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 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5118 pf->flags |= I40E_FLAG_DCB_ENABLED;
5119 dev_dbg(&pf->pdev->dev,
5120 "DCBX offload is supported for this PF.\n");
5123 dev_info(&pf->pdev->dev,
5124 "Query for DCB configuration failed, err %s aq_err %s\n",
5125 i40e_stat_str(&pf->hw, err),
5126 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5132 #endif /* CONFIG_I40E_DCB */
5133 #define SPEED_SIZE 14
5136 * i40e_print_link_message - print link up or down
5137 * @vsi: the VSI for which link needs a message
5139 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5141 char *speed = "Unknown";
5142 char *fc = "Unknown";
5144 if (vsi->current_isup == isup)
5146 vsi->current_isup = isup;
5148 netdev_info(vsi->netdev, "NIC Link is Down\n");
5152 /* Warn user if link speed on NPAR enabled partition is not at
5155 if (vsi->back->hw.func_caps.npar_enable &&
5156 (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5157 vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5158 netdev_warn(vsi->netdev,
5159 "The partition detected link speed that is less than 10Gbps\n");
5161 switch (vsi->back->hw.phy.link_info.link_speed) {
5162 case I40E_LINK_SPEED_40GB:
5165 case I40E_LINK_SPEED_20GB:
5168 case I40E_LINK_SPEED_10GB:
5171 case I40E_LINK_SPEED_1GB:
5174 case I40E_LINK_SPEED_100MB:
5181 switch (vsi->back->hw.fc.current_mode) {
5185 case I40E_FC_TX_PAUSE:
5188 case I40E_FC_RX_PAUSE:
5196 netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5201 * i40e_up_complete - Finish the last steps of bringing up a connection
5202 * @vsi: the VSI being configured
5204 static int i40e_up_complete(struct i40e_vsi *vsi)
5206 struct i40e_pf *pf = vsi->back;
5209 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5210 i40e_vsi_configure_msix(vsi);
5212 i40e_configure_msi_and_legacy(vsi);
5215 err = i40e_vsi_control_rings(vsi, true);
5219 clear_bit(__I40E_DOWN, &vsi->state);
5220 i40e_napi_enable_all(vsi);
5221 i40e_vsi_enable_irq(vsi);
5223 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5225 i40e_print_link_message(vsi, true);
5226 netif_tx_start_all_queues(vsi->netdev);
5227 netif_carrier_on(vsi->netdev);
5228 } else if (vsi->netdev) {
5229 i40e_print_link_message(vsi, false);
5230 /* need to check for qualified module here*/
5231 if ((pf->hw.phy.link_info.link_info &
5232 I40E_AQ_MEDIA_AVAILABLE) &&
5233 (!(pf->hw.phy.link_info.an_info &
5234 I40E_AQ_QUALIFIED_MODULE)))
5235 netdev_err(vsi->netdev,
5236 "the driver failed to link because an unqualified module was detected.");
5239 /* replay FDIR SB filters */
5240 if (vsi->type == I40E_VSI_FDIR) {
5241 /* reset fd counters */
5242 pf->fd_add_err = pf->fd_atr_cnt = 0;
5243 if (pf->fd_tcp_rule > 0) {
5244 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5245 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5246 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5247 pf->fd_tcp_rule = 0;
5249 i40e_fdir_filter_restore(vsi);
5252 /* On the next run of the service_task, notify any clients of the new
5255 pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5256 i40e_service_event_schedule(pf);
5262 * i40e_vsi_reinit_locked - Reset the VSI
5263 * @vsi: the VSI being configured
5265 * Rebuild the ring structs after some configuration
5266 * has changed, e.g. MTU size.
5268 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5270 struct i40e_pf *pf = vsi->back;
5272 WARN_ON(in_interrupt());
5273 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5274 usleep_range(1000, 2000);
5278 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5282 * i40e_up - Bring the connection back up after being down
5283 * @vsi: the VSI being configured
5285 int i40e_up(struct i40e_vsi *vsi)
5289 err = i40e_vsi_configure(vsi);
5291 err = i40e_up_complete(vsi);
5297 * i40e_down - Shutdown the connection processing
5298 * @vsi: the VSI being stopped
5300 void i40e_down(struct i40e_vsi *vsi)
5304 /* It is assumed that the caller of this function
5305 * sets the vsi->state __I40E_DOWN bit.
5308 netif_carrier_off(vsi->netdev);
5309 netif_tx_disable(vsi->netdev);
5311 i40e_vsi_disable_irq(vsi);
5312 i40e_vsi_control_rings(vsi, false);
5313 i40e_napi_disable_all(vsi);
5315 for (i = 0; i < vsi->num_queue_pairs; i++) {
5316 i40e_clean_tx_ring(vsi->tx_rings[i]);
5317 i40e_clean_rx_ring(vsi->rx_rings[i]);
5320 i40e_notify_client_of_netdev_close(vsi, false);
5325 * i40e_setup_tc - configure multiple traffic classes
5326 * @netdev: net device to configure
5327 * @tc: number of traffic classes to enable
5329 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5331 struct i40e_netdev_priv *np = netdev_priv(netdev);
5332 struct i40e_vsi *vsi = np->vsi;
5333 struct i40e_pf *pf = vsi->back;
5338 /* Check if DCB enabled to continue */
5339 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5340 netdev_info(netdev, "DCB is not enabled for adapter\n");
5344 /* Check if MFP enabled */
5345 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5346 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5350 /* Check whether tc count is within enabled limit */
5351 if (tc > i40e_pf_get_num_tc(pf)) {
5352 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5356 /* Generate TC map for number of tc requested */
5357 for (i = 0; i < tc; i++)
5358 enabled_tc |= BIT(i);
5360 /* Requesting same TC configuration as already enabled */
5361 if (enabled_tc == vsi->tc_config.enabled_tc)
5364 /* Quiesce VSI queues */
5365 i40e_quiesce_vsi(vsi);
5367 /* Configure VSI for enabled TCs */
5368 ret = i40e_vsi_config_tc(vsi, enabled_tc);
5370 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5376 i40e_unquiesce_vsi(vsi);
5383 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5384 struct tc_to_netdev *tc)
5386 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5387 struct tc_to_netdev *tc)
5390 if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5392 return i40e_setup_tc(netdev, tc->tc);
5396 * i40e_open - Called when a network interface is made active
5397 * @netdev: network interface device structure
5399 * The open entry point is called when a network interface is made
5400 * active by the system (IFF_UP). At this point all resources needed
5401 * for transmit and receive operations are allocated, the interrupt
5402 * handler is registered with the OS, the netdev watchdog subtask is
5403 * enabled, and the stack is notified that the interface is ready.
5405 * Returns 0 on success, negative value on failure
5407 int i40e_open(struct net_device *netdev)
5409 struct i40e_netdev_priv *np = netdev_priv(netdev);
5410 struct i40e_vsi *vsi = np->vsi;
5411 struct i40e_pf *pf = vsi->back;
5414 /* disallow open during test or if eeprom is broken */
5415 if (test_bit(__I40E_TESTING, &pf->state) ||
5416 test_bit(__I40E_BAD_EEPROM, &pf->state))
5419 netif_carrier_off(netdev);
5421 err = i40e_vsi_open(vsi);
5425 /* configure global TSO hardware offload settings */
5426 wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5427 TCP_FLAG_FIN) >> 16);
5428 wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5430 TCP_FLAG_CWR) >> 16);
5431 wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5433 udp_tunnel_get_rx_info(netdev);
5440 * @vsi: the VSI to open
5442 * Finish initialization of the VSI.
5444 * Returns 0 on success, negative value on failure
5446 int i40e_vsi_open(struct i40e_vsi *vsi)
5448 struct i40e_pf *pf = vsi->back;
5449 char int_name[I40E_INT_NAME_STR_LEN];
5452 /* allocate descriptors */
5453 err = i40e_vsi_setup_tx_resources(vsi);
5456 err = i40e_vsi_setup_rx_resources(vsi);
5460 err = i40e_vsi_configure(vsi);
5465 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5466 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5467 err = i40e_vsi_request_irq(vsi, int_name);
5471 /* Notify the stack of the actual queue counts. */
5472 err = netif_set_real_num_tx_queues(vsi->netdev,
5473 vsi->num_queue_pairs);
5475 goto err_set_queues;
5477 err = netif_set_real_num_rx_queues(vsi->netdev,
5478 vsi->num_queue_pairs);
5480 goto err_set_queues;
5482 } else if (vsi->type == I40E_VSI_FDIR) {
5483 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5484 dev_driver_string(&pf->pdev->dev),
5485 dev_name(&pf->pdev->dev));
5486 err = i40e_vsi_request_irq(vsi, int_name);
5493 err = i40e_up_complete(vsi);
5495 goto err_up_complete;
5502 i40e_vsi_free_irq(vsi);
5504 i40e_vsi_free_rx_resources(vsi);
5506 i40e_vsi_free_tx_resources(vsi);
5507 if (vsi == pf->vsi[pf->lan_vsi])
5508 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5514 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5515 * @pf: Pointer to PF
5517 * This function destroys the hlist where all the Flow Director
5518 * filters were saved.
5520 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5522 struct i40e_fdir_filter *filter;
5523 struct hlist_node *node2;
5525 hlist_for_each_entry_safe(filter, node2,
5526 &pf->fdir_filter_list, fdir_node) {
5527 hlist_del(&filter->fdir_node);
5530 pf->fdir_pf_active_filters = 0;
5534 * i40e_close - Disables a network interface
5535 * @netdev: network interface device structure
5537 * The close entry point is called when an interface is de-activated
5538 * by the OS. The hardware is still under the driver's control, but
5539 * this netdev interface is disabled.
5541 * Returns 0, this is not allowed to fail
5543 int i40e_close(struct net_device *netdev)
5545 struct i40e_netdev_priv *np = netdev_priv(netdev);
5546 struct i40e_vsi *vsi = np->vsi;
5548 i40e_vsi_close(vsi);
5554 * i40e_do_reset - Start a PF or Core Reset sequence
5555 * @pf: board private structure
5556 * @reset_flags: which reset is requested
5558 * The essential difference in resets is that the PF Reset
5559 * doesn't clear the packet buffers, doesn't reset the PE
5560 * firmware, and doesn't bother the other PFs on the chip.
5562 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5566 WARN_ON(in_interrupt());
5569 /* do the biggest reset indicated */
5570 if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5572 /* Request a Global Reset
5574 * This will start the chip's countdown to the actual full
5575 * chip reset event, and a warning interrupt to be sent
5576 * to all PFs, including the requestor. Our handler
5577 * for the warning interrupt will deal with the shutdown
5578 * and recovery of the switch setup.
5580 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5581 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5582 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5583 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5585 } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5587 /* Request a Core Reset
5589 * Same as Global Reset, except does *not* include the MAC/PHY
5591 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5592 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5593 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5594 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5595 i40e_flush(&pf->hw);
5597 } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5599 /* Request a PF Reset
5601 * Resets only the PF-specific registers
5603 * This goes directly to the tear-down and rebuild of
5604 * the switch, since we need to do all the recovery as
5605 * for the Core Reset.
5607 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5608 i40e_handle_reset_warning(pf);
5610 } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5613 /* Find the VSI(s) that requested a re-init */
5614 dev_info(&pf->pdev->dev,
5615 "VSI reinit requested\n");
5616 for (v = 0; v < pf->num_alloc_vsi; v++) {
5617 struct i40e_vsi *vsi = pf->vsi[v];
5620 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5621 i40e_vsi_reinit_locked(pf->vsi[v]);
5622 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5625 } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5628 /* Find the VSI(s) that needs to be brought down */
5629 dev_info(&pf->pdev->dev, "VSI down requested\n");
5630 for (v = 0; v < pf->num_alloc_vsi; v++) {
5631 struct i40e_vsi *vsi = pf->vsi[v];
5634 test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5635 set_bit(__I40E_DOWN, &vsi->state);
5637 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5641 dev_info(&pf->pdev->dev,
5642 "bad reset request 0x%08x\n", reset_flags);
5646 #ifdef CONFIG_I40E_DCB
5648 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5649 * @pf: board private structure
5650 * @old_cfg: current DCB config
5651 * @new_cfg: new DCB config
5653 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5654 struct i40e_dcbx_config *old_cfg,
5655 struct i40e_dcbx_config *new_cfg)
5657 bool need_reconfig = false;
5659 /* Check if ETS configuration has changed */
5660 if (memcmp(&new_cfg->etscfg,
5662 sizeof(new_cfg->etscfg))) {
5663 /* If Priority Table has changed reconfig is needed */
5664 if (memcmp(&new_cfg->etscfg.prioritytable,
5665 &old_cfg->etscfg.prioritytable,
5666 sizeof(new_cfg->etscfg.prioritytable))) {
5667 need_reconfig = true;
5668 dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5671 if (memcmp(&new_cfg->etscfg.tcbwtable,
5672 &old_cfg->etscfg.tcbwtable,
5673 sizeof(new_cfg->etscfg.tcbwtable)))
5674 dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5676 if (memcmp(&new_cfg->etscfg.tsatable,
5677 &old_cfg->etscfg.tsatable,
5678 sizeof(new_cfg->etscfg.tsatable)))
5679 dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5682 /* Check if PFC configuration has changed */
5683 if (memcmp(&new_cfg->pfc,
5685 sizeof(new_cfg->pfc))) {
5686 need_reconfig = true;
5687 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5690 /* Check if APP Table has changed */
5691 if (memcmp(&new_cfg->app,
5693 sizeof(new_cfg->app))) {
5694 need_reconfig = true;
5695 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5698 dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5699 return need_reconfig;
5703 * i40e_handle_lldp_event - Handle LLDP Change MIB event
5704 * @pf: board private structure
5705 * @e: event info posted on ARQ
5707 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5708 struct i40e_arq_event_info *e)
5710 struct i40e_aqc_lldp_get_mib *mib =
5711 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5712 struct i40e_hw *hw = &pf->hw;
5713 struct i40e_dcbx_config tmp_dcbx_cfg;
5714 bool need_reconfig = false;
5718 /* Not DCB capable or capability disabled */
5719 if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5722 /* Ignore if event is not for Nearest Bridge */
5723 type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5724 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5725 dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5726 if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5729 /* Check MIB Type and return if event for Remote MIB update */
5730 type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5731 dev_dbg(&pf->pdev->dev,
5732 "LLDP event mib type %s\n", type ? "remote" : "local");
5733 if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5734 /* Update the remote cached instance and return */
5735 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5736 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5737 &hw->remote_dcbx_config);
5741 /* Store the old configuration */
5742 tmp_dcbx_cfg = hw->local_dcbx_config;
5744 /* Reset the old DCBx configuration data */
5745 memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5746 /* Get updated DCBX data from firmware */
5747 ret = i40e_get_dcb_config(&pf->hw);
5749 dev_info(&pf->pdev->dev,
5750 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5751 i40e_stat_str(&pf->hw, ret),
5752 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5756 /* No change detected in DCBX configs */
5757 if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5758 sizeof(tmp_dcbx_cfg))) {
5759 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5763 need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5764 &hw->local_dcbx_config);
5766 i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5771 /* Enable DCB tagging only when more than one TC */
5772 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5773 pf->flags |= I40E_FLAG_DCB_ENABLED;
5775 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5777 set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5778 /* Reconfiguration needed quiesce all VSIs */
5779 i40e_pf_quiesce_all_vsi(pf);
5781 /* Changes in configuration update VEB/VSI */
5782 i40e_dcb_reconfigure(pf);
5784 ret = i40e_resume_port_tx(pf);
5786 clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5787 /* In case of error no point in resuming VSIs */
5791 /* Wait for the PF's queues to be disabled */
5792 ret = i40e_pf_wait_queues_disabled(pf);
5794 /* Schedule PF reset to recover */
5795 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5796 i40e_service_event_schedule(pf);
5798 i40e_pf_unquiesce_all_vsi(pf);
5799 /* Notify the client for the DCB changes */
5800 i40e_notify_client_of_l2_param_changes(pf->vsi[pf->lan_vsi]);
5806 #endif /* CONFIG_I40E_DCB */
5809 * i40e_do_reset_safe - Protected reset path for userland calls.
5810 * @pf: board private structure
5811 * @reset_flags: which reset is requested
5814 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5817 i40e_do_reset(pf, reset_flags);
5822 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5823 * @pf: board private structure
5824 * @e: event info posted on ARQ
5826 * Handler for LAN Queue Overflow Event generated by the firmware for PF
5829 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5830 struct i40e_arq_event_info *e)
5832 struct i40e_aqc_lan_overflow *data =
5833 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5834 u32 queue = le32_to_cpu(data->prtdcb_rupto);
5835 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5836 struct i40e_hw *hw = &pf->hw;
5840 dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5843 /* Queue belongs to VF, find the VF and issue VF reset */
5844 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5845 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5846 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5847 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5848 vf_id -= hw->func_caps.vf_base_id;
5849 vf = &pf->vf[vf_id];
5850 i40e_vc_notify_vf_reset(vf);
5851 /* Allow VF to process pending reset notification */
5853 i40e_reset_vf(vf, false);
5858 * i40e_service_event_complete - Finish up the service event
5859 * @pf: board private structure
5861 static void i40e_service_event_complete(struct i40e_pf *pf)
5863 WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5865 /* flush memory to make sure state is correct before next watchog */
5866 smp_mb__before_atomic();
5867 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5871 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5872 * @pf: board private structure
5874 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5878 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5879 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5884 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5885 * @pf: board private structure
5887 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5891 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5892 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5893 ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5894 I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5899 * i40e_get_global_fd_count - Get total FD filters programmed on device
5900 * @pf: board private structure
5902 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5906 val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5907 fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5908 ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5909 I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5914 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5915 * @pf: board private structure
5917 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5919 struct i40e_fdir_filter *filter;
5920 u32 fcnt_prog, fcnt_avail;
5921 struct hlist_node *node;
5923 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5926 /* Check if, FD SB or ATR was auto disabled and if there is enough room
5929 fcnt_prog = i40e_get_global_fd_count(pf);
5930 fcnt_avail = pf->fdir_pf_filter_count;
5931 if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5932 (pf->fd_add_err == 0) ||
5933 (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5934 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5935 (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5936 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5937 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5938 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5941 /* Wait for some more space to be available to turn on ATR */
5942 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5943 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5944 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5945 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5946 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5947 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5951 /* if hw had a problem adding a filter, delete it */
5952 if (pf->fd_inv > 0) {
5953 hlist_for_each_entry_safe(filter, node,
5954 &pf->fdir_filter_list, fdir_node) {
5955 if (filter->fd_id == pf->fd_inv) {
5956 hlist_del(&filter->fdir_node);
5958 pf->fdir_pf_active_filters--;
5964 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5965 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5967 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5968 * @pf: board private structure
5970 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5972 unsigned long min_flush_time;
5973 int flush_wait_retry = 50;
5974 bool disable_atr = false;
5978 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5981 if (!time_after(jiffies, pf->fd_flush_timestamp +
5982 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5985 /* If the flush is happening too quick and we have mostly SB rules we
5986 * should not re-enable ATR for some time.
5988 min_flush_time = pf->fd_flush_timestamp +
5989 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5990 fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5992 if (!(time_after(jiffies, min_flush_time)) &&
5993 (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5994 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5995 dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5999 pf->fd_flush_timestamp = jiffies;
6000 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6001 /* flush all filters */
6002 wr32(&pf->hw, I40E_PFQF_CTL_1,
6003 I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6004 i40e_flush(&pf->hw);
6008 /* Check FD flush status every 5-6msec */
6009 usleep_range(5000, 6000);
6010 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6011 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6013 } while (flush_wait_retry--);
6014 if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6015 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6017 /* replay sideband filters */
6018 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6020 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
6021 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6022 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6023 dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6028 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6029 * @pf: board private structure
6031 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6033 return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6036 /* We can see up to 256 filter programming desc in transit if the filters are
6037 * being applied really fast; before we see the first
6038 * filter miss error on Rx queue 0. Accumulating enough error messages before
6039 * reacting will make sure we don't cause flush too often.
6041 #define I40E_MAX_FD_PROGRAM_ERROR 256
6044 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6045 * @pf: board private structure
6047 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6050 /* if interface is down do nothing */
6051 if (test_bit(__I40E_DOWN, &pf->state))
6054 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
6057 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6058 i40e_fdir_flush_and_replay(pf);
6060 i40e_fdir_check_and_reenable(pf);
6065 * i40e_vsi_link_event - notify VSI of a link event
6066 * @vsi: vsi to be notified
6067 * @link_up: link up or down
6069 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6071 if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6074 switch (vsi->type) {
6079 if (!vsi->netdev || !vsi->netdev_registered)
6083 netif_carrier_on(vsi->netdev);
6084 netif_tx_wake_all_queues(vsi->netdev);
6086 netif_carrier_off(vsi->netdev);
6087 netif_tx_stop_all_queues(vsi->netdev);
6091 case I40E_VSI_SRIOV:
6092 case I40E_VSI_VMDQ2:
6094 case I40E_VSI_IWARP:
6095 case I40E_VSI_MIRROR:
6097 /* there is no notification for other VSIs */
6103 * i40e_veb_link_event - notify elements on the veb of a link event
6104 * @veb: veb to be notified
6105 * @link_up: link up or down
6107 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6112 if (!veb || !veb->pf)
6116 /* depth first... */
6117 for (i = 0; i < I40E_MAX_VEB; i++)
6118 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6119 i40e_veb_link_event(pf->veb[i], link_up);
6121 /* ... now the local VSIs */
6122 for (i = 0; i < pf->num_alloc_vsi; i++)
6123 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6124 i40e_vsi_link_event(pf->vsi[i], link_up);
6128 * i40e_link_event - Update netif_carrier status
6129 * @pf: board private structure
6131 static void i40e_link_event(struct i40e_pf *pf)
6133 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6134 u8 new_link_speed, old_link_speed;
6136 bool new_link, old_link;
6138 /* save off old link status information */
6139 pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6141 /* set this to force the get_link_status call to refresh state */
6142 pf->hw.phy.get_link_info = true;
6144 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6146 status = i40e_get_link_status(&pf->hw, &new_link);
6148 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6153 old_link_speed = pf->hw.phy.link_info_old.link_speed;
6154 new_link_speed = pf->hw.phy.link_info.link_speed;
6156 if (new_link == old_link &&
6157 new_link_speed == old_link_speed &&
6158 (test_bit(__I40E_DOWN, &vsi->state) ||
6159 new_link == netif_carrier_ok(vsi->netdev)))
6162 if (!test_bit(__I40E_DOWN, &vsi->state))
6163 i40e_print_link_message(vsi, new_link);
6165 /* Notify the base of the switch tree connected to
6166 * the link. Floating VEBs are not notified.
6168 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6169 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6171 i40e_vsi_link_event(vsi, new_link);
6174 i40e_vc_notify_link_state(pf);
6176 if (pf->flags & I40E_FLAG_PTP)
6177 i40e_ptp_set_increment(pf);
6181 * i40e_watchdog_subtask - periodic checks not using event driven response
6182 * @pf: board private structure
6184 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6188 /* if interface is down do nothing */
6189 if (test_bit(__I40E_DOWN, &pf->state) ||
6190 test_bit(__I40E_CONFIG_BUSY, &pf->state))
6193 /* make sure we don't do these things too often */
6194 if (time_before(jiffies, (pf->service_timer_previous +
6195 pf->service_timer_period)))
6197 pf->service_timer_previous = jiffies;
6199 if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6200 i40e_link_event(pf);
6202 /* Update the stats for active netdevs so the network stack
6203 * can look at updated numbers whenever it cares to
6205 for (i = 0; i < pf->num_alloc_vsi; i++)
6206 if (pf->vsi[i] && pf->vsi[i]->netdev)
6207 i40e_update_stats(pf->vsi[i]);
6209 if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6210 /* Update the stats for the active switching components */
6211 for (i = 0; i < I40E_MAX_VEB; i++)
6213 i40e_update_veb_stats(pf->veb[i]);
6216 i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6220 * i40e_reset_subtask - Set up for resetting the device and driver
6221 * @pf: board private structure
6223 static void i40e_reset_subtask(struct i40e_pf *pf)
6225 u32 reset_flags = 0;
6228 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6229 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6230 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6232 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6233 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6234 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6236 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6237 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6238 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6240 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6241 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6242 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6244 if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6245 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6246 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6249 /* If there's a recovery already waiting, it takes
6250 * precedence before starting a new reset sequence.
6252 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6253 i40e_handle_reset_warning(pf);
6257 /* If we're already down or resetting, just bail */
6259 !test_bit(__I40E_DOWN, &pf->state) &&
6260 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6261 i40e_do_reset(pf, reset_flags);
6268 * i40e_handle_link_event - Handle link event
6269 * @pf: board private structure
6270 * @e: event info posted on ARQ
6272 static void i40e_handle_link_event(struct i40e_pf *pf,
6273 struct i40e_arq_event_info *e)
6275 struct i40e_aqc_get_link_status *status =
6276 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6278 /* Do a new status request to re-enable LSE reporting
6279 * and load new status information into the hw struct
6280 * This completely ignores any state information
6281 * in the ARQ event info, instead choosing to always
6282 * issue the AQ update link status command.
6284 i40e_link_event(pf);
6286 /* check for unqualified module, if link is down */
6287 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6288 (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6289 (!(status->link_info & I40E_AQ_LINK_UP)))
6290 dev_err(&pf->pdev->dev,
6291 "The driver failed to link because an unqualified module was detected.\n");
6295 * i40e_clean_adminq_subtask - Clean the AdminQ rings
6296 * @pf: board private structure
6298 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6300 struct i40e_arq_event_info event;
6301 struct i40e_hw *hw = &pf->hw;
6308 /* Do not run clean AQ when PF reset fails */
6309 if (test_bit(__I40E_RESET_FAILED, &pf->state))
6312 /* check for error indications */
6313 val = rd32(&pf->hw, pf->hw.aq.arq.len);
6315 if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6316 if (hw->debug_mask & I40E_DEBUG_AQ)
6317 dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6318 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6320 if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6321 if (hw->debug_mask & I40E_DEBUG_AQ)
6322 dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6323 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6324 pf->arq_overflows++;
6326 if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6327 if (hw->debug_mask & I40E_DEBUG_AQ)
6328 dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6329 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6332 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6334 val = rd32(&pf->hw, pf->hw.aq.asq.len);
6336 if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6337 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6338 dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6339 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6341 if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6342 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6343 dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6344 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6346 if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6347 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6348 dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6349 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6352 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6354 event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6355 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6360 ret = i40e_clean_arq_element(hw, &event, &pending);
6361 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6364 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6368 opcode = le16_to_cpu(event.desc.opcode);
6371 case i40e_aqc_opc_get_link_status:
6372 i40e_handle_link_event(pf, &event);
6374 case i40e_aqc_opc_send_msg_to_pf:
6375 ret = i40e_vc_process_vf_msg(pf,
6376 le16_to_cpu(event.desc.retval),
6377 le32_to_cpu(event.desc.cookie_high),
6378 le32_to_cpu(event.desc.cookie_low),
6382 case i40e_aqc_opc_lldp_update_mib:
6383 dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6384 #ifdef CONFIG_I40E_DCB
6386 ret = i40e_handle_lldp_event(pf, &event);
6388 #endif /* CONFIG_I40E_DCB */
6390 case i40e_aqc_opc_event_lan_overflow:
6391 dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6392 i40e_handle_lan_overflow_event(pf, &event);
6394 case i40e_aqc_opc_send_msg_to_peer:
6395 dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6397 case i40e_aqc_opc_nvm_erase:
6398 case i40e_aqc_opc_nvm_update:
6399 case i40e_aqc_opc_oem_post_update:
6400 i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6401 "ARQ NVM operation 0x%04x completed\n",
6405 dev_info(&pf->pdev->dev,
6406 "ARQ: Unknown event 0x%04x ignored\n",
6410 } while (pending && (i++ < pf->adminq_work_limit));
6412 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6413 /* re-enable Admin queue interrupt cause */
6414 val = rd32(hw, I40E_PFINT_ICR0_ENA);
6415 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6416 wr32(hw, I40E_PFINT_ICR0_ENA, val);
6419 kfree(event.msg_buf);
6423 * i40e_verify_eeprom - make sure eeprom is good to use
6424 * @pf: board private structure
6426 static void i40e_verify_eeprom(struct i40e_pf *pf)
6430 err = i40e_diag_eeprom_test(&pf->hw);
6432 /* retry in case of garbage read */
6433 err = i40e_diag_eeprom_test(&pf->hw);
6435 dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6437 set_bit(__I40E_BAD_EEPROM, &pf->state);
6441 if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6442 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6443 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6448 * i40e_enable_pf_switch_lb
6449 * @pf: pointer to the PF structure
6451 * enable switch loop back or die - no point in a return value
6453 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6455 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6456 struct i40e_vsi_context ctxt;
6459 ctxt.seid = pf->main_vsi_seid;
6460 ctxt.pf_num = pf->hw.pf_id;
6462 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6464 dev_info(&pf->pdev->dev,
6465 "couldn't get PF vsi config, err %s aq_err %s\n",
6466 i40e_stat_str(&pf->hw, ret),
6467 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6470 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6471 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6472 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6474 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6476 dev_info(&pf->pdev->dev,
6477 "update vsi switch failed, err %s aq_err %s\n",
6478 i40e_stat_str(&pf->hw, ret),
6479 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6484 * i40e_disable_pf_switch_lb
6485 * @pf: pointer to the PF structure
6487 * disable switch loop back or die - no point in a return value
6489 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6491 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6492 struct i40e_vsi_context ctxt;
6495 ctxt.seid = pf->main_vsi_seid;
6496 ctxt.pf_num = pf->hw.pf_id;
6498 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6500 dev_info(&pf->pdev->dev,
6501 "couldn't get PF vsi config, err %s aq_err %s\n",
6502 i40e_stat_str(&pf->hw, ret),
6503 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6506 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6507 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6508 ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6510 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6512 dev_info(&pf->pdev->dev,
6513 "update vsi switch failed, err %s aq_err %s\n",
6514 i40e_stat_str(&pf->hw, ret),
6515 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6520 * i40e_config_bridge_mode - Configure the HW bridge mode
6521 * @veb: pointer to the bridge instance
6523 * Configure the loop back mode for the LAN VSI that is downlink to the
6524 * specified HW bridge instance. It is expected this function is called
6525 * when a new HW bridge is instantiated.
6527 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6529 struct i40e_pf *pf = veb->pf;
6531 if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6532 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6533 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6534 if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6535 i40e_disable_pf_switch_lb(pf);
6537 i40e_enable_pf_switch_lb(pf);
6541 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6542 * @veb: pointer to the VEB instance
6544 * This is a recursive function that first builds the attached VSIs then
6545 * recurses in to build the next layer of VEB. We track the connections
6546 * through our own index numbers because the seid's from the HW could
6547 * change across the reset.
6549 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6551 struct i40e_vsi *ctl_vsi = NULL;
6552 struct i40e_pf *pf = veb->pf;
6556 /* build VSI that owns this VEB, temporarily attached to base VEB */
6557 for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6559 pf->vsi[v]->veb_idx == veb->idx &&
6560 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6561 ctl_vsi = pf->vsi[v];
6566 dev_info(&pf->pdev->dev,
6567 "missing owner VSI for veb_idx %d\n", veb->idx);
6569 goto end_reconstitute;
6571 if (ctl_vsi != pf->vsi[pf->lan_vsi])
6572 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6573 ret = i40e_add_vsi(ctl_vsi);
6575 dev_info(&pf->pdev->dev,
6576 "rebuild of veb_idx %d owner VSI failed: %d\n",
6578 goto end_reconstitute;
6580 i40e_vsi_reset_stats(ctl_vsi);
6582 /* create the VEB in the switch and move the VSI onto the VEB */
6583 ret = i40e_add_veb(veb, ctl_vsi);
6585 goto end_reconstitute;
6587 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6588 veb->bridge_mode = BRIDGE_MODE_VEB;
6590 veb->bridge_mode = BRIDGE_MODE_VEPA;
6591 i40e_config_bridge_mode(veb);
6593 /* create the remaining VSIs attached to this VEB */
6594 for (v = 0; v < pf->num_alloc_vsi; v++) {
6595 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6598 if (pf->vsi[v]->veb_idx == veb->idx) {
6599 struct i40e_vsi *vsi = pf->vsi[v];
6601 vsi->uplink_seid = veb->seid;
6602 ret = i40e_add_vsi(vsi);
6604 dev_info(&pf->pdev->dev,
6605 "rebuild of vsi_idx %d failed: %d\n",
6607 goto end_reconstitute;
6609 i40e_vsi_reset_stats(vsi);
6613 /* create any VEBs attached to this VEB - RECURSION */
6614 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6615 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6616 pf->veb[veb_idx]->uplink_seid = veb->seid;
6617 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6628 * i40e_get_capabilities - get info about the HW
6629 * @pf: the PF struct
6631 static int i40e_get_capabilities(struct i40e_pf *pf)
6633 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6638 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6640 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6644 /* this loads the data into the hw struct for us */
6645 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6647 i40e_aqc_opc_list_func_capabilities,
6649 /* data loaded, buffer no longer needed */
6652 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6653 /* retry with a larger buffer */
6654 buf_len = data_size;
6655 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6656 dev_info(&pf->pdev->dev,
6657 "capability discovery failed, err %s aq_err %s\n",
6658 i40e_stat_str(&pf->hw, err),
6659 i40e_aq_str(&pf->hw,
6660 pf->hw.aq.asq_last_status));
6665 if (pf->hw.debug_mask & I40E_DEBUG_USER)
6666 dev_info(&pf->pdev->dev,
6667 "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",
6668 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6669 pf->hw.func_caps.num_msix_vectors,
6670 pf->hw.func_caps.num_msix_vectors_vf,
6671 pf->hw.func_caps.fd_filters_guaranteed,
6672 pf->hw.func_caps.fd_filters_best_effort,
6673 pf->hw.func_caps.num_tx_qp,
6674 pf->hw.func_caps.num_vsis);
6676 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6677 + pf->hw.func_caps.num_vfs)
6678 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6679 dev_info(&pf->pdev->dev,
6680 "got num_vsis %d, setting num_vsis to %d\n",
6681 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6682 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6688 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6691 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6692 * @pf: board private structure
6694 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6696 struct i40e_vsi *vsi;
6699 /* quick workaround for an NVM issue that leaves a critical register
6702 if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6703 static const u32 hkey[] = {
6704 0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6705 0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6706 0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6709 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6710 wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6713 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6716 /* find existing VSI and see if it needs configuring */
6718 for (i = 0; i < pf->num_alloc_vsi; i++) {
6719 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6725 /* create a new VSI if none exists */
6727 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6728 pf->vsi[pf->lan_vsi]->seid, 0);
6730 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6731 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6736 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6740 * i40e_fdir_teardown - release the Flow Director resources
6741 * @pf: board private structure
6743 static void i40e_fdir_teardown(struct i40e_pf *pf)
6747 i40e_fdir_filter_exit(pf);
6748 for (i = 0; i < pf->num_alloc_vsi; i++) {
6749 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6750 i40e_vsi_release(pf->vsi[i]);
6757 * i40e_prep_for_reset - prep for the core to reset
6758 * @pf: board private structure
6760 * Close up the VFs and other things in prep for PF Reset.
6762 static void i40e_prep_for_reset(struct i40e_pf *pf)
6764 struct i40e_hw *hw = &pf->hw;
6765 i40e_status ret = 0;
6768 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6769 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6771 if (i40e_check_asq_alive(&pf->hw))
6772 i40e_vc_notify_reset(pf);
6774 dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6776 /* quiesce the VSIs and their queues that are not already DOWN */
6777 i40e_pf_quiesce_all_vsi(pf);
6779 for (v = 0; v < pf->num_alloc_vsi; v++) {
6781 pf->vsi[v]->seid = 0;
6784 i40e_shutdown_adminq(&pf->hw);
6786 /* call shutdown HMC */
6787 if (hw->hmc.hmc_obj) {
6788 ret = i40e_shutdown_lan_hmc(hw);
6790 dev_warn(&pf->pdev->dev,
6791 "shutdown_lan_hmc failed: %d\n", ret);
6796 * i40e_send_version - update firmware with driver version
6799 static void i40e_send_version(struct i40e_pf *pf)
6801 struct i40e_driver_version dv;
6803 dv.major_version = DRV_VERSION_MAJOR;
6804 dv.minor_version = DRV_VERSION_MINOR;
6805 dv.build_version = DRV_VERSION_BUILD;
6806 dv.subbuild_version = 0;
6807 strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6808 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6812 * i40e_reset_and_rebuild - reset and rebuild using a saved config
6813 * @pf: board private structure
6814 * @reinit: if the Main VSI needs to re-initialized.
6816 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6818 struct i40e_hw *hw = &pf->hw;
6819 u8 set_fc_aq_fail = 0;
6824 /* Now we wait for GRST to settle out.
6825 * We don't have to delete the VEBs or VSIs from the hw switch
6826 * because the reset will make them disappear.
6828 ret = i40e_pf_reset(hw);
6830 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6831 set_bit(__I40E_RESET_FAILED, &pf->state);
6832 goto clear_recovery;
6836 if (test_bit(__I40E_DOWN, &pf->state))
6837 goto clear_recovery;
6838 dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6840 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6841 ret = i40e_init_adminq(&pf->hw);
6843 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6844 i40e_stat_str(&pf->hw, ret),
6845 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6846 goto clear_recovery;
6849 /* re-verify the eeprom if we just had an EMP reset */
6850 if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6851 i40e_verify_eeprom(pf);
6853 i40e_clear_pxe_mode(hw);
6854 ret = i40e_get_capabilities(pf);
6856 goto end_core_reset;
6858 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6859 hw->func_caps.num_rx_qp,
6860 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6862 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6863 goto end_core_reset;
6865 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6867 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6868 goto end_core_reset;
6871 #ifdef CONFIG_I40E_DCB
6872 ret = i40e_init_pf_dcb(pf);
6874 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6875 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6876 /* Continue without DCB enabled */
6878 #endif /* CONFIG_I40E_DCB */
6880 i40e_init_pf_fcoe(pf);
6883 /* do basic switch setup */
6884 ret = i40e_setup_pf_switch(pf, reinit);
6886 goto end_core_reset;
6888 /* The driver only wants link up/down and module qualification
6889 * reports from firmware. Note the negative logic.
6891 ret = i40e_aq_set_phy_int_mask(&pf->hw,
6892 ~(I40E_AQ_EVENT_LINK_UPDOWN |
6893 I40E_AQ_EVENT_MEDIA_NA |
6894 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6896 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6897 i40e_stat_str(&pf->hw, ret),
6898 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6900 /* make sure our flow control settings are restored */
6901 ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6903 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6904 i40e_stat_str(&pf->hw, ret),
6905 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6907 /* Rebuild the VSIs and VEBs that existed before reset.
6908 * They are still in our local switch element arrays, so only
6909 * need to rebuild the switch model in the HW.
6911 * If there were VEBs but the reconstitution failed, we'll try
6912 * try to recover minimal use by getting the basic PF VSI working.
6914 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6915 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6916 /* find the one VEB connected to the MAC, and find orphans */
6917 for (v = 0; v < I40E_MAX_VEB; v++) {
6921 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6922 pf->veb[v]->uplink_seid == 0) {
6923 ret = i40e_reconstitute_veb(pf->veb[v]);
6928 /* If Main VEB failed, we're in deep doodoo,
6929 * so give up rebuilding the switch and set up
6930 * for minimal rebuild of PF VSI.
6931 * If orphan failed, we'll report the error
6932 * but try to keep going.
6934 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6935 dev_info(&pf->pdev->dev,
6936 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6938 pf->vsi[pf->lan_vsi]->uplink_seid
6941 } else if (pf->veb[v]->uplink_seid == 0) {
6942 dev_info(&pf->pdev->dev,
6943 "rebuild of orphan VEB failed: %d\n",
6950 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6951 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6952 /* no VEB, so rebuild only the Main VSI */
6953 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6955 dev_info(&pf->pdev->dev,
6956 "rebuild of Main VSI failed: %d\n", ret);
6957 goto end_core_reset;
6961 /* Reconfigure hardware for allowing smaller MSS in the case
6962 * of TSO, so that we avoid the MDD being fired and causing
6963 * a reset in the case of small MSS+TSO.
6965 #define I40E_REG_MSS 0x000E64DC
6966 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6967 #define I40E_64BYTE_MSS 0x400000
6968 val = rd32(hw, I40E_REG_MSS);
6969 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6970 val &= ~I40E_REG_MSS_MIN_MASK;
6971 val |= I40E_64BYTE_MSS;
6972 wr32(hw, I40E_REG_MSS, val);
6975 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6977 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6979 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6980 i40e_stat_str(&pf->hw, ret),
6981 i40e_aq_str(&pf->hw,
6982 pf->hw.aq.asq_last_status));
6984 /* reinit the misc interrupt */
6985 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6986 ret = i40e_setup_misc_vector(pf);
6988 /* Add a filter to drop all Flow control frames from any VSI from being
6989 * transmitted. By doing so we stop a malicious VF from sending out
6990 * PAUSE or PFC frames and potentially controlling traffic for other
6992 * The FW can still send Flow control frames if enabled.
6994 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6997 /* restart the VSIs that were rebuilt and running before the reset */
6998 i40e_pf_unquiesce_all_vsi(pf);
7000 if (pf->num_alloc_vfs) {
7001 for (v = 0; v < pf->num_alloc_vfs; v++)
7002 i40e_reset_vf(&pf->vf[v], true);
7005 /* tell the firmware that we're starting */
7006 i40e_send_version(pf);
7009 clear_bit(__I40E_RESET_FAILED, &pf->state);
7011 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
7015 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7016 * @pf: board private structure
7018 * Close up the VFs and other things in prep for a Core Reset,
7019 * then get ready to rebuild the world.
7021 static void i40e_handle_reset_warning(struct i40e_pf *pf)
7023 i40e_prep_for_reset(pf);
7024 i40e_reset_and_rebuild(pf, false);
7028 * i40e_handle_mdd_event
7029 * @pf: pointer to the PF structure
7031 * Called from the MDD irq handler to identify possibly malicious vfs
7033 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7035 struct i40e_hw *hw = &pf->hw;
7036 bool mdd_detected = false;
7037 bool pf_mdd_detected = false;
7042 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7045 /* find what triggered the MDD event */
7046 reg = rd32(hw, I40E_GL_MDET_TX);
7047 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7048 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7049 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7050 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7051 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7052 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7053 I40E_GL_MDET_TX_EVENT_SHIFT;
7054 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7055 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7056 pf->hw.func_caps.base_queue;
7057 if (netif_msg_tx_err(pf))
7058 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7059 event, queue, pf_num, vf_num);
7060 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7061 mdd_detected = true;
7063 reg = rd32(hw, I40E_GL_MDET_RX);
7064 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7065 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7066 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7067 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7068 I40E_GL_MDET_RX_EVENT_SHIFT;
7069 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7070 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7071 pf->hw.func_caps.base_queue;
7072 if (netif_msg_rx_err(pf))
7073 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7074 event, queue, func);
7075 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7076 mdd_detected = true;
7080 reg = rd32(hw, I40E_PF_MDET_TX);
7081 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7082 wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7083 dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7084 pf_mdd_detected = true;
7086 reg = rd32(hw, I40E_PF_MDET_RX);
7087 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7088 wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7089 dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7090 pf_mdd_detected = true;
7092 /* Queue belongs to the PF, initiate a reset */
7093 if (pf_mdd_detected) {
7094 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7095 i40e_service_event_schedule(pf);
7099 /* see if one of the VFs needs its hand slapped */
7100 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7102 reg = rd32(hw, I40E_VP_MDET_TX(i));
7103 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7104 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7105 vf->num_mdd_events++;
7106 dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7110 reg = rd32(hw, I40E_VP_MDET_RX(i));
7111 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7112 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7113 vf->num_mdd_events++;
7114 dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7118 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7119 dev_info(&pf->pdev->dev,
7120 "Too many MDD events on VF %d, disabled\n", i);
7121 dev_info(&pf->pdev->dev,
7122 "Use PF Control I/F to re-enable the VF\n");
7123 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7127 /* re-enable mdd interrupt cause */
7128 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7129 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7130 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7131 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7136 * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7137 * @pf: board private structure
7139 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7141 struct i40e_hw *hw = &pf->hw;
7146 if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7149 pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7151 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7152 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7153 pf->pending_udp_bitmap &= ~BIT_ULL(i);
7154 port = pf->udp_ports[i].index;
7156 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7157 pf->udp_ports[i].type,
7160 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7163 dev_dbg(&pf->pdev->dev,
7164 "%s %s port %d, index %d failed, err %s aq_err %s\n",
7165 pf->udp_ports[i].type ? "vxlan" : "geneve",
7166 port ? "add" : "delete",
7168 i40e_stat_str(&pf->hw, ret),
7169 i40e_aq_str(&pf->hw,
7170 pf->hw.aq.asq_last_status));
7171 pf->udp_ports[i].index = 0;
7178 * i40e_service_task - Run the driver's async subtasks
7179 * @work: pointer to work_struct containing our data
7181 static void i40e_service_task(struct work_struct *work)
7183 struct i40e_pf *pf = container_of(work,
7186 unsigned long start_time = jiffies;
7188 /* don't bother with service tasks if a reset is in progress */
7189 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7190 i40e_service_event_complete(pf);
7194 i40e_detect_recover_hung(pf);
7195 i40e_sync_filters_subtask(pf);
7196 i40e_reset_subtask(pf);
7197 i40e_handle_mdd_event(pf);
7198 i40e_vc_process_vflr_event(pf);
7199 i40e_watchdog_subtask(pf);
7200 i40e_fdir_reinit_subtask(pf);
7201 i40e_client_subtask(pf);
7202 i40e_sync_filters_subtask(pf);
7203 i40e_sync_udp_filters_subtask(pf);
7204 i40e_clean_adminq_subtask(pf);
7206 i40e_service_event_complete(pf);
7208 /* If the tasks have taken longer than one timer cycle or there
7209 * is more work to be done, reschedule the service task now
7210 * rather than wait for the timer to tick again.
7212 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7213 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
7214 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
7215 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7216 i40e_service_event_schedule(pf);
7220 * i40e_service_timer - timer callback
7221 * @data: pointer to PF struct
7223 static void i40e_service_timer(unsigned long data)
7225 struct i40e_pf *pf = (struct i40e_pf *)data;
7227 mod_timer(&pf->service_timer,
7228 round_jiffies(jiffies + pf->service_timer_period));
7229 i40e_service_event_schedule(pf);
7233 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7234 * @vsi: the VSI being configured
7236 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7238 struct i40e_pf *pf = vsi->back;
7240 switch (vsi->type) {
7242 vsi->alloc_queue_pairs = pf->num_lan_qps;
7243 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7244 I40E_REQ_DESCRIPTOR_MULTIPLE);
7245 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7246 vsi->num_q_vectors = pf->num_lan_msix;
7248 vsi->num_q_vectors = 1;
7253 vsi->alloc_queue_pairs = 1;
7254 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7255 I40E_REQ_DESCRIPTOR_MULTIPLE);
7256 vsi->num_q_vectors = pf->num_fdsb_msix;
7259 case I40E_VSI_VMDQ2:
7260 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7261 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7262 I40E_REQ_DESCRIPTOR_MULTIPLE);
7263 vsi->num_q_vectors = pf->num_vmdq_msix;
7266 case I40E_VSI_SRIOV:
7267 vsi->alloc_queue_pairs = pf->num_vf_qps;
7268 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7269 I40E_REQ_DESCRIPTOR_MULTIPLE);
7274 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7275 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7276 I40E_REQ_DESCRIPTOR_MULTIPLE);
7277 vsi->num_q_vectors = pf->num_fcoe_msix;
7280 #endif /* I40E_FCOE */
7290 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7291 * @type: VSI pointer
7292 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7294 * On error: returns error code (negative)
7295 * On success: returns 0
7297 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7302 /* allocate memory for both Tx and Rx ring pointers */
7303 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7304 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7307 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7309 if (alloc_qvectors) {
7310 /* allocate memory for q_vector pointers */
7311 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7312 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7313 if (!vsi->q_vectors) {
7321 kfree(vsi->tx_rings);
7326 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7327 * @pf: board private structure
7328 * @type: type of VSI
7330 * On error: returns error code (negative)
7331 * On success: returns vsi index in PF (positive)
7333 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7336 struct i40e_vsi *vsi;
7340 /* Need to protect the allocation of the VSIs at the PF level */
7341 mutex_lock(&pf->switch_mutex);
7343 /* VSI list may be fragmented if VSI creation/destruction has
7344 * been happening. We can afford to do a quick scan to look
7345 * for any free VSIs in the list.
7347 * find next empty vsi slot, looping back around if necessary
7350 while (i < pf->num_alloc_vsi && pf->vsi[i])
7352 if (i >= pf->num_alloc_vsi) {
7354 while (i < pf->next_vsi && pf->vsi[i])
7358 if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7359 vsi_idx = i; /* Found one! */
7362 goto unlock_pf; /* out of VSI slots! */
7366 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7373 set_bit(__I40E_DOWN, &vsi->state);
7376 vsi->int_rate_limit = 0;
7377 vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7378 pf->rss_table_size : 64;
7379 vsi->netdev_registered = false;
7380 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7381 INIT_LIST_HEAD(&vsi->mac_filter_list);
7382 vsi->irqs_ready = false;
7384 ret = i40e_set_num_rings_in_vsi(vsi);
7388 ret = i40e_vsi_alloc_arrays(vsi, true);
7392 /* Setup default MSIX irq handler for VSI */
7393 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7395 /* Initialize VSI lock */
7396 spin_lock_init(&vsi->mac_filter_list_lock);
7397 pf->vsi[vsi_idx] = vsi;
7402 pf->next_vsi = i - 1;
7405 mutex_unlock(&pf->switch_mutex);
7410 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7411 * @type: VSI pointer
7412 * @free_qvectors: a bool to specify if q_vectors need to be freed.
7414 * On error: returns error code (negative)
7415 * On success: returns 0
7417 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7419 /* free the ring and vector containers */
7420 if (free_qvectors) {
7421 kfree(vsi->q_vectors);
7422 vsi->q_vectors = NULL;
7424 kfree(vsi->tx_rings);
7425 vsi->tx_rings = NULL;
7426 vsi->rx_rings = NULL;
7430 * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7432 * @vsi: Pointer to VSI structure
7434 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7439 kfree(vsi->rss_hkey_user);
7440 vsi->rss_hkey_user = NULL;
7442 kfree(vsi->rss_lut_user);
7443 vsi->rss_lut_user = NULL;
7447 * i40e_vsi_clear - Deallocate the VSI provided
7448 * @vsi: the VSI being un-configured
7450 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7461 mutex_lock(&pf->switch_mutex);
7462 if (!pf->vsi[vsi->idx]) {
7463 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7464 vsi->idx, vsi->idx, vsi, vsi->type);
7468 if (pf->vsi[vsi->idx] != vsi) {
7469 dev_err(&pf->pdev->dev,
7470 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7471 pf->vsi[vsi->idx]->idx,
7473 pf->vsi[vsi->idx]->type,
7474 vsi->idx, vsi, vsi->type);
7478 /* updates the PF for this cleared vsi */
7479 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7480 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7482 i40e_vsi_free_arrays(vsi, true);
7483 i40e_clear_rss_config_user(vsi);
7485 pf->vsi[vsi->idx] = NULL;
7486 if (vsi->idx < pf->next_vsi)
7487 pf->next_vsi = vsi->idx;
7490 mutex_unlock(&pf->switch_mutex);
7498 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7499 * @vsi: the VSI being cleaned
7501 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7505 if (vsi->tx_rings && vsi->tx_rings[0]) {
7506 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7507 kfree_rcu(vsi->tx_rings[i], rcu);
7508 vsi->tx_rings[i] = NULL;
7509 vsi->rx_rings[i] = NULL;
7515 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7516 * @vsi: the VSI being configured
7518 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7520 struct i40e_ring *tx_ring, *rx_ring;
7521 struct i40e_pf *pf = vsi->back;
7524 /* Set basic values in the rings to be used later during open() */
7525 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7526 /* allocate space for both Tx and Rx in one shot */
7527 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7531 tx_ring->queue_index = i;
7532 tx_ring->reg_idx = vsi->base_queue + i;
7533 tx_ring->ring_active = false;
7535 tx_ring->netdev = vsi->netdev;
7536 tx_ring->dev = &pf->pdev->dev;
7537 tx_ring->count = vsi->num_desc;
7539 tx_ring->dcb_tc = 0;
7540 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7541 tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7542 tx_ring->tx_itr_setting = pf->tx_itr_default;
7543 vsi->tx_rings[i] = tx_ring;
7545 rx_ring = &tx_ring[1];
7546 rx_ring->queue_index = i;
7547 rx_ring->reg_idx = vsi->base_queue + i;
7548 rx_ring->ring_active = false;
7550 rx_ring->netdev = vsi->netdev;
7551 rx_ring->dev = &pf->pdev->dev;
7552 rx_ring->count = vsi->num_desc;
7554 rx_ring->dcb_tc = 0;
7555 rx_ring->rx_itr_setting = pf->rx_itr_default;
7556 vsi->rx_rings[i] = rx_ring;
7562 i40e_vsi_clear_rings(vsi);
7567 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7568 * @pf: board private structure
7569 * @vectors: the number of MSI-X vectors to request
7571 * Returns the number of vectors reserved, or error
7573 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7575 vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7576 I40E_MIN_MSIX, vectors);
7578 dev_info(&pf->pdev->dev,
7579 "MSI-X vector reservation failed: %d\n", vectors);
7587 * i40e_init_msix - Setup the MSIX capability
7588 * @pf: board private structure
7590 * Work with the OS to set up the MSIX vectors needed.
7592 * Returns the number of vectors reserved or negative on failure
7594 static int i40e_init_msix(struct i40e_pf *pf)
7596 struct i40e_hw *hw = &pf->hw;
7600 int iwarp_requested = 0;
7602 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7605 /* The number of vectors we'll request will be comprised of:
7606 * - Add 1 for "other" cause for Admin Queue events, etc.
7607 * - The number of LAN queue pairs
7608 * - Queues being used for RSS.
7609 * We don't need as many as max_rss_size vectors.
7610 * use rss_size instead in the calculation since that
7611 * is governed by number of cpus in the system.
7612 * - assumes symmetric Tx/Rx pairing
7613 * - The number of VMDq pairs
7614 * - The CPU count within the NUMA node if iWARP is enabled
7616 * - The number of FCOE qps.
7618 * Once we count this up, try the request.
7620 * If we can't get what we want, we'll simplify to nearly nothing
7621 * and try again. If that still fails, we punt.
7623 vectors_left = hw->func_caps.num_msix_vectors;
7626 /* reserve one vector for miscellaneous handler */
7632 /* reserve vectors for the main PF traffic queues */
7633 pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7634 vectors_left -= pf->num_lan_msix;
7635 v_budget += pf->num_lan_msix;
7637 /* reserve one vector for sideband flow director */
7638 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7640 pf->num_fdsb_msix = 1;
7644 pf->num_fdsb_msix = 0;
7645 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7650 /* can we reserve enough for FCoE? */
7651 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7653 pf->num_fcoe_msix = 0;
7654 else if (vectors_left >= pf->num_fcoe_qps)
7655 pf->num_fcoe_msix = pf->num_fcoe_qps;
7657 pf->num_fcoe_msix = 1;
7658 v_budget += pf->num_fcoe_msix;
7659 vectors_left -= pf->num_fcoe_msix;
7663 /* can we reserve enough for iWARP? */
7664 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7666 pf->num_iwarp_msix = 0;
7667 else if (vectors_left < pf->num_iwarp_msix)
7668 pf->num_iwarp_msix = 1;
7669 v_budget += pf->num_iwarp_msix;
7670 vectors_left -= pf->num_iwarp_msix;
7673 /* any vectors left over go for VMDq support */
7674 if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7675 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7676 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7678 /* if we're short on vectors for what's desired, we limit
7679 * the queues per vmdq. If this is still more than are
7680 * available, the user will need to change the number of
7681 * queues/vectors used by the PF later with the ethtool
7684 if (vmdq_vecs < vmdq_vecs_wanted)
7685 pf->num_vmdq_qps = 1;
7686 pf->num_vmdq_msix = pf->num_vmdq_qps;
7688 v_budget += vmdq_vecs;
7689 vectors_left -= vmdq_vecs;
7692 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7694 if (!pf->msix_entries)
7697 for (i = 0; i < v_budget; i++)
7698 pf->msix_entries[i].entry = i;
7699 v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7701 if (v_actual != v_budget) {
7702 /* If we have limited resources, we will start with no vectors
7703 * for the special features and then allocate vectors to some
7704 * of these features based on the policy and at the end disable
7705 * the features that did not get any vectors.
7707 iwarp_requested = pf->num_iwarp_msix;
7708 pf->num_iwarp_msix = 0;
7710 pf->num_fcoe_qps = 0;
7711 pf->num_fcoe_msix = 0;
7713 pf->num_vmdq_msix = 0;
7716 if (v_actual < I40E_MIN_MSIX) {
7717 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7718 kfree(pf->msix_entries);
7719 pf->msix_entries = NULL;
7722 } else if (v_actual == I40E_MIN_MSIX) {
7723 /* Adjust for minimal MSIX use */
7724 pf->num_vmdq_vsis = 0;
7725 pf->num_vmdq_qps = 0;
7726 pf->num_lan_qps = 1;
7727 pf->num_lan_msix = 1;
7729 } else if (v_actual != v_budget) {
7732 /* reserve the misc vector */
7735 /* Scale vector usage down */
7736 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
7737 pf->num_vmdq_vsis = 1;
7738 pf->num_vmdq_qps = 1;
7739 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7741 /* partition out the remaining vectors */
7744 pf->num_lan_msix = 1;
7747 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7748 pf->num_lan_msix = 1;
7749 pf->num_iwarp_msix = 1;
7751 pf->num_lan_msix = 2;
7754 /* give one vector to FCoE */
7755 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7756 pf->num_lan_msix = 1;
7757 pf->num_fcoe_msix = 1;
7762 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7763 pf->num_iwarp_msix = min_t(int, (vec / 3),
7765 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7766 I40E_DEFAULT_NUM_VMDQ_VSI);
7768 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7769 I40E_DEFAULT_NUM_VMDQ_VSI);
7771 pf->num_lan_msix = min_t(int,
7772 (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7775 /* give one vector to FCoE */
7776 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7777 pf->num_fcoe_msix = 1;
7785 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7786 (pf->num_vmdq_msix == 0)) {
7787 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7788 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7791 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7792 (pf->num_iwarp_msix == 0)) {
7793 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7794 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7798 if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7799 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7800 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7807 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7808 * @vsi: the VSI being configured
7809 * @v_idx: index of the vector in the vsi struct
7810 * @cpu: cpu to be used on affinity_mask
7812 * We allocate one q_vector. If allocation fails we return -ENOMEM.
7814 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7816 struct i40e_q_vector *q_vector;
7818 /* allocate q_vector */
7819 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7823 q_vector->vsi = vsi;
7824 q_vector->v_idx = v_idx;
7825 cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7828 netif_napi_add(vsi->netdev, &q_vector->napi,
7829 i40e_napi_poll, NAPI_POLL_WEIGHT);
7831 q_vector->rx.latency_range = I40E_LOW_LATENCY;
7832 q_vector->tx.latency_range = I40E_LOW_LATENCY;
7834 /* tie q_vector and vsi together */
7835 vsi->q_vectors[v_idx] = q_vector;
7841 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7842 * @vsi: the VSI being configured
7844 * We allocate one q_vector per queue interrupt. If allocation fails we
7847 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7849 struct i40e_pf *pf = vsi->back;
7850 int err, v_idx, num_q_vectors, current_cpu;
7852 /* if not MSIX, give the one vector only to the LAN VSI */
7853 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7854 num_q_vectors = vsi->num_q_vectors;
7855 else if (vsi == pf->vsi[pf->lan_vsi])
7860 current_cpu = cpumask_first(cpu_online_mask);
7862 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7863 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7866 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7867 if (unlikely(current_cpu >= nr_cpu_ids))
7868 current_cpu = cpumask_first(cpu_online_mask);
7875 i40e_free_q_vector(vsi, v_idx);
7881 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7882 * @pf: board private structure to initialize
7884 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7889 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7890 vectors = i40e_init_msix(pf);
7892 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
7893 I40E_FLAG_IWARP_ENABLED |
7895 I40E_FLAG_FCOE_ENABLED |
7897 I40E_FLAG_RSS_ENABLED |
7898 I40E_FLAG_DCB_CAPABLE |
7899 I40E_FLAG_SRIOV_ENABLED |
7900 I40E_FLAG_FD_SB_ENABLED |
7901 I40E_FLAG_FD_ATR_ENABLED |
7902 I40E_FLAG_VMDQ_ENABLED);
7904 /* rework the queue expectations without MSIX */
7905 i40e_determine_queue_usage(pf);
7909 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7910 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7911 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7912 vectors = pci_enable_msi(pf->pdev);
7914 dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7916 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7918 vectors = 1; /* one MSI or Legacy vector */
7921 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7922 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7924 /* set up vector assignment tracking */
7925 size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7926 pf->irq_pile = kzalloc(size, GFP_KERNEL);
7927 if (!pf->irq_pile) {
7928 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7931 pf->irq_pile->num_entries = vectors;
7932 pf->irq_pile->search_hint = 0;
7934 /* track first vector for misc interrupts, ignore return */
7935 (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7941 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7942 * @pf: board private structure
7944 * This sets up the handler for MSIX 0, which is used to manage the
7945 * non-queue interrupts, e.g. AdminQ and errors. This is not used
7946 * when in MSI or Legacy interrupt mode.
7948 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7950 struct i40e_hw *hw = &pf->hw;
7953 /* Only request the irq if this is the first time through, and
7954 * not when we're rebuilding after a Reset
7956 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7957 err = request_irq(pf->msix_entries[0].vector,
7958 i40e_intr, 0, pf->int_name, pf);
7960 dev_info(&pf->pdev->dev,
7961 "request_irq for %s failed: %d\n",
7967 i40e_enable_misc_int_causes(pf);
7969 /* associate no queues to the misc vector */
7970 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7971 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7975 i40e_irq_dynamic_enable_icr0(pf, true);
7981 * i40e_config_rss_aq - Prepare for RSS using AQ commands
7982 * @vsi: vsi structure
7983 * @seed: RSS hash seed
7985 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7986 u8 *lut, u16 lut_size)
7988 struct i40e_aqc_get_set_rss_key_data rss_key;
7989 struct i40e_pf *pf = vsi->back;
7990 struct i40e_hw *hw = &pf->hw;
7991 bool pf_lut = false;
7995 memcpy(&rss_key, seed, sizeof(rss_key));
7997 rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
8001 /* Populate the LUT with max no. of queues in round robin fashion */
8002 for (i = 0; i < vsi->rss_table_size; i++)
8003 rss_lut[i] = i % vsi->rss_size;
8005 ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
8007 dev_info(&pf->pdev->dev,
8008 "Cannot set RSS key, err %s aq_err %s\n",
8009 i40e_stat_str(&pf->hw, ret),
8010 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8011 goto config_rss_aq_out;
8014 if (vsi->type == I40E_VSI_MAIN)
8017 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
8018 vsi->rss_table_size);
8020 dev_info(&pf->pdev->dev,
8021 "Cannot set RSS lut, err %s aq_err %s\n",
8022 i40e_stat_str(&pf->hw, ret),
8023 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8031 * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8032 * @vsi: VSI structure
8034 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8036 u8 seed[I40E_HKEY_ARRAY_SIZE];
8037 struct i40e_pf *pf = vsi->back;
8041 if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8044 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8048 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8049 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8050 vsi->rss_size = min_t(int, pf->alloc_rss_size, vsi->num_queue_pairs);
8051 ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8058 * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8059 * @vsi: Pointer to vsi structure
8060 * @seed: Buffter to store the hash keys
8061 * @lut: Buffer to store the lookup table entries
8062 * @lut_size: Size of buffer to store the lookup table entries
8064 * Return 0 on success, negative on failure
8066 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8067 u8 *lut, u16 lut_size)
8069 struct i40e_pf *pf = vsi->back;
8070 struct i40e_hw *hw = &pf->hw;
8074 ret = i40e_aq_get_rss_key(hw, vsi->id,
8075 (struct i40e_aqc_get_set_rss_key_data *)seed);
8077 dev_info(&pf->pdev->dev,
8078 "Cannot get RSS key, err %s aq_err %s\n",
8079 i40e_stat_str(&pf->hw, ret),
8080 i40e_aq_str(&pf->hw,
8081 pf->hw.aq.asq_last_status));
8087 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8089 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8091 dev_info(&pf->pdev->dev,
8092 "Cannot get RSS lut, err %s aq_err %s\n",
8093 i40e_stat_str(&pf->hw, ret),
8094 i40e_aq_str(&pf->hw,
8095 pf->hw.aq.asq_last_status));
8104 * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8105 * @vsi: Pointer to vsi structure
8106 * @seed: RSS hash seed
8107 * @lut: Lookup table
8108 * @lut_size: Lookup table size
8110 * Returns 0 on success, negative on failure
8112 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8113 const u8 *lut, u16 lut_size)
8115 struct i40e_pf *pf = vsi->back;
8116 struct i40e_hw *hw = &pf->hw;
8117 u16 vf_id = vsi->vf_id;
8120 /* Fill out hash function seed */
8122 u32 *seed_dw = (u32 *)seed;
8124 if (vsi->type == I40E_VSI_MAIN) {
8125 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8126 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8128 } else if (vsi->type == I40E_VSI_SRIOV) {
8129 for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8130 i40e_write_rx_ctl(hw,
8131 I40E_VFQF_HKEY1(i, vf_id),
8134 dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8139 u32 *lut_dw = (u32 *)lut;
8141 if (vsi->type == I40E_VSI_MAIN) {
8142 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8144 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8145 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8146 } else if (vsi->type == I40E_VSI_SRIOV) {
8147 if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8149 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8150 i40e_write_rx_ctl(hw,
8151 I40E_VFQF_HLUT1(i, vf_id),
8154 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8163 * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8164 * @vsi: Pointer to VSI structure
8165 * @seed: Buffer to store the keys
8166 * @lut: Buffer to store the lookup table entries
8167 * @lut_size: Size of buffer to store the lookup table entries
8169 * Returns 0 on success, negative on failure
8171 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8172 u8 *lut, u16 lut_size)
8174 struct i40e_pf *pf = vsi->back;
8175 struct i40e_hw *hw = &pf->hw;
8179 u32 *seed_dw = (u32 *)seed;
8181 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8182 seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8185 u32 *lut_dw = (u32 *)lut;
8187 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8189 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8190 lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8197 * i40e_config_rss - Configure RSS keys and lut
8198 * @vsi: Pointer to VSI structure
8199 * @seed: RSS hash seed
8200 * @lut: Lookup table
8201 * @lut_size: Lookup table size
8203 * Returns 0 on success, negative on failure
8205 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8207 struct i40e_pf *pf = vsi->back;
8209 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8210 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8212 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8216 * i40e_get_rss - Get RSS keys and lut
8217 * @vsi: Pointer to VSI structure
8218 * @seed: Buffer to store the keys
8219 * @lut: Buffer to store the lookup table entries
8220 * lut_size: Size of buffer to store the lookup table entries
8222 * Returns 0 on success, negative on failure
8224 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8226 struct i40e_pf *pf = vsi->back;
8228 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8229 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8231 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8235 * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8236 * @pf: Pointer to board private structure
8237 * @lut: Lookup table
8238 * @rss_table_size: Lookup table size
8239 * @rss_size: Range of queue number for hashing
8241 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8242 u16 rss_table_size, u16 rss_size)
8246 for (i = 0; i < rss_table_size; i++)
8247 lut[i] = i % rss_size;
8251 * i40e_pf_config_rss - Prepare for RSS if used
8252 * @pf: board private structure
8254 static int i40e_pf_config_rss(struct i40e_pf *pf)
8256 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8257 u8 seed[I40E_HKEY_ARRAY_SIZE];
8259 struct i40e_hw *hw = &pf->hw;
8264 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8265 hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8266 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8267 hena |= i40e_pf_get_default_rss_hena(pf);
8269 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8270 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8272 /* Determine the RSS table size based on the hardware capabilities */
8273 reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8274 reg_val = (pf->rss_table_size == 512) ?
8275 (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8276 (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8277 i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8279 /* Determine the RSS size of the VSI */
8281 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8282 vsi->num_queue_pairs);
8284 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8288 /* Use user configured lut if there is one, otherwise use default */
8289 if (vsi->rss_lut_user)
8290 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8292 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8294 /* Use user configured hash key if there is one, otherwise
8297 if (vsi->rss_hkey_user)
8298 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8300 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8301 ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8308 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8309 * @pf: board private structure
8310 * @queue_count: the requested queue count for rss.
8312 * returns 0 if rss is not enabled, if enabled returns the final rss queue
8313 * count which may be different from the requested queue count.
8315 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8317 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8320 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8323 new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8325 if (queue_count != vsi->num_queue_pairs) {
8326 vsi->req_queue_pairs = queue_count;
8327 i40e_prep_for_reset(pf);
8329 pf->alloc_rss_size = new_rss_size;
8331 i40e_reset_and_rebuild(pf, true);
8333 /* Discard the user configured hash keys and lut, if less
8334 * queues are enabled.
8336 if (queue_count < vsi->rss_size) {
8337 i40e_clear_rss_config_user(vsi);
8338 dev_dbg(&pf->pdev->dev,
8339 "discard user configured hash keys and lut\n");
8342 /* Reset vsi->rss_size, as number of enabled queues changed */
8343 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8344 vsi->num_queue_pairs);
8346 i40e_pf_config_rss(pf);
8348 dev_info(&pf->pdev->dev, "RSS count/HW max RSS count: %d/%d\n",
8349 pf->alloc_rss_size, pf->rss_size_max);
8350 return pf->alloc_rss_size;
8354 * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8355 * @pf: board private structure
8357 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8360 bool min_valid, max_valid;
8363 status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8364 &min_valid, &max_valid);
8368 pf->npar_min_bw = min_bw;
8370 pf->npar_max_bw = max_bw;
8377 * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8378 * @pf: board private structure
8380 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8382 struct i40e_aqc_configure_partition_bw_data bw_data;
8385 /* Set the valid bit for this PF */
8386 bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8387 bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8388 bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8390 /* Set the new bandwidths */
8391 status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8397 * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8398 * @pf: board private structure
8400 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8402 /* Commit temporary BW setting to permanent NVM image */
8403 enum i40e_admin_queue_err last_aq_status;
8407 if (pf->hw.partition_id != 1) {
8408 dev_info(&pf->pdev->dev,
8409 "Commit BW only works on partition 1! This is partition %d",
8410 pf->hw.partition_id);
8411 ret = I40E_NOT_SUPPORTED;
8415 /* Acquire NVM for read access */
8416 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8417 last_aq_status = pf->hw.aq.asq_last_status;
8419 dev_info(&pf->pdev->dev,
8420 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8421 i40e_stat_str(&pf->hw, ret),
8422 i40e_aq_str(&pf->hw, last_aq_status));
8426 /* Read word 0x10 of NVM - SW compatibility word 1 */
8427 ret = i40e_aq_read_nvm(&pf->hw,
8428 I40E_SR_NVM_CONTROL_WORD,
8429 0x10, sizeof(nvm_word), &nvm_word,
8431 /* Save off last admin queue command status before releasing
8434 last_aq_status = pf->hw.aq.asq_last_status;
8435 i40e_release_nvm(&pf->hw);
8437 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8438 i40e_stat_str(&pf->hw, ret),
8439 i40e_aq_str(&pf->hw, last_aq_status));
8443 /* Wait a bit for NVM release to complete */
8446 /* Acquire NVM for write access */
8447 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8448 last_aq_status = pf->hw.aq.asq_last_status;
8450 dev_info(&pf->pdev->dev,
8451 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8452 i40e_stat_str(&pf->hw, ret),
8453 i40e_aq_str(&pf->hw, last_aq_status));
8456 /* Write it back out unchanged to initiate update NVM,
8457 * which will force a write of the shadow (alt) RAM to
8458 * the NVM - thus storing the bandwidth values permanently.
8460 ret = i40e_aq_update_nvm(&pf->hw,
8461 I40E_SR_NVM_CONTROL_WORD,
8462 0x10, sizeof(nvm_word),
8463 &nvm_word, true, NULL);
8464 /* Save off last admin queue command status before releasing
8467 last_aq_status = pf->hw.aq.asq_last_status;
8468 i40e_release_nvm(&pf->hw);
8470 dev_info(&pf->pdev->dev,
8471 "BW settings NOT SAVED, err %s aq_err %s\n",
8472 i40e_stat_str(&pf->hw, ret),
8473 i40e_aq_str(&pf->hw, last_aq_status));
8480 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8481 * @pf: board private structure to initialize
8483 * i40e_sw_init initializes the Adapter private data structure.
8484 * Fields are initialized based on PCI device information and
8485 * OS network device settings (MTU size).
8487 static int i40e_sw_init(struct i40e_pf *pf)
8492 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8493 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8494 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8495 if (I40E_DEBUG_USER & debug)
8496 pf->hw.debug_mask = debug;
8497 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8498 I40E_DEFAULT_MSG_ENABLE);
8501 /* Set default capability flags */
8502 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8503 I40E_FLAG_MSI_ENABLED |
8504 I40E_FLAG_MSIX_ENABLED;
8506 /* Set default ITR */
8507 pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8508 pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8510 /* Depending on PF configurations, it is possible that the RSS
8511 * maximum might end up larger than the available queues
8513 pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8514 pf->alloc_rss_size = 1;
8515 pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8516 pf->rss_size_max = min_t(int, pf->rss_size_max,
8517 pf->hw.func_caps.num_tx_qp);
8518 if (pf->hw.func_caps.rss) {
8519 pf->flags |= I40E_FLAG_RSS_ENABLED;
8520 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8524 /* MFP mode enabled */
8525 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8526 pf->flags |= I40E_FLAG_MFP_ENABLED;
8527 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8528 if (i40e_get_npar_bw_setting(pf))
8529 dev_warn(&pf->pdev->dev,
8530 "Could not get NPAR bw settings\n");
8532 dev_info(&pf->pdev->dev,
8533 "Min BW = %8.8x, Max BW = %8.8x\n",
8534 pf->npar_min_bw, pf->npar_max_bw);
8537 /* FW/NVM is not yet fixed in this regard */
8538 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8539 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8540 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8541 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8542 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8543 pf->hw.num_partitions > 1)
8544 dev_info(&pf->pdev->dev,
8545 "Flow Director Sideband mode Disabled in MFP mode\n");
8547 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8548 pf->fdir_pf_filter_count =
8549 pf->hw.func_caps.fd_filters_guaranteed;
8550 pf->hw.fdir_shared_filter_count =
8551 pf->hw.func_caps.fd_filters_best_effort;
8554 if (i40e_is_mac_710(&pf->hw) &&
8555 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8556 (pf->hw.aq.fw_maj_ver < 4))) {
8557 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8558 /* No DCB support for FW < v4.33 */
8559 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8562 /* Disable FW LLDP if FW < v4.3 */
8563 if (i40e_is_mac_710(&pf->hw) &&
8564 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8565 (pf->hw.aq.fw_maj_ver < 4)))
8566 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8568 /* Use the FW Set LLDP MIB API if FW > v4.40 */
8569 if (i40e_is_mac_710(&pf->hw) &&
8570 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8571 (pf->hw.aq.fw_maj_ver >= 5)))
8572 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8574 if (pf->hw.func_caps.vmdq) {
8575 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8576 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8577 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8580 if (pf->hw.func_caps.iwarp) {
8581 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8582 /* IWARP needs one extra vector for CQP just like MISC.*/
8583 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8587 i40e_init_pf_fcoe(pf);
8589 #endif /* I40E_FCOE */
8590 #ifdef CONFIG_PCI_IOV
8591 if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8592 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8593 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8594 pf->num_req_vfs = min_t(int,
8595 pf->hw.func_caps.num_vfs,
8598 #endif /* CONFIG_PCI_IOV */
8599 if (pf->hw.mac.type == I40E_MAC_X722) {
8600 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8601 I40E_FLAG_128_QP_RSS_CAPABLE |
8602 I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8603 I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8604 I40E_FLAG_WB_ON_ITR_CAPABLE |
8605 I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8606 I40E_FLAG_NO_PCI_LINK_CHECK |
8607 I40E_FLAG_100M_SGMII_CAPABLE |
8608 I40E_FLAG_USE_SET_LLDP_MIB |
8609 I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8610 } else if ((pf->hw.aq.api_maj_ver > 1) ||
8611 ((pf->hw.aq.api_maj_ver == 1) &&
8612 (pf->hw.aq.api_min_ver > 4))) {
8613 /* Supported in FW API version higher than 1.4 */
8614 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8615 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8617 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8620 pf->eeprom_version = 0xDEAD;
8621 pf->lan_veb = I40E_NO_VEB;
8622 pf->lan_vsi = I40E_NO_VSI;
8624 /* By default FW has this off for performance reasons */
8625 pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8627 /* set up queue assignment tracking */
8628 size = sizeof(struct i40e_lump_tracking)
8629 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8630 pf->qp_pile = kzalloc(size, GFP_KERNEL);
8635 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8636 pf->qp_pile->search_hint = 0;
8638 pf->tx_timeout_recovery_level = 1;
8640 mutex_init(&pf->switch_mutex);
8642 /* If NPAR is enabled nudge the Tx scheduler */
8643 if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8644 i40e_set_npar_bw_setting(pf);
8651 * i40e_set_ntuple - set the ntuple feature flag and take action
8652 * @pf: board private structure to initialize
8653 * @features: the feature set that the stack is suggesting
8655 * returns a bool to indicate if reset needs to happen
8657 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8659 bool need_reset = false;
8661 /* Check if Flow Director n-tuple support was enabled or disabled. If
8662 * the state changed, we need to reset.
8664 if (features & NETIF_F_NTUPLE) {
8665 /* Enable filters and mark for reset */
8666 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8668 /* enable FD_SB only if there is MSI-X vector */
8669 if (pf->num_fdsb_msix > 0)
8670 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8672 /* turn off filters, mark for reset and clear SW filter list */
8673 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8675 i40e_fdir_filter_exit(pf);
8677 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8678 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8679 /* reset fd counters */
8680 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8681 pf->fdir_pf_active_filters = 0;
8682 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8683 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8684 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8685 /* if ATR was auto disabled it can be re-enabled. */
8686 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8687 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8688 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8694 * i40e_set_features - set the netdev feature flags
8695 * @netdev: ptr to the netdev being adjusted
8696 * @features: the feature set that the stack is suggesting
8698 static int i40e_set_features(struct net_device *netdev,
8699 netdev_features_t features)
8701 struct i40e_netdev_priv *np = netdev_priv(netdev);
8702 struct i40e_vsi *vsi = np->vsi;
8703 struct i40e_pf *pf = vsi->back;
8706 if (features & NETIF_F_HW_VLAN_CTAG_RX)
8707 i40e_vlan_stripping_enable(vsi);
8709 i40e_vlan_stripping_disable(vsi);
8711 need_reset = i40e_set_ntuple(pf, features);
8714 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8720 * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8721 * @pf: board private structure
8722 * @port: The UDP port to look up
8724 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8726 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8730 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8731 if (pf->udp_ports[i].index == port)
8739 * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8740 * @netdev: This physical port's netdev
8741 * @ti: Tunnel endpoint information
8743 static void i40e_udp_tunnel_add(struct net_device *netdev,
8744 struct udp_tunnel_info *ti)
8746 struct i40e_netdev_priv *np = netdev_priv(netdev);
8747 struct i40e_vsi *vsi = np->vsi;
8748 struct i40e_pf *pf = vsi->back;
8749 __be16 port = ti->port;
8753 idx = i40e_get_udp_port_idx(pf, port);
8755 /* Check if port already exists */
8756 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8757 netdev_info(netdev, "port %d already offloaded\n",
8762 /* Now check if there is space to add the new port */
8763 next_idx = i40e_get_udp_port_idx(pf, 0);
8765 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8766 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8772 case UDP_TUNNEL_TYPE_VXLAN:
8773 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8775 case UDP_TUNNEL_TYPE_GENEVE:
8776 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8778 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8784 /* New port: add it and mark its index in the bitmap */
8785 pf->udp_ports[next_idx].index = port;
8786 pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8787 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8791 * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
8792 * @netdev: This physical port's netdev
8793 * @ti: Tunnel endpoint information
8795 static void i40e_udp_tunnel_del(struct net_device *netdev,
8796 struct udp_tunnel_info *ti)
8798 struct i40e_netdev_priv *np = netdev_priv(netdev);
8799 struct i40e_vsi *vsi = np->vsi;
8800 struct i40e_pf *pf = vsi->back;
8801 __be16 port = ti->port;
8804 idx = i40e_get_udp_port_idx(pf, port);
8806 /* Check if port already exists */
8807 if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
8811 case UDP_TUNNEL_TYPE_VXLAN:
8812 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
8815 case UDP_TUNNEL_TYPE_GENEVE:
8816 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
8823 /* if port exists, set it to 0 (mark for deletion)
8824 * and make it pending
8826 pf->udp_ports[idx].index = 0;
8827 pf->pending_udp_bitmap |= BIT_ULL(idx);
8828 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8832 netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
8836 static int i40e_get_phys_port_id(struct net_device *netdev,
8837 struct netdev_phys_item_id *ppid)
8839 struct i40e_netdev_priv *np = netdev_priv(netdev);
8840 struct i40e_pf *pf = np->vsi->back;
8841 struct i40e_hw *hw = &pf->hw;
8843 if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8846 ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8847 memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8853 * i40e_ndo_fdb_add - add an entry to the hardware database
8854 * @ndm: the input from the stack
8855 * @tb: pointer to array of nladdr (unused)
8856 * @dev: the net device pointer
8857 * @addr: the MAC address entry being added
8858 * @flags: instructions from stack about fdb operation
8860 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8861 struct net_device *dev,
8862 const unsigned char *addr, u16 vid,
8865 struct i40e_netdev_priv *np = netdev_priv(dev);
8866 struct i40e_pf *pf = np->vsi->back;
8869 if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8873 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8877 /* Hardware does not support aging addresses so if a
8878 * ndm_state is given only allow permanent addresses
8880 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8881 netdev_info(dev, "FDB only supports static addresses\n");
8885 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8886 err = dev_uc_add_excl(dev, addr);
8887 else if (is_multicast_ether_addr(addr))
8888 err = dev_mc_add_excl(dev, addr);
8892 /* Only return duplicate errors if NLM_F_EXCL is set */
8893 if (err == -EEXIST && !(flags & NLM_F_EXCL))
8900 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8901 * @dev: the netdev being configured
8902 * @nlh: RTNL message
8904 * Inserts a new hardware bridge if not already created and
8905 * enables the bridging mode requested (VEB or VEPA). If the
8906 * hardware bridge has already been inserted and the request
8907 * is to change the mode then that requires a PF reset to
8908 * allow rebuild of the components with required hardware
8909 * bridge mode enabled.
8911 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8912 struct nlmsghdr *nlh,
8915 struct i40e_netdev_priv *np = netdev_priv(dev);
8916 struct i40e_vsi *vsi = np->vsi;
8917 struct i40e_pf *pf = vsi->back;
8918 struct i40e_veb *veb = NULL;
8919 struct nlattr *attr, *br_spec;
8922 /* Only for PF VSI for now */
8923 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8926 /* Find the HW bridge for PF VSI */
8927 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8928 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8932 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8934 nla_for_each_nested(attr, br_spec, rem) {
8937 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8940 mode = nla_get_u16(attr);
8941 if ((mode != BRIDGE_MODE_VEPA) &&
8942 (mode != BRIDGE_MODE_VEB))
8945 /* Insert a new HW bridge */
8947 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8948 vsi->tc_config.enabled_tc);
8950 veb->bridge_mode = mode;
8951 i40e_config_bridge_mode(veb);
8953 /* No Bridge HW offload available */
8957 } else if (mode != veb->bridge_mode) {
8958 /* Existing HW bridge but different mode needs reset */
8959 veb->bridge_mode = mode;
8960 /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8961 if (mode == BRIDGE_MODE_VEB)
8962 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8964 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8965 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8974 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8977 * @seq: RTNL message seq #
8978 * @dev: the netdev being configured
8979 * @filter_mask: unused
8980 * @nlflags: netlink flags passed in
8982 * Return the mode in which the hardware bridge is operating in
8985 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8986 struct net_device *dev,
8987 u32 __always_unused filter_mask,
8990 struct i40e_netdev_priv *np = netdev_priv(dev);
8991 struct i40e_vsi *vsi = np->vsi;
8992 struct i40e_pf *pf = vsi->back;
8993 struct i40e_veb *veb = NULL;
8996 /* Only for PF VSI for now */
8997 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9000 /* Find the HW bridge for the PF VSI */
9001 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9002 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9009 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9010 nlflags, 0, 0, filter_mask, NULL);
9013 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9014 * inner mac plus all inner ethertypes.
9016 #define I40E_MAX_TUNNEL_HDR_LEN 128
9018 * i40e_features_check - Validate encapsulated packet conforms to limits
9020 * @dev: This physical port's netdev
9021 * @features: Offload features that the stack believes apply
9023 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9024 struct net_device *dev,
9025 netdev_features_t features)
9027 if (skb->encapsulation &&
9028 ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9029 I40E_MAX_TUNNEL_HDR_LEN))
9030 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9035 static const struct net_device_ops i40e_netdev_ops = {
9036 .ndo_open = i40e_open,
9037 .ndo_stop = i40e_close,
9038 .ndo_start_xmit = i40e_lan_xmit_frame,
9039 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
9040 .ndo_set_rx_mode = i40e_set_rx_mode,
9041 .ndo_validate_addr = eth_validate_addr,
9042 .ndo_set_mac_address = i40e_set_mac,
9043 .ndo_change_mtu = i40e_change_mtu,
9044 .ndo_do_ioctl = i40e_ioctl,
9045 .ndo_tx_timeout = i40e_tx_timeout,
9046 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
9047 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
9048 #ifdef CONFIG_NET_POLL_CONTROLLER
9049 .ndo_poll_controller = i40e_netpoll,
9051 .ndo_setup_tc = __i40e_setup_tc,
9053 .ndo_fcoe_enable = i40e_fcoe_enable,
9054 .ndo_fcoe_disable = i40e_fcoe_disable,
9056 .ndo_set_features = i40e_set_features,
9057 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
9058 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
9059 .ndo_set_vf_rate = i40e_ndo_set_vf_bw,
9060 .ndo_get_vf_config = i40e_ndo_get_vf_config,
9061 .ndo_set_vf_link_state = i40e_ndo_set_vf_link_state,
9062 .ndo_set_vf_spoofchk = i40e_ndo_set_vf_spoofchk,
9063 .ndo_set_vf_trust = i40e_ndo_set_vf_trust,
9064 .ndo_udp_tunnel_add = i40e_udp_tunnel_add,
9065 .ndo_udp_tunnel_del = i40e_udp_tunnel_del,
9066 .ndo_get_phys_port_id = i40e_get_phys_port_id,
9067 .ndo_fdb_add = i40e_ndo_fdb_add,
9068 .ndo_features_check = i40e_features_check,
9069 .ndo_bridge_getlink = i40e_ndo_bridge_getlink,
9070 .ndo_bridge_setlink = i40e_ndo_bridge_setlink,
9074 * i40e_config_netdev - Setup the netdev flags
9075 * @vsi: the VSI being configured
9077 * Returns 0 on success, negative value on failure
9079 static int i40e_config_netdev(struct i40e_vsi *vsi)
9081 struct i40e_pf *pf = vsi->back;
9082 struct i40e_hw *hw = &pf->hw;
9083 struct i40e_netdev_priv *np;
9084 struct net_device *netdev;
9085 u8 mac_addr[ETH_ALEN];
9088 etherdev_size = sizeof(struct i40e_netdev_priv);
9089 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9093 vsi->netdev = netdev;
9094 np = netdev_priv(netdev);
9097 netdev->hw_enc_features |= NETIF_F_SG |
9101 NETIF_F_SOFT_FEATURES |
9106 NETIF_F_GSO_GRE_CSUM |
9107 NETIF_F_GSO_IPXIP4 |
9108 NETIF_F_GSO_IPXIP6 |
9109 NETIF_F_GSO_UDP_TUNNEL |
9110 NETIF_F_GSO_UDP_TUNNEL_CSUM |
9111 NETIF_F_GSO_PARTIAL |
9117 if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9118 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9120 netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9122 /* record features VLANs can make use of */
9123 netdev->vlan_features |= netdev->hw_enc_features |
9124 NETIF_F_TSO_MANGLEID;
9126 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9127 netdev->hw_features |= NETIF_F_NTUPLE;
9129 netdev->hw_features |= netdev->hw_enc_features |
9130 NETIF_F_HW_VLAN_CTAG_TX |
9131 NETIF_F_HW_VLAN_CTAG_RX;
9133 netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9134 netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9136 if (vsi->type == I40E_VSI_MAIN) {
9137 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9138 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9139 /* The following steps are necessary to prevent reception
9140 * of tagged packets - some older NVM configurations load a
9141 * default a MAC-VLAN filter that accepts any tagged packet
9142 * which must be replaced by a normal filter.
9144 i40e_rm_default_mac_filter(vsi, mac_addr);
9145 spin_lock_bh(&vsi->mac_filter_list_lock);
9146 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, true);
9147 spin_unlock_bh(&vsi->mac_filter_list_lock);
9149 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9150 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9151 pf->vsi[pf->lan_vsi]->netdev->name);
9152 random_ether_addr(mac_addr);
9154 spin_lock_bh(&vsi->mac_filter_list_lock);
9155 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9156 spin_unlock_bh(&vsi->mac_filter_list_lock);
9159 ether_addr_copy(netdev->dev_addr, mac_addr);
9160 ether_addr_copy(netdev->perm_addr, mac_addr);
9162 netdev->priv_flags |= IFF_UNICAST_FLT;
9163 netdev->priv_flags |= IFF_SUPP_NOFCS;
9164 /* Setup netdev TC information */
9165 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9167 netdev->netdev_ops = &i40e_netdev_ops;
9168 netdev->watchdog_timeo = 5 * HZ;
9169 i40e_set_ethtool_ops(netdev);
9171 i40e_fcoe_config_netdev(netdev, vsi);
9178 * i40e_vsi_delete - Delete a VSI from the switch
9179 * @vsi: the VSI being removed
9181 * Returns 0 on success, negative value on failure
9183 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9185 /* remove default VSI is not allowed */
9186 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9189 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9193 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9194 * @vsi: the VSI being queried
9196 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9198 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9200 struct i40e_veb *veb;
9201 struct i40e_pf *pf = vsi->back;
9203 /* Uplink is not a bridge so default to VEB */
9204 if (vsi->veb_idx == I40E_NO_VEB)
9207 veb = pf->veb[vsi->veb_idx];
9209 dev_info(&pf->pdev->dev,
9210 "There is no veb associated with the bridge\n");
9214 /* Uplink is a bridge in VEPA mode */
9215 if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9218 /* Uplink is a bridge in VEB mode */
9222 /* VEPA is now default bridge, so return 0 */
9227 * i40e_add_vsi - Add a VSI to the switch
9228 * @vsi: the VSI being configured
9230 * This initializes a VSI context depending on the VSI type to be added and
9231 * passes it down to the add_vsi aq command.
9233 static int i40e_add_vsi(struct i40e_vsi *vsi)
9236 i40e_status aq_ret = 0;
9237 struct i40e_pf *pf = vsi->back;
9238 struct i40e_hw *hw = &pf->hw;
9239 struct i40e_vsi_context ctxt;
9240 struct i40e_mac_filter *f, *ftmp;
9242 u8 enabled_tc = 0x1; /* TC0 enabled */
9245 memset(&ctxt, 0, sizeof(ctxt));
9246 switch (vsi->type) {
9248 /* The PF's main VSI is already setup as part of the
9249 * device initialization, so we'll not bother with
9250 * the add_vsi call, but we will retrieve the current
9253 ctxt.seid = pf->main_vsi_seid;
9254 ctxt.pf_num = pf->hw.pf_id;
9256 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9257 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9259 dev_info(&pf->pdev->dev,
9260 "couldn't get PF vsi config, err %s aq_err %s\n",
9261 i40e_stat_str(&pf->hw, ret),
9262 i40e_aq_str(&pf->hw,
9263 pf->hw.aq.asq_last_status));
9266 vsi->info = ctxt.info;
9267 vsi->info.valid_sections = 0;
9269 vsi->seid = ctxt.seid;
9270 vsi->id = ctxt.vsi_number;
9272 enabled_tc = i40e_pf_get_tc_map(pf);
9274 /* MFP mode setup queue map and update VSI */
9275 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9276 !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9277 memset(&ctxt, 0, sizeof(ctxt));
9278 ctxt.seid = pf->main_vsi_seid;
9279 ctxt.pf_num = pf->hw.pf_id;
9281 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9282 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9284 dev_info(&pf->pdev->dev,
9285 "update vsi failed, err %s aq_err %s\n",
9286 i40e_stat_str(&pf->hw, ret),
9287 i40e_aq_str(&pf->hw,
9288 pf->hw.aq.asq_last_status));
9292 /* update the local VSI info queue map */
9293 i40e_vsi_update_queue_map(vsi, &ctxt);
9294 vsi->info.valid_sections = 0;
9296 /* Default/Main VSI is only enabled for TC0
9297 * reconfigure it to enable all TCs that are
9298 * available on the port in SFP mode.
9299 * For MFP case the iSCSI PF would use this
9300 * flow to enable LAN+iSCSI TC.
9302 ret = i40e_vsi_config_tc(vsi, enabled_tc);
9304 dev_info(&pf->pdev->dev,
9305 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9307 i40e_stat_str(&pf->hw, ret),
9308 i40e_aq_str(&pf->hw,
9309 pf->hw.aq.asq_last_status));
9316 ctxt.pf_num = hw->pf_id;
9318 ctxt.uplink_seid = vsi->uplink_seid;
9319 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9320 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9321 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9322 (i40e_is_vsi_uplink_mode_veb(vsi))) {
9323 ctxt.info.valid_sections |=
9324 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9325 ctxt.info.switch_id =
9326 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9328 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9331 case I40E_VSI_VMDQ2:
9332 ctxt.pf_num = hw->pf_id;
9334 ctxt.uplink_seid = vsi->uplink_seid;
9335 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9336 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9338 /* This VSI is connected to VEB so the switch_id
9339 * should be set to zero by default.
9341 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9342 ctxt.info.valid_sections |=
9343 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9344 ctxt.info.switch_id =
9345 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9348 /* Setup the VSI tx/rx queue map for TC0 only for now */
9349 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9352 case I40E_VSI_SRIOV:
9353 ctxt.pf_num = hw->pf_id;
9354 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9355 ctxt.uplink_seid = vsi->uplink_seid;
9356 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9357 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9359 /* This VSI is connected to VEB so the switch_id
9360 * should be set to zero by default.
9362 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9363 ctxt.info.valid_sections |=
9364 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9365 ctxt.info.switch_id =
9366 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9369 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9370 ctxt.info.valid_sections |=
9371 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9372 ctxt.info.queueing_opt_flags |=
9373 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9374 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9377 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9378 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9379 if (pf->vf[vsi->vf_id].spoofchk) {
9380 ctxt.info.valid_sections |=
9381 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9382 ctxt.info.sec_flags |=
9383 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9384 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9386 /* Setup the VSI tx/rx queue map for TC0 only for now */
9387 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9392 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9394 dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9399 #endif /* I40E_FCOE */
9400 case I40E_VSI_IWARP:
9401 /* send down message to iWARP */
9408 if (vsi->type != I40E_VSI_MAIN) {
9409 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9411 dev_info(&vsi->back->pdev->dev,
9412 "add vsi failed, err %s aq_err %s\n",
9413 i40e_stat_str(&pf->hw, ret),
9414 i40e_aq_str(&pf->hw,
9415 pf->hw.aq.asq_last_status));
9419 vsi->info = ctxt.info;
9420 vsi->info.valid_sections = 0;
9421 vsi->seid = ctxt.seid;
9422 vsi->id = ctxt.vsi_number;
9424 /* Except FDIR VSI, for all othet VSI set the broadcast filter */
9425 if (vsi->type != I40E_VSI_FDIR) {
9426 aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9428 ret = i40e_aq_rc_to_posix(aq_ret,
9429 hw->aq.asq_last_status);
9430 dev_info(&pf->pdev->dev,
9431 "set brdcast promisc failed, err %s, aq_err %s\n",
9432 i40e_stat_str(hw, aq_ret),
9433 i40e_aq_str(hw, hw->aq.asq_last_status));
9437 vsi->active_filters = 0;
9438 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9439 spin_lock_bh(&vsi->mac_filter_list_lock);
9440 /* If macvlan filters already exist, force them to get loaded */
9441 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9442 f->state = I40E_FILTER_NEW;
9445 spin_unlock_bh(&vsi->mac_filter_list_lock);
9448 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9449 pf->flags |= I40E_FLAG_FILTER_SYNC;
9452 /* Update VSI BW information */
9453 ret = i40e_vsi_get_bw_info(vsi);
9455 dev_info(&pf->pdev->dev,
9456 "couldn't get vsi bw info, err %s aq_err %s\n",
9457 i40e_stat_str(&pf->hw, ret),
9458 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9459 /* VSI is already added so not tearing that up */
9468 * i40e_vsi_release - Delete a VSI and free its resources
9469 * @vsi: the VSI being removed
9471 * Returns 0 on success or < 0 on error
9473 int i40e_vsi_release(struct i40e_vsi *vsi)
9475 struct i40e_mac_filter *f, *ftmp;
9476 struct i40e_veb *veb = NULL;
9483 /* release of a VEB-owner or last VSI is not allowed */
9484 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9485 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9486 vsi->seid, vsi->uplink_seid);
9489 if (vsi == pf->vsi[pf->lan_vsi] &&
9490 !test_bit(__I40E_DOWN, &pf->state)) {
9491 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9495 uplink_seid = vsi->uplink_seid;
9496 if (vsi->type != I40E_VSI_SRIOV) {
9497 if (vsi->netdev_registered) {
9498 vsi->netdev_registered = false;
9500 /* results in a call to i40e_close() */
9501 unregister_netdev(vsi->netdev);
9504 i40e_vsi_close(vsi);
9506 i40e_vsi_disable_irq(vsi);
9509 spin_lock_bh(&vsi->mac_filter_list_lock);
9510 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9511 i40e_del_filter(vsi, f->macaddr, f->vlan,
9512 f->is_vf, f->is_netdev);
9513 spin_unlock_bh(&vsi->mac_filter_list_lock);
9515 i40e_sync_vsi_filters(vsi);
9517 i40e_vsi_delete(vsi);
9518 i40e_vsi_free_q_vectors(vsi);
9520 free_netdev(vsi->netdev);
9523 i40e_vsi_clear_rings(vsi);
9524 i40e_vsi_clear(vsi);
9526 /* If this was the last thing on the VEB, except for the
9527 * controlling VSI, remove the VEB, which puts the controlling
9528 * VSI onto the next level down in the switch.
9530 * Well, okay, there's one more exception here: don't remove
9531 * the orphan VEBs yet. We'll wait for an explicit remove request
9532 * from up the network stack.
9534 for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9536 pf->vsi[i]->uplink_seid == uplink_seid &&
9537 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9538 n++; /* count the VSIs */
9541 for (i = 0; i < I40E_MAX_VEB; i++) {
9544 if (pf->veb[i]->uplink_seid == uplink_seid)
9545 n++; /* count the VEBs */
9546 if (pf->veb[i]->seid == uplink_seid)
9549 if (n == 0 && veb && veb->uplink_seid != 0)
9550 i40e_veb_release(veb);
9556 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9557 * @vsi: ptr to the VSI
9559 * This should only be called after i40e_vsi_mem_alloc() which allocates the
9560 * corresponding SW VSI structure and initializes num_queue_pairs for the
9561 * newly allocated VSI.
9563 * Returns 0 on success or negative on failure
9565 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9568 struct i40e_pf *pf = vsi->back;
9570 if (vsi->q_vectors[0]) {
9571 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9576 if (vsi->base_vector) {
9577 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9578 vsi->seid, vsi->base_vector);
9582 ret = i40e_vsi_alloc_q_vectors(vsi);
9584 dev_info(&pf->pdev->dev,
9585 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9586 vsi->num_q_vectors, vsi->seid, ret);
9587 vsi->num_q_vectors = 0;
9588 goto vector_setup_out;
9591 /* In Legacy mode, we do not have to get any other vector since we
9592 * piggyback on the misc/ICR0 for queue interrupts.
9594 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9596 if (vsi->num_q_vectors)
9597 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9598 vsi->num_q_vectors, vsi->idx);
9599 if (vsi->base_vector < 0) {
9600 dev_info(&pf->pdev->dev,
9601 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9602 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9603 i40e_vsi_free_q_vectors(vsi);
9605 goto vector_setup_out;
9613 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9614 * @vsi: pointer to the vsi.
9616 * This re-allocates a vsi's queue resources.
9618 * Returns pointer to the successfully allocated and configured VSI sw struct
9619 * on success, otherwise returns NULL on failure.
9621 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9632 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9633 i40e_vsi_clear_rings(vsi);
9635 i40e_vsi_free_arrays(vsi, false);
9636 i40e_set_num_rings_in_vsi(vsi);
9637 ret = i40e_vsi_alloc_arrays(vsi, false);
9641 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9643 dev_info(&pf->pdev->dev,
9644 "failed to get tracking for %d queues for VSI %d err %d\n",
9645 vsi->alloc_queue_pairs, vsi->seid, ret);
9648 vsi->base_queue = ret;
9650 /* Update the FW view of the VSI. Force a reset of TC and queue
9651 * layout configurations.
9653 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9654 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9655 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9656 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9657 if (vsi->type == I40E_VSI_MAIN)
9658 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9660 /* assign it some queues */
9661 ret = i40e_alloc_rings(vsi);
9665 /* map all of the rings to the q_vectors */
9666 i40e_vsi_map_rings_to_vectors(vsi);
9670 i40e_vsi_free_q_vectors(vsi);
9671 if (vsi->netdev_registered) {
9672 vsi->netdev_registered = false;
9673 unregister_netdev(vsi->netdev);
9674 free_netdev(vsi->netdev);
9677 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9679 i40e_vsi_clear(vsi);
9684 * i40e_vsi_setup - Set up a VSI by a given type
9685 * @pf: board private structure
9687 * @uplink_seid: the switch element to link to
9688 * @param1: usage depends upon VSI type. For VF types, indicates VF id
9690 * This allocates the sw VSI structure and its queue resources, then add a VSI
9691 * to the identified VEB.
9693 * Returns pointer to the successfully allocated and configure VSI sw struct on
9694 * success, otherwise returns NULL on failure.
9696 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9697 u16 uplink_seid, u32 param1)
9699 struct i40e_vsi *vsi = NULL;
9700 struct i40e_veb *veb = NULL;
9704 /* The requested uplink_seid must be either
9705 * - the PF's port seid
9706 * no VEB is needed because this is the PF
9707 * or this is a Flow Director special case VSI
9708 * - seid of an existing VEB
9709 * - seid of a VSI that owns an existing VEB
9710 * - seid of a VSI that doesn't own a VEB
9711 * a new VEB is created and the VSI becomes the owner
9712 * - seid of the PF VSI, which is what creates the first VEB
9713 * this is a special case of the previous
9715 * Find which uplink_seid we were given and create a new VEB if needed
9717 for (i = 0; i < I40E_MAX_VEB; i++) {
9718 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9724 if (!veb && uplink_seid != pf->mac_seid) {
9726 for (i = 0; i < pf->num_alloc_vsi; i++) {
9727 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9733 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9738 if (vsi->uplink_seid == pf->mac_seid)
9739 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9740 vsi->tc_config.enabled_tc);
9741 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9742 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9743 vsi->tc_config.enabled_tc);
9745 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9746 dev_info(&vsi->back->pdev->dev,
9747 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9750 /* We come up by default in VEPA mode if SRIOV is not
9751 * already enabled, in which case we can't force VEPA
9754 if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9755 veb->bridge_mode = BRIDGE_MODE_VEPA;
9756 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9758 i40e_config_bridge_mode(veb);
9760 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9761 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9765 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9769 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9770 uplink_seid = veb->seid;
9773 /* get vsi sw struct */
9774 v_idx = i40e_vsi_mem_alloc(pf, type);
9777 vsi = pf->vsi[v_idx];
9781 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9783 if (type == I40E_VSI_MAIN)
9784 pf->lan_vsi = v_idx;
9785 else if (type == I40E_VSI_SRIOV)
9786 vsi->vf_id = param1;
9787 /* assign it some queues */
9788 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9791 dev_info(&pf->pdev->dev,
9792 "failed to get tracking for %d queues for VSI %d err=%d\n",
9793 vsi->alloc_queue_pairs, vsi->seid, ret);
9796 vsi->base_queue = ret;
9798 /* get a VSI from the hardware */
9799 vsi->uplink_seid = uplink_seid;
9800 ret = i40e_add_vsi(vsi);
9804 switch (vsi->type) {
9805 /* setup the netdev if needed */
9807 /* Apply relevant filters if a platform-specific mac
9808 * address was selected.
9810 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9811 ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9813 dev_warn(&pf->pdev->dev,
9814 "could not set up macaddr; err %d\n",
9818 case I40E_VSI_VMDQ2:
9820 ret = i40e_config_netdev(vsi);
9823 ret = register_netdev(vsi->netdev);
9826 vsi->netdev_registered = true;
9827 netif_carrier_off(vsi->netdev);
9828 #ifdef CONFIG_I40E_DCB
9829 /* Setup DCB netlink interface */
9830 i40e_dcbnl_setup(vsi);
9831 #endif /* CONFIG_I40E_DCB */
9835 /* set up vectors and rings if needed */
9836 ret = i40e_vsi_setup_vectors(vsi);
9840 ret = i40e_alloc_rings(vsi);
9844 /* map all of the rings to the q_vectors */
9845 i40e_vsi_map_rings_to_vectors(vsi);
9847 i40e_vsi_reset_stats(vsi);
9851 /* no netdev or rings for the other VSI types */
9855 if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9856 (vsi->type == I40E_VSI_VMDQ2)) {
9857 ret = i40e_vsi_config_rss(vsi);
9862 i40e_vsi_free_q_vectors(vsi);
9864 if (vsi->netdev_registered) {
9865 vsi->netdev_registered = false;
9866 unregister_netdev(vsi->netdev);
9867 free_netdev(vsi->netdev);
9871 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9873 i40e_vsi_clear(vsi);
9879 * i40e_veb_get_bw_info - Query VEB BW information
9880 * @veb: the veb to query
9882 * Query the Tx scheduler BW configuration data for given VEB
9884 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9886 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9887 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9888 struct i40e_pf *pf = veb->pf;
9889 struct i40e_hw *hw = &pf->hw;
9894 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9897 dev_info(&pf->pdev->dev,
9898 "query veb bw config failed, err %s aq_err %s\n",
9899 i40e_stat_str(&pf->hw, ret),
9900 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9904 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9907 dev_info(&pf->pdev->dev,
9908 "query veb bw ets config failed, err %s aq_err %s\n",
9909 i40e_stat_str(&pf->hw, ret),
9910 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9914 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9915 veb->bw_max_quanta = ets_data.tc_bw_max;
9916 veb->is_abs_credits = bw_data.absolute_credits_enable;
9917 veb->enabled_tc = ets_data.tc_valid_bits;
9918 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9919 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9920 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9921 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9922 veb->bw_tc_limit_credits[i] =
9923 le16_to_cpu(bw_data.tc_bw_limits[i]);
9924 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9932 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9933 * @pf: board private structure
9935 * On error: returns error code (negative)
9936 * On success: returns vsi index in PF (positive)
9938 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9941 struct i40e_veb *veb;
9944 /* Need to protect the allocation of switch elements at the PF level */
9945 mutex_lock(&pf->switch_mutex);
9947 /* VEB list may be fragmented if VEB creation/destruction has
9948 * been happening. We can afford to do a quick scan to look
9949 * for any free slots in the list.
9951 * find next empty veb slot, looping back around if necessary
9954 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9956 if (i >= I40E_MAX_VEB) {
9958 goto err_alloc_veb; /* out of VEB slots! */
9961 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9968 veb->enabled_tc = 1;
9973 mutex_unlock(&pf->switch_mutex);
9978 * i40e_switch_branch_release - Delete a branch of the switch tree
9979 * @branch: where to start deleting
9981 * This uses recursion to find the tips of the branch to be
9982 * removed, deleting until we get back to and can delete this VEB.
9984 static void i40e_switch_branch_release(struct i40e_veb *branch)
9986 struct i40e_pf *pf = branch->pf;
9987 u16 branch_seid = branch->seid;
9988 u16 veb_idx = branch->idx;
9991 /* release any VEBs on this VEB - RECURSION */
9992 for (i = 0; i < I40E_MAX_VEB; i++) {
9995 if (pf->veb[i]->uplink_seid == branch->seid)
9996 i40e_switch_branch_release(pf->veb[i]);
9999 /* Release the VSIs on this VEB, but not the owner VSI.
10001 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10002 * the VEB itself, so don't use (*branch) after this loop.
10004 for (i = 0; i < pf->num_alloc_vsi; i++) {
10007 if (pf->vsi[i]->uplink_seid == branch_seid &&
10008 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10009 i40e_vsi_release(pf->vsi[i]);
10013 /* There's one corner case where the VEB might not have been
10014 * removed, so double check it here and remove it if needed.
10015 * This case happens if the veb was created from the debugfs
10016 * commands and no VSIs were added to it.
10018 if (pf->veb[veb_idx])
10019 i40e_veb_release(pf->veb[veb_idx]);
10023 * i40e_veb_clear - remove veb struct
10024 * @veb: the veb to remove
10026 static void i40e_veb_clear(struct i40e_veb *veb)
10032 struct i40e_pf *pf = veb->pf;
10034 mutex_lock(&pf->switch_mutex);
10035 if (pf->veb[veb->idx] == veb)
10036 pf->veb[veb->idx] = NULL;
10037 mutex_unlock(&pf->switch_mutex);
10044 * i40e_veb_release - Delete a VEB and free its resources
10045 * @veb: the VEB being removed
10047 void i40e_veb_release(struct i40e_veb *veb)
10049 struct i40e_vsi *vsi = NULL;
10050 struct i40e_pf *pf;
10055 /* find the remaining VSI and check for extras */
10056 for (i = 0; i < pf->num_alloc_vsi; i++) {
10057 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10063 dev_info(&pf->pdev->dev,
10064 "can't remove VEB %d with %d VSIs left\n",
10069 /* move the remaining VSI to uplink veb */
10070 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10071 if (veb->uplink_seid) {
10072 vsi->uplink_seid = veb->uplink_seid;
10073 if (veb->uplink_seid == pf->mac_seid)
10074 vsi->veb_idx = I40E_NO_VEB;
10076 vsi->veb_idx = veb->veb_idx;
10079 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10080 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10083 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10084 i40e_veb_clear(veb);
10088 * i40e_add_veb - create the VEB in the switch
10089 * @veb: the VEB to be instantiated
10090 * @vsi: the controlling VSI
10092 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10094 struct i40e_pf *pf = veb->pf;
10095 bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10098 ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10099 veb->enabled_tc, false,
10100 &veb->seid, enable_stats, NULL);
10102 /* get a VEB from the hardware */
10104 dev_info(&pf->pdev->dev,
10105 "couldn't add VEB, err %s aq_err %s\n",
10106 i40e_stat_str(&pf->hw, ret),
10107 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10111 /* get statistics counter */
10112 ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10113 &veb->stats_idx, NULL, NULL, NULL);
10115 dev_info(&pf->pdev->dev,
10116 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10117 i40e_stat_str(&pf->hw, ret),
10118 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10121 ret = i40e_veb_get_bw_info(veb);
10123 dev_info(&pf->pdev->dev,
10124 "couldn't get VEB bw info, err %s aq_err %s\n",
10125 i40e_stat_str(&pf->hw, ret),
10126 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10127 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10131 vsi->uplink_seid = veb->seid;
10132 vsi->veb_idx = veb->idx;
10133 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10139 * i40e_veb_setup - Set up a VEB
10140 * @pf: board private structure
10141 * @flags: VEB setup flags
10142 * @uplink_seid: the switch element to link to
10143 * @vsi_seid: the initial VSI seid
10144 * @enabled_tc: Enabled TC bit-map
10146 * This allocates the sw VEB structure and links it into the switch
10147 * It is possible and legal for this to be a duplicate of an already
10148 * existing VEB. It is also possible for both uplink and vsi seids
10149 * to be zero, in order to create a floating VEB.
10151 * Returns pointer to the successfully allocated VEB sw struct on
10152 * success, otherwise returns NULL on failure.
10154 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10155 u16 uplink_seid, u16 vsi_seid,
10158 struct i40e_veb *veb, *uplink_veb = NULL;
10159 int vsi_idx, veb_idx;
10162 /* if one seid is 0, the other must be 0 to create a floating relay */
10163 if ((uplink_seid == 0 || vsi_seid == 0) &&
10164 (uplink_seid + vsi_seid != 0)) {
10165 dev_info(&pf->pdev->dev,
10166 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10167 uplink_seid, vsi_seid);
10171 /* make sure there is such a vsi and uplink */
10172 for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10173 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10175 if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10176 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10181 if (uplink_seid && uplink_seid != pf->mac_seid) {
10182 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10183 if (pf->veb[veb_idx] &&
10184 pf->veb[veb_idx]->seid == uplink_seid) {
10185 uplink_veb = pf->veb[veb_idx];
10190 dev_info(&pf->pdev->dev,
10191 "uplink seid %d not found\n", uplink_seid);
10196 /* get veb sw struct */
10197 veb_idx = i40e_veb_mem_alloc(pf);
10200 veb = pf->veb[veb_idx];
10201 veb->flags = flags;
10202 veb->uplink_seid = uplink_seid;
10203 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10204 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10206 /* create the VEB in the switch */
10207 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10210 if (vsi_idx == pf->lan_vsi)
10211 pf->lan_veb = veb->idx;
10216 i40e_veb_clear(veb);
10222 * i40e_setup_pf_switch_element - set PF vars based on switch type
10223 * @pf: board private structure
10224 * @ele: element we are building info from
10225 * @num_reported: total number of elements
10226 * @printconfig: should we print the contents
10228 * helper function to assist in extracting a few useful SEID values.
10230 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10231 struct i40e_aqc_switch_config_element_resp *ele,
10232 u16 num_reported, bool printconfig)
10234 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10235 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10236 u8 element_type = ele->element_type;
10237 u16 seid = le16_to_cpu(ele->seid);
10240 dev_info(&pf->pdev->dev,
10241 "type=%d seid=%d uplink=%d downlink=%d\n",
10242 element_type, seid, uplink_seid, downlink_seid);
10244 switch (element_type) {
10245 case I40E_SWITCH_ELEMENT_TYPE_MAC:
10246 pf->mac_seid = seid;
10248 case I40E_SWITCH_ELEMENT_TYPE_VEB:
10250 if (uplink_seid != pf->mac_seid)
10252 if (pf->lan_veb == I40E_NO_VEB) {
10255 /* find existing or else empty VEB */
10256 for (v = 0; v < I40E_MAX_VEB; v++) {
10257 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10262 if (pf->lan_veb == I40E_NO_VEB) {
10263 v = i40e_veb_mem_alloc(pf);
10270 pf->veb[pf->lan_veb]->seid = seid;
10271 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10272 pf->veb[pf->lan_veb]->pf = pf;
10273 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10275 case I40E_SWITCH_ELEMENT_TYPE_VSI:
10276 if (num_reported != 1)
10278 /* This is immediately after a reset so we can assume this is
10281 pf->mac_seid = uplink_seid;
10282 pf->pf_seid = downlink_seid;
10283 pf->main_vsi_seid = seid;
10285 dev_info(&pf->pdev->dev,
10286 "pf_seid=%d main_vsi_seid=%d\n",
10287 pf->pf_seid, pf->main_vsi_seid);
10289 case I40E_SWITCH_ELEMENT_TYPE_PF:
10290 case I40E_SWITCH_ELEMENT_TYPE_VF:
10291 case I40E_SWITCH_ELEMENT_TYPE_EMP:
10292 case I40E_SWITCH_ELEMENT_TYPE_BMC:
10293 case I40E_SWITCH_ELEMENT_TYPE_PE:
10294 case I40E_SWITCH_ELEMENT_TYPE_PA:
10295 /* ignore these for now */
10298 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10299 element_type, seid);
10305 * i40e_fetch_switch_configuration - Get switch config from firmware
10306 * @pf: board private structure
10307 * @printconfig: should we print the contents
10309 * Get the current switch configuration from the device and
10310 * extract a few useful SEID values.
10312 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10314 struct i40e_aqc_get_switch_config_resp *sw_config;
10320 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10324 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10326 u16 num_reported, num_total;
10328 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10332 dev_info(&pf->pdev->dev,
10333 "get switch config failed err %s aq_err %s\n",
10334 i40e_stat_str(&pf->hw, ret),
10335 i40e_aq_str(&pf->hw,
10336 pf->hw.aq.asq_last_status));
10341 num_reported = le16_to_cpu(sw_config->header.num_reported);
10342 num_total = le16_to_cpu(sw_config->header.num_total);
10345 dev_info(&pf->pdev->dev,
10346 "header: %d reported %d total\n",
10347 num_reported, num_total);
10349 for (i = 0; i < num_reported; i++) {
10350 struct i40e_aqc_switch_config_element_resp *ele =
10351 &sw_config->element[i];
10353 i40e_setup_pf_switch_element(pf, ele, num_reported,
10356 } while (next_seid != 0);
10363 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10364 * @pf: board private structure
10365 * @reinit: if the Main VSI needs to re-initialized.
10367 * Returns 0 on success, negative value on failure
10369 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10374 /* find out what's out there already */
10375 ret = i40e_fetch_switch_configuration(pf, false);
10377 dev_info(&pf->pdev->dev,
10378 "couldn't fetch switch config, err %s aq_err %s\n",
10379 i40e_stat_str(&pf->hw, ret),
10380 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10383 i40e_pf_reset_stats(pf);
10385 /* set the switch config bit for the whole device to
10386 * support limited promisc or true promisc
10387 * when user requests promisc. The default is limited
10391 if ((pf->hw.pf_id == 0) &&
10392 !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10393 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10395 if (pf->hw.pf_id == 0) {
10398 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10399 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10401 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10402 dev_info(&pf->pdev->dev,
10403 "couldn't set switch config bits, err %s aq_err %s\n",
10404 i40e_stat_str(&pf->hw, ret),
10405 i40e_aq_str(&pf->hw,
10406 pf->hw.aq.asq_last_status));
10407 /* not a fatal problem, just keep going */
10411 /* first time setup */
10412 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10413 struct i40e_vsi *vsi = NULL;
10416 /* Set up the PF VSI associated with the PF's main VSI
10417 * that is already in the HW switch
10419 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10420 uplink_seid = pf->veb[pf->lan_veb]->seid;
10422 uplink_seid = pf->mac_seid;
10423 if (pf->lan_vsi == I40E_NO_VSI)
10424 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10426 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10428 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10429 i40e_fdir_teardown(pf);
10433 /* force a reset of TC and queue layout configurations */
10434 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10436 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10437 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10438 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10440 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10442 i40e_fdir_sb_setup(pf);
10444 /* Setup static PF queue filter control settings */
10445 ret = i40e_setup_pf_filter_control(pf);
10447 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10449 /* Failure here should not stop continuing other steps */
10452 /* enable RSS in the HW, even for only one queue, as the stack can use
10455 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10456 i40e_pf_config_rss(pf);
10458 /* fill in link information and enable LSE reporting */
10459 i40e_update_link_info(&pf->hw);
10460 i40e_link_event(pf);
10462 /* Initialize user-specific link properties */
10463 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10464 I40E_AQ_AN_COMPLETED) ? true : false);
10472 * i40e_determine_queue_usage - Work out queue distribution
10473 * @pf: board private structure
10475 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10479 pf->num_lan_qps = 0;
10481 pf->num_fcoe_qps = 0;
10484 /* Find the max queues to be put into basic use. We'll always be
10485 * using TC0, whether or not DCB is running, and TC0 will get the
10488 queues_left = pf->hw.func_caps.num_tx_qp;
10490 if ((queues_left == 1) ||
10491 !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10492 /* one qp for PF, no queues for anything else */
10494 pf->alloc_rss_size = pf->num_lan_qps = 1;
10496 /* make sure all the fancies are disabled */
10497 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10498 I40E_FLAG_IWARP_ENABLED |
10500 I40E_FLAG_FCOE_ENABLED |
10502 I40E_FLAG_FD_SB_ENABLED |
10503 I40E_FLAG_FD_ATR_ENABLED |
10504 I40E_FLAG_DCB_CAPABLE |
10505 I40E_FLAG_SRIOV_ENABLED |
10506 I40E_FLAG_VMDQ_ENABLED);
10507 } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10508 I40E_FLAG_FD_SB_ENABLED |
10509 I40E_FLAG_FD_ATR_ENABLED |
10510 I40E_FLAG_DCB_CAPABLE))) {
10511 /* one qp for PF */
10512 pf->alloc_rss_size = pf->num_lan_qps = 1;
10513 queues_left -= pf->num_lan_qps;
10515 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10516 I40E_FLAG_IWARP_ENABLED |
10518 I40E_FLAG_FCOE_ENABLED |
10520 I40E_FLAG_FD_SB_ENABLED |
10521 I40E_FLAG_FD_ATR_ENABLED |
10522 I40E_FLAG_DCB_ENABLED |
10523 I40E_FLAG_VMDQ_ENABLED);
10525 /* Not enough queues for all TCs */
10526 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10527 (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10528 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10529 dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10531 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10532 num_online_cpus());
10533 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10534 pf->hw.func_caps.num_tx_qp);
10536 queues_left -= pf->num_lan_qps;
10540 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10541 if (I40E_DEFAULT_FCOE <= queues_left) {
10542 pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10543 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10544 pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10546 pf->num_fcoe_qps = 0;
10547 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10548 dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10551 queues_left -= pf->num_fcoe_qps;
10555 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10556 if (queues_left > 1) {
10557 queues_left -= 1; /* save 1 queue for FD */
10559 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10560 dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10564 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10565 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10566 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10567 (queues_left / pf->num_vf_qps));
10568 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10571 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10572 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10573 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10574 (queues_left / pf->num_vmdq_qps));
10575 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10578 pf->queues_left = queues_left;
10579 dev_dbg(&pf->pdev->dev,
10580 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10581 pf->hw.func_caps.num_tx_qp,
10582 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10583 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10584 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10587 dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10592 * i40e_setup_pf_filter_control - Setup PF static filter control
10593 * @pf: PF to be setup
10595 * i40e_setup_pf_filter_control sets up a PF's initial filter control
10596 * settings. If PE/FCoE are enabled then it will also set the per PF
10597 * based filter sizes required for them. It also enables Flow director,
10598 * ethertype and macvlan type filter settings for the pf.
10600 * Returns 0 on success, negative on failure
10602 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10604 struct i40e_filter_control_settings *settings = &pf->filter_settings;
10606 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10608 /* Flow Director is enabled */
10609 if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10610 settings->enable_fdir = true;
10612 /* Ethtype and MACVLAN filters enabled for PF */
10613 settings->enable_ethtype = true;
10614 settings->enable_macvlan = true;
10616 if (i40e_set_filter_control(&pf->hw, settings))
10622 #define INFO_STRING_LEN 255
10623 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10624 static void i40e_print_features(struct i40e_pf *pf)
10626 struct i40e_hw *hw = &pf->hw;
10630 buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10634 i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10635 #ifdef CONFIG_PCI_IOV
10636 i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10638 i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10639 pf->hw.func_caps.num_vsis,
10640 pf->vsi[pf->lan_vsi]->num_queue_pairs);
10641 if (pf->flags & I40E_FLAG_RSS_ENABLED)
10642 i += snprintf(&buf[i], REMAIN(i), " RSS");
10643 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10644 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10645 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10646 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10647 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10649 if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10650 i += snprintf(&buf[i], REMAIN(i), " DCB");
10651 i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10652 i += snprintf(&buf[i], REMAIN(i), " Geneve");
10653 if (pf->flags & I40E_FLAG_PTP)
10654 i += snprintf(&buf[i], REMAIN(i), " PTP");
10656 if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10657 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10659 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10660 i += snprintf(&buf[i], REMAIN(i), " VEB");
10662 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10664 dev_info(&pf->pdev->dev, "%s\n", buf);
10666 WARN_ON(i > INFO_STRING_LEN);
10670 * i40e_get_platform_mac_addr - get platform-specific MAC address
10672 * @pdev: PCI device information struct
10673 * @pf: board private structure
10675 * Look up the MAC address in Open Firmware on systems that support it,
10676 * and use IDPROM on SPARC if no OF address is found. On return, the
10677 * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10678 * has been selected.
10680 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10682 pf->flags &= ~I40E_FLAG_PF_MAC;
10683 if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10684 pf->flags |= I40E_FLAG_PF_MAC;
10688 * i40e_probe - Device initialization routine
10689 * @pdev: PCI device information struct
10690 * @ent: entry in i40e_pci_tbl
10692 * i40e_probe initializes a PF identified by a pci_dev structure.
10693 * The OS initialization, configuring of the PF private structure,
10694 * and a hardware reset occur.
10696 * Returns 0 on success, negative on failure
10698 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10700 struct i40e_aq_get_phy_abilities_resp abilities;
10701 struct i40e_pf *pf;
10702 struct i40e_hw *hw;
10703 static u16 pfs_found;
10711 err = pci_enable_device_mem(pdev);
10715 /* set up for high or low dma */
10716 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10718 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10720 dev_err(&pdev->dev,
10721 "DMA configuration failed: 0x%x\n", err);
10726 /* set up pci connections */
10727 err = pci_request_mem_regions(pdev, i40e_driver_name);
10729 dev_info(&pdev->dev,
10730 "pci_request_selected_regions failed %d\n", err);
10734 pci_enable_pcie_error_reporting(pdev);
10735 pci_set_master(pdev);
10737 /* Now that we have a PCI connection, we need to do the
10738 * low level device setup. This is primarily setting up
10739 * the Admin Queue structures and then querying for the
10740 * device's current profile information.
10742 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10749 set_bit(__I40E_DOWN, &pf->state);
10754 pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10755 I40E_MAX_CSR_SPACE);
10757 hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10758 if (!hw->hw_addr) {
10760 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10761 (unsigned int)pci_resource_start(pdev, 0),
10762 pf->ioremap_len, err);
10765 hw->vendor_id = pdev->vendor;
10766 hw->device_id = pdev->device;
10767 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10768 hw->subsystem_vendor_id = pdev->subsystem_vendor;
10769 hw->subsystem_device_id = pdev->subsystem_device;
10770 hw->bus.device = PCI_SLOT(pdev->devfn);
10771 hw->bus.func = PCI_FUNC(pdev->devfn);
10772 pf->instance = pfs_found;
10774 /* set up the locks for the AQ, do this only once in probe
10775 * and destroy them only once in remove
10777 mutex_init(&hw->aq.asq_mutex);
10778 mutex_init(&hw->aq.arq_mutex);
10781 pf->msg_enable = pf->hw.debug_mask;
10782 pf->msg_enable = debug;
10785 /* do a special CORER for clearing PXE mode once at init */
10786 if (hw->revision_id == 0 &&
10787 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10788 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10793 i40e_clear_pxe_mode(hw);
10796 /* Reset here to make sure all is clean and to define PF 'n' */
10798 err = i40e_pf_reset(hw);
10800 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10805 hw->aq.num_arq_entries = I40E_AQ_LEN;
10806 hw->aq.num_asq_entries = I40E_AQ_LEN;
10807 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10808 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10809 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10811 snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10813 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10815 err = i40e_init_shared_code(hw);
10817 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10822 /* set up a default setting for link flow control */
10823 pf->hw.fc.requested_mode = I40E_FC_NONE;
10825 err = i40e_init_adminq(hw);
10827 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10828 dev_info(&pdev->dev,
10829 "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");
10831 dev_info(&pdev->dev,
10832 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10837 /* provide nvm, fw, api versions */
10838 dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10839 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10840 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10841 i40e_nvm_version_str(hw));
10843 if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10844 hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10845 dev_info(&pdev->dev,
10846 "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");
10847 else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10848 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10849 dev_info(&pdev->dev,
10850 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10852 i40e_verify_eeprom(pf);
10854 /* Rev 0 hardware was never productized */
10855 if (hw->revision_id < 1)
10856 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");
10858 i40e_clear_pxe_mode(hw);
10859 err = i40e_get_capabilities(pf);
10861 goto err_adminq_setup;
10863 err = i40e_sw_init(pf);
10865 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10869 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10870 hw->func_caps.num_rx_qp,
10871 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10873 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10874 goto err_init_lan_hmc;
10877 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10879 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10881 goto err_configure_lan_hmc;
10884 /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10885 * Ignore error return codes because if it was already disabled via
10886 * hardware settings this will fail
10888 if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10889 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10890 i40e_aq_stop_lldp(hw, true, NULL);
10893 i40e_get_mac_addr(hw, hw->mac.addr);
10894 /* allow a platform config to override the HW addr */
10895 i40e_get_platform_mac_addr(pdev, pf);
10896 if (!is_valid_ether_addr(hw->mac.addr)) {
10897 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10901 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10902 ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10903 i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10904 if (is_valid_ether_addr(hw->mac.port_addr))
10905 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10907 err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10909 dev_info(&pdev->dev,
10910 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10911 if (!is_valid_ether_addr(hw->mac.san_addr)) {
10912 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10914 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10916 dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10917 #endif /* I40E_FCOE */
10919 pci_set_drvdata(pdev, pf);
10920 pci_save_state(pdev);
10921 #ifdef CONFIG_I40E_DCB
10922 err = i40e_init_pf_dcb(pf);
10924 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10925 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10926 /* Continue without DCB enabled */
10928 #endif /* CONFIG_I40E_DCB */
10930 /* set up periodic task facility */
10931 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10932 pf->service_timer_period = HZ;
10934 INIT_WORK(&pf->service_task, i40e_service_task);
10935 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10936 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10938 /* NVM bit on means WoL disabled for the port */
10939 i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10940 if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10941 pf->wol_en = false;
10944 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10946 /* set up the main switch operations */
10947 i40e_determine_queue_usage(pf);
10948 err = i40e_init_interrupt_scheme(pf);
10950 goto err_switch_setup;
10952 /* The number of VSIs reported by the FW is the minimum guaranteed
10953 * to us; HW supports far more and we share the remaining pool with
10954 * the other PFs. We allocate space for more than the guarantee with
10955 * the understanding that we might not get them all later.
10957 if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10958 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10960 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10962 /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10963 pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
10967 goto err_switch_setup;
10970 #ifdef CONFIG_PCI_IOV
10971 /* prep for VF support */
10972 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10973 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10974 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10975 if (pci_num_vf(pdev))
10976 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10979 err = i40e_setup_pf_switch(pf, false);
10981 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10985 /* Make sure flow control is set according to current settings */
10986 err = i40e_set_fc(hw, &set_fc_aq_fail, true);
10987 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
10988 dev_dbg(&pf->pdev->dev,
10989 "Set fc with err %s aq_err %s on get_phy_cap\n",
10990 i40e_stat_str(hw, err),
10991 i40e_aq_str(hw, hw->aq.asq_last_status));
10992 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
10993 dev_dbg(&pf->pdev->dev,
10994 "Set fc with err %s aq_err %s on set_phy_config\n",
10995 i40e_stat_str(hw, err),
10996 i40e_aq_str(hw, hw->aq.asq_last_status));
10997 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
10998 dev_dbg(&pf->pdev->dev,
10999 "Set fc with err %s aq_err %s on get_link_info\n",
11000 i40e_stat_str(hw, err),
11001 i40e_aq_str(hw, hw->aq.asq_last_status));
11003 /* if FDIR VSI was set up, start it now */
11004 for (i = 0; i < pf->num_alloc_vsi; i++) {
11005 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11006 i40e_vsi_open(pf->vsi[i]);
11011 /* The driver only wants link up/down and module qualification
11012 * reports from firmware. Note the negative logic.
11014 err = i40e_aq_set_phy_int_mask(&pf->hw,
11015 ~(I40E_AQ_EVENT_LINK_UPDOWN |
11016 I40E_AQ_EVENT_MEDIA_NA |
11017 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11019 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11020 i40e_stat_str(&pf->hw, err),
11021 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11023 /* Reconfigure hardware for allowing smaller MSS in the case
11024 * of TSO, so that we avoid the MDD being fired and causing
11025 * a reset in the case of small MSS+TSO.
11027 val = rd32(hw, I40E_REG_MSS);
11028 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11029 val &= ~I40E_REG_MSS_MIN_MASK;
11030 val |= I40E_64BYTE_MSS;
11031 wr32(hw, I40E_REG_MSS, val);
11034 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11036 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11038 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11039 i40e_stat_str(&pf->hw, err),
11040 i40e_aq_str(&pf->hw,
11041 pf->hw.aq.asq_last_status));
11043 /* The main driver is (mostly) up and happy. We need to set this state
11044 * before setting up the misc vector or we get a race and the vector
11045 * ends up disabled forever.
11047 clear_bit(__I40E_DOWN, &pf->state);
11049 /* In case of MSIX we are going to setup the misc vector right here
11050 * to handle admin queue events etc. In case of legacy and MSI
11051 * the misc functionality and queue processing is combined in
11052 * the same vector and that gets setup at open.
11054 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11055 err = i40e_setup_misc_vector(pf);
11057 dev_info(&pdev->dev,
11058 "setup of misc vector failed: %d\n", err);
11063 #ifdef CONFIG_PCI_IOV
11064 /* prep for VF support */
11065 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11066 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11067 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11068 /* disable link interrupts for VFs */
11069 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11070 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11071 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11074 if (pci_num_vf(pdev)) {
11075 dev_info(&pdev->dev,
11076 "Active VFs found, allocating resources.\n");
11077 err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11079 dev_info(&pdev->dev,
11080 "Error %d allocating resources for existing VFs\n",
11084 #endif /* CONFIG_PCI_IOV */
11086 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11087 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11088 pf->num_iwarp_msix,
11089 I40E_IWARP_IRQ_PILE_ID);
11090 if (pf->iwarp_base_vector < 0) {
11091 dev_info(&pdev->dev,
11092 "failed to get tracking for %d vectors for IWARP err=%d\n",
11093 pf->num_iwarp_msix, pf->iwarp_base_vector);
11094 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11098 i40e_dbg_pf_init(pf);
11100 /* tell the firmware that we're starting */
11101 i40e_send_version(pf);
11103 /* since everything's happy, start the service_task timer */
11104 mod_timer(&pf->service_timer,
11105 round_jiffies(jiffies + pf->service_timer_period));
11107 /* add this PF to client device list and launch a client service task */
11108 err = i40e_lan_add_device(pf);
11110 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11114 /* create FCoE interface */
11115 i40e_fcoe_vsi_setup(pf);
11118 #define PCI_SPEED_SIZE 8
11119 #define PCI_WIDTH_SIZE 8
11120 /* Devices on the IOSF bus do not have this information
11121 * and will report PCI Gen 1 x 1 by default so don't bother
11124 if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11125 char speed[PCI_SPEED_SIZE] = "Unknown";
11126 char width[PCI_WIDTH_SIZE] = "Unknown";
11128 /* Get the negotiated link width and speed from PCI config
11131 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11134 i40e_set_pci_config_data(hw, link_status);
11136 switch (hw->bus.speed) {
11137 case i40e_bus_speed_8000:
11138 strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11139 case i40e_bus_speed_5000:
11140 strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11141 case i40e_bus_speed_2500:
11142 strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11146 switch (hw->bus.width) {
11147 case i40e_bus_width_pcie_x8:
11148 strncpy(width, "8", PCI_WIDTH_SIZE); break;
11149 case i40e_bus_width_pcie_x4:
11150 strncpy(width, "4", PCI_WIDTH_SIZE); break;
11151 case i40e_bus_width_pcie_x2:
11152 strncpy(width, "2", PCI_WIDTH_SIZE); break;
11153 case i40e_bus_width_pcie_x1:
11154 strncpy(width, "1", PCI_WIDTH_SIZE); break;
11159 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11162 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11163 hw->bus.speed < i40e_bus_speed_8000) {
11164 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11165 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11169 /* get the requested speeds from the fw */
11170 err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11172 dev_dbg(&pf->pdev->dev, "get requested speeds ret = %s last_status = %s\n",
11173 i40e_stat_str(&pf->hw, err),
11174 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11175 pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11177 /* get the supported phy types from the fw */
11178 err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11180 dev_dbg(&pf->pdev->dev, "get supported phy types ret = %s last_status = %s\n",
11181 i40e_stat_str(&pf->hw, err),
11182 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11183 pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11185 /* Add a filter to drop all Flow control frames from any VSI from being
11186 * transmitted. By doing so we stop a malicious VF from sending out
11187 * PAUSE or PFC frames and potentially controlling traffic for other
11189 * The FW can still send Flow control frames if enabled.
11191 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11192 pf->main_vsi_seid);
11194 if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11195 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11196 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11198 /* print a string summarizing features */
11199 i40e_print_features(pf);
11203 /* Unwind what we've done if something failed in the setup */
11205 set_bit(__I40E_DOWN, &pf->state);
11206 i40e_clear_interrupt_scheme(pf);
11209 i40e_reset_interrupt_capability(pf);
11210 del_timer_sync(&pf->service_timer);
11212 err_configure_lan_hmc:
11213 (void)i40e_shutdown_lan_hmc(hw);
11215 kfree(pf->qp_pile);
11219 iounmap(hw->hw_addr);
11223 pci_disable_pcie_error_reporting(pdev);
11224 pci_release_mem_regions(pdev);
11227 pci_disable_device(pdev);
11232 * i40e_remove - Device removal routine
11233 * @pdev: PCI device information struct
11235 * i40e_remove is called by the PCI subsystem to alert the driver
11236 * that is should release a PCI device. This could be caused by a
11237 * Hot-Plug event, or because the driver is going to be removed from
11240 static void i40e_remove(struct pci_dev *pdev)
11242 struct i40e_pf *pf = pci_get_drvdata(pdev);
11243 struct i40e_hw *hw = &pf->hw;
11244 i40e_status ret_code;
11247 i40e_dbg_pf_exit(pf);
11251 /* Disable RSS in hw */
11252 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11253 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11255 /* no more scheduling of any task */
11256 set_bit(__I40E_SUSPENDED, &pf->state);
11257 set_bit(__I40E_DOWN, &pf->state);
11258 if (pf->service_timer.data)
11259 del_timer_sync(&pf->service_timer);
11260 if (pf->service_task.func)
11261 cancel_work_sync(&pf->service_task);
11263 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11265 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11268 i40e_fdir_teardown(pf);
11270 /* If there is a switch structure or any orphans, remove them.
11271 * This will leave only the PF's VSI remaining.
11273 for (i = 0; i < I40E_MAX_VEB; i++) {
11277 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11278 pf->veb[i]->uplink_seid == 0)
11279 i40e_switch_branch_release(pf->veb[i]);
11282 /* Now we can shutdown the PF's VSI, just before we kill
11285 if (pf->vsi[pf->lan_vsi])
11286 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11288 /* remove attached clients */
11289 ret_code = i40e_lan_del_device(pf);
11291 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11295 /* shutdown and destroy the HMC */
11296 if (hw->hmc.hmc_obj) {
11297 ret_code = i40e_shutdown_lan_hmc(hw);
11299 dev_warn(&pdev->dev,
11300 "Failed to destroy the HMC resources: %d\n",
11304 /* shutdown the adminq */
11305 ret_code = i40e_shutdown_adminq(hw);
11307 dev_warn(&pdev->dev,
11308 "Failed to destroy the Admin Queue resources: %d\n",
11311 /* destroy the locks only once, here */
11312 mutex_destroy(&hw->aq.arq_mutex);
11313 mutex_destroy(&hw->aq.asq_mutex);
11315 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11316 i40e_clear_interrupt_scheme(pf);
11317 for (i = 0; i < pf->num_alloc_vsi; i++) {
11319 i40e_vsi_clear_rings(pf->vsi[i]);
11320 i40e_vsi_clear(pf->vsi[i]);
11325 for (i = 0; i < I40E_MAX_VEB; i++) {
11330 kfree(pf->qp_pile);
11333 iounmap(hw->hw_addr);
11335 pci_release_mem_regions(pdev);
11337 pci_disable_pcie_error_reporting(pdev);
11338 pci_disable_device(pdev);
11342 * i40e_pci_error_detected - warning that something funky happened in PCI land
11343 * @pdev: PCI device information struct
11345 * Called to warn that something happened and the error handling steps
11346 * are in progress. Allows the driver to quiesce things, be ready for
11349 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11350 enum pci_channel_state error)
11352 struct i40e_pf *pf = pci_get_drvdata(pdev);
11354 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11356 /* shutdown all operations */
11357 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11359 i40e_prep_for_reset(pf);
11363 /* Request a slot reset */
11364 return PCI_ERS_RESULT_NEED_RESET;
11368 * i40e_pci_error_slot_reset - a PCI slot reset just happened
11369 * @pdev: PCI device information struct
11371 * Called to find if the driver can work with the device now that
11372 * the pci slot has been reset. If a basic connection seems good
11373 * (registers are readable and have sane content) then return a
11374 * happy little PCI_ERS_RESULT_xxx.
11376 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11378 struct i40e_pf *pf = pci_get_drvdata(pdev);
11379 pci_ers_result_t result;
11383 dev_dbg(&pdev->dev, "%s\n", __func__);
11384 if (pci_enable_device_mem(pdev)) {
11385 dev_info(&pdev->dev,
11386 "Cannot re-enable PCI device after reset.\n");
11387 result = PCI_ERS_RESULT_DISCONNECT;
11389 pci_set_master(pdev);
11390 pci_restore_state(pdev);
11391 pci_save_state(pdev);
11392 pci_wake_from_d3(pdev, false);
11394 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11396 result = PCI_ERS_RESULT_RECOVERED;
11398 result = PCI_ERS_RESULT_DISCONNECT;
11401 err = pci_cleanup_aer_uncorrect_error_status(pdev);
11403 dev_info(&pdev->dev,
11404 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11406 /* non-fatal, continue */
11413 * i40e_pci_error_resume - restart operations after PCI error recovery
11414 * @pdev: PCI device information struct
11416 * Called to allow the driver to bring things back up after PCI error
11417 * and/or reset recovery has finished.
11419 static void i40e_pci_error_resume(struct pci_dev *pdev)
11421 struct i40e_pf *pf = pci_get_drvdata(pdev);
11423 dev_dbg(&pdev->dev, "%s\n", __func__);
11424 if (test_bit(__I40E_SUSPENDED, &pf->state))
11428 i40e_handle_reset_warning(pf);
11433 * i40e_shutdown - PCI callback for shutting down
11434 * @pdev: PCI device information struct
11436 static void i40e_shutdown(struct pci_dev *pdev)
11438 struct i40e_pf *pf = pci_get_drvdata(pdev);
11439 struct i40e_hw *hw = &pf->hw;
11441 set_bit(__I40E_SUSPENDED, &pf->state);
11442 set_bit(__I40E_DOWN, &pf->state);
11444 i40e_prep_for_reset(pf);
11447 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11448 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11450 del_timer_sync(&pf->service_timer);
11451 cancel_work_sync(&pf->service_task);
11452 i40e_fdir_teardown(pf);
11455 i40e_prep_for_reset(pf);
11458 wr32(hw, I40E_PFPM_APM,
11459 (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11460 wr32(hw, I40E_PFPM_WUFC,
11461 (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11463 i40e_clear_interrupt_scheme(pf);
11465 if (system_state == SYSTEM_POWER_OFF) {
11466 pci_wake_from_d3(pdev, pf->wol_en);
11467 pci_set_power_state(pdev, PCI_D3hot);
11473 * i40e_suspend - PCI callback for moving to D3
11474 * @pdev: PCI device information struct
11476 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11478 struct i40e_pf *pf = pci_get_drvdata(pdev);
11479 struct i40e_hw *hw = &pf->hw;
11482 set_bit(__I40E_SUSPENDED, &pf->state);
11483 set_bit(__I40E_DOWN, &pf->state);
11486 i40e_prep_for_reset(pf);
11489 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11490 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11492 i40e_stop_misc_vector(pf);
11494 retval = pci_save_state(pdev);
11498 pci_wake_from_d3(pdev, pf->wol_en);
11499 pci_set_power_state(pdev, PCI_D3hot);
11505 * i40e_resume - PCI callback for waking up from D3
11506 * @pdev: PCI device information struct
11508 static int i40e_resume(struct pci_dev *pdev)
11510 struct i40e_pf *pf = pci_get_drvdata(pdev);
11513 pci_set_power_state(pdev, PCI_D0);
11514 pci_restore_state(pdev);
11515 /* pci_restore_state() clears dev->state_saves, so
11516 * call pci_save_state() again to restore it.
11518 pci_save_state(pdev);
11520 err = pci_enable_device_mem(pdev);
11522 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11525 pci_set_master(pdev);
11527 /* no wakeup events while running */
11528 pci_wake_from_d3(pdev, false);
11530 /* handling the reset will rebuild the device state */
11531 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11532 clear_bit(__I40E_DOWN, &pf->state);
11534 i40e_reset_and_rebuild(pf, false);
11542 static const struct pci_error_handlers i40e_err_handler = {
11543 .error_detected = i40e_pci_error_detected,
11544 .slot_reset = i40e_pci_error_slot_reset,
11545 .resume = i40e_pci_error_resume,
11548 static struct pci_driver i40e_driver = {
11549 .name = i40e_driver_name,
11550 .id_table = i40e_pci_tbl,
11551 .probe = i40e_probe,
11552 .remove = i40e_remove,
11554 .suspend = i40e_suspend,
11555 .resume = i40e_resume,
11557 .shutdown = i40e_shutdown,
11558 .err_handler = &i40e_err_handler,
11559 .sriov_configure = i40e_pci_sriov_configure,
11563 * i40e_init_module - Driver registration routine
11565 * i40e_init_module is the first routine called when the driver is
11566 * loaded. All it does is register with the PCI subsystem.
11568 static int __init i40e_init_module(void)
11570 pr_info("%s: %s - version %s\n", i40e_driver_name,
11571 i40e_driver_string, i40e_driver_version_str);
11572 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11574 /* we will see if single thread per module is enough for now,
11575 * it can't be any worse than using the system workqueue which
11576 * was already single threaded
11578 i40e_wq = create_singlethread_workqueue(i40e_driver_name);
11580 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11585 return pci_register_driver(&i40e_driver);
11587 module_init(i40e_init_module);
11590 * i40e_exit_module - Driver exit cleanup routine
11592 * i40e_exit_module is called just before the driver is removed
11595 static void __exit i40e_exit_module(void)
11597 pci_unregister_driver(&i40e_driver);
11598 destroy_workqueue(i40e_wq);
11601 module_exit(i40e_exit_module);