1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 - 2014 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 ******************************************************************************/
29 #include "i40e_diag.h"
30 #ifdef CONFIG_I40E_VXLAN
31 #include <net/vxlan.h>
34 const char i40e_driver_name[] = "i40e";
35 static const char i40e_driver_string[] =
36 "Intel(R) Ethernet Connection XL710 Network Driver";
40 #define DRV_VERSION_MAJOR 0
41 #define DRV_VERSION_MINOR 4
42 #define DRV_VERSION_BUILD 3
43 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
44 __stringify(DRV_VERSION_MINOR) "." \
45 __stringify(DRV_VERSION_BUILD) DRV_KERN
46 const char i40e_driver_version_str[] = DRV_VERSION;
47 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
49 /* a bit of forward declarations */
50 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
51 static void i40e_handle_reset_warning(struct i40e_pf *pf);
52 static int i40e_add_vsi(struct i40e_vsi *vsi);
53 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
54 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
55 static int i40e_setup_misc_vector(struct i40e_pf *pf);
56 static void i40e_determine_queue_usage(struct i40e_pf *pf);
57 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
58 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
59 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
61 /* i40e_pci_tbl - PCI Device ID Table
63 * Last entry must be all 0s
65 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
66 * Class, Class Mask, private data (not used) }
68 static DEFINE_PCI_DEVICE_TABLE(i40e_pci_tbl) = {
69 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
70 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X710), 0},
71 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
72 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_A), 0},
73 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
74 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
75 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_D), 0},
76 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
77 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
78 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
79 /* required last entry */
82 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
84 #define I40E_MAX_VF_COUNT 128
85 static int debug = -1;
86 module_param(debug, int, 0);
87 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
89 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
90 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
91 MODULE_LICENSE("GPL");
92 MODULE_VERSION(DRV_VERSION);
95 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
96 * @hw: pointer to the HW structure
97 * @mem: ptr to mem struct to fill out
98 * @size: size of memory requested
99 * @alignment: what to align the allocation to
101 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
102 u64 size, u32 alignment)
104 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
106 mem->size = ALIGN(size, alignment);
107 mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
108 &mem->pa, GFP_KERNEL);
116 * i40e_free_dma_mem_d - OS specific memory free for shared code
117 * @hw: pointer to the HW structure
118 * @mem: ptr to mem struct to free
120 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
122 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
124 dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
133 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
134 * @hw: pointer to the HW structure
135 * @mem: ptr to mem struct to fill out
136 * @size: size of memory requested
138 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
142 mem->va = kzalloc(size, GFP_KERNEL);
151 * i40e_free_virt_mem_d - OS specific memory free for shared code
152 * @hw: pointer to the HW structure
153 * @mem: ptr to mem struct to free
155 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
157 /* it's ok to kfree a NULL pointer */
166 * i40e_get_lump - find a lump of free generic resource
167 * @pf: board private structure
168 * @pile: the pile of resource to search
169 * @needed: the number of items needed
170 * @id: an owner id to stick on the items assigned
172 * Returns the base item index of the lump, or negative for error
174 * The search_hint trick and lack of advanced fit-finding only work
175 * because we're highly likely to have all the same size lump requests.
176 * Linear search time and any fragmentation should be minimal.
178 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
184 if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
185 dev_info(&pf->pdev->dev,
186 "param err: pile=%p needed=%d id=0x%04x\n",
191 /* start the linear search with an imperfect hint */
192 i = pile->search_hint;
193 while (i < pile->num_entries) {
194 /* skip already allocated entries */
195 if (pile->list[i] & I40E_PILE_VALID_BIT) {
200 /* do we have enough in this lump? */
201 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
202 if (pile->list[i+j] & I40E_PILE_VALID_BIT)
207 /* there was enough, so assign it to the requestor */
208 for (j = 0; j < needed; j++)
209 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
211 pile->search_hint = i + j;
214 /* not enough, so skip over it and continue looking */
223 * i40e_put_lump - return a lump of generic resource
224 * @pile: the pile of resource to search
225 * @index: the base item index
226 * @id: the owner id of the items assigned
228 * Returns the count of items in the lump
230 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
232 int valid_id = (id | I40E_PILE_VALID_BIT);
236 if (!pile || index >= pile->num_entries)
240 i < pile->num_entries && pile->list[i] == valid_id;
246 if (count && index < pile->search_hint)
247 pile->search_hint = index;
253 * i40e_service_event_schedule - Schedule the service task to wake up
254 * @pf: board private structure
256 * If not already scheduled, this puts the task into the work queue
258 static void i40e_service_event_schedule(struct i40e_pf *pf)
260 if (!test_bit(__I40E_DOWN, &pf->state) &&
261 !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
262 !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
263 schedule_work(&pf->service_task);
267 * i40e_tx_timeout - Respond to a Tx Hang
268 * @netdev: network interface device structure
270 * If any port has noticed a Tx timeout, it is likely that the whole
271 * device is munged, not just the one netdev port, so go for the full
274 static void i40e_tx_timeout(struct net_device *netdev)
276 struct i40e_netdev_priv *np = netdev_priv(netdev);
277 struct i40e_vsi *vsi = np->vsi;
278 struct i40e_pf *pf = vsi->back;
280 pf->tx_timeout_count++;
282 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
283 pf->tx_timeout_recovery_level = 0;
284 pf->tx_timeout_last_recovery = jiffies;
285 netdev_info(netdev, "tx_timeout recovery level %d\n",
286 pf->tx_timeout_recovery_level);
288 switch (pf->tx_timeout_recovery_level) {
290 /* disable and re-enable queues for the VSI */
291 if (in_interrupt()) {
292 set_bit(__I40E_REINIT_REQUESTED, &pf->state);
293 set_bit(__I40E_REINIT_REQUESTED, &vsi->state);
295 i40e_vsi_reinit_locked(vsi);
299 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
302 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
305 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
308 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
309 set_bit(__I40E_DOWN, &vsi->state);
313 i40e_service_event_schedule(pf);
314 pf->tx_timeout_recovery_level++;
318 * i40e_release_rx_desc - Store the new tail and head values
319 * @rx_ring: ring to bump
320 * @val: new head index
322 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
324 rx_ring->next_to_use = val;
326 /* Force memory writes to complete before letting h/w
327 * know there are new descriptors to fetch. (Only
328 * applicable for weak-ordered memory model archs,
332 writel(val, rx_ring->tail);
336 * i40e_get_vsi_stats_struct - Get System Network Statistics
337 * @vsi: the VSI we care about
339 * Returns the address of the device statistics structure.
340 * The statistics are actually updated from the service task.
342 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
344 return &vsi->net_stats;
348 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
349 * @netdev: network interface device structure
351 * Returns the address of the device statistics structure.
352 * The statistics are actually updated from the service task.
354 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
355 struct net_device *netdev,
356 struct rtnl_link_stats64 *stats)
358 struct i40e_netdev_priv *np = netdev_priv(netdev);
359 struct i40e_ring *tx_ring, *rx_ring;
360 struct i40e_vsi *vsi = np->vsi;
361 struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
364 if (test_bit(__I40E_DOWN, &vsi->state))
371 for (i = 0; i < vsi->num_queue_pairs; i++) {
375 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
380 start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
381 packets = tx_ring->stats.packets;
382 bytes = tx_ring->stats.bytes;
383 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
385 stats->tx_packets += packets;
386 stats->tx_bytes += bytes;
387 rx_ring = &tx_ring[1];
390 start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
391 packets = rx_ring->stats.packets;
392 bytes = rx_ring->stats.bytes;
393 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
395 stats->rx_packets += packets;
396 stats->rx_bytes += bytes;
400 /* following stats updated by i40e_watchdog_subtask() */
401 stats->multicast = vsi_stats->multicast;
402 stats->tx_errors = vsi_stats->tx_errors;
403 stats->tx_dropped = vsi_stats->tx_dropped;
404 stats->rx_errors = vsi_stats->rx_errors;
405 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
406 stats->rx_length_errors = vsi_stats->rx_length_errors;
412 * i40e_vsi_reset_stats - Resets all stats of the given vsi
413 * @vsi: the VSI to have its stats reset
415 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
417 struct rtnl_link_stats64 *ns;
423 ns = i40e_get_vsi_stats_struct(vsi);
424 memset(ns, 0, sizeof(*ns));
425 memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
426 memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
427 memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
428 if (vsi->rx_rings && vsi->rx_rings[0]) {
429 for (i = 0; i < vsi->num_queue_pairs; i++) {
430 memset(&vsi->rx_rings[i]->stats, 0 ,
431 sizeof(vsi->rx_rings[i]->stats));
432 memset(&vsi->rx_rings[i]->rx_stats, 0 ,
433 sizeof(vsi->rx_rings[i]->rx_stats));
434 memset(&vsi->tx_rings[i]->stats, 0 ,
435 sizeof(vsi->tx_rings[i]->stats));
436 memset(&vsi->tx_rings[i]->tx_stats, 0,
437 sizeof(vsi->tx_rings[i]->tx_stats));
440 vsi->stat_offsets_loaded = false;
444 * i40e_pf_reset_stats - Reset all of the stats for the given pf
445 * @pf: the PF to be reset
447 void i40e_pf_reset_stats(struct i40e_pf *pf)
449 memset(&pf->stats, 0, sizeof(pf->stats));
450 memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
451 pf->stat_offsets_loaded = false;
455 * i40e_stat_update48 - read and update a 48 bit stat from the chip
456 * @hw: ptr to the hardware info
457 * @hireg: the high 32 bit reg to read
458 * @loreg: the low 32 bit reg to read
459 * @offset_loaded: has the initial offset been loaded yet
460 * @offset: ptr to current offset value
461 * @stat: ptr to the stat
463 * Since the device stats are not reset at PFReset, they likely will not
464 * be zeroed when the driver starts. We'll save the first values read
465 * and use them as offsets to be subtracted from the raw values in order
466 * to report stats that count from zero. In the process, we also manage
467 * the potential roll-over.
469 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
470 bool offset_loaded, u64 *offset, u64 *stat)
474 if (hw->device_id == I40E_DEV_ID_QEMU) {
475 new_data = rd32(hw, loreg);
476 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
478 new_data = rd64(hw, loreg);
482 if (likely(new_data >= *offset))
483 *stat = new_data - *offset;
485 *stat = (new_data + ((u64)1 << 48)) - *offset;
486 *stat &= 0xFFFFFFFFFFFFULL;
490 * i40e_stat_update32 - read and update a 32 bit stat from the chip
491 * @hw: ptr to the hardware info
492 * @reg: the hw reg to read
493 * @offset_loaded: has the initial offset been loaded yet
494 * @offset: ptr to current offset value
495 * @stat: ptr to the stat
497 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
498 bool offset_loaded, u64 *offset, u64 *stat)
502 new_data = rd32(hw, reg);
505 if (likely(new_data >= *offset))
506 *stat = (u32)(new_data - *offset);
508 *stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
512 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
513 * @vsi: the VSI to be updated
515 void i40e_update_eth_stats(struct i40e_vsi *vsi)
517 int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
518 struct i40e_pf *pf = vsi->back;
519 struct i40e_hw *hw = &pf->hw;
520 struct i40e_eth_stats *oes;
521 struct i40e_eth_stats *es; /* device's eth stats */
523 es = &vsi->eth_stats;
524 oes = &vsi->eth_stats_offsets;
526 /* Gather up the stats that the hw collects */
527 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
528 vsi->stat_offsets_loaded,
529 &oes->tx_errors, &es->tx_errors);
530 i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
531 vsi->stat_offsets_loaded,
532 &oes->rx_discards, &es->rx_discards);
533 i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
534 vsi->stat_offsets_loaded,
535 &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
536 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
537 vsi->stat_offsets_loaded,
538 &oes->tx_errors, &es->tx_errors);
540 i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
541 I40E_GLV_GORCL(stat_idx),
542 vsi->stat_offsets_loaded,
543 &oes->rx_bytes, &es->rx_bytes);
544 i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
545 I40E_GLV_UPRCL(stat_idx),
546 vsi->stat_offsets_loaded,
547 &oes->rx_unicast, &es->rx_unicast);
548 i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
549 I40E_GLV_MPRCL(stat_idx),
550 vsi->stat_offsets_loaded,
551 &oes->rx_multicast, &es->rx_multicast);
552 i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
553 I40E_GLV_BPRCL(stat_idx),
554 vsi->stat_offsets_loaded,
555 &oes->rx_broadcast, &es->rx_broadcast);
557 i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
558 I40E_GLV_GOTCL(stat_idx),
559 vsi->stat_offsets_loaded,
560 &oes->tx_bytes, &es->tx_bytes);
561 i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
562 I40E_GLV_UPTCL(stat_idx),
563 vsi->stat_offsets_loaded,
564 &oes->tx_unicast, &es->tx_unicast);
565 i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
566 I40E_GLV_MPTCL(stat_idx),
567 vsi->stat_offsets_loaded,
568 &oes->tx_multicast, &es->tx_multicast);
569 i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
570 I40E_GLV_BPTCL(stat_idx),
571 vsi->stat_offsets_loaded,
572 &oes->tx_broadcast, &es->tx_broadcast);
573 vsi->stat_offsets_loaded = true;
577 * i40e_update_veb_stats - Update Switch component statistics
578 * @veb: the VEB being updated
580 static void i40e_update_veb_stats(struct i40e_veb *veb)
582 struct i40e_pf *pf = veb->pf;
583 struct i40e_hw *hw = &pf->hw;
584 struct i40e_eth_stats *oes;
585 struct i40e_eth_stats *es; /* device's eth stats */
588 idx = veb->stats_idx;
590 oes = &veb->stats_offsets;
592 /* Gather up the stats that the hw collects */
593 i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
594 veb->stat_offsets_loaded,
595 &oes->tx_discards, &es->tx_discards);
596 if (hw->revision_id > 0)
597 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
598 veb->stat_offsets_loaded,
599 &oes->rx_unknown_protocol,
600 &es->rx_unknown_protocol);
601 i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
602 veb->stat_offsets_loaded,
603 &oes->rx_bytes, &es->rx_bytes);
604 i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
605 veb->stat_offsets_loaded,
606 &oes->rx_unicast, &es->rx_unicast);
607 i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
608 veb->stat_offsets_loaded,
609 &oes->rx_multicast, &es->rx_multicast);
610 i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
611 veb->stat_offsets_loaded,
612 &oes->rx_broadcast, &es->rx_broadcast);
614 i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
615 veb->stat_offsets_loaded,
616 &oes->tx_bytes, &es->tx_bytes);
617 i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
618 veb->stat_offsets_loaded,
619 &oes->tx_unicast, &es->tx_unicast);
620 i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
621 veb->stat_offsets_loaded,
622 &oes->tx_multicast, &es->tx_multicast);
623 i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
624 veb->stat_offsets_loaded,
625 &oes->tx_broadcast, &es->tx_broadcast);
626 veb->stat_offsets_loaded = true;
630 * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
631 * @pf: the corresponding PF
633 * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
635 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
637 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
638 struct i40e_hw_port_stats *nsd = &pf->stats;
639 struct i40e_hw *hw = &pf->hw;
643 if ((hw->fc.current_mode != I40E_FC_FULL) &&
644 (hw->fc.current_mode != I40E_FC_RX_PAUSE))
647 xoff = nsd->link_xoff_rx;
648 i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
649 pf->stat_offsets_loaded,
650 &osd->link_xoff_rx, &nsd->link_xoff_rx);
652 /* No new LFC xoff rx */
653 if (!(nsd->link_xoff_rx - xoff))
656 /* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
657 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
658 struct i40e_vsi *vsi = pf->vsi[v];
663 for (i = 0; i < vsi->num_queue_pairs; i++) {
664 struct i40e_ring *ring = vsi->tx_rings[i];
665 clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
671 * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
672 * @pf: the corresponding PF
674 * Update the Rx XOFF counter (PAUSE frames) in PFC mode
676 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
678 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
679 struct i40e_hw_port_stats *nsd = &pf->stats;
680 bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
681 struct i40e_dcbx_config *dcb_cfg;
682 struct i40e_hw *hw = &pf->hw;
686 dcb_cfg = &hw->local_dcbx_config;
688 /* See if DCB enabled with PFC TC */
689 if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
690 !(dcb_cfg->pfc.pfcenable)) {
691 i40e_update_link_xoff_rx(pf);
695 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
696 u64 prio_xoff = nsd->priority_xoff_rx[i];
697 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
698 pf->stat_offsets_loaded,
699 &osd->priority_xoff_rx[i],
700 &nsd->priority_xoff_rx[i]);
702 /* No new PFC xoff rx */
703 if (!(nsd->priority_xoff_rx[i] - prio_xoff))
705 /* Get the TC for given priority */
706 tc = dcb_cfg->etscfg.prioritytable[i];
710 /* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
711 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
712 struct i40e_vsi *vsi = pf->vsi[v];
717 for (i = 0; i < vsi->num_queue_pairs; i++) {
718 struct i40e_ring *ring = vsi->tx_rings[i];
722 clear_bit(__I40E_HANG_CHECK_ARMED,
729 * i40e_update_vsi_stats - Update the vsi statistics counters.
730 * @vsi: the VSI to be updated
732 * There are a few instances where we store the same stat in a
733 * couple of different structs. This is partly because we have
734 * the netdev stats that need to be filled out, which is slightly
735 * different from the "eth_stats" defined by the chip and used in
736 * VF communications. We sort it out here.
738 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
740 struct i40e_pf *pf = vsi->back;
741 struct rtnl_link_stats64 *ons;
742 struct rtnl_link_stats64 *ns; /* netdev stats */
743 struct i40e_eth_stats *oes;
744 struct i40e_eth_stats *es; /* device's eth stats */
745 u32 tx_restart, tx_busy;
751 if (test_bit(__I40E_DOWN, &vsi->state) ||
752 test_bit(__I40E_CONFIG_BUSY, &pf->state))
755 ns = i40e_get_vsi_stats_struct(vsi);
756 ons = &vsi->net_stats_offsets;
757 es = &vsi->eth_stats;
758 oes = &vsi->eth_stats_offsets;
760 /* Gather up the netdev and vsi stats that the driver collects
761 * on the fly during packet processing
765 tx_restart = tx_busy = 0;
769 for (q = 0; q < vsi->num_queue_pairs; q++) {
775 p = ACCESS_ONCE(vsi->tx_rings[q]);
778 start = u64_stats_fetch_begin_irq(&p->syncp);
779 packets = p->stats.packets;
780 bytes = p->stats.bytes;
781 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
784 tx_restart += p->tx_stats.restart_queue;
785 tx_busy += p->tx_stats.tx_busy;
787 /* Rx queue is part of the same block as Tx queue */
790 start = u64_stats_fetch_begin_irq(&p->syncp);
791 packets = p->stats.packets;
792 bytes = p->stats.bytes;
793 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
796 rx_buf += p->rx_stats.alloc_buff_failed;
797 rx_page += p->rx_stats.alloc_page_failed;
800 vsi->tx_restart = tx_restart;
801 vsi->tx_busy = tx_busy;
802 vsi->rx_page_failed = rx_page;
803 vsi->rx_buf_failed = rx_buf;
805 ns->rx_packets = rx_p;
807 ns->tx_packets = tx_p;
810 /* update netdev stats from eth stats */
811 i40e_update_eth_stats(vsi);
812 ons->tx_errors = oes->tx_errors;
813 ns->tx_errors = es->tx_errors;
814 ons->multicast = oes->rx_multicast;
815 ns->multicast = es->rx_multicast;
816 ons->rx_dropped = oes->rx_discards;
817 ns->rx_dropped = es->rx_discards;
818 ons->tx_dropped = oes->tx_discards;
819 ns->tx_dropped = es->tx_discards;
821 /* pull in a couple PF stats if this is the main vsi */
822 if (vsi == pf->vsi[pf->lan_vsi]) {
823 ns->rx_crc_errors = pf->stats.crc_errors;
824 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
825 ns->rx_length_errors = pf->stats.rx_length_errors;
830 * i40e_update_pf_stats - Update the pf statistics counters.
831 * @pf: the PF to be updated
833 static void i40e_update_pf_stats(struct i40e_pf *pf)
835 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
836 struct i40e_hw_port_stats *nsd = &pf->stats;
837 struct i40e_hw *hw = &pf->hw;
841 i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
842 I40E_GLPRT_GORCL(hw->port),
843 pf->stat_offsets_loaded,
844 &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
845 i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
846 I40E_GLPRT_GOTCL(hw->port),
847 pf->stat_offsets_loaded,
848 &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
849 i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
850 pf->stat_offsets_loaded,
851 &osd->eth.rx_discards,
852 &nsd->eth.rx_discards);
853 i40e_stat_update32(hw, I40E_GLPRT_TDPC(hw->port),
854 pf->stat_offsets_loaded,
855 &osd->eth.tx_discards,
856 &nsd->eth.tx_discards);
858 i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
859 I40E_GLPRT_UPRCL(hw->port),
860 pf->stat_offsets_loaded,
861 &osd->eth.rx_unicast,
862 &nsd->eth.rx_unicast);
863 i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
864 I40E_GLPRT_MPRCL(hw->port),
865 pf->stat_offsets_loaded,
866 &osd->eth.rx_multicast,
867 &nsd->eth.rx_multicast);
868 i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
869 I40E_GLPRT_BPRCL(hw->port),
870 pf->stat_offsets_loaded,
871 &osd->eth.rx_broadcast,
872 &nsd->eth.rx_broadcast);
873 i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
874 I40E_GLPRT_UPTCL(hw->port),
875 pf->stat_offsets_loaded,
876 &osd->eth.tx_unicast,
877 &nsd->eth.tx_unicast);
878 i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
879 I40E_GLPRT_MPTCL(hw->port),
880 pf->stat_offsets_loaded,
881 &osd->eth.tx_multicast,
882 &nsd->eth.tx_multicast);
883 i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
884 I40E_GLPRT_BPTCL(hw->port),
885 pf->stat_offsets_loaded,
886 &osd->eth.tx_broadcast,
887 &nsd->eth.tx_broadcast);
889 i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
890 pf->stat_offsets_loaded,
891 &osd->tx_dropped_link_down,
892 &nsd->tx_dropped_link_down);
894 i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
895 pf->stat_offsets_loaded,
896 &osd->crc_errors, &nsd->crc_errors);
898 i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
899 pf->stat_offsets_loaded,
900 &osd->illegal_bytes, &nsd->illegal_bytes);
902 i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
903 pf->stat_offsets_loaded,
904 &osd->mac_local_faults,
905 &nsd->mac_local_faults);
906 i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
907 pf->stat_offsets_loaded,
908 &osd->mac_remote_faults,
909 &nsd->mac_remote_faults);
911 i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
912 pf->stat_offsets_loaded,
913 &osd->rx_length_errors,
914 &nsd->rx_length_errors);
916 i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
917 pf->stat_offsets_loaded,
918 &osd->link_xon_rx, &nsd->link_xon_rx);
919 i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
920 pf->stat_offsets_loaded,
921 &osd->link_xon_tx, &nsd->link_xon_tx);
922 i40e_update_prio_xoff_rx(pf); /* handles I40E_GLPRT_LXOFFRXC */
923 i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
924 pf->stat_offsets_loaded,
925 &osd->link_xoff_tx, &nsd->link_xoff_tx);
927 for (i = 0; i < 8; i++) {
928 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
929 pf->stat_offsets_loaded,
930 &osd->priority_xon_rx[i],
931 &nsd->priority_xon_rx[i]);
932 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
933 pf->stat_offsets_loaded,
934 &osd->priority_xon_tx[i],
935 &nsd->priority_xon_tx[i]);
936 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
937 pf->stat_offsets_loaded,
938 &osd->priority_xoff_tx[i],
939 &nsd->priority_xoff_tx[i]);
940 i40e_stat_update32(hw,
941 I40E_GLPRT_RXON2OFFCNT(hw->port, i),
942 pf->stat_offsets_loaded,
943 &osd->priority_xon_2_xoff[i],
944 &nsd->priority_xon_2_xoff[i]);
947 i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
948 I40E_GLPRT_PRC64L(hw->port),
949 pf->stat_offsets_loaded,
950 &osd->rx_size_64, &nsd->rx_size_64);
951 i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
952 I40E_GLPRT_PRC127L(hw->port),
953 pf->stat_offsets_loaded,
954 &osd->rx_size_127, &nsd->rx_size_127);
955 i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
956 I40E_GLPRT_PRC255L(hw->port),
957 pf->stat_offsets_loaded,
958 &osd->rx_size_255, &nsd->rx_size_255);
959 i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
960 I40E_GLPRT_PRC511L(hw->port),
961 pf->stat_offsets_loaded,
962 &osd->rx_size_511, &nsd->rx_size_511);
963 i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
964 I40E_GLPRT_PRC1023L(hw->port),
965 pf->stat_offsets_loaded,
966 &osd->rx_size_1023, &nsd->rx_size_1023);
967 i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
968 I40E_GLPRT_PRC1522L(hw->port),
969 pf->stat_offsets_loaded,
970 &osd->rx_size_1522, &nsd->rx_size_1522);
971 i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
972 I40E_GLPRT_PRC9522L(hw->port),
973 pf->stat_offsets_loaded,
974 &osd->rx_size_big, &nsd->rx_size_big);
976 i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
977 I40E_GLPRT_PTC64L(hw->port),
978 pf->stat_offsets_loaded,
979 &osd->tx_size_64, &nsd->tx_size_64);
980 i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
981 I40E_GLPRT_PTC127L(hw->port),
982 pf->stat_offsets_loaded,
983 &osd->tx_size_127, &nsd->tx_size_127);
984 i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
985 I40E_GLPRT_PTC255L(hw->port),
986 pf->stat_offsets_loaded,
987 &osd->tx_size_255, &nsd->tx_size_255);
988 i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
989 I40E_GLPRT_PTC511L(hw->port),
990 pf->stat_offsets_loaded,
991 &osd->tx_size_511, &nsd->tx_size_511);
992 i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
993 I40E_GLPRT_PTC1023L(hw->port),
994 pf->stat_offsets_loaded,
995 &osd->tx_size_1023, &nsd->tx_size_1023);
996 i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
997 I40E_GLPRT_PTC1522L(hw->port),
998 pf->stat_offsets_loaded,
999 &osd->tx_size_1522, &nsd->tx_size_1522);
1000 i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1001 I40E_GLPRT_PTC9522L(hw->port),
1002 pf->stat_offsets_loaded,
1003 &osd->tx_size_big, &nsd->tx_size_big);
1005 i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1006 pf->stat_offsets_loaded,
1007 &osd->rx_undersize, &nsd->rx_undersize);
1008 i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1009 pf->stat_offsets_loaded,
1010 &osd->rx_fragments, &nsd->rx_fragments);
1011 i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1012 pf->stat_offsets_loaded,
1013 &osd->rx_oversize, &nsd->rx_oversize);
1014 i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1015 pf->stat_offsets_loaded,
1016 &osd->rx_jabber, &nsd->rx_jabber);
1018 val = rd32(hw, I40E_PRTPM_EEE_STAT);
1019 nsd->tx_lpi_status =
1020 (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1021 I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1022 nsd->rx_lpi_status =
1023 (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1024 I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1025 i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1026 pf->stat_offsets_loaded,
1027 &osd->tx_lpi_count, &nsd->tx_lpi_count);
1028 i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1029 pf->stat_offsets_loaded,
1030 &osd->rx_lpi_count, &nsd->rx_lpi_count);
1032 pf->stat_offsets_loaded = true;
1036 * i40e_update_stats - Update the various statistics counters.
1037 * @vsi: the VSI to be updated
1039 * Update the various stats for this VSI and its related entities.
1041 void i40e_update_stats(struct i40e_vsi *vsi)
1043 struct i40e_pf *pf = vsi->back;
1045 if (vsi == pf->vsi[pf->lan_vsi])
1046 i40e_update_pf_stats(pf);
1048 i40e_update_vsi_stats(vsi);
1052 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1053 * @vsi: the VSI to be searched
1054 * @macaddr: the MAC address
1056 * @is_vf: make sure its a vf filter, else doesn't matter
1057 * @is_netdev: make sure its a netdev filter, else doesn't matter
1059 * Returns ptr to the filter object or NULL
1061 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1062 u8 *macaddr, s16 vlan,
1063 bool is_vf, bool is_netdev)
1065 struct i40e_mac_filter *f;
1067 if (!vsi || !macaddr)
1070 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1071 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1072 (vlan == f->vlan) &&
1073 (!is_vf || f->is_vf) &&
1074 (!is_netdev || f->is_netdev))
1081 * i40e_find_mac - Find a mac addr in the macvlan filters list
1082 * @vsi: the VSI to be searched
1083 * @macaddr: the MAC address we are searching for
1084 * @is_vf: make sure its a vf filter, else doesn't matter
1085 * @is_netdev: make sure its a netdev filter, else doesn't matter
1087 * Returns the first filter with the provided MAC address or NULL if
1088 * MAC address was not found
1090 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1091 bool is_vf, bool is_netdev)
1093 struct i40e_mac_filter *f;
1095 if (!vsi || !macaddr)
1098 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1099 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1100 (!is_vf || f->is_vf) &&
1101 (!is_netdev || f->is_netdev))
1108 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1109 * @vsi: the VSI to be searched
1111 * Returns true if VSI is in vlan mode or false otherwise
1113 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1115 struct i40e_mac_filter *f;
1117 /* Only -1 for all the filters denotes not in vlan mode
1118 * so we have to go through all the list in order to make sure
1120 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1129 * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1130 * @vsi: the VSI to be searched
1131 * @macaddr: the mac address to be filtered
1132 * @is_vf: true if it is a vf
1133 * @is_netdev: true if it is a netdev
1135 * Goes through all the macvlan filters and adds a
1136 * macvlan filter for each unique vlan that already exists
1138 * Returns first filter found on success, else NULL
1140 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1141 bool is_vf, bool is_netdev)
1143 struct i40e_mac_filter *f;
1145 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1146 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1147 is_vf, is_netdev)) {
1148 if (!i40e_add_filter(vsi, macaddr, f->vlan,
1154 return list_first_entry_or_null(&vsi->mac_filter_list,
1155 struct i40e_mac_filter, list);
1159 * i40e_add_filter - Add a mac/vlan filter to the VSI
1160 * @vsi: the VSI to be searched
1161 * @macaddr: the MAC address
1163 * @is_vf: make sure its a vf filter, else doesn't matter
1164 * @is_netdev: make sure its a netdev filter, else doesn't matter
1166 * Returns ptr to the filter object or NULL when no memory available.
1168 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1169 u8 *macaddr, s16 vlan,
1170 bool is_vf, bool is_netdev)
1172 struct i40e_mac_filter *f;
1174 if (!vsi || !macaddr)
1177 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1179 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1181 goto add_filter_out;
1183 memcpy(f->macaddr, macaddr, ETH_ALEN);
1187 INIT_LIST_HEAD(&f->list);
1188 list_add(&f->list, &vsi->mac_filter_list);
1191 /* increment counter and add a new flag if needed */
1197 } else if (is_netdev) {
1198 if (!f->is_netdev) {
1199 f->is_netdev = true;
1206 /* changed tells sync_filters_subtask to
1207 * push the filter down to the firmware
1210 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1211 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1219 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1220 * @vsi: the VSI to be searched
1221 * @macaddr: the MAC address
1223 * @is_vf: make sure it's a vf filter, else doesn't matter
1224 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1226 void i40e_del_filter(struct i40e_vsi *vsi,
1227 u8 *macaddr, s16 vlan,
1228 bool is_vf, bool is_netdev)
1230 struct i40e_mac_filter *f;
1232 if (!vsi || !macaddr)
1235 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1236 if (!f || f->counter == 0)
1244 } else if (is_netdev) {
1246 f->is_netdev = false;
1250 /* make sure we don't remove a filter in use by vf or netdev */
1252 min_f += (f->is_vf ? 1 : 0);
1253 min_f += (f->is_netdev ? 1 : 0);
1255 if (f->counter > min_f)
1259 /* counter == 0 tells sync_filters_subtask to
1260 * remove the filter from the firmware's list
1262 if (f->counter == 0) {
1264 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1265 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1270 * i40e_set_mac - NDO callback to set mac address
1271 * @netdev: network interface device structure
1272 * @p: pointer to an address structure
1274 * Returns 0 on success, negative on failure
1276 static int i40e_set_mac(struct net_device *netdev, void *p)
1278 struct i40e_netdev_priv *np = netdev_priv(netdev);
1279 struct i40e_vsi *vsi = np->vsi;
1280 struct sockaddr *addr = p;
1281 struct i40e_mac_filter *f;
1283 if (!is_valid_ether_addr(addr->sa_data))
1284 return -EADDRNOTAVAIL;
1286 netdev_info(netdev, "set mac address=%pM\n", addr->sa_data);
1288 if (ether_addr_equal(netdev->dev_addr, addr->sa_data))
1291 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1292 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1293 return -EADDRNOTAVAIL;
1295 if (vsi->type == I40E_VSI_MAIN) {
1297 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1298 I40E_AQC_WRITE_TYPE_LAA_ONLY,
1299 addr->sa_data, NULL);
1302 "Addr change for Main VSI failed: %d\n",
1304 return -EADDRNOTAVAIL;
1307 memcpy(vsi->back->hw.mac.addr, addr->sa_data, netdev->addr_len);
1310 /* In order to be sure to not drop any packets, add the new address
1311 * then delete the old one.
1313 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY, false, false);
1317 i40e_sync_vsi_filters(vsi);
1318 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY, false, false);
1319 i40e_sync_vsi_filters(vsi);
1321 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1327 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1328 * @vsi: the VSI being setup
1329 * @ctxt: VSI context structure
1330 * @enabled_tc: Enabled TCs bitmap
1331 * @is_add: True if called before Add VSI
1333 * Setup VSI queue mapping for enabled traffic classes.
1335 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1336 struct i40e_vsi_context *ctxt,
1340 struct i40e_pf *pf = vsi->back;
1350 sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1353 if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1354 /* Find numtc from enabled TC bitmap */
1355 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1356 if (enabled_tc & (1 << i)) /* TC is enabled */
1360 dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1364 /* At least TC0 is enabled in case of non-DCB case */
1368 vsi->tc_config.numtc = numtc;
1369 vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1370 /* Number of queues per enabled TC */
1371 num_tc_qps = rounddown_pow_of_two(vsi->alloc_queue_pairs/numtc);
1372 num_tc_qps = min_t(int, num_tc_qps, I40E_MAX_QUEUES_PER_TC);
1374 /* Setup queue offset/count for all TCs for given VSI */
1375 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1376 /* See if the given TC is enabled for the given VSI */
1377 if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1380 switch (vsi->type) {
1382 qcount = min_t(int, pf->rss_size, num_tc_qps);
1385 case I40E_VSI_SRIOV:
1386 case I40E_VSI_VMDQ2:
1388 qcount = num_tc_qps;
1392 vsi->tc_config.tc_info[i].qoffset = offset;
1393 vsi->tc_config.tc_info[i].qcount = qcount;
1395 /* find the power-of-2 of the number of queue pairs */
1398 while (num_qps && ((1 << pow) < qcount)) {
1403 vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1405 (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1406 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1410 /* TC is not enabled so set the offset to
1411 * default queue and allocate one queue
1414 vsi->tc_config.tc_info[i].qoffset = 0;
1415 vsi->tc_config.tc_info[i].qcount = 1;
1416 vsi->tc_config.tc_info[i].netdev_tc = 0;
1420 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1423 /* Set actual Tx/Rx queue pairs */
1424 vsi->num_queue_pairs = offset;
1426 /* Scheduler section valid can only be set for ADD VSI */
1428 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1430 ctxt->info.up_enable_bits = enabled_tc;
1432 if (vsi->type == I40E_VSI_SRIOV) {
1433 ctxt->info.mapping_flags |=
1434 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1435 for (i = 0; i < vsi->num_queue_pairs; i++)
1436 ctxt->info.queue_mapping[i] =
1437 cpu_to_le16(vsi->base_queue + i);
1439 ctxt->info.mapping_flags |=
1440 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1441 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1443 ctxt->info.valid_sections |= cpu_to_le16(sections);
1447 * i40e_set_rx_mode - NDO callback to set the netdev filters
1448 * @netdev: network interface device structure
1450 static void i40e_set_rx_mode(struct net_device *netdev)
1452 struct i40e_netdev_priv *np = netdev_priv(netdev);
1453 struct i40e_mac_filter *f, *ftmp;
1454 struct i40e_vsi *vsi = np->vsi;
1455 struct netdev_hw_addr *uca;
1456 struct netdev_hw_addr *mca;
1457 struct netdev_hw_addr *ha;
1459 /* add addr if not already in the filter list */
1460 netdev_for_each_uc_addr(uca, netdev) {
1461 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1462 if (i40e_is_vsi_in_vlan(vsi))
1463 i40e_put_mac_in_vlan(vsi, uca->addr,
1466 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1471 netdev_for_each_mc_addr(mca, netdev) {
1472 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1473 if (i40e_is_vsi_in_vlan(vsi))
1474 i40e_put_mac_in_vlan(vsi, mca->addr,
1477 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1482 /* remove filter if not in netdev list */
1483 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1489 if (is_multicast_ether_addr(f->macaddr)) {
1490 netdev_for_each_mc_addr(mca, netdev) {
1491 if (ether_addr_equal(mca->addr, f->macaddr)) {
1497 netdev_for_each_uc_addr(uca, netdev) {
1498 if (ether_addr_equal(uca->addr, f->macaddr)) {
1504 for_each_dev_addr(netdev, ha) {
1505 if (ether_addr_equal(ha->addr, f->macaddr)) {
1513 vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1516 /* check for other flag changes */
1517 if (vsi->current_netdev_flags != vsi->netdev->flags) {
1518 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1519 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1524 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1525 * @vsi: ptr to the VSI
1527 * Push any outstanding VSI filter changes through the AdminQ.
1529 * Returns 0 or error value
1531 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1533 struct i40e_mac_filter *f, *ftmp;
1534 bool promisc_forced_on = false;
1535 bool add_happened = false;
1536 int filter_list_len = 0;
1537 u32 changed_flags = 0;
1538 i40e_status aq_ret = 0;
1544 /* empty array typed pointers, kcalloc later */
1545 struct i40e_aqc_add_macvlan_element_data *add_list;
1546 struct i40e_aqc_remove_macvlan_element_data *del_list;
1548 while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1549 usleep_range(1000, 2000);
1553 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1554 vsi->current_netdev_flags = vsi->netdev->flags;
1557 if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1558 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1560 filter_list_len = pf->hw.aq.asq_buf_size /
1561 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1562 del_list = kcalloc(filter_list_len,
1563 sizeof(struct i40e_aqc_remove_macvlan_element_data),
1568 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1572 if (f->counter != 0)
1577 /* add to delete list */
1578 memcpy(del_list[num_del].mac_addr,
1579 f->macaddr, ETH_ALEN);
1580 del_list[num_del].vlan_tag =
1581 cpu_to_le16((u16)(f->vlan ==
1582 I40E_VLAN_ANY ? 0 : f->vlan));
1584 cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1585 del_list[num_del].flags = cmd_flags;
1588 /* unlink from filter list */
1592 /* flush a full buffer */
1593 if (num_del == filter_list_len) {
1594 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1595 vsi->seid, del_list, num_del,
1598 memset(del_list, 0, sizeof(*del_list));
1601 dev_info(&pf->pdev->dev,
1602 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1604 pf->hw.aq.asq_last_status);
1608 aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1609 del_list, num_del, NULL);
1613 dev_info(&pf->pdev->dev,
1614 "ignoring delete macvlan error, err %d, aq_err %d\n",
1615 aq_ret, pf->hw.aq.asq_last_status);
1621 /* do all the adds now */
1622 filter_list_len = pf->hw.aq.asq_buf_size /
1623 sizeof(struct i40e_aqc_add_macvlan_element_data),
1624 add_list = kcalloc(filter_list_len,
1625 sizeof(struct i40e_aqc_add_macvlan_element_data),
1630 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1634 if (f->counter == 0)
1637 add_happened = true;
1640 /* add to add array */
1641 memcpy(add_list[num_add].mac_addr,
1642 f->macaddr, ETH_ALEN);
1643 add_list[num_add].vlan_tag =
1645 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1646 add_list[num_add].queue_number = 0;
1648 cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1649 add_list[num_add].flags = cpu_to_le16(cmd_flags);
1652 /* flush a full buffer */
1653 if (num_add == filter_list_len) {
1654 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1661 memset(add_list, 0, sizeof(*add_list));
1665 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1666 add_list, num_add, NULL);
1672 if (add_happened && (!aq_ret)) {
1674 } else if (add_happened && (aq_ret)) {
1675 dev_info(&pf->pdev->dev,
1676 "add filter failed, err %d, aq_err %d\n",
1677 aq_ret, pf->hw.aq.asq_last_status);
1678 if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1679 !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1681 promisc_forced_on = true;
1682 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1684 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1689 /* check for changes in promiscuous modes */
1690 if (changed_flags & IFF_ALLMULTI) {
1691 bool cur_multipromisc;
1692 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1693 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1698 dev_info(&pf->pdev->dev,
1699 "set multi promisc failed, err %d, aq_err %d\n",
1700 aq_ret, pf->hw.aq.asq_last_status);
1702 if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1704 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1705 test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1707 aq_ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1711 dev_info(&pf->pdev->dev,
1712 "set uni promisc failed, err %d, aq_err %d\n",
1713 aq_ret, pf->hw.aq.asq_last_status);
1714 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
1718 dev_info(&pf->pdev->dev,
1719 "set brdcast promisc failed, err %d, aq_err %d\n",
1720 aq_ret, pf->hw.aq.asq_last_status);
1723 clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1728 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1729 * @pf: board private structure
1731 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1735 if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1737 pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1739 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
1741 (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1742 i40e_sync_vsi_filters(pf->vsi[v]);
1747 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1748 * @netdev: network interface device structure
1749 * @new_mtu: new value for maximum frame size
1751 * Returns 0 on success, negative on failure
1753 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1755 struct i40e_netdev_priv *np = netdev_priv(netdev);
1756 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
1757 struct i40e_vsi *vsi = np->vsi;
1759 /* MTU < 68 is an error and causes problems on some kernels */
1760 if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1763 netdev_info(netdev, "changing MTU from %d to %d\n",
1764 netdev->mtu, new_mtu);
1765 netdev->mtu = new_mtu;
1766 if (netif_running(netdev))
1767 i40e_vsi_reinit_locked(vsi);
1773 * i40e_ioctl - Access the hwtstamp interface
1774 * @netdev: network interface device structure
1775 * @ifr: interface request data
1776 * @cmd: ioctl command
1778 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
1780 struct i40e_netdev_priv *np = netdev_priv(netdev);
1781 struct i40e_pf *pf = np->vsi->back;
1785 return i40e_ptp_get_ts_config(pf, ifr);
1787 return i40e_ptp_set_ts_config(pf, ifr);
1794 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1795 * @vsi: the vsi being adjusted
1797 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1799 struct i40e_vsi_context ctxt;
1802 if ((vsi->info.valid_sections &
1803 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1804 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1805 return; /* already enabled */
1807 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1808 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1809 I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1811 ctxt.seid = vsi->seid;
1812 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1813 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1815 dev_info(&vsi->back->pdev->dev,
1816 "%s: update vsi failed, aq_err=%d\n",
1817 __func__, vsi->back->hw.aq.asq_last_status);
1822 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1823 * @vsi: the vsi being adjusted
1825 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
1827 struct i40e_vsi_context ctxt;
1830 if ((vsi->info.valid_sections &
1831 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1832 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
1833 I40E_AQ_VSI_PVLAN_EMOD_MASK))
1834 return; /* already disabled */
1836 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1837 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1838 I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
1840 ctxt.seid = vsi->seid;
1841 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1842 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1844 dev_info(&vsi->back->pdev->dev,
1845 "%s: update vsi failed, aq_err=%d\n",
1846 __func__, vsi->back->hw.aq.asq_last_status);
1851 * i40e_vlan_rx_register - Setup or shutdown vlan offload
1852 * @netdev: network interface to be adjusted
1853 * @features: netdev features to test if VLAN offload is enabled or not
1855 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
1857 struct i40e_netdev_priv *np = netdev_priv(netdev);
1858 struct i40e_vsi *vsi = np->vsi;
1860 if (features & NETIF_F_HW_VLAN_CTAG_RX)
1861 i40e_vlan_stripping_enable(vsi);
1863 i40e_vlan_stripping_disable(vsi);
1867 * i40e_vsi_add_vlan - Add vsi membership for given vlan
1868 * @vsi: the vsi being configured
1869 * @vid: vlan id to be added (0 = untagged only , -1 = any)
1871 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
1873 struct i40e_mac_filter *f, *add_f;
1874 bool is_netdev, is_vf;
1876 is_vf = (vsi->type == I40E_VSI_SRIOV);
1877 is_netdev = !!(vsi->netdev);
1880 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
1883 dev_info(&vsi->back->pdev->dev,
1884 "Could not add vlan filter %d for %pM\n",
1885 vid, vsi->netdev->dev_addr);
1890 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1891 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1893 dev_info(&vsi->back->pdev->dev,
1894 "Could not add vlan filter %d for %pM\n",
1900 /* Now if we add a vlan tag, make sure to check if it is the first
1901 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
1902 * with 0, so we now accept untagged and specified tagged traffic
1903 * (and not any taged and untagged)
1906 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
1908 is_vf, is_netdev)) {
1909 i40e_del_filter(vsi, vsi->netdev->dev_addr,
1910 I40E_VLAN_ANY, is_vf, is_netdev);
1911 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
1914 dev_info(&vsi->back->pdev->dev,
1915 "Could not add filter 0 for %pM\n",
1916 vsi->netdev->dev_addr);
1922 /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
1923 if (vid > 0 && !vsi->info.pvid) {
1924 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1925 if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1926 is_vf, is_netdev)) {
1927 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1929 add_f = i40e_add_filter(vsi, f->macaddr,
1930 0, is_vf, is_netdev);
1932 dev_info(&vsi->back->pdev->dev,
1933 "Could not add filter 0 for %pM\n",
1941 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1942 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1945 return i40e_sync_vsi_filters(vsi);
1949 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
1950 * @vsi: the vsi being configured
1951 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
1953 * Return: 0 on success or negative otherwise
1955 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
1957 struct net_device *netdev = vsi->netdev;
1958 struct i40e_mac_filter *f, *add_f;
1959 bool is_vf, is_netdev;
1960 int filter_count = 0;
1962 is_vf = (vsi->type == I40E_VSI_SRIOV);
1963 is_netdev = !!(netdev);
1966 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
1968 list_for_each_entry(f, &vsi->mac_filter_list, list)
1969 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1971 /* go through all the filters for this VSI and if there is only
1972 * vid == 0 it means there are no other filters, so vid 0 must
1973 * be replaced with -1. This signifies that we should from now
1974 * on accept any traffic (with any tag present, or untagged)
1976 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1979 ether_addr_equal(netdev->dev_addr, f->macaddr))
1987 if (!filter_count && is_netdev) {
1988 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
1989 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1992 dev_info(&vsi->back->pdev->dev,
1993 "Could not add filter %d for %pM\n",
1994 I40E_VLAN_ANY, netdev->dev_addr);
1999 if (!filter_count) {
2000 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2001 i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2002 add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2005 dev_info(&vsi->back->pdev->dev,
2006 "Could not add filter %d for %pM\n",
2007 I40E_VLAN_ANY, f->macaddr);
2013 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2014 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2017 return i40e_sync_vsi_filters(vsi);
2021 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2022 * @netdev: network interface to be adjusted
2023 * @vid: vlan id to be added
2025 * net_device_ops implementation for adding vlan ids
2027 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2028 __always_unused __be16 proto, u16 vid)
2030 struct i40e_netdev_priv *np = netdev_priv(netdev);
2031 struct i40e_vsi *vsi = np->vsi;
2037 netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2039 /* If the network stack called us with vid = 0 then
2040 * it is asking to receive priority tagged packets with
2041 * vlan id 0. Our HW receives them by default when configured
2042 * to receive untagged packets so there is no need to add an
2043 * extra filter for vlan 0 tagged packets.
2046 ret = i40e_vsi_add_vlan(vsi, vid);
2048 if (!ret && (vid < VLAN_N_VID))
2049 set_bit(vid, vsi->active_vlans);
2055 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2056 * @netdev: network interface to be adjusted
2057 * @vid: vlan id to be removed
2059 * net_device_ops implementation for removing vlan ids
2061 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2062 __always_unused __be16 proto, u16 vid)
2064 struct i40e_netdev_priv *np = netdev_priv(netdev);
2065 struct i40e_vsi *vsi = np->vsi;
2067 netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2069 /* return code is ignored as there is nothing a user
2070 * can do about failure to remove and a log message was
2071 * already printed from the other function
2073 i40e_vsi_kill_vlan(vsi, vid);
2075 clear_bit(vid, vsi->active_vlans);
2081 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2082 * @vsi: the vsi being brought back up
2084 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2091 i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2093 for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2094 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2099 * i40e_vsi_add_pvid - Add pvid for the VSI
2100 * @vsi: the vsi being adjusted
2101 * @vid: the vlan id to set as a PVID
2103 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2105 struct i40e_vsi_context ctxt;
2108 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2109 vsi->info.pvid = cpu_to_le16(vid);
2110 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2111 I40E_AQ_VSI_PVLAN_INSERT_PVID |
2112 I40E_AQ_VSI_PVLAN_EMOD_STR;
2114 ctxt.seid = vsi->seid;
2115 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
2116 aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2118 dev_info(&vsi->back->pdev->dev,
2119 "%s: update vsi failed, aq_err=%d\n",
2120 __func__, vsi->back->hw.aq.asq_last_status);
2128 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2129 * @vsi: the vsi being adjusted
2131 * Just use the vlan_rx_register() service to put it back to normal
2133 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2135 i40e_vlan_stripping_disable(vsi);
2141 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2142 * @vsi: ptr to the VSI
2144 * If this function returns with an error, then it's possible one or
2145 * more of the rings is populated (while the rest are not). It is the
2146 * callers duty to clean those orphaned rings.
2148 * Return 0 on success, negative on failure
2150 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2154 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2155 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2161 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2162 * @vsi: ptr to the VSI
2164 * Free VSI's transmit software resources
2166 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2173 for (i = 0; i < vsi->num_queue_pairs; i++)
2174 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2175 i40e_free_tx_resources(vsi->tx_rings[i]);
2179 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2180 * @vsi: ptr to the VSI
2182 * If this function returns with an error, then it's possible one or
2183 * more of the rings is populated (while the rest are not). It is the
2184 * callers duty to clean those orphaned rings.
2186 * Return 0 on success, negative on failure
2188 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2192 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2193 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2198 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2199 * @vsi: ptr to the VSI
2201 * Free all receive software resources
2203 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2210 for (i = 0; i < vsi->num_queue_pairs; i++)
2211 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2212 i40e_free_rx_resources(vsi->rx_rings[i]);
2216 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2217 * @ring: The Tx ring to configure
2219 * Configure the Tx descriptor ring in the HMC context.
2221 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2223 struct i40e_vsi *vsi = ring->vsi;
2224 u16 pf_q = vsi->base_queue + ring->queue_index;
2225 struct i40e_hw *hw = &vsi->back->hw;
2226 struct i40e_hmc_obj_txq tx_ctx;
2227 i40e_status err = 0;
2230 /* some ATR related tx ring init */
2231 if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2232 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2233 ring->atr_count = 0;
2235 ring->atr_sample_rate = 0;
2238 /* initialize XPS */
2239 if (ring->q_vector && ring->netdev &&
2240 vsi->tc_config.numtc <= 1 &&
2241 !test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2242 netif_set_xps_queue(ring->netdev,
2243 &ring->q_vector->affinity_mask,
2246 /* clear the context structure first */
2247 memset(&tx_ctx, 0, sizeof(tx_ctx));
2249 tx_ctx.new_context = 1;
2250 tx_ctx.base = (ring->dma / 128);
2251 tx_ctx.qlen = ring->count;
2252 tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2253 I40E_FLAG_FD_ATR_ENABLED));
2254 tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2255 /* FDIR VSI tx ring can still use RS bit and writebacks */
2256 if (vsi->type != I40E_VSI_FDIR)
2257 tx_ctx.head_wb_ena = 1;
2258 tx_ctx.head_wb_addr = ring->dma +
2259 (ring->count * sizeof(struct i40e_tx_desc));
2261 /* As part of VSI creation/update, FW allocates certain
2262 * Tx arbitration queue sets for each TC enabled for
2263 * the VSI. The FW returns the handles to these queue
2264 * sets as part of the response buffer to Add VSI,
2265 * Update VSI, etc. AQ commands. It is expected that
2266 * these queue set handles be associated with the Tx
2267 * queues by the driver as part of the TX queue context
2268 * initialization. This has to be done regardless of
2269 * DCB as by default everything is mapped to TC0.
2271 tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2272 tx_ctx.rdylist_act = 0;
2274 /* clear the context in the HMC */
2275 err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2277 dev_info(&vsi->back->pdev->dev,
2278 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2279 ring->queue_index, pf_q, err);
2283 /* set the context in the HMC */
2284 err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2286 dev_info(&vsi->back->pdev->dev,
2287 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2288 ring->queue_index, pf_q, err);
2292 /* Now associate this queue with this PCI function */
2293 if (vsi->type == I40E_VSI_VMDQ2)
2294 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2296 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2297 qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2298 I40E_QTX_CTL_PF_INDX_MASK);
2299 wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2302 clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2304 /* cache tail off for easier writes later */
2305 ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2311 * i40e_configure_rx_ring - Configure a receive ring context
2312 * @ring: The Rx ring to configure
2314 * Configure the Rx descriptor ring in the HMC context.
2316 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2318 struct i40e_vsi *vsi = ring->vsi;
2319 u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2320 u16 pf_q = vsi->base_queue + ring->queue_index;
2321 struct i40e_hw *hw = &vsi->back->hw;
2322 struct i40e_hmc_obj_rxq rx_ctx;
2323 i40e_status err = 0;
2327 /* clear the context structure first */
2328 memset(&rx_ctx, 0, sizeof(rx_ctx));
2330 ring->rx_buf_len = vsi->rx_buf_len;
2331 ring->rx_hdr_len = vsi->rx_hdr_len;
2333 rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2334 rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2336 rx_ctx.base = (ring->dma / 128);
2337 rx_ctx.qlen = ring->count;
2339 if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2340 set_ring_16byte_desc_enabled(ring);
2346 rx_ctx.dtype = vsi->dtype;
2348 set_ring_ps_enabled(ring);
2349 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2 |
2351 I40E_RX_SPLIT_TCP_UDP |
2354 rx_ctx.hsplit_0 = 0;
2357 rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2358 (chain_len * ring->rx_buf_len));
2359 rx_ctx.tphrdesc_ena = 1;
2360 rx_ctx.tphwdesc_ena = 1;
2361 rx_ctx.tphdata_ena = 1;
2362 rx_ctx.tphhead_ena = 1;
2363 if (hw->revision_id == 0)
2364 rx_ctx.lrxqthresh = 0;
2366 rx_ctx.lrxqthresh = 2;
2367 rx_ctx.crcstrip = 1;
2370 /* set the prefena field to 1 because the manual says to */
2373 /* clear the context in the HMC */
2374 err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2376 dev_info(&vsi->back->pdev->dev,
2377 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2378 ring->queue_index, pf_q, err);
2382 /* set the context in the HMC */
2383 err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2385 dev_info(&vsi->back->pdev->dev,
2386 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2387 ring->queue_index, pf_q, err);
2391 /* cache tail for quicker writes, and clear the reg before use */
2392 ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2393 writel(0, ring->tail);
2395 i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2401 * i40e_vsi_configure_tx - Configure the VSI for Tx
2402 * @vsi: VSI structure describing this set of rings and resources
2404 * Configure the Tx VSI for operation.
2406 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2411 for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2412 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2418 * i40e_vsi_configure_rx - Configure the VSI for Rx
2419 * @vsi: the VSI being configured
2421 * Configure the Rx VSI for operation.
2423 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2428 if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2429 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2430 + ETH_FCS_LEN + VLAN_HLEN;
2432 vsi->max_frame = I40E_RXBUFFER_2048;
2434 /* figure out correct receive buffer length */
2435 switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2436 I40E_FLAG_RX_PS_ENABLED)) {
2437 case I40E_FLAG_RX_1BUF_ENABLED:
2438 vsi->rx_hdr_len = 0;
2439 vsi->rx_buf_len = vsi->max_frame;
2440 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2442 case I40E_FLAG_RX_PS_ENABLED:
2443 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2444 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2445 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2448 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2449 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2450 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2454 /* round up for the chip's needs */
2455 vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2456 (1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2457 vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2458 (1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2460 /* set up individual rings */
2461 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2462 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2468 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2469 * @vsi: ptr to the VSI
2471 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2473 struct i40e_ring *tx_ring, *rx_ring;
2474 u16 qoffset, qcount;
2477 if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED))
2480 for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2481 if (!(vsi->tc_config.enabled_tc & (1 << n)))
2484 qoffset = vsi->tc_config.tc_info[n].qoffset;
2485 qcount = vsi->tc_config.tc_info[n].qcount;
2486 for (i = qoffset; i < (qoffset + qcount); i++) {
2487 rx_ring = vsi->rx_rings[i];
2488 tx_ring = vsi->tx_rings[i];
2489 rx_ring->dcb_tc = n;
2490 tx_ring->dcb_tc = n;
2496 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2497 * @vsi: ptr to the VSI
2499 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2502 i40e_set_rx_mode(vsi->netdev);
2506 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
2507 * @vsi: Pointer to the targeted VSI
2509 * This function replays the hlist on the hw where all the SB Flow Director
2510 * filters were saved.
2512 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
2514 struct i40e_fdir_filter *filter;
2515 struct i40e_pf *pf = vsi->back;
2516 struct hlist_node *node;
2518 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
2521 hlist_for_each_entry_safe(filter, node,
2522 &pf->fdir_filter_list, fdir_node) {
2523 i40e_add_del_fdir(vsi, filter, true);
2528 * i40e_vsi_configure - Set up the VSI for action
2529 * @vsi: the VSI being configured
2531 static int i40e_vsi_configure(struct i40e_vsi *vsi)
2535 i40e_set_vsi_rx_mode(vsi);
2536 i40e_restore_vlan(vsi);
2537 i40e_vsi_config_dcb_rings(vsi);
2538 err = i40e_vsi_configure_tx(vsi);
2540 err = i40e_vsi_configure_rx(vsi);
2546 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2547 * @vsi: the VSI being configured
2549 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2551 struct i40e_pf *pf = vsi->back;
2552 struct i40e_q_vector *q_vector;
2553 struct i40e_hw *hw = &pf->hw;
2559 /* The interrupt indexing is offset by 1 in the PFINT_ITRn
2560 * and PFINT_LNKLSTn registers, e.g.:
2561 * PFINT_ITRn[0..n-1] gets msix-1..msix-n (qpair interrupts)
2563 qp = vsi->base_queue;
2564 vector = vsi->base_vector;
2565 for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
2566 q_vector = vsi->q_vectors[i];
2567 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2568 q_vector->rx.latency_range = I40E_LOW_LATENCY;
2569 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2571 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2572 q_vector->tx.latency_range = I40E_LOW_LATENCY;
2573 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2576 /* Linked list for the queuepairs assigned to this vector */
2577 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2578 for (q = 0; q < q_vector->num_ringpairs; q++) {
2579 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2580 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2581 (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2582 (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2584 << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2586 wr32(hw, I40E_QINT_RQCTL(qp), val);
2588 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2589 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2590 (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2591 ((qp+1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2593 << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2595 /* Terminate the linked list */
2596 if (q == (q_vector->num_ringpairs - 1))
2597 val |= (I40E_QUEUE_END_OF_LIST
2598 << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2600 wr32(hw, I40E_QINT_TQCTL(qp), val);
2609 * i40e_enable_misc_int_causes - enable the non-queue interrupts
2610 * @hw: ptr to the hardware info
2612 static void i40e_enable_misc_int_causes(struct i40e_hw *hw)
2616 /* clear things first */
2617 wr32(hw, I40E_PFINT_ICR0_ENA, 0); /* disable all */
2618 rd32(hw, I40E_PFINT_ICR0); /* read to clear */
2620 val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK |
2621 I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK |
2622 I40E_PFINT_ICR0_ENA_GRST_MASK |
2623 I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2624 I40E_PFINT_ICR0_ENA_GPIO_MASK |
2625 I40E_PFINT_ICR0_ENA_TIMESYNC_MASK |
2626 I40E_PFINT_ICR0_ENA_HMC_ERR_MASK |
2627 I40E_PFINT_ICR0_ENA_VFLR_MASK |
2628 I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2630 wr32(hw, I40E_PFINT_ICR0_ENA, val);
2632 /* SW_ITR_IDX = 0, but don't change INTENA */
2633 wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
2634 I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
2636 /* OTHER_ITR_IDX = 0 */
2637 wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2641 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2642 * @vsi: the VSI being configured
2644 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2646 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
2647 struct i40e_pf *pf = vsi->back;
2648 struct i40e_hw *hw = &pf->hw;
2651 /* set the ITR configuration */
2652 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2653 q_vector->rx.latency_range = I40E_LOW_LATENCY;
2654 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2655 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2656 q_vector->tx.latency_range = I40E_LOW_LATENCY;
2657 wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2659 i40e_enable_misc_int_causes(hw);
2661 /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2662 wr32(hw, I40E_PFINT_LNKLST0, 0);
2664 /* Associate the queue pair to the vector and enable the queue int */
2665 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2666 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2667 (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2669 wr32(hw, I40E_QINT_RQCTL(0), val);
2671 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2672 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2673 (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2675 wr32(hw, I40E_QINT_TQCTL(0), val);
2680 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
2681 * @pf: board private structure
2683 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
2685 struct i40e_hw *hw = &pf->hw;
2687 wr32(hw, I40E_PFINT_DYN_CTL0,
2688 I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2693 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2694 * @pf: board private structure
2696 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2698 struct i40e_hw *hw = &pf->hw;
2701 val = I40E_PFINT_DYN_CTL0_INTENA_MASK |
2702 I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2703 (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2705 wr32(hw, I40E_PFINT_DYN_CTL0, val);
2710 * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2711 * @vsi: pointer to a vsi
2712 * @vector: enable a particular Hw Interrupt vector
2714 void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2716 struct i40e_pf *pf = vsi->back;
2717 struct i40e_hw *hw = &pf->hw;
2720 val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2721 I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2722 (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2723 wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2724 /* skip the flush */
2728 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2729 * @irq: interrupt number
2730 * @data: pointer to a q_vector
2732 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2734 struct i40e_q_vector *q_vector = data;
2736 if (!q_vector->tx.ring && !q_vector->rx.ring)
2739 napi_schedule(&q_vector->napi);
2745 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
2746 * @vsi: the VSI being configured
2747 * @basename: name for the vector
2749 * Allocates MSI-X vectors and requests interrupts from the kernel.
2751 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
2753 int q_vectors = vsi->num_q_vectors;
2754 struct i40e_pf *pf = vsi->back;
2755 int base = vsi->base_vector;
2760 for (vector = 0; vector < q_vectors; vector++) {
2761 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
2763 if (q_vector->tx.ring && q_vector->rx.ring) {
2764 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2765 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2767 } else if (q_vector->rx.ring) {
2768 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2769 "%s-%s-%d", basename, "rx", rx_int_idx++);
2770 } else if (q_vector->tx.ring) {
2771 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2772 "%s-%s-%d", basename, "tx", tx_int_idx++);
2774 /* skip this unused q_vector */
2777 err = request_irq(pf->msix_entries[base + vector].vector,
2783 dev_info(&pf->pdev->dev,
2784 "%s: request_irq failed, error: %d\n",
2786 goto free_queue_irqs;
2788 /* assign the mask for this irq */
2789 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2790 &q_vector->affinity_mask);
2793 vsi->irqs_ready = true;
2799 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2801 free_irq(pf->msix_entries[base + vector].vector,
2802 &(vsi->q_vectors[vector]));
2808 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
2809 * @vsi: the VSI being un-configured
2811 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
2813 struct i40e_pf *pf = vsi->back;
2814 struct i40e_hw *hw = &pf->hw;
2815 int base = vsi->base_vector;
2818 for (i = 0; i < vsi->num_queue_pairs; i++) {
2819 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
2820 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
2823 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2824 for (i = vsi->base_vector;
2825 i < (vsi->num_q_vectors + vsi->base_vector); i++)
2826 wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
2829 for (i = 0; i < vsi->num_q_vectors; i++)
2830 synchronize_irq(pf->msix_entries[i + base].vector);
2832 /* Legacy and MSI mode - this stops all interrupt handling */
2833 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
2834 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
2836 synchronize_irq(pf->pdev->irq);
2841 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
2842 * @vsi: the VSI being configured
2844 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
2846 struct i40e_pf *pf = vsi->back;
2849 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2850 for (i = vsi->base_vector;
2851 i < (vsi->num_q_vectors + vsi->base_vector); i++)
2852 i40e_irq_dynamic_enable(vsi, i);
2854 i40e_irq_dynamic_enable_icr0(pf);
2857 i40e_flush(&pf->hw);
2862 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
2863 * @pf: board private structure
2865 static void i40e_stop_misc_vector(struct i40e_pf *pf)
2868 wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
2869 i40e_flush(&pf->hw);
2873 * i40e_intr - MSI/Legacy and non-queue interrupt handler
2874 * @irq: interrupt number
2875 * @data: pointer to a q_vector
2877 * This is the handler used for all MSI/Legacy interrupts, and deals
2878 * with both queue and non-queue interrupts. This is also used in
2879 * MSIX mode to handle the non-queue interrupts.
2881 static irqreturn_t i40e_intr(int irq, void *data)
2883 struct i40e_pf *pf = (struct i40e_pf *)data;
2884 struct i40e_hw *hw = &pf->hw;
2885 irqreturn_t ret = IRQ_NONE;
2886 u32 icr0, icr0_remaining;
2889 icr0 = rd32(hw, I40E_PFINT_ICR0);
2890 ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
2892 /* if sharing a legacy IRQ, we might get called w/o an intr pending */
2893 if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
2896 /* if interrupt but no bits showing, must be SWINT */
2897 if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
2898 (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
2901 /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
2902 if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
2904 /* temporarily disable queue cause for NAPI processing */
2905 u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
2906 qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
2907 wr32(hw, I40E_QINT_RQCTL(0), qval);
2909 qval = rd32(hw, I40E_QINT_TQCTL(0));
2910 qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
2911 wr32(hw, I40E_QINT_TQCTL(0), qval);
2913 if (!test_bit(__I40E_DOWN, &pf->state))
2914 napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0]->napi);
2917 if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
2918 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2919 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
2922 if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
2923 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
2924 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
2927 if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
2928 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
2929 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
2932 if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
2933 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
2934 set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
2935 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
2936 val = rd32(hw, I40E_GLGEN_RSTAT);
2937 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
2938 >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
2939 if (val == I40E_RESET_CORER) {
2941 } else if (val == I40E_RESET_GLOBR) {
2943 } else if (val == I40E_RESET_EMPR) {
2945 set_bit(__I40E_EMP_RESET_REQUESTED, &pf->state);
2949 if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
2950 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
2951 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
2954 if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
2955 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
2957 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
2958 icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
2959 i40e_ptp_tx_hwtstamp(pf);
2963 /* If a critical error is pending we have no choice but to reset the
2965 * Report and mask out any remaining unexpected interrupts.
2967 icr0_remaining = icr0 & ena_mask;
2968 if (icr0_remaining) {
2969 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
2971 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
2972 (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
2973 (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
2974 dev_info(&pf->pdev->dev, "device will be reset\n");
2975 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2976 i40e_service_event_schedule(pf);
2978 ena_mask &= ~icr0_remaining;
2983 /* re-enable interrupt causes */
2984 wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
2985 if (!test_bit(__I40E_DOWN, &pf->state)) {
2986 i40e_service_event_schedule(pf);
2987 i40e_irq_dynamic_enable_icr0(pf);
2994 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
2995 * @tx_ring: tx ring to clean
2996 * @budget: how many cleans we're allowed
2998 * Returns true if there's any budget left (e.g. the clean is finished)
3000 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3002 struct i40e_vsi *vsi = tx_ring->vsi;
3003 u16 i = tx_ring->next_to_clean;
3004 struct i40e_tx_buffer *tx_buf;
3005 struct i40e_tx_desc *tx_desc;
3007 tx_buf = &tx_ring->tx_bi[i];
3008 tx_desc = I40E_TX_DESC(tx_ring, i);
3009 i -= tx_ring->count;
3012 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3014 /* if next_to_watch is not set then there is no work pending */
3018 /* prevent any other reads prior to eop_desc */
3019 read_barrier_depends();
3021 /* if the descriptor isn't done, no work yet to do */
3022 if (!(eop_desc->cmd_type_offset_bsz &
3023 cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3026 /* clear next_to_watch to prevent false hangs */
3027 tx_buf->next_to_watch = NULL;
3029 /* unmap skb header data */
3030 dma_unmap_single(tx_ring->dev,
3031 dma_unmap_addr(tx_buf, dma),
3032 dma_unmap_len(tx_buf, len),
3035 dma_unmap_len_set(tx_buf, len, 0);
3038 /* move to the next desc and buffer to clean */
3043 i -= tx_ring->count;
3044 tx_buf = tx_ring->tx_bi;
3045 tx_desc = I40E_TX_DESC(tx_ring, 0);
3048 /* update budget accounting */
3050 } while (likely(budget));
3052 i += tx_ring->count;
3053 tx_ring->next_to_clean = i;
3055 if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED) {
3056 i40e_irq_dynamic_enable(vsi,
3057 tx_ring->q_vector->v_idx + vsi->base_vector);
3063 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3064 * @irq: interrupt number
3065 * @data: pointer to a q_vector
3067 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3069 struct i40e_q_vector *q_vector = data;
3070 struct i40e_vsi *vsi;
3072 if (!q_vector->tx.ring)
3075 vsi = q_vector->tx.ring->vsi;
3076 i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3082 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3083 * @vsi: the VSI being configured
3084 * @v_idx: vector index
3085 * @qp_idx: queue pair index
3087 static void map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3089 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3090 struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3091 struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3093 tx_ring->q_vector = q_vector;
3094 tx_ring->next = q_vector->tx.ring;
3095 q_vector->tx.ring = tx_ring;
3096 q_vector->tx.count++;
3098 rx_ring->q_vector = q_vector;
3099 rx_ring->next = q_vector->rx.ring;
3100 q_vector->rx.ring = rx_ring;
3101 q_vector->rx.count++;
3105 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3106 * @vsi: the VSI being configured
3108 * This function maps descriptor rings to the queue-specific vectors
3109 * we were allotted through the MSI-X enabling code. Ideally, we'd have
3110 * one vector per queue pair, but on a constrained vector budget, we
3111 * group the queue pairs as "efficiently" as possible.
3113 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3115 int qp_remaining = vsi->num_queue_pairs;
3116 int q_vectors = vsi->num_q_vectors;
3121 /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3122 * group them so there are multiple queues per vector.
3124 for (; v_start < q_vectors && qp_remaining; v_start++) {
3125 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3127 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3129 q_vector->num_ringpairs = num_ringpairs;
3131 q_vector->rx.count = 0;
3132 q_vector->tx.count = 0;
3133 q_vector->rx.ring = NULL;
3134 q_vector->tx.ring = NULL;
3136 while (num_ringpairs--) {
3137 map_vector_to_qp(vsi, v_start, qp_idx);
3145 * i40e_vsi_request_irq - Request IRQ from the OS
3146 * @vsi: the VSI being configured
3147 * @basename: name for the vector
3149 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3151 struct i40e_pf *pf = vsi->back;
3154 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3155 err = i40e_vsi_request_irq_msix(vsi, basename);
3156 else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3157 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3158 pf->misc_int_name, pf);
3160 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3161 pf->misc_int_name, pf);
3164 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3169 #ifdef CONFIG_NET_POLL_CONTROLLER
3171 * i40e_netpoll - A Polling 'interrupt'handler
3172 * @netdev: network interface device structure
3174 * This is used by netconsole to send skbs without having to re-enable
3175 * interrupts. It's not called while the normal interrupt routine is executing.
3177 static void i40e_netpoll(struct net_device *netdev)
3179 struct i40e_netdev_priv *np = netdev_priv(netdev);
3180 struct i40e_vsi *vsi = np->vsi;
3181 struct i40e_pf *pf = vsi->back;
3184 /* if interface is down do nothing */
3185 if (test_bit(__I40E_DOWN, &vsi->state))
3188 pf->flags |= I40E_FLAG_IN_NETPOLL;
3189 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3190 for (i = 0; i < vsi->num_q_vectors; i++)
3191 i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3193 i40e_intr(pf->pdev->irq, netdev);
3195 pf->flags &= ~I40E_FLAG_IN_NETPOLL;
3200 * i40e_vsi_control_tx - Start or stop a VSI's rings
3201 * @vsi: the VSI being configured
3202 * @enable: start or stop the rings
3204 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3206 struct i40e_pf *pf = vsi->back;
3207 struct i40e_hw *hw = &pf->hw;
3211 pf_q = vsi->base_queue;
3212 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3214 /* warn the TX unit of coming changes */
3215 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3219 for (j = 0; j < 50; j++) {
3220 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3221 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3222 ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3224 usleep_range(1000, 2000);
3226 /* Skip if the queue is already in the requested state */
3227 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3230 /* turn on/off the queue */
3232 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3233 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3235 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3238 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3240 /* wait for the change to finish */
3241 for (j = 0; j < 10; j++) {
3242 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3243 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3249 dev_info(&pf->pdev->dev, "Tx ring %d %sable timeout\n",
3250 pf_q, (enable ? "en" : "dis"));
3255 if (hw->revision_id == 0)
3262 * i40e_vsi_control_rx - Start or stop a VSI's rings
3263 * @vsi: the VSI being configured
3264 * @enable: start or stop the rings
3266 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3268 struct i40e_pf *pf = vsi->back;
3269 struct i40e_hw *hw = &pf->hw;
3273 pf_q = vsi->base_queue;
3274 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3275 for (j = 0; j < 50; j++) {
3276 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3277 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3278 ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3280 usleep_range(1000, 2000);
3283 /* Skip if the queue is already in the requested state */
3284 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3287 /* turn on/off the queue */
3289 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3291 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3292 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3294 /* wait for the change to finish */
3295 for (j = 0; j < 10; j++) {
3296 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3298 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3304 dev_info(&pf->pdev->dev, "Rx ring %d %sable timeout\n",
3305 pf_q, (enable ? "en" : "dis"));
3314 * i40e_vsi_control_rings - Start or stop a VSI's rings
3315 * @vsi: the VSI being configured
3316 * @enable: start or stop the rings
3318 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3322 /* do rx first for enable and last for disable */
3324 ret = i40e_vsi_control_rx(vsi, request);
3327 ret = i40e_vsi_control_tx(vsi, request);
3329 /* Ignore return value, we need to shutdown whatever we can */
3330 i40e_vsi_control_tx(vsi, request);
3331 i40e_vsi_control_rx(vsi, request);
3338 * i40e_vsi_free_irq - Free the irq association with the OS
3339 * @vsi: the VSI being configured
3341 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3343 struct i40e_pf *pf = vsi->back;
3344 struct i40e_hw *hw = &pf->hw;
3345 int base = vsi->base_vector;
3349 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3350 if (!vsi->q_vectors)
3353 if (!vsi->irqs_ready)
3356 vsi->irqs_ready = false;
3357 for (i = 0; i < vsi->num_q_vectors; i++) {
3358 u16 vector = i + base;
3360 /* free only the irqs that were actually requested */
3361 if (!vsi->q_vectors[i] ||
3362 !vsi->q_vectors[i]->num_ringpairs)
3365 /* clear the affinity_mask in the IRQ descriptor */
3366 irq_set_affinity_hint(pf->msix_entries[vector].vector,
3368 free_irq(pf->msix_entries[vector].vector,
3371 /* Tear down the interrupt queue link list
3373 * We know that they come in pairs and always
3374 * the Rx first, then the Tx. To clear the
3375 * link list, stick the EOL value into the
3376 * next_q field of the registers.
3378 val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3379 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3380 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3381 val |= I40E_QUEUE_END_OF_LIST
3382 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3383 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3385 while (qp != I40E_QUEUE_END_OF_LIST) {
3388 val = rd32(hw, I40E_QINT_RQCTL(qp));
3390 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
3391 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3392 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3393 I40E_QINT_RQCTL_INTEVENT_MASK);
3395 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3396 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3398 wr32(hw, I40E_QINT_RQCTL(qp), val);
3400 val = rd32(hw, I40E_QINT_TQCTL(qp));
3402 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3403 >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3405 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
3406 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3407 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3408 I40E_QINT_TQCTL_INTEVENT_MASK);
3410 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3411 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3413 wr32(hw, I40E_QINT_TQCTL(qp), val);
3418 free_irq(pf->pdev->irq, pf);
3420 val = rd32(hw, I40E_PFINT_LNKLST0);
3421 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3422 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3423 val |= I40E_QUEUE_END_OF_LIST
3424 << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3425 wr32(hw, I40E_PFINT_LNKLST0, val);
3427 val = rd32(hw, I40E_QINT_RQCTL(qp));
3428 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
3429 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3430 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3431 I40E_QINT_RQCTL_INTEVENT_MASK);
3433 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3434 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3436 wr32(hw, I40E_QINT_RQCTL(qp), val);
3438 val = rd32(hw, I40E_QINT_TQCTL(qp));
3440 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
3441 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3442 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3443 I40E_QINT_TQCTL_INTEVENT_MASK);
3445 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3446 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3448 wr32(hw, I40E_QINT_TQCTL(qp), val);
3453 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
3454 * @vsi: the VSI being configured
3455 * @v_idx: Index of vector to be freed
3457 * This function frees the memory allocated to the q_vector. In addition if
3458 * NAPI is enabled it will delete any references to the NAPI struct prior
3459 * to freeing the q_vector.
3461 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
3463 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3464 struct i40e_ring *ring;
3469 /* disassociate q_vector from rings */
3470 i40e_for_each_ring(ring, q_vector->tx)
3471 ring->q_vector = NULL;
3473 i40e_for_each_ring(ring, q_vector->rx)
3474 ring->q_vector = NULL;
3476 /* only VSI w/ an associated netdev is set up w/ NAPI */
3478 netif_napi_del(&q_vector->napi);
3480 vsi->q_vectors[v_idx] = NULL;
3482 kfree_rcu(q_vector, rcu);
3486 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3487 * @vsi: the VSI being un-configured
3489 * This frees the memory allocated to the q_vectors and
3490 * deletes references to the NAPI struct.
3492 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3496 for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
3497 i40e_free_q_vector(vsi, v_idx);
3501 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3502 * @pf: board private structure
3504 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3506 /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3507 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3508 pci_disable_msix(pf->pdev);
3509 kfree(pf->msix_entries);
3510 pf->msix_entries = NULL;
3511 } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3512 pci_disable_msi(pf->pdev);
3514 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3518 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3519 * @pf: board private structure
3521 * We go through and clear interrupt specific resources and reset the structure
3522 * to pre-load conditions
3524 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3528 i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3529 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
3531 i40e_vsi_free_q_vectors(pf->vsi[i]);
3532 i40e_reset_interrupt_capability(pf);
3536 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3537 * @vsi: the VSI being configured
3539 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3546 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3547 napi_enable(&vsi->q_vectors[q_idx]->napi);
3551 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3552 * @vsi: the VSI being configured
3554 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3561 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3562 napi_disable(&vsi->q_vectors[q_idx]->napi);
3566 * i40e_vsi_close - Shut down a VSI
3567 * @vsi: the vsi to be quelled
3569 static void i40e_vsi_close(struct i40e_vsi *vsi)
3571 if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
3573 i40e_vsi_free_irq(vsi);
3574 i40e_vsi_free_tx_resources(vsi);
3575 i40e_vsi_free_rx_resources(vsi);
3579 * i40e_quiesce_vsi - Pause a given VSI
3580 * @vsi: the VSI being paused
3582 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3584 if (test_bit(__I40E_DOWN, &vsi->state))
3587 set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3588 if (vsi->netdev && netif_running(vsi->netdev)) {
3589 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3591 i40e_vsi_close(vsi);
3596 * i40e_unquiesce_vsi - Resume a given VSI
3597 * @vsi: the VSI being resumed
3599 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3601 if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3604 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3605 if (vsi->netdev && netif_running(vsi->netdev))
3606 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3608 i40e_vsi_open(vsi); /* this clears the DOWN bit */
3612 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3615 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3619 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3621 i40e_quiesce_vsi(pf->vsi[v]);
3626 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3629 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3633 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3635 i40e_unquiesce_vsi(pf->vsi[v]);
3640 * i40e_dcb_get_num_tc - Get the number of TCs from DCBx config
3641 * @dcbcfg: the corresponding DCBx configuration structure
3643 * Return the number of TCs from given DCBx configuration
3645 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
3650 /* Scan the ETS Config Priority Table to find
3651 * traffic class enabled for a given priority
3652 * and use the traffic class index to get the
3653 * number of traffic classes enabled
3655 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3656 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
3657 num_tc = dcbcfg->etscfg.prioritytable[i];
3660 /* Traffic class index starts from zero so
3661 * increment to return the actual count
3667 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
3668 * @dcbcfg: the corresponding DCBx configuration structure
3670 * Query the current DCB configuration and return the number of
3671 * traffic classes enabled from the given DCBX config
3673 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
3675 u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
3679 for (i = 0; i < num_tc; i++)
3680 enabled_tc |= 1 << i;
3686 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
3687 * @pf: PF being queried
3689 * Return number of traffic classes enabled for the given PF
3691 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
3693 struct i40e_hw *hw = &pf->hw;
3696 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3698 /* If DCB is not enabled then always in single TC */
3699 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3702 /* MFP mode return count of enabled TCs for this PF */
3703 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3704 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3705 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3706 if (enabled_tc & (1 << i))
3712 /* SFP mode will be enabled for all TCs on port */
3713 return i40e_dcb_get_num_tc(dcbcfg);
3717 * i40e_pf_get_default_tc - Get bitmap for first enabled TC
3718 * @pf: PF being queried
3720 * Return a bitmap for first enabled traffic class for this PF.
3722 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
3724 u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3728 return 0x1; /* TC0 */
3730 /* Find the first enabled TC */
3731 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3732 if (enabled_tc & (1 << i))
3740 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
3741 * @pf: PF being queried
3743 * Return a bitmap for enabled traffic classes for this PF.
3745 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
3747 /* If DCB is not enabled for this PF then just return default TC */
3748 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3749 return i40e_pf_get_default_tc(pf);
3751 /* MFP mode will have enabled TCs set by FW */
3752 if (pf->flags & I40E_FLAG_MFP_ENABLED)
3753 return pf->hw.func_caps.enabled_tcmap;
3755 /* SFP mode we want PF to be enabled for all TCs */
3756 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
3760 * i40e_vsi_get_bw_info - Query VSI BW Information
3761 * @vsi: the VSI being queried
3763 * Returns 0 on success, negative value on failure
3765 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
3767 struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
3768 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
3769 struct i40e_pf *pf = vsi->back;
3770 struct i40e_hw *hw = &pf->hw;
3775 /* Get the VSI level BW configuration */
3776 aq_ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
3778 dev_info(&pf->pdev->dev,
3779 "couldn't get pf vsi bw config, err %d, aq_err %d\n",
3780 aq_ret, pf->hw.aq.asq_last_status);
3784 /* Get the VSI level BW configuration per TC */
3785 aq_ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
3788 dev_info(&pf->pdev->dev,
3789 "couldn't get pf vsi ets bw config, err %d, aq_err %d\n",
3790 aq_ret, pf->hw.aq.asq_last_status);
3794 if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
3795 dev_info(&pf->pdev->dev,
3796 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
3797 bw_config.tc_valid_bits,
3798 bw_ets_config.tc_valid_bits);
3799 /* Still continuing */
3802 vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
3803 vsi->bw_max_quanta = bw_config.max_bw;
3804 tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
3805 (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
3806 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3807 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
3808 vsi->bw_ets_limit_credits[i] =
3809 le16_to_cpu(bw_ets_config.credits[i]);
3810 /* 3 bits out of 4 for each TC */
3811 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
3818 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
3819 * @vsi: the VSI being configured
3820 * @enabled_tc: TC bitmap
3821 * @bw_credits: BW shared credits per TC
3823 * Returns 0 on success, negative value on failure
3825 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
3828 struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
3832 bw_data.tc_valid_bits = enabled_tc;
3833 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3834 bw_data.tc_bw_credits[i] = bw_share[i];
3836 aq_ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
3839 dev_info(&vsi->back->pdev->dev,
3840 "AQ command Config VSI BW allocation per TC failed = %d\n",
3841 vsi->back->hw.aq.asq_last_status);
3845 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3846 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
3852 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
3853 * @vsi: the VSI being configured
3854 * @enabled_tc: TC map to be enabled
3857 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3859 struct net_device *netdev = vsi->netdev;
3860 struct i40e_pf *pf = vsi->back;
3861 struct i40e_hw *hw = &pf->hw;
3864 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3870 netdev_reset_tc(netdev);
3874 /* Set up actual enabled TCs on the VSI */
3875 if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
3878 /* set per TC queues for the VSI */
3879 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3880 /* Only set TC queues for enabled tcs
3882 * e.g. For a VSI that has TC0 and TC3 enabled the
3883 * enabled_tc bitmap would be 0x00001001; the driver
3884 * will set the numtc for netdev as 2 that will be
3885 * referenced by the netdev layer as TC 0 and 1.
3887 if (vsi->tc_config.enabled_tc & (1 << i))
3888 netdev_set_tc_queue(netdev,
3889 vsi->tc_config.tc_info[i].netdev_tc,
3890 vsi->tc_config.tc_info[i].qcount,
3891 vsi->tc_config.tc_info[i].qoffset);
3894 /* Assign UP2TC map for the VSI */
3895 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3896 /* Get the actual TC# for the UP */
3897 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
3898 /* Get the mapped netdev TC# for the UP */
3899 netdev_tc = vsi->tc_config.tc_info[ets_tc].netdev_tc;
3900 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3905 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
3906 * @vsi: the VSI being configured
3907 * @ctxt: the ctxt buffer returned from AQ VSI update param command
3909 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
3910 struct i40e_vsi_context *ctxt)
3912 /* copy just the sections touched not the entire info
3913 * since not all sections are valid as returned by
3916 vsi->info.mapping_flags = ctxt->info.mapping_flags;
3917 memcpy(&vsi->info.queue_mapping,
3918 &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
3919 memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
3920 sizeof(vsi->info.tc_mapping));
3924 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
3925 * @vsi: VSI to be configured
3926 * @enabled_tc: TC bitmap
3928 * This configures a particular VSI for TCs that are mapped to the
3929 * given TC bitmap. It uses default bandwidth share for TCs across
3930 * VSIs to configure TC for a particular VSI.
3933 * It is expected that the VSI queues have been quisced before calling
3936 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3938 u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
3939 struct i40e_vsi_context ctxt;
3943 /* Check if enabled_tc is same as existing or new TCs */
3944 if (vsi->tc_config.enabled_tc == enabled_tc)
3947 /* Enable ETS TCs with equal BW Share for now across all VSIs */
3948 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3949 if (enabled_tc & (1 << i))
3953 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
3955 dev_info(&vsi->back->pdev->dev,
3956 "Failed configuring TC map %d for VSI %d\n",
3957 enabled_tc, vsi->seid);
3961 /* Update Queue Pairs Mapping for currently enabled UPs */
3962 ctxt.seid = vsi->seid;
3963 ctxt.pf_num = vsi->back->hw.pf_id;
3965 ctxt.uplink_seid = vsi->uplink_seid;
3966 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
3967 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
3969 /* Update the VSI after updating the VSI queue-mapping information */
3970 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
3972 dev_info(&vsi->back->pdev->dev,
3973 "update vsi failed, aq_err=%d\n",
3974 vsi->back->hw.aq.asq_last_status);
3977 /* update the local VSI info with updated queue map */
3978 i40e_vsi_update_queue_map(vsi, &ctxt);
3979 vsi->info.valid_sections = 0;
3981 /* Update current VSI BW information */
3982 ret = i40e_vsi_get_bw_info(vsi);
3984 dev_info(&vsi->back->pdev->dev,
3985 "Failed updating vsi bw info, aq_err=%d\n",
3986 vsi->back->hw.aq.asq_last_status);
3990 /* Update the netdev TC setup */
3991 i40e_vsi_config_netdev_tc(vsi, enabled_tc);
3997 * i40e_veb_config_tc - Configure TCs for given VEB
3999 * @enabled_tc: TC bitmap
4001 * Configures given TC bitmap for VEB (switching) element
4003 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4005 struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4006 struct i40e_pf *pf = veb->pf;
4010 /* No TCs or already enabled TCs just return */
4011 if (!enabled_tc || veb->enabled_tc == enabled_tc)
4014 bw_data.tc_valid_bits = enabled_tc;
4015 /* bw_data.absolute_credits is not set (relative) */
4017 /* Enable ETS TCs with equal BW Share for now */
4018 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4019 if (enabled_tc & (1 << i))
4020 bw_data.tc_bw_share_credits[i] = 1;
4023 ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4026 dev_info(&pf->pdev->dev,
4027 "veb bw config failed, aq_err=%d\n",
4028 pf->hw.aq.asq_last_status);
4032 /* Update the BW information */
4033 ret = i40e_veb_get_bw_info(veb);
4035 dev_info(&pf->pdev->dev,
4036 "Failed getting veb bw config, aq_err=%d\n",
4037 pf->hw.aq.asq_last_status);
4044 #ifdef CONFIG_I40E_DCB
4046 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4049 * Reconfigure VEB/VSIs on a given PF; it is assumed that
4050 * the caller would've quiesce all the VSIs before calling
4053 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4059 /* Enable the TCs available on PF to all VEBs */
4060 tc_map = i40e_pf_get_tc_map(pf);
4061 for (v = 0; v < I40E_MAX_VEB; v++) {
4064 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4066 dev_info(&pf->pdev->dev,
4067 "Failed configuring TC for VEB seid=%d\n",
4069 /* Will try to configure as many components */
4073 /* Update each VSI */
4074 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4078 /* - Enable all TCs for the LAN VSI
4079 * - For all others keep them at TC0 for now
4081 if (v == pf->lan_vsi)
4082 tc_map = i40e_pf_get_tc_map(pf);
4084 tc_map = i40e_pf_get_default_tc(pf);
4086 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4088 dev_info(&pf->pdev->dev,
4089 "Failed configuring TC for VSI seid=%d\n",
4091 /* Will try to configure as many components */
4093 /* Re-configure VSI vectors based on updated TC map */
4094 i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4095 if (pf->vsi[v]->netdev)
4096 i40e_dcbnl_set_all(pf->vsi[v]);
4102 * i40e_init_pf_dcb - Initialize DCB configuration
4103 * @pf: PF being configured
4105 * Query the current DCB configuration and cache it
4106 * in the hardware structure
4108 static int i40e_init_pf_dcb(struct i40e_pf *pf)
4110 struct i40e_hw *hw = &pf->hw;
4113 if (pf->hw.func_caps.npar_enable)
4116 /* Get the initial DCB configuration */
4117 err = i40e_init_dcb(hw);
4119 /* Device/Function is not DCBX capable */
4120 if ((!hw->func_caps.dcb) ||
4121 (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
4122 dev_info(&pf->pdev->dev,
4123 "DCBX offload is not supported or is disabled for this PF.\n");
4125 if (pf->flags & I40E_FLAG_MFP_ENABLED)
4129 /* When status is not DISABLED then DCBX in FW */
4130 pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
4131 DCB_CAP_DCBX_VER_IEEE;
4132 pf->flags |= I40E_FLAG_DCB_ENABLED;
4135 dev_info(&pf->pdev->dev, "AQ Querying DCB configuration failed: %d\n",
4136 pf->hw.aq.asq_last_status);
4142 #endif /* CONFIG_I40E_DCB */
4143 #define SPEED_SIZE 14
4146 * i40e_print_link_message - print link up or down
4147 * @vsi: the VSI for which link needs a message
4149 static void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
4151 char speed[SPEED_SIZE] = "Unknown";
4152 char fc[FC_SIZE] = "RX/TX";
4155 netdev_info(vsi->netdev, "NIC Link is Down\n");
4159 switch (vsi->back->hw.phy.link_info.link_speed) {
4160 case I40E_LINK_SPEED_40GB:
4161 strncpy(speed, "40 Gbps", SPEED_SIZE);
4163 case I40E_LINK_SPEED_10GB:
4164 strncpy(speed, "10 Gbps", SPEED_SIZE);
4166 case I40E_LINK_SPEED_1GB:
4167 strncpy(speed, "1000 Mbps", SPEED_SIZE);
4173 switch (vsi->back->hw.fc.current_mode) {
4175 strncpy(fc, "RX/TX", FC_SIZE);
4177 case I40E_FC_TX_PAUSE:
4178 strncpy(fc, "TX", FC_SIZE);
4180 case I40E_FC_RX_PAUSE:
4181 strncpy(fc, "RX", FC_SIZE);
4184 strncpy(fc, "None", FC_SIZE);
4188 netdev_info(vsi->netdev, "NIC Link is Up %s Full Duplex, Flow Control: %s\n",
4193 * i40e_up_complete - Finish the last steps of bringing up a connection
4194 * @vsi: the VSI being configured
4196 static int i40e_up_complete(struct i40e_vsi *vsi)
4198 struct i40e_pf *pf = vsi->back;
4201 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4202 i40e_vsi_configure_msix(vsi);
4204 i40e_configure_msi_and_legacy(vsi);
4207 err = i40e_vsi_control_rings(vsi, true);
4211 clear_bit(__I40E_DOWN, &vsi->state);
4212 i40e_napi_enable_all(vsi);
4213 i40e_vsi_enable_irq(vsi);
4215 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
4217 i40e_print_link_message(vsi, true);
4218 netif_tx_start_all_queues(vsi->netdev);
4219 netif_carrier_on(vsi->netdev);
4220 } else if (vsi->netdev) {
4221 i40e_print_link_message(vsi, false);
4224 /* replay FDIR SB filters */
4225 if (vsi->type == I40E_VSI_FDIR)
4226 i40e_fdir_filter_restore(vsi);
4227 i40e_service_event_schedule(pf);
4233 * i40e_vsi_reinit_locked - Reset the VSI
4234 * @vsi: the VSI being configured
4236 * Rebuild the ring structs after some configuration
4237 * has changed, e.g. MTU size.
4239 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
4241 struct i40e_pf *pf = vsi->back;
4243 WARN_ON(in_interrupt());
4244 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
4245 usleep_range(1000, 2000);
4248 /* Give a VF some time to respond to the reset. The
4249 * two second wait is based upon the watchdog cycle in
4252 if (vsi->type == I40E_VSI_SRIOV)
4255 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
4259 * i40e_up - Bring the connection back up after being down
4260 * @vsi: the VSI being configured
4262 int i40e_up(struct i40e_vsi *vsi)
4266 err = i40e_vsi_configure(vsi);
4268 err = i40e_up_complete(vsi);
4274 * i40e_down - Shutdown the connection processing
4275 * @vsi: the VSI being stopped
4277 void i40e_down(struct i40e_vsi *vsi)
4281 /* It is assumed that the caller of this function
4282 * sets the vsi->state __I40E_DOWN bit.
4285 netif_carrier_off(vsi->netdev);
4286 netif_tx_disable(vsi->netdev);
4288 i40e_vsi_disable_irq(vsi);
4289 i40e_vsi_control_rings(vsi, false);
4290 i40e_napi_disable_all(vsi);
4292 for (i = 0; i < vsi->num_queue_pairs; i++) {
4293 i40e_clean_tx_ring(vsi->tx_rings[i]);
4294 i40e_clean_rx_ring(vsi->rx_rings[i]);
4299 * i40e_setup_tc - configure multiple traffic classes
4300 * @netdev: net device to configure
4301 * @tc: number of traffic classes to enable
4303 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
4305 struct i40e_netdev_priv *np = netdev_priv(netdev);
4306 struct i40e_vsi *vsi = np->vsi;
4307 struct i40e_pf *pf = vsi->back;
4312 /* Check if DCB enabled to continue */
4313 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
4314 netdev_info(netdev, "DCB is not enabled for adapter\n");
4318 /* Check if MFP enabled */
4319 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
4320 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
4324 /* Check whether tc count is within enabled limit */
4325 if (tc > i40e_pf_get_num_tc(pf)) {
4326 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
4330 /* Generate TC map for number of tc requested */
4331 for (i = 0; i < tc; i++)
4332 enabled_tc |= (1 << i);
4334 /* Requesting same TC configuration as already enabled */
4335 if (enabled_tc == vsi->tc_config.enabled_tc)
4338 /* Quiesce VSI queues */
4339 i40e_quiesce_vsi(vsi);
4341 /* Configure VSI for enabled TCs */
4342 ret = i40e_vsi_config_tc(vsi, enabled_tc);
4344 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
4350 i40e_unquiesce_vsi(vsi);
4357 * i40e_open - Called when a network interface is made active
4358 * @netdev: network interface device structure
4360 * The open entry point is called when a network interface is made
4361 * active by the system (IFF_UP). At this point all resources needed
4362 * for transmit and receive operations are allocated, the interrupt
4363 * handler is registered with the OS, the netdev watchdog subtask is
4364 * enabled, and the stack is notified that the interface is ready.
4366 * Returns 0 on success, negative value on failure
4368 static int i40e_open(struct net_device *netdev)
4370 struct i40e_netdev_priv *np = netdev_priv(netdev);
4371 struct i40e_vsi *vsi = np->vsi;
4372 struct i40e_pf *pf = vsi->back;
4375 /* disallow open during test or if eeprom is broken */
4376 if (test_bit(__I40E_TESTING, &pf->state) ||
4377 test_bit(__I40E_BAD_EEPROM, &pf->state))
4380 netif_carrier_off(netdev);
4382 err = i40e_vsi_open(vsi);
4386 /* configure global TSO hardware offload settings */
4387 wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
4388 TCP_FLAG_FIN) >> 16);
4389 wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
4391 TCP_FLAG_CWR) >> 16);
4392 wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
4394 #ifdef CONFIG_I40E_VXLAN
4395 vxlan_get_rx_port(netdev);
4403 * @vsi: the VSI to open
4405 * Finish initialization of the VSI.
4407 * Returns 0 on success, negative value on failure
4409 int i40e_vsi_open(struct i40e_vsi *vsi)
4411 struct i40e_pf *pf = vsi->back;
4412 char int_name[IFNAMSIZ];
4415 /* allocate descriptors */
4416 err = i40e_vsi_setup_tx_resources(vsi);
4419 err = i40e_vsi_setup_rx_resources(vsi);
4423 err = i40e_vsi_configure(vsi);
4428 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
4429 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
4430 err = i40e_vsi_request_irq(vsi, int_name);
4434 /* Notify the stack of the actual queue counts. */
4435 err = netif_set_real_num_tx_queues(vsi->netdev,
4436 vsi->num_queue_pairs);
4438 goto err_set_queues;
4440 err = netif_set_real_num_rx_queues(vsi->netdev,
4441 vsi->num_queue_pairs);
4443 goto err_set_queues;
4445 } else if (vsi->type == I40E_VSI_FDIR) {
4446 snprintf(int_name, sizeof(int_name) - 1, "%s-fdir",
4447 dev_driver_string(&pf->pdev->dev));
4448 err = i40e_vsi_request_irq(vsi, int_name);
4454 err = i40e_up_complete(vsi);
4456 goto err_up_complete;
4463 i40e_vsi_free_irq(vsi);
4465 i40e_vsi_free_rx_resources(vsi);
4467 i40e_vsi_free_tx_resources(vsi);
4468 if (vsi == pf->vsi[pf->lan_vsi])
4469 i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
4475 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
4476 * @pf: Pointer to pf
4478 * This function destroys the hlist where all the Flow Director
4479 * filters were saved.
4481 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
4483 struct i40e_fdir_filter *filter;
4484 struct hlist_node *node2;
4486 hlist_for_each_entry_safe(filter, node2,
4487 &pf->fdir_filter_list, fdir_node) {
4488 hlist_del(&filter->fdir_node);
4491 pf->fdir_pf_active_filters = 0;
4495 * i40e_close - Disables a network interface
4496 * @netdev: network interface device structure
4498 * The close entry point is called when an interface is de-activated
4499 * by the OS. The hardware is still under the driver's control, but
4500 * this netdev interface is disabled.
4502 * Returns 0, this is not allowed to fail
4504 static int i40e_close(struct net_device *netdev)
4506 struct i40e_netdev_priv *np = netdev_priv(netdev);
4507 struct i40e_vsi *vsi = np->vsi;
4509 i40e_vsi_close(vsi);
4515 * i40e_do_reset - Start a PF or Core Reset sequence
4516 * @pf: board private structure
4517 * @reset_flags: which reset is requested
4519 * The essential difference in resets is that the PF Reset
4520 * doesn't clear the packet buffers, doesn't reset the PE
4521 * firmware, and doesn't bother the other PFs on the chip.
4523 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
4527 WARN_ON(in_interrupt());
4529 if (i40e_check_asq_alive(&pf->hw))
4530 i40e_vc_notify_reset(pf);
4532 /* do the biggest reset indicated */
4533 if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
4535 /* Request a Global Reset
4537 * This will start the chip's countdown to the actual full
4538 * chip reset event, and a warning interrupt to be sent
4539 * to all PFs, including the requestor. Our handler
4540 * for the warning interrupt will deal with the shutdown
4541 * and recovery of the switch setup.
4543 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
4544 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4545 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
4546 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4548 } else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
4550 /* Request a Core Reset
4552 * Same as Global Reset, except does *not* include the MAC/PHY
4554 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
4555 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4556 val |= I40E_GLGEN_RTRIG_CORER_MASK;
4557 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4558 i40e_flush(&pf->hw);
4560 } else if (reset_flags & (1 << __I40E_EMP_RESET_REQUESTED)) {
4562 /* Request a Firmware Reset
4564 * Same as Global reset, plus restarting the
4565 * embedded firmware engine.
4567 /* enable EMP Reset */
4568 val = rd32(&pf->hw, I40E_GLGEN_RSTENA_EMP);
4569 val |= I40E_GLGEN_RSTENA_EMP_EMP_RST_ENA_MASK;
4570 wr32(&pf->hw, I40E_GLGEN_RSTENA_EMP, val);
4572 /* force the reset */
4573 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4574 val |= I40E_GLGEN_RTRIG_EMPFWR_MASK;
4575 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4576 i40e_flush(&pf->hw);
4578 } else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
4580 /* Request a PF Reset
4582 * Resets only the PF-specific registers
4584 * This goes directly to the tear-down and rebuild of
4585 * the switch, since we need to do all the recovery as
4586 * for the Core Reset.
4588 dev_dbg(&pf->pdev->dev, "PFR requested\n");
4589 i40e_handle_reset_warning(pf);
4591 } else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
4594 /* Find the VSI(s) that requested a re-init */
4595 dev_info(&pf->pdev->dev,
4596 "VSI reinit requested\n");
4597 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4598 struct i40e_vsi *vsi = pf->vsi[v];
4600 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
4601 i40e_vsi_reinit_locked(pf->vsi[v]);
4602 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
4606 /* no further action needed, so return now */
4609 dev_info(&pf->pdev->dev,
4610 "bad reset request 0x%08x\n", reset_flags);
4615 #ifdef CONFIG_I40E_DCB
4617 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
4618 * @pf: board private structure
4619 * @old_cfg: current DCB config
4620 * @new_cfg: new DCB config
4622 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
4623 struct i40e_dcbx_config *old_cfg,
4624 struct i40e_dcbx_config *new_cfg)
4626 bool need_reconfig = false;
4628 /* Check if ETS configuration has changed */
4629 if (memcmp(&new_cfg->etscfg,
4631 sizeof(new_cfg->etscfg))) {
4632 /* If Priority Table has changed reconfig is needed */
4633 if (memcmp(&new_cfg->etscfg.prioritytable,
4634 &old_cfg->etscfg.prioritytable,
4635 sizeof(new_cfg->etscfg.prioritytable))) {
4636 need_reconfig = true;
4637 dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
4640 if (memcmp(&new_cfg->etscfg.tcbwtable,
4641 &old_cfg->etscfg.tcbwtable,
4642 sizeof(new_cfg->etscfg.tcbwtable)))
4643 dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
4645 if (memcmp(&new_cfg->etscfg.tsatable,
4646 &old_cfg->etscfg.tsatable,
4647 sizeof(new_cfg->etscfg.tsatable)))
4648 dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
4651 /* Check if PFC configuration has changed */
4652 if (memcmp(&new_cfg->pfc,
4654 sizeof(new_cfg->pfc))) {
4655 need_reconfig = true;
4656 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
4659 /* Check if APP Table has changed */
4660 if (memcmp(&new_cfg->app,
4662 sizeof(new_cfg->app))) {
4663 need_reconfig = true;
4664 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
4667 return need_reconfig;
4671 * i40e_handle_lldp_event - Handle LLDP Change MIB event
4672 * @pf: board private structure
4673 * @e: event info posted on ARQ
4675 static int i40e_handle_lldp_event(struct i40e_pf *pf,
4676 struct i40e_arq_event_info *e)
4678 struct i40e_aqc_lldp_get_mib *mib =
4679 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
4680 struct i40e_hw *hw = &pf->hw;
4681 struct i40e_dcbx_config *dcbx_cfg = &hw->local_dcbx_config;
4682 struct i40e_dcbx_config tmp_dcbx_cfg;
4683 bool need_reconfig = false;
4687 /* Ignore if event is not for Nearest Bridge */
4688 type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
4689 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
4690 if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
4693 /* Check MIB Type and return if event for Remote MIB update */
4694 type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
4695 if (type == I40E_AQ_LLDP_MIB_REMOTE) {
4696 /* Update the remote cached instance and return */
4697 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
4698 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
4699 &hw->remote_dcbx_config);
4703 /* Convert/store the DCBX data from LLDPDU temporarily */
4704 memset(&tmp_dcbx_cfg, 0, sizeof(tmp_dcbx_cfg));
4705 ret = i40e_lldp_to_dcb_config(e->msg_buf, &tmp_dcbx_cfg);
4707 /* Error in LLDPDU parsing return */
4708 dev_info(&pf->pdev->dev, "Failed parsing LLDPDU from event buffer\n");
4712 /* No change detected in DCBX configs */
4713 if (!memcmp(&tmp_dcbx_cfg, dcbx_cfg, sizeof(tmp_dcbx_cfg))) {
4714 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
4718 need_reconfig = i40e_dcb_need_reconfig(pf, dcbx_cfg, &tmp_dcbx_cfg);
4720 i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg);
4722 /* Overwrite the new configuration */
4723 *dcbx_cfg = tmp_dcbx_cfg;
4728 /* Reconfiguration needed quiesce all VSIs */
4729 i40e_pf_quiesce_all_vsi(pf);
4731 /* Changes in configuration update VEB/VSI */
4732 i40e_dcb_reconfigure(pf);
4734 i40e_pf_unquiesce_all_vsi(pf);
4738 #endif /* CONFIG_I40E_DCB */
4741 * i40e_do_reset_safe - Protected reset path for userland calls.
4742 * @pf: board private structure
4743 * @reset_flags: which reset is requested
4746 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
4749 i40e_do_reset(pf, reset_flags);
4754 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
4755 * @pf: board private structure
4756 * @e: event info posted on ARQ
4758 * Handler for LAN Queue Overflow Event generated by the firmware for PF
4761 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
4762 struct i40e_arq_event_info *e)
4764 struct i40e_aqc_lan_overflow *data =
4765 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
4766 u32 queue = le32_to_cpu(data->prtdcb_rupto);
4767 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
4768 struct i40e_hw *hw = &pf->hw;
4772 dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
4775 /* Queue belongs to VF, find the VF and issue VF reset */
4776 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
4777 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
4778 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
4779 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
4780 vf_id -= hw->func_caps.vf_base_id;
4781 vf = &pf->vf[vf_id];
4782 i40e_vc_notify_vf_reset(vf);
4783 /* Allow VF to process pending reset notification */
4785 i40e_reset_vf(vf, false);
4790 * i40e_service_event_complete - Finish up the service event
4791 * @pf: board private structure
4793 static void i40e_service_event_complete(struct i40e_pf *pf)
4795 BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
4797 /* flush memory to make sure state is correct before next watchog */
4798 smp_mb__before_clear_bit();
4799 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
4803 * i40e_get_current_fd_count - Get the count of FD filters programmed in the HW
4804 * @pf: board private structure
4806 int i40e_get_current_fd_count(struct i40e_pf *pf)
4809 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
4810 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
4811 ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
4812 I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
4817 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
4818 * @pf: board private structure
4820 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
4822 u32 fcnt_prog, fcnt_avail;
4824 /* Check if, FD SB or ATR was auto disabled and if there is enough room
4827 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
4828 (pf->flags & I40E_FLAG_FD_SB_ENABLED))
4830 fcnt_prog = i40e_get_current_fd_count(pf);
4831 fcnt_avail = i40e_get_fd_cnt_all(pf);
4832 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) {
4833 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
4834 (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
4835 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
4836 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
4839 /* Wait for some more space to be available to turn on ATR */
4840 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
4841 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
4842 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
4843 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
4844 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
4850 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
4851 * @pf: board private structure
4853 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
4855 if (!(pf->flags & I40E_FLAG_FDIR_REQUIRES_REINIT))
4858 /* if interface is down do nothing */
4859 if (test_bit(__I40E_DOWN, &pf->state))
4861 i40e_fdir_check_and_reenable(pf);
4863 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
4864 (pf->flags & I40E_FLAG_FD_SB_ENABLED))
4865 pf->flags &= ~I40E_FLAG_FDIR_REQUIRES_REINIT;
4869 * i40e_vsi_link_event - notify VSI of a link event
4870 * @vsi: vsi to be notified
4871 * @link_up: link up or down
4873 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
4878 switch (vsi->type) {
4880 if (!vsi->netdev || !vsi->netdev_registered)
4884 netif_carrier_on(vsi->netdev);
4885 netif_tx_wake_all_queues(vsi->netdev);
4887 netif_carrier_off(vsi->netdev);
4888 netif_tx_stop_all_queues(vsi->netdev);
4892 case I40E_VSI_SRIOV:
4895 case I40E_VSI_VMDQ2:
4897 case I40E_VSI_MIRROR:
4899 /* there is no notification for other VSIs */
4905 * i40e_veb_link_event - notify elements on the veb of a link event
4906 * @veb: veb to be notified
4907 * @link_up: link up or down
4909 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
4914 if (!veb || !veb->pf)
4918 /* depth first... */
4919 for (i = 0; i < I40E_MAX_VEB; i++)
4920 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
4921 i40e_veb_link_event(pf->veb[i], link_up);
4923 /* ... now the local VSIs */
4924 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4925 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
4926 i40e_vsi_link_event(pf->vsi[i], link_up);
4930 * i40e_link_event - Update netif_carrier status
4931 * @pf: board private structure
4933 static void i40e_link_event(struct i40e_pf *pf)
4935 bool new_link, old_link;
4937 new_link = (pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP);
4938 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
4940 if (new_link == old_link)
4942 if (!test_bit(__I40E_DOWN, &pf->vsi[pf->lan_vsi]->state))
4943 i40e_print_link_message(pf->vsi[pf->lan_vsi], new_link);
4945 /* Notify the base of the switch tree connected to
4946 * the link. Floating VEBs are not notified.
4948 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
4949 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
4951 i40e_vsi_link_event(pf->vsi[pf->lan_vsi], new_link);
4954 i40e_vc_notify_link_state(pf);
4956 if (pf->flags & I40E_FLAG_PTP)
4957 i40e_ptp_set_increment(pf);
4961 * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
4962 * @pf: board private structure
4964 * Set the per-queue flags to request a check for stuck queues in the irq
4965 * clean functions, then force interrupts to be sure the irq clean is called.
4967 static void i40e_check_hang_subtask(struct i40e_pf *pf)
4971 /* If we're down or resetting, just bail */
4972 if (test_bit(__I40E_CONFIG_BUSY, &pf->state))
4975 /* for each VSI/netdev
4977 * set the check flag
4979 * force an interrupt
4981 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4982 struct i40e_vsi *vsi = pf->vsi[v];
4986 test_bit(__I40E_DOWN, &vsi->state) ||
4987 (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
4990 for (i = 0; i < vsi->num_queue_pairs; i++) {
4991 set_check_for_tx_hang(vsi->tx_rings[i]);
4992 if (test_bit(__I40E_HANG_CHECK_ARMED,
4993 &vsi->tx_rings[i]->state))
4998 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
4999 wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
5000 (I40E_PFINT_DYN_CTL0_INTENA_MASK |
5001 I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK));
5003 u16 vec = vsi->base_vector - 1;
5004 u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
5005 I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK);
5006 for (i = 0; i < vsi->num_q_vectors; i++, vec++)
5007 wr32(&vsi->back->hw,
5008 I40E_PFINT_DYN_CTLN(vec), val);
5010 i40e_flush(&vsi->back->hw);
5016 * i40e_watchdog_subtask - Check and bring link up
5017 * @pf: board private structure
5019 static void i40e_watchdog_subtask(struct i40e_pf *pf)
5023 /* if interface is down do nothing */
5024 if (test_bit(__I40E_DOWN, &pf->state) ||
5025 test_bit(__I40E_CONFIG_BUSY, &pf->state))
5028 /* Update the stats for active netdevs so the network stack
5029 * can look at updated numbers whenever it cares to
5031 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
5032 if (pf->vsi[i] && pf->vsi[i]->netdev)
5033 i40e_update_stats(pf->vsi[i]);
5035 /* Update the stats for the active switching components */
5036 for (i = 0; i < I40E_MAX_VEB; i++)
5038 i40e_update_veb_stats(pf->veb[i]);
5040 i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
5044 * i40e_reset_subtask - Set up for resetting the device and driver
5045 * @pf: board private structure
5047 static void i40e_reset_subtask(struct i40e_pf *pf)
5049 u32 reset_flags = 0;
5052 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
5053 reset_flags |= (1 << __I40E_REINIT_REQUESTED);
5054 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
5056 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
5057 reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
5058 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5060 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
5061 reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
5062 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
5064 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
5065 reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
5066 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
5069 /* If there's a recovery already waiting, it takes
5070 * precedence before starting a new reset sequence.
5072 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
5073 i40e_handle_reset_warning(pf);
5077 /* If we're already down or resetting, just bail */
5079 !test_bit(__I40E_DOWN, &pf->state) &&
5080 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
5081 i40e_do_reset(pf, reset_flags);
5088 * i40e_handle_link_event - Handle link event
5089 * @pf: board private structure
5090 * @e: event info posted on ARQ
5092 static void i40e_handle_link_event(struct i40e_pf *pf,
5093 struct i40e_arq_event_info *e)
5095 struct i40e_hw *hw = &pf->hw;
5096 struct i40e_aqc_get_link_status *status =
5097 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
5098 struct i40e_link_status *hw_link_info = &hw->phy.link_info;
5100 /* save off old link status information */
5101 memcpy(&pf->hw.phy.link_info_old, hw_link_info,
5102 sizeof(pf->hw.phy.link_info_old));
5104 /* update link status */
5105 hw_link_info->phy_type = (enum i40e_aq_phy_type)status->phy_type;
5106 hw_link_info->link_speed = (enum i40e_aq_link_speed)status->link_speed;
5107 hw_link_info->link_info = status->link_info;
5108 hw_link_info->an_info = status->an_info;
5109 hw_link_info->ext_info = status->ext_info;
5110 hw_link_info->lse_enable =
5111 le16_to_cpu(status->command_flags) &
5114 /* process the event */
5115 i40e_link_event(pf);
5117 /* Do a new status request to re-enable LSE reporting
5118 * and load new status information into the hw struct,
5119 * then see if the status changed while processing the
5122 i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
5123 i40e_link_event(pf);
5127 * i40e_clean_adminq_subtask - Clean the AdminQ rings
5128 * @pf: board private structure
5130 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
5132 struct i40e_arq_event_info event;
5133 struct i40e_hw *hw = &pf->hw;
5139 if (!test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state))
5142 event.msg_size = I40E_MAX_AQ_BUF_SIZE;
5143 event.msg_buf = kzalloc(event.msg_size, GFP_KERNEL);
5148 event.msg_size = I40E_MAX_AQ_BUF_SIZE; /* reinit each time */
5149 ret = i40e_clean_arq_element(hw, &event, &pending);
5150 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK) {
5151 dev_info(&pf->pdev->dev, "No ARQ event found\n");
5154 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
5158 opcode = le16_to_cpu(event.desc.opcode);
5161 case i40e_aqc_opc_get_link_status:
5162 i40e_handle_link_event(pf, &event);
5164 case i40e_aqc_opc_send_msg_to_pf:
5165 ret = i40e_vc_process_vf_msg(pf,
5166 le16_to_cpu(event.desc.retval),
5167 le32_to_cpu(event.desc.cookie_high),
5168 le32_to_cpu(event.desc.cookie_low),
5172 case i40e_aqc_opc_lldp_update_mib:
5173 dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
5174 #ifdef CONFIG_I40E_DCB
5176 ret = i40e_handle_lldp_event(pf, &event);
5178 #endif /* CONFIG_I40E_DCB */
5180 case i40e_aqc_opc_event_lan_overflow:
5181 dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
5182 i40e_handle_lan_overflow_event(pf, &event);
5184 case i40e_aqc_opc_send_msg_to_peer:
5185 dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
5188 dev_info(&pf->pdev->dev,
5189 "ARQ Error: Unknown event 0x%04x received\n",
5193 } while (pending && (i++ < pf->adminq_work_limit));
5195 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
5196 /* re-enable Admin queue interrupt cause */
5197 val = rd32(hw, I40E_PFINT_ICR0_ENA);
5198 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
5199 wr32(hw, I40E_PFINT_ICR0_ENA, val);
5202 kfree(event.msg_buf);
5206 * i40e_verify_eeprom - make sure eeprom is good to use
5207 * @pf: board private structure
5209 static void i40e_verify_eeprom(struct i40e_pf *pf)
5213 err = i40e_diag_eeprom_test(&pf->hw);
5215 /* retry in case of garbage read */
5216 err = i40e_diag_eeprom_test(&pf->hw);
5218 dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
5220 set_bit(__I40E_BAD_EEPROM, &pf->state);
5224 if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
5225 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
5226 clear_bit(__I40E_BAD_EEPROM, &pf->state);
5231 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
5232 * @veb: pointer to the VEB instance
5234 * This is a recursive function that first builds the attached VSIs then
5235 * recurses in to build the next layer of VEB. We track the connections
5236 * through our own index numbers because the seid's from the HW could
5237 * change across the reset.
5239 static int i40e_reconstitute_veb(struct i40e_veb *veb)
5241 struct i40e_vsi *ctl_vsi = NULL;
5242 struct i40e_pf *pf = veb->pf;
5246 /* build VSI that owns this VEB, temporarily attached to base VEB */
5247 for (v = 0; v < pf->hw.func_caps.num_vsis && !ctl_vsi; v++) {
5249 pf->vsi[v]->veb_idx == veb->idx &&
5250 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
5251 ctl_vsi = pf->vsi[v];
5256 dev_info(&pf->pdev->dev,
5257 "missing owner VSI for veb_idx %d\n", veb->idx);
5259 goto end_reconstitute;
5261 if (ctl_vsi != pf->vsi[pf->lan_vsi])
5262 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
5263 ret = i40e_add_vsi(ctl_vsi);
5265 dev_info(&pf->pdev->dev,
5266 "rebuild of owner VSI failed: %d\n", ret);
5267 goto end_reconstitute;
5269 i40e_vsi_reset_stats(ctl_vsi);
5271 /* create the VEB in the switch and move the VSI onto the VEB */
5272 ret = i40e_add_veb(veb, ctl_vsi);
5274 goto end_reconstitute;
5276 /* create the remaining VSIs attached to this VEB */
5277 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
5278 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
5281 if (pf->vsi[v]->veb_idx == veb->idx) {
5282 struct i40e_vsi *vsi = pf->vsi[v];
5283 vsi->uplink_seid = veb->seid;
5284 ret = i40e_add_vsi(vsi);
5286 dev_info(&pf->pdev->dev,
5287 "rebuild of vsi_idx %d failed: %d\n",
5289 goto end_reconstitute;
5291 i40e_vsi_reset_stats(vsi);
5295 /* create any VEBs attached to this VEB - RECURSION */
5296 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
5297 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
5298 pf->veb[veb_idx]->uplink_seid = veb->seid;
5299 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
5310 * i40e_get_capabilities - get info about the HW
5311 * @pf: the PF struct
5313 static int i40e_get_capabilities(struct i40e_pf *pf)
5315 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
5320 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
5322 cap_buf = kzalloc(buf_len, GFP_KERNEL);
5326 /* this loads the data into the hw struct for us */
5327 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
5329 i40e_aqc_opc_list_func_capabilities,
5331 /* data loaded, buffer no longer needed */
5334 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
5335 /* retry with a larger buffer */
5336 buf_len = data_size;
5337 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
5338 dev_info(&pf->pdev->dev,
5339 "capability discovery failed: aq=%d\n",
5340 pf->hw.aq.asq_last_status);
5345 if (((pf->hw.aq.fw_maj_ver == 2) && (pf->hw.aq.fw_min_ver < 22)) ||
5346 (pf->hw.aq.fw_maj_ver < 2)) {
5347 pf->hw.func_caps.num_msix_vectors++;
5348 pf->hw.func_caps.num_msix_vectors_vf++;
5351 if (pf->hw.debug_mask & I40E_DEBUG_USER)
5352 dev_info(&pf->pdev->dev,
5353 "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",
5354 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
5355 pf->hw.func_caps.num_msix_vectors,
5356 pf->hw.func_caps.num_msix_vectors_vf,
5357 pf->hw.func_caps.fd_filters_guaranteed,
5358 pf->hw.func_caps.fd_filters_best_effort,
5359 pf->hw.func_caps.num_tx_qp,
5360 pf->hw.func_caps.num_vsis);
5362 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
5363 + pf->hw.func_caps.num_vfs)
5364 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
5365 dev_info(&pf->pdev->dev,
5366 "got num_vsis %d, setting num_vsis to %d\n",
5367 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
5368 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
5374 static int i40e_vsi_clear(struct i40e_vsi *vsi);
5377 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
5378 * @pf: board private structure
5380 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
5382 struct i40e_vsi *vsi;
5385 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
5388 /* find existing VSI and see if it needs configuring */
5390 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
5391 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
5397 /* create a new VSI if none exists */
5399 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
5400 pf->vsi[pf->lan_vsi]->seid, 0);
5402 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
5403 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
5408 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
5412 * i40e_fdir_teardown - release the Flow Director resources
5413 * @pf: board private structure
5415 static void i40e_fdir_teardown(struct i40e_pf *pf)
5419 i40e_fdir_filter_exit(pf);
5420 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
5421 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
5422 i40e_vsi_release(pf->vsi[i]);
5429 * i40e_prep_for_reset - prep for the core to reset
5430 * @pf: board private structure
5432 * Close up the VFs and other things in prep for pf Reset.
5434 static int i40e_prep_for_reset(struct i40e_pf *pf)
5436 struct i40e_hw *hw = &pf->hw;
5437 i40e_status ret = 0;
5440 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
5441 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
5444 dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
5446 /* quiesce the VSIs and their queues that are not already DOWN */
5447 i40e_pf_quiesce_all_vsi(pf);
5449 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
5451 pf->vsi[v]->seid = 0;
5454 i40e_shutdown_adminq(&pf->hw);
5456 /* call shutdown HMC */
5457 if (hw->hmc.hmc_obj) {
5458 ret = i40e_shutdown_lan_hmc(hw);
5460 dev_warn(&pf->pdev->dev,
5461 "shutdown_lan_hmc failed: %d\n", ret);
5462 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
5469 * i40e_send_version - update firmware with driver version
5472 static void i40e_send_version(struct i40e_pf *pf)
5474 struct i40e_driver_version dv;
5476 dv.major_version = DRV_VERSION_MAJOR;
5477 dv.minor_version = DRV_VERSION_MINOR;
5478 dv.build_version = DRV_VERSION_BUILD;
5479 dv.subbuild_version = 0;
5480 strncpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
5481 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
5485 * i40e_reset_and_rebuild - reset and rebuild using a saved config
5486 * @pf: board private structure
5487 * @reinit: if the Main VSI needs to re-initialized.
5489 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
5491 struct i40e_hw *hw = &pf->hw;
5495 /* Now we wait for GRST to settle out.
5496 * We don't have to delete the VEBs or VSIs from the hw switch
5497 * because the reset will make them disappear.
5499 ret = i40e_pf_reset(hw);
5501 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
5502 goto end_core_reset;
5506 if (test_bit(__I40E_DOWN, &pf->state))
5507 goto end_core_reset;
5508 dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
5510 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
5511 ret = i40e_init_adminq(&pf->hw);
5513 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
5514 goto end_core_reset;
5517 /* re-verify the eeprom if we just had an EMP reset */
5518 if (test_bit(__I40E_EMP_RESET_REQUESTED, &pf->state)) {
5519 clear_bit(__I40E_EMP_RESET_REQUESTED, &pf->state);
5520 i40e_verify_eeprom(pf);
5523 ret = i40e_get_capabilities(pf);
5525 dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
5527 goto end_core_reset;
5530 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
5531 hw->func_caps.num_rx_qp,
5532 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
5534 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
5535 goto end_core_reset;
5537 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
5539 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
5540 goto end_core_reset;
5543 #ifdef CONFIG_I40E_DCB
5544 ret = i40e_init_pf_dcb(pf);
5546 dev_info(&pf->pdev->dev, "init_pf_dcb failed: %d\n", ret);
5547 goto end_core_reset;
5549 #endif /* CONFIG_I40E_DCB */
5551 /* do basic switch setup */
5552 ret = i40e_setup_pf_switch(pf, reinit);
5554 goto end_core_reset;
5556 /* Rebuild the VSIs and VEBs that existed before reset.
5557 * They are still in our local switch element arrays, so only
5558 * need to rebuild the switch model in the HW.
5560 * If there were VEBs but the reconstitution failed, we'll try
5561 * try to recover minimal use by getting the basic PF VSI working.
5563 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
5564 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
5565 /* find the one VEB connected to the MAC, and find orphans */
5566 for (v = 0; v < I40E_MAX_VEB; v++) {
5570 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
5571 pf->veb[v]->uplink_seid == 0) {
5572 ret = i40e_reconstitute_veb(pf->veb[v]);
5577 /* If Main VEB failed, we're in deep doodoo,
5578 * so give up rebuilding the switch and set up
5579 * for minimal rebuild of PF VSI.
5580 * If orphan failed, we'll report the error
5581 * but try to keep going.
5583 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
5584 dev_info(&pf->pdev->dev,
5585 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
5587 pf->vsi[pf->lan_vsi]->uplink_seid
5590 } else if (pf->veb[v]->uplink_seid == 0) {
5591 dev_info(&pf->pdev->dev,
5592 "rebuild of orphan VEB failed: %d\n",
5599 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
5600 dev_info(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
5601 /* no VEB, so rebuild only the Main VSI */
5602 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
5604 dev_info(&pf->pdev->dev,
5605 "rebuild of Main VSI failed: %d\n", ret);
5606 goto end_core_reset;
5610 /* reinit the misc interrupt */
5611 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5612 ret = i40e_setup_misc_vector(pf);
5614 /* restart the VSIs that were rebuilt and running before the reset */
5615 i40e_pf_unquiesce_all_vsi(pf);
5617 if (pf->num_alloc_vfs) {
5618 for (v = 0; v < pf->num_alloc_vfs; v++)
5619 i40e_reset_vf(&pf->vf[v], true);
5622 /* tell the firmware that we're starting */
5623 i40e_send_version(pf);
5625 dev_info(&pf->pdev->dev, "reset complete\n");
5628 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
5632 * i40e_handle_reset_warning - prep for the pf to reset, reset and rebuild
5633 * @pf: board private structure
5635 * Close up the VFs and other things in prep for a Core Reset,
5636 * then get ready to rebuild the world.
5638 static void i40e_handle_reset_warning(struct i40e_pf *pf)
5642 ret = i40e_prep_for_reset(pf);
5644 i40e_reset_and_rebuild(pf, false);
5648 * i40e_handle_mdd_event
5649 * @pf: pointer to the pf structure
5651 * Called from the MDD irq handler to identify possibly malicious vfs
5653 static void i40e_handle_mdd_event(struct i40e_pf *pf)
5655 struct i40e_hw *hw = &pf->hw;
5656 bool mdd_detected = false;
5661 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
5664 /* find what triggered the MDD event */
5665 reg = rd32(hw, I40E_GL_MDET_TX);
5666 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
5667 u8 func = (reg & I40E_GL_MDET_TX_FUNCTION_MASK)
5668 >> I40E_GL_MDET_TX_FUNCTION_SHIFT;
5669 u8 event = (reg & I40E_GL_MDET_TX_EVENT_SHIFT)
5670 >> I40E_GL_MDET_TX_EVENT_SHIFT;
5671 u8 queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK)
5672 >> I40E_GL_MDET_TX_QUEUE_SHIFT;
5673 dev_info(&pf->pdev->dev,
5674 "Malicious Driver Detection event 0x%02x on TX queue %d of function 0x%02x\n",
5675 event, queue, func);
5676 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
5677 mdd_detected = true;
5679 reg = rd32(hw, I40E_GL_MDET_RX);
5680 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
5681 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK)
5682 >> I40E_GL_MDET_RX_FUNCTION_SHIFT;
5683 u8 event = (reg & I40E_GL_MDET_RX_EVENT_SHIFT)
5684 >> I40E_GL_MDET_RX_EVENT_SHIFT;
5685 u8 queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK)
5686 >> I40E_GL_MDET_RX_QUEUE_SHIFT;
5687 dev_info(&pf->pdev->dev,
5688 "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
5689 event, queue, func);
5690 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
5691 mdd_detected = true;
5694 /* see if one of the VFs needs its hand slapped */
5695 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
5697 reg = rd32(hw, I40E_VP_MDET_TX(i));
5698 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
5699 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
5700 vf->num_mdd_events++;
5701 dev_info(&pf->pdev->dev, "MDD TX event on VF %d\n", i);
5704 reg = rd32(hw, I40E_VP_MDET_RX(i));
5705 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
5706 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
5707 vf->num_mdd_events++;
5708 dev_info(&pf->pdev->dev, "MDD RX event on VF %d\n", i);
5711 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
5712 dev_info(&pf->pdev->dev,
5713 "Too many MDD events on VF %d, disabled\n", i);
5714 dev_info(&pf->pdev->dev,
5715 "Use PF Control I/F to re-enable the VF\n");
5716 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
5720 /* re-enable mdd interrupt cause */
5721 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
5722 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
5723 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
5724 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
5728 #ifdef CONFIG_I40E_VXLAN
5730 * i40e_sync_vxlan_filters_subtask - Sync the VSI filter list with HW
5731 * @pf: board private structure
5733 static void i40e_sync_vxlan_filters_subtask(struct i40e_pf *pf)
5735 struct i40e_hw *hw = &pf->hw;
5741 if (!(pf->flags & I40E_FLAG_VXLAN_FILTER_SYNC))
5744 pf->flags &= ~I40E_FLAG_VXLAN_FILTER_SYNC;
5746 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
5747 if (pf->pending_vxlan_bitmap & (1 << i)) {
5748 pf->pending_vxlan_bitmap &= ~(1 << i);
5749 port = pf->vxlan_ports[i];
5751 i40e_aq_add_udp_tunnel(hw, ntohs(port),
5752 I40E_AQC_TUNNEL_TYPE_VXLAN,
5753 &filter_index, NULL)
5754 : i40e_aq_del_udp_tunnel(hw, i, NULL);
5757 dev_info(&pf->pdev->dev, "Failed to execute AQ command for %s port %d with index %d\n",
5758 port ? "adding" : "deleting",
5759 ntohs(port), port ? i : i);
5761 pf->vxlan_ports[i] = 0;
5763 dev_info(&pf->pdev->dev, "%s port %d with AQ command with index %d\n",
5764 port ? "Added" : "Deleted",
5765 ntohs(port), port ? i : filter_index);
5773 * i40e_service_task - Run the driver's async subtasks
5774 * @work: pointer to work_struct containing our data
5776 static void i40e_service_task(struct work_struct *work)
5778 struct i40e_pf *pf = container_of(work,
5781 unsigned long start_time = jiffies;
5783 i40e_reset_subtask(pf);
5784 i40e_handle_mdd_event(pf);
5785 i40e_vc_process_vflr_event(pf);
5786 i40e_watchdog_subtask(pf);
5787 i40e_fdir_reinit_subtask(pf);
5788 i40e_check_hang_subtask(pf);
5789 i40e_sync_filters_subtask(pf);
5790 #ifdef CONFIG_I40E_VXLAN
5791 i40e_sync_vxlan_filters_subtask(pf);
5793 i40e_clean_adminq_subtask(pf);
5795 i40e_service_event_complete(pf);
5797 /* If the tasks have taken longer than one timer cycle or there
5798 * is more work to be done, reschedule the service task now
5799 * rather than wait for the timer to tick again.
5801 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
5802 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
5803 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
5804 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
5805 i40e_service_event_schedule(pf);
5809 * i40e_service_timer - timer callback
5810 * @data: pointer to PF struct
5812 static void i40e_service_timer(unsigned long data)
5814 struct i40e_pf *pf = (struct i40e_pf *)data;
5816 mod_timer(&pf->service_timer,
5817 round_jiffies(jiffies + pf->service_timer_period));
5818 i40e_service_event_schedule(pf);
5822 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
5823 * @vsi: the VSI being configured
5825 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
5827 struct i40e_pf *pf = vsi->back;
5829 switch (vsi->type) {
5831 vsi->alloc_queue_pairs = pf->num_lan_qps;
5832 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5833 I40E_REQ_DESCRIPTOR_MULTIPLE);
5834 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5835 vsi->num_q_vectors = pf->num_lan_msix;
5837 vsi->num_q_vectors = 1;
5842 vsi->alloc_queue_pairs = 1;
5843 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
5844 I40E_REQ_DESCRIPTOR_MULTIPLE);
5845 vsi->num_q_vectors = 1;
5848 case I40E_VSI_VMDQ2:
5849 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
5850 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5851 I40E_REQ_DESCRIPTOR_MULTIPLE);
5852 vsi->num_q_vectors = pf->num_vmdq_msix;
5855 case I40E_VSI_SRIOV:
5856 vsi->alloc_queue_pairs = pf->num_vf_qps;
5857 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5858 I40E_REQ_DESCRIPTOR_MULTIPLE);
5870 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
5871 * @type: VSI pointer
5872 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
5874 * On error: returns error code (negative)
5875 * On success: returns 0
5877 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
5882 /* allocate memory for both Tx and Rx ring pointers */
5883 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
5884 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
5887 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
5889 if (alloc_qvectors) {
5890 /* allocate memory for q_vector pointers */
5891 size = sizeof(struct i40e_q_vectors *) * vsi->num_q_vectors;
5892 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
5893 if (!vsi->q_vectors) {
5901 kfree(vsi->tx_rings);
5906 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
5907 * @pf: board private structure
5908 * @type: type of VSI
5910 * On error: returns error code (negative)
5911 * On success: returns vsi index in PF (positive)
5913 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
5916 struct i40e_vsi *vsi;
5920 /* Need to protect the allocation of the VSIs at the PF level */
5921 mutex_lock(&pf->switch_mutex);
5923 /* VSI list may be fragmented if VSI creation/destruction has
5924 * been happening. We can afford to do a quick scan to look
5925 * for any free VSIs in the list.
5927 * find next empty vsi slot, looping back around if necessary
5930 while (i < pf->hw.func_caps.num_vsis && pf->vsi[i])
5932 if (i >= pf->hw.func_caps.num_vsis) {
5934 while (i < pf->next_vsi && pf->vsi[i])
5938 if (i < pf->hw.func_caps.num_vsis && !pf->vsi[i]) {
5939 vsi_idx = i; /* Found one! */
5942 goto unlock_pf; /* out of VSI slots! */
5946 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
5953 set_bit(__I40E_DOWN, &vsi->state);
5956 vsi->rx_itr_setting = pf->rx_itr_default;
5957 vsi->tx_itr_setting = pf->tx_itr_default;
5958 vsi->netdev_registered = false;
5959 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
5960 INIT_LIST_HEAD(&vsi->mac_filter_list);
5961 vsi->irqs_ready = false;
5963 ret = i40e_set_num_rings_in_vsi(vsi);
5967 ret = i40e_vsi_alloc_arrays(vsi, true);
5971 /* Setup default MSIX irq handler for VSI */
5972 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
5974 pf->vsi[vsi_idx] = vsi;
5979 pf->next_vsi = i - 1;
5982 mutex_unlock(&pf->switch_mutex);
5987 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
5988 * @type: VSI pointer
5989 * @free_qvectors: a bool to specify if q_vectors need to be freed.
5991 * On error: returns error code (negative)
5992 * On success: returns 0
5994 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
5996 /* free the ring and vector containers */
5997 if (free_qvectors) {
5998 kfree(vsi->q_vectors);
5999 vsi->q_vectors = NULL;
6001 kfree(vsi->tx_rings);
6002 vsi->tx_rings = NULL;
6003 vsi->rx_rings = NULL;
6007 * i40e_vsi_clear - Deallocate the VSI provided
6008 * @vsi: the VSI being un-configured
6010 static int i40e_vsi_clear(struct i40e_vsi *vsi)
6021 mutex_lock(&pf->switch_mutex);
6022 if (!pf->vsi[vsi->idx]) {
6023 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
6024 vsi->idx, vsi->idx, vsi, vsi->type);
6028 if (pf->vsi[vsi->idx] != vsi) {
6029 dev_err(&pf->pdev->dev,
6030 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
6031 pf->vsi[vsi->idx]->idx,
6033 pf->vsi[vsi->idx]->type,
6034 vsi->idx, vsi, vsi->type);
6038 /* updates the pf for this cleared vsi */
6039 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
6040 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
6042 i40e_vsi_free_arrays(vsi, true);
6044 pf->vsi[vsi->idx] = NULL;
6045 if (vsi->idx < pf->next_vsi)
6046 pf->next_vsi = vsi->idx;
6049 mutex_unlock(&pf->switch_mutex);
6057 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
6058 * @vsi: the VSI being cleaned
6060 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
6064 if (vsi->tx_rings && vsi->tx_rings[0]) {
6065 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6066 kfree_rcu(vsi->tx_rings[i], rcu);
6067 vsi->tx_rings[i] = NULL;
6068 vsi->rx_rings[i] = NULL;
6074 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
6075 * @vsi: the VSI being configured
6077 static int i40e_alloc_rings(struct i40e_vsi *vsi)
6079 struct i40e_ring *tx_ring, *rx_ring;
6080 struct i40e_pf *pf = vsi->back;
6083 /* Set basic values in the rings to be used later during open() */
6084 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6085 /* allocate space for both Tx and Rx in one shot */
6086 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
6090 tx_ring->queue_index = i;
6091 tx_ring->reg_idx = vsi->base_queue + i;
6092 tx_ring->ring_active = false;
6094 tx_ring->netdev = vsi->netdev;
6095 tx_ring->dev = &pf->pdev->dev;
6096 tx_ring->count = vsi->num_desc;
6098 tx_ring->dcb_tc = 0;
6099 vsi->tx_rings[i] = tx_ring;
6101 rx_ring = &tx_ring[1];
6102 rx_ring->queue_index = i;
6103 rx_ring->reg_idx = vsi->base_queue + i;
6104 rx_ring->ring_active = false;
6106 rx_ring->netdev = vsi->netdev;
6107 rx_ring->dev = &pf->pdev->dev;
6108 rx_ring->count = vsi->num_desc;
6110 rx_ring->dcb_tc = 0;
6111 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
6112 set_ring_16byte_desc_enabled(rx_ring);
6114 clear_ring_16byte_desc_enabled(rx_ring);
6115 vsi->rx_rings[i] = rx_ring;
6121 i40e_vsi_clear_rings(vsi);
6126 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
6127 * @pf: board private structure
6128 * @vectors: the number of MSI-X vectors to request
6130 * Returns the number of vectors reserved, or error
6132 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
6134 vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
6135 I40E_MIN_MSIX, vectors);
6137 dev_info(&pf->pdev->dev,
6138 "MSI-X vector reservation failed: %d\n", vectors);
6146 * i40e_init_msix - Setup the MSIX capability
6147 * @pf: board private structure
6149 * Work with the OS to set up the MSIX vectors needed.
6151 * Returns 0 on success, negative on failure
6153 static int i40e_init_msix(struct i40e_pf *pf)
6155 i40e_status err = 0;
6156 struct i40e_hw *hw = &pf->hw;
6160 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
6163 /* The number of vectors we'll request will be comprised of:
6164 * - Add 1 for "other" cause for Admin Queue events, etc.
6165 * - The number of LAN queue pairs
6166 * - Queues being used for RSS.
6167 * We don't need as many as max_rss_size vectors.
6168 * use rss_size instead in the calculation since that
6169 * is governed by number of cpus in the system.
6170 * - assumes symmetric Tx/Rx pairing
6171 * - The number of VMDq pairs
6172 * Once we count this up, try the request.
6174 * If we can't get what we want, we'll simplify to nearly nothing
6175 * and try again. If that still fails, we punt.
6177 pf->num_lan_msix = pf->num_lan_qps - (pf->rss_size_max - pf->rss_size);
6178 pf->num_vmdq_msix = pf->num_vmdq_qps;
6179 v_budget = 1 + pf->num_lan_msix;
6180 v_budget += (pf->num_vmdq_vsis * pf->num_vmdq_msix);
6181 if (pf->flags & I40E_FLAG_FD_SB_ENABLED)
6184 /* Scale down if necessary, and the rings will share vectors */
6185 v_budget = min_t(int, v_budget, hw->func_caps.num_msix_vectors);
6187 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
6189 if (!pf->msix_entries)
6192 for (i = 0; i < v_budget; i++)
6193 pf->msix_entries[i].entry = i;
6194 vec = i40e_reserve_msix_vectors(pf, v_budget);
6195 if (vec < I40E_MIN_MSIX) {
6196 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
6197 kfree(pf->msix_entries);
6198 pf->msix_entries = NULL;
6201 } else if (vec == I40E_MIN_MSIX) {
6202 /* Adjust for minimal MSIX use */
6203 dev_info(&pf->pdev->dev, "Features disabled, not enough MSI-X vectors\n");
6204 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
6205 pf->num_vmdq_vsis = 0;
6206 pf->num_vmdq_qps = 0;
6207 pf->num_vmdq_msix = 0;
6208 pf->num_lan_qps = 1;
6209 pf->num_lan_msix = 1;
6211 } else if (vec != v_budget) {
6212 /* Scale vector usage down */
6213 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
6214 vec--; /* reserve the misc vector */
6216 /* partition out the remaining vectors */
6219 pf->num_vmdq_vsis = 1;
6220 pf->num_lan_msix = 1;
6223 pf->num_vmdq_vsis = 1;
6224 pf->num_lan_msix = 2;
6227 pf->num_lan_msix = min_t(int, (vec / 2),
6229 pf->num_vmdq_vsis = min_t(int, (vec - pf->num_lan_msix),
6230 I40E_DEFAULT_NUM_VMDQ_VSI);
6239 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
6240 * @vsi: the VSI being configured
6241 * @v_idx: index of the vector in the vsi struct
6243 * We allocate one q_vector. If allocation fails we return -ENOMEM.
6245 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
6247 struct i40e_q_vector *q_vector;
6249 /* allocate q_vector */
6250 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
6254 q_vector->vsi = vsi;
6255 q_vector->v_idx = v_idx;
6256 cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
6258 netif_napi_add(vsi->netdev, &q_vector->napi,
6259 i40e_napi_poll, vsi->work_limit);
6261 q_vector->rx.latency_range = I40E_LOW_LATENCY;
6262 q_vector->tx.latency_range = I40E_LOW_LATENCY;
6264 /* tie q_vector and vsi together */
6265 vsi->q_vectors[v_idx] = q_vector;
6271 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
6272 * @vsi: the VSI being configured
6274 * We allocate one q_vector per queue interrupt. If allocation fails we
6277 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
6279 struct i40e_pf *pf = vsi->back;
6280 int v_idx, num_q_vectors;
6283 /* if not MSIX, give the one vector only to the LAN VSI */
6284 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6285 num_q_vectors = vsi->num_q_vectors;
6286 else if (vsi == pf->vsi[pf->lan_vsi])
6291 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
6292 err = i40e_vsi_alloc_q_vector(vsi, v_idx);
6301 i40e_free_q_vector(vsi, v_idx);
6307 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
6308 * @pf: board private structure to initialize
6310 static void i40e_init_interrupt_scheme(struct i40e_pf *pf)
6314 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
6315 err = i40e_init_msix(pf);
6317 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
6318 I40E_FLAG_RSS_ENABLED |
6319 I40E_FLAG_DCB_ENABLED |
6320 I40E_FLAG_SRIOV_ENABLED |
6321 I40E_FLAG_FD_SB_ENABLED |
6322 I40E_FLAG_FD_ATR_ENABLED |
6323 I40E_FLAG_VMDQ_ENABLED);
6325 /* rework the queue expectations without MSIX */
6326 i40e_determine_queue_usage(pf);
6330 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
6331 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
6332 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
6333 err = pci_enable_msi(pf->pdev);
6335 dev_info(&pf->pdev->dev, "MSI init failed - %d\n", err);
6336 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
6340 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
6341 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
6343 /* track first vector for misc interrupts */
6344 err = i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT-1);
6348 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
6349 * @pf: board private structure
6351 * This sets up the handler for MSIX 0, which is used to manage the
6352 * non-queue interrupts, e.g. AdminQ and errors. This is not used
6353 * when in MSI or Legacy interrupt mode.
6355 static int i40e_setup_misc_vector(struct i40e_pf *pf)
6357 struct i40e_hw *hw = &pf->hw;
6360 /* Only request the irq if this is the first time through, and
6361 * not when we're rebuilding after a Reset
6363 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
6364 err = request_irq(pf->msix_entries[0].vector,
6365 i40e_intr, 0, pf->misc_int_name, pf);
6367 dev_info(&pf->pdev->dev,
6368 "request_irq for %s failed: %d\n",
6369 pf->misc_int_name, err);
6374 i40e_enable_misc_int_causes(hw);
6376 /* associate no queues to the misc vector */
6377 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
6378 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
6382 i40e_irq_dynamic_enable_icr0(pf);
6388 * i40e_config_rss - Prepare for RSS if used
6389 * @pf: board private structure
6391 static int i40e_config_rss(struct i40e_pf *pf)
6393 /* Set of random keys generated using kernel random number generator */
6394 static const u32 seed[I40E_PFQF_HKEY_MAX_INDEX + 1] = {0x41b01687,
6395 0x183cfd8c, 0xce880440, 0x580cbc3c, 0x35897377,
6396 0x328b25e1, 0x4fa98922, 0xb7d90c14, 0xd5bad70d,
6397 0xcd15a2c1, 0xe8580225, 0x4a1e9d11, 0xfe5731be};
6398 struct i40e_hw *hw = &pf->hw;
6403 /* Fill out hash function seed */
6404 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
6405 wr32(hw, I40E_PFQF_HKEY(i), seed[i]);
6407 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
6408 hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
6409 ((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
6410 hena |= I40E_DEFAULT_RSS_HENA;
6411 wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
6412 wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
6414 /* Populate the LUT with max no. of queues in round robin fashion */
6415 for (i = 0, j = 0; i < pf->hw.func_caps.rss_table_size; i++, j++) {
6417 /* The assumption is that lan qp count will be the highest
6418 * qp count for any PF VSI that needs RSS.
6419 * If multiple VSIs need RSS support, all the qp counts
6420 * for those VSIs should be a power of 2 for RSS to work.
6421 * If LAN VSI is the only consumer for RSS then this requirement
6424 if (j == pf->rss_size)
6426 /* lut = 4-byte sliding window of 4 lut entries */
6427 lut = (lut << 8) | (j &
6428 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
6429 /* On i = 3, we have 4 entries in lut; write to the register */
6431 wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
6439 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
6440 * @pf: board private structure
6441 * @queue_count: the requested queue count for rss.
6443 * returns 0 if rss is not enabled, if enabled returns the final rss queue
6444 * count which may be different from the requested queue count.
6446 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
6448 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
6451 queue_count = min_t(int, queue_count, pf->rss_size_max);
6452 queue_count = rounddown_pow_of_two(queue_count);
6454 if (queue_count != pf->rss_size) {
6455 i40e_prep_for_reset(pf);
6457 pf->rss_size = queue_count;
6459 i40e_reset_and_rebuild(pf, true);
6460 i40e_config_rss(pf);
6462 dev_info(&pf->pdev->dev, "RSS count: %d\n", pf->rss_size);
6463 return pf->rss_size;
6467 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
6468 * @pf: board private structure to initialize
6470 * i40e_sw_init initializes the Adapter private data structure.
6471 * Fields are initialized based on PCI device information and
6472 * OS network device settings (MTU size).
6474 static int i40e_sw_init(struct i40e_pf *pf)
6479 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
6480 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
6481 pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
6482 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
6483 if (I40E_DEBUG_USER & debug)
6484 pf->hw.debug_mask = debug;
6485 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
6486 I40E_DEFAULT_MSG_ENABLE);
6489 /* Set default capability flags */
6490 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
6491 I40E_FLAG_MSI_ENABLED |
6492 I40E_FLAG_MSIX_ENABLED |
6493 I40E_FLAG_RX_1BUF_ENABLED;
6495 /* Set default ITR */
6496 pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
6497 pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
6499 /* Depending on PF configurations, it is possible that the RSS
6500 * maximum might end up larger than the available queues
6502 pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
6503 pf->rss_size_max = min_t(int, pf->rss_size_max,
6504 pf->hw.func_caps.num_tx_qp);
6505 if (pf->hw.func_caps.rss) {
6506 pf->flags |= I40E_FLAG_RSS_ENABLED;
6507 pf->rss_size = min_t(int, pf->rss_size_max, num_online_cpus());
6508 pf->rss_size = rounddown_pow_of_two(pf->rss_size);
6513 /* MFP mode enabled */
6514 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
6515 pf->flags |= I40E_FLAG_MFP_ENABLED;
6516 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
6519 /* FW/NVM is not yet fixed in this regard */
6520 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
6521 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
6522 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
6523 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
6524 if (!(pf->flags & I40E_FLAG_MFP_ENABLED)) {
6525 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
6527 dev_info(&pf->pdev->dev,
6528 "Flow Director Sideband mode Disabled in MFP mode\n");
6530 pf->fdir_pf_filter_count =
6531 pf->hw.func_caps.fd_filters_guaranteed;
6532 pf->hw.fdir_shared_filter_count =
6533 pf->hw.func_caps.fd_filters_best_effort;
6536 if (pf->hw.func_caps.vmdq) {
6537 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
6538 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
6539 pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
6542 #ifdef CONFIG_PCI_IOV
6543 if (pf->hw.func_caps.num_vfs) {
6544 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
6545 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
6546 pf->num_req_vfs = min_t(int,
6547 pf->hw.func_caps.num_vfs,
6550 #endif /* CONFIG_PCI_IOV */
6551 pf->eeprom_version = 0xDEAD;
6552 pf->lan_veb = I40E_NO_VEB;
6553 pf->lan_vsi = I40E_NO_VSI;
6555 /* set up queue assignment tracking */
6556 size = sizeof(struct i40e_lump_tracking)
6557 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
6558 pf->qp_pile = kzalloc(size, GFP_KERNEL);
6563 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
6564 pf->qp_pile->search_hint = 0;
6566 /* set up vector assignment tracking */
6567 size = sizeof(struct i40e_lump_tracking)
6568 + (sizeof(u16) * pf->hw.func_caps.num_msix_vectors);
6569 pf->irq_pile = kzalloc(size, GFP_KERNEL);
6570 if (!pf->irq_pile) {
6575 pf->irq_pile->num_entries = pf->hw.func_caps.num_msix_vectors;
6576 pf->irq_pile->search_hint = 0;
6578 mutex_init(&pf->switch_mutex);
6585 * i40e_set_ntuple - set the ntuple feature flag and take action
6586 * @pf: board private structure to initialize
6587 * @features: the feature set that the stack is suggesting
6589 * returns a bool to indicate if reset needs to happen
6591 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
6593 bool need_reset = false;
6595 /* Check if Flow Director n-tuple support was enabled or disabled. If
6596 * the state changed, we need to reset.
6598 if (features & NETIF_F_NTUPLE) {
6599 /* Enable filters and mark for reset */
6600 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6602 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
6604 /* turn off filters, mark for reset and clear SW filter list */
6605 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
6607 i40e_fdir_filter_exit(pf);
6609 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6610 /* if ATR was disabled it can be re-enabled. */
6611 if (!(pf->flags & I40E_FLAG_FD_ATR_ENABLED))
6612 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
6618 * i40e_set_features - set the netdev feature flags
6619 * @netdev: ptr to the netdev being adjusted
6620 * @features: the feature set that the stack is suggesting
6622 static int i40e_set_features(struct net_device *netdev,
6623 netdev_features_t features)
6625 struct i40e_netdev_priv *np = netdev_priv(netdev);
6626 struct i40e_vsi *vsi = np->vsi;
6627 struct i40e_pf *pf = vsi->back;
6630 if (features & NETIF_F_HW_VLAN_CTAG_RX)
6631 i40e_vlan_stripping_enable(vsi);
6633 i40e_vlan_stripping_disable(vsi);
6635 need_reset = i40e_set_ntuple(pf, features);
6638 i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
6643 #ifdef CONFIG_I40E_VXLAN
6645 * i40e_get_vxlan_port_idx - Lookup a possibly offloaded for Rx UDP port
6646 * @pf: board private structure
6647 * @port: The UDP port to look up
6649 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
6651 static u8 i40e_get_vxlan_port_idx(struct i40e_pf *pf, __be16 port)
6655 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
6656 if (pf->vxlan_ports[i] == port)
6664 * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
6665 * @netdev: This physical port's netdev
6666 * @sa_family: Socket Family that VXLAN is notifying us about
6667 * @port: New UDP port number that VXLAN started listening to
6669 static void i40e_add_vxlan_port(struct net_device *netdev,
6670 sa_family_t sa_family, __be16 port)
6672 struct i40e_netdev_priv *np = netdev_priv(netdev);
6673 struct i40e_vsi *vsi = np->vsi;
6674 struct i40e_pf *pf = vsi->back;
6678 if (sa_family == AF_INET6)
6681 idx = i40e_get_vxlan_port_idx(pf, port);
6683 /* Check if port already exists */
6684 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
6685 netdev_info(netdev, "Port %d already offloaded\n", ntohs(port));
6689 /* Now check if there is space to add the new port */
6690 next_idx = i40e_get_vxlan_port_idx(pf, 0);
6692 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
6693 netdev_info(netdev, "Maximum number of UDP ports reached, not adding port %d\n",
6698 /* New port: add it and mark its index in the bitmap */
6699 pf->vxlan_ports[next_idx] = port;
6700 pf->pending_vxlan_bitmap |= (1 << next_idx);
6702 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
6706 * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
6707 * @netdev: This physical port's netdev
6708 * @sa_family: Socket Family that VXLAN is notifying us about
6709 * @port: UDP port number that VXLAN stopped listening to
6711 static void i40e_del_vxlan_port(struct net_device *netdev,
6712 sa_family_t sa_family, __be16 port)
6714 struct i40e_netdev_priv *np = netdev_priv(netdev);
6715 struct i40e_vsi *vsi = np->vsi;
6716 struct i40e_pf *pf = vsi->back;
6719 if (sa_family == AF_INET6)
6722 idx = i40e_get_vxlan_port_idx(pf, port);
6724 /* Check if port already exists */
6725 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
6726 /* if port exists, set it to 0 (mark for deletion)
6727 * and make it pending
6729 pf->vxlan_ports[idx] = 0;
6731 pf->pending_vxlan_bitmap |= (1 << idx);
6733 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
6735 netdev_warn(netdev, "Port %d was not found, not deleting\n",
6742 #ifdef USE_CONST_DEV_UC_CHAR
6743 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
6744 struct net_device *dev,
6745 const unsigned char *addr,
6748 static int i40e_ndo_fdb_add(struct ndmsg *ndm,
6749 struct net_device *dev,
6750 unsigned char *addr,
6754 struct i40e_netdev_priv *np = netdev_priv(dev);
6755 struct i40e_pf *pf = np->vsi->back;
6758 if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
6761 /* Hardware does not support aging addresses so if a
6762 * ndm_state is given only allow permanent addresses
6764 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
6765 netdev_info(dev, "FDB only supports static addresses\n");
6769 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
6770 err = dev_uc_add_excl(dev, addr);
6771 else if (is_multicast_ether_addr(addr))
6772 err = dev_mc_add_excl(dev, addr);
6776 /* Only return duplicate errors if NLM_F_EXCL is set */
6777 if (err == -EEXIST && !(flags & NLM_F_EXCL))
6783 #ifndef USE_DEFAULT_FDB_DEL_DUMP
6784 #ifdef USE_CONST_DEV_UC_CHAR
6785 static int i40e_ndo_fdb_del(struct ndmsg *ndm,
6786 struct net_device *dev,
6787 const unsigned char *addr)
6789 static int i40e_ndo_fdb_del(struct ndmsg *ndm,
6790 struct net_device *dev,
6791 unsigned char *addr)
6794 struct i40e_netdev_priv *np = netdev_priv(dev);
6795 struct i40e_pf *pf = np->vsi->back;
6796 int err = -EOPNOTSUPP;
6798 if (ndm->ndm_state & NUD_PERMANENT) {
6799 netdev_info(dev, "FDB only supports static addresses\n");
6803 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
6804 if (is_unicast_ether_addr(addr))
6805 err = dev_uc_del(dev, addr);
6806 else if (is_multicast_ether_addr(addr))
6807 err = dev_mc_del(dev, addr);
6815 static int i40e_ndo_fdb_dump(struct sk_buff *skb,
6816 struct netlink_callback *cb,
6817 struct net_device *dev,
6820 struct i40e_netdev_priv *np = netdev_priv(dev);
6821 struct i40e_pf *pf = np->vsi->back;
6823 if (pf->flags & I40E_FLAG_SRIOV_ENABLED)
6824 idx = ndo_dflt_fdb_dump(skb, cb, dev, idx);
6829 #endif /* USE_DEFAULT_FDB_DEL_DUMP */
6830 #endif /* HAVE_FDB_OPS */
6831 static const struct net_device_ops i40e_netdev_ops = {
6832 .ndo_open = i40e_open,
6833 .ndo_stop = i40e_close,
6834 .ndo_start_xmit = i40e_lan_xmit_frame,
6835 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
6836 .ndo_set_rx_mode = i40e_set_rx_mode,
6837 .ndo_validate_addr = eth_validate_addr,
6838 .ndo_set_mac_address = i40e_set_mac,
6839 .ndo_change_mtu = i40e_change_mtu,
6840 .ndo_do_ioctl = i40e_ioctl,
6841 .ndo_tx_timeout = i40e_tx_timeout,
6842 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
6843 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
6844 #ifdef CONFIG_NET_POLL_CONTROLLER
6845 .ndo_poll_controller = i40e_netpoll,
6847 .ndo_setup_tc = i40e_setup_tc,
6848 .ndo_set_features = i40e_set_features,
6849 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
6850 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
6851 .ndo_set_vf_rate = i40e_ndo_set_vf_bw,
6852 .ndo_get_vf_config = i40e_ndo_get_vf_config,
6853 .ndo_set_vf_link_state = i40e_ndo_set_vf_link_state,
6854 #ifdef CONFIG_I40E_VXLAN
6855 .ndo_add_vxlan_port = i40e_add_vxlan_port,
6856 .ndo_del_vxlan_port = i40e_del_vxlan_port,
6859 .ndo_fdb_add = i40e_ndo_fdb_add,
6860 #ifndef USE_DEFAULT_FDB_DEL_DUMP
6861 .ndo_fdb_del = i40e_ndo_fdb_del,
6862 .ndo_fdb_dump = i40e_ndo_fdb_dump,
6868 * i40e_config_netdev - Setup the netdev flags
6869 * @vsi: the VSI being configured
6871 * Returns 0 on success, negative value on failure
6873 static int i40e_config_netdev(struct i40e_vsi *vsi)
6875 u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
6876 struct i40e_pf *pf = vsi->back;
6877 struct i40e_hw *hw = &pf->hw;
6878 struct i40e_netdev_priv *np;
6879 struct net_device *netdev;
6880 u8 mac_addr[ETH_ALEN];
6883 etherdev_size = sizeof(struct i40e_netdev_priv);
6884 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
6888 vsi->netdev = netdev;
6889 np = netdev_priv(netdev);
6892 netdev->hw_enc_features |= NETIF_F_IP_CSUM |
6893 NETIF_F_GSO_UDP_TUNNEL |
6896 netdev->features = NETIF_F_SG |
6900 NETIF_F_GSO_UDP_TUNNEL |
6901 NETIF_F_HW_VLAN_CTAG_TX |
6902 NETIF_F_HW_VLAN_CTAG_RX |
6903 NETIF_F_HW_VLAN_CTAG_FILTER |
6912 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
6913 netdev->features |= NETIF_F_NTUPLE;
6915 /* copy netdev features into list of user selectable features */
6916 netdev->hw_features |= netdev->features;
6918 if (vsi->type == I40E_VSI_MAIN) {
6919 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
6920 memcpy(mac_addr, hw->mac.perm_addr, ETH_ALEN);
6922 /* relate the VSI_VMDQ name to the VSI_MAIN name */
6923 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
6924 pf->vsi[pf->lan_vsi]->netdev->name);
6925 random_ether_addr(mac_addr);
6926 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
6928 i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
6930 memcpy(netdev->dev_addr, mac_addr, ETH_ALEN);
6931 memcpy(netdev->perm_addr, mac_addr, ETH_ALEN);
6932 /* vlan gets same features (except vlan offload)
6933 * after any tweaks for specific VSI types
6935 netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
6936 NETIF_F_HW_VLAN_CTAG_RX |
6937 NETIF_F_HW_VLAN_CTAG_FILTER);
6938 netdev->priv_flags |= IFF_UNICAST_FLT;
6939 netdev->priv_flags |= IFF_SUPP_NOFCS;
6940 /* Setup netdev TC information */
6941 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
6943 netdev->netdev_ops = &i40e_netdev_ops;
6944 netdev->watchdog_timeo = 5 * HZ;
6945 i40e_set_ethtool_ops(netdev);
6951 * i40e_vsi_delete - Delete a VSI from the switch
6952 * @vsi: the VSI being removed
6954 * Returns 0 on success, negative value on failure
6956 static void i40e_vsi_delete(struct i40e_vsi *vsi)
6958 /* remove default VSI is not allowed */
6959 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
6962 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
6966 * i40e_add_vsi - Add a VSI to the switch
6967 * @vsi: the VSI being configured
6969 * This initializes a VSI context depending on the VSI type to be added and
6970 * passes it down to the add_vsi aq command.
6972 static int i40e_add_vsi(struct i40e_vsi *vsi)
6975 struct i40e_mac_filter *f, *ftmp;
6976 struct i40e_pf *pf = vsi->back;
6977 struct i40e_hw *hw = &pf->hw;
6978 struct i40e_vsi_context ctxt;
6979 u8 enabled_tc = 0x1; /* TC0 enabled */
6982 memset(&ctxt, 0, sizeof(ctxt));
6983 switch (vsi->type) {
6985 /* The PF's main VSI is already setup as part of the
6986 * device initialization, so we'll not bother with
6987 * the add_vsi call, but we will retrieve the current
6990 ctxt.seid = pf->main_vsi_seid;
6991 ctxt.pf_num = pf->hw.pf_id;
6993 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6994 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6996 dev_info(&pf->pdev->dev,
6997 "couldn't get pf vsi config, err %d, aq_err %d\n",
6998 ret, pf->hw.aq.asq_last_status);
7001 memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
7002 vsi->info.valid_sections = 0;
7004 vsi->seid = ctxt.seid;
7005 vsi->id = ctxt.vsi_number;
7007 enabled_tc = i40e_pf_get_tc_map(pf);
7009 /* MFP mode setup queue map and update VSI */
7010 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
7011 memset(&ctxt, 0, sizeof(ctxt));
7012 ctxt.seid = pf->main_vsi_seid;
7013 ctxt.pf_num = pf->hw.pf_id;
7015 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
7016 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
7018 dev_info(&pf->pdev->dev,
7019 "update vsi failed, aq_err=%d\n",
7020 pf->hw.aq.asq_last_status);
7024 /* update the local VSI info queue map */
7025 i40e_vsi_update_queue_map(vsi, &ctxt);
7026 vsi->info.valid_sections = 0;
7028 /* Default/Main VSI is only enabled for TC0
7029 * reconfigure it to enable all TCs that are
7030 * available on the port in SFP mode.
7032 ret = i40e_vsi_config_tc(vsi, enabled_tc);
7034 dev_info(&pf->pdev->dev,
7035 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
7037 pf->hw.aq.asq_last_status);
7044 ctxt.pf_num = hw->pf_id;
7046 ctxt.uplink_seid = vsi->uplink_seid;
7047 ctxt.connection_type = 0x1; /* regular data port */
7048 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
7049 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
7052 case I40E_VSI_VMDQ2:
7053 ctxt.pf_num = hw->pf_id;
7055 ctxt.uplink_seid = vsi->uplink_seid;
7056 ctxt.connection_type = 0x1; /* regular data port */
7057 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
7059 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
7061 /* This VSI is connected to VEB so the switch_id
7062 * should be set to zero by default.
7064 ctxt.info.switch_id = 0;
7065 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
7066 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
7068 /* Setup the VSI tx/rx queue map for TC0 only for now */
7069 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
7072 case I40E_VSI_SRIOV:
7073 ctxt.pf_num = hw->pf_id;
7074 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
7075 ctxt.uplink_seid = vsi->uplink_seid;
7076 ctxt.connection_type = 0x1; /* regular data port */
7077 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
7079 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
7081 /* This VSI is connected to VEB so the switch_id
7082 * should be set to zero by default.
7084 ctxt.info.switch_id = cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
7086 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
7087 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
7088 /* Setup the VSI tx/rx queue map for TC0 only for now */
7089 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
7096 if (vsi->type != I40E_VSI_MAIN) {
7097 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
7099 dev_info(&vsi->back->pdev->dev,
7100 "add vsi failed, aq_err=%d\n",
7101 vsi->back->hw.aq.asq_last_status);
7105 memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
7106 vsi->info.valid_sections = 0;
7107 vsi->seid = ctxt.seid;
7108 vsi->id = ctxt.vsi_number;
7111 /* If macvlan filters already exist, force them to get loaded */
7112 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
7117 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
7118 pf->flags |= I40E_FLAG_FILTER_SYNC;
7121 /* Update VSI BW information */
7122 ret = i40e_vsi_get_bw_info(vsi);
7124 dev_info(&pf->pdev->dev,
7125 "couldn't get vsi bw info, err %d, aq_err %d\n",
7126 ret, pf->hw.aq.asq_last_status);
7127 /* VSI is already added so not tearing that up */
7136 * i40e_vsi_release - Delete a VSI and free its resources
7137 * @vsi: the VSI being removed
7139 * Returns 0 on success or < 0 on error
7141 int i40e_vsi_release(struct i40e_vsi *vsi)
7143 struct i40e_mac_filter *f, *ftmp;
7144 struct i40e_veb *veb = NULL;
7151 /* release of a VEB-owner or last VSI is not allowed */
7152 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
7153 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
7154 vsi->seid, vsi->uplink_seid);
7157 if (vsi == pf->vsi[pf->lan_vsi] &&
7158 !test_bit(__I40E_DOWN, &pf->state)) {
7159 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
7163 uplink_seid = vsi->uplink_seid;
7164 if (vsi->type != I40E_VSI_SRIOV) {
7165 if (vsi->netdev_registered) {
7166 vsi->netdev_registered = false;
7168 /* results in a call to i40e_close() */
7169 unregister_netdev(vsi->netdev);
7172 i40e_vsi_close(vsi);
7174 i40e_vsi_disable_irq(vsi);
7177 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
7178 i40e_del_filter(vsi, f->macaddr, f->vlan,
7179 f->is_vf, f->is_netdev);
7180 i40e_sync_vsi_filters(vsi);
7182 i40e_vsi_delete(vsi);
7183 i40e_vsi_free_q_vectors(vsi);
7185 free_netdev(vsi->netdev);
7188 i40e_vsi_clear_rings(vsi);
7189 i40e_vsi_clear(vsi);
7191 /* If this was the last thing on the VEB, except for the
7192 * controlling VSI, remove the VEB, which puts the controlling
7193 * VSI onto the next level down in the switch.
7195 * Well, okay, there's one more exception here: don't remove
7196 * the orphan VEBs yet. We'll wait for an explicit remove request
7197 * from up the network stack.
7199 for (n = 0, i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7201 pf->vsi[i]->uplink_seid == uplink_seid &&
7202 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
7203 n++; /* count the VSIs */
7206 for (i = 0; i < I40E_MAX_VEB; i++) {
7209 if (pf->veb[i]->uplink_seid == uplink_seid)
7210 n++; /* count the VEBs */
7211 if (pf->veb[i]->seid == uplink_seid)
7214 if (n == 0 && veb && veb->uplink_seid != 0)
7215 i40e_veb_release(veb);
7221 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
7222 * @vsi: ptr to the VSI
7224 * This should only be called after i40e_vsi_mem_alloc() which allocates the
7225 * corresponding SW VSI structure and initializes num_queue_pairs for the
7226 * newly allocated VSI.
7228 * Returns 0 on success or negative on failure
7230 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
7233 struct i40e_pf *pf = vsi->back;
7235 if (vsi->q_vectors[0]) {
7236 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
7241 if (vsi->base_vector) {
7242 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
7243 vsi->seid, vsi->base_vector);
7247 ret = i40e_vsi_alloc_q_vectors(vsi);
7249 dev_info(&pf->pdev->dev,
7250 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
7251 vsi->num_q_vectors, vsi->seid, ret);
7252 vsi->num_q_vectors = 0;
7253 goto vector_setup_out;
7256 if (vsi->num_q_vectors)
7257 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
7258 vsi->num_q_vectors, vsi->idx);
7259 if (vsi->base_vector < 0) {
7260 dev_info(&pf->pdev->dev,
7261 "failed to get queue tracking for VSI %d, err=%d\n",
7262 vsi->seid, vsi->base_vector);
7263 i40e_vsi_free_q_vectors(vsi);
7265 goto vector_setup_out;
7273 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
7274 * @vsi: pointer to the vsi.
7276 * This re-allocates a vsi's queue resources.
7278 * Returns pointer to the successfully allocated and configured VSI sw struct
7279 * on success, otherwise returns NULL on failure.
7281 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
7283 struct i40e_pf *pf = vsi->back;
7287 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7288 i40e_vsi_clear_rings(vsi);
7290 i40e_vsi_free_arrays(vsi, false);
7291 i40e_set_num_rings_in_vsi(vsi);
7292 ret = i40e_vsi_alloc_arrays(vsi, false);
7296 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
7298 dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
7302 vsi->base_queue = ret;
7304 /* Update the FW view of the VSI. Force a reset of TC and queue
7305 * layout configurations.
7307 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
7308 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
7309 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
7310 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
7312 /* assign it some queues */
7313 ret = i40e_alloc_rings(vsi);
7317 /* map all of the rings to the q_vectors */
7318 i40e_vsi_map_rings_to_vectors(vsi);
7322 i40e_vsi_free_q_vectors(vsi);
7323 if (vsi->netdev_registered) {
7324 vsi->netdev_registered = false;
7325 unregister_netdev(vsi->netdev);
7326 free_netdev(vsi->netdev);
7329 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
7331 i40e_vsi_clear(vsi);
7336 * i40e_vsi_setup - Set up a VSI by a given type
7337 * @pf: board private structure
7339 * @uplink_seid: the switch element to link to
7340 * @param1: usage depends upon VSI type. For VF types, indicates VF id
7342 * This allocates the sw VSI structure and its queue resources, then add a VSI
7343 * to the identified VEB.
7345 * Returns pointer to the successfully allocated and configure VSI sw struct on
7346 * success, otherwise returns NULL on failure.
7348 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
7349 u16 uplink_seid, u32 param1)
7351 struct i40e_vsi *vsi = NULL;
7352 struct i40e_veb *veb = NULL;
7356 /* The requested uplink_seid must be either
7357 * - the PF's port seid
7358 * no VEB is needed because this is the PF
7359 * or this is a Flow Director special case VSI
7360 * - seid of an existing VEB
7361 * - seid of a VSI that owns an existing VEB
7362 * - seid of a VSI that doesn't own a VEB
7363 * a new VEB is created and the VSI becomes the owner
7364 * - seid of the PF VSI, which is what creates the first VEB
7365 * this is a special case of the previous
7367 * Find which uplink_seid we were given and create a new VEB if needed
7369 for (i = 0; i < I40E_MAX_VEB; i++) {
7370 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
7376 if (!veb && uplink_seid != pf->mac_seid) {
7378 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7379 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
7385 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
7390 if (vsi->uplink_seid == pf->mac_seid)
7391 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
7392 vsi->tc_config.enabled_tc);
7393 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
7394 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
7395 vsi->tc_config.enabled_tc);
7397 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
7398 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
7402 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
7406 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
7407 uplink_seid = veb->seid;
7410 /* get vsi sw struct */
7411 v_idx = i40e_vsi_mem_alloc(pf, type);
7414 vsi = pf->vsi[v_idx];
7418 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
7420 if (type == I40E_VSI_MAIN)
7421 pf->lan_vsi = v_idx;
7422 else if (type == I40E_VSI_SRIOV)
7423 vsi->vf_id = param1;
7424 /* assign it some queues */
7425 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
7428 dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
7432 vsi->base_queue = ret;
7434 /* get a VSI from the hardware */
7435 vsi->uplink_seid = uplink_seid;
7436 ret = i40e_add_vsi(vsi);
7440 switch (vsi->type) {
7441 /* setup the netdev if needed */
7443 case I40E_VSI_VMDQ2:
7444 ret = i40e_config_netdev(vsi);
7447 ret = register_netdev(vsi->netdev);
7450 vsi->netdev_registered = true;
7451 netif_carrier_off(vsi->netdev);
7452 #ifdef CONFIG_I40E_DCB
7453 /* Setup DCB netlink interface */
7454 i40e_dcbnl_setup(vsi);
7455 #endif /* CONFIG_I40E_DCB */
7459 /* set up vectors and rings if needed */
7460 ret = i40e_vsi_setup_vectors(vsi);
7464 ret = i40e_alloc_rings(vsi);
7468 /* map all of the rings to the q_vectors */
7469 i40e_vsi_map_rings_to_vectors(vsi);
7471 i40e_vsi_reset_stats(vsi);
7475 /* no netdev or rings for the other VSI types */
7482 i40e_vsi_free_q_vectors(vsi);
7484 if (vsi->netdev_registered) {
7485 vsi->netdev_registered = false;
7486 unregister_netdev(vsi->netdev);
7487 free_netdev(vsi->netdev);
7491 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
7493 i40e_vsi_clear(vsi);
7499 * i40e_veb_get_bw_info - Query VEB BW information
7500 * @veb: the veb to query
7502 * Query the Tx scheduler BW configuration data for given VEB
7504 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
7506 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
7507 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
7508 struct i40e_pf *pf = veb->pf;
7509 struct i40e_hw *hw = &pf->hw;
7514 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
7517 dev_info(&pf->pdev->dev,
7518 "query veb bw config failed, aq_err=%d\n",
7519 hw->aq.asq_last_status);
7523 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
7526 dev_info(&pf->pdev->dev,
7527 "query veb bw ets config failed, aq_err=%d\n",
7528 hw->aq.asq_last_status);
7532 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
7533 veb->bw_max_quanta = ets_data.tc_bw_max;
7534 veb->is_abs_credits = bw_data.absolute_credits_enable;
7535 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
7536 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
7537 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
7538 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
7539 veb->bw_tc_limit_credits[i] =
7540 le16_to_cpu(bw_data.tc_bw_limits[i]);
7541 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
7549 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
7550 * @pf: board private structure
7552 * On error: returns error code (negative)
7553 * On success: returns vsi index in PF (positive)
7555 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
7558 struct i40e_veb *veb;
7561 /* Need to protect the allocation of switch elements at the PF level */
7562 mutex_lock(&pf->switch_mutex);
7564 /* VEB list may be fragmented if VEB creation/destruction has
7565 * been happening. We can afford to do a quick scan to look
7566 * for any free slots in the list.
7568 * find next empty veb slot, looping back around if necessary
7571 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
7573 if (i >= I40E_MAX_VEB) {
7575 goto err_alloc_veb; /* out of VEB slots! */
7578 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
7585 veb->enabled_tc = 1;
7590 mutex_unlock(&pf->switch_mutex);
7595 * i40e_switch_branch_release - Delete a branch of the switch tree
7596 * @branch: where to start deleting
7598 * This uses recursion to find the tips of the branch to be
7599 * removed, deleting until we get back to and can delete this VEB.
7601 static void i40e_switch_branch_release(struct i40e_veb *branch)
7603 struct i40e_pf *pf = branch->pf;
7604 u16 branch_seid = branch->seid;
7605 u16 veb_idx = branch->idx;
7608 /* release any VEBs on this VEB - RECURSION */
7609 for (i = 0; i < I40E_MAX_VEB; i++) {
7612 if (pf->veb[i]->uplink_seid == branch->seid)
7613 i40e_switch_branch_release(pf->veb[i]);
7616 /* Release the VSIs on this VEB, but not the owner VSI.
7618 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
7619 * the VEB itself, so don't use (*branch) after this loop.
7621 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7624 if (pf->vsi[i]->uplink_seid == branch_seid &&
7625 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
7626 i40e_vsi_release(pf->vsi[i]);
7630 /* There's one corner case where the VEB might not have been
7631 * removed, so double check it here and remove it if needed.
7632 * This case happens if the veb was created from the debugfs
7633 * commands and no VSIs were added to it.
7635 if (pf->veb[veb_idx])
7636 i40e_veb_release(pf->veb[veb_idx]);
7640 * i40e_veb_clear - remove veb struct
7641 * @veb: the veb to remove
7643 static void i40e_veb_clear(struct i40e_veb *veb)
7649 struct i40e_pf *pf = veb->pf;
7651 mutex_lock(&pf->switch_mutex);
7652 if (pf->veb[veb->idx] == veb)
7653 pf->veb[veb->idx] = NULL;
7654 mutex_unlock(&pf->switch_mutex);
7661 * i40e_veb_release - Delete a VEB and free its resources
7662 * @veb: the VEB being removed
7664 void i40e_veb_release(struct i40e_veb *veb)
7666 struct i40e_vsi *vsi = NULL;
7672 /* find the remaining VSI and check for extras */
7673 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7674 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
7680 dev_info(&pf->pdev->dev,
7681 "can't remove VEB %d with %d VSIs left\n",
7686 /* move the remaining VSI to uplink veb */
7687 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
7688 if (veb->uplink_seid) {
7689 vsi->uplink_seid = veb->uplink_seid;
7690 if (veb->uplink_seid == pf->mac_seid)
7691 vsi->veb_idx = I40E_NO_VEB;
7693 vsi->veb_idx = veb->veb_idx;
7696 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
7697 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
7700 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
7701 i40e_veb_clear(veb);
7705 * i40e_add_veb - create the VEB in the switch
7706 * @veb: the VEB to be instantiated
7707 * @vsi: the controlling VSI
7709 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
7711 bool is_default = false;
7712 bool is_cloud = false;
7715 /* get a VEB from the hardware */
7716 ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
7717 veb->enabled_tc, is_default,
7718 is_cloud, &veb->seid, NULL);
7720 dev_info(&veb->pf->pdev->dev,
7721 "couldn't add VEB, err %d, aq_err %d\n",
7722 ret, veb->pf->hw.aq.asq_last_status);
7726 /* get statistics counter */
7727 ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
7728 &veb->stats_idx, NULL, NULL, NULL);
7730 dev_info(&veb->pf->pdev->dev,
7731 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
7732 ret, veb->pf->hw.aq.asq_last_status);
7735 ret = i40e_veb_get_bw_info(veb);
7737 dev_info(&veb->pf->pdev->dev,
7738 "couldn't get VEB bw info, err %d, aq_err %d\n",
7739 ret, veb->pf->hw.aq.asq_last_status);
7740 i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
7744 vsi->uplink_seid = veb->seid;
7745 vsi->veb_idx = veb->idx;
7746 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
7752 * i40e_veb_setup - Set up a VEB
7753 * @pf: board private structure
7754 * @flags: VEB setup flags
7755 * @uplink_seid: the switch element to link to
7756 * @vsi_seid: the initial VSI seid
7757 * @enabled_tc: Enabled TC bit-map
7759 * This allocates the sw VEB structure and links it into the switch
7760 * It is possible and legal for this to be a duplicate of an already
7761 * existing VEB. It is also possible for both uplink and vsi seids
7762 * to be zero, in order to create a floating VEB.
7764 * Returns pointer to the successfully allocated VEB sw struct on
7765 * success, otherwise returns NULL on failure.
7767 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
7768 u16 uplink_seid, u16 vsi_seid,
7771 struct i40e_veb *veb, *uplink_veb = NULL;
7772 int vsi_idx, veb_idx;
7775 /* if one seid is 0, the other must be 0 to create a floating relay */
7776 if ((uplink_seid == 0 || vsi_seid == 0) &&
7777 (uplink_seid + vsi_seid != 0)) {
7778 dev_info(&pf->pdev->dev,
7779 "one, not both seid's are 0: uplink=%d vsi=%d\n",
7780 uplink_seid, vsi_seid);
7784 /* make sure there is such a vsi and uplink */
7785 for (vsi_idx = 0; vsi_idx < pf->hw.func_caps.num_vsis; vsi_idx++)
7786 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
7788 if (vsi_idx >= pf->hw.func_caps.num_vsis && vsi_seid != 0) {
7789 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
7794 if (uplink_seid && uplink_seid != pf->mac_seid) {
7795 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
7796 if (pf->veb[veb_idx] &&
7797 pf->veb[veb_idx]->seid == uplink_seid) {
7798 uplink_veb = pf->veb[veb_idx];
7803 dev_info(&pf->pdev->dev,
7804 "uplink seid %d not found\n", uplink_seid);
7809 /* get veb sw struct */
7810 veb_idx = i40e_veb_mem_alloc(pf);
7813 veb = pf->veb[veb_idx];
7815 veb->uplink_seid = uplink_seid;
7816 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
7817 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
7819 /* create the VEB in the switch */
7820 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
7823 if (vsi_idx == pf->lan_vsi)
7824 pf->lan_veb = veb->idx;
7829 i40e_veb_clear(veb);
7835 * i40e_setup_pf_switch_element - set pf vars based on switch type
7836 * @pf: board private structure
7837 * @ele: element we are building info from
7838 * @num_reported: total number of elements
7839 * @printconfig: should we print the contents
7841 * helper function to assist in extracting a few useful SEID values.
7843 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
7844 struct i40e_aqc_switch_config_element_resp *ele,
7845 u16 num_reported, bool printconfig)
7847 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
7848 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
7849 u8 element_type = ele->element_type;
7850 u16 seid = le16_to_cpu(ele->seid);
7853 dev_info(&pf->pdev->dev,
7854 "type=%d seid=%d uplink=%d downlink=%d\n",
7855 element_type, seid, uplink_seid, downlink_seid);
7857 switch (element_type) {
7858 case I40E_SWITCH_ELEMENT_TYPE_MAC:
7859 pf->mac_seid = seid;
7861 case I40E_SWITCH_ELEMENT_TYPE_VEB:
7863 if (uplink_seid != pf->mac_seid)
7865 if (pf->lan_veb == I40E_NO_VEB) {
7868 /* find existing or else empty VEB */
7869 for (v = 0; v < I40E_MAX_VEB; v++) {
7870 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
7875 if (pf->lan_veb == I40E_NO_VEB) {
7876 v = i40e_veb_mem_alloc(pf);
7883 pf->veb[pf->lan_veb]->seid = seid;
7884 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
7885 pf->veb[pf->lan_veb]->pf = pf;
7886 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
7888 case I40E_SWITCH_ELEMENT_TYPE_VSI:
7889 if (num_reported != 1)
7891 /* This is immediately after a reset so we can assume this is
7894 pf->mac_seid = uplink_seid;
7895 pf->pf_seid = downlink_seid;
7896 pf->main_vsi_seid = seid;
7898 dev_info(&pf->pdev->dev,
7899 "pf_seid=%d main_vsi_seid=%d\n",
7900 pf->pf_seid, pf->main_vsi_seid);
7902 case I40E_SWITCH_ELEMENT_TYPE_PF:
7903 case I40E_SWITCH_ELEMENT_TYPE_VF:
7904 case I40E_SWITCH_ELEMENT_TYPE_EMP:
7905 case I40E_SWITCH_ELEMENT_TYPE_BMC:
7906 case I40E_SWITCH_ELEMENT_TYPE_PE:
7907 case I40E_SWITCH_ELEMENT_TYPE_PA:
7908 /* ignore these for now */
7911 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
7912 element_type, seid);
7918 * i40e_fetch_switch_configuration - Get switch config from firmware
7919 * @pf: board private structure
7920 * @printconfig: should we print the contents
7922 * Get the current switch configuration from the device and
7923 * extract a few useful SEID values.
7925 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
7927 struct i40e_aqc_get_switch_config_resp *sw_config;
7933 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
7937 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
7939 u16 num_reported, num_total;
7941 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
7945 dev_info(&pf->pdev->dev,
7946 "get switch config failed %d aq_err=%x\n",
7947 ret, pf->hw.aq.asq_last_status);
7952 num_reported = le16_to_cpu(sw_config->header.num_reported);
7953 num_total = le16_to_cpu(sw_config->header.num_total);
7956 dev_info(&pf->pdev->dev,
7957 "header: %d reported %d total\n",
7958 num_reported, num_total);
7961 int sz = sizeof(*sw_config) * num_reported;
7963 kfree(pf->sw_config);
7964 pf->sw_config = kzalloc(sz, GFP_KERNEL);
7966 memcpy(pf->sw_config, sw_config, sz);
7969 for (i = 0; i < num_reported; i++) {
7970 struct i40e_aqc_switch_config_element_resp *ele =
7971 &sw_config->element[i];
7973 i40e_setup_pf_switch_element(pf, ele, num_reported,
7976 } while (next_seid != 0);
7983 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
7984 * @pf: board private structure
7985 * @reinit: if the Main VSI needs to re-initialized.
7987 * Returns 0 on success, negative value on failure
7989 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
7991 u32 rxfc = 0, txfc = 0, rxfc_reg;
7994 /* find out what's out there already */
7995 ret = i40e_fetch_switch_configuration(pf, false);
7997 dev_info(&pf->pdev->dev,
7998 "couldn't fetch switch config, err %d, aq_err %d\n",
7999 ret, pf->hw.aq.asq_last_status);
8002 i40e_pf_reset_stats(pf);
8004 /* first time setup */
8005 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
8006 struct i40e_vsi *vsi = NULL;
8009 /* Set up the PF VSI associated with the PF's main VSI
8010 * that is already in the HW switch
8012 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
8013 uplink_seid = pf->veb[pf->lan_veb]->seid;
8015 uplink_seid = pf->mac_seid;
8016 if (pf->lan_vsi == I40E_NO_VSI)
8017 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
8019 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
8021 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
8022 i40e_fdir_teardown(pf);
8026 /* force a reset of TC and queue layout configurations */
8027 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
8028 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
8029 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
8030 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
8032 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
8034 i40e_fdir_sb_setup(pf);
8036 /* Setup static PF queue filter control settings */
8037 ret = i40e_setup_pf_filter_control(pf);
8039 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
8041 /* Failure here should not stop continuing other steps */
8044 /* enable RSS in the HW, even for only one queue, as the stack can use
8047 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
8048 i40e_config_rss(pf);
8050 /* fill in link information and enable LSE reporting */
8051 i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
8052 i40e_link_event(pf);
8054 /* Initialize user-specific link properties */
8055 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
8056 I40E_AQ_AN_COMPLETED) ? true : false);
8057 /* requested_mode is set in probe or by ethtool */
8058 if (!pf->fc_autoneg_status)
8061 if ((pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX) &&
8062 (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX))
8063 pf->hw.fc.current_mode = I40E_FC_FULL;
8064 else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX)
8065 pf->hw.fc.current_mode = I40E_FC_TX_PAUSE;
8066 else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX)
8067 pf->hw.fc.current_mode = I40E_FC_RX_PAUSE;
8069 pf->hw.fc.current_mode = I40E_FC_NONE;
8071 /* sync the flow control settings with the auto-neg values */
8072 switch (pf->hw.fc.current_mode) {
8077 case I40E_FC_TX_PAUSE:
8081 case I40E_FC_RX_PAUSE:
8086 case I40E_FC_DEFAULT:
8093 /* no default case, we have to handle all possibilities here */
8096 wr32(&pf->hw, I40E_PRTDCB_FCCFG, txfc << I40E_PRTDCB_FCCFG_TFCE_SHIFT);
8098 rxfc_reg = rd32(&pf->hw, I40E_PRTDCB_MFLCN) &
8099 ~I40E_PRTDCB_MFLCN_RFCE_MASK;
8100 rxfc_reg |= (rxfc << I40E_PRTDCB_MFLCN_RFCE_SHIFT);
8102 wr32(&pf->hw, I40E_PRTDCB_MFLCN, rxfc_reg);
8107 /* disable L2 flow control, user can turn it on if they wish */
8108 wr32(&pf->hw, I40E_PRTDCB_FCCFG, 0);
8109 wr32(&pf->hw, I40E_PRTDCB_MFLCN, rd32(&pf->hw, I40E_PRTDCB_MFLCN) &
8110 ~I40E_PRTDCB_MFLCN_RFCE_MASK);
8119 * i40e_determine_queue_usage - Work out queue distribution
8120 * @pf: board private structure
8122 static void i40e_determine_queue_usage(struct i40e_pf *pf)
8126 pf->num_lan_qps = 0;
8128 /* Find the max queues to be put into basic use. We'll always be
8129 * using TC0, whether or not DCB is running, and TC0 will get the
8132 queues_left = pf->hw.func_caps.num_tx_qp;
8134 if ((queues_left == 1) ||
8135 !(pf->flags & I40E_FLAG_MSIX_ENABLED) ||
8136 !(pf->flags & (I40E_FLAG_RSS_ENABLED | I40E_FLAG_FD_SB_ENABLED |
8137 I40E_FLAG_DCB_ENABLED))) {
8138 /* one qp for PF, no queues for anything else */
8140 pf->rss_size = pf->num_lan_qps = 1;
8142 /* make sure all the fancies are disabled */
8143 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
8144 I40E_FLAG_FD_SB_ENABLED |
8145 I40E_FLAG_FD_ATR_ENABLED |
8146 I40E_FLAG_DCB_ENABLED |
8147 I40E_FLAG_SRIOV_ENABLED |
8148 I40E_FLAG_VMDQ_ENABLED);
8150 /* Not enough queues for all TCs */
8151 if ((pf->flags & I40E_FLAG_DCB_ENABLED) &&
8152 (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
8153 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8154 dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
8156 pf->num_lan_qps = pf->rss_size_max;
8157 queues_left -= pf->num_lan_qps;
8160 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8161 if (queues_left > 1) {
8162 queues_left -= 1; /* save 1 queue for FD */
8164 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8165 dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
8169 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
8170 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
8171 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
8172 (queues_left / pf->num_vf_qps));
8173 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
8176 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
8177 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
8178 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
8179 (queues_left / pf->num_vmdq_qps));
8180 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
8183 pf->queues_left = queues_left;
8187 * i40e_setup_pf_filter_control - Setup PF static filter control
8188 * @pf: PF to be setup
8190 * i40e_setup_pf_filter_control sets up a pf's initial filter control
8191 * settings. If PE/FCoE are enabled then it will also set the per PF
8192 * based filter sizes required for them. It also enables Flow director,
8193 * ethertype and macvlan type filter settings for the pf.
8195 * Returns 0 on success, negative on failure
8197 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
8199 struct i40e_filter_control_settings *settings = &pf->filter_settings;
8201 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
8203 /* Flow Director is enabled */
8204 if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
8205 settings->enable_fdir = true;
8207 /* Ethtype and MACVLAN filters enabled for PF */
8208 settings->enable_ethtype = true;
8209 settings->enable_macvlan = true;
8211 if (i40e_set_filter_control(&pf->hw, settings))
8217 #define INFO_STRING_LEN 255
8218 static void i40e_print_features(struct i40e_pf *pf)
8220 struct i40e_hw *hw = &pf->hw;
8223 string = kzalloc(INFO_STRING_LEN, GFP_KERNEL);
8225 dev_err(&pf->pdev->dev, "Features string allocation failed\n");
8231 buf += sprintf(string, "Features: PF-id[%d] ", hw->pf_id);
8232 #ifdef CONFIG_PCI_IOV
8233 buf += sprintf(buf, "VFs: %d ", pf->num_req_vfs);
8235 buf += sprintf(buf, "VSIs: %d QP: %d ", pf->hw.func_caps.num_vsis,
8236 pf->vsi[pf->lan_vsi]->num_queue_pairs);
8238 if (pf->flags & I40E_FLAG_RSS_ENABLED)
8239 buf += sprintf(buf, "RSS ");
8240 buf += sprintf(buf, "FDir ");
8241 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
8242 buf += sprintf(buf, "ATR ");
8243 if (pf->flags & I40E_FLAG_FD_SB_ENABLED)
8244 buf += sprintf(buf, "NTUPLE ");
8245 if (pf->flags & I40E_FLAG_DCB_ENABLED)
8246 buf += sprintf(buf, "DCB ");
8247 if (pf->flags & I40E_FLAG_PTP)
8248 buf += sprintf(buf, "PTP ");
8250 BUG_ON(buf > (string + INFO_STRING_LEN));
8251 dev_info(&pf->pdev->dev, "%s\n", string);
8256 * i40e_probe - Device initialization routine
8257 * @pdev: PCI device information struct
8258 * @ent: entry in i40e_pci_tbl
8260 * i40e_probe initializes a pf identified by a pci_dev structure.
8261 * The OS initialization, configuring of the pf private structure,
8262 * and a hardware reset occur.
8264 * Returns 0 on success, negative on failure
8266 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
8270 static u16 pfs_found;
8276 err = pci_enable_device_mem(pdev);
8280 /* set up for high or low dma */
8281 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
8283 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
8286 "DMA configuration failed: 0x%x\n", err);
8291 /* set up pci connections */
8292 err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
8293 IORESOURCE_MEM), i40e_driver_name);
8295 dev_info(&pdev->dev,
8296 "pci_request_selected_regions failed %d\n", err);
8300 pci_enable_pcie_error_reporting(pdev);
8301 pci_set_master(pdev);
8303 /* Now that we have a PCI connection, we need to do the
8304 * low level device setup. This is primarily setting up
8305 * the Admin Queue structures and then querying for the
8306 * device's current profile information.
8308 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
8315 set_bit(__I40E_DOWN, &pf->state);
8319 hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
8320 pci_resource_len(pdev, 0));
8323 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
8324 (unsigned int)pci_resource_start(pdev, 0),
8325 (unsigned int)pci_resource_len(pdev, 0), err);
8328 hw->vendor_id = pdev->vendor;
8329 hw->device_id = pdev->device;
8330 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
8331 hw->subsystem_vendor_id = pdev->subsystem_vendor;
8332 hw->subsystem_device_id = pdev->subsystem_device;
8333 hw->bus.device = PCI_SLOT(pdev->devfn);
8334 hw->bus.func = PCI_FUNC(pdev->devfn);
8335 pf->instance = pfs_found;
8337 /* do a special CORER for clearing PXE mode once at init */
8338 if (hw->revision_id == 0 &&
8339 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
8340 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
8345 i40e_clear_pxe_mode(hw);
8348 /* Reset here to make sure all is clean and to define PF 'n' */
8349 err = i40e_pf_reset(hw);
8351 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
8356 hw->aq.num_arq_entries = I40E_AQ_LEN;
8357 hw->aq.num_asq_entries = I40E_AQ_LEN;
8358 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
8359 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
8360 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
8361 snprintf(pf->misc_int_name, sizeof(pf->misc_int_name) - 1,
8363 dev_driver_string(&pf->pdev->dev), pf->hw.pf_id);
8365 err = i40e_init_shared_code(hw);
8367 dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
8371 /* set up a default setting for link flow control */
8372 pf->hw.fc.requested_mode = I40E_FC_NONE;
8374 err = i40e_init_adminq(hw);
8375 dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
8377 dev_info(&pdev->dev,
8378 "init_adminq failed: %d expecting API %02x.%02x\n",
8380 I40E_FW_API_VERSION_MAJOR, I40E_FW_API_VERSION_MINOR);
8384 i40e_verify_eeprom(pf);
8386 /* Rev 0 hardware was never productized */
8387 if (hw->revision_id < 1)
8388 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");
8390 i40e_clear_pxe_mode(hw);
8391 err = i40e_get_capabilities(pf);
8393 goto err_adminq_setup;
8395 err = i40e_sw_init(pf);
8397 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
8401 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
8402 hw->func_caps.num_rx_qp,
8403 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
8405 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
8406 goto err_init_lan_hmc;
8409 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
8411 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
8413 goto err_configure_lan_hmc;
8416 i40e_get_mac_addr(hw, hw->mac.addr);
8417 if (!is_valid_ether_addr(hw->mac.addr)) {
8418 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
8422 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
8423 memcpy(hw->mac.perm_addr, hw->mac.addr, ETH_ALEN);
8425 pci_set_drvdata(pdev, pf);
8426 pci_save_state(pdev);
8427 #ifdef CONFIG_I40E_DCB
8428 err = i40e_init_pf_dcb(pf);
8430 dev_info(&pdev->dev, "init_pf_dcb failed: %d\n", err);
8431 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
8432 /* Continue without DCB enabled */
8434 #endif /* CONFIG_I40E_DCB */
8436 /* set up periodic task facility */
8437 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
8438 pf->service_timer_period = HZ;
8440 INIT_WORK(&pf->service_task, i40e_service_task);
8441 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
8442 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
8443 pf->link_check_timeout = jiffies;
8445 /* WoL defaults to disabled */
8447 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
8449 /* set up the main switch operations */
8450 i40e_determine_queue_usage(pf);
8451 i40e_init_interrupt_scheme(pf);
8453 /* Set up the *vsi struct based on the number of VSIs in the HW,
8454 * and set up our local tracking of the MAIN PF vsi.
8456 len = sizeof(struct i40e_vsi *) * pf->hw.func_caps.num_vsis;
8457 pf->vsi = kzalloc(len, GFP_KERNEL);
8460 goto err_switch_setup;
8463 err = i40e_setup_pf_switch(pf, false);
8465 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
8468 /* if FDIR VSI was set up, start it now */
8469 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
8470 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
8471 i40e_vsi_open(pf->vsi[i]);
8476 /* The main driver is (mostly) up and happy. We need to set this state
8477 * before setting up the misc vector or we get a race and the vector
8478 * ends up disabled forever.
8480 clear_bit(__I40E_DOWN, &pf->state);
8482 /* In case of MSIX we are going to setup the misc vector right here
8483 * to handle admin queue events etc. In case of legacy and MSI
8484 * the misc functionality and queue processing is combined in
8485 * the same vector and that gets setup at open.
8487 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8488 err = i40e_setup_misc_vector(pf);
8490 dev_info(&pdev->dev,
8491 "setup of misc vector failed: %d\n", err);
8496 #ifdef CONFIG_PCI_IOV
8497 /* prep for VF support */
8498 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
8499 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
8500 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
8503 /* disable link interrupts for VFs */
8504 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
8505 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
8506 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
8509 if (pci_num_vf(pdev)) {
8510 dev_info(&pdev->dev,
8511 "Active VFs found, allocating resources.\n");
8512 err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
8514 dev_info(&pdev->dev,
8515 "Error %d allocating resources for existing VFs\n",
8519 #endif /* CONFIG_PCI_IOV */
8523 i40e_dbg_pf_init(pf);
8525 /* tell the firmware that we're starting */
8526 i40e_send_version(pf);
8528 /* since everything's happy, start the service_task timer */
8529 mod_timer(&pf->service_timer,
8530 round_jiffies(jiffies + pf->service_timer_period));
8532 /* Get the negotiated link width and speed from PCI config space */
8533 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA, &link_status);
8535 i40e_set_pci_config_data(hw, link_status);
8537 dev_info(&pdev->dev, "PCI-Express: %s %s\n",
8538 (hw->bus.speed == i40e_bus_speed_8000 ? "Speed 8.0GT/s" :
8539 hw->bus.speed == i40e_bus_speed_5000 ? "Speed 5.0GT/s" :
8540 hw->bus.speed == i40e_bus_speed_2500 ? "Speed 2.5GT/s" :
8542 (hw->bus.width == i40e_bus_width_pcie_x8 ? "Width x8" :
8543 hw->bus.width == i40e_bus_width_pcie_x4 ? "Width x4" :
8544 hw->bus.width == i40e_bus_width_pcie_x2 ? "Width x2" :
8545 hw->bus.width == i40e_bus_width_pcie_x1 ? "Width x1" :
8548 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
8549 hw->bus.speed < i40e_bus_speed_8000) {
8550 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
8551 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
8554 /* print a string summarizing features */
8555 i40e_print_features(pf);
8559 /* Unwind what we've done if something failed in the setup */
8561 set_bit(__I40E_DOWN, &pf->state);
8562 i40e_clear_interrupt_scheme(pf);
8565 i40e_reset_interrupt_capability(pf);
8566 del_timer_sync(&pf->service_timer);
8568 err_configure_lan_hmc:
8569 (void)i40e_shutdown_lan_hmc(hw);
8572 kfree(pf->irq_pile);
8575 (void)i40e_shutdown_adminq(hw);
8577 iounmap(hw->hw_addr);
8581 pci_disable_pcie_error_reporting(pdev);
8582 pci_release_selected_regions(pdev,
8583 pci_select_bars(pdev, IORESOURCE_MEM));
8586 pci_disable_device(pdev);
8591 * i40e_remove - Device removal routine
8592 * @pdev: PCI device information struct
8594 * i40e_remove is called by the PCI subsystem to alert the driver
8595 * that is should release a PCI device. This could be caused by a
8596 * Hot-Plug event, or because the driver is going to be removed from
8599 static void i40e_remove(struct pci_dev *pdev)
8601 struct i40e_pf *pf = pci_get_drvdata(pdev);
8602 i40e_status ret_code;
8606 i40e_dbg_pf_exit(pf);
8610 /* no more scheduling of any task */
8611 set_bit(__I40E_DOWN, &pf->state);
8612 del_timer_sync(&pf->service_timer);
8613 cancel_work_sync(&pf->service_task);
8615 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
8617 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
8620 i40e_fdir_teardown(pf);
8622 /* If there is a switch structure or any orphans, remove them.
8623 * This will leave only the PF's VSI remaining.
8625 for (i = 0; i < I40E_MAX_VEB; i++) {
8629 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
8630 pf->veb[i]->uplink_seid == 0)
8631 i40e_switch_branch_release(pf->veb[i]);
8634 /* Now we can shutdown the PF's VSI, just before we kill
8637 if (pf->vsi[pf->lan_vsi])
8638 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
8640 i40e_stop_misc_vector(pf);
8641 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
8642 synchronize_irq(pf->msix_entries[0].vector);
8643 free_irq(pf->msix_entries[0].vector, pf);
8646 /* shutdown and destroy the HMC */
8647 if (pf->hw.hmc.hmc_obj) {
8648 ret_code = i40e_shutdown_lan_hmc(&pf->hw);
8650 dev_warn(&pdev->dev,
8651 "Failed to destroy the HMC resources: %d\n",
8655 /* shutdown the adminq */
8656 ret_code = i40e_shutdown_adminq(&pf->hw);
8658 dev_warn(&pdev->dev,
8659 "Failed to destroy the Admin Queue resources: %d\n",
8662 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
8663 i40e_clear_interrupt_scheme(pf);
8664 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
8666 i40e_vsi_clear_rings(pf->vsi[i]);
8667 i40e_vsi_clear(pf->vsi[i]);
8672 for (i = 0; i < I40E_MAX_VEB; i++) {
8678 kfree(pf->irq_pile);
8679 kfree(pf->sw_config);
8682 /* force a PF reset to clean anything leftover */
8683 reg = rd32(&pf->hw, I40E_PFGEN_CTRL);
8684 wr32(&pf->hw, I40E_PFGEN_CTRL, (reg | I40E_PFGEN_CTRL_PFSWR_MASK));
8685 i40e_flush(&pf->hw);
8687 iounmap(pf->hw.hw_addr);
8689 pci_release_selected_regions(pdev,
8690 pci_select_bars(pdev, IORESOURCE_MEM));
8692 pci_disable_pcie_error_reporting(pdev);
8693 pci_disable_device(pdev);
8697 * i40e_pci_error_detected - warning that something funky happened in PCI land
8698 * @pdev: PCI device information struct
8700 * Called to warn that something happened and the error handling steps
8701 * are in progress. Allows the driver to quiesce things, be ready for
8704 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
8705 enum pci_channel_state error)
8707 struct i40e_pf *pf = pci_get_drvdata(pdev);
8709 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
8711 /* shutdown all operations */
8712 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
8714 i40e_prep_for_reset(pf);
8718 /* Request a slot reset */
8719 return PCI_ERS_RESULT_NEED_RESET;
8723 * i40e_pci_error_slot_reset - a PCI slot reset just happened
8724 * @pdev: PCI device information struct
8726 * Called to find if the driver can work with the device now that
8727 * the pci slot has been reset. If a basic connection seems good
8728 * (registers are readable and have sane content) then return a
8729 * happy little PCI_ERS_RESULT_xxx.
8731 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
8733 struct i40e_pf *pf = pci_get_drvdata(pdev);
8734 pci_ers_result_t result;
8738 dev_info(&pdev->dev, "%s\n", __func__);
8739 if (pci_enable_device_mem(pdev)) {
8740 dev_info(&pdev->dev,
8741 "Cannot re-enable PCI device after reset.\n");
8742 result = PCI_ERS_RESULT_DISCONNECT;
8744 pci_set_master(pdev);
8745 pci_restore_state(pdev);
8746 pci_save_state(pdev);
8747 pci_wake_from_d3(pdev, false);
8749 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
8751 result = PCI_ERS_RESULT_RECOVERED;
8753 result = PCI_ERS_RESULT_DISCONNECT;
8756 err = pci_cleanup_aer_uncorrect_error_status(pdev);
8758 dev_info(&pdev->dev,
8759 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
8761 /* non-fatal, continue */
8768 * i40e_pci_error_resume - restart operations after PCI error recovery
8769 * @pdev: PCI device information struct
8771 * Called to allow the driver to bring things back up after PCI error
8772 * and/or reset recovery has finished.
8774 static void i40e_pci_error_resume(struct pci_dev *pdev)
8776 struct i40e_pf *pf = pci_get_drvdata(pdev);
8778 dev_info(&pdev->dev, "%s\n", __func__);
8779 if (test_bit(__I40E_SUSPENDED, &pf->state))
8783 i40e_handle_reset_warning(pf);
8788 * i40e_shutdown - PCI callback for shutting down
8789 * @pdev: PCI device information struct
8791 static void i40e_shutdown(struct pci_dev *pdev)
8793 struct i40e_pf *pf = pci_get_drvdata(pdev);
8794 struct i40e_hw *hw = &pf->hw;
8796 set_bit(__I40E_SUSPENDED, &pf->state);
8797 set_bit(__I40E_DOWN, &pf->state);
8799 i40e_prep_for_reset(pf);
8802 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
8803 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
8805 if (system_state == SYSTEM_POWER_OFF) {
8806 pci_wake_from_d3(pdev, pf->wol_en);
8807 pci_set_power_state(pdev, PCI_D3hot);
8813 * i40e_suspend - PCI callback for moving to D3
8814 * @pdev: PCI device information struct
8816 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
8818 struct i40e_pf *pf = pci_get_drvdata(pdev);
8819 struct i40e_hw *hw = &pf->hw;
8821 set_bit(__I40E_SUSPENDED, &pf->state);
8822 set_bit(__I40E_DOWN, &pf->state);
8824 i40e_prep_for_reset(pf);
8827 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
8828 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
8830 pci_wake_from_d3(pdev, pf->wol_en);
8831 pci_set_power_state(pdev, PCI_D3hot);
8837 * i40e_resume - PCI callback for waking up from D3
8838 * @pdev: PCI device information struct
8840 static int i40e_resume(struct pci_dev *pdev)
8842 struct i40e_pf *pf = pci_get_drvdata(pdev);
8845 pci_set_power_state(pdev, PCI_D0);
8846 pci_restore_state(pdev);
8847 /* pci_restore_state() clears dev->state_saves, so
8848 * call pci_save_state() again to restore it.
8850 pci_save_state(pdev);
8852 err = pci_enable_device_mem(pdev);
8855 "%s: Cannot enable PCI device from suspend\n",
8859 pci_set_master(pdev);
8861 /* no wakeup events while running */
8862 pci_wake_from_d3(pdev, false);
8864 /* handling the reset will rebuild the device state */
8865 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
8866 clear_bit(__I40E_DOWN, &pf->state);
8868 i40e_reset_and_rebuild(pf, false);
8876 static const struct pci_error_handlers i40e_err_handler = {
8877 .error_detected = i40e_pci_error_detected,
8878 .slot_reset = i40e_pci_error_slot_reset,
8879 .resume = i40e_pci_error_resume,
8882 static struct pci_driver i40e_driver = {
8883 .name = i40e_driver_name,
8884 .id_table = i40e_pci_tbl,
8885 .probe = i40e_probe,
8886 .remove = i40e_remove,
8888 .suspend = i40e_suspend,
8889 .resume = i40e_resume,
8891 .shutdown = i40e_shutdown,
8892 .err_handler = &i40e_err_handler,
8893 .sriov_configure = i40e_pci_sriov_configure,
8897 * i40e_init_module - Driver registration routine
8899 * i40e_init_module is the first routine called when the driver is
8900 * loaded. All it does is register with the PCI subsystem.
8902 static int __init i40e_init_module(void)
8904 pr_info("%s: %s - version %s\n", i40e_driver_name,
8905 i40e_driver_string, i40e_driver_version_str);
8906 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
8908 return pci_register_driver(&i40e_driver);
8910 module_init(i40e_init_module);
8913 * i40e_exit_module - Driver exit cleanup routine
8915 * i40e_exit_module is called just before the driver is removed
8918 static void __exit i40e_exit_module(void)
8920 pci_unregister_driver(&i40e_driver);
8923 module_exit(i40e_exit_module);