353 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6356 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6358 event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6359 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6364 ret = i40e_clean_arq_element(hw, &event, &pending);
6365 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6368 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6372 opcode = le16_to_cpu(event.desc.opcode);
6375 case i40e_aqc_opc_get_link_status:
6376 i40e_handle_link_event(pf, &event);
6378 case i40e_aqc_opc_send_msg_to_pf:
6379 ret = i40e_vc_process_vf_msg(pf,
6380 le16_to_cpu(event.desc.retval),
6381 le32_to_cpu(event.desc.cookie_high),
6382 le32_to_cpu(event.desc.cookie_low),
6386 case i40e_aqc_opc_lldp_update_mib:
6387 dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6388 #ifdef CONFIG_I40E_DCB
6390 ret = i40e_handle_lldp_event(pf, &event);
6392 #endif /* CONFIG_I40E_DCB */
6394 case i40e_aqc_opc_event_lan_overflow:
6395 dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6396 i40e_handle_lan_overflow_event(pf, &event);
6398 case i40e_aqc_opc_send_msg_to_peer:
6399 dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6401 case i40e_aqc_opc_nvm_erase:
6402 case i40e_aqc_opc_nvm_update:
6403 case i40e_aqc_opc_oem_post_update:
6404 i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6405 "ARQ NVM operation 0x%04x completed\n",
6409 dev_info(&pf->pdev->dev,
6410 "ARQ: Unknown event 0x%04x ignored\n",
6414 } while (pending && (i++ < pf->adminq_work_limit));
6416 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6417 /* re-enable Admin queue interrupt cause */
6418 val = rd32(hw, I40E_PFINT_ICR0_ENA);
6419 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6420 wr32(hw, I40E_PFINT_ICR0_ENA, val);
6423 kfree(event.msg_buf);
6427 * i40e_verify_eeprom - make sure eeprom is good to use
6428 * @pf: board private structure
6430 static void i40e_verify_eeprom(struct i40e_pf *pf)
6434 err = i40e_diag_eeprom_test(&pf->hw);
6436 /* retry in case of garbage read */
6437 err = i40e_diag_eeprom_test(&pf->hw);
6439 dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6441 set_bit(__I40E_BAD_EEPROM, &pf->state);
6445 if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6446 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6447 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6452 * i40e_enable_pf_switch_lb
6453 * @pf: pointer to the PF structure
6455 * enable switch loop back or die - no point in a return value
6457 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6459 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6460 struct i40e_vsi_context ctxt;
6463 ctxt.seid = pf->main_vsi_seid;
6464 ctxt.pf_num = pf->hw.pf_id;
6466 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6468 dev_info(&pf->pdev->dev,
6469 "couldn't get PF vsi config, err %s aq_err %s\n",
6470 i40e_stat_str(&pf->hw, ret),
6471 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6474 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6475 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6476 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6478 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6480 dev_info(&pf->pdev->dev,
6481 "update vsi switch failed, err %s aq_err %s\n",
6482 i40e_stat_str(&pf->hw, ret),
6483 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6488 * i40e_disable_pf_switch_lb
6489 * @pf: pointer to the PF structure
6491 * disable switch loop back or die - no point in a return value
6493 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6495 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6496 struct i40e_vsi_context ctxt;
6499 ctxt.seid = pf->main_vsi_seid;
6500 ctxt.pf_num = pf->hw.pf_id;
6502 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6504 dev_info(&pf->pdev->dev,
6505 "couldn't get PF vsi config, err %s aq_err %s\n",
6506 i40e_stat_str(&pf->hw, ret),
6507 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6510 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6511 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6512 ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6514 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6516 dev_info(&pf->pdev->dev,
6517 "update vsi switch failed, err %s aq_err %s\n",
6518 i40e_stat_str(&pf->hw, ret),
6519 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6524 * i40e_config_bridge_mode - Configure the HW bridge mode
6525 * @veb: pointer to the bridge instance
6527 * Configure the loop back mode for the LAN VSI that is downlink to the
6528 * specified HW bridge instance. It is expected this function is called
6529 * when a new HW bridge is instantiated.
6531 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6533 struct i40e_pf *pf = veb->pf;
6535 if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6536 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6537 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6538 if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6539 i40e_disable_pf_switch_lb(pf);
6541 i40e_enable_pf_switch_lb(pf);
6545 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6546 * @veb: pointer to the VEB instance
6548 * This is a recursive function that first builds the attached VSIs then
6549 * recurses in to build the next layer of VEB. We track the connections
6550 * through our own index numbers because the seid's from the HW could
6551 * change across the reset.
6553 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6555 struct i40e_vsi *ctl_vsi = NULL;
6556 struct i40e_pf *pf = veb->pf;
6560 /* build VSI that owns this VEB, temporarily attached to base VEB */
6561 for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6563 pf->vsi[v]->veb_idx == veb->idx &&
6564 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6565 ctl_vsi = pf->vsi[v];
6570 dev_info(&pf->pdev->dev,
6571 "missing owner VSI for veb_idx %d\n", veb->idx);
6573 goto end_reconstitute;
6575 if (ctl_vsi != pf->vsi[pf->lan_vsi])
6576 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6577 ret = i40e_add_vsi(ctl_vsi);
6579 dev_info(&pf->pdev->dev,
6580 "rebuild of veb_idx %d owner VSI failed: %d\n",
6582 goto end_reconstitute;
6584 i40e_vsi_reset_stats(ctl_vsi);
6586 /* create the VEB in the switch and move the VSI onto the VEB */
6587 ret = i40e_add_veb(veb, ctl_vsi);
6589 goto end_reconstitute;
6591 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6592 veb->bridge_mode = BRIDGE_MODE_VEB;
6594 veb->bridge_mode = BRIDGE_MODE_VEPA;
6595 i40e_config_bridge_mode(veb);
6597 /* create the remaining VSIs attached to this VEB */
6598 for (v = 0; v < pf->num_alloc_vsi; v++) {
6599 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6602 if (pf->vsi[v]->veb_idx == veb->idx) {
6603 struct i40e_vsi *vsi = pf->vsi[v];
6605 vsi->uplink_seid = veb->seid;
6606 ret = i40e_add_vsi(vsi);
6608 dev_info(&pf->pdev->dev,
6609 "rebuild of vsi_idx %d failed: %d\n",
6611 goto end_reconstitute;
6613 i40e_vsi_reset_stats(vsi);
6617 /* create any VEBs attached to this VEB - RECURSION */
6618 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6619 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6620 pf->veb[veb_idx]->uplink_seid = veb->seid;
6621 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6632 * i40e_get_capabilities - get info about the HW
6633 * @pf: the PF struct
6635 static int i40e_get_capabilities(struct i40e_pf *pf)
6637 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6642 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6644 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6648 /* this loads the data into the hw struct for us */
6649 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6651 i40e_aqc_opc_list_func_capabilities,
6653 /* data loaded, buffer no longer needed */
6656 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6657 /* retry with a larger buffer */
6658 buf_len = data_size;
6659 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6660 dev_info(&pf->pdev->dev,
6661 "capability discovery failed, err %s aq_err %s\n",
6662 i40e_stat_str(&pf->hw, err),
6663 i40e_aq_str(&pf->hw,
6664 pf->hw.aq.asq_last_status));
6669 if (pf->hw.debug_mask & I40E_DEBUG_USER)
6670 dev_info(&pf->pdev->dev,
6671 "pf=%d, num_vfs=%d, msix_pf=%d, msix_vf=%d, fd_g=%d, fd_b=%d, pf_max_q=%d num_vsi=%d\n",
6672 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6673 pf->hw.func_caps.num_msix_vectors,
6674 pf->hw.func_caps.num_msix_vectors_vf,
6675 pf->hw.func_caps.fd_filters_guaranteed,
6676 pf->hw.func_caps.fd_filters_best_effort,
6677 pf->hw.func_caps.num_tx_qp,
6678 pf->hw.func_caps.num_vsis);
6680 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6681 + pf->hw.func_caps.num_vfs)
6682 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6683 dev_info(&pf->pdev->dev,
6684 "got num_vsis %d, setting num_vsis to %d\n",
6685 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6686 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6692 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6695 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6696 * @pf: board private structure
6698 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6700 struct i40e_vsi *vsi;
6703 /* quick workaround for an NVM issue that leaves a critical register
6706 if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6707 static const u32 hkey[] = {
6708 0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6709 0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6710 0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6713 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6714 wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6717 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6720 /* find existing VSI and see if it needs configuring */
6722 for (i = 0; i < pf->num_alloc_vsi; i++) {
6723 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6729 /* create a new VSI if none exists */
6731 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6732 pf->vsi[pf->lan_vsi]->seid, 0);
6734 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6735 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6740 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6744 * i40e_fdir_teardown - release the Flow Director resources
6745 * @pf: board private structure
6747 static void i40e_fdir_teardown(struct i40e_pf *pf)
6751 i40e_fdir_filter_exit(pf);
6752 for (i = 0; i < pf->num_alloc_vsi; i++) {
6753 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6754 i40e_vsi_release(pf->vsi[i]);
6761 * i40e_prep_for_reset - prep for the core to reset
6762 * @pf: board private structure
6764 * Close up the VFs and other things in prep for PF Reset.
6766 static void i40e_prep_for_reset(struct i40e_pf *pf)
6768 struct i40e_hw *hw = &pf->hw;
6769 i40e_status ret = 0;
6772 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6773 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6775 if (i40e_check_asq_alive(&pf->hw))
6776 i40e_vc_notify_reset(pf);
6778 dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6780 /* quiesce the VSIs and their queues that are not already DOWN */
6781 i40e_pf_quiesce_all_vsi(pf);
6783 for (v = 0; v < pf->num_alloc_vsi; v++) {
6785 pf->vsi[v]->seid = 0;
6788 i40e_shutdown_adminq(&pf->hw);
6790 /* call shutdown HMC */
6791 if (hw->hmc.hmc_obj) {
6792 ret = i40e_shutdown_lan_hmc(hw);
6794 dev_warn(&pf->pdev->dev,
6795 "shutdown_lan_hmc failed: %d\n", ret);
6800 * i40e_send_version - update firmware with driver version
6803 static void i40e_send_version(struct i40e_pf *pf)
6805 struct i40e_driver_version dv;
6807 dv.major_version = DRV_VERSION_MAJOR;
6808 dv.minor_version = DRV_VERSION_MINOR;
6809 dv.build_version = DRV_VERSION_BUILD;
6810 dv.subbuild_version = 0;
6811 strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6812 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6816 * i40e_reset_and_rebuild - reset and rebuild using a saved config
6817 * @pf: board private structure
6818 * @reinit: if the Main VSI needs to re-initialized.
6820 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6822 struct i40e_hw *hw = &pf->hw;
6823 u8 set_fc_aq_fail = 0;
6828 /* Now we wait for GRST to settle out.
6829 * We don't have to delete the VEBs or VSIs from the hw switch
6830 * because the reset will make them disappear.
6832 ret = i40e_pf_reset(hw);
6834 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6835 set_bit(__I40E_RESET_FAILED, &pf->state);
6836 goto clear_recovery;
6840 if (test_bit(__I40E_DOWN, &pf->state))
6841 goto clear_recovery;
6842 dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6844 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6845 ret = i40e_init_adminq(&pf->hw);
6847 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6848 i40e_stat_str(&pf->hw, ret),
6849 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6850 goto clear_recovery;
6853 /* re-verify the eeprom if we just had an EMP reset */
6854 if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6855 i40e_verify_eeprom(pf);
6857 i40e_clear_pxe_mode(hw);
6858 ret = i40e_get_capabilities(pf);
6860 goto end_core_reset;
6862 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6863 hw->func_caps.num_rx_qp,
6864 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6866 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6867 goto end_core_reset;
6869 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6871 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6872 goto end_core_reset;
6875 #ifdef CONFIG_I40E_DCB
6876 ret = i40e_init_pf_dcb(pf);
6878 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6879 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6880 /* Continue without DCB enabled */
6882 #endif /* CONFIG_I40E_DCB */
6884 i40e_init_pf_fcoe(pf);
6887 /* do basic switch setup */
6888 ret = i40e_setup_pf_switch(pf, reinit);
6890 goto end_core_reset;
6892 /* The driver only wants link up/down and module qualification
6893 * reports from firmware. Note the negative logic.
6895 ret = i40e_aq_set_phy_int_mask(&pf->hw,
6896 ~(I40E_AQ_EVENT_LINK_UPDOWN |
6897 I40E_AQ_EVENT_MEDIA_NA |
6898 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6900 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6901 i40e_stat_str(&pf->hw, ret),
6902 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6904 /* make sure our flow control settings are restored */
6905 ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6907 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6908 i40e_stat_str(&pf->hw, ret),
6909 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6911 /* Rebuild the VSIs and VEBs that existed before reset.
6912 * They are still in our local switch element arrays, so only
6913 * need to rebuild the switch model in the HW.
6915 * If there were VEBs but the reconstitution failed, we'll try
6916 * try to recover minimal use by getting the basic PF VSI working.
6918 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6919 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6920 /* find the one VEB connected to the MAC, and find orphans */
6921 for (v = 0; v < I40E_MAX_VEB; v++) {
6925 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6926 pf->veb[v]->uplink_seid == 0) {
6927 ret = i40e_reconstitute_veb(pf->veb[v]);
6932 /* If Main VEB failed, we're in deep doodoo,
6933 * so give up rebuilding the switch and set up
6934 * for minimal rebuild of PF VSI.
6935 * If orphan failed, we'll report the error
6936 * but try to keep going.
6938 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6939 dev_info(&pf->pdev->dev,
6940 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6942 pf->vsi[pf->lan_vsi]->uplink_seid
6945 } else if (pf->veb[v]->uplink_seid == 0) {
6946 dev_info(&pf->pdev->dev,
6947 "rebuild of orphan VEB failed: %d\n",
6954 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6955 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6956 /* no VEB, so rebuild only the Main VSI */
6957 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6959 dev_info(&pf->pdev->dev,
6960 "rebuild of Main VSI failed: %d\n", ret);
6961 goto end_core_reset;
6965 /* Reconfigure hardware for allowing smaller MSS in the case
6966 * of TSO, so that we avoid the MDD being fired and causing
6967 * a reset in the case of small MSS+TSO.
6969 #define I40E_REG_MSS 0x000E64DC
6970 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6971 #define I40E_64BYTE_MSS 0x400000
6972 val = rd32(hw, I40E_REG_MSS);
6973 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6974 val &= ~I40E_REG_MSS_MIN_MASK;
6975 val |= I40E_64BYTE_MSS;
6976 wr32(hw, I40E_REG_MSS, val);
6979 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6981 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6983 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6984 i40e_stat_str(&pf->hw, ret),
6985 i40e_aq_str(&pf->hw,
6986 pf->hw.aq.asq_last_status));
6988 /* reinit the misc interrupt */
6989 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6990 ret = i40e_setup_misc_vector(pf);
6992 /* Add a filter to drop all Flow control frames from any VSI from being
6993 * transmitted. By doing so we stop a malicious VF from sending out
6994 * PAUSE or PFC frames and potentially controlling traffic for other
6996 * The FW can still send Flow control frames if enabled.
6998 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
7001 /* restart the VSIs that were rebuilt and running before the reset */
7002 i40e_pf_unquiesce_all_vsi(pf);
7004 if (pf->num_alloc_vfs) {
7005 for (v = 0; v < pf->num_alloc_vfs; v++)
7006 i40e_reset_vf(&pf->vf[v], true);
7009 /* tell the firmware that we're starting */
7010 i40e_send_version(pf);
7013 clear_bit(__I40E_RESET_FAILED, &pf->state);
7015 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
7019 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7020 * @pf: board private structure
7022 * Close up the VFs and other things in prep for a Core Reset,
7023 * then get ready to rebuild the world.
7025 static void i40e_handle_reset_warning(struct i40e_pf *pf)
7027 i40e_prep_for_reset(pf);
7028 i40e_reset_and_rebuild(pf, false);
7032 * i40e_handle_mdd_event
7033 * @pf: pointer to the PF structure
7035 * Called from the MDD irq handler to identify possibly malicious vfs
7037 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7039 struct i40e_hw *hw = &pf->hw;
7040 bool mdd_detected = false;
7041 bool pf_mdd_detected = false;
7046 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7049 /* find what triggered the MDD event */
7050 reg = rd32(hw, I40E_GL_MDET_TX);
7051 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7052 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7053 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7054 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7055 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7056 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7057 I40E_GL_MDET_TX_EVENT_SHIFT;
7058 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7059 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7060 pf->hw.func_caps.base_queue;
7061 if (netif_msg_tx_err(pf))
7062 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7063 event, queue, pf_num, vf_num);
7064 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7065 mdd_detected = true;
7067 reg = rd32(hw, I40E_GL_MDET_RX);
7068 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7069 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7070 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7071 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7072 I40E_GL_MDET_RX_EVENT_SHIFT;
7073 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7074 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7075 pf->hw.func_caps.base_queue;
7076 if (netif_msg_rx_err(pf))
7077 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7078 event, queue, func);
7079 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7080 mdd_detected = true;
7084 reg = rd32(hw, I40E_PF_MDET_TX);
7085 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7086 wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7087 dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7088 pf_mdd_detected = true;
7090 reg = rd32(hw, I40E_PF_MDET_RX);
7091 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7092 wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7093 dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7094 pf_mdd_detected = true;
7096 /* Queue belongs to the PF, initiate a reset */
7097 if (pf_mdd_detected) {
7098 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7099 i40e_service_event_schedule(pf);
7103 /* see if one of the VFs needs its hand slapped */
7104 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7106 reg = rd32(hw, I40E_VP_MDET_TX(i));
7107 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7108 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7109 vf->num_mdd_events++;
7110 dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7114 reg = rd32(hw, I40E_VP_MDET_RX(i));
7115 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7116 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7117 vf->num_mdd_events++;
7118 dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7122 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7123 dev_info(&pf->pdev->dev,
7124 "Too many MDD events on VF %d, disabled\n", i);
7125 dev_info(&pf->pdev->dev,
7126 "Use PF Control I/F to re-enable the VF\n");
7127 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7131 /* re-enable mdd interrupt cause */
7132 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7133 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7134 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7135 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7140 * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7141 * @pf: board private structure
7143 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7145 struct i40e_hw *hw = &pf->hw;
7150 if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7153 pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7155 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7156 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7157 pf->pending_udp_bitmap &= ~BIT_ULL(i);
7158 port = pf->udp_ports[i].index;
7160 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7161 pf->udp_ports[i].type,
7164 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7167 dev_dbg(&pf->pdev->dev,
7168 "%s %s port %d, index %d failed, err %s aq_err %s\n",
7169 pf->udp_ports[i].type ? "vxlan" : "geneve",
7170 port ? "add" : "delete",
7172 i40e_stat_str(&pf->hw, ret),
7173 i40e_aq_str(&pf->hw,
7174 pf->hw.aq.asq_last_status));
7175 pf->udp_ports[i].index = 0;
7182 * i40e_service_task - Run the driver's async subtasks
7183 * @work: pointer to work_struct containing our data
7185 static void i40e_service_task(struct work_struct *work)
7187 struct i40e_pf *pf = container_of(work,
7190 unsigned long start_time = jiffies;
7192 /* don't bother with service tasks if a reset is in progress */
7193 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7194 i40e_service_event_complete(pf);
7198 i40e_detect_recover_hung(pf);
7199 i40e_sync_filters_subtask(pf);
7200 i40e_reset_subtask(pf);
7201 i40e_handle_mdd_event(pf);
7202 i40e_vc_process_vflr_event(pf);
7203 i40e_watchdog_subtask(pf);
7204 i40e_fdir_reinit_subtask(pf);
7205 i40e_client_subtask(pf);
7206 i40e_sync_filters_subtask(pf);
7207 i40e_sync_udp_filters_subtask(pf);
7208 i40e_clean_adminq_subtask(pf);
7210 i40e_service_event_complete(pf);
7212 /* If the tasks have taken longer than one timer cycle or there
7213 * is more work to be done, reschedule the service task now
7214 * rather than wait for the timer to tick again.
7216 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7217 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
7218 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
7219 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7220 i40e_service_event_schedule(pf);
7224 * i40e_service_timer - timer callback
7225 * @data: pointer to PF struct
7227 static void i40e_service_timer(unsigned long data)
7229 struct i40e_pf *pf = (struct i40e_pf *)data;
7231 mod_timer(&pf->service_timer,
7232 round_jiffies(jiffies + pf->service_timer_period));
7233 i40e_service_event_schedule(pf);
7237 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7238 * @vsi: the VSI being configured
7240 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7242 struct i40e_pf *pf = vsi->back;
7244 switch (vsi->type) {
7246 vsi->alloc_queue_pairs = pf->num_lan_qps;
7247 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7248 I40E_REQ_DESCRIPTOR_MULTIPLE);
7249 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7250 vsi->num_q_vectors = pf->num_lan_msix;
7252 vsi->num_q_vectors = 1;
7257 vsi->alloc_queue_pairs = 1;
7258 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7259 I40E_REQ_DESCRIPTOR_MULTIPLE);
7260 vsi->num_q_vectors = pf->num_fdsb_msix;
7263 case I40E_VSI_VMDQ2:
7264 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7265 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7266 I40E_REQ_DESCRIPTOR_MULTIPLE);
7267 vsi->num_q_vectors = pf->num_vmdq_msix;
7270 case I40E_VSI_SRIOV:
7271 vsi->alloc_queue_pairs = pf->num_vf_qps;
7272 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7273 I40E_REQ_DESCRIPTOR_MULTIPLE);
7278 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7279 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7280 I40E_REQ_DESCRIPTOR_MULTIPLE);
7281 vsi->num_q_vectors = pf->num_fcoe_msix;
7284 #endif /* I40E_FCOE */
7294 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7295 * @type: VSI pointer
7296 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7298 * On error: returns error code (negative)
7299 * On success: returns 0
7301 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7306 /* allocate memory for both Tx and Rx ring pointers */
7307 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7308 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7311 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7313 if (alloc_qvectors) {
7314 /* allocate memory for q_vector pointers */
7315 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7316 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7317 if (!vsi->q_vectors) {
7325 kfree(vsi->tx_rings);
7330 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7331 * @pf: board private structure
7332 * @type: type of VSI
7334 * On error: returns error code (negative)
7335 * On success: returns vsi index in PF (positive)
7337 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7340 struct i40e_vsi *vsi;
7344 /* Need to protect the allocation of the VSIs at the PF level */
7345 mutex_lock(&pf->switch_mutex);
7347 /* VSI list may be fragmented if VSI creation/destruction has
7348 * been happening. We can afford to do a quick scan to look
7349 * for any free VSIs in the list.
7351 * find next empty vsi slot, looping back around if necessary
7354 while (i < pf->num_alloc_vsi && pf->vsi[i])
7356 if (i >= pf->num_alloc_vsi) {
7358 while (i < pf->next_vsi && pf->vsi[i])
7362 if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7363 vsi_idx = i; /* Found one! */
7366 goto unlock_pf; /* out of VSI slots! */
7370 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7377 set_bit(__I40E_DOWN, &vsi->state);
7380 vsi->int_rate_limit = 0;
7381 vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7382 pf->rss_table_size : 64;
7383 vsi->netdev_registered = false;
7384 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7385 INIT_LIST_HEAD(&vsi->mac_filter_list);
7386 vsi->irqs_ready = false;
7388 ret = i40e_set_num_rings_in_vsi(vsi);
7392 ret = i40e_vsi_alloc_arrays(vsi, true);
7396 /* Setup default MSIX irq handler for VSI */
7397 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7399 /* Initialize VSI lock */
7400 spin_lock_init(&vsi->mac_filter_list_lock);
7401 pf->vsi[vsi_idx] = vsi;
7406 pf->next_vsi = i - 1;
7409 mutex_unlock(&pf->switch_mutex);
7414 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7415 * @type: VSI pointer
7416 * @free_qvectors: a bool to specify if q_vectors need to be freed.
7418 * On error: returns error code (negative)
7419 * On success: returns 0
7421 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7423 /* free the ring and vector containers */
7424 if (free_qvectors) {
7425 kfree(vsi->q_vectors);
7426 vsi->q_vectors = NULL;
7428 kfree(vsi->tx_rings);
7429 vsi->tx_rings = NULL;
7430 vsi->rx_rings = NULL;
7434 * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7436 * @vsi: Pointer to VSI structure
7438 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7443 kfree(vsi->rss_hkey_user);
7444 vsi->rss_hkey_user = NULL;
7446 kfree(vsi->rss_lut_user);
7447 vsi->rss_lut_user = NULL;
7451 * i40e_vsi_clear - Deallocate the VSI provided
7452 * @vsi: the VSI being un-configured
7454 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7465 mutex_lock(&pf->switch_mutex);
7466 if (!pf->vsi[vsi->idx]) {
7467 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7468 vsi->idx, vsi->idx, vsi, vsi->type);
7472 if (pf->vsi[vsi->idx] != vsi) {
7473 dev_err(&pf->pdev->dev,
7474 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7475 pf->vsi[vsi->idx]->idx,
7477 pf->vsi[vsi->idx]->type,
7478 vsi->idx, vsi, vsi->type);
7482 /* updates the PF for this cleared vsi */
7483 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7484 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7486 i40e_vsi_free_arrays(vsi, true);
7487 i40e_clear_rss_config_user(vsi);
7489 pf->vsi[vsi->idx] = NULL;
7490 if (vsi->idx < pf->next_vsi)
7491 pf->next_vsi = vsi->idx;
7494 mutex_unlock(&pf->switch_mutex);
7502 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7503 * @vsi: the VSI being cleaned
7505 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7509 if (vsi->tx_rings && vsi->tx_rings[0]) {
7510 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7511 kfree_rcu(vsi->tx_rings[i], rcu);
7512 vsi->tx_rings[i] = NULL;
7513 vsi->rx_rings[i] = NULL;
7519 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7520 * @vsi: the VSI being configured
7522 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7524 struct i40e_ring *tx_ring, *rx_ring;
7525 struct i40e_pf *pf = vsi->back;
7528 /* Set basic values in the rings to be used later during open() */
7529 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7530 /* allocate space for both Tx and Rx in one shot */
7531 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7535 tx_ring->queue_index = i;
7536 tx_ring->reg_idx = vsi->base_queue + i;
7537 tx_ring->ring_active = false;
7539 tx_ring->netdev = vsi->netdev;
7540 tx_ring->dev = &pf->pdev->dev;
7541 tx_ring->count = vsi->num_desc;
7543 tx_ring->dcb_tc = 0;
7544 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7545 tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7546 tx_ring->tx_itr_setting = pf->tx_itr_default;
7547 vsi->tx_rings[i] = tx_ring;
7549 rx_ring = &tx_ring[1];
7550 rx_ring->queue_index = i;
7551 rx_ring->reg_idx = vsi->base_queue + i;
7552 rx_ring->ring_active = false;
7554 rx_ring->netdev = vsi->netdev;
7555 rx_ring->dev = &pf->pdev->dev;
7556 rx_ring->count = vsi->num_desc;
7558 rx_ring->dcb_tc = 0;
7559 rx_ring->rx_itr_setting = pf->rx_itr_default;
7560 vsi->rx_rings[i] = rx_ring;
7566 i40e_vsi_clear_rings(vsi);
7571 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7572 * @pf: board private structure
7573 * @vectors: the number of MSI-X vectors to request
7575 * Returns the number of vectors reserved, or error
7577 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7579 vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7580 I40E_MIN_MSIX, vectors);
7582 dev_info(&pf->pdev->dev,
7583 "MSI-X vector reservation failed: %d\n", vectors);
7591 * i40e_init_msix - Setup the MSIX capability
7592 * @pf: board private structure
7594 * Work with the OS to set up the MSIX vectors needed.
7596 * Returns the number of vectors reserved or negative on failure
7598 static int i40e_init_msix(struct i40e_pf *pf)
7600 struct i40e_hw *hw = &pf->hw;
7604 int iwarp_requested = 0;
7606 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7609 /* The number of vectors we'll request will be comprised of:
7610 * - Add 1 for "other" cause for Admin Queue events, etc.
7611 * - The number of LAN queue pairs
7612 * - Queues being used for RSS.
7613 * We don't need as many as max_rss_size vectors.
7614 * use rss_size instead in the calculation since that
7615 * is governed by number of cpus in the system.
7616 * - assumes symmetric Tx/Rx pairing
7617 * - The number of VMDq pairs
7618 * - The CPU count within the NUMA node if iWARP is enabled
7620 * - The number of FCOE qps.
7622 * Once we count this up, try the request.
7624 * If we can't get what we want, we'll simplify to nearly nothing
7625 * and try again. If that still fails, we punt.
7627 vectors_left = hw->func_caps.num_msix_vectors;
7630 /* reserve one vector for miscellaneous handler */
7636 /* reserve vectors for the main PF traffic queues */
7637 pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7638 vectors_left -= pf->num_lan_msix;
7639 v_budget += pf->num_lan_msix;
7641 /* reserve one vector for sideband flow director */
7642 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7644 pf->num_fdsb_msix = 1;
7648 pf->num_fdsb_msix = 0;
7649 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7654 /* can we reserve enough for FCoE? */
7655 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7657 pf->num_fcoe_msix = 0;
7658 else if (vectors_left >= pf->num_fcoe_qps)
7659 pf->num_fcoe_msix = pf->num_fcoe_qps;
7661 pf->num_fcoe_msix = 1;
7662 v_budget += pf->num_fcoe_msix;
7663 vectors_left -= pf->num_fcoe_msix;
7667 /* can we reserve enough for iWARP? */
7668 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7670 pf->num_iwarp_msix = 0;
7671 else if (vectors_left < pf->num_iwarp_msix)
7672 pf->num_iwarp_msix = 1;
7673 v_budget += pf->num_iwarp_msix;
7674 vectors_left -= pf->num_iwarp_msix;
7677 /* any vectors left over go for VMDq support */
7678 if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7679 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7680 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7682 /* if we're short on vectors for what's desired, we limit
7683 * the queues per vmdq. If this is still more than are
7684 * available, the user will need to change the number of
7685 * queues/vectors used by the PF later with the ethtool
7688 if (vmdq_vecs < vmdq_vecs_wanted)
7689 pf->num_vmdq_qps = 1;
7690 pf->num_vmdq_msix = pf->num_vmdq_qps;
7692 v_budget += vmdq_vecs;
7693 vectors_left -= vmdq_vecs;
7696 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7698 if (!pf->msix_entries)
7701 for (i = 0; i < v_budget; i++)
7702 pf->msix_entries[i].entry = i;
7703 v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7705 if (v_actual != v_budget) {
7706 /* If we have limited resources, we will start with no vectors
7707 * for the special features and then allocate vectors to some
7708 * of these features based on the policy and at the end disable
7709 * the features that did not get any vectors.
7711 iwarp_requested = pf->num_iwarp_msix;
7712 pf->num_iwarp_msix = 0;
7714 pf->num_fcoe_qps = 0;
7715 pf->num_fcoe_msix = 0;
7717 pf->num_vmdq_msix = 0;
7720 if (v_actual < I40E_MIN_MSIX) {
7721 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7722 kfree(pf->msix_entries);
7723 pf->msix_entries = NULL;
7726 } else if (v_actual == I40E_MIN_MSIX) {
7727 /* Adjust for minimal MSIX use */
7728 pf->num_vmdq_vsis = 0;
7729 pf->num_vmdq_qps = 0;
7730 pf->num_lan_qps = 1;
7731 pf->num_lan_msix = 1;
7733 } else if (v_actual != v_budget) {
7736 /* reserve the misc vector */
7739 /* Scale vector usage down */
7740 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
7741 pf->num_vmdq_vsis = 1;
7742 pf->num_vmdq_qps = 1;
7743 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7745 /* partition out the remaining vectors */
7748 pf->num_lan_msix = 1;
7751 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7752 pf->num_lan_msix = 1;
7753 pf->num_iwarp_msix = 1;
7755 pf->num_lan_msix = 2;
7758 /* give one vector to FCoE */
7759 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7760 pf->num_lan_msix = 1;
7761 pf->num_fcoe_msix = 1;
7766 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7767 pf->num_iwarp_msix = min_t(int, (vec / 3),
7769 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7770 I40E_DEFAULT_NUM_VMDQ_VSI);
7772 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7773 I40E_DEFAULT_NUM_VMDQ_VSI);
7775 pf->num_lan_msix = min_t(int,
7776 (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7779 /* give one vector to FCoE */
7780 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7781 pf->num_fcoe_msix = 1;
7789 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7790 (pf->num_vmdq_msix == 0)) {
7791 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7792 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7795 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7796 (pf->num_iwarp_msix == 0)) {
7797 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7798 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7802 if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7803 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7804 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7811 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7812 * @vsi: the VSI being configured
7813 * @v_idx: index of the vector in the vsi struct
7814 * @cpu: cpu to be used on affinity_mask
7816 * We allocate one q_vector. If allocation fails we return -ENOMEM.
7818 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7820 struct i40e_q_vector *q_vector;
7822 /* allocate q_vector */
7823 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7827 q_vector->vsi = vsi;
7828 q_vector->v_idx = v_idx;
7829 cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7832 netif_napi_add(vsi->netdev, &q_vector->napi,
7833 i40e_napi_poll, NAPI_POLL_WEIGHT);
7835 q_vector->rx.latency_range = I40E_LOW_LATENCY;
7836 q_vector->tx.latency_range = I40E_LOW_LATENCY;
7838 /* tie q_vector and vsi together */
7839 vsi->q_vectors[v_idx] = q_vector;
7845 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7846 * @vsi: the VSI being configured
7848 * We allocate one q_vector per queue interrupt. If allocation fails we
7851 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7853 struct i40e_pf *pf = vsi->back;
7854 int err, v_idx, num_q_vectors, current_cpu;
7856 /* if not MSIX, give the one vector only to the LAN VSI */
7857 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7858 num_q_vectors = vsi->num_q_vectors;
7859 else if (vsi == pf->vsi[pf->lan_vsi])
7864 current_cpu = cpumask_first(cpu_online_mask);
7866 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7867 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7870 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7871 if (unlikely(current_cpu >= nr_cpu_ids))
7872 current_cpu = cpumask_first(cpu_online_mask);
7879 i40e_free_q_vector(vsi, v_idx);
7885 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7886 * @pf: board private structure to initialize
7888 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7893 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7894 vectors = i40e_init_msix(pf);
7896 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
7897 I40E_FLAG_IWARP_ENABLED |
7899 I40E_FLAG_FCOE_ENABLED |
7901 I40E_FLAG_RSS_ENABLED |
7902 I40E_FLAG_DCB_CAPABLE |
7903 I40E_FLAG_DCB_ENABLED |
7904 I40E_FLAG_SRIOV_ENABLED |
7905 I40E_FLAG_FD_SB_ENABLED |
7906 I40E_FLAG_FD_ATR_ENABLED |
7907 I40E_FLAG_VMDQ_ENABLED);
7909 /* rework the queue expectations without MSIX */
7910 i40e_determine_queue_usage(pf);
7914 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7915 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7916 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7917 vectors = pci_enable_msi(pf->pdev);
7919 dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7921 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7923 vectors = 1; /* one MSI or Legacy vector */
7926 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7927 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7929 /* set up vector assignment tracking */
7930 size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7931 pf->irq_pile = kzalloc(size, GFP_KERNEL);
7932 if (!pf->irq_pile) {
7933 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7936 pf->irq_pile->num_entries = vectors;
7937 pf->irq_pile->search_hint = 0;
7939 /* track first vector for misc interrupts, ignore return */
7940 (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7946 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7947 * @pf: board private structure
7949 * This sets up the handler for MSIX 0, which is used to manage the
7950 * non-queue interrupts, e.g. AdminQ and errors. This is not used
7951 * when in MSI or Legacy interrupt mode.
7953 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7955 struct i40e_hw *hw = &pf->hw;
7958 /* Only request the irq if this is the first time through, and
7959 * not when we're rebuilding after a Reset
7961 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7962 err = request_irq(pf->msix_entries[0].vector,
7963 i40e_intr, 0, pf->int_name, pf);
7965 dev_info(&pf->pdev->dev,
7966 "request_irq for %s failed: %d\n",
7972 i40e_enable_misc_int_causes(pf);
7974 /* associate no queues to the misc vector */
7975 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7976 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7980 i40e_irq_dynamic_enable_icr0(pf, true);
7986 * i40e_config_rss_aq - Prepare for RSS using AQ commands
7987 * @vsi: vsi structure
7988 * @seed: RSS hash seed
7990 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7991 u8 *lut, u16 lut_size)
7993 struct i40e_aqc_get_set_rss_key_data rss_key;
7994 struct i40e_pf *pf = vsi->back;
7995 struct i40e_hw *hw = &pf->hw;
7996 bool pf_lut = false;
8000 memcpy(&rss_key, seed, sizeof(rss_key));
8002 rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
8006 /* Populate the LUT with max no. of queues in round robin fashion */
8007 for (i = 0; i < vsi->rss_table_size; i++)
8008 rss_lut[i] = i % vsi->rss_size;
8010 ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
8012 dev_info(&pf->pdev->dev,
8013 "Cannot set RSS key, err %s aq_err %s\n",
8014 i40e_stat_str(&pf->hw, ret),
8015 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8016 goto config_rss_aq_out;
8019 if (vsi->type == I40E_VSI_MAIN)
8022 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
8023 vsi->rss_table_size);
8025 dev_info(&pf->pdev->dev,
8026 "Cannot set RSS lut, err %s aq_err %s\n",
8027 i40e_stat_str(&pf->hw, ret),
8028 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8036 * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8037 * @vsi: VSI structure
8039 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8041 u8 seed[I40E_HKEY_ARRAY_SIZE];
8042 struct i40e_pf *pf = vsi->back;
8046 if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8049 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8053 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8054 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8055 vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
8056 ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8063 * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8064 * @vsi: Pointer to vsi structure
8065 * @seed: Buffter to store the hash keys
8066 * @lut: Buffer to store the lookup table entries
8067 * @lut_size: Size of buffer to store the lookup table entries
8069 * Return 0 on success, negative on failure
8071 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8072 u8 *lut, u16 lut_size)
8074 struct i40e_pf *pf = vsi->back;
8075 struct i40e_hw *hw = &pf->hw;
8079 ret = i40e_aq_get_rss_key(hw, vsi->id,
8080 (struct i40e_aqc_get_set_rss_key_data *)seed);
8082 dev_info(&pf->pdev->dev,
8083 "Cannot get RSS key, err %s aq_err %s\n",
8084 i40e_stat_str(&pf->hw, ret),
8085 i40e_aq_str(&pf->hw,
8086 pf->hw.aq.asq_last_status));
8092 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8094 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8096 dev_info(&pf->pdev->dev,
8097 "Cannot get RSS lut, err %s aq_err %s\n",
8098 i40e_stat_str(&pf->hw, ret),
8099 i40e_aq_str(&pf->hw,
8100 pf->hw.aq.asq_last_status));
8109 * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8110 * @vsi: Pointer to vsi structure
8111 * @seed: RSS hash seed
8112 * @lut: Lookup table
8113 * @lut_size: Lookup table size
8115 * Returns 0 on success, negative on failure
8117 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8118 const u8 *lut, u16 lut_size)
8120 struct i40e_pf *pf = vsi->back;
8121 struct i40e_hw *hw = &pf->hw;
8122 u16 vf_id = vsi->vf_id;
8125 /* Fill out hash function seed */
8127 u32 *seed_dw = (u32 *)seed;
8129 if (vsi->type == I40E_VSI_MAIN) {
8130 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8131 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8133 } else if (vsi->type == I40E_VSI_SRIOV) {
8134 for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8135 i40e_write_rx_ctl(hw,
8136 I40E_VFQF_HKEY1(i, vf_id),
8139 dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8144 u32 *lut_dw = (u32 *)lut;
8146 if (vsi->type == I40E_VSI_MAIN) {
8147 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8149 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8150 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8151 } else if (vsi->type == I40E_VSI_SRIOV) {
8152 if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8154 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8155 i40e_write_rx_ctl(hw,
8156 I40E_VFQF_HLUT1(i, vf_id),
8159 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8168 * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8169 * @vsi: Pointer to VSI structure
8170 * @seed: Buffer to store the keys
8171 * @lut: Buffer to store the lookup table entries
8172 * @lut_size: Size of buffer to store the lookup table entries
8174 * Returns 0 on success, negative on failure
8176 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8177 u8 *lut, u16 lut_size)
8179 struct i40e_pf *pf = vsi->back;
8180 struct i40e_hw *hw = &pf->hw;
8184 u32 *seed_dw = (u32 *)seed;
8186 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8187 seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8190 u32 *lut_dw = (u32 *)lut;
8192 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8194 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8195 lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8202 * i40e_config_rss - Configure RSS keys and lut
8203 * @vsi: Pointer to VSI structure
8204 * @seed: RSS hash seed
8205 * @lut: Lookup table
8206 * @lut_size: Lookup table size
8208 * Returns 0 on success, negative on failure
8210 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8212 struct i40e_pf *pf = vsi->back;
8214 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8215 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8217 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8221 * i40e_get_rss - Get RSS keys and lut
8222 * @vsi: Pointer to VSI structure
8223 * @seed: Buffer to store the keys
8224 * @lut: Buffer to store the lookup table entries
8225 * lut_size: Size of buffer to store the lookup table entries
8227 * Returns 0 on success, negative on failure
8229 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8231 struct i40e_pf *pf = vsi->back;
8233 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8234 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8236 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8240 * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8241 * @pf: Pointer to board private structure
8242 * @lut: Lookup table
8243 * @rss_table_size: Lookup table size
8244 * @rss_size: Range of queue number for hashing
8246 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8247 u16 rss_table_size, u16 rss_size)
8251 for (i = 0; i < rss_table_size; i++)
8252 lut[i] = i % rss_size;
8256 * i40e_pf_config_rss - Prepare for RSS if used
8257 * @pf: board private structure
8259 static int i40e_pf_config_rss(struct i40e_pf *pf)
8261 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8262 u8 seed[I40E_HKEY_ARRAY_SIZE];
8264 struct i40e_hw *hw = &pf->hw;
8269 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8270 hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8271 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8272 hena |= i40e_pf_get_default_rss_hena(pf);
8274 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8275 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8277 /* Determine the RSS table size based on the hardware capabilities */
8278 reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8279 reg_val = (pf->rss_table_size == 512) ?
8280 (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8281 (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8282 i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8284 /* Determine the RSS size of the VSI */
8286 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8287 vsi->num_queue_pairs);
8289 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8293 /* Use user configured lut if there is one, otherwise use default */
8294 if (vsi->rss_lut_user)
8295 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8297 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8299 /* Use user configured hash key if there is one, otherwise
8302 if (vsi->rss_hkey_user)
8303 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8305 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8306 ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8313 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8314 * @pf: board private structure
8315 * @queue_count: the requested queue count for rss.
8317 * returns 0 if rss is not enabled, if enabled returns the final rss queue
8318 * count which may be different from the requested queue count.
8320 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8322 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8325 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8328 new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8330 if (queue_count != vsi->num_queue_pairs) {
8331 vsi->req_queue_pairs = queue_count;
8332 i40e_prep_for_reset(pf);
8334 pf->alloc_rss_size = new_rss_size;
8336 i40e_reset_and_rebuild(pf, true);
8338 /* Discard the user configured hash keys and lut, if less
8339 * queues are enabled.
8341 if (queue_count < vsi->rss_size) {
8342 i40e_clear_rss_config_user(vsi);
8343 dev_dbg(&pf->pdev->dev,
8344 "discard user configured hash keys and lut\n");
8347 /* Reset vsi->rss_size, as number of enabled queues changed */
8348 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8349 vsi->num_queue_pairs);
8351 i40e_pf_config_rss(pf);
8353 dev_info(&pf->pdev->dev, "RSS count/HW max RSS count: %d/%d\n",
8354 pf->alloc_rss_size, pf->rss_size_max);
8355 return pf->alloc_rss_size;
8359 * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8360 * @pf: board private structure
8362 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8365 bool min_valid, max_valid;
8368 status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8369 &min_valid, &max_valid);
8373 pf->npar_min_bw = min_bw;
8375 pf->npar_max_bw = max_bw;
8382 * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8383 * @pf: board private structure
8385 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8387 struct i40e_aqc_configure_partition_bw_data bw_data;
8390 /* Set the valid bit for this PF */
8391 bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8392 bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8393 bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8395 /* Set the new bandwidths */
8396 status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8402 * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8403 * @pf: board private structure
8405 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8407 /* Commit temporary BW setting to permanent NVM image */
8408 enum i40e_admin_queue_err last_aq_status;
8412 if (pf->hw.partition_id != 1) {
8413 dev_info(&pf->pdev->dev,
8414 "Commit BW only works on partition 1! This is partition %d",
8415 pf->hw.partition_id);
8416 ret = I40E_NOT_SUPPORTED;
8420 /* Acquire NVM for read access */
8421 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8422 last_aq_status = pf->hw.aq.asq_last_status;
8424 dev_info(&pf->pdev->dev,
8425 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8426 i40e_stat_str(&pf->hw, ret),
8427 i40e_aq_str(&pf->hw, last_aq_status));
8431 /* Read word 0x10 of NVM - SW compatibility word 1 */
8432 ret = i40e_aq_read_nvm(&pf->hw,
8433 I40E_SR_NVM_CONTROL_WORD,
8434 0x10, sizeof(nvm_word), &nvm_word,
8436 /* Save off last admin queue command status before releasing
8439 last_aq_status = pf->hw.aq.asq_last_status;
8440 i40e_release_nvm(&pf->hw);
8442 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8443 i40e_stat_str(&pf->hw, ret),
8444 i40e_aq_str(&pf->hw, last_aq_status));
8448 /* Wait a bit for NVM release to complete */
8451 /* Acquire NVM for write access */
8452 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8453 last_aq_status = pf->hw.aq.asq_last_status;
8455 dev_info(&pf->pdev->dev,
8456 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8457 i40e_stat_str(&pf->hw, ret),
8458 i40e_aq_str(&pf->hw, last_aq_status));
8461 /* Write it back out unchanged to initiate update NVM,
8462 * which will force a write of the shadow (alt) RAM to
8463 * the NVM - thus storing the bandwidth values permanently.
8465 ret = i40e_aq_update_nvm(&pf->hw,
8466 I40E_SR_NVM_CONTROL_WORD,
8467 0x10, sizeof(nvm_word),
8468 &nvm_word, true, NULL);
8469 /* Save off last admin queue command status before releasing
8472 last_aq_status = pf->hw.aq.asq_last_status;
8473 i40e_release_nvm(&pf->hw);
8475 dev_info(&pf->pdev->dev,
8476 "BW settings NOT SAVED, err %s aq_err %s\n",
8477 i40e_stat_str(&pf->hw, ret),
8478 i40e_aq_str(&pf->hw, last_aq_status));
8485 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8486 * @pf: board private structure to initialize
8488 * i40e_sw_init initializes the Adapter private data structure.
8489 * Fields are initialized based on PCI device information and
8490 * OS network device settings (MTU size).
8492 static int i40e_sw_init(struct i40e_pf *pf)
8497 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8498 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8499 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8500 if (I40E_DEBUG_USER & debug)
8501 pf->hw.debug_mask = debug;
8502 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8503 I40E_DEFAULT_MSG_ENABLE);
8506 /* Set default capability flags */
8507 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8508 I40E_FLAG_MSI_ENABLED |
8509 I40E_FLAG_MSIX_ENABLED;
8511 /* Set default ITR */
8512 pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8513 pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8515 /* Depending on PF configurations, it is possible that the RSS
8516 * maximum might end up larger than the available queues
8518 pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8519 pf->alloc_rss_size = 1;
8520 pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8521 pf->rss_size_max = min_t(int, pf->rss_size_max,
8522 pf->hw.func_caps.num_tx_qp);
8523 if (pf->hw.func_caps.rss) {
8524 pf->flags |= I40E_FLAG_RSS_ENABLED;
8525 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8529 /* MFP mode enabled */
8530 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8531 pf->flags |= I40E_FLAG_MFP_ENABLED;
8532 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8533 if (i40e_get_npar_bw_setting(pf))
8534 dev_warn(&pf->pdev->dev,
8535 "Could not get NPAR bw settings\n");
8537 dev_info(&pf->pdev->dev,
8538 "Min BW = %8.8x, Max BW = %8.8x\n",
8539 pf->npar_min_bw, pf->npar_max_bw);
8542 /* FW/NVM is not yet fixed in this regard */
8543 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8544 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8545 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8546 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8547 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8548 pf->hw.num_partitions > 1)
8549 dev_info(&pf->pdev->dev,
8550 "Flow Director Sideband mode Disabled in MFP mode\n");
8552 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8553 pf->fdir_pf_filter_count =
8554 pf->hw.func_caps.fd_filters_guaranteed;
8555 pf->hw.fdir_shared_filter_count =
8556 pf->hw.func_caps.fd_filters_best_effort;
8559 if (i40e_is_mac_710(&pf->hw) &&
8560 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8561 (pf->hw.aq.fw_maj_ver < 4))) {
8562 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8563 /* No DCB support for FW < v4.33 */
8564 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8567 /* Disable FW LLDP if FW < v4.3 */
8568 if (i40e_is_mac_710(&pf->hw) &&
8569 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8570 (pf->hw.aq.fw_maj_ver < 4)))
8571 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8573 /* Use the FW Set LLDP MIB API if FW > v4.40 */
8574 if (i40e_is_mac_710(&pf->hw) &&
8575 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8576 (pf->hw.aq.fw_maj_ver >= 5)))
8577 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8579 if (pf->hw.func_caps.vmdq) {
8580 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8581 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8582 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8585 if (pf->hw.func_caps.iwarp) {
8586 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8587 /* IWARP needs one extra vector for CQP just like MISC.*/
8588 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8592 i40e_init_pf_fcoe(pf);
8594 #endif /* I40E_FCOE */
8595 #ifdef CONFIG_PCI_IOV
8596 if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8597 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8598 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8599 pf->num_req_vfs = min_t(int,
8600 pf->hw.func_caps.num_vfs,
8603 #endif /* CONFIG_PCI_IOV */
8604 if (pf->hw.mac.type == I40E_MAC_X722) {
8605 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8606 I40E_FLAG_128_QP_RSS_CAPABLE |
8607 I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8608 I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8609 I40E_FLAG_WB_ON_ITR_CAPABLE |
8610 I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8611 I40E_FLAG_NO_PCI_LINK_CHECK |
8612 I40E_FLAG_100M_SGMII_CAPABLE |
8613 I40E_FLAG_USE_SET_LLDP_MIB |
8614 I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8615 } else if ((pf->hw.aq.api_maj_ver > 1) ||
8616 ((pf->hw.aq.api_maj_ver == 1) &&
8617 (pf->hw.aq.api_min_ver > 4))) {
8618 /* Supported in FW API version higher than 1.4 */
8619 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8620 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8622 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8625 pf->eeprom_version = 0xDEAD;
8626 pf->lan_veb = I40E_NO_VEB;
8627 pf->lan_vsi = I40E_NO_VSI;
8629 /* By default FW has this off for performance reasons */
8630 pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8632 /* set up queue assignment tracking */
8633 size = sizeof(struct i40e_lump_tracking)
8634 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8635 pf->qp_pile = kzalloc(size, GFP_KERNEL);
8640 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8641 pf->qp_pile->search_hint = 0;
8643 pf->tx_timeout_recovery_level = 1;
8645 mutex_init(&pf->switch_mutex);
8647 /* If NPAR is enabled nudge the Tx scheduler */
8648 if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8649 i40e_set_npar_bw_setting(pf);
8656 * i40e_set_ntuple - set the ntuple feature flag and take action
8657 * @pf: board private structure to initialize
8658 * @features: the feature set that the stack is suggesting
8660 * returns a bool to indicate if reset needs to happen
8662 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8664 bool need_reset = false;
8666 /* Check if Flow Director n-tuple support was enabled or disabled. If
8667 * the state changed, we need to reset.
8669 if (features & NETIF_F_NTUPLE) {
8670 /* Enable filters and mark for reset */
8671 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8673 /* enable FD_SB only if there is MSI-X vector */
8674 if (pf->num_fdsb_msix > 0)
8675 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8677 /* turn off filters, mark for reset and clear SW filter list */
8678 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8680 i40e_fdir_filter_exit(pf);
8682 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8683 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8684 /* reset fd counters */
8685 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8686 pf->fdir_pf_active_filters = 0;
8687 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8688 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8689 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8690 /* if ATR was auto disabled it can be re-enabled. */
8691 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8692 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8693 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8699 * i40e_set_features - set the netdev feature flags
8700 * @netdev: ptr to the netdev being adjusted
8701 * @features: the feature set that the stack is suggesting
8703 static int i40e_set_features(struct net_device *netdev,
8704 netdev_features_t features)
8706 struct i40e_netdev_priv *np = netdev_priv(netdev);
8707 struct i40e_vsi *vsi = np->vsi;
8708 struct i40e_pf *pf = vsi->back;
8711 if (features & NETIF_F_HW_VLAN_CTAG_RX)
8712 i40e_vlan_stripping_enable(vsi);
8714 i40e_vlan_stripping_disable(vsi);
8716 need_reset = i40e_set_ntuple(pf, features);
8719 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8725 * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8726 * @pf: board private structure
8727 * @port: The UDP port to look up
8729 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8731 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8735 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8736 if (pf->udp_ports[i].index == port)
8744 * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8745 * @netdev: This physical port's netdev
8746 * @ti: Tunnel endpoint information
8748 static void i40e_udp_tunnel_add(struct net_device *netdev,
8749 struct udp_tunnel_info *ti)
8751 struct i40e_netdev_priv *np = netdev_priv(netdev);
8752 struct i40e_vsi *vsi = np->vsi;
8753 struct i40e_pf *pf = vsi->back;
8754 __be16 port = ti->port;
8758 idx = i40e_get_udp_port_idx(pf, port);
8760 /* Check if port already exists */
8761 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8762 netdev_info(netdev, "port %d already offloaded\n",
8767 /* Now check if there is space to add the new port */
8768 next_idx = i40e_get_udp_port_idx(pf, 0);
8770 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8771 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8777 case UDP_TUNNEL_TYPE_VXLAN:
8778 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8780 case UDP_TUNNEL_TYPE_GENEVE:
8781 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8783 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8789 /* New port: add it and mark its index in the bitmap */
8790 pf->udp_ports[next_idx].index = port;
8791 pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8792 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8796 * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
8797 * @netdev: This physical port's netdev
8798 * @ti: Tunnel endpoint information
8800 static void i40e_udp_tunnel_del(struct net_device *netdev,
8801 struct udp_tunnel_info *ti)
8803 struct i40e_netdev_priv *np = netdev_priv(netdev);
8804 struct i40e_vsi *vsi = np->vsi;
8805 struct i40e_pf *pf = vsi->back;
8806 __be16 port = ti->port;
8809 idx = i40e_get_udp_port_idx(pf, port);
8811 /* Check if port already exists */
8812 if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
8816 case UDP_TUNNEL_TYPE_VXLAN:
8817 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
8820 case UDP_TUNNEL_TYPE_GENEVE:
8821 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
8828 /* if port exists, set it to 0 (mark for deletion)
8829 * and make it pending
8831 pf->udp_ports[idx].index = 0;
8832 pf->pending_udp_bitmap |= BIT_ULL(idx);
8833 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8837 netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
8841 static int i40e_get_phys_port_id(struct net_device *netdev,
8842 struct netdev_phys_item_id *ppid)
8844 struct i40e_netdev_priv *np = netdev_priv(netdev);
8845 struct i40e_pf *pf = np->vsi->back;
8846 struct i40e_hw *hw = &pf->hw;
8848 if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8851 ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8852 memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8858 * i40e_ndo_fdb_add - add an entry to the hardware database
8859 * @ndm: the input from the stack
8860 * @tb: pointer to array of nladdr (unused)
8861 * @dev: the net device pointer
8862 * @addr: the MAC address entry being added
8863 * @flags: instructions from stack about fdb operation
8865 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8866 struct net_device *dev,
8867 const unsigned char *addr, u16 vid,
8870 struct i40e_netdev_priv *np = netdev_priv(dev);
8871 struct i40e_pf *pf = np->vsi->back;
8874 if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8878 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8882 /* Hardware does not support aging addresses so if a
8883 * ndm_state is given only allow permanent addresses
8885 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8886 netdev_info(dev, "FDB only supports static addresses\n");
8890 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8891 err = dev_uc_add_excl(dev, addr);
8892 else if (is_multicast_ether_addr(addr))
8893 err = dev_mc_add_excl(dev, addr);
8897 /* Only return duplicate errors if NLM_F_EXCL is set */
8898 if (err == -EEXIST && !(flags & NLM_F_EXCL))
8905 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8906 * @dev: the netdev being configured
8907 * @nlh: RTNL message
8909 * Inserts a new hardware bridge if not already created and
8910 * enables the bridging mode requested (VEB or VEPA). If the
8911 * hardware bridge has already been inserted and the request
8912 * is to change the mode then that requires a PF reset to
8913 * allow rebuild of the components with required hardware
8914 * bridge mode enabled.
8916 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8917 struct nlmsghdr *nlh,
8920 struct i40e_netdev_priv *np = netdev_priv(dev);
8921 struct i40e_vsi *vsi = np->vsi;
8922 struct i40e_pf *pf = vsi->back;
8923 struct i40e_veb *veb = NULL;
8924 struct nlattr *attr, *br_spec;
8927 /* Only for PF VSI for now */
8928 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8931 /* Find the HW bridge for PF VSI */
8932 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8933 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8937 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8939 nla_for_each_nested(attr, br_spec, rem) {
8942 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8945 mode = nla_get_u16(attr);
8946 if ((mode != BRIDGE_MODE_VEPA) &&
8947 (mode != BRIDGE_MODE_VEB))
8950 /* Insert a new HW bridge */
8952 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8953 vsi->tc_config.enabled_tc);
8955 veb->bridge_mode = mode;
8956 i40e_config_bridge_mode(veb);
8958 /* No Bridge HW offload available */
8962 } else if (mode != veb->bridge_mode) {
8963 /* Existing HW bridge but different mode needs reset */
8964 veb->bridge_mode = mode;
8965 /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8966 if (mode == BRIDGE_MODE_VEB)
8967 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8969 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8970 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8979 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8982 * @seq: RTNL message seq #
8983 * @dev: the netdev being configured
8984 * @filter_mask: unused
8985 * @nlflags: netlink flags passed in
8987 * Return the mode in which the hardware bridge is operating in
8990 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8991 struct net_device *dev,
8992 u32 __always_unused filter_mask,
8995 struct i40e_netdev_priv *np = netdev_priv(dev);
8996 struct i40e_vsi *vsi = np->vsi;
8997 struct i40e_pf *pf = vsi->back;
8998 struct i40e_veb *veb = NULL;
9001 /* Only for PF VSI for now */
9002 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9005 /* Find the HW bridge for the PF VSI */
9006 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9007 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9014 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9015 nlflags, 0, 0, filter_mask, NULL);
9018 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9019 * inner mac plus all inner ethertypes.
9021 #define I40E_MAX_TUNNEL_HDR_LEN 128
9023 * i40e_features_check - Validate encapsulated packet conforms to limits
9025 * @dev: This physical port's netdev
9026 * @features: Offload features that the stack believes apply
9028 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9029 struct net_device *dev,
9030 netdev_features_t features)
9032 if (skb->encapsulation &&
9033 ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9034 I40E_MAX_TUNNEL_HDR_LEN))
9035 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9040 static const struct net_device_ops i40e_netdev_ops = {
9041 .ndo_open = i40e_open,
9042 .ndo_stop = i40e_close,
9043 .ndo_start_xmit = i40e_lan_xmit_frame,
9044 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
9045 .ndo_set_rx_mode = i40e_set_rx_mode,
9046 .ndo_validate_addr = eth_validate_addr,
9047 .ndo_set_mac_address = i40e_set_mac,
9048 .ndo_change_mtu = i40e_change_mtu,
9049 .ndo_do_ioctl = i40e_ioctl,
9050 .ndo_tx_timeout = i40e_tx_timeout,
9051 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
9052 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
9053 #ifdef CONFIG_NET_POLL_CONTROLLER
9054 .ndo_poll_controller = i40e_netpoll,
9056 .ndo_setup_tc = __i40e_setup_tc,
9058 .ndo_fcoe_enable = i40e_fcoe_enable,
9059 .ndo_fcoe_disable = i40e_fcoe_disable,
9061 .ndo_set_features = i40e_set_features,
9062 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
9063 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
9064 .ndo_set_vf_rate = i40e_ndo_set_vf_bw,
9065 .ndo_get_vf_config = i40e_ndo_get_vf_config,
9066 .ndo_set_vf_link_state = i40e_ndo_set_vf_link_state,
9067 .ndo_set_vf_spoofchk = i40e_ndo_set_vf_spoofchk,
9068 .ndo_set_vf_trust = i40e_ndo_set_vf_trust,
9069 .ndo_udp_tunnel_add = i40e_udp_tunnel_add,
9070 .ndo_udp_tunnel_del = i40e_udp_tunnel_del,
9071 .ndo_get_phys_port_id = i40e_get_phys_port_id,
9072 .ndo_fdb_add = i40e_ndo_fdb_add,
9073 .ndo_features_check = i40e_features_check,
9074 .ndo_bridge_getlink = i40e_ndo_bridge_getlink,
9075 .ndo_bridge_setlink = i40e_ndo_bridge_setlink,
9079 * i40e_config_netdev - Setup the netdev flags
9080 * @vsi: the VSI being configured
9082 * Returns 0 on success, negative value on failure
9084 static int i40e_config_netdev(struct i40e_vsi *vsi)
9086 struct i40e_pf *pf = vsi->back;
9087 struct i40e_hw *hw = &pf->hw;
9088 struct i40e_netdev_priv *np;
9089 struct net_device *netdev;
9090 u8 mac_addr[ETH_ALEN];
9093 etherdev_size = sizeof(struct i40e_netdev_priv);
9094 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9098 vsi->netdev = netdev;
9099 np = netdev_priv(netdev);
9102 netdev->hw_enc_features |= NETIF_F_SG |
9106 NETIF_F_SOFT_FEATURES |
9111 NETIF_F_GSO_GRE_CSUM |
9112 NETIF_F_GSO_IPXIP4 |
9113 NETIF_F_GSO_IPXIP6 |
9114 NETIF_F_GSO_UDP_TUNNEL |
9115 NETIF_F_GSO_UDP_TUNNEL_CSUM |
9116 NETIF_F_GSO_PARTIAL |
9122 if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9123 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9125 netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9127 /* record features VLANs can make use of */
9128 netdev->vlan_features |= netdev->hw_enc_features |
9129 NETIF_F_TSO_MANGLEID;
9131 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9132 netdev->hw_features |= NETIF_F_NTUPLE;
9134 netdev->hw_features |= netdev->hw_enc_features |
9135 NETIF_F_HW_VLAN_CTAG_TX |
9136 NETIF_F_HW_VLAN_CTAG_RX;
9138 netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9139 netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9141 if (vsi->type == I40E_VSI_MAIN) {
9142 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9143 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9144 /* The following steps are necessary to prevent reception
9145 * of tagged packets - some older NVM configurations load a
9146 * default a MAC-VLAN filter that accepts any tagged packet
9147 * which must be replaced by a normal filter.
9149 i40e_rm_default_mac_filter(vsi, mac_addr);
9150 spin_lock_bh(&vsi->mac_filter_list_lock);
9151 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, true);
9152 spin_unlock_bh(&vsi->mac_filter_list_lock);
9154 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9155 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9156 pf->vsi[pf->lan_vsi]->netdev->name);
9157 random_ether_addr(mac_addr);
9159 spin_lock_bh(&vsi->mac_filter_list_lock);
9160 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9161 spin_unlock_bh(&vsi->mac_filter_list_lock);
9164 ether_addr_copy(netdev->dev_addr, mac_addr);
9165 ether_addr_copy(netdev->perm_addr, mac_addr);
9167 netdev->priv_flags |= IFF_UNICAST_FLT;
9168 netdev->priv_flags |= IFF_SUPP_NOFCS;
9169 /* Setup netdev TC information */
9170 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9172 netdev->netdev_ops = &i40e_netdev_ops;
9173 netdev->watchdog_timeo = 5 * HZ;
9174 i40e_set_ethtool_ops(netdev);
9176 i40e_fcoe_config_netdev(netdev, vsi);
9183 * i40e_vsi_delete - Delete a VSI from the switch
9184 * @vsi: the VSI being removed
9186 * Returns 0 on success, negative value on failure
9188 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9190 /* remove default VSI is not allowed */
9191 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9194 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9198 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9199 * @vsi: the VSI being queried
9201 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9203 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9205 struct i40e_veb *veb;
9206 struct i40e_pf *pf = vsi->back;
9208 /* Uplink is not a bridge so default to VEB */
9209 if (vsi->veb_idx == I40E_NO_VEB)
9212 veb = pf->veb[vsi->veb_idx];
9214 dev_info(&pf->pdev->dev,
9215 "There is no veb associated with the bridge\n");
9219 /* Uplink is a bridge in VEPA mode */
9220 if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9223 /* Uplink is a bridge in VEB mode */
9227 /* VEPA is now default bridge, so return 0 */
9232 * i40e_add_vsi - Add a VSI to the switch
9233 * @vsi: the VSI being configured
9235 * This initializes a VSI context depending on the VSI type to be added and
9236 * passes it down to the add_vsi aq command.
9238 static int i40e_add_vsi(struct i40e_vsi *vsi)
9241 i40e_status aq_ret = 0;
9242 struct i40e_pf *pf = vsi->back;
9243 struct i40e_hw *hw = &pf->hw;
9244 struct i40e_vsi_context ctxt;
9245 struct i40e_mac_filter *f, *ftmp;
9247 u8 enabled_tc = 0x1; /* TC0 enabled */
9250 memset(&ctxt, 0, sizeof(ctxt));
9251 switch (vsi->type) {
9253 /* The PF's main VSI is already setup as part of the
9254 * device initialization, so we'll not bother with
9255 * the add_vsi call, but we will retrieve the current
9258 ctxt.seid = pf->main_vsi_seid;
9259 ctxt.pf_num = pf->hw.pf_id;
9261 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9262 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9264 dev_info(&pf->pdev->dev,
9265 "couldn't get PF vsi config, err %s aq_err %s\n",
9266 i40e_stat_str(&pf->hw, ret),
9267 i40e_aq_str(&pf->hw,
9268 pf->hw.aq.asq_last_status));
9271 vsi->info = ctxt.info;
9272 vsi->info.valid_sections = 0;
9274 vsi->seid = ctxt.seid;
9275 vsi->id = ctxt.vsi_number;
9277 enabled_tc = i40e_pf_get_tc_map(pf);
9279 /* MFP mode setup queue map and update VSI */
9280 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9281 !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9282 memset(&ctxt, 0, sizeof(ctxt));
9283 ctxt.seid = pf->main_vsi_seid;
9284 ctxt.pf_num = pf->hw.pf_id;
9286 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9287 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9289 dev_info(&pf->pdev->dev,
9290 "update vsi failed, err %s aq_err %s\n",
9291 i40e_stat_str(&pf->hw, ret),
9292 i40e_aq_str(&pf->hw,
9293 pf->hw.aq.asq_last_status));
9297 /* update the local VSI info queue map */
9298 i40e_vsi_update_queue_map(vsi, &ctxt);
9299 vsi->info.valid_sections = 0;
9301 /* Default/Main VSI is only enabled for TC0
9302 * reconfigure it to enable all TCs that are
9303 * available on the port in SFP mode.
9304 * For MFP case the iSCSI PF would use this
9305 * flow to enable LAN+iSCSI TC.
9307 ret = i40e_vsi_config_tc(vsi, enabled_tc);
9309 dev_info(&pf->pdev->dev,
9310 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9312 i40e_stat_str(&pf->hw, ret),
9313 i40e_aq_str(&pf->hw,
9314 pf->hw.aq.asq_last_status));
9321 ctxt.pf_num = hw->pf_id;
9323 ctxt.uplink_seid = vsi->uplink_seid;
9324 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9325 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9326 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9327 (i40e_is_vsi_uplink_mode_veb(vsi))) {
9328 ctxt.info.valid_sections |=
9329 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9330 ctxt.info.switch_id =
9331 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9333 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9336 case I40E_VSI_VMDQ2:
9337 ctxt.pf_num = hw->pf_id;
9339 ctxt.uplink_seid = vsi->uplink_seid;
9340 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9341 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9343 /* This VSI is connected to VEB so the switch_id
9344 * should be set to zero by default.
9346 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9347 ctxt.info.valid_sections |=
9348 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9349 ctxt.info.switch_id =
9350 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9353 /* Setup the VSI tx/rx queue map for TC0 only for now */
9354 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9357 case I40E_VSI_SRIOV:
9358 ctxt.pf_num = hw->pf_id;
9359 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9360 ctxt.uplink_seid = vsi->uplink_seid;
9361 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9362 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9364 /* This VSI is connected to VEB so the switch_id
9365 * should be set to zero by default.
9367 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9368 ctxt.info.valid_sections |=
9369 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9370 ctxt.info.switch_id =
9371 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9374 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9375 ctxt.info.valid_sections |=
9376 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9377 ctxt.info.queueing_opt_flags |=
9378 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9379 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9382 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9383 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9384 if (pf->vf[vsi->vf_id].spoofchk) {
9385 ctxt.info.valid_sections |=
9386 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9387 ctxt.info.sec_flags |=
9388 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9389 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9391 /* Setup the VSI tx/rx queue map for TC0 only for now */
9392 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9397 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9399 dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9404 #endif /* I40E_FCOE */
9405 case I40E_VSI_IWARP:
9406 /* send down message to iWARP */
9413 if (vsi->type != I40E_VSI_MAIN) {
9414 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9416 dev_info(&vsi->back->pdev->dev,
9417 "add vsi failed, err %s aq_err %s\n",
9418 i40e_stat_str(&pf->hw, ret),
9419 i40e_aq_str(&pf->hw,
9420 pf->hw.aq.asq_last_status));
9424 vsi->info = ctxt.info;
9425 vsi->info.valid_sections = 0;
9426 vsi->seid = ctxt.seid;
9427 vsi->id = ctxt.vsi_number;
9429 /* Except FDIR VSI, for all othet VSI set the broadcast filter */
9430 if (vsi->type != I40E_VSI_FDIR) {
9431 aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9433 ret = i40e_aq_rc_to_posix(aq_ret,
9434 hw->aq.asq_last_status);
9435 dev_info(&pf->pdev->dev,
9436 "set brdcast promisc failed, err %s, aq_err %s\n",
9437 i40e_stat_str(hw, aq_ret),
9438 i40e_aq_str(hw, hw->aq.asq_last_status));
9442 vsi->active_filters = 0;
9443 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9444 spin_lock_bh(&vsi->mac_filter_list_lock);
9445 /* If macvlan filters already exist, force them to get loaded */
9446 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9447 f->state = I40E_FILTER_NEW;
9450 spin_unlock_bh(&vsi->mac_filter_list_lock);
9453 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9454 pf->flags |= I40E_FLAG_FILTER_SYNC;
9457 /* Update VSI BW information */
9458 ret = i40e_vsi_get_bw_info(vsi);
9460 dev_info(&pf->pdev->dev,
9461 "couldn't get vsi bw info, err %s aq_err %s\n",
9462 i40e_stat_str(&pf->hw, ret),
9463 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9464 /* VSI is already added so not tearing that up */
9473 * i40e_vsi_release - Delete a VSI and free its resources
9474 * @vsi: the VSI being removed
9476 * Returns 0 on success or < 0 on error
9478 int i40e_vsi_release(struct i40e_vsi *vsi)
9480 struct i40e_mac_filter *f, *ftmp;
9481 struct i40e_veb *veb = NULL;
9488 /* release of a VEB-owner or last VSI is not allowed */
9489 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9490 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9491 vsi->seid, vsi->uplink_seid);
9494 if (vsi == pf->vsi[pf->lan_vsi] &&
9495 !test_bit(__I40E_DOWN, &pf->state)) {
9496 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9500 uplink_seid = vsi->uplink_seid;
9501 if (vsi->type != I40E_VSI_SRIOV) {
9502 if (vsi->netdev_registered) {
9503 vsi->netdev_registered = false;
9505 /* results in a call to i40e_close() */
9506 unregister_netdev(vsi->netdev);
9509 i40e_vsi_close(vsi);
9511 i40e_vsi_disable_irq(vsi);
9514 spin_lock_bh(&vsi->mac_filter_list_lock);
9515 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9516 i40e_del_filter(vsi, f->macaddr, f->vlan,
9517 f->is_vf, f->is_netdev);
9518 spin_unlock_bh(&vsi->mac_filter_list_lock);
9520 i40e_sync_vsi_filters(vsi);
9522 i40e_vsi_delete(vsi);
9523 i40e_vsi_free_q_vectors(vsi);
9525 free_netdev(vsi->netdev);
9528 i40e_vsi_clear_rings(vsi);
9529 i40e_vsi_clear(vsi);
9531 /* If this was the last thing on the VEB, except for the
9532 * controlling VSI, remove the VEB, which puts the controlling
9533 * VSI onto the next level down in the switch.
9535 * Well, okay, there's one more exception here: don't remove
9536 * the orphan VEBs yet. We'll wait for an explicit remove request
9537 * from up the network stack.
9539 for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9541 pf->vsi[i]->uplink_seid == uplink_seid &&
9542 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9543 n++; /* count the VSIs */
9546 for (i = 0; i < I40E_MAX_VEB; i++) {
9549 if (pf->veb[i]->uplink_seid == uplink_seid)
9550 n++; /* count the VEBs */
9551 if (pf->veb[i]->seid == uplink_seid)
9554 if (n == 0 && veb && veb->uplink_seid != 0)
9555 i40e_veb_release(veb);
9561 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9562 * @vsi: ptr to the VSI
9564 * This should only be called after i40e_vsi_mem_alloc() which allocates the
9565 * corresponding SW VSI structure and initializes num_queue_pairs for the
9566 * newly allocated VSI.
9568 * Returns 0 on success or negative on failure
9570 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9573 struct i40e_pf *pf = vsi->back;
9575 if (vsi->q_vectors[0]) {
9576 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9581 if (vsi->base_vector) {
9582 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9583 vsi->seid, vsi->base_vector);
9587 ret = i40e_vsi_alloc_q_vectors(vsi);
9589 dev_info(&pf->pdev->dev,
9590 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9591 vsi->num_q_vectors, vsi->seid, ret);
9592 vsi->num_q_vectors = 0;
9593 goto vector_setup_out;
9596 /* In Legacy mode, we do not have to get any other vector since we
9597 * piggyback on the misc/ICR0 for queue interrupts.
9599 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9601 if (vsi->num_q_vectors)
9602 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9603 vsi->num_q_vectors, vsi->idx);
9604 if (vsi->base_vector < 0) {
9605 dev_info(&pf->pdev->dev,
9606 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9607 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9608 i40e_vsi_free_q_vectors(vsi);
9610 goto vector_setup_out;
9618 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9619 * @vsi: pointer to the vsi.
9621 * This re-allocates a vsi's queue resources.
9623 * Returns pointer to the successfully allocated and configured VSI sw struct
9624 * on success, otherwise returns NULL on failure.
9626 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9637 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9638 i40e_vsi_clear_rings(vsi);
9640 i40e_vsi_free_arrays(vsi, false);
9641 i40e_set_num_rings_in_vsi(vsi);
9642 ret = i40e_vsi_alloc_arrays(vsi, false);
9646 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9648 dev_info(&pf->pdev->dev,
9649 "failed to get tracking for %d queues for VSI %d err %d\n",
9650 vsi->alloc_queue_pairs, vsi->seid, ret);
9653 vsi->base_queue = ret;
9655 /* Update the FW view of the VSI. Force a reset of TC and queue
9656 * layout configurations.
9658 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9659 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9660 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9661 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9662 if (vsi->type == I40E_VSI_MAIN)
9663 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9665 /* assign it some queues */
9666 ret = i40e_alloc_rings(vsi);
9670 /* map all of the rings to the q_vectors */
9671 i40e_vsi_map_rings_to_vectors(vsi);
9675 i40e_vsi_free_q_vectors(vsi);
9676 if (vsi->netdev_registered) {
9677 vsi->netdev_registered = false;
9678 unregister_netdev(vsi->netdev);
9679 free_netdev(vsi->netdev);
9682 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9684 i40e_vsi_clear(vsi);
9689 * i40e_vsi_setup - Set up a VSI by a given type
9690 * @pf: board private structure
9692 * @uplink_seid: the switch element to link to
9693 * @param1: usage depends upon VSI type. For VF types, indicates VF id
9695 * This allocates the sw VSI structure and its queue resources, then add a VSI
9696 * to the identified VEB.
9698 * Returns pointer to the successfully allocated and configure VSI sw struct on
9699 * success, otherwise returns NULL on failure.
9701 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9702 u16 uplink_seid, u32 param1)
9704 struct i40e_vsi *vsi = NULL;
9705 struct i40e_veb *veb = NULL;
9709 /* The requested uplink_seid must be either
9710 * - the PF's port seid
9711 * no VEB is needed because this is the PF
9712 * or this is a Flow Director special case VSI
9713 * - seid of an existing VEB
9714 * - seid of a VSI that owns an existing VEB
9715 * - seid of a VSI that doesn't own a VEB
9716 * a new VEB is created and the VSI becomes the owner
9717 * - seid of the PF VSI, which is what creates the first VEB
9718 * this is a special case of the previous
9720 * Find which uplink_seid we were given and create a new VEB if needed
9722 for (i = 0; i < I40E_MAX_VEB; i++) {
9723 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9729 if (!veb && uplink_seid != pf->mac_seid) {
9731 for (i = 0; i < pf->num_alloc_vsi; i++) {
9732 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9738 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9743 if (vsi->uplink_seid == pf->mac_seid)
9744 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9745 vsi->tc_config.enabled_tc);
9746 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9747 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9748 vsi->tc_config.enabled_tc);
9750 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9751 dev_info(&vsi->back->pdev->dev,
9752 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9755 /* We come up by default in VEPA mode if SRIOV is not
9756 * already enabled, in which case we can't force VEPA
9759 if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9760 veb->bridge_mode = BRIDGE_MODE_VEPA;
9761 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9763 i40e_config_bridge_mode(veb);
9765 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9766 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9770 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9774 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9775 uplink_seid = veb->seid;
9778 /* get vsi sw struct */
9779 v_idx = i40e_vsi_mem_alloc(pf, type);
9782 vsi = pf->vsi[v_idx];
9786 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9788 if (type == I40E_VSI_MAIN)
9789 pf->lan_vsi = v_idx;
9790 else if (type == I40E_VSI_SRIOV)
9791 vsi->vf_id = param1;
9792 /* assign it some queues */
9793 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9796 dev_info(&pf->pdev->dev,
9797 "failed to get tracking for %d queues for VSI %d err=%d\n",
9798 vsi->alloc_queue_pairs, vsi->seid, ret);
9801 vsi->base_queue = ret;
9803 /* get a VSI from the hardware */
9804 vsi->uplink_seid = uplink_seid;
9805 ret = i40e_add_vsi(vsi);
9809 switch (vsi->type) {
9810 /* setup the netdev if needed */
9812 /* Apply relevant filters if a platform-specific mac
9813 * address was selected.
9815 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9816 ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9818 dev_warn(&pf->pdev->dev,
9819 "could not set up macaddr; err %d\n",
9823 case I40E_VSI_VMDQ2:
9825 ret = i40e_config_netdev(vsi);
9828 ret = register_netdev(vsi->netdev);
9831 vsi->netdev_registered = true;
9832 netif_carrier_off(vsi->netdev);
9833 #ifdef CONFIG_I40E_DCB
9834 /* Setup DCB netlink interface */
9835 i40e_dcbnl_setup(vsi);
9836 #endif /* CONFIG_I40E_DCB */
9840 /* set up vectors and rings if needed */
9841 ret = i40e_vsi_setup_vectors(vsi);
9845 ret = i40e_alloc_rings(vsi);
9849 /* map all of the rings to the q_vectors */
9850 i40e_vsi_map_rings_to_vectors(vsi);
9852 i40e_vsi_reset_stats(vsi);
9856 /* no netdev or rings for the other VSI types */
9860 if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9861 (vsi->type == I40E_VSI_VMDQ2)) {
9862 ret = i40e_vsi_config_rss(vsi);
9867 i40e_vsi_free_q_vectors(vsi);
9869 if (vsi->netdev_registered) {
9870 vsi->netdev_registered = false;
9871 unregister_netdev(vsi->netdev);
9872 free_netdev(vsi->netdev);
9876 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9878 i40e_vsi_clear(vsi);
9884 * i40e_veb_get_bw_info - Query VEB BW information
9885 * @veb: the veb to query
9887 * Query the Tx scheduler BW configuration data for given VEB
9889 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9891 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9892 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9893 struct i40e_pf *pf = veb->pf;
9894 struct i40e_hw *hw = &pf->hw;
9899 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9902 dev_info(&pf->pdev->dev,
9903 "query veb bw config failed, err %s aq_err %s\n",
9904 i40e_stat_str(&pf->hw, ret),
9905 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9909 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9912 dev_info(&pf->pdev->dev,
9913 "query veb bw ets config failed, err %s aq_err %s\n",
9914 i40e_stat_str(&pf->hw, ret),
9915 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9919 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9920 veb->bw_max_quanta = ets_data.tc_bw_max;
9921 veb->is_abs_credits = bw_data.absolute_credits_enable;
9922 veb->enabled_tc = ets_data.tc_valid_bits;
9923 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9924 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9925 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9926 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9927 veb->bw_tc_limit_credits[i] =
9928 le16_to_cpu(bw_data.tc_bw_limits[i]);
9929 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9937 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9938 * @pf: board private structure
9940 * On error: returns error code (negative)
9941 * On success: returns vsi index in PF (positive)
9943 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9946 struct i40e_veb *veb;
9949 /* Need to protect the allocation of switch elements at the PF level */
9950 mutex_lock(&pf->switch_mutex);
9952 /* VEB list may be fragmented if VEB creation/destruction has
9953 * been happening. We can afford to do a quick scan to look
9954 * for any free slots in the list.
9956 * find next empty veb slot, looping back around if necessary
9959 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9961 if (i >= I40E_MAX_VEB) {
9963 goto err_alloc_veb; /* out of VEB slots! */
9966 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9973 veb->enabled_tc = 1;
9978 mutex_unlock(&pf->switch_mutex);
9983 * i40e_switch_branch_release - Delete a branch of the switch tree
9984 * @branch: where to start deleting
9986 * This uses recursion to find the tips of the branch to be
9987 * removed, deleting until we get back to and can delete this VEB.
9989 static void i40e_switch_branch_release(struct i40e_veb *branch)
9991 struct i40e_pf *pf = branch->pf;
9992 u16 branch_seid = branch->seid;
9993 u16 veb_idx = branch->idx;
9996 /* release any VEBs on this VEB - RECURSION */
9997 for (i = 0; i < I40E_MAX_VEB; i++) {
10000 if (pf->veb[i]->uplink_seid == branch->seid)
10001 i40e_switch_branch_release(pf->veb[i]);
10004 /* Release the VSIs on this VEB, but not the owner VSI.
10006 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10007 * the VEB itself, so don't use (*branch) after this loop.
10009 for (i = 0; i < pf->num_alloc_vsi; i++) {
10012 if (pf->vsi[i]->uplink_seid == branch_seid &&
10013 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10014 i40e_vsi_release(pf->vsi[i]);
10018 /* There's one corner case where the VEB might not have been
10019 * removed, so double check it here and remove it if needed.
10020 * This case happens if the veb was created from the debugfs
10021 * commands and no VSIs were added to it.
10023 if (pf->veb[veb_idx])
10024 i40e_veb_release(pf->veb[veb_idx]);
10028 * i40e_veb_clear - remove veb struct
10029 * @veb: the veb to remove
10031 static void i40e_veb_clear(struct i40e_veb *veb)
10037 struct i40e_pf *pf = veb->pf;
10039 mutex_lock(&pf->switch_mutex);
10040 if (pf->veb[veb->idx] == veb)
10041 pf->veb[veb->idx] = NULL;
10042 mutex_unlock(&pf->switch_mutex);
10049 * i40e_veb_release - Delete a VEB and free its resources
10050 * @veb: the VEB being removed
10052 void i40e_veb_release(struct i40e_veb *veb)
10054 struct i40e_vsi *vsi = NULL;
10055 struct i40e_pf *pf;
10060 /* find the remaining VSI and check for extras */
10061 for (i = 0; i < pf->num_alloc_vsi; i++) {
10062 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10068 dev_info(&pf->pdev->dev,
10069 "can't remove VEB %d with %d VSIs left\n",
10074 /* move the remaining VSI to uplink veb */
10075 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10076 if (veb->uplink_seid) {
10077 vsi->uplink_seid = veb->uplink_seid;
10078 if (veb->uplink_seid == pf->mac_seid)
10079 vsi->veb_idx = I40E_NO_VEB;
10081 vsi->veb_idx = veb->veb_idx;
10084 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10085 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10088 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10089 i40e_veb_clear(veb);
10093 * i40e_add_veb - create the VEB in the switch
10094 * @veb: the VEB to be instantiated
10095 * @vsi: the controlling VSI
10097 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10099 struct i40e_pf *pf = veb->pf;
10100 bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10103 ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10104 veb->enabled_tc, false,
10105 &veb->seid, enable_stats, NULL);
10107 /* get a VEB from the hardware */
10109 dev_info(&pf->pdev->dev,
10110 "couldn't add VEB, err %s aq_err %s\n",
10111 i40e_stat_str(&pf->hw, ret),
10112 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10116 /* get statistics counter */
10117 ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10118 &veb->stats_idx, NULL, NULL, NULL);
10120 dev_info(&pf->pdev->dev,
10121 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10122 i40e_stat_str(&pf->hw, ret),
10123 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10126 ret = i40e_veb_get_bw_info(veb);
10128 dev_info(&pf->pdev->dev,
10129 "couldn't get VEB bw info, err %s aq_err %s\n",
10130 i40e_stat_str(&pf->hw, ret),
10131 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10132 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10136 vsi->uplink_seid = veb->seid;
10137 vsi->veb_idx = veb->idx;
10138 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10144 * i40e_veb_setup - Set up a VEB
10145 * @pf: board private structure
10146 * @flags: VEB setup flags
10147 * @uplink_seid: the switch element to link to
10148 * @vsi_seid: the initial VSI seid
10149 * @enabled_tc: Enabled TC bit-map
10151 * This allocates the sw VEB structure and links it into the switch
10152 * It is possible and legal for this to be a duplicate of an already
10153 * existing VEB. It is also possible for both uplink and vsi seids
10154 * to be zero, in order to create a floating VEB.
10156 * Returns pointer to the successfully allocated VEB sw struct on
10157 * success, otherwise returns NULL on failure.
10159 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10160 u16 uplink_seid, u16 vsi_seid,
10163 struct i40e_veb *veb, *uplink_veb = NULL;
10164 int vsi_idx, veb_idx;
10167 /* if one seid is 0, the other must be 0 to create a floating relay */
10168 if ((uplink_seid == 0 || vsi_seid == 0) &&
10169 (uplink_seid + vsi_seid != 0)) {
10170 dev_info(&pf->pdev->dev,
10171 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10172 uplink_seid, vsi_seid);
10176 /* make sure there is such a vsi and uplink */
10177 for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10178 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10180 if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10181 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10186 if (uplink_seid && uplink_seid != pf->mac_seid) {
10187 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10188 if (pf->veb[veb_idx] &&
10189 pf->veb[veb_idx]->seid == uplink_seid) {
10190 uplink_veb = pf->veb[veb_idx];
10195 dev_info(&pf->pdev->dev,
10196 "uplink seid %d not found\n", uplink_seid);
10201 /* get veb sw struct */
10202 veb_idx = i40e_veb_mem_alloc(pf);
10205 veb = pf->veb[veb_idx];
10206 veb->flags = flags;
10207 veb->uplink_seid = uplink_seid;
10208 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10209 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10211 /* create the VEB in the switch */
10212 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10215 if (vsi_idx == pf->lan_vsi)
10216 pf->lan_veb = veb->idx;
10221 i40e_veb_clear(veb);
10227 * i40e_setup_pf_switch_element - set PF vars based on switch type
10228 * @pf: board private structure
10229 * @ele: element we are building info from
10230 * @num_reported: total number of elements
10231 * @printconfig: should we print the contents
10233 * helper function to assist in extracting a few useful SEID values.
10235 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10236 struct i40e_aqc_switch_config_element_resp *ele,
10237 u16 num_reported, bool printconfig)
10239 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10240 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10241 u8 element_type = ele->element_type;
10242 u16 seid = le16_to_cpu(ele->seid);
10245 dev_info(&pf->pdev->dev,
10246 "type=%d seid=%d uplink=%d downlink=%d\n",
10247 element_type, seid, uplink_seid, downlink_seid);
10249 switch (element_type) {
10250 case I40E_SWITCH_ELEMENT_TYPE_MAC:
10251 pf->mac_seid = seid;
10253 case I40E_SWITCH_ELEMENT_TYPE_VEB:
10255 if (uplink_seid != pf->mac_seid)
10257 if (pf->lan_veb == I40E_NO_VEB) {
10260 /* find existing or else empty VEB */
10261 for (v = 0; v < I40E_MAX_VEB; v++) {
10262 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10267 if (pf->lan_veb == I40E_NO_VEB) {
10268 v = i40e_veb_mem_alloc(pf);
10275 pf->veb[pf->lan_veb]->seid = seid;
10276 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10277 pf->veb[pf->lan_veb]->pf = pf;
10278 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10280 case I40E_SWITCH_ELEMENT_TYPE_VSI:
10281 if (num_reported != 1)
10283 /* This is immediately after a reset so we can assume this is
10286 pf->mac_seid = uplink_seid;
10287 pf->pf_seid = downlink_seid;
10288 pf->main_vsi_seid = seid;
10290 dev_info(&pf->pdev->dev,
10291 "pf_seid=%d main_vsi_seid=%d\n",
10292 pf->pf_seid, pf->main_vsi_seid);
10294 case I40E_SWITCH_ELEMENT_TYPE_PF:
10295 case I40E_SWITCH_ELEMENT_TYPE_VF:
10296 case I40E_SWITCH_ELEMENT_TYPE_EMP:
10297 case I40E_SWITCH_ELEMENT_TYPE_BMC:
10298 case I40E_SWITCH_ELEMENT_TYPE_PE:
10299 case I40E_SWITCH_ELEMENT_TYPE_PA:
10300 /* ignore these for now */
10303 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10304 element_type, seid);
10310 * i40e_fetch_switch_configuration - Get switch config from firmware
10311 * @pf: board private structure
10312 * @printconfig: should we print the contents
10314 * Get the current switch configuration from the device and
10315 * extract a few useful SEID values.
10317 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10319 struct i40e_aqc_get_switch_config_resp *sw_config;
10325 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10329 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10331 u16 num_reported, num_total;
10333 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10337 dev_info(&pf->pdev->dev,
10338 "get switch config failed err %s aq_err %s\n",
10339 i40e_stat_str(&pf->hw, ret),
10340 i40e_aq_str(&pf->hw,
10341 pf->hw.aq.asq_last_status));
10346 num_reported = le16_to_cpu(sw_config->header.num_reported);
10347 num_total = le16_to_cpu(sw_config->header.num_total);
10350 dev_info(&pf->pdev->dev,
10351 "header: %d reported %d total\n",
10352 num_reported, num_total);
10354 for (i = 0; i < num_reported; i++) {
10355 struct i40e_aqc_switch_config_element_resp *ele =
10356 &sw_config->element[i];
10358 i40e_setup_pf_switch_element(pf, ele, num_reported,
10361 } while (next_seid != 0);
10368 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10369 * @pf: board private structure
10370 * @reinit: if the Main VSI needs to re-initialized.
10372 * Returns 0 on success, negative value on failure
10374 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10379 /* find out what's out there already */
10380 ret = i40e_fetch_switch_configuration(pf, false);
10382 dev_info(&pf->pdev->dev,
10383 "couldn't fetch switch config, err %s aq_err %s\n",
10384 i40e_stat_str(&pf->hw, ret),
10385 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10388 i40e_pf_reset_stats(pf);
10390 /* set the switch config bit for the whole device to
10391 * support limited promisc or true promisc
10392 * when user requests promisc. The default is limited
10396 if ((pf->hw.pf_id == 0) &&
10397 !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10398 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10400 if (pf->hw.pf_id == 0) {
10403 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10404 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10406 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10407 dev_info(&pf->pdev->dev,
10408 "couldn't set switch config bits, err %s aq_err %s\n",
10409 i40e_stat_str(&pf->hw, ret),
10410 i40e_aq_str(&pf->hw,
10411 pf->hw.aq.asq_last_status));
10412 /* not a fatal problem, just keep going */
10416 /* first time setup */
10417 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10418 struct i40e_vsi *vsi = NULL;
10421 /* Set up the PF VSI associated with the PF's main VSI
10422 * that is already in the HW switch
10424 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10425 uplink_seid = pf->veb[pf->lan_veb]->seid;
10427 uplink_seid = pf->mac_seid;
10428 if (pf->lan_vsi == I40E_NO_VSI)
10429 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10431 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10433 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10434 i40e_fdir_teardown(pf);
10438 /* force a reset of TC and queue layout configurations */
10439 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10441 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10442 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10443 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10445 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10447 i40e_fdir_sb_setup(pf);
10449 /* Setup static PF queue filter control settings */
10450 ret = i40e_setup_pf_filter_control(pf);
10452 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10454 /* Failure here should not stop continuing other steps */
10457 /* enable RSS in the HW, even for only one queue, as the stack can use
10460 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10461 i40e_pf_config_rss(pf);
10463 /* fill in link information and enable LSE reporting */
10464 i40e_update_link_info(&pf->hw);
10465 i40e_link_event(pf);
10467 /* Initialize user-specific link properties */
10468 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10469 I40E_AQ_AN_COMPLETED) ? true : false);
10477 * i40e_determine_queue_usage - Work out queue distribution
10478 * @pf: board private structure
10480 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10484 pf->num_lan_qps = 0;
10486 pf->num_fcoe_qps = 0;
10489 /* Find the max queues to be put into basic use. We'll always be
10490 * using TC0, whether or not DCB is running, and TC0 will get the
10493 queues_left = pf->hw.func_caps.num_tx_qp;
10495 if ((queues_left == 1) ||
10496 !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10497 /* one qp for PF, no queues for anything else */
10499 pf->alloc_rss_size = pf->num_lan_qps = 1;
10501 /* make sure all the fancies are disabled */
10502 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10503 I40E_FLAG_IWARP_ENABLED |
10505 I40E_FLAG_FCOE_ENABLED |
10507 I40E_FLAG_FD_SB_ENABLED |
10508 I40E_FLAG_FD_ATR_ENABLED |
10509 I40E_FLAG_DCB_CAPABLE |
10510 I40E_FLAG_DCB_ENABLED |
10511 I40E_FLAG_SRIOV_ENABLED |
10512 I40E_FLAG_VMDQ_ENABLED);
10513 } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10514 I40E_FLAG_FD_SB_ENABLED |
10515 I40E_FLAG_FD_ATR_ENABLED |
10516 I40E_FLAG_DCB_CAPABLE))) {
10517 /* one qp for PF */
10518 pf->alloc_rss_size = pf->num_lan_qps = 1;
10519 queues_left -= pf->num_lan_qps;
10521 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10522 I40E_FLAG_IWARP_ENABLED |
10524 I40E_FLAG_FCOE_ENABLED |
10526 I40E_FLAG_FD_SB_ENABLED |
10527 I40E_FLAG_FD_ATR_ENABLED |
10528 I40E_FLAG_DCB_ENABLED |
10529 I40E_FLAG_VMDQ_ENABLED);
10531 /* Not enough queues for all TCs */
10532 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10533 (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10534 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10535 I40E_FLAG_DCB_ENABLED);
10536 dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10538 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10539 num_online_cpus());
10540 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10541 pf->hw.func_caps.num_tx_qp);
10543 queues_left -= pf->num_lan_qps;
10547 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10548 if (I40E_DEFAULT_FCOE <= queues_left) {
10549 pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10550 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10551 pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10553 pf->num_fcoe_qps = 0;
10554 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10555 dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10558 queues_left -= pf->num_fcoe_qps;
10562 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10563 if (queues_left > 1) {
10564 queues_left -= 1; /* save 1 queue for FD */
10566 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10567 dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10571 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10572 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10573 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10574 (queues_left / pf->num_vf_qps));
10575 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10578 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10579 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10580 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10581 (queues_left / pf->num_vmdq_qps));
10582 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10585 pf->queues_left = queues_left;
10586 dev_dbg(&pf->pdev->dev,
10587 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10588 pf->hw.func_caps.num_tx_qp,
10589 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10590 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10591 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10594 dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10599 * i40e_setup_pf_filter_control - Setup PF static filter control
10600 * @pf: PF to be setup
10602 * i40e_setup_pf_filter_control sets up a PF's initial filter control
10603 * settings. If PE/FCoE are enabled then it will also set the per PF
10604 * based filter sizes required for them. It also enables Flow director,
10605 * ethertype and macvlan type filter settings for the pf.
10607 * Returns 0 on success, negative on failure
10609 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10611 struct i40e_filter_control_settings *settings = &pf->filter_settings;
10613 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10615 /* Flow Director is enabled */
10616 if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10617 settings->enable_fdir = true;
10619 /* Ethtype and MACVLAN filters enabled for PF */
10620 settings->enable_ethtype = true;
10621 settings->enable_macvlan = true;
10623 if (i40e_set_filter_control(&pf->hw, settings))
10629 #define INFO_STRING_LEN 255
10630 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10631 static void i40e_print_features(struct i40e_pf *pf)
10633 struct i40e_hw *hw = &pf->hw;
10637 buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10641 i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10642 #ifdef CONFIG_PCI_IOV
10643 i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10645 i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10646 pf->hw.func_caps.num_vsis,
10647 pf->vsi[pf->lan_vsi]->num_queue_pairs);
10648 if (pf->flags & I40E_FLAG_RSS_ENABLED)
10649 i += snprintf(&buf[i], REMAIN(i), " RSS");
10650 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10651 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10652 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10653 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10654 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10656 if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10657 i += snprintf(&buf[i], REMAIN(i), " DCB");
10658 i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10659 i += snprintf(&buf[i], REMAIN(i), " Geneve");
10660 if (pf->flags & I40E_FLAG_PTP)
10661 i += snprintf(&buf[i], REMAIN(i), " PTP");
10663 if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10664 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10666 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10667 i += snprintf(&buf[i], REMAIN(i), " VEB");
10669 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10671 dev_info(&pf->pdev->dev, "%s\n", buf);
10673 WARN_ON(i > INFO_STRING_LEN);
10677 * i40e_get_platform_mac_addr - get platform-specific MAC address
10679 * @pdev: PCI device information struct
10680 * @pf: board private structure
10682 * Look up the MAC address in Open Firmware on systems that support it,
10683 * and use IDPROM on SPARC if no OF address is found. On return, the
10684 * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10685 * has been selected.
10687 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10689 pf->flags &= ~I40E_FLAG_PF_MAC;
10690 if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10691 pf->flags |= I40E_FLAG_PF_MAC;
10695 * i40e_probe - Device initialization routine
10696 * @pdev: PCI device information struct
10697 * @ent: entry in i40e_pci_tbl
10699 * i40e_probe initializes a PF identified by a pci_dev structure.
10700 * The OS initialization, configuring of the PF private structure,
10701 * and a hardware reset occur.
10703 * Returns 0 on success, negative on failure
10705 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10707 struct i40e_aq_get_phy_abilities_resp abilities;
10708 struct i40e_pf *pf;
10709 struct i40e_hw *hw;
10710 static u16 pfs_found;
10718 err = pci_enable_device_mem(pdev);
10722 /* set up for high or low dma */
10723 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10725 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10727 dev_err(&pdev->dev,
10728 "DMA configuration failed: 0x%x\n", err);
10733 /* set up pci connections */
10734 err = pci_request_mem_regions(pdev, i40e_driver_name);
10736 dev_info(&pdev->dev,
10737 "pci_request_selected_regions failed %d\n", err);
10741 pci_enable_pcie_error_reporting(pdev);
10742 pci_set_master(pdev);
10744 /* Now that we have a PCI connection, we need to do the
10745 * low level device setup. This is primarily setting up
10746 * the Admin Queue structures and then querying for the
10747 * device's current profile information.
10749 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10756 set_bit(__I40E_DOWN, &pf->state);
10761 pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10762 I40E_MAX_CSR_SPACE);
10764 hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10765 if (!hw->hw_addr) {
10767 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10768 (unsigned int)pci_resource_start(pdev, 0),
10769 pf->ioremap_len, err);
10772 hw->vendor_id = pdev->vendor;
10773 hw->device_id = pdev->device;
10774 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10775 hw->subsystem_vendor_id = pdev->subsystem_vendor;
10776 hw->subsystem_device_id = pdev->subsystem_device;
10777 hw->bus.device = PCI_SLOT(pdev->devfn);
10778 hw->bus.func = PCI_FUNC(pdev->devfn);
10779 pf->instance = pfs_found;
10781 /* set up the locks for the AQ, do this only once in probe
10782 * and destroy them only once in remove
10784 mutex_init(&hw->aq.asq_mutex);
10785 mutex_init(&hw->aq.arq_mutex);
10788 pf->msg_enable = pf->hw.debug_mask;
10789 pf->msg_enable = debug;
10792 /* do a special CORER for clearing PXE mode once at init */
10793 if (hw->revision_id == 0 &&
10794 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10795 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10800 i40e_clear_pxe_mode(hw);
10803 /* Reset here to make sure all is clean and to define PF 'n' */
10805 err = i40e_pf_reset(hw);
10807 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10812 hw->aq.num_arq_entries = I40E_AQ_LEN;
10813 hw->aq.num_asq_entries = I40E_AQ_LEN;
10814 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10815 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10816 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10818 snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10820 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10822 err = i40e_init_shared_code(hw);
10824 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10829 /* set up a default setting for link flow control */
10830 pf->hw.fc.requested_mode = I40E_FC_NONE;
10832 err = i40e_init_adminq(hw);
10834 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10835 dev_info(&pdev->dev,
10836 "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
10838 dev_info(&pdev->dev,
10839 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10844 /* provide nvm, fw, api versions */
10845 dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10846 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10847 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10848 i40e_nvm_version_str(hw));
10850 if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10851 hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10852 dev_info(&pdev->dev,
10853 "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
10854 else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10855 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10856 dev_info(&pdev->dev,
10857 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10859 i40e_verify_eeprom(pf);
10861 /* Rev 0 hardware was never productized */
10862 if (hw->revision_id < 1)
10863 dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
10865 i40e_clear_pxe_mode(hw);
10866 err = i40e_get_capabilities(pf);
10868 goto err_adminq_setup;
10870 err = i40e_sw_init(pf);
10872 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10876 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10877 hw->func_caps.num_rx_qp,
10878 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10880 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10881 goto err_init_lan_hmc;
10884 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10886 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10888 goto err_configure_lan_hmc;
10891 /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10892 * Ignore error return codes because if it was already disabled via
10893 * hardware settings this will fail
10895 if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10896 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10897 i40e_aq_stop_lldp(hw, true, NULL);
10900 i40e_get_mac_addr(hw, hw->mac.addr);
10901 /* allow a platform config to override the HW addr */
10902 i40e_get_platform_mac_addr(pdev, pf);
10903 if (!is_valid_ether_addr(hw->mac.addr)) {
10904 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10908 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10909 ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10910 i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10911 if (is_valid_ether_addr(hw->mac.port_addr))
10912 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10914 err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10916 dev_info(&pdev->dev,
10917 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10918 if (!is_valid_ether_addr(hw->mac.san_addr)) {
10919 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10921 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10923 dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10924 #endif /* I40E_FCOE */
10926 pci_set_drvdata(pdev, pf);
10927 pci_save_state(pdev);
10928 #ifdef CONFIG_I40E_DCB
10929 err = i40e_init_pf_dcb(pf);
10931 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10932 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE & I40E_FLAG_DCB_ENABLED);
10933 /* Continue without DCB enabled */
10935 #endif /* CONFIG_I40E_DCB */
10937 /* set up periodic task facility */
10938 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10939 pf->service_timer_period = HZ;
10941 INIT_WORK(&pf->service_task, i40e_service_task);
10942 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10943 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10945 /* NVM bit on means WoL disabled for the port */
10946 i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10947 if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10948 pf->wol_en = false;
10951 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10953 /* set up the main switch operations */
10954 i40e_determine_queue_usage(pf);
10955 err = i40e_init_interrupt_scheme(pf);
10957 goto err_switch_setup;
10959 /* The number of VSIs reported by the FW is the minimum guaranteed
10960 * to us; HW supports far more and we share the remaining pool with
10961 * the other PFs. We allocate space for more than the guarantee with
10962 * the understanding that we might not get them all later.
10964 if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10965 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10967 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10969 /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10970 pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
10974 goto err_switch_setup;
10977 #ifdef CONFIG_PCI_IOV
10978 /* prep for VF support */
10979 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10980 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10981 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10982 if (pci_num_vf(pdev))
10983 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10986 err = i40e_setup_pf_switch(pf, false);
10988 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10992 /* Make sure flow control is set according to current settings */
10993 err = i40e_set_fc(hw, &set_fc_aq_fail, true);
10994 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
10995 dev_dbg(&pf->pdev->dev,
10996 "Set fc with err %s aq_err %s on get_phy_cap\n",
10997 i40e_stat_str(hw, err),
10998 i40e_aq_str(hw, hw->aq.asq_last_status));
10999 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11000 dev_dbg(&pf->pdev->dev,
11001 "Set fc with err %s aq_err %s on set_phy_config\n",
11002 i40e_stat_str(hw, err),
11003 i40e_aq_str(hw, hw->aq.asq_last_status));
11004 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11005 dev_dbg(&pf->pdev->dev,
11006 "Set fc with err %s aq_err %s on get_link_info\n",
11007 i40e_stat_str(hw, err),
11008 i40e_aq_str(hw, hw->aq.asq_last_status));
11010 /* if FDIR VSI was set up, start it now */
11011 for (i = 0; i < pf->num_alloc_vsi; i++) {
11012 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11013 i40e_vsi_open(pf->vsi[i]);
11018 /* The driver only wants link up/down and module qualification
11019 * reports from firmware. Note the negative logic.
11021 err = i40e_aq_set_phy_int_mask(&pf->hw,
11022 ~(I40E_AQ_EVENT_LINK_UPDOWN |
11023 I40E_AQ_EVENT_MEDIA_NA |
11024 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11026 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11027 i40e_stat_str(&pf->hw, err),
11028 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11030 /* Reconfigure hardware for allowing smaller MSS in the case
11031 * of TSO, so that we avoid the MDD being fired and causing
11032 * a reset in the case of small MSS+TSO.
11034 val = rd32(hw, I40E_REG_MSS);
11035 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11036 val &= ~I40E_REG_MSS_MIN_MASK;
11037 val |= I40E_64BYTE_MSS;
11038 wr32(hw, I40E_REG_MSS, val);
11041 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11043 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11045 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11046 i40e_stat_str(&pf->hw, err),
11047 i40e_aq_str(&pf->hw,
11048 pf->hw.aq.asq_last_status));
11050 /* The main driver is (mostly) up and happy. We need to set this state
11051 * before setting up the misc vector or we get a race and the vector
11052 * ends up disabled forever.
11054 clear_bit(__I40E_DOWN, &pf->state);
11056 /* In case of MSIX we are going to setup the misc vector right here
11057 * to handle admin queue events etc. In case of legacy and MSI
11058 * the misc functionality and queue processing is combined in
11059 * the same vector and that gets setup at open.
11061 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11062 err = i40e_setup_misc_vector(pf);
11064 dev_info(&pdev->dev,
11065 "setup of misc vector failed: %d\n", err);
11070 #ifdef CONFIG_PCI_IOV
11071 /* prep for VF support */
11072 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11073 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11074 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11075 /* disable link interrupts for VFs */
11076 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11077 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11078 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11081 if (pci_num_vf(pdev)) {
11082 dev_info(&pdev->dev,
11083 "Active VFs found, allocating resources.\n");
11084 err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11086 dev_info(&pdev->dev,
11087 "Error %d allocating resources for existing VFs\n",
11091 #endif /* CONFIG_PCI_IOV */
11093 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11094 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11095 pf->num_iwarp_msix,
11096 I40E_IWARP_IRQ_PILE_ID);
11097 if (pf->iwarp_base_vector < 0) {
11098 dev_info(&pdev->dev,
11099 "failed to get tracking for %d vectors for IWARP err=%d\n",
11100 pf->num_iwarp_msix, pf->iwarp_base_vector);
11101 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11105 i40e_dbg_pf_init(pf);
11107 /* tell the firmware that we're starting */
11108 i40e_send_version(pf);
11110 /* since everything's happy, start the service_task timer */
11111 mod_timer(&pf->service_timer,
11112 round_jiffies(jiffies + pf->service_timer_period));
11114 /* add this PF to client device list and launch a client service task */
11115 err = i40e_lan_add_device(pf);
11117 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11121 /* create FCoE interface */
11122 i40e_fcoe_vsi_setup(pf);
11125 #define PCI_SPEED_SIZE 8
11126 #define PCI_WIDTH_SIZE 8
11127 /* Devices on the IOSF bus do not have this information
11128 * and will report PCI Gen 1 x 1 by default so don't bother
11131 if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11132 char speed[PCI_SPEED_SIZE] = "Unknown";
11133 char width[PCI_WIDTH_SIZE] = "Unknown";
11135 /* Get the negotiated link width and speed from PCI config
11138 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11141 i40e_set_pci_config_data(hw, link_status);
11143 switch (hw->bus.speed) {
11144 case i40e_bus_speed_8000:
11145 strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11146 case i40e_bus_speed_5000:
11147 strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11148 case i40e_bus_speed_2500:
11149 strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11153 switch (hw->bus.width) {
11154 case i40e_bus_width_pcie_x8:
11155 strncpy(width, "8", PCI_WIDTH_SIZE); break;
11156 case i40e_bus_width_pcie_x4:
11157 strncpy(width, "4", PCI_WIDTH_SIZE); break;
11158 case i40e_bus_width_pcie_x2:
11159 strncpy(width, "2", PCI_WIDTH_SIZE); break;
11160 case i40e_bus_width_pcie_x1:
11161 strncpy(width, "1", PCI_WIDTH_SIZE); break;
11166 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11169 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11170 hw->bus.speed < i40e_bus_speed_8000) {
11171 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11172 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11176 /* get the requested speeds from the fw */
11177 err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11179 dev_dbg(&pf->pdev->dev, "get requested speeds ret = %s last_status = %s\n",
11180 i40e_stat_str(&pf->hw, err),
11181 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11182 pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11184 /* get the supported phy types from the fw */
11185 err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11187 dev_dbg(&pf->pdev->dev, "get supported phy types ret = %s last_status = %s\n",
11188 i40e_stat_str(&pf->hw, err),
11189 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11190 pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11192 /* Add a filter to drop all Flow control frames from any VSI from being
11193 * transmitted. By doing so we stop a malicious VF from sending out
11194 * PAUSE or PFC frames and potentially controlling traffic for other
11196 * The FW can still send Flow control frames if enabled.
11198 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11199 pf->main_vsi_seid);
11201 if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11202 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11203 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11205 /* print a string summarizing features */
11206 i40e_print_features(pf);
11210 /* Unwind what we've done if something failed in the setup */
11212 set_bit(__I40E_DOWN, &pf->state);
11213 i40e_clear_interrupt_scheme(pf);
11216 i40e_reset_interrupt_capability(pf);
11217 del_timer_sync(&pf->service_timer);
11219 err_configure_lan_hmc:
11220 (void)i40e_shutdown_lan_hmc(hw);
11222 kfree(pf->qp_pile);
11226 iounmap(hw->hw_addr);
11230 pci_disable_pcie_error_reporting(pdev);
11231 pci_release_mem_regions(pdev);
11234 pci_disable_device(pdev);
11239 * i40e_remove - Device removal routine
11240 * @pdev: PCI device information struct
11242 * i40e_remove is called by the PCI subsystem to alert the driver
11243 * that is should release a PCI device. This could be caused by a
11244 * Hot-Plug event, or because the driver is going to be removed from
11247 static void i40e_remove(struct pci_dev *pdev)
11249 struct i40e_pf *pf = pci_get_drvdata(pdev);
11250 struct i40e_hw *hw = &pf->hw;
11251 i40e_status ret_code;
11254 i40e_dbg_pf_exit(pf);
11258 /* Disable RSS in hw */
11259 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11260 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11262 /* no more scheduling of any task */
11263 set_bit(__I40E_SUSPENDED, &pf->state);
11264 set_bit(__I40E_DOWN, &pf->state);
11265 if (pf->service_timer.data)
11266 del_timer_sync(&pf->service_timer);
11267 if (pf->service_task.func)
11268 cancel_work_sync(&pf->service_task);
11270 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11272 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11275 i40e_fdir_teardown(pf);
11277 /* If there is a switch structure or any orphans, remove them.
11278 * This will leave only the PF's VSI remaining.
11280 for (i = 0; i < I40E_MAX_VEB; i++) {
11284 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11285 pf->veb[i]->uplink_seid == 0)
11286 i40e_switch_branch_release(pf->veb[i]);
11289 /* Now we can shutdown the PF's VSI, just before we kill
11292 if (pf->vsi[pf->lan_vsi])
11293 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11295 /* remove attached clients */
11296 ret_code = i40e_lan_del_device(pf);
11298 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11302 /* shutdown and destroy the HMC */
11303 if (hw->hmc.hmc_obj) {
11304 ret_code = i40e_shutdown_lan_hmc(hw);
11306 dev_warn(&pdev->dev,
11307 "Failed to destroy the HMC resources: %d\n",
11311 /* shutdown the adminq */
11312 ret_code = i40e_shutdown_adminq(hw);
11314 dev_warn(&pdev->dev,
11315 "Failed to destroy the Admin Queue resources: %d\n",
11318 /* destroy the locks only once, here */
11319 mutex_destroy(&hw->aq.arq_mutex);
11320 mutex_destroy(&hw->aq.asq_mutex);
11322 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11323 i40e_clear_interrupt_scheme(pf);
11324 for (i = 0; i < pf->num_alloc_vsi; i++) {
11326 i40e_vsi_clear_rings(pf->vsi[i]);
11327 i40e_vsi_clear(pf->vsi[i]);
11332 for (i = 0; i < I40E_MAX_VEB; i++) {
11337 kfree(pf->qp_pile);
11340 iounmap(hw->hw_addr);
11342 pci_release_mem_regions(pdev);
11344 pci_disable_pcie_error_reporting(pdev);
11345 pci_disable_device(pdev);
11349 * i40e_pci_error_detected - warning that something funky happened in PCI land
11350 * @pdev: PCI device information struct
11352 * Called to warn that something happened and the error handling steps
11353 * are in progress. Allows the driver to quiesce things, be ready for
11356 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11357 enum pci_channel_state error)
11359 struct i40e_pf *pf = pci_get_drvdata(pdev);
11361 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11363 /* shutdown all operations */
11364 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11366 i40e_prep_for_reset(pf);
11370 /* Request a slot reset */
11371 return PCI_ERS_RESULT_NEED_RESET;
11375 * i40e_pci_error_slot_reset - a PCI slot reset just happened
11376 * @pdev: PCI device information struct
11378 * Called to find if the driver can work with the device now that
11379 * the pci slot has been reset. If a basic connection seems good
11380 * (registers are readable and have sane content) then return a
11381 * happy little PCI_ERS_RESULT_xxx.
11383 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11385 struct i40e_pf *pf = pci_get_drvdata(pdev);
11386 pci_ers_result_t result;
11390 dev_dbg(&pdev->dev, "%s\n", __func__);
11391 if (pci_enable_device_mem(pdev)) {
11392 dev_info(&pdev->dev,
11393 "Cannot re-enable PCI device after reset.\n");
11394 result = PCI_ERS_RESULT_DISCONNECT;
11396 pci_set_master(pdev);
11397 pci_restore_state(pdev);
11398 pci_save_state(pdev);
11399 pci_wake_from_d3(pdev, false);
11401 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11403 result = PCI_ERS_RESULT_RECOVERED;
11405 result = PCI_ERS_RESULT_DISCONNECT;
11408 err = pci_cleanup_aer_uncorrect_error_status(pdev);
11410 dev_info(&pdev->dev,
11411 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11413 /* non-fatal, continue */
11420 * i40e_pci_error_resume - restart operations after PCI error recovery
11421 * @pdev: PCI device information struct
11423 * Called to allow the driver to bring things back up after PCI error
11424 * and/or reset recovery has finished.
11426 static void i40e_pci_error_resume(struct pci_dev *pdev)
11428 struct i40e_pf *pf = pci_get_drvdata(pdev);
11430 dev_dbg(&pdev->dev, "%s\n", __func__);
11431 if (test_bit(__I40E_SUSPENDED, &pf->state))
11435 i40e_handle_reset_warning(pf);
11440 * i40e_shutdown - PCI callback for shutting down
11441 * @pdev: PCI device information struct
11443 static void i40e_shutdown(struct pci_dev *pdev)
11445 struct i40e_pf *pf = pci_get_drvdata(pdev);
11446 struct i40e_hw *hw = &pf->hw;
11448 set_bit(__I40E_SUSPENDED, &pf->state);
11449 set_bit(__I40E_DOWN, &pf->state);
11451 i40e_prep_for_reset(pf);
11454 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11455 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11457 del_timer_sync(&pf->service_timer);
11458 cancel_work_sync(&pf->service_task);
11459 i40e_fdir_teardown(pf);
11462 i40e_prep_for_reset(pf);
11465 wr32(hw, I40E_PFPM_APM,
11466 (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11467 wr32(hw, I40E_PFPM_WUFC,
11468 (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11470 i40e_clear_interrupt_scheme(pf);
11472 if (system_state == SYSTEM_POWER_OFF) {
11473 pci_wake_from_d3(pdev, pf->wol_en);
11474 pci_set_power_state(pdev, PCI_D3hot);
11480 * i40e_suspend - PCI callback for moving to D3
11481 * @pdev: PCI device information struct
11483 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11485 struct i40e_pf *pf = pci_get_drvdata(pdev);
11486 struct i40e_hw *hw = &pf->hw;
11489 set_bit(__I40E_SUSPENDED, &pf->state);
11490 set_bit(__I40E_DOWN, &pf->state);
11493 i40e_prep_for_reset(pf);
11496 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11497 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11499 i40e_stop_misc_vector(pf);
11501 retval = pci_save_state(pdev);
11505 pci_wake_from_d3(pdev, pf->wol_en);
11506 pci_set_power_state(pdev, PCI_D3hot);
11512 * i40e_resume - PCI callback for waking up from D3
11513 * @pdev: PCI device information struct
11515 static int i40e_resume(struct pci_dev *pdev)
11517 struct i40e_pf *pf = pci_get_drvdata(pdev);
11520 pci_set_power_state(pdev, PCI_D0);
11521 pci_restore_state(pdev);
11522 /* pci_restore_state() clears dev->state_saves, so
11523 * call pci_save_state() again to restore it.
11525 pci_save_state(pdev);
11527 err = pci_enable_device_mem(pdev);
11529 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11532 pci_set_master(pdev);
11534 /* no wakeup events while running */
11535 pci_wake_from_d3(pdev, false);
11537 /* handling the reset will rebuild the device state */
11538 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11539 clear_bit(__I40E_DOWN, &pf->state);
11541 i40e_reset_and_rebuild(pf, false);
11549 static const struct pci_error_handlers i40e_err_handler = {
11550 .error_detected = i40e_pci_error_detected,
11551 .slot_reset = i40e_pci_error_slot_reset,
11552 .resume = i40e_pci_error_resume,
11555 static struct pci_driver i40e_driver = {
11556 .name = i40e_driver_name,
11557 .id_table = i40e_pci_tbl,
11558 .probe = i40e_probe,
11559 .remove = i40e_remove,
11561 .suspend = i40e_suspend,
11562 .resume = i40e_resume,
11564 .shutdown = i40e_shutdown,
11565 .err_handler = &i40e_err_handler,
11566 .sriov_configure = i40e_pci_sriov_configure,
11570 * i40e_init_module - Driver registration routine
11572 * i40e_init_module is the first routine called when the driver is
11573 * loaded. All it does is register with the PCI subsystem.
11575 static int __init i40e_init_module(void)
11577 pr_info("%s: %s - version %s\n", i40e_driver_name,
11578 i40e_driver_string, i40e_driver_version_str);
11579 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11581 /* we will see if single thread per module is enough for now,
11582 * it can't be any worse than using the system workqueue which
11583 * was already single threaded
11585 i40e_wq = create_singlethread_workqueue(i40e_driver_name);
11587 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11592 return pci_register_driver(&i40e_driver);
11594 module_init(i40e_init_module);
11597 * i40e_exit_module - Driver exit cleanup routine
11599 * i40e_exit_module is called just before the driver is removed
11602 static void __exit i40e_exit_module(void)
11604 pci_unregister_driver(&i40e_driver);
11605 destroy_workqueue(i40e_wq);
11608 module_exit(i40e_exit_module);