1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 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 with
16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
19 * The full GNU General Public License is included in this distribution in
20 * the file called "COPYING".
22 * Contact Information:
23 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
26 ******************************************************************************/
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 3
42 #define DRV_VERSION_BUILD 25
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 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);
59 /* i40e_pci_tbl - PCI Device ID Table
61 * Last entry must be all 0s
63 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
64 * Class, Class Mask, private data (not used) }
66 static DEFINE_PCI_DEVICE_TABLE(i40e_pci_tbl) = {
67 {PCI_VDEVICE(INTEL, I40E_SFP_XL710_DEVICE_ID), 0},
68 {PCI_VDEVICE(INTEL, I40E_SFP_X710_DEVICE_ID), 0},
69 {PCI_VDEVICE(INTEL, I40E_QEMU_DEVICE_ID), 0},
70 {PCI_VDEVICE(INTEL, I40E_KX_A_DEVICE_ID), 0},
71 {PCI_VDEVICE(INTEL, I40E_KX_B_DEVICE_ID), 0},
72 {PCI_VDEVICE(INTEL, I40E_KX_C_DEVICE_ID), 0},
73 {PCI_VDEVICE(INTEL, I40E_KX_D_DEVICE_ID), 0},
74 {PCI_VDEVICE(INTEL, I40E_QSFP_A_DEVICE_ID), 0},
75 {PCI_VDEVICE(INTEL, I40E_QSFP_B_DEVICE_ID), 0},
76 {PCI_VDEVICE(INTEL, I40E_QSFP_C_DEVICE_ID), 0},
77 /* required last entry */
80 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
82 #define I40E_MAX_VF_COUNT 128
83 static int debug = -1;
84 module_param(debug, int, 0);
85 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
87 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
88 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
89 MODULE_LICENSE("GPL");
90 MODULE_VERSION(DRV_VERSION);
93 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
94 * @hw: pointer to the HW structure
95 * @mem: ptr to mem struct to fill out
96 * @size: size of memory requested
97 * @alignment: what to align the allocation to
99 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
100 u64 size, u32 alignment)
102 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
104 mem->size = ALIGN(size, alignment);
105 mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
106 &mem->pa, GFP_KERNEL);
114 * i40e_free_dma_mem_d - OS specific memory free for shared code
115 * @hw: pointer to the HW structure
116 * @mem: ptr to mem struct to free
118 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
120 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
122 dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
131 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
132 * @hw: pointer to the HW structure
133 * @mem: ptr to mem struct to fill out
134 * @size: size of memory requested
136 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
140 mem->va = kzalloc(size, GFP_KERNEL);
149 * i40e_free_virt_mem_d - OS specific memory free for shared code
150 * @hw: pointer to the HW structure
151 * @mem: ptr to mem struct to free
153 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
155 /* it's ok to kfree a NULL pointer */
164 * i40e_get_lump - find a lump of free generic resource
165 * @pf: board private structure
166 * @pile: the pile of resource to search
167 * @needed: the number of items needed
168 * @id: an owner id to stick on the items assigned
170 * Returns the base item index of the lump, or negative for error
172 * The search_hint trick and lack of advanced fit-finding only work
173 * because we're highly likely to have all the same size lump requests.
174 * Linear search time and any fragmentation should be minimal.
176 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
182 if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
183 dev_info(&pf->pdev->dev,
184 "param err: pile=%p needed=%d id=0x%04x\n",
189 /* start the linear search with an imperfect hint */
190 i = pile->search_hint;
191 while (i < pile->num_entries) {
192 /* skip already allocated entries */
193 if (pile->list[i] & I40E_PILE_VALID_BIT) {
198 /* do we have enough in this lump? */
199 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
200 if (pile->list[i+j] & I40E_PILE_VALID_BIT)
205 /* there was enough, so assign it to the requestor */
206 for (j = 0; j < needed; j++)
207 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
209 pile->search_hint = i + j;
212 /* not enough, so skip over it and continue looking */
221 * i40e_put_lump - return a lump of generic resource
222 * @pile: the pile of resource to search
223 * @index: the base item index
224 * @id: the owner id of the items assigned
226 * Returns the count of items in the lump
228 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
230 int valid_id = (id | I40E_PILE_VALID_BIT);
234 if (!pile || index >= pile->num_entries)
238 i < pile->num_entries && pile->list[i] == valid_id;
244 if (count && index < pile->search_hint)
245 pile->search_hint = index;
251 * i40e_service_event_schedule - Schedule the service task to wake up
252 * @pf: board private structure
254 * If not already scheduled, this puts the task into the work queue
256 static void i40e_service_event_schedule(struct i40e_pf *pf)
258 if (!test_bit(__I40E_DOWN, &pf->state) &&
259 !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
260 !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
261 schedule_work(&pf->service_task);
265 * i40e_tx_timeout - Respond to a Tx Hang
266 * @netdev: network interface device structure
268 * If any port has noticed a Tx timeout, it is likely that the whole
269 * device is munged, not just the one netdev port, so go for the full
272 static void i40e_tx_timeout(struct net_device *netdev)
274 struct i40e_netdev_priv *np = netdev_priv(netdev);
275 struct i40e_vsi *vsi = np->vsi;
276 struct i40e_pf *pf = vsi->back;
278 pf->tx_timeout_count++;
280 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
281 pf->tx_timeout_recovery_level = 0;
282 pf->tx_timeout_last_recovery = jiffies;
283 netdev_info(netdev, "tx_timeout recovery level %d\n",
284 pf->tx_timeout_recovery_level);
286 switch (pf->tx_timeout_recovery_level) {
288 /* disable and re-enable queues for the VSI */
289 if (in_interrupt()) {
290 set_bit(__I40E_REINIT_REQUESTED, &pf->state);
291 set_bit(__I40E_REINIT_REQUESTED, &vsi->state);
293 i40e_vsi_reinit_locked(vsi);
297 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
300 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
303 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
306 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
310 i40e_service_event_schedule(pf);
311 pf->tx_timeout_recovery_level++;
315 * i40e_release_rx_desc - Store the new tail and head values
316 * @rx_ring: ring to bump
317 * @val: new head index
319 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
321 rx_ring->next_to_use = val;
323 /* Force memory writes to complete before letting h/w
324 * know there are new descriptors to fetch. (Only
325 * applicable for weak-ordered memory model archs,
329 writel(val, rx_ring->tail);
333 * i40e_get_vsi_stats_struct - Get System Network Statistics
334 * @vsi: the VSI we care about
336 * Returns the address of the device statistics structure.
337 * The statistics are actually updated from the service task.
339 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
341 return &vsi->net_stats;
345 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
346 * @netdev: network interface device structure
348 * Returns the address of the device statistics structure.
349 * The statistics are actually updated from the service task.
351 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
352 struct net_device *netdev,
353 struct rtnl_link_stats64 *stats)
355 struct i40e_netdev_priv *np = netdev_priv(netdev);
356 struct i40e_vsi *vsi = np->vsi;
357 struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
361 if (test_bit(__I40E_DOWN, &vsi->state))
368 for (i = 0; i < vsi->num_queue_pairs; i++) {
369 struct i40e_ring *tx_ring, *rx_ring;
373 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
378 start = u64_stats_fetch_begin_bh(&tx_ring->syncp);
379 packets = tx_ring->stats.packets;
380 bytes = tx_ring->stats.bytes;
381 } while (u64_stats_fetch_retry_bh(&tx_ring->syncp, start));
383 stats->tx_packets += packets;
384 stats->tx_bytes += bytes;
385 rx_ring = &tx_ring[1];
388 start = u64_stats_fetch_begin_bh(&rx_ring->syncp);
389 packets = rx_ring->stats.packets;
390 bytes = rx_ring->stats.bytes;
391 } while (u64_stats_fetch_retry_bh(&rx_ring->syncp, start));
393 stats->rx_packets += packets;
394 stats->rx_bytes += bytes;
398 /* following stats updated by ixgbe_watchdog_task() */
399 stats->multicast = vsi_stats->multicast;
400 stats->tx_errors = vsi_stats->tx_errors;
401 stats->tx_dropped = vsi_stats->tx_dropped;
402 stats->rx_errors = vsi_stats->rx_errors;
403 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
404 stats->rx_length_errors = vsi_stats->rx_length_errors;
410 * i40e_vsi_reset_stats - Resets all stats of the given vsi
411 * @vsi: the VSI to have its stats reset
413 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
415 struct rtnl_link_stats64 *ns;
421 ns = i40e_get_vsi_stats_struct(vsi);
422 memset(ns, 0, sizeof(*ns));
423 memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
424 memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
425 memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
427 for (i = 0; i < vsi->num_queue_pairs; i++) {
428 memset(&vsi->rx_rings[i]->stats, 0 ,
429 sizeof(vsi->rx_rings[i]->stats));
430 memset(&vsi->rx_rings[i]->rx_stats, 0 ,
431 sizeof(vsi->rx_rings[i]->rx_stats));
432 memset(&vsi->tx_rings[i]->stats, 0 ,
433 sizeof(vsi->tx_rings[i]->stats));
434 memset(&vsi->tx_rings[i]->tx_stats, 0,
435 sizeof(vsi->tx_rings[i]->tx_stats));
437 vsi->stat_offsets_loaded = false;
441 * i40e_pf_reset_stats - Reset all of the stats for the given pf
442 * @pf: the PF to be reset
444 void i40e_pf_reset_stats(struct i40e_pf *pf)
446 memset(&pf->stats, 0, sizeof(pf->stats));
447 memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
448 pf->stat_offsets_loaded = false;
452 * i40e_stat_update48 - read and update a 48 bit stat from the chip
453 * @hw: ptr to the hardware info
454 * @hireg: the high 32 bit reg to read
455 * @loreg: the low 32 bit reg to read
456 * @offset_loaded: has the initial offset been loaded yet
457 * @offset: ptr to current offset value
458 * @stat: ptr to the stat
460 * Since the device stats are not reset at PFReset, they likely will not
461 * be zeroed when the driver starts. We'll save the first values read
462 * and use them as offsets to be subtracted from the raw values in order
463 * to report stats that count from zero. In the process, we also manage
464 * the potential roll-over.
466 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
467 bool offset_loaded, u64 *offset, u64 *stat)
471 if (hw->device_id == I40E_QEMU_DEVICE_ID) {
472 new_data = rd32(hw, loreg);
473 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
475 new_data = rd64(hw, loreg);
479 if (likely(new_data >= *offset))
480 *stat = new_data - *offset;
482 *stat = (new_data + ((u64)1 << 48)) - *offset;
483 *stat &= 0xFFFFFFFFFFFFULL;
487 * i40e_stat_update32 - read and update a 32 bit stat from the chip
488 * @hw: ptr to the hardware info
489 * @reg: the hw reg to read
490 * @offset_loaded: has the initial offset been loaded yet
491 * @offset: ptr to current offset value
492 * @stat: ptr to the stat
494 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
495 bool offset_loaded, u64 *offset, u64 *stat)
499 new_data = rd32(hw, reg);
502 if (likely(new_data >= *offset))
503 *stat = (u32)(new_data - *offset);
505 *stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
509 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
510 * @vsi: the VSI to be updated
512 void i40e_update_eth_stats(struct i40e_vsi *vsi)
514 int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
515 struct i40e_pf *pf = vsi->back;
516 struct i40e_hw *hw = &pf->hw;
517 struct i40e_eth_stats *oes;
518 struct i40e_eth_stats *es; /* device's eth stats */
520 es = &vsi->eth_stats;
521 oes = &vsi->eth_stats_offsets;
523 /* Gather up the stats that the hw collects */
524 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
525 vsi->stat_offsets_loaded,
526 &oes->tx_errors, &es->tx_errors);
527 i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
528 vsi->stat_offsets_loaded,
529 &oes->rx_discards, &es->rx_discards);
531 i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
532 I40E_GLV_GORCL(stat_idx),
533 vsi->stat_offsets_loaded,
534 &oes->rx_bytes, &es->rx_bytes);
535 i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
536 I40E_GLV_UPRCL(stat_idx),
537 vsi->stat_offsets_loaded,
538 &oes->rx_unicast, &es->rx_unicast);
539 i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
540 I40E_GLV_MPRCL(stat_idx),
541 vsi->stat_offsets_loaded,
542 &oes->rx_multicast, &es->rx_multicast);
543 i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
544 I40E_GLV_BPRCL(stat_idx),
545 vsi->stat_offsets_loaded,
546 &oes->rx_broadcast, &es->rx_broadcast);
548 i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
549 I40E_GLV_GOTCL(stat_idx),
550 vsi->stat_offsets_loaded,
551 &oes->tx_bytes, &es->tx_bytes);
552 i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
553 I40E_GLV_UPTCL(stat_idx),
554 vsi->stat_offsets_loaded,
555 &oes->tx_unicast, &es->tx_unicast);
556 i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
557 I40E_GLV_MPTCL(stat_idx),
558 vsi->stat_offsets_loaded,
559 &oes->tx_multicast, &es->tx_multicast);
560 i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
561 I40E_GLV_BPTCL(stat_idx),
562 vsi->stat_offsets_loaded,
563 &oes->tx_broadcast, &es->tx_broadcast);
564 vsi->stat_offsets_loaded = true;
568 * i40e_update_veb_stats - Update Switch component statistics
569 * @veb: the VEB being updated
571 static void i40e_update_veb_stats(struct i40e_veb *veb)
573 struct i40e_pf *pf = veb->pf;
574 struct i40e_hw *hw = &pf->hw;
575 struct i40e_eth_stats *oes;
576 struct i40e_eth_stats *es; /* device's eth stats */
579 idx = veb->stats_idx;
581 oes = &veb->stats_offsets;
583 /* Gather up the stats that the hw collects */
584 i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
585 veb->stat_offsets_loaded,
586 &oes->tx_discards, &es->tx_discards);
587 if (hw->revision_id > 0)
588 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
589 veb->stat_offsets_loaded,
590 &oes->rx_unknown_protocol,
591 &es->rx_unknown_protocol);
592 i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
593 veb->stat_offsets_loaded,
594 &oes->rx_bytes, &es->rx_bytes);
595 i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
596 veb->stat_offsets_loaded,
597 &oes->rx_unicast, &es->rx_unicast);
598 i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
599 veb->stat_offsets_loaded,
600 &oes->rx_multicast, &es->rx_multicast);
601 i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
602 veb->stat_offsets_loaded,
603 &oes->rx_broadcast, &es->rx_broadcast);
605 i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
606 veb->stat_offsets_loaded,
607 &oes->tx_bytes, &es->tx_bytes);
608 i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
609 veb->stat_offsets_loaded,
610 &oes->tx_unicast, &es->tx_unicast);
611 i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
612 veb->stat_offsets_loaded,
613 &oes->tx_multicast, &es->tx_multicast);
614 i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
615 veb->stat_offsets_loaded,
616 &oes->tx_broadcast, &es->tx_broadcast);
617 veb->stat_offsets_loaded = true;
621 * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
622 * @pf: the corresponding PF
624 * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
626 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
628 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
629 struct i40e_hw_port_stats *nsd = &pf->stats;
630 struct i40e_hw *hw = &pf->hw;
634 if ((hw->fc.current_mode != I40E_FC_FULL) &&
635 (hw->fc.current_mode != I40E_FC_RX_PAUSE))
638 xoff = nsd->link_xoff_rx;
639 i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
640 pf->stat_offsets_loaded,
641 &osd->link_xoff_rx, &nsd->link_xoff_rx);
643 /* No new LFC xoff rx */
644 if (!(nsd->link_xoff_rx - xoff))
647 /* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
648 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
649 struct i40e_vsi *vsi = pf->vsi[v];
654 for (i = 0; i < vsi->num_queue_pairs; i++) {
655 struct i40e_ring *ring = vsi->tx_rings[i];
656 clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
662 * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
663 * @pf: the corresponding PF
665 * Update the Rx XOFF counter (PAUSE frames) in PFC mode
667 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
669 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
670 struct i40e_hw_port_stats *nsd = &pf->stats;
671 bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
672 struct i40e_dcbx_config *dcb_cfg;
673 struct i40e_hw *hw = &pf->hw;
677 dcb_cfg = &hw->local_dcbx_config;
679 /* See if DCB enabled with PFC TC */
680 if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
681 !(dcb_cfg->pfc.pfcenable)) {
682 i40e_update_link_xoff_rx(pf);
686 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
687 u64 prio_xoff = nsd->priority_xoff_rx[i];
688 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
689 pf->stat_offsets_loaded,
690 &osd->priority_xoff_rx[i],
691 &nsd->priority_xoff_rx[i]);
693 /* No new PFC xoff rx */
694 if (!(nsd->priority_xoff_rx[i] - prio_xoff))
696 /* Get the TC for given priority */
697 tc = dcb_cfg->etscfg.prioritytable[i];
701 /* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
702 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
703 struct i40e_vsi *vsi = pf->vsi[v];
708 for (i = 0; i < vsi->num_queue_pairs; i++) {
709 struct i40e_ring *ring = vsi->tx_rings[i];
713 clear_bit(__I40E_HANG_CHECK_ARMED,
720 * i40e_update_stats - Update the board statistics counters.
721 * @vsi: the VSI to be updated
723 * There are a few instances where we store the same stat in a
724 * couple of different structs. This is partly because we have
725 * the netdev stats that need to be filled out, which is slightly
726 * different from the "eth_stats" defined by the chip and used in
727 * VF communications. We sort it all out here in a central place.
729 void i40e_update_stats(struct i40e_vsi *vsi)
731 struct i40e_pf *pf = vsi->back;
732 struct i40e_hw *hw = &pf->hw;
733 struct rtnl_link_stats64 *ons;
734 struct rtnl_link_stats64 *ns; /* netdev stats */
735 struct i40e_eth_stats *oes;
736 struct i40e_eth_stats *es; /* device's eth stats */
737 u32 tx_restart, tx_busy;
744 if (test_bit(__I40E_DOWN, &vsi->state) ||
745 test_bit(__I40E_CONFIG_BUSY, &pf->state))
748 ns = i40e_get_vsi_stats_struct(vsi);
749 ons = &vsi->net_stats_offsets;
750 es = &vsi->eth_stats;
751 oes = &vsi->eth_stats_offsets;
753 /* Gather up the netdev and vsi stats that the driver collects
754 * on the fly during packet processing
758 tx_restart = tx_busy = 0;
762 for (q = 0; q < vsi->num_queue_pairs; q++) {
768 p = ACCESS_ONCE(vsi->tx_rings[q]);
771 start = u64_stats_fetch_begin_bh(&p->syncp);
772 packets = p->stats.packets;
773 bytes = p->stats.bytes;
774 } while (u64_stats_fetch_retry_bh(&p->syncp, start));
777 tx_restart += p->tx_stats.restart_queue;
778 tx_busy += p->tx_stats.tx_busy;
780 /* Rx queue is part of the same block as Tx queue */
783 start = u64_stats_fetch_begin_bh(&p->syncp);
784 packets = p->stats.packets;
785 bytes = p->stats.bytes;
786 } while (u64_stats_fetch_retry_bh(&p->syncp, start));
789 rx_buf += p->rx_stats.alloc_rx_buff_failed;
790 rx_page += p->rx_stats.alloc_rx_page_failed;
793 vsi->tx_restart = tx_restart;
794 vsi->tx_busy = tx_busy;
795 vsi->rx_page_failed = rx_page;
796 vsi->rx_buf_failed = rx_buf;
798 ns->rx_packets = rx_p;
800 ns->tx_packets = tx_p;
803 i40e_update_eth_stats(vsi);
804 /* update netdev stats from eth stats */
805 ons->rx_errors = oes->rx_errors;
806 ns->rx_errors = es->rx_errors;
807 ons->tx_errors = oes->tx_errors;
808 ns->tx_errors = es->tx_errors;
809 ons->multicast = oes->rx_multicast;
810 ns->multicast = es->rx_multicast;
811 ons->tx_dropped = oes->tx_discards;
812 ns->tx_dropped = es->tx_discards;
814 /* Get the port data only if this is the main PF VSI */
815 if (vsi == pf->vsi[pf->lan_vsi]) {
816 struct i40e_hw_port_stats *nsd = &pf->stats;
817 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
819 i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
820 I40E_GLPRT_GORCL(hw->port),
821 pf->stat_offsets_loaded,
822 &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
823 i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
824 I40E_GLPRT_GOTCL(hw->port),
825 pf->stat_offsets_loaded,
826 &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
827 i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
828 pf->stat_offsets_loaded,
829 &osd->eth.rx_discards,
830 &nsd->eth.rx_discards);
831 i40e_stat_update32(hw, I40E_GLPRT_TDPC(hw->port),
832 pf->stat_offsets_loaded,
833 &osd->eth.tx_discards,
834 &nsd->eth.tx_discards);
835 i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
836 I40E_GLPRT_MPRCL(hw->port),
837 pf->stat_offsets_loaded,
838 &osd->eth.rx_multicast,
839 &nsd->eth.rx_multicast);
841 i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
842 pf->stat_offsets_loaded,
843 &osd->tx_dropped_link_down,
844 &nsd->tx_dropped_link_down);
846 i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
847 pf->stat_offsets_loaded,
848 &osd->crc_errors, &nsd->crc_errors);
849 ns->rx_crc_errors = nsd->crc_errors;
851 i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
852 pf->stat_offsets_loaded,
853 &osd->illegal_bytes, &nsd->illegal_bytes);
854 ns->rx_errors = nsd->crc_errors
855 + nsd->illegal_bytes;
857 i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
858 pf->stat_offsets_loaded,
859 &osd->mac_local_faults,
860 &nsd->mac_local_faults);
861 i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
862 pf->stat_offsets_loaded,
863 &osd->mac_remote_faults,
864 &nsd->mac_remote_faults);
866 i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
867 pf->stat_offsets_loaded,
868 &osd->rx_length_errors,
869 &nsd->rx_length_errors);
870 ns->rx_length_errors = nsd->rx_length_errors;
872 i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
873 pf->stat_offsets_loaded,
874 &osd->link_xon_rx, &nsd->link_xon_rx);
875 i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
876 pf->stat_offsets_loaded,
877 &osd->link_xon_tx, &nsd->link_xon_tx);
878 i40e_update_prio_xoff_rx(pf); /* handles I40E_GLPRT_LXOFFRXC */
879 i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
880 pf->stat_offsets_loaded,
881 &osd->link_xoff_tx, &nsd->link_xoff_tx);
883 for (i = 0; i < 8; i++) {
884 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
885 pf->stat_offsets_loaded,
886 &osd->priority_xon_rx[i],
887 &nsd->priority_xon_rx[i]);
888 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
889 pf->stat_offsets_loaded,
890 &osd->priority_xon_tx[i],
891 &nsd->priority_xon_tx[i]);
892 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
893 pf->stat_offsets_loaded,
894 &osd->priority_xoff_tx[i],
895 &nsd->priority_xoff_tx[i]);
896 i40e_stat_update32(hw,
897 I40E_GLPRT_RXON2OFFCNT(hw->port, i),
898 pf->stat_offsets_loaded,
899 &osd->priority_xon_2_xoff[i],
900 &nsd->priority_xon_2_xoff[i]);
903 i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
904 I40E_GLPRT_PRC64L(hw->port),
905 pf->stat_offsets_loaded,
906 &osd->rx_size_64, &nsd->rx_size_64);
907 i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
908 I40E_GLPRT_PRC127L(hw->port),
909 pf->stat_offsets_loaded,
910 &osd->rx_size_127, &nsd->rx_size_127);
911 i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
912 I40E_GLPRT_PRC255L(hw->port),
913 pf->stat_offsets_loaded,
914 &osd->rx_size_255, &nsd->rx_size_255);
915 i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
916 I40E_GLPRT_PRC511L(hw->port),
917 pf->stat_offsets_loaded,
918 &osd->rx_size_511, &nsd->rx_size_511);
919 i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
920 I40E_GLPRT_PRC1023L(hw->port),
921 pf->stat_offsets_loaded,
922 &osd->rx_size_1023, &nsd->rx_size_1023);
923 i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
924 I40E_GLPRT_PRC1522L(hw->port),
925 pf->stat_offsets_loaded,
926 &osd->rx_size_1522, &nsd->rx_size_1522);
927 i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
928 I40E_GLPRT_PRC9522L(hw->port),
929 pf->stat_offsets_loaded,
930 &osd->rx_size_big, &nsd->rx_size_big);
932 i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
933 I40E_GLPRT_PTC64L(hw->port),
934 pf->stat_offsets_loaded,
935 &osd->tx_size_64, &nsd->tx_size_64);
936 i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
937 I40E_GLPRT_PTC127L(hw->port),
938 pf->stat_offsets_loaded,
939 &osd->tx_size_127, &nsd->tx_size_127);
940 i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
941 I40E_GLPRT_PTC255L(hw->port),
942 pf->stat_offsets_loaded,
943 &osd->tx_size_255, &nsd->tx_size_255);
944 i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
945 I40E_GLPRT_PTC511L(hw->port),
946 pf->stat_offsets_loaded,
947 &osd->tx_size_511, &nsd->tx_size_511);
948 i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
949 I40E_GLPRT_PTC1023L(hw->port),
950 pf->stat_offsets_loaded,
951 &osd->tx_size_1023, &nsd->tx_size_1023);
952 i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
953 I40E_GLPRT_PTC1522L(hw->port),
954 pf->stat_offsets_loaded,
955 &osd->tx_size_1522, &nsd->tx_size_1522);
956 i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
957 I40E_GLPRT_PTC9522L(hw->port),
958 pf->stat_offsets_loaded,
959 &osd->tx_size_big, &nsd->tx_size_big);
961 i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
962 pf->stat_offsets_loaded,
963 &osd->rx_undersize, &nsd->rx_undersize);
964 i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
965 pf->stat_offsets_loaded,
966 &osd->rx_fragments, &nsd->rx_fragments);
967 i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
968 pf->stat_offsets_loaded,
969 &osd->rx_oversize, &nsd->rx_oversize);
970 i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
971 pf->stat_offsets_loaded,
972 &osd->rx_jabber, &nsd->rx_jabber);
975 pf->stat_offsets_loaded = true;
979 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
980 * @vsi: the VSI to be searched
981 * @macaddr: the MAC address
983 * @is_vf: make sure its a vf filter, else doesn't matter
984 * @is_netdev: make sure its a netdev filter, else doesn't matter
986 * Returns ptr to the filter object or NULL
988 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
989 u8 *macaddr, s16 vlan,
990 bool is_vf, bool is_netdev)
992 struct i40e_mac_filter *f;
994 if (!vsi || !macaddr)
997 list_for_each_entry(f, &vsi->mac_filter_list, list) {
998 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1000 (!is_vf || f->is_vf) &&
1001 (!is_netdev || f->is_netdev))
1008 * i40e_find_mac - Find a mac addr in the macvlan filters list
1009 * @vsi: the VSI to be searched
1010 * @macaddr: the MAC address we are searching for
1011 * @is_vf: make sure its a vf filter, else doesn't matter
1012 * @is_netdev: make sure its a netdev filter, else doesn't matter
1014 * Returns the first filter with the provided MAC address or NULL if
1015 * MAC address was not found
1017 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1018 bool is_vf, bool is_netdev)
1020 struct i40e_mac_filter *f;
1022 if (!vsi || !macaddr)
1025 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1026 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1027 (!is_vf || f->is_vf) &&
1028 (!is_netdev || f->is_netdev))
1035 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1036 * @vsi: the VSI to be searched
1038 * Returns true if VSI is in vlan mode or false otherwise
1040 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1042 struct i40e_mac_filter *f;
1044 /* Only -1 for all the filters denotes not in vlan mode
1045 * so we have to go through all the list in order to make sure
1047 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1056 * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1057 * @vsi: the VSI to be searched
1058 * @macaddr: the mac address to be filtered
1059 * @is_vf: true if it is a vf
1060 * @is_netdev: true if it is a netdev
1062 * Goes through all the macvlan filters and adds a
1063 * macvlan filter for each unique vlan that already exists
1065 * Returns first filter found on success, else NULL
1067 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1068 bool is_vf, bool is_netdev)
1070 struct i40e_mac_filter *f;
1072 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1073 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1074 is_vf, is_netdev)) {
1075 if (!i40e_add_filter(vsi, macaddr, f->vlan,
1081 return list_first_entry_or_null(&vsi->mac_filter_list,
1082 struct i40e_mac_filter, list);
1086 * i40e_add_filter - Add a mac/vlan filter to the VSI
1087 * @vsi: the VSI to be searched
1088 * @macaddr: the MAC address
1090 * @is_vf: make sure its a vf filter, else doesn't matter
1091 * @is_netdev: make sure its a netdev filter, else doesn't matter
1093 * Returns ptr to the filter object or NULL when no memory available.
1095 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1096 u8 *macaddr, s16 vlan,
1097 bool is_vf, bool is_netdev)
1099 struct i40e_mac_filter *f;
1101 if (!vsi || !macaddr)
1104 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1106 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1108 goto add_filter_out;
1110 memcpy(f->macaddr, macaddr, ETH_ALEN);
1114 INIT_LIST_HEAD(&f->list);
1115 list_add(&f->list, &vsi->mac_filter_list);
1118 /* increment counter and add a new flag if needed */
1124 } else if (is_netdev) {
1125 if (!f->is_netdev) {
1126 f->is_netdev = true;
1133 /* changed tells sync_filters_subtask to
1134 * push the filter down to the firmware
1137 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1138 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1146 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1147 * @vsi: the VSI to be searched
1148 * @macaddr: the MAC address
1150 * @is_vf: make sure it's a vf filter, else doesn't matter
1151 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1153 void i40e_del_filter(struct i40e_vsi *vsi,
1154 u8 *macaddr, s16 vlan,
1155 bool is_vf, bool is_netdev)
1157 struct i40e_mac_filter *f;
1159 if (!vsi || !macaddr)
1162 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1163 if (!f || f->counter == 0)
1171 } else if (is_netdev) {
1173 f->is_netdev = false;
1177 /* make sure we don't remove a filter in use by vf or netdev */
1179 min_f += (f->is_vf ? 1 : 0);
1180 min_f += (f->is_netdev ? 1 : 0);
1182 if (f->counter > min_f)
1186 /* counter == 0 tells sync_filters_subtask to
1187 * remove the filter from the firmware's list
1189 if (f->counter == 0) {
1191 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1192 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1197 * i40e_set_mac - NDO callback to set mac address
1198 * @netdev: network interface device structure
1199 * @p: pointer to an address structure
1201 * Returns 0 on success, negative on failure
1203 static int i40e_set_mac(struct net_device *netdev, void *p)
1205 struct i40e_netdev_priv *np = netdev_priv(netdev);
1206 struct i40e_vsi *vsi = np->vsi;
1207 struct sockaddr *addr = p;
1208 struct i40e_mac_filter *f;
1210 if (!is_valid_ether_addr(addr->sa_data))
1211 return -EADDRNOTAVAIL;
1213 netdev_info(netdev, "set mac address=%pM\n", addr->sa_data);
1215 if (ether_addr_equal(netdev->dev_addr, addr->sa_data))
1218 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1219 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1220 return -EADDRNOTAVAIL;
1222 if (vsi->type == I40E_VSI_MAIN) {
1224 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1225 I40E_AQC_WRITE_TYPE_LAA_ONLY,
1226 addr->sa_data, NULL);
1229 "Addr change for Main VSI failed: %d\n",
1231 return -EADDRNOTAVAIL;
1234 memcpy(vsi->back->hw.mac.addr, addr->sa_data, netdev->addr_len);
1237 /* In order to be sure to not drop any packets, add the new address
1238 * then delete the old one.
1240 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY, false, false);
1244 i40e_sync_vsi_filters(vsi);
1245 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY, false, false);
1246 i40e_sync_vsi_filters(vsi);
1248 memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1254 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1255 * @vsi: the VSI being setup
1256 * @ctxt: VSI context structure
1257 * @enabled_tc: Enabled TCs bitmap
1258 * @is_add: True if called before Add VSI
1260 * Setup VSI queue mapping for enabled traffic classes.
1262 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1263 struct i40e_vsi_context *ctxt,
1267 struct i40e_pf *pf = vsi->back;
1276 sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1279 if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1280 /* Find numtc from enabled TC bitmap */
1281 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1282 if (enabled_tc & (1 << i)) /* TC is enabled */
1286 dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1290 /* At least TC0 is enabled in case of non-DCB case */
1294 vsi->tc_config.numtc = numtc;
1295 vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1297 /* Setup queue offset/count for all TCs for given VSI */
1298 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1299 /* See if the given TC is enabled for the given VSI */
1300 if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1303 vsi->tc_config.tc_info[i].qoffset = offset;
1304 switch (vsi->type) {
1307 qcount = pf->rss_size;
1309 qcount = pf->num_tc_qps;
1310 vsi->tc_config.tc_info[i].qcount = qcount;
1313 case I40E_VSI_SRIOV:
1314 case I40E_VSI_VMDQ2:
1316 qcount = vsi->alloc_queue_pairs;
1317 vsi->tc_config.tc_info[i].qcount = qcount;
1322 /* find the power-of-2 of the number of queue pairs */
1323 num_qps = vsi->tc_config.tc_info[i].qcount;
1326 ((1 << pow) < vsi->tc_config.tc_info[i].qcount)) {
1331 vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1333 (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1334 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1336 offset += vsi->tc_config.tc_info[i].qcount;
1338 /* TC is not enabled so set the offset to
1339 * default queue and allocate one queue
1342 vsi->tc_config.tc_info[i].qoffset = 0;
1343 vsi->tc_config.tc_info[i].qcount = 1;
1344 vsi->tc_config.tc_info[i].netdev_tc = 0;
1348 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1351 /* Set actual Tx/Rx queue pairs */
1352 vsi->num_queue_pairs = offset;
1354 /* Scheduler section valid can only be set for ADD VSI */
1356 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1358 ctxt->info.up_enable_bits = enabled_tc;
1360 if (vsi->type == I40E_VSI_SRIOV) {
1361 ctxt->info.mapping_flags |=
1362 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1363 for (i = 0; i < vsi->num_queue_pairs; i++)
1364 ctxt->info.queue_mapping[i] =
1365 cpu_to_le16(vsi->base_queue + i);
1367 ctxt->info.mapping_flags |=
1368 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1369 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1371 ctxt->info.valid_sections |= cpu_to_le16(sections);
1375 * i40e_set_rx_mode - NDO callback to set the netdev filters
1376 * @netdev: network interface device structure
1378 static void i40e_set_rx_mode(struct net_device *netdev)
1380 struct i40e_netdev_priv *np = netdev_priv(netdev);
1381 struct i40e_mac_filter *f, *ftmp;
1382 struct i40e_vsi *vsi = np->vsi;
1383 struct netdev_hw_addr *uca;
1384 struct netdev_hw_addr *mca;
1385 struct netdev_hw_addr *ha;
1387 /* add addr if not already in the filter list */
1388 netdev_for_each_uc_addr(uca, netdev) {
1389 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1390 if (i40e_is_vsi_in_vlan(vsi))
1391 i40e_put_mac_in_vlan(vsi, uca->addr,
1394 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1399 netdev_for_each_mc_addr(mca, netdev) {
1400 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1401 if (i40e_is_vsi_in_vlan(vsi))
1402 i40e_put_mac_in_vlan(vsi, mca->addr,
1405 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1410 /* remove filter if not in netdev list */
1411 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1417 if (is_multicast_ether_addr(f->macaddr)) {
1418 netdev_for_each_mc_addr(mca, netdev) {
1419 if (ether_addr_equal(mca->addr, f->macaddr)) {
1425 netdev_for_each_uc_addr(uca, netdev) {
1426 if (ether_addr_equal(uca->addr, f->macaddr)) {
1432 for_each_dev_addr(netdev, ha) {
1433 if (ether_addr_equal(ha->addr, f->macaddr)) {
1441 vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1444 /* check for other flag changes */
1445 if (vsi->current_netdev_flags != vsi->netdev->flags) {
1446 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1447 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1452 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1453 * @vsi: ptr to the VSI
1455 * Push any outstanding VSI filter changes through the AdminQ.
1457 * Returns 0 or error value
1459 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1461 struct i40e_mac_filter *f, *ftmp;
1462 bool promisc_forced_on = false;
1463 bool add_happened = false;
1464 int filter_list_len = 0;
1465 u32 changed_flags = 0;
1466 i40e_status aq_ret = 0;
1472 /* empty array typed pointers, kcalloc later */
1473 struct i40e_aqc_add_macvlan_element_data *add_list;
1474 struct i40e_aqc_remove_macvlan_element_data *del_list;
1476 while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1477 usleep_range(1000, 2000);
1481 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1482 vsi->current_netdev_flags = vsi->netdev->flags;
1485 if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1486 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1488 filter_list_len = pf->hw.aq.asq_buf_size /
1489 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1490 del_list = kcalloc(filter_list_len,
1491 sizeof(struct i40e_aqc_remove_macvlan_element_data),
1496 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1500 if (f->counter != 0)
1505 /* add to delete list */
1506 memcpy(del_list[num_del].mac_addr,
1507 f->macaddr, ETH_ALEN);
1508 del_list[num_del].vlan_tag =
1509 cpu_to_le16((u16)(f->vlan ==
1510 I40E_VLAN_ANY ? 0 : f->vlan));
1512 cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1513 del_list[num_del].flags = cmd_flags;
1516 /* unlink from filter list */
1520 /* flush a full buffer */
1521 if (num_del == filter_list_len) {
1522 aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1523 vsi->seid, del_list, num_del,
1526 memset(del_list, 0, sizeof(*del_list));
1529 dev_info(&pf->pdev->dev,
1530 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1532 pf->hw.aq.asq_last_status);
1536 aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1537 del_list, num_del, NULL);
1541 dev_info(&pf->pdev->dev,
1542 "ignoring delete macvlan error, err %d, aq_err %d\n",
1543 aq_ret, pf->hw.aq.asq_last_status);
1549 /* do all the adds now */
1550 filter_list_len = pf->hw.aq.asq_buf_size /
1551 sizeof(struct i40e_aqc_add_macvlan_element_data),
1552 add_list = kcalloc(filter_list_len,
1553 sizeof(struct i40e_aqc_add_macvlan_element_data),
1558 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1562 if (f->counter == 0)
1565 add_happened = true;
1568 /* add to add array */
1569 memcpy(add_list[num_add].mac_addr,
1570 f->macaddr, ETH_ALEN);
1571 add_list[num_add].vlan_tag =
1573 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1574 add_list[num_add].queue_number = 0;
1576 cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1577 add_list[num_add].flags = cpu_to_le16(cmd_flags);
1580 /* flush a full buffer */
1581 if (num_add == filter_list_len) {
1582 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1589 memset(add_list, 0, sizeof(*add_list));
1593 aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1594 add_list, num_add, NULL);
1600 if (add_happened && (!aq_ret)) {
1602 } else if (add_happened && (aq_ret)) {
1603 dev_info(&pf->pdev->dev,
1604 "add filter failed, err %d, aq_err %d\n",
1605 aq_ret, pf->hw.aq.asq_last_status);
1606 if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1607 !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1609 promisc_forced_on = true;
1610 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1612 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1617 /* check for changes in promiscuous modes */
1618 if (changed_flags & IFF_ALLMULTI) {
1619 bool cur_multipromisc;
1620 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1621 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1626 dev_info(&pf->pdev->dev,
1627 "set multi promisc failed, err %d, aq_err %d\n",
1628 aq_ret, pf->hw.aq.asq_last_status);
1630 if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1632 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1633 test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1635 aq_ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1639 dev_info(&pf->pdev->dev,
1640 "set uni promisc failed, err %d, aq_err %d\n",
1641 aq_ret, pf->hw.aq.asq_last_status);
1642 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
1646 dev_info(&pf->pdev->dev,
1647 "set brdcast promisc failed, err %d, aq_err %d\n",
1648 aq_ret, pf->hw.aq.asq_last_status);
1651 clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1656 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1657 * @pf: board private structure
1659 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1663 if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1665 pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1667 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
1669 (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1670 i40e_sync_vsi_filters(pf->vsi[v]);
1675 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1676 * @netdev: network interface device structure
1677 * @new_mtu: new value for maximum frame size
1679 * Returns 0 on success, negative on failure
1681 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1683 struct i40e_netdev_priv *np = netdev_priv(netdev);
1684 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
1685 struct i40e_vsi *vsi = np->vsi;
1687 /* MTU < 68 is an error and causes problems on some kernels */
1688 if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1691 netdev_info(netdev, "changing MTU from %d to %d\n",
1692 netdev->mtu, new_mtu);
1693 netdev->mtu = new_mtu;
1694 if (netif_running(netdev))
1695 i40e_vsi_reinit_locked(vsi);
1701 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1702 * @vsi: the vsi being adjusted
1704 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1706 struct i40e_vsi_context ctxt;
1709 if ((vsi->info.valid_sections &
1710 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1711 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1712 return; /* already enabled */
1714 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1715 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1716 I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1718 ctxt.seid = vsi->seid;
1719 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1720 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1722 dev_info(&vsi->back->pdev->dev,
1723 "%s: update vsi failed, aq_err=%d\n",
1724 __func__, vsi->back->hw.aq.asq_last_status);
1729 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1730 * @vsi: the vsi being adjusted
1732 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
1734 struct i40e_vsi_context ctxt;
1737 if ((vsi->info.valid_sections &
1738 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1739 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
1740 I40E_AQ_VSI_PVLAN_EMOD_MASK))
1741 return; /* already disabled */
1743 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1744 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1745 I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
1747 ctxt.seid = vsi->seid;
1748 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1749 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1751 dev_info(&vsi->back->pdev->dev,
1752 "%s: update vsi failed, aq_err=%d\n",
1753 __func__, vsi->back->hw.aq.asq_last_status);
1758 * i40e_vlan_rx_register - Setup or shutdown vlan offload
1759 * @netdev: network interface to be adjusted
1760 * @features: netdev features to test if VLAN offload is enabled or not
1762 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
1764 struct i40e_netdev_priv *np = netdev_priv(netdev);
1765 struct i40e_vsi *vsi = np->vsi;
1767 if (features & NETIF_F_HW_VLAN_CTAG_RX)
1768 i40e_vlan_stripping_enable(vsi);
1770 i40e_vlan_stripping_disable(vsi);
1774 * i40e_vsi_add_vlan - Add vsi membership for given vlan
1775 * @vsi: the vsi being configured
1776 * @vid: vlan id to be added (0 = untagged only , -1 = any)
1778 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
1780 struct i40e_mac_filter *f, *add_f;
1781 bool is_netdev, is_vf;
1783 is_vf = (vsi->type == I40E_VSI_SRIOV);
1784 is_netdev = !!(vsi->netdev);
1787 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
1790 dev_info(&vsi->back->pdev->dev,
1791 "Could not add vlan filter %d for %pM\n",
1792 vid, vsi->netdev->dev_addr);
1797 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1798 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1800 dev_info(&vsi->back->pdev->dev,
1801 "Could not add vlan filter %d for %pM\n",
1807 /* Now if we add a vlan tag, make sure to check if it is the first
1808 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
1809 * with 0, so we now accept untagged and specified tagged traffic
1810 * (and not any taged and untagged)
1813 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
1815 is_vf, is_netdev)) {
1816 i40e_del_filter(vsi, vsi->netdev->dev_addr,
1817 I40E_VLAN_ANY, is_vf, is_netdev);
1818 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
1821 dev_info(&vsi->back->pdev->dev,
1822 "Could not add filter 0 for %pM\n",
1823 vsi->netdev->dev_addr);
1828 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1829 if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1830 is_vf, is_netdev)) {
1831 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1833 add_f = i40e_add_filter(vsi, f->macaddr,
1834 0, is_vf, is_netdev);
1836 dev_info(&vsi->back->pdev->dev,
1837 "Could not add filter 0 for %pM\n",
1845 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1846 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1849 return i40e_sync_vsi_filters(vsi);
1853 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
1854 * @vsi: the vsi being configured
1855 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
1857 * Return: 0 on success or negative otherwise
1859 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
1861 struct net_device *netdev = vsi->netdev;
1862 struct i40e_mac_filter *f, *add_f;
1863 bool is_vf, is_netdev;
1864 int filter_count = 0;
1866 is_vf = (vsi->type == I40E_VSI_SRIOV);
1867 is_netdev = !!(netdev);
1870 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
1872 list_for_each_entry(f, &vsi->mac_filter_list, list)
1873 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1875 /* go through all the filters for this VSI and if there is only
1876 * vid == 0 it means there are no other filters, so vid 0 must
1877 * be replaced with -1. This signifies that we should from now
1878 * on accept any traffic (with any tag present, or untagged)
1880 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1883 ether_addr_equal(netdev->dev_addr, f->macaddr))
1891 if (!filter_count && is_netdev) {
1892 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
1893 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1896 dev_info(&vsi->back->pdev->dev,
1897 "Could not add filter %d for %pM\n",
1898 I40E_VLAN_ANY, netdev->dev_addr);
1903 if (!filter_count) {
1904 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1905 i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
1906 add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1909 dev_info(&vsi->back->pdev->dev,
1910 "Could not add filter %d for %pM\n",
1911 I40E_VLAN_ANY, f->macaddr);
1917 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1918 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1921 return i40e_sync_vsi_filters(vsi);
1925 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
1926 * @netdev: network interface to be adjusted
1927 * @vid: vlan id to be added
1929 * net_device_ops implementation for adding vlan ids
1931 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
1932 __always_unused __be16 proto, u16 vid)
1934 struct i40e_netdev_priv *np = netdev_priv(netdev);
1935 struct i40e_vsi *vsi = np->vsi;
1941 netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
1943 /* If the network stack called us with vid = 0, we should
1944 * indicate to i40e_vsi_add_vlan() that we want to receive
1945 * any traffic (i.e. with any vlan tag, or untagged)
1947 ret = i40e_vsi_add_vlan(vsi, vid ? vid : I40E_VLAN_ANY);
1949 if (!ret && (vid < VLAN_N_VID))
1950 set_bit(vid, vsi->active_vlans);
1956 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
1957 * @netdev: network interface to be adjusted
1958 * @vid: vlan id to be removed
1960 * net_device_ops implementation for adding vlan ids
1962 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
1963 __always_unused __be16 proto, u16 vid)
1965 struct i40e_netdev_priv *np = netdev_priv(netdev);
1966 struct i40e_vsi *vsi = np->vsi;
1968 netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
1970 /* return code is ignored as there is nothing a user
1971 * can do about failure to remove and a log message was
1972 * already printed from the other function
1974 i40e_vsi_kill_vlan(vsi, vid);
1976 clear_bit(vid, vsi->active_vlans);
1982 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
1983 * @vsi: the vsi being brought back up
1985 static void i40e_restore_vlan(struct i40e_vsi *vsi)
1992 i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
1994 for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
1995 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2000 * i40e_vsi_add_pvid - Add pvid for the VSI
2001 * @vsi: the vsi being adjusted
2002 * @vid: the vlan id to set as a PVID
2004 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2006 struct i40e_vsi_context ctxt;
2009 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2010 vsi->info.pvid = cpu_to_le16(vid);
2011 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2012 I40E_AQ_VSI_PVLAN_INSERT_PVID |
2013 I40E_AQ_VSI_PVLAN_EMOD_STR;
2015 ctxt.seid = vsi->seid;
2016 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
2017 aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2019 dev_info(&vsi->back->pdev->dev,
2020 "%s: update vsi failed, aq_err=%d\n",
2021 __func__, vsi->back->hw.aq.asq_last_status);
2029 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2030 * @vsi: the vsi being adjusted
2032 * Just use the vlan_rx_register() service to put it back to normal
2034 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2036 i40e_vlan_stripping_disable(vsi);
2042 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2043 * @vsi: ptr to the VSI
2045 * If this function returns with an error, then it's possible one or
2046 * more of the rings is populated (while the rest are not). It is the
2047 * callers duty to clean those orphaned rings.
2049 * Return 0 on success, negative on failure
2051 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2055 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2056 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2062 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2063 * @vsi: ptr to the VSI
2065 * Free VSI's transmit software resources
2067 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2071 for (i = 0; i < vsi->num_queue_pairs; i++)
2072 if (vsi->tx_rings[i]->desc)
2073 i40e_free_tx_resources(vsi->tx_rings[i]);
2077 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2078 * @vsi: ptr to the VSI
2080 * If this function returns with an error, then it's possible one or
2081 * more of the rings is populated (while the rest are not). It is the
2082 * callers duty to clean those orphaned rings.
2084 * Return 0 on success, negative on failure
2086 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2090 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2091 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2096 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2097 * @vsi: ptr to the VSI
2099 * Free all receive software resources
2101 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2105 for (i = 0; i < vsi->num_queue_pairs; i++)
2106 if (vsi->rx_rings[i]->desc)
2107 i40e_free_rx_resources(vsi->rx_rings[i]);
2111 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2112 * @ring: The Tx ring to configure
2114 * Configure the Tx descriptor ring in the HMC context.
2116 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2118 struct i40e_vsi *vsi = ring->vsi;
2119 u16 pf_q = vsi->base_queue + ring->queue_index;
2120 struct i40e_hw *hw = &vsi->back->hw;
2121 struct i40e_hmc_obj_txq tx_ctx;
2122 i40e_status err = 0;
2125 /* some ATR related tx ring init */
2126 if (vsi->back->flags & I40E_FLAG_FDIR_ATR_ENABLED) {
2127 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2128 ring->atr_count = 0;
2130 ring->atr_sample_rate = 0;
2133 /* initialize XPS */
2134 if (ring->q_vector && ring->netdev &&
2135 !test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2136 netif_set_xps_queue(ring->netdev,
2137 &ring->q_vector->affinity_mask,
2140 /* clear the context structure first */
2141 memset(&tx_ctx, 0, sizeof(tx_ctx));
2143 tx_ctx.new_context = 1;
2144 tx_ctx.base = (ring->dma / 128);
2145 tx_ctx.qlen = ring->count;
2146 tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FDIR_ENABLED |
2147 I40E_FLAG_FDIR_ATR_ENABLED));
2149 /* As part of VSI creation/update, FW allocates certain
2150 * Tx arbitration queue sets for each TC enabled for
2151 * the VSI. The FW returns the handles to these queue
2152 * sets as part of the response buffer to Add VSI,
2153 * Update VSI, etc. AQ commands. It is expected that
2154 * these queue set handles be associated with the Tx
2155 * queues by the driver as part of the TX queue context
2156 * initialization. This has to be done regardless of
2157 * DCB as by default everything is mapped to TC0.
2159 tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2160 tx_ctx.rdylist_act = 0;
2162 /* clear the context in the HMC */
2163 err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2165 dev_info(&vsi->back->pdev->dev,
2166 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2167 ring->queue_index, pf_q, err);
2171 /* set the context in the HMC */
2172 err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2174 dev_info(&vsi->back->pdev->dev,
2175 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2176 ring->queue_index, pf_q, err);
2180 /* Now associate this queue with this PCI function */
2181 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2182 qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2183 I40E_QTX_CTL_PF_INDX_MASK);
2184 wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2187 clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2189 /* cache tail off for easier writes later */
2190 ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2196 * i40e_configure_rx_ring - Configure a receive ring context
2197 * @ring: The Rx ring to configure
2199 * Configure the Rx descriptor ring in the HMC context.
2201 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2203 struct i40e_vsi *vsi = ring->vsi;
2204 u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2205 u16 pf_q = vsi->base_queue + ring->queue_index;
2206 struct i40e_hw *hw = &vsi->back->hw;
2207 struct i40e_hmc_obj_rxq rx_ctx;
2208 i40e_status err = 0;
2212 /* clear the context structure first */
2213 memset(&rx_ctx, 0, sizeof(rx_ctx));
2215 ring->rx_buf_len = vsi->rx_buf_len;
2216 ring->rx_hdr_len = vsi->rx_hdr_len;
2218 rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2219 rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2221 rx_ctx.base = (ring->dma / 128);
2222 rx_ctx.qlen = ring->count;
2224 if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2225 set_ring_16byte_desc_enabled(ring);
2231 rx_ctx.dtype = vsi->dtype;
2233 set_ring_ps_enabled(ring);
2234 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2 |
2236 I40E_RX_SPLIT_TCP_UDP |
2239 rx_ctx.hsplit_0 = 0;
2242 rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2243 (chain_len * ring->rx_buf_len));
2244 rx_ctx.tphrdesc_ena = 1;
2245 rx_ctx.tphwdesc_ena = 1;
2246 rx_ctx.tphdata_ena = 1;
2247 rx_ctx.tphhead_ena = 1;
2248 if (hw->revision_id == 0)
2249 rx_ctx.lrxqthresh = 0;
2251 rx_ctx.lrxqthresh = 2;
2252 rx_ctx.crcstrip = 1;
2256 /* clear the context in the HMC */
2257 err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2259 dev_info(&vsi->back->pdev->dev,
2260 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2261 ring->queue_index, pf_q, err);
2265 /* set the context in the HMC */
2266 err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2268 dev_info(&vsi->back->pdev->dev,
2269 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2270 ring->queue_index, pf_q, err);
2274 /* cache tail for quicker writes, and clear the reg before use */
2275 ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2276 writel(0, ring->tail);
2278 i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2284 * i40e_vsi_configure_tx - Configure the VSI for Tx
2285 * @vsi: VSI structure describing this set of rings and resources
2287 * Configure the Tx VSI for operation.
2289 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2294 for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2295 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2301 * i40e_vsi_configure_rx - Configure the VSI for Rx
2302 * @vsi: the VSI being configured
2304 * Configure the Rx VSI for operation.
2306 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2311 if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2312 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2313 + ETH_FCS_LEN + VLAN_HLEN;
2315 vsi->max_frame = I40E_RXBUFFER_2048;
2317 /* figure out correct receive buffer length */
2318 switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2319 I40E_FLAG_RX_PS_ENABLED)) {
2320 case I40E_FLAG_RX_1BUF_ENABLED:
2321 vsi->rx_hdr_len = 0;
2322 vsi->rx_buf_len = vsi->max_frame;
2323 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2325 case I40E_FLAG_RX_PS_ENABLED:
2326 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2327 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2328 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2331 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2332 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2333 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2337 /* round up for the chip's needs */
2338 vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2339 (1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2340 vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2341 (1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2343 /* set up individual rings */
2344 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2345 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2351 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2352 * @vsi: ptr to the VSI
2354 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2356 u16 qoffset, qcount;
2359 if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED))
2362 for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2363 if (!(vsi->tc_config.enabled_tc & (1 << n)))
2366 qoffset = vsi->tc_config.tc_info[n].qoffset;
2367 qcount = vsi->tc_config.tc_info[n].qcount;
2368 for (i = qoffset; i < (qoffset + qcount); i++) {
2369 struct i40e_ring *rx_ring = vsi->rx_rings[i];
2370 struct i40e_ring *tx_ring = vsi->tx_rings[i];
2371 rx_ring->dcb_tc = n;
2372 tx_ring->dcb_tc = n;
2378 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2379 * @vsi: ptr to the VSI
2381 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2384 i40e_set_rx_mode(vsi->netdev);
2388 * i40e_vsi_configure - Set up the VSI for action
2389 * @vsi: the VSI being configured
2391 static int i40e_vsi_configure(struct i40e_vsi *vsi)
2395 i40e_set_vsi_rx_mode(vsi);
2396 i40e_restore_vlan(vsi);
2397 i40e_vsi_config_dcb_rings(vsi);
2398 err = i40e_vsi_configure_tx(vsi);
2400 err = i40e_vsi_configure_rx(vsi);
2406 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2407 * @vsi: the VSI being configured
2409 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2411 struct i40e_pf *pf = vsi->back;
2412 struct i40e_q_vector *q_vector;
2413 struct i40e_hw *hw = &pf->hw;
2419 /* The interrupt indexing is offset by 1 in the PFINT_ITRn
2420 * and PFINT_LNKLSTn registers, e.g.:
2421 * PFINT_ITRn[0..n-1] gets msix-1..msix-n (qpair interrupts)
2423 qp = vsi->base_queue;
2424 vector = vsi->base_vector;
2425 for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
2426 q_vector = vsi->q_vectors[i];
2427 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2428 q_vector->rx.latency_range = I40E_LOW_LATENCY;
2429 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2431 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2432 q_vector->tx.latency_range = I40E_LOW_LATENCY;
2433 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2436 /* Linked list for the queuepairs assigned to this vector */
2437 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2438 for (q = 0; q < q_vector->num_ringpairs; q++) {
2439 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2440 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2441 (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2442 (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2444 << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2446 wr32(hw, I40E_QINT_RQCTL(qp), val);
2448 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2449 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2450 (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2451 ((qp+1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2453 << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2455 /* Terminate the linked list */
2456 if (q == (q_vector->num_ringpairs - 1))
2457 val |= (I40E_QUEUE_END_OF_LIST
2458 << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2460 wr32(hw, I40E_QINT_TQCTL(qp), val);
2469 * i40e_enable_misc_int_causes - enable the non-queue interrupts
2470 * @hw: ptr to the hardware info
2472 static void i40e_enable_misc_int_causes(struct i40e_hw *hw)
2476 /* clear things first */
2477 wr32(hw, I40E_PFINT_ICR0_ENA, 0); /* disable all */
2478 rd32(hw, I40E_PFINT_ICR0); /* read to clear */
2480 val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK |
2481 I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK |
2482 I40E_PFINT_ICR0_ENA_GRST_MASK |
2483 I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2484 I40E_PFINT_ICR0_ENA_GPIO_MASK |
2485 I40E_PFINT_ICR0_ENA_STORM_DETECT_MASK |
2486 I40E_PFINT_ICR0_ENA_HMC_ERR_MASK |
2487 I40E_PFINT_ICR0_ENA_VFLR_MASK |
2488 I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2490 wr32(hw, I40E_PFINT_ICR0_ENA, val);
2492 /* SW_ITR_IDX = 0, but don't change INTENA */
2493 wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
2494 I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
2496 /* OTHER_ITR_IDX = 0 */
2497 wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2501 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2502 * @vsi: the VSI being configured
2504 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2506 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
2507 struct i40e_pf *pf = vsi->back;
2508 struct i40e_hw *hw = &pf->hw;
2511 /* set the ITR configuration */
2512 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2513 q_vector->rx.latency_range = I40E_LOW_LATENCY;
2514 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2515 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2516 q_vector->tx.latency_range = I40E_LOW_LATENCY;
2517 wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2519 i40e_enable_misc_int_causes(hw);
2521 /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2522 wr32(hw, I40E_PFINT_LNKLST0, 0);
2524 /* Associate the queue pair to the vector and enable the q int */
2525 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2526 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2527 (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2529 wr32(hw, I40E_QINT_RQCTL(0), val);
2531 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2532 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2533 (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2535 wr32(hw, I40E_QINT_TQCTL(0), val);
2540 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
2541 * @pf: board private structure
2543 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
2545 struct i40e_hw *hw = &pf->hw;
2547 wr32(hw, I40E_PFINT_DYN_CTL0,
2548 I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2553 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2554 * @pf: board private structure
2556 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2558 struct i40e_hw *hw = &pf->hw;
2561 val = I40E_PFINT_DYN_CTL0_INTENA_MASK |
2562 I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2563 (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2565 wr32(hw, I40E_PFINT_DYN_CTL0, val);
2570 * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2571 * @vsi: pointer to a vsi
2572 * @vector: enable a particular Hw Interrupt vector
2574 void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2576 struct i40e_pf *pf = vsi->back;
2577 struct i40e_hw *hw = &pf->hw;
2580 val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2581 I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2582 (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2583 wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2584 /* skip the flush */
2588 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2589 * @irq: interrupt number
2590 * @data: pointer to a q_vector
2592 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2594 struct i40e_q_vector *q_vector = data;
2596 if (!q_vector->tx.ring && !q_vector->rx.ring)
2599 napi_schedule(&q_vector->napi);
2605 * i40e_fdir_clean_rings - Interrupt Handler for FDIR rings
2606 * @irq: interrupt number
2607 * @data: pointer to a q_vector
2609 static irqreturn_t i40e_fdir_clean_rings(int irq, void *data)
2611 struct i40e_q_vector *q_vector = data;
2613 if (!q_vector->tx.ring && !q_vector->rx.ring)
2616 pr_info("fdir ring cleaning needed\n");
2622 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
2623 * @vsi: the VSI being configured
2624 * @basename: name for the vector
2626 * Allocates MSI-X vectors and requests interrupts from the kernel.
2628 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
2630 int q_vectors = vsi->num_q_vectors;
2631 struct i40e_pf *pf = vsi->back;
2632 int base = vsi->base_vector;
2637 for (vector = 0; vector < q_vectors; vector++) {
2638 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
2640 if (q_vector->tx.ring && q_vector->rx.ring) {
2641 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2642 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2644 } else if (q_vector->rx.ring) {
2645 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2646 "%s-%s-%d", basename, "rx", rx_int_idx++);
2647 } else if (q_vector->tx.ring) {
2648 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2649 "%s-%s-%d", basename, "tx", tx_int_idx++);
2651 /* skip this unused q_vector */
2654 err = request_irq(pf->msix_entries[base + vector].vector,
2660 dev_info(&pf->pdev->dev,
2661 "%s: request_irq failed, error: %d\n",
2663 goto free_queue_irqs;
2665 /* assign the mask for this irq */
2666 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2667 &q_vector->affinity_mask);
2675 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2677 free_irq(pf->msix_entries[base + vector].vector,
2678 &(vsi->q_vectors[vector]));
2684 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
2685 * @vsi: the VSI being un-configured
2687 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
2689 struct i40e_pf *pf = vsi->back;
2690 struct i40e_hw *hw = &pf->hw;
2691 int base = vsi->base_vector;
2694 for (i = 0; i < vsi->num_queue_pairs; i++) {
2695 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
2696 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
2699 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2700 for (i = vsi->base_vector;
2701 i < (vsi->num_q_vectors + vsi->base_vector); i++)
2702 wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
2705 for (i = 0; i < vsi->num_q_vectors; i++)
2706 synchronize_irq(pf->msix_entries[i + base].vector);
2708 /* Legacy and MSI mode - this stops all interrupt handling */
2709 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
2710 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
2712 synchronize_irq(pf->pdev->irq);
2717 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
2718 * @vsi: the VSI being configured
2720 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
2722 struct i40e_pf *pf = vsi->back;
2725 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2726 for (i = vsi->base_vector;
2727 i < (vsi->num_q_vectors + vsi->base_vector); i++)
2728 i40e_irq_dynamic_enable(vsi, i);
2730 i40e_irq_dynamic_enable_icr0(pf);
2733 i40e_flush(&pf->hw);
2738 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
2739 * @pf: board private structure
2741 static void i40e_stop_misc_vector(struct i40e_pf *pf)
2744 wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
2745 i40e_flush(&pf->hw);
2749 * i40e_intr - MSI/Legacy and non-queue interrupt handler
2750 * @irq: interrupt number
2751 * @data: pointer to a q_vector
2753 * This is the handler used for all MSI/Legacy interrupts, and deals
2754 * with both queue and non-queue interrupts. This is also used in
2755 * MSIX mode to handle the non-queue interrupts.
2757 static irqreturn_t i40e_intr(int irq, void *data)
2759 struct i40e_pf *pf = (struct i40e_pf *)data;
2760 struct i40e_hw *hw = &pf->hw;
2761 u32 icr0, icr0_remaining;
2764 icr0 = rd32(hw, I40E_PFINT_ICR0);
2766 /* if sharing a legacy IRQ, we might get called w/o an intr pending */
2767 if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
2770 ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
2772 /* if interrupt but no bits showing, must be SWINT */
2773 if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
2774 (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
2777 /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
2778 if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
2780 /* temporarily disable queue cause for NAPI processing */
2781 u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
2782 qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
2783 wr32(hw, I40E_QINT_RQCTL(0), qval);
2785 qval = rd32(hw, I40E_QINT_TQCTL(0));
2786 qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
2787 wr32(hw, I40E_QINT_TQCTL(0), qval);
2789 if (!test_bit(__I40E_DOWN, &pf->state))
2790 napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0]->napi);
2793 if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
2794 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2795 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
2798 if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
2799 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
2800 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
2803 if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
2804 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
2805 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
2808 if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
2809 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
2810 set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
2811 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
2812 val = rd32(hw, I40E_GLGEN_RSTAT);
2813 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
2814 >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
2815 if (val == I40E_RESET_CORER)
2817 else if (val == I40E_RESET_GLOBR)
2819 else if (val == I40E_RESET_EMPR)
2823 if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
2824 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
2825 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
2828 /* If a critical error is pending we have no choice but to reset the
2830 * Report and mask out any remaining unexpected interrupts.
2832 icr0_remaining = icr0 & ena_mask;
2833 if (icr0_remaining) {
2834 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
2836 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
2837 (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
2838 (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK) ||
2839 (icr0_remaining & I40E_PFINT_ICR0_MAL_DETECT_MASK)) {
2840 dev_info(&pf->pdev->dev, "device will be reset\n");
2841 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2842 i40e_service_event_schedule(pf);
2844 ena_mask &= ~icr0_remaining;
2847 /* re-enable interrupt causes */
2848 wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
2849 if (!test_bit(__I40E_DOWN, &pf->state)) {
2850 i40e_service_event_schedule(pf);
2851 i40e_irq_dynamic_enable_icr0(pf);
2858 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
2859 * @vsi: the VSI being configured
2860 * @v_idx: vector index
2861 * @qp_idx: queue pair index
2863 static void map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
2865 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
2866 struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
2867 struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
2869 tx_ring->q_vector = q_vector;
2870 tx_ring->next = q_vector->tx.ring;
2871 q_vector->tx.ring = tx_ring;
2872 q_vector->tx.count++;
2874 rx_ring->q_vector = q_vector;
2875 rx_ring->next = q_vector->rx.ring;
2876 q_vector->rx.ring = rx_ring;
2877 q_vector->rx.count++;
2881 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
2882 * @vsi: the VSI being configured
2884 * This function maps descriptor rings to the queue-specific vectors
2885 * we were allotted through the MSI-X enabling code. Ideally, we'd have
2886 * one vector per queue pair, but on a constrained vector budget, we
2887 * group the queue pairs as "efficiently" as possible.
2889 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
2891 int qp_remaining = vsi->num_queue_pairs;
2892 int q_vectors = vsi->num_q_vectors;
2897 /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
2898 * group them so there are multiple queues per vector.
2900 for (; v_start < q_vectors && qp_remaining; v_start++) {
2901 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
2903 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
2905 q_vector->num_ringpairs = num_ringpairs;
2907 q_vector->rx.count = 0;
2908 q_vector->tx.count = 0;
2909 q_vector->rx.ring = NULL;
2910 q_vector->tx.ring = NULL;
2912 while (num_ringpairs--) {
2913 map_vector_to_qp(vsi, v_start, qp_idx);
2921 * i40e_vsi_request_irq - Request IRQ from the OS
2922 * @vsi: the VSI being configured
2923 * @basename: name for the vector
2925 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
2927 struct i40e_pf *pf = vsi->back;
2930 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
2931 err = i40e_vsi_request_irq_msix(vsi, basename);
2932 else if (pf->flags & I40E_FLAG_MSI_ENABLED)
2933 err = request_irq(pf->pdev->irq, i40e_intr, 0,
2934 pf->misc_int_name, pf);
2936 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
2937 pf->misc_int_name, pf);
2940 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
2945 #ifdef CONFIG_NET_POLL_CONTROLLER
2947 * i40e_netpoll - A Polling 'interrupt'handler
2948 * @netdev: network interface device structure
2950 * This is used by netconsole to send skbs without having to re-enable
2951 * interrupts. It's not called while the normal interrupt routine is executing.
2953 static void i40e_netpoll(struct net_device *netdev)
2955 struct i40e_netdev_priv *np = netdev_priv(netdev);
2956 struct i40e_vsi *vsi = np->vsi;
2957 struct i40e_pf *pf = vsi->back;
2960 /* if interface is down do nothing */
2961 if (test_bit(__I40E_DOWN, &vsi->state))
2964 pf->flags |= I40E_FLAG_IN_NETPOLL;
2965 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2966 for (i = 0; i < vsi->num_q_vectors; i++)
2967 i40e_msix_clean_rings(0, vsi->q_vectors[i]);
2969 i40e_intr(pf->pdev->irq, netdev);
2971 pf->flags &= ~I40E_FLAG_IN_NETPOLL;
2976 * i40e_vsi_control_tx - Start or stop a VSI's rings
2977 * @vsi: the VSI being configured
2978 * @enable: start or stop the rings
2980 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
2982 struct i40e_pf *pf = vsi->back;
2983 struct i40e_hw *hw = &pf->hw;
2987 pf_q = vsi->base_queue;
2988 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
2991 usleep_range(1000, 2000);
2992 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
2993 } while (j-- && ((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT)
2994 ^ (tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT)) & 1);
2996 /* Skip if the queue is already in the requested state */
2997 if (enable && (tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
2999 if (!enable && !(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3002 /* turn on/off the queue */
3004 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK |
3005 I40E_QTX_ENA_QENA_STAT_MASK;
3007 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3009 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3011 /* wait for the change to finish */
3012 for (j = 0; j < 10; j++) {
3013 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3015 if ((tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3018 if (!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3025 dev_info(&pf->pdev->dev, "Tx ring %d %sable timeout\n",
3026 pf_q, (enable ? "en" : "dis"));
3031 if (hw->revision_id == 0)
3038 * i40e_vsi_control_rx - Start or stop a VSI's rings
3039 * @vsi: the VSI being configured
3040 * @enable: start or stop the rings
3042 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3044 struct i40e_pf *pf = vsi->back;
3045 struct i40e_hw *hw = &pf->hw;
3049 pf_q = vsi->base_queue;
3050 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3053 usleep_range(1000, 2000);
3054 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3055 } while (j-- && ((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT)
3056 ^ (rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT)) & 1);
3060 if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3063 /* is !STAT set ? */
3064 if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3068 /* turn on/off the queue */
3070 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK |
3071 I40E_QRX_ENA_QENA_STAT_MASK;
3073 rx_reg &= ~(I40E_QRX_ENA_QENA_REQ_MASK |
3074 I40E_QRX_ENA_QENA_STAT_MASK);
3075 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3077 /* wait for the change to finish */
3078 for (j = 0; j < 10; j++) {
3079 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3082 if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3085 if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3092 dev_info(&pf->pdev->dev, "Rx ring %d %sable timeout\n",
3093 pf_q, (enable ? "en" : "dis"));
3102 * i40e_vsi_control_rings - Start or stop a VSI's rings
3103 * @vsi: the VSI being configured
3104 * @enable: start or stop the rings
3106 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3110 /* do rx first for enable and last for disable */
3112 ret = i40e_vsi_control_rx(vsi, request);
3115 ret = i40e_vsi_control_tx(vsi, request);
3117 ret = i40e_vsi_control_tx(vsi, request);
3120 ret = i40e_vsi_control_rx(vsi, request);
3127 * i40e_vsi_free_irq - Free the irq association with the OS
3128 * @vsi: the VSI being configured
3130 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3132 struct i40e_pf *pf = vsi->back;
3133 struct i40e_hw *hw = &pf->hw;
3134 int base = vsi->base_vector;
3138 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3139 if (!vsi->q_vectors)
3142 for (i = 0; i < vsi->num_q_vectors; i++) {
3143 u16 vector = i + base;
3145 /* free only the irqs that were actually requested */
3146 if (!vsi->q_vectors[i] ||
3147 !vsi->q_vectors[i]->num_ringpairs)
3150 /* clear the affinity_mask in the IRQ descriptor */
3151 irq_set_affinity_hint(pf->msix_entries[vector].vector,
3153 free_irq(pf->msix_entries[vector].vector,
3156 /* Tear down the interrupt queue link list
3158 * We know that they come in pairs and always
3159 * the Rx first, then the Tx. To clear the
3160 * link list, stick the EOL value into the
3161 * next_q field of the registers.
3163 val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3164 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3165 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3166 val |= I40E_QUEUE_END_OF_LIST
3167 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3168 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3170 while (qp != I40E_QUEUE_END_OF_LIST) {
3173 val = rd32(hw, I40E_QINT_RQCTL(qp));
3175 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
3176 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3177 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3178 I40E_QINT_RQCTL_INTEVENT_MASK);
3180 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3181 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3183 wr32(hw, I40E_QINT_RQCTL(qp), val);
3185 val = rd32(hw, I40E_QINT_TQCTL(qp));
3187 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3188 >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3190 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
3191 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3192 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3193 I40E_QINT_TQCTL_INTEVENT_MASK);
3195 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3196 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3198 wr32(hw, I40E_QINT_TQCTL(qp), val);
3203 free_irq(pf->pdev->irq, pf);
3205 val = rd32(hw, I40E_PFINT_LNKLST0);
3206 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3207 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3208 val |= I40E_QUEUE_END_OF_LIST
3209 << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3210 wr32(hw, I40E_PFINT_LNKLST0, val);
3212 val = rd32(hw, I40E_QINT_RQCTL(qp));
3213 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
3214 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3215 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3216 I40E_QINT_RQCTL_INTEVENT_MASK);
3218 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3219 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3221 wr32(hw, I40E_QINT_RQCTL(qp), val);
3223 val = rd32(hw, I40E_QINT_TQCTL(qp));
3225 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
3226 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3227 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3228 I40E_QINT_TQCTL_INTEVENT_MASK);
3230 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3231 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3233 wr32(hw, I40E_QINT_TQCTL(qp), val);
3238 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
3239 * @vsi: the VSI being configured
3240 * @v_idx: Index of vector to be freed
3242 * This function frees the memory allocated to the q_vector. In addition if
3243 * NAPI is enabled it will delete any references to the NAPI struct prior
3244 * to freeing the q_vector.
3246 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
3248 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3249 struct i40e_ring *ring;
3254 /* disassociate q_vector from rings */
3255 i40e_for_each_ring(ring, q_vector->tx)
3256 ring->q_vector = NULL;
3258 i40e_for_each_ring(ring, q_vector->rx)
3259 ring->q_vector = NULL;
3261 /* only VSI w/ an associated netdev is set up w/ NAPI */
3263 netif_napi_del(&q_vector->napi);
3265 vsi->q_vectors[v_idx] = NULL;
3267 kfree_rcu(q_vector, rcu);
3271 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3272 * @vsi: the VSI being un-configured
3274 * This frees the memory allocated to the q_vectors and
3275 * deletes references to the NAPI struct.
3277 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3281 for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
3282 i40e_free_q_vector(vsi, v_idx);
3286 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3287 * @pf: board private structure
3289 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3291 /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3292 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3293 pci_disable_msix(pf->pdev);
3294 kfree(pf->msix_entries);
3295 pf->msix_entries = NULL;
3296 } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3297 pci_disable_msi(pf->pdev);
3299 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3303 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3304 * @pf: board private structure
3306 * We go through and clear interrupt specific resources and reset the structure
3307 * to pre-load conditions
3309 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3313 i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3314 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
3316 i40e_vsi_free_q_vectors(pf->vsi[i]);
3317 i40e_reset_interrupt_capability(pf);
3321 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3322 * @vsi: the VSI being configured
3324 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3331 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3332 napi_enable(&vsi->q_vectors[q_idx]->napi);
3336 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3337 * @vsi: the VSI being configured
3339 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3346 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3347 napi_disable(&vsi->q_vectors[q_idx]->napi);
3351 * i40e_quiesce_vsi - Pause a given VSI
3352 * @vsi: the VSI being paused
3354 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3356 if (test_bit(__I40E_DOWN, &vsi->state))
3359 set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3360 if (vsi->netdev && netif_running(vsi->netdev)) {
3361 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3363 set_bit(__I40E_DOWN, &vsi->state);
3369 * i40e_unquiesce_vsi - Resume a given VSI
3370 * @vsi: the VSI being resumed
3372 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3374 if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3377 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3378 if (vsi->netdev && netif_running(vsi->netdev))
3379 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3381 i40e_up(vsi); /* this clears the DOWN bit */
3385 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3388 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3392 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3394 i40e_quiesce_vsi(pf->vsi[v]);
3399 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3402 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3406 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3408 i40e_unquiesce_vsi(pf->vsi[v]);
3413 * i40e_dcb_get_num_tc - Get the number of TCs from DCBx config
3414 * @dcbcfg: the corresponding DCBx configuration structure
3416 * Return the number of TCs from given DCBx configuration
3418 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
3423 /* Scan the ETS Config Priority Table to find
3424 * traffic class enabled for a given priority
3425 * and use the traffic class index to get the
3426 * number of traffic classes enabled
3428 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3429 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
3430 num_tc = dcbcfg->etscfg.prioritytable[i];
3433 /* Traffic class index starts from zero so
3434 * increment to return the actual count
3440 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
3441 * @dcbcfg: the corresponding DCBx configuration structure
3443 * Query the current DCB configuration and return the number of
3444 * traffic classes enabled from the given DCBX config
3446 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
3448 u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
3452 for (i = 0; i < num_tc; i++)
3453 enabled_tc |= 1 << i;
3459 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
3460 * @pf: PF being queried
3462 * Return number of traffic classes enabled for the given PF
3464 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
3466 struct i40e_hw *hw = &pf->hw;
3469 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3471 /* If DCB is not enabled then always in single TC */
3472 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3475 /* MFP mode return count of enabled TCs for this PF */
3476 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3477 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3478 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3479 if (enabled_tc & (1 << i))
3485 /* SFP mode will be enabled for all TCs on port */
3486 return i40e_dcb_get_num_tc(dcbcfg);
3490 * i40e_pf_get_default_tc - Get bitmap for first enabled TC
3491 * @pf: PF being queried
3493 * Return a bitmap for first enabled traffic class for this PF.
3495 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
3497 u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3501 return 0x1; /* TC0 */
3503 /* Find the first enabled TC */
3504 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3505 if (enabled_tc & (1 << i))
3513 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
3514 * @pf: PF being queried
3516 * Return a bitmap for enabled traffic classes for this PF.
3518 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
3520 /* If DCB is not enabled for this PF then just return default TC */
3521 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3522 return i40e_pf_get_default_tc(pf);
3524 /* MFP mode will have enabled TCs set by FW */
3525 if (pf->flags & I40E_FLAG_MFP_ENABLED)
3526 return pf->hw.func_caps.enabled_tcmap;
3528 /* SFP mode we want PF to be enabled for all TCs */
3529 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
3533 * i40e_vsi_get_bw_info - Query VSI BW Information
3534 * @vsi: the VSI being queried
3536 * Returns 0 on success, negative value on failure
3538 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
3540 struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
3541 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
3542 struct i40e_pf *pf = vsi->back;
3543 struct i40e_hw *hw = &pf->hw;
3548 /* Get the VSI level BW configuration */
3549 aq_ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
3551 dev_info(&pf->pdev->dev,
3552 "couldn't get pf vsi bw config, err %d, aq_err %d\n",
3553 aq_ret, pf->hw.aq.asq_last_status);
3557 /* Get the VSI level BW configuration per TC */
3558 aq_ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
3561 dev_info(&pf->pdev->dev,
3562 "couldn't get pf vsi ets bw config, err %d, aq_err %d\n",
3563 aq_ret, pf->hw.aq.asq_last_status);
3567 if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
3568 dev_info(&pf->pdev->dev,
3569 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
3570 bw_config.tc_valid_bits,
3571 bw_ets_config.tc_valid_bits);
3572 /* Still continuing */
3575 vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
3576 vsi->bw_max_quanta = bw_config.max_bw;
3577 tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
3578 (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
3579 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3580 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
3581 vsi->bw_ets_limit_credits[i] =
3582 le16_to_cpu(bw_ets_config.credits[i]);
3583 /* 3 bits out of 4 for each TC */
3584 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
3591 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
3592 * @vsi: the VSI being configured
3593 * @enabled_tc: TC bitmap
3594 * @bw_credits: BW shared credits per TC
3596 * Returns 0 on success, negative value on failure
3598 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
3601 struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
3605 bw_data.tc_valid_bits = enabled_tc;
3606 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3607 bw_data.tc_bw_credits[i] = bw_share[i];
3609 aq_ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
3612 dev_info(&vsi->back->pdev->dev,
3613 "%s: AQ command Config VSI BW allocation per TC failed = %d\n",
3614 __func__, vsi->back->hw.aq.asq_last_status);
3618 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3619 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
3625 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
3626 * @vsi: the VSI being configured
3627 * @enabled_tc: TC map to be enabled
3630 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3632 struct net_device *netdev = vsi->netdev;
3633 struct i40e_pf *pf = vsi->back;
3634 struct i40e_hw *hw = &pf->hw;
3637 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3643 netdev_reset_tc(netdev);
3647 /* Set up actual enabled TCs on the VSI */
3648 if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
3651 /* set per TC queues for the VSI */
3652 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3653 /* Only set TC queues for enabled tcs
3655 * e.g. For a VSI that has TC0 and TC3 enabled the
3656 * enabled_tc bitmap would be 0x00001001; the driver
3657 * will set the numtc for netdev as 2 that will be
3658 * referenced by the netdev layer as TC 0 and 1.
3660 if (vsi->tc_config.enabled_tc & (1 << i))
3661 netdev_set_tc_queue(netdev,
3662 vsi->tc_config.tc_info[i].netdev_tc,
3663 vsi->tc_config.tc_info[i].qcount,
3664 vsi->tc_config.tc_info[i].qoffset);
3667 /* Assign UP2TC map for the VSI */
3668 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3669 /* Get the actual TC# for the UP */
3670 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
3671 /* Get the mapped netdev TC# for the UP */
3672 netdev_tc = vsi->tc_config.tc_info[ets_tc].netdev_tc;
3673 netdev_set_prio_tc_map(netdev, i, netdev_tc);
3678 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
3679 * @vsi: the VSI being configured
3680 * @ctxt: the ctxt buffer returned from AQ VSI update param command
3682 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
3683 struct i40e_vsi_context *ctxt)
3685 /* copy just the sections touched not the entire info
3686 * since not all sections are valid as returned by
3689 vsi->info.mapping_flags = ctxt->info.mapping_flags;
3690 memcpy(&vsi->info.queue_mapping,
3691 &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
3692 memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
3693 sizeof(vsi->info.tc_mapping));
3697 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
3698 * @vsi: VSI to be configured
3699 * @enabled_tc: TC bitmap
3701 * This configures a particular VSI for TCs that are mapped to the
3702 * given TC bitmap. It uses default bandwidth share for TCs across
3703 * VSIs to configure TC for a particular VSI.
3706 * It is expected that the VSI queues have been quisced before calling
3709 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3711 u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
3712 struct i40e_vsi_context ctxt;
3716 /* Check if enabled_tc is same as existing or new TCs */
3717 if (vsi->tc_config.enabled_tc == enabled_tc)
3720 /* Enable ETS TCs with equal BW Share for now across all VSIs */
3721 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3722 if (enabled_tc & (1 << i))
3726 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
3728 dev_info(&vsi->back->pdev->dev,
3729 "Failed configuring TC map %d for VSI %d\n",
3730 enabled_tc, vsi->seid);
3734 /* Update Queue Pairs Mapping for currently enabled UPs */
3735 ctxt.seid = vsi->seid;
3736 ctxt.pf_num = vsi->back->hw.pf_id;
3738 ctxt.uplink_seid = vsi->uplink_seid;
3739 memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
3740 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
3742 /* Update the VSI after updating the VSI queue-mapping information */
3743 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
3745 dev_info(&vsi->back->pdev->dev,
3746 "update vsi failed, aq_err=%d\n",
3747 vsi->back->hw.aq.asq_last_status);
3750 /* update the local VSI info with updated queue map */
3751 i40e_vsi_update_queue_map(vsi, &ctxt);
3752 vsi->info.valid_sections = 0;
3754 /* Update current VSI BW information */
3755 ret = i40e_vsi_get_bw_info(vsi);
3757 dev_info(&vsi->back->pdev->dev,
3758 "Failed updating vsi bw info, aq_err=%d\n",
3759 vsi->back->hw.aq.asq_last_status);
3763 /* Update the netdev TC setup */
3764 i40e_vsi_config_netdev_tc(vsi, enabled_tc);
3770 * i40e_up_complete - Finish the last steps of bringing up a connection
3771 * @vsi: the VSI being configured
3773 static int i40e_up_complete(struct i40e_vsi *vsi)
3775 struct i40e_pf *pf = vsi->back;
3778 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3779 i40e_vsi_configure_msix(vsi);
3781 i40e_configure_msi_and_legacy(vsi);
3784 err = i40e_vsi_control_rings(vsi, true);
3788 clear_bit(__I40E_DOWN, &vsi->state);
3789 i40e_napi_enable_all(vsi);
3790 i40e_vsi_enable_irq(vsi);
3792 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
3794 netdev_info(vsi->netdev, "NIC Link is Up\n");
3795 netif_tx_start_all_queues(vsi->netdev);
3796 netif_carrier_on(vsi->netdev);
3797 } else if (vsi->netdev) {
3798 netdev_info(vsi->netdev, "NIC Link is Down\n");
3800 i40e_service_event_schedule(pf);
3806 * i40e_vsi_reinit_locked - Reset the VSI
3807 * @vsi: the VSI being configured
3809 * Rebuild the ring structs after some configuration
3810 * has changed, e.g. MTU size.
3812 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
3814 struct i40e_pf *pf = vsi->back;
3816 WARN_ON(in_interrupt());
3817 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
3818 usleep_range(1000, 2000);
3821 /* Give a VF some time to respond to the reset. The
3822 * two second wait is based upon the watchdog cycle in
3825 if (vsi->type == I40E_VSI_SRIOV)
3828 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
3832 * i40e_up - Bring the connection back up after being down
3833 * @vsi: the VSI being configured
3835 int i40e_up(struct i40e_vsi *vsi)
3839 err = i40e_vsi_configure(vsi);
3841 err = i40e_up_complete(vsi);
3847 * i40e_down - Shutdown the connection processing
3848 * @vsi: the VSI being stopped
3850 void i40e_down(struct i40e_vsi *vsi)
3854 /* It is assumed that the caller of this function
3855 * sets the vsi->state __I40E_DOWN bit.
3858 netif_carrier_off(vsi->netdev);
3859 netif_tx_disable(vsi->netdev);
3861 i40e_vsi_disable_irq(vsi);
3862 i40e_vsi_control_rings(vsi, false);
3863 i40e_napi_disable_all(vsi);
3865 for (i = 0; i < vsi->num_queue_pairs; i++) {
3866 i40e_clean_tx_ring(vsi->tx_rings[i]);
3867 i40e_clean_rx_ring(vsi->rx_rings[i]);
3872 * i40e_setup_tc - configure multiple traffic classes
3873 * @netdev: net device to configure
3874 * @tc: number of traffic classes to enable
3876 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
3878 struct i40e_netdev_priv *np = netdev_priv(netdev);
3879 struct i40e_vsi *vsi = np->vsi;
3880 struct i40e_pf *pf = vsi->back;
3885 /* Check if DCB enabled to continue */
3886 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
3887 netdev_info(netdev, "DCB is not enabled for adapter\n");
3891 /* Check if MFP enabled */
3892 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3893 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
3897 /* Check whether tc count is within enabled limit */
3898 if (tc > i40e_pf_get_num_tc(pf)) {
3899 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
3903 /* Generate TC map for number of tc requested */
3904 for (i = 0; i < tc; i++)
3905 enabled_tc |= (1 << i);
3907 /* Requesting same TC configuration as already enabled */
3908 if (enabled_tc == vsi->tc_config.enabled_tc)
3911 /* Quiesce VSI queues */
3912 i40e_quiesce_vsi(vsi);
3914 /* Configure VSI for enabled TCs */
3915 ret = i40e_vsi_config_tc(vsi, enabled_tc);
3917 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
3923 i40e_unquiesce_vsi(vsi);
3930 * i40e_open - Called when a network interface is made active
3931 * @netdev: network interface device structure
3933 * The open entry point is called when a network interface is made
3934 * active by the system (IFF_UP). At this point all resources needed
3935 * for transmit and receive operations are allocated, the interrupt
3936 * handler is registered with the OS, the netdev watchdog subtask is
3937 * enabled, and the stack is notified that the interface is ready.
3939 * Returns 0 on success, negative value on failure
3941 static int i40e_open(struct net_device *netdev)
3943 struct i40e_netdev_priv *np = netdev_priv(netdev);
3944 struct i40e_vsi *vsi = np->vsi;
3945 struct i40e_pf *pf = vsi->back;
3946 char int_name[IFNAMSIZ];
3949 /* disallow open during test */
3950 if (test_bit(__I40E_TESTING, &pf->state))
3953 netif_carrier_off(netdev);
3955 /* allocate descriptors */
3956 err = i40e_vsi_setup_tx_resources(vsi);
3959 err = i40e_vsi_setup_rx_resources(vsi);
3963 err = i40e_vsi_configure(vsi);
3967 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3968 dev_driver_string(&pf->pdev->dev), netdev->name);
3969 err = i40e_vsi_request_irq(vsi, int_name);
3973 /* Notify the stack of the actual queue counts. */
3974 err = netif_set_real_num_tx_queues(netdev, vsi->num_queue_pairs);
3976 goto err_set_queues;
3978 err = netif_set_real_num_rx_queues(netdev, vsi->num_queue_pairs);
3980 goto err_set_queues;
3982 err = i40e_up_complete(vsi);
3984 goto err_up_complete;
3986 #ifdef CONFIG_I40E_VXLAN
3987 vxlan_get_rx_port(netdev);
3995 i40e_vsi_free_irq(vsi);
3997 i40e_vsi_free_rx_resources(vsi);
3999 i40e_vsi_free_tx_resources(vsi);
4000 if (vsi == pf->vsi[pf->lan_vsi])
4001 i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
4007 * i40e_close - Disables a network interface
4008 * @netdev: network interface device structure
4010 * The close entry point is called when an interface is de-activated
4011 * by the OS. The hardware is still under the driver's control, but
4012 * this netdev interface is disabled.
4014 * Returns 0, this is not allowed to fail
4016 static int i40e_close(struct net_device *netdev)
4018 struct i40e_netdev_priv *np = netdev_priv(netdev);
4019 struct i40e_vsi *vsi = np->vsi;
4021 if (test_and_set_bit(__I40E_DOWN, &vsi->state))
4025 i40e_vsi_free_irq(vsi);
4027 i40e_vsi_free_tx_resources(vsi);
4028 i40e_vsi_free_rx_resources(vsi);
4034 * i40e_do_reset - Start a PF or Core Reset sequence
4035 * @pf: board private structure
4036 * @reset_flags: which reset is requested
4038 * The essential difference in resets is that the PF Reset
4039 * doesn't clear the packet buffers, doesn't reset the PE
4040 * firmware, and doesn't bother the other PFs on the chip.
4042 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
4046 WARN_ON(in_interrupt());
4048 /* do the biggest reset indicated */
4049 if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
4051 /* Request a Global Reset
4053 * This will start the chip's countdown to the actual full
4054 * chip reset event, and a warning interrupt to be sent
4055 * to all PFs, including the requestor. Our handler
4056 * for the warning interrupt will deal with the shutdown
4057 * and recovery of the switch setup.
4059 dev_info(&pf->pdev->dev, "GlobalR requested\n");
4060 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4061 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
4062 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4064 } else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
4066 /* Request a Core Reset
4068 * Same as Global Reset, except does *not* include the MAC/PHY
4070 dev_info(&pf->pdev->dev, "CoreR requested\n");
4071 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4072 val |= I40E_GLGEN_RTRIG_CORER_MASK;
4073 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4074 i40e_flush(&pf->hw);
4076 } else if (reset_flags & (1 << __I40E_EMP_RESET_REQUESTED)) {
4078 /* Request a Firmware Reset
4080 * Same as Global reset, plus restarting the
4081 * embedded firmware engine.
4083 /* enable EMP Reset */
4084 val = rd32(&pf->hw, I40E_GLGEN_RSTENA_EMP);
4085 val |= I40E_GLGEN_RSTENA_EMP_EMP_RST_ENA_MASK;
4086 wr32(&pf->hw, I40E_GLGEN_RSTENA_EMP, val);
4088 /* force the reset */
4089 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4090 val |= I40E_GLGEN_RTRIG_EMPFWR_MASK;
4091 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4092 i40e_flush(&pf->hw);
4094 } else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
4096 /* Request a PF Reset
4098 * Resets only the PF-specific registers
4100 * This goes directly to the tear-down and rebuild of
4101 * the switch, since we need to do all the recovery as
4102 * for the Core Reset.
4104 dev_info(&pf->pdev->dev, "PFR requested\n");
4105 i40e_handle_reset_warning(pf);
4107 } else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
4110 /* Find the VSI(s) that requested a re-init */
4111 dev_info(&pf->pdev->dev,
4112 "VSI reinit requested\n");
4113 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4114 struct i40e_vsi *vsi = pf->vsi[v];
4116 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
4117 i40e_vsi_reinit_locked(pf->vsi[v]);
4118 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
4122 /* no further action needed, so return now */
4125 dev_info(&pf->pdev->dev,
4126 "bad reset request 0x%08x\n", reset_flags);
4132 * i40e_do_reset_safe - Protected reset path for userland calls.
4133 * @pf: board private structure
4134 * @reset_flags: which reset is requested
4137 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
4140 i40e_do_reset(pf, reset_flags);
4145 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
4146 * @pf: board private structure
4147 * @e: event info posted on ARQ
4149 * Handler for LAN Queue Overflow Event generated by the firmware for PF
4152 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
4153 struct i40e_arq_event_info *e)
4155 struct i40e_aqc_lan_overflow *data =
4156 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
4157 u32 queue = le32_to_cpu(data->prtdcb_rupto);
4158 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
4159 struct i40e_hw *hw = &pf->hw;
4163 dev_info(&pf->pdev->dev, "%s: Rx Queue Number = %d QTX_CTL=0x%08x\n",
4164 __func__, queue, qtx_ctl);
4166 /* Queue belongs to VF, find the VF and issue VF reset */
4167 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
4168 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
4169 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
4170 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
4171 vf_id -= hw->func_caps.vf_base_id;
4172 vf = &pf->vf[vf_id];
4173 i40e_vc_notify_vf_reset(vf);
4174 /* Allow VF to process pending reset notification */
4176 i40e_reset_vf(vf, false);
4181 * i40e_service_event_complete - Finish up the service event
4182 * @pf: board private structure
4184 static void i40e_service_event_complete(struct i40e_pf *pf)
4186 BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
4188 /* flush memory to make sure state is correct before next watchog */
4189 smp_mb__before_clear_bit();
4190 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
4194 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
4195 * @pf: board private structure
4197 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
4199 if (!(pf->flags & I40E_FLAG_FDIR_REQUIRES_REINIT))
4202 pf->flags &= ~I40E_FLAG_FDIR_REQUIRES_REINIT;
4204 /* if interface is down do nothing */
4205 if (test_bit(__I40E_DOWN, &pf->state))
4210 * i40e_vsi_link_event - notify VSI of a link event
4211 * @vsi: vsi to be notified
4212 * @link_up: link up or down
4214 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
4219 switch (vsi->type) {
4221 if (!vsi->netdev || !vsi->netdev_registered)
4225 netif_carrier_on(vsi->netdev);
4226 netif_tx_wake_all_queues(vsi->netdev);
4228 netif_carrier_off(vsi->netdev);
4229 netif_tx_stop_all_queues(vsi->netdev);
4233 case I40E_VSI_SRIOV:
4236 case I40E_VSI_VMDQ2:
4238 case I40E_VSI_MIRROR:
4240 /* there is no notification for other VSIs */
4246 * i40e_veb_link_event - notify elements on the veb of a link event
4247 * @veb: veb to be notified
4248 * @link_up: link up or down
4250 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
4255 if (!veb || !veb->pf)
4259 /* depth first... */
4260 for (i = 0; i < I40E_MAX_VEB; i++)
4261 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
4262 i40e_veb_link_event(pf->veb[i], link_up);
4264 /* ... now the local VSIs */
4265 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4266 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
4267 i40e_vsi_link_event(pf->vsi[i], link_up);
4271 * i40e_link_event - Update netif_carrier status
4272 * @pf: board private structure
4274 static void i40e_link_event(struct i40e_pf *pf)
4276 bool new_link, old_link;
4278 new_link = (pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP);
4279 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
4281 if (new_link == old_link)
4284 if (!test_bit(__I40E_DOWN, &pf->vsi[pf->lan_vsi]->state))
4285 netdev_info(pf->vsi[pf->lan_vsi]->netdev,
4286 "NIC Link is %s\n", (new_link ? "Up" : "Down"));
4288 /* Notify the base of the switch tree connected to
4289 * the link. Floating VEBs are not notified.
4291 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
4292 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
4294 i40e_vsi_link_event(pf->vsi[pf->lan_vsi], new_link);
4297 i40e_vc_notify_link_state(pf);
4301 * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
4302 * @pf: board private structure
4304 * Set the per-queue flags to request a check for stuck queues in the irq
4305 * clean functions, then force interrupts to be sure the irq clean is called.
4307 static void i40e_check_hang_subtask(struct i40e_pf *pf)
4311 /* If we're down or resetting, just bail */
4312 if (test_bit(__I40E_CONFIG_BUSY, &pf->state))
4315 /* for each VSI/netdev
4317 * set the check flag
4319 * force an interrupt
4321 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4322 struct i40e_vsi *vsi = pf->vsi[v];
4326 test_bit(__I40E_DOWN, &vsi->state) ||
4327 (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
4330 for (i = 0; i < vsi->num_queue_pairs; i++) {
4331 set_check_for_tx_hang(vsi->tx_rings[i]);
4332 if (test_bit(__I40E_HANG_CHECK_ARMED,
4333 &vsi->tx_rings[i]->state))
4338 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
4339 wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
4340 (I40E_PFINT_DYN_CTL0_INTENA_MASK |
4341 I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK));
4343 u16 vec = vsi->base_vector - 1;
4344 u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
4345 I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK);
4346 for (i = 0; i < vsi->num_q_vectors; i++, vec++)
4347 wr32(&vsi->back->hw,
4348 I40E_PFINT_DYN_CTLN(vec), val);
4350 i40e_flush(&vsi->back->hw);
4356 * i40e_watchdog_subtask - Check and bring link up
4357 * @pf: board private structure
4359 static void i40e_watchdog_subtask(struct i40e_pf *pf)
4363 /* if interface is down do nothing */
4364 if (test_bit(__I40E_DOWN, &pf->state) ||
4365 test_bit(__I40E_CONFIG_BUSY, &pf->state))
4368 /* Update the stats for active netdevs so the network stack
4369 * can look at updated numbers whenever it cares to
4371 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4372 if (pf->vsi[i] && pf->vsi[i]->netdev)
4373 i40e_update_stats(pf->vsi[i]);
4375 /* Update the stats for the active switching components */
4376 for (i = 0; i < I40E_MAX_VEB; i++)
4378 i40e_update_veb_stats(pf->veb[i]);
4382 * i40e_reset_subtask - Set up for resetting the device and driver
4383 * @pf: board private structure
4385 static void i40e_reset_subtask(struct i40e_pf *pf)
4387 u32 reset_flags = 0;
4390 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
4391 reset_flags |= (1 << __I40E_REINIT_REQUESTED);
4392 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
4394 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
4395 reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
4396 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4398 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
4399 reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
4400 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
4402 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
4403 reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
4404 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
4407 /* If there's a recovery already waiting, it takes
4408 * precedence before starting a new reset sequence.
4410 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
4411 i40e_handle_reset_warning(pf);
4415 /* If we're already down or resetting, just bail */
4417 !test_bit(__I40E_DOWN, &pf->state) &&
4418 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
4419 i40e_do_reset(pf, reset_flags);
4426 * i40e_handle_link_event - Handle link event
4427 * @pf: board private structure
4428 * @e: event info posted on ARQ
4430 static void i40e_handle_link_event(struct i40e_pf *pf,
4431 struct i40e_arq_event_info *e)
4433 struct i40e_hw *hw = &pf->hw;
4434 struct i40e_aqc_get_link_status *status =
4435 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
4436 struct i40e_link_status *hw_link_info = &hw->phy.link_info;
4438 /* save off old link status information */
4439 memcpy(&pf->hw.phy.link_info_old, hw_link_info,
4440 sizeof(pf->hw.phy.link_info_old));
4442 /* update link status */
4443 hw_link_info->phy_type = (enum i40e_aq_phy_type)status->phy_type;
4444 hw_link_info->link_speed = (enum i40e_aq_link_speed)status->link_speed;
4445 hw_link_info->link_info = status->link_info;
4446 hw_link_info->an_info = status->an_info;
4447 hw_link_info->ext_info = status->ext_info;
4448 hw_link_info->lse_enable =
4449 le16_to_cpu(status->command_flags) &
4452 /* process the event */
4453 i40e_link_event(pf);
4455 /* Do a new status request to re-enable LSE reporting
4456 * and load new status information into the hw struct,
4457 * then see if the status changed while processing the
4460 i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
4461 i40e_link_event(pf);
4465 * i40e_clean_adminq_subtask - Clean the AdminQ rings
4466 * @pf: board private structure
4468 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
4470 struct i40e_arq_event_info event;
4471 struct i40e_hw *hw = &pf->hw;
4477 if (!test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state))
4480 event.msg_size = I40E_MAX_AQ_BUF_SIZE;
4481 event.msg_buf = kzalloc(event.msg_size, GFP_KERNEL);
4486 event.msg_size = I40E_MAX_AQ_BUF_SIZE; /* reinit each time */
4487 ret = i40e_clean_arq_element(hw, &event, &pending);
4488 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK) {
4489 dev_info(&pf->pdev->dev, "No ARQ event found\n");
4492 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
4496 opcode = le16_to_cpu(event.desc.opcode);
4499 case i40e_aqc_opc_get_link_status:
4500 i40e_handle_link_event(pf, &event);
4502 case i40e_aqc_opc_send_msg_to_pf:
4503 ret = i40e_vc_process_vf_msg(pf,
4504 le16_to_cpu(event.desc.retval),
4505 le32_to_cpu(event.desc.cookie_high),
4506 le32_to_cpu(event.desc.cookie_low),
4510 case i40e_aqc_opc_lldp_update_mib:
4511 dev_info(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
4513 case i40e_aqc_opc_event_lan_overflow:
4514 dev_info(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
4515 i40e_handle_lan_overflow_event(pf, &event);
4518 dev_info(&pf->pdev->dev,
4519 "ARQ Error: Unknown event %d received\n",
4523 } while (pending && (i++ < pf->adminq_work_limit));
4525 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
4526 /* re-enable Admin queue interrupt cause */
4527 val = rd32(hw, I40E_PFINT_ICR0_ENA);
4528 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
4529 wr32(hw, I40E_PFINT_ICR0_ENA, val);
4532 kfree(event.msg_buf);
4536 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
4537 * @veb: pointer to the VEB instance
4539 * This is a recursive function that first builds the attached VSIs then
4540 * recurses in to build the next layer of VEB. We track the connections
4541 * through our own index numbers because the seid's from the HW could
4542 * change across the reset.
4544 static int i40e_reconstitute_veb(struct i40e_veb *veb)
4546 struct i40e_vsi *ctl_vsi = NULL;
4547 struct i40e_pf *pf = veb->pf;
4551 /* build VSI that owns this VEB, temporarily attached to base VEB */
4552 for (v = 0; v < pf->hw.func_caps.num_vsis && !ctl_vsi; v++) {
4554 pf->vsi[v]->veb_idx == veb->idx &&
4555 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
4556 ctl_vsi = pf->vsi[v];
4561 dev_info(&pf->pdev->dev,
4562 "missing owner VSI for veb_idx %d\n", veb->idx);
4564 goto end_reconstitute;
4566 if (ctl_vsi != pf->vsi[pf->lan_vsi])
4567 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
4568 ret = i40e_add_vsi(ctl_vsi);
4570 dev_info(&pf->pdev->dev,
4571 "rebuild of owner VSI failed: %d\n", ret);
4572 goto end_reconstitute;
4574 i40e_vsi_reset_stats(ctl_vsi);
4576 /* create the VEB in the switch and move the VSI onto the VEB */
4577 ret = i40e_add_veb(veb, ctl_vsi);
4579 goto end_reconstitute;
4581 /* create the remaining VSIs attached to this VEB */
4582 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4583 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
4586 if (pf->vsi[v]->veb_idx == veb->idx) {
4587 struct i40e_vsi *vsi = pf->vsi[v];
4588 vsi->uplink_seid = veb->seid;
4589 ret = i40e_add_vsi(vsi);
4591 dev_info(&pf->pdev->dev,
4592 "rebuild of vsi_idx %d failed: %d\n",
4594 goto end_reconstitute;
4596 i40e_vsi_reset_stats(vsi);
4600 /* create any VEBs attached to this VEB - RECURSION */
4601 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
4602 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
4603 pf->veb[veb_idx]->uplink_seid = veb->seid;
4604 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
4615 * i40e_get_capabilities - get info about the HW
4616 * @pf: the PF struct
4618 static int i40e_get_capabilities(struct i40e_pf *pf)
4620 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
4625 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
4627 cap_buf = kzalloc(buf_len, GFP_KERNEL);
4631 /* this loads the data into the hw struct for us */
4632 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
4634 i40e_aqc_opc_list_func_capabilities,
4636 /* data loaded, buffer no longer needed */
4639 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
4640 /* retry with a larger buffer */
4641 buf_len = data_size;
4642 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
4643 dev_info(&pf->pdev->dev,
4644 "capability discovery failed: aq=%d\n",
4645 pf->hw.aq.asq_last_status);
4650 if (pf->hw.revision_id == 0 && pf->hw.func_caps.npar_enable) {
4651 pf->hw.func_caps.num_msix_vectors += 1;
4652 pf->hw.func_caps.num_tx_qp =
4653 min_t(int, pf->hw.func_caps.num_tx_qp,
4657 if (pf->hw.debug_mask & I40E_DEBUG_USER)
4658 dev_info(&pf->pdev->dev,
4659 "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",
4660 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
4661 pf->hw.func_caps.num_msix_vectors,
4662 pf->hw.func_caps.num_msix_vectors_vf,
4663 pf->hw.func_caps.fd_filters_guaranteed,
4664 pf->hw.func_caps.fd_filters_best_effort,
4665 pf->hw.func_caps.num_tx_qp,
4666 pf->hw.func_caps.num_vsis);
4668 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
4669 + pf->hw.func_caps.num_vfs)
4670 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
4671 dev_info(&pf->pdev->dev,
4672 "got num_vsis %d, setting num_vsis to %d\n",
4673 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
4674 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
4681 * i40e_fdir_setup - initialize the Flow Director resources
4682 * @pf: board private structure
4684 static void i40e_fdir_setup(struct i40e_pf *pf)
4686 struct i40e_vsi *vsi;
4687 bool new_vsi = false;
4690 if (!(pf->flags & (I40E_FLAG_FDIR_ENABLED |
4691 I40E_FLAG_FDIR_ATR_ENABLED)))
4694 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
4696 /* find existing or make new FDIR VSI */
4698 for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4699 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR)
4702 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR, pf->mac_seid, 0);
4704 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
4705 pf->flags &= ~I40E_FLAG_FDIR_ENABLED;
4710 WARN_ON(vsi->base_queue != I40E_FDIR_RING);
4711 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_rings);
4713 err = i40e_vsi_setup_tx_resources(vsi);
4715 err = i40e_vsi_setup_rx_resources(vsi);
4717 err = i40e_vsi_configure(vsi);
4718 if (!err && new_vsi) {
4719 char int_name[IFNAMSIZ + 9];
4720 snprintf(int_name, sizeof(int_name) - 1, "%s-fdir",
4721 dev_driver_string(&pf->pdev->dev));
4722 err = i40e_vsi_request_irq(vsi, int_name);
4725 err = i40e_up_complete(vsi);
4727 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4731 * i40e_fdir_teardown - release the Flow Director resources
4732 * @pf: board private structure
4734 static void i40e_fdir_teardown(struct i40e_pf *pf)
4738 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
4739 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
4740 i40e_vsi_release(pf->vsi[i]);
4747 * i40e_prep_for_reset - prep for the core to reset
4748 * @pf: board private structure
4750 * Close up the VFs and other things in prep for pf Reset.
4752 static int i40e_prep_for_reset(struct i40e_pf *pf)
4754 struct i40e_hw *hw = &pf->hw;
4758 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
4759 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
4762 dev_info(&pf->pdev->dev, "Tearing down internal switch for reset\n");
4764 if (i40e_check_asq_alive(hw))
4765 i40e_vc_notify_reset(pf);
4767 /* quiesce the VSIs and their queues that are not already DOWN */
4768 i40e_pf_quiesce_all_vsi(pf);
4770 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4772 pf->vsi[v]->seid = 0;
4775 i40e_shutdown_adminq(&pf->hw);
4777 /* call shutdown HMC */
4778 ret = i40e_shutdown_lan_hmc(hw);
4780 dev_info(&pf->pdev->dev, "shutdown_lan_hmc failed: %d\n", ret);
4781 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
4787 * i40e_reset_and_rebuild - reset and rebuid using a saved config
4788 * @pf: board private structure
4789 * @reinit: if the Main VSI needs to re-initialized.
4791 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
4793 struct i40e_driver_version dv;
4794 struct i40e_hw *hw = &pf->hw;
4798 /* Now we wait for GRST to settle out.
4799 * We don't have to delete the VEBs or VSIs from the hw switch
4800 * because the reset will make them disappear.
4802 ret = i40e_pf_reset(hw);
4804 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
4807 if (test_bit(__I40E_DOWN, &pf->state))
4808 goto end_core_reset;
4809 dev_info(&pf->pdev->dev, "Rebuilding internal switch\n");
4811 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
4812 ret = i40e_init_adminq(&pf->hw);
4814 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
4815 goto end_core_reset;
4818 ret = i40e_get_capabilities(pf);
4820 dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
4822 goto end_core_reset;
4825 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
4826 hw->func_caps.num_rx_qp,
4827 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
4829 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
4830 goto end_core_reset;
4832 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
4834 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
4835 goto end_core_reset;
4838 /* do basic switch setup */
4839 ret = i40e_setup_pf_switch(pf, reinit);
4841 goto end_core_reset;
4843 /* Rebuild the VSIs and VEBs that existed before reset.
4844 * They are still in our local switch element arrays, so only
4845 * need to rebuild the switch model in the HW.
4847 * If there were VEBs but the reconstitution failed, we'll try
4848 * try to recover minimal use by getting the basic PF VSI working.
4850 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
4851 dev_info(&pf->pdev->dev, "attempting to rebuild switch\n");
4852 /* find the one VEB connected to the MAC, and find orphans */
4853 for (v = 0; v < I40E_MAX_VEB; v++) {
4857 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
4858 pf->veb[v]->uplink_seid == 0) {
4859 ret = i40e_reconstitute_veb(pf->veb[v]);
4864 /* If Main VEB failed, we're in deep doodoo,
4865 * so give up rebuilding the switch and set up
4866 * for minimal rebuild of PF VSI.
4867 * If orphan failed, we'll report the error
4868 * but try to keep going.
4870 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
4871 dev_info(&pf->pdev->dev,
4872 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
4874 pf->vsi[pf->lan_vsi]->uplink_seid
4877 } else if (pf->veb[v]->uplink_seid == 0) {
4878 dev_info(&pf->pdev->dev,
4879 "rebuild of orphan VEB failed: %d\n",
4886 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
4887 dev_info(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
4888 /* no VEB, so rebuild only the Main VSI */
4889 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
4891 dev_info(&pf->pdev->dev,
4892 "rebuild of Main VSI failed: %d\n", ret);
4893 goto end_core_reset;
4897 /* reinit the misc interrupt */
4898 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4899 ret = i40e_setup_misc_vector(pf);
4901 /* restart the VSIs that were rebuilt and running before the reset */
4902 i40e_pf_unquiesce_all_vsi(pf);
4904 /* tell the firmware that we're starting */
4905 dv.major_version = DRV_VERSION_MAJOR;
4906 dv.minor_version = DRV_VERSION_MINOR;
4907 dv.build_version = DRV_VERSION_BUILD;
4908 dv.subbuild_version = 0;
4909 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
4911 dev_info(&pf->pdev->dev, "PF reset done\n");
4914 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
4918 * i40e_handle_reset_warning - prep for the pf to reset, reset and rebuild
4919 * @pf: board private structure
4921 * Close up the VFs and other things in prep for a Core Reset,
4922 * then get ready to rebuild the world.
4924 static void i40e_handle_reset_warning(struct i40e_pf *pf)
4928 ret = i40e_prep_for_reset(pf);
4930 i40e_reset_and_rebuild(pf, false);
4934 * i40e_handle_mdd_event
4935 * @pf: pointer to the pf structure
4937 * Called from the MDD irq handler to identify possibly malicious vfs
4939 static void i40e_handle_mdd_event(struct i40e_pf *pf)
4941 struct i40e_hw *hw = &pf->hw;
4942 bool mdd_detected = false;
4947 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
4950 /* find what triggered the MDD event */
4951 reg = rd32(hw, I40E_GL_MDET_TX);
4952 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
4953 u8 func = (reg & I40E_GL_MDET_TX_FUNCTION_MASK)
4954 >> I40E_GL_MDET_TX_FUNCTION_SHIFT;
4955 u8 event = (reg & I40E_GL_MDET_TX_EVENT_SHIFT)
4956 >> I40E_GL_MDET_TX_EVENT_SHIFT;
4957 u8 queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK)
4958 >> I40E_GL_MDET_TX_QUEUE_SHIFT;
4959 dev_info(&pf->pdev->dev,
4960 "Malicious Driver Detection TX event 0x%02x on q %d of function 0x%02x\n",
4961 event, queue, func);
4962 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
4963 mdd_detected = true;
4965 reg = rd32(hw, I40E_GL_MDET_RX);
4966 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
4967 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK)
4968 >> I40E_GL_MDET_RX_FUNCTION_SHIFT;
4969 u8 event = (reg & I40E_GL_MDET_RX_EVENT_SHIFT)
4970 >> I40E_GL_MDET_RX_EVENT_SHIFT;
4971 u8 queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK)
4972 >> I40E_GL_MDET_RX_QUEUE_SHIFT;
4973 dev_info(&pf->pdev->dev,
4974 "Malicious Driver Detection RX event 0x%02x on q %d of function 0x%02x\n",
4975 event, queue, func);
4976 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
4977 mdd_detected = true;
4980 /* see if one of the VFs needs its hand slapped */
4981 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
4983 reg = rd32(hw, I40E_VP_MDET_TX(i));
4984 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
4985 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
4986 vf->num_mdd_events++;
4987 dev_info(&pf->pdev->dev, "MDD TX event on VF %d\n", i);
4990 reg = rd32(hw, I40E_VP_MDET_RX(i));
4991 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
4992 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
4993 vf->num_mdd_events++;
4994 dev_info(&pf->pdev->dev, "MDD RX event on VF %d\n", i);
4997 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
4998 dev_info(&pf->pdev->dev,
4999 "Too many MDD events on VF %d, disabled\n", i);
5000 dev_info(&pf->pdev->dev,
5001 "Use PF Control I/F to re-enable the VF\n");
5002 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
5006 /* re-enable mdd interrupt cause */
5007 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
5008 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
5009 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
5010 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
5014 #ifdef CONFIG_I40E_VXLAN
5016 * i40e_sync_vxlan_filters_subtask - Sync the VSI filter list with HW
5017 * @pf: board private structure
5019 static void i40e_sync_vxlan_filters_subtask(struct i40e_pf *pf)
5021 const int vxlan_hdr_qwords = 4;
5022 struct i40e_hw *hw = &pf->hw;
5028 if (!(pf->flags & I40E_FLAG_VXLAN_FILTER_SYNC))
5031 pf->flags &= ~I40E_FLAG_VXLAN_FILTER_SYNC;
5033 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
5034 if (pf->pending_vxlan_bitmap & (1 << i)) {
5035 pf->pending_vxlan_bitmap &= ~(1 << i);
5036 port = pf->vxlan_ports[i];
5038 i40e_aq_add_udp_tunnel(hw, ntohs(port),
5040 I40E_AQC_TUNNEL_TYPE_VXLAN,
5041 &filter_index, NULL)
5042 : i40e_aq_del_udp_tunnel(hw, i, NULL);
5045 dev_info(&pf->pdev->dev, "Failed to execute AQ command for %s port %d with index %d\n",
5046 port ? "adding" : "deleting",
5047 ntohs(port), port ? i : i);
5049 pf->vxlan_ports[i] = 0;
5051 dev_info(&pf->pdev->dev, "%s port %d with AQ command with index %d\n",
5052 port ? "Added" : "Deleted",
5053 ntohs(port), port ? i : filter_index);
5061 * i40e_service_task - Run the driver's async subtasks
5062 * @work: pointer to work_struct containing our data
5064 static void i40e_service_task(struct work_struct *work)
5066 struct i40e_pf *pf = container_of(work,
5069 unsigned long start_time = jiffies;
5071 i40e_reset_subtask(pf);
5072 i40e_handle_mdd_event(pf);
5073 i40e_vc_process_vflr_event(pf);
5074 i40e_watchdog_subtask(pf);
5075 i40e_fdir_reinit_subtask(pf);
5076 i40e_check_hang_subtask(pf);
5077 i40e_sync_filters_subtask(pf);
5078 #ifdef CONFIG_I40E_VXLAN
5079 i40e_sync_vxlan_filters_subtask(pf);
5081 i40e_clean_adminq_subtask(pf);
5083 i40e_service_event_complete(pf);
5085 /* If the tasks have taken longer than one timer cycle or there
5086 * is more work to be done, reschedule the service task now
5087 * rather than wait for the timer to tick again.
5089 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
5090 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
5091 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
5092 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
5093 i40e_service_event_schedule(pf);
5097 * i40e_service_timer - timer callback
5098 * @data: pointer to PF struct
5100 static void i40e_service_timer(unsigned long data)
5102 struct i40e_pf *pf = (struct i40e_pf *)data;
5104 mod_timer(&pf->service_timer,
5105 round_jiffies(jiffies + pf->service_timer_period));
5106 i40e_service_event_schedule(pf);
5110 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
5111 * @vsi: the VSI being configured
5113 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
5115 struct i40e_pf *pf = vsi->back;
5117 switch (vsi->type) {
5119 vsi->alloc_queue_pairs = pf->num_lan_qps;
5120 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5121 I40E_REQ_DESCRIPTOR_MULTIPLE);
5122 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5123 vsi->num_q_vectors = pf->num_lan_msix;
5125 vsi->num_q_vectors = 1;
5130 vsi->alloc_queue_pairs = 1;
5131 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
5132 I40E_REQ_DESCRIPTOR_MULTIPLE);
5133 vsi->num_q_vectors = 1;
5136 case I40E_VSI_VMDQ2:
5137 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
5138 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5139 I40E_REQ_DESCRIPTOR_MULTIPLE);
5140 vsi->num_q_vectors = pf->num_vmdq_msix;
5143 case I40E_VSI_SRIOV:
5144 vsi->alloc_queue_pairs = pf->num_vf_qps;
5145 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
5146 I40E_REQ_DESCRIPTOR_MULTIPLE);
5158 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
5159 * @type: VSI pointer
5160 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
5162 * On error: returns error code (negative)
5163 * On success: returns 0
5165 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
5170 /* allocate memory for both Tx and Rx ring pointers */
5171 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
5172 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
5175 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
5177 if (alloc_qvectors) {
5178 /* allocate memory for q_vector pointers */
5179 size = sizeof(struct i40e_q_vectors *) * vsi->num_q_vectors;
5180 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
5181 if (!vsi->q_vectors) {
5189 kfree(vsi->tx_rings);
5194 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
5195 * @pf: board private structure
5196 * @type: type of VSI
5198 * On error: returns error code (negative)
5199 * On success: returns vsi index in PF (positive)
5201 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
5204 struct i40e_vsi *vsi;
5208 /* Need to protect the allocation of the VSIs at the PF level */
5209 mutex_lock(&pf->switch_mutex);
5211 /* VSI list may be fragmented if VSI creation/destruction has
5212 * been happening. We can afford to do a quick scan to look
5213 * for any free VSIs in the list.
5215 * find next empty vsi slot, looping back around if necessary
5218 while (i < pf->hw.func_caps.num_vsis && pf->vsi[i])
5220 if (i >= pf->hw.func_caps.num_vsis) {
5222 while (i < pf->next_vsi && pf->vsi[i])
5226 if (i < pf->hw.func_caps.num_vsis && !pf->vsi[i]) {
5227 vsi_idx = i; /* Found one! */
5230 goto unlock_pf; /* out of VSI slots! */
5234 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
5241 set_bit(__I40E_DOWN, &vsi->state);
5244 vsi->rx_itr_setting = pf->rx_itr_default;
5245 vsi->tx_itr_setting = pf->tx_itr_default;
5246 vsi->netdev_registered = false;
5247 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
5248 INIT_LIST_HEAD(&vsi->mac_filter_list);
5250 ret = i40e_set_num_rings_in_vsi(vsi);
5254 ret = i40e_vsi_alloc_arrays(vsi, true);
5258 /* Setup default MSIX irq handler for VSI */
5259 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
5261 pf->vsi[vsi_idx] = vsi;
5266 pf->next_vsi = i - 1;
5269 mutex_unlock(&pf->switch_mutex);
5274 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
5275 * @type: VSI pointer
5276 * @free_qvectors: a bool to specify if q_vectors need to be freed.
5278 * On error: returns error code (negative)
5279 * On success: returns 0
5281 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
5283 /* free the ring and vector containers */
5284 if (free_qvectors) {
5285 kfree(vsi->q_vectors);
5286 vsi->q_vectors = NULL;
5288 kfree(vsi->tx_rings);
5289 vsi->tx_rings = NULL;
5290 vsi->rx_rings = NULL;
5294 * i40e_vsi_clear - Deallocate the VSI provided
5295 * @vsi: the VSI being un-configured
5297 static int i40e_vsi_clear(struct i40e_vsi *vsi)
5308 mutex_lock(&pf->switch_mutex);
5309 if (!pf->vsi[vsi->idx]) {
5310 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
5311 vsi->idx, vsi->idx, vsi, vsi->type);
5315 if (pf->vsi[vsi->idx] != vsi) {
5316 dev_err(&pf->pdev->dev,
5317 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
5318 pf->vsi[vsi->idx]->idx,
5320 pf->vsi[vsi->idx]->type,
5321 vsi->idx, vsi, vsi->type);
5325 /* updates the pf for this cleared vsi */
5326 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
5327 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
5329 i40e_vsi_free_arrays(vsi, true);
5331 pf->vsi[vsi->idx] = NULL;
5332 if (vsi->idx < pf->next_vsi)
5333 pf->next_vsi = vsi->idx;
5336 mutex_unlock(&pf->switch_mutex);
5344 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
5345 * @vsi: the VSI being cleaned
5347 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
5351 if (vsi->tx_rings[0]) {
5352 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
5353 kfree_rcu(vsi->tx_rings[i], rcu);
5354 vsi->tx_rings[i] = NULL;
5355 vsi->rx_rings[i] = NULL;
5361 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
5362 * @vsi: the VSI being configured
5364 static int i40e_alloc_rings(struct i40e_vsi *vsi)
5366 struct i40e_pf *pf = vsi->back;
5369 /* Set basic values in the rings to be used later during open() */
5370 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
5371 struct i40e_ring *tx_ring;
5372 struct i40e_ring *rx_ring;
5374 /* allocate space for both Tx and Rx in one shot */
5375 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
5379 tx_ring->queue_index = i;
5380 tx_ring->reg_idx = vsi->base_queue + i;
5381 tx_ring->ring_active = false;
5383 tx_ring->netdev = vsi->netdev;
5384 tx_ring->dev = &pf->pdev->dev;
5385 tx_ring->count = vsi->num_desc;
5387 tx_ring->dcb_tc = 0;
5388 vsi->tx_rings[i] = tx_ring;
5390 rx_ring = &tx_ring[1];
5391 rx_ring->queue_index = i;
5392 rx_ring->reg_idx = vsi->base_queue + i;
5393 rx_ring->ring_active = false;
5395 rx_ring->netdev = vsi->netdev;
5396 rx_ring->dev = &pf->pdev->dev;
5397 rx_ring->count = vsi->num_desc;
5399 rx_ring->dcb_tc = 0;
5400 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
5401 set_ring_16byte_desc_enabled(rx_ring);
5403 clear_ring_16byte_desc_enabled(rx_ring);
5404 vsi->rx_rings[i] = rx_ring;
5410 i40e_vsi_clear_rings(vsi);
5415 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
5416 * @pf: board private structure
5417 * @vectors: the number of MSI-X vectors to request
5419 * Returns the number of vectors reserved, or error
5421 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
5425 pf->num_msix_entries = 0;
5426 while (vectors >= I40E_MIN_MSIX) {
5427 err = pci_enable_msix(pf->pdev, pf->msix_entries, vectors);
5430 pf->num_msix_entries = vectors;
5432 } else if (err < 0) {
5434 dev_info(&pf->pdev->dev,
5435 "MSI-X vector reservation failed: %d\n", err);
5439 /* err > 0 is the hint for retry */
5440 dev_info(&pf->pdev->dev,
5441 "MSI-X vectors wanted %d, retrying with %d\n",
5447 if (vectors > 0 && vectors < I40E_MIN_MSIX) {
5448 dev_info(&pf->pdev->dev,
5449 "Couldn't get enough vectors, only %d available\n",
5458 * i40e_init_msix - Setup the MSIX capability
5459 * @pf: board private structure
5461 * Work with the OS to set up the MSIX vectors needed.
5463 * Returns 0 on success, negative on failure
5465 static int i40e_init_msix(struct i40e_pf *pf)
5467 i40e_status err = 0;
5468 struct i40e_hw *hw = &pf->hw;
5472 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
5475 /* The number of vectors we'll request will be comprised of:
5476 * - Add 1 for "other" cause for Admin Queue events, etc.
5477 * - The number of LAN queue pairs
5478 * - Queues being used for RSS.
5479 * We don't need as many as max_rss_size vectors.
5480 * use rss_size instead in the calculation since that
5481 * is governed by number of cpus in the system.
5482 * - assumes symmetric Tx/Rx pairing
5483 * - The number of VMDq pairs
5484 * Once we count this up, try the request.
5486 * If we can't get what we want, we'll simplify to nearly nothing
5487 * and try again. If that still fails, we punt.
5489 pf->num_lan_msix = pf->num_lan_qps - (pf->rss_size_max - pf->rss_size);
5490 pf->num_vmdq_msix = pf->num_vmdq_qps;
5491 v_budget = 1 + pf->num_lan_msix;
5492 v_budget += (pf->num_vmdq_vsis * pf->num_vmdq_msix);
5493 if (pf->flags & I40E_FLAG_FDIR_ENABLED)
5496 /* Scale down if necessary, and the rings will share vectors */
5497 v_budget = min_t(int, v_budget, hw->func_caps.num_msix_vectors);
5499 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
5501 if (!pf->msix_entries)
5504 for (i = 0; i < v_budget; i++)
5505 pf->msix_entries[i].entry = i;
5506 vec = i40e_reserve_msix_vectors(pf, v_budget);
5507 if (vec < I40E_MIN_MSIX) {
5508 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
5509 kfree(pf->msix_entries);
5510 pf->msix_entries = NULL;
5513 } else if (vec == I40E_MIN_MSIX) {
5514 /* Adjust for minimal MSIX use */
5515 dev_info(&pf->pdev->dev, "Features disabled, not enough MSIX vectors\n");
5516 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
5517 pf->num_vmdq_vsis = 0;
5518 pf->num_vmdq_qps = 0;
5519 pf->num_vmdq_msix = 0;
5520 pf->num_lan_qps = 1;
5521 pf->num_lan_msix = 1;
5523 } else if (vec != v_budget) {
5524 /* Scale vector usage down */
5525 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
5526 vec--; /* reserve the misc vector */
5528 /* partition out the remaining vectors */
5531 pf->num_vmdq_vsis = 1;
5532 pf->num_lan_msix = 1;
5535 pf->num_vmdq_vsis = 1;
5536 pf->num_lan_msix = 2;
5539 pf->num_lan_msix = min_t(int, (vec / 2),
5541 pf->num_vmdq_vsis = min_t(int, (vec - pf->num_lan_msix),
5542 I40E_DEFAULT_NUM_VMDQ_VSI);
5551 * i40e_alloc_q_vector - Allocate memory for a single interrupt vector
5552 * @vsi: the VSI being configured
5553 * @v_idx: index of the vector in the vsi struct
5555 * We allocate one q_vector. If allocation fails we return -ENOMEM.
5557 static int i40e_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
5559 struct i40e_q_vector *q_vector;
5561 /* allocate q_vector */
5562 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
5566 q_vector->vsi = vsi;
5567 q_vector->v_idx = v_idx;
5568 cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
5570 netif_napi_add(vsi->netdev, &q_vector->napi,
5571 i40e_napi_poll, vsi->work_limit);
5573 q_vector->rx.latency_range = I40E_LOW_LATENCY;
5574 q_vector->tx.latency_range = I40E_LOW_LATENCY;
5576 /* tie q_vector and vsi together */
5577 vsi->q_vectors[v_idx] = q_vector;
5583 * i40e_alloc_q_vectors - Allocate memory for interrupt vectors
5584 * @vsi: the VSI being configured
5586 * We allocate one q_vector per queue interrupt. If allocation fails we
5589 static int i40e_alloc_q_vectors(struct i40e_vsi *vsi)
5591 struct i40e_pf *pf = vsi->back;
5592 int v_idx, num_q_vectors;
5595 /* if not MSIX, give the one vector only to the LAN VSI */
5596 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5597 num_q_vectors = vsi->num_q_vectors;
5598 else if (vsi == pf->vsi[pf->lan_vsi])
5603 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
5604 err = i40e_alloc_q_vector(vsi, v_idx);
5613 i40e_free_q_vector(vsi, v_idx);
5619 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
5620 * @pf: board private structure to initialize
5622 static void i40e_init_interrupt_scheme(struct i40e_pf *pf)
5626 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
5627 err = i40e_init_msix(pf);
5629 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
5630 I40E_FLAG_RSS_ENABLED |
5631 I40E_FLAG_DCB_ENABLED |
5632 I40E_FLAG_SRIOV_ENABLED |
5633 I40E_FLAG_FDIR_ENABLED |
5634 I40E_FLAG_FDIR_ATR_ENABLED |
5635 I40E_FLAG_VMDQ_ENABLED);
5637 /* rework the queue expectations without MSIX */
5638 i40e_determine_queue_usage(pf);
5642 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
5643 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
5644 dev_info(&pf->pdev->dev, "MSIX not available, trying MSI\n");
5645 err = pci_enable_msi(pf->pdev);
5647 dev_info(&pf->pdev->dev, "MSI init failed - %d\n", err);
5648 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
5652 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
5653 dev_info(&pf->pdev->dev, "MSIX and MSI not available, falling back to Legacy IRQ\n");
5655 /* track first vector for misc interrupts */
5656 err = i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT-1);
5660 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
5661 * @pf: board private structure
5663 * This sets up the handler for MSIX 0, which is used to manage the
5664 * non-queue interrupts, e.g. AdminQ and errors. This is not used
5665 * when in MSI or Legacy interrupt mode.
5667 static int i40e_setup_misc_vector(struct i40e_pf *pf)
5669 struct i40e_hw *hw = &pf->hw;
5672 /* Only request the irq if this is the first time through, and
5673 * not when we're rebuilding after a Reset
5675 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
5676 err = request_irq(pf->msix_entries[0].vector,
5677 i40e_intr, 0, pf->misc_int_name, pf);
5679 dev_info(&pf->pdev->dev,
5680 "request_irq for msix_misc failed: %d\n", err);
5685 i40e_enable_misc_int_causes(hw);
5687 /* associate no queues to the misc vector */
5688 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
5689 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
5693 i40e_irq_dynamic_enable_icr0(pf);
5699 * i40e_config_rss - Prepare for RSS if used
5700 * @pf: board private structure
5702 static int i40e_config_rss(struct i40e_pf *pf)
5704 /* Set of random keys generated using kernel random number generator */
5705 static const u32 seed[I40E_PFQF_HKEY_MAX_INDEX + 1] = {0x41b01687,
5706 0x183cfd8c, 0xce880440, 0x580cbc3c, 0x35897377,
5707 0x328b25e1, 0x4fa98922, 0xb7d90c14, 0xd5bad70d,
5708 0xcd15a2c1, 0xe8580225, 0x4a1e9d11, 0xfe5731be};
5709 struct i40e_hw *hw = &pf->hw;
5714 /* Fill out hash function seed */
5715 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
5716 wr32(hw, I40E_PFQF_HKEY(i), seed[i]);
5718 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
5719 hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
5720 ((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
5721 hena |= I40E_DEFAULT_RSS_HENA;
5722 wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
5723 wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
5725 /* Populate the LUT with max no. of queues in round robin fashion */
5726 for (i = 0, j = 0; i < pf->hw.func_caps.rss_table_size; i++, j++) {
5728 /* The assumption is that lan qp count will be the highest
5729 * qp count for any PF VSI that needs RSS.
5730 * If multiple VSIs need RSS support, all the qp counts
5731 * for those VSIs should be a power of 2 for RSS to work.
5732 * If LAN VSI is the only consumer for RSS then this requirement
5735 if (j == pf->rss_size)
5737 /* lut = 4-byte sliding window of 4 lut entries */
5738 lut = (lut << 8) | (j &
5739 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
5740 /* On i = 3, we have 4 entries in lut; write to the register */
5742 wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
5750 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
5751 * @pf: board private structure
5752 * @queue_count: the requested queue count for rss.
5754 * returns 0 if rss is not enabled, if enabled returns the final rss queue
5755 * count which may be different from the requested queue count.
5757 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
5759 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
5762 queue_count = min_t(int, queue_count, pf->rss_size_max);
5763 queue_count = rounddown_pow_of_two(queue_count);
5765 if (queue_count != pf->rss_size) {
5766 if (pf->queues_left < (queue_count - pf->rss_size)) {
5767 dev_info(&pf->pdev->dev,
5768 "Not enough queues to do RSS on %d queues: remaining queues %d\n",
5769 queue_count, pf->queues_left);
5770 return pf->rss_size;
5772 i40e_prep_for_reset(pf);
5774 pf->num_lan_qps += (queue_count - pf->rss_size);
5775 pf->queues_left -= (queue_count - pf->rss_size);
5776 pf->rss_size = queue_count;
5778 i40e_reset_and_rebuild(pf, true);
5779 i40e_config_rss(pf);
5781 dev_info(&pf->pdev->dev, "RSS count: %d\n", pf->rss_size);
5782 return pf->rss_size;
5786 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
5787 * @pf: board private structure to initialize
5789 * i40e_sw_init initializes the Adapter private data structure.
5790 * Fields are initialized based on PCI device information and
5791 * OS network device settings (MTU size).
5793 static int i40e_sw_init(struct i40e_pf *pf)
5798 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
5799 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
5800 pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
5801 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
5802 if (I40E_DEBUG_USER & debug)
5803 pf->hw.debug_mask = debug;
5804 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
5805 I40E_DEFAULT_MSG_ENABLE);
5808 /* Set default capability flags */
5809 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
5810 I40E_FLAG_MSI_ENABLED |
5811 I40E_FLAG_MSIX_ENABLED |
5812 I40E_FLAG_RX_PS_ENABLED |
5813 I40E_FLAG_RX_1BUF_ENABLED;
5815 /* Depending on PF configurations, it is possible that the RSS
5816 * maximum might end up larger than the available queues
5818 pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
5819 pf->rss_size_max = min_t(int, pf->rss_size_max,
5820 pf->hw.func_caps.num_tx_qp);
5821 if (pf->hw.func_caps.rss) {
5822 pf->flags |= I40E_FLAG_RSS_ENABLED;
5823 pf->rss_size = min_t(int, pf->rss_size_max, num_online_cpus());
5828 if (pf->hw.func_caps.dcb)
5829 pf->num_tc_qps = I40E_DEFAULT_QUEUES_PER_TC;
5833 if (pf->hw.func_caps.fd) {
5834 /* FW/NVM is not yet fixed in this regard */
5835 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
5836 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
5837 pf->flags |= I40E_FLAG_FDIR_ATR_ENABLED;
5838 dev_info(&pf->pdev->dev,
5839 "Flow Director ATR mode Enabled\n");
5840 pf->flags |= I40E_FLAG_FDIR_ENABLED;
5841 dev_info(&pf->pdev->dev,
5842 "Flow Director Side Band mode Enabled\n");
5843 pf->fdir_pf_filter_count =
5844 pf->hw.func_caps.fd_filters_guaranteed;
5847 pf->fdir_pf_filter_count = 0;
5850 if (pf->hw.func_caps.vmdq) {
5851 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
5852 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
5853 pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
5856 /* MFP mode enabled */
5857 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
5858 pf->flags |= I40E_FLAG_MFP_ENABLED;
5859 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
5862 #ifdef CONFIG_PCI_IOV
5863 if (pf->hw.func_caps.num_vfs) {
5864 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
5865 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
5866 pf->num_req_vfs = min_t(int,
5867 pf->hw.func_caps.num_vfs,
5869 dev_info(&pf->pdev->dev,
5870 "Number of VFs being requested for PF[%d] = %d\n",
5871 pf->hw.pf_id, pf->num_req_vfs);
5873 #endif /* CONFIG_PCI_IOV */
5874 pf->eeprom_version = 0xDEAD;
5875 pf->lan_veb = I40E_NO_VEB;
5876 pf->lan_vsi = I40E_NO_VSI;
5878 /* set up queue assignment tracking */
5879 size = sizeof(struct i40e_lump_tracking)
5880 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
5881 pf->qp_pile = kzalloc(size, GFP_KERNEL);
5886 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
5887 pf->qp_pile->search_hint = 0;
5889 /* set up vector assignment tracking */
5890 size = sizeof(struct i40e_lump_tracking)
5891 + (sizeof(u16) * pf->hw.func_caps.num_msix_vectors);
5892 pf->irq_pile = kzalloc(size, GFP_KERNEL);
5893 if (!pf->irq_pile) {
5898 pf->irq_pile->num_entries = pf->hw.func_caps.num_msix_vectors;
5899 pf->irq_pile->search_hint = 0;
5901 mutex_init(&pf->switch_mutex);
5908 * i40e_set_features - set the netdev feature flags
5909 * @netdev: ptr to the netdev being adjusted
5910 * @features: the feature set that the stack is suggesting
5912 static int i40e_set_features(struct net_device *netdev,
5913 netdev_features_t features)
5915 struct i40e_netdev_priv *np = netdev_priv(netdev);
5916 struct i40e_vsi *vsi = np->vsi;
5918 if (features & NETIF_F_HW_VLAN_CTAG_RX)
5919 i40e_vlan_stripping_enable(vsi);
5921 i40e_vlan_stripping_disable(vsi);
5926 #ifdef CONFIG_I40E_VXLAN
5928 * i40e_get_vxlan_port_idx - Lookup a possibly offloaded for Rx UDP port
5929 * @pf: board private structure
5930 * @port: The UDP port to look up
5932 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
5934 static u8 i40e_get_vxlan_port_idx(struct i40e_pf *pf, __be16 port)
5938 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
5939 if (pf->vxlan_ports[i] == port)
5947 * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
5948 * @netdev: This physical port's netdev
5949 * @sa_family: Socket Family that VXLAN is notifying us about
5950 * @port: New UDP port number that VXLAN started listening to
5952 static void i40e_add_vxlan_port(struct net_device *netdev,
5953 sa_family_t sa_family, __be16 port)
5955 struct i40e_netdev_priv *np = netdev_priv(netdev);
5956 struct i40e_vsi *vsi = np->vsi;
5957 struct i40e_pf *pf = vsi->back;
5961 if (sa_family == AF_INET6)
5964 idx = i40e_get_vxlan_port_idx(pf, port);
5966 /* Check if port already exists */
5967 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
5968 netdev_info(netdev, "Port %d already offloaded\n", ntohs(port));
5972 /* Now check if there is space to add the new port */
5973 next_idx = i40e_get_vxlan_port_idx(pf, 0);
5975 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
5976 netdev_info(netdev, "Maximum number of UDP ports reached, not adding port %d\n",
5981 /* New port: add it and mark its index in the bitmap */
5982 pf->vxlan_ports[next_idx] = port;
5983 pf->pending_vxlan_bitmap |= (1 << next_idx);
5985 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
5989 * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
5990 * @netdev: This physical port's netdev
5991 * @sa_family: Socket Family that VXLAN is notifying us about
5992 * @port: UDP port number that VXLAN stopped listening to
5994 static void i40e_del_vxlan_port(struct net_device *netdev,
5995 sa_family_t sa_family, __be16 port)
5997 struct i40e_netdev_priv *np = netdev_priv(netdev);
5998 struct i40e_vsi *vsi = np->vsi;
5999 struct i40e_pf *pf = vsi->back;
6002 if (sa_family == AF_INET6)
6005 idx = i40e_get_vxlan_port_idx(pf, port);
6007 /* Check if port already exists */
6008 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
6009 /* if port exists, set it to 0 (mark for deletion)
6010 * and make it pending
6012 pf->vxlan_ports[idx] = 0;
6014 pf->pending_vxlan_bitmap |= (1 << idx);
6016 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
6018 netdev_warn(netdev, "Port %d was not found, not deleting\n",
6024 static const struct net_device_ops i40e_netdev_ops = {
6025 .ndo_open = i40e_open,
6026 .ndo_stop = i40e_close,
6027 .ndo_start_xmit = i40e_lan_xmit_frame,
6028 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
6029 .ndo_set_rx_mode = i40e_set_rx_mode,
6030 .ndo_validate_addr = eth_validate_addr,
6031 .ndo_set_mac_address = i40e_set_mac,
6032 .ndo_change_mtu = i40e_change_mtu,
6033 .ndo_tx_timeout = i40e_tx_timeout,
6034 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
6035 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
6036 #ifdef CONFIG_NET_POLL_CONTROLLER
6037 .ndo_poll_controller = i40e_netpoll,
6039 .ndo_setup_tc = i40e_setup_tc,
6040 .ndo_set_features = i40e_set_features,
6041 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
6042 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
6043 .ndo_set_vf_tx_rate = i40e_ndo_set_vf_bw,
6044 .ndo_get_vf_config = i40e_ndo_get_vf_config,
6045 #ifdef CONFIG_I40E_VXLAN
6046 .ndo_add_vxlan_port = i40e_add_vxlan_port,
6047 .ndo_del_vxlan_port = i40e_del_vxlan_port,
6052 * i40e_config_netdev - Setup the netdev flags
6053 * @vsi: the VSI being configured
6055 * Returns 0 on success, negative value on failure
6057 static int i40e_config_netdev(struct i40e_vsi *vsi)
6059 u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
6060 struct i40e_pf *pf = vsi->back;
6061 struct i40e_hw *hw = &pf->hw;
6062 struct i40e_netdev_priv *np;
6063 struct net_device *netdev;
6064 u8 mac_addr[ETH_ALEN];
6067 etherdev_size = sizeof(struct i40e_netdev_priv);
6068 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
6072 vsi->netdev = netdev;
6073 np = netdev_priv(netdev);
6076 netdev->hw_enc_features = NETIF_F_IP_CSUM |
6077 NETIF_F_GSO_UDP_TUNNEL |
6081 netdev->features = NETIF_F_SG |
6085 NETIF_F_GSO_UDP_TUNNEL |
6086 NETIF_F_HW_VLAN_CTAG_TX |
6087 NETIF_F_HW_VLAN_CTAG_RX |
6088 NETIF_F_HW_VLAN_CTAG_FILTER |
6096 /* copy netdev features into list of user selectable features */
6097 netdev->hw_features |= netdev->features;
6099 if (vsi->type == I40E_VSI_MAIN) {
6100 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
6101 memcpy(mac_addr, hw->mac.perm_addr, ETH_ALEN);
6103 /* relate the VSI_VMDQ name to the VSI_MAIN name */
6104 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
6105 pf->vsi[pf->lan_vsi]->netdev->name);
6106 random_ether_addr(mac_addr);
6107 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
6109 i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
6111 memcpy(netdev->dev_addr, mac_addr, ETH_ALEN);
6112 memcpy(netdev->perm_addr, mac_addr, ETH_ALEN);
6113 /* vlan gets same features (except vlan offload)
6114 * after any tweaks for specific VSI types
6116 netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
6117 NETIF_F_HW_VLAN_CTAG_RX |
6118 NETIF_F_HW_VLAN_CTAG_FILTER);
6119 netdev->priv_flags |= IFF_UNICAST_FLT;
6120 netdev->priv_flags |= IFF_SUPP_NOFCS;
6121 /* Setup netdev TC information */
6122 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
6124 netdev->netdev_ops = &i40e_netdev_ops;
6125 netdev->watchdog_timeo = 5 * HZ;
6126 i40e_set_ethtool_ops(netdev);
6132 * i40e_vsi_delete - Delete a VSI from the switch
6133 * @vsi: the VSI being removed
6135 * Returns 0 on success, negative value on failure
6137 static void i40e_vsi_delete(struct i40e_vsi *vsi)
6139 /* remove default VSI is not allowed */
6140 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
6143 /* there is no HW VSI for FDIR */
6144 if (vsi->type == I40E_VSI_FDIR)
6147 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
6152 * i40e_add_vsi - Add a VSI to the switch
6153 * @vsi: the VSI being configured
6155 * This initializes a VSI context depending on the VSI type to be added and
6156 * passes it down to the add_vsi aq command.
6158 static int i40e_add_vsi(struct i40e_vsi *vsi)
6161 struct i40e_mac_filter *f, *ftmp;
6162 struct i40e_pf *pf = vsi->back;
6163 struct i40e_hw *hw = &pf->hw;
6164 struct i40e_vsi_context ctxt;
6165 u8 enabled_tc = 0x1; /* TC0 enabled */
6168 memset(&ctxt, 0, sizeof(ctxt));
6169 switch (vsi->type) {
6171 /* The PF's main VSI is already setup as part of the
6172 * device initialization, so we'll not bother with
6173 * the add_vsi call, but we will retrieve the current
6176 ctxt.seid = pf->main_vsi_seid;
6177 ctxt.pf_num = pf->hw.pf_id;
6179 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6180 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6182 dev_info(&pf->pdev->dev,
6183 "couldn't get pf vsi config, err %d, aq_err %d\n",
6184 ret, pf->hw.aq.asq_last_status);
6187 memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
6188 vsi->info.valid_sections = 0;
6190 vsi->seid = ctxt.seid;
6191 vsi->id = ctxt.vsi_number;
6193 enabled_tc = i40e_pf_get_tc_map(pf);
6195 /* MFP mode setup queue map and update VSI */
6196 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
6197 memset(&ctxt, 0, sizeof(ctxt));
6198 ctxt.seid = pf->main_vsi_seid;
6199 ctxt.pf_num = pf->hw.pf_id;
6201 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
6202 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
6204 dev_info(&pf->pdev->dev,
6205 "update vsi failed, aq_err=%d\n",
6206 pf->hw.aq.asq_last_status);
6210 /* update the local VSI info queue map */
6211 i40e_vsi_update_queue_map(vsi, &ctxt);
6212 vsi->info.valid_sections = 0;
6214 /* Default/Main VSI is only enabled for TC0
6215 * reconfigure it to enable all TCs that are
6216 * available on the port in SFP mode.
6218 ret = i40e_vsi_config_tc(vsi, enabled_tc);
6220 dev_info(&pf->pdev->dev,
6221 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
6223 pf->hw.aq.asq_last_status);
6230 /* no queue mapping or actual HW VSI needed */
6231 vsi->info.valid_sections = 0;
6234 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
6238 case I40E_VSI_VMDQ2:
6239 ctxt.pf_num = hw->pf_id;
6241 ctxt.uplink_seid = vsi->uplink_seid;
6242 ctxt.connection_type = 0x1; /* regular data port */
6243 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
6245 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6247 /* This VSI is connected to VEB so the switch_id
6248 * should be set to zero by default.
6250 ctxt.info.switch_id = 0;
6251 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
6252 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6254 /* Setup the VSI tx/rx queue map for TC0 only for now */
6255 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
6258 case I40E_VSI_SRIOV:
6259 ctxt.pf_num = hw->pf_id;
6260 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
6261 ctxt.uplink_seid = vsi->uplink_seid;
6262 ctxt.connection_type = 0x1; /* regular data port */
6263 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
6265 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6267 /* This VSI is connected to VEB so the switch_id
6268 * should be set to zero by default.
6270 ctxt.info.switch_id = cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6272 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
6273 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
6274 /* Setup the VSI tx/rx queue map for TC0 only for now */
6275 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
6282 if (vsi->type != I40E_VSI_MAIN) {
6283 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
6285 dev_info(&vsi->back->pdev->dev,
6286 "add vsi failed, aq_err=%d\n",
6287 vsi->back->hw.aq.asq_last_status);
6291 memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
6292 vsi->info.valid_sections = 0;
6293 vsi->seid = ctxt.seid;
6294 vsi->id = ctxt.vsi_number;
6297 /* If macvlan filters already exist, force them to get loaded */
6298 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
6303 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
6304 pf->flags |= I40E_FLAG_FILTER_SYNC;
6307 /* Update VSI BW information */
6308 ret = i40e_vsi_get_bw_info(vsi);
6310 dev_info(&pf->pdev->dev,
6311 "couldn't get vsi bw info, err %d, aq_err %d\n",
6312 ret, pf->hw.aq.asq_last_status);
6313 /* VSI is already added so not tearing that up */
6322 * i40e_vsi_release - Delete a VSI and free its resources
6323 * @vsi: the VSI being removed
6325 * Returns 0 on success or < 0 on error
6327 int i40e_vsi_release(struct i40e_vsi *vsi)
6329 struct i40e_mac_filter *f, *ftmp;
6330 struct i40e_veb *veb = NULL;
6337 /* release of a VEB-owner or last VSI is not allowed */
6338 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
6339 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
6340 vsi->seid, vsi->uplink_seid);
6343 if (vsi == pf->vsi[pf->lan_vsi] &&
6344 !test_bit(__I40E_DOWN, &pf->state)) {
6345 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
6349 uplink_seid = vsi->uplink_seid;
6350 if (vsi->type != I40E_VSI_SRIOV) {
6351 if (vsi->netdev_registered) {
6352 vsi->netdev_registered = false;
6354 /* results in a call to i40e_close() */
6355 unregister_netdev(vsi->netdev);
6356 free_netdev(vsi->netdev);
6360 if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
6362 i40e_vsi_free_irq(vsi);
6363 i40e_vsi_free_tx_resources(vsi);
6364 i40e_vsi_free_rx_resources(vsi);
6366 i40e_vsi_disable_irq(vsi);
6369 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
6370 i40e_del_filter(vsi, f->macaddr, f->vlan,
6371 f->is_vf, f->is_netdev);
6372 i40e_sync_vsi_filters(vsi);
6374 i40e_vsi_delete(vsi);
6375 i40e_vsi_free_q_vectors(vsi);
6376 i40e_vsi_clear_rings(vsi);
6377 i40e_vsi_clear(vsi);
6379 /* If this was the last thing on the VEB, except for the
6380 * controlling VSI, remove the VEB, which puts the controlling
6381 * VSI onto the next level down in the switch.
6383 * Well, okay, there's one more exception here: don't remove
6384 * the orphan VEBs yet. We'll wait for an explicit remove request
6385 * from up the network stack.
6387 for (n = 0, i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6389 pf->vsi[i]->uplink_seid == uplink_seid &&
6390 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
6391 n++; /* count the VSIs */
6394 for (i = 0; i < I40E_MAX_VEB; i++) {
6397 if (pf->veb[i]->uplink_seid == uplink_seid)
6398 n++; /* count the VEBs */
6399 if (pf->veb[i]->seid == uplink_seid)
6402 if (n == 0 && veb && veb->uplink_seid != 0)
6403 i40e_veb_release(veb);
6409 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
6410 * @vsi: ptr to the VSI
6412 * This should only be called after i40e_vsi_mem_alloc() which allocates the
6413 * corresponding SW VSI structure and initializes num_queue_pairs for the
6414 * newly allocated VSI.
6416 * Returns 0 on success or negative on failure
6418 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
6421 struct i40e_pf *pf = vsi->back;
6423 if (vsi->q_vectors[0]) {
6424 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
6429 if (vsi->base_vector) {
6430 dev_info(&pf->pdev->dev,
6431 "VSI %d has non-zero base vector %d\n",
6432 vsi->seid, vsi->base_vector);
6436 ret = i40e_alloc_q_vectors(vsi);
6438 dev_info(&pf->pdev->dev,
6439 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
6440 vsi->num_q_vectors, vsi->seid, ret);
6441 vsi->num_q_vectors = 0;
6442 goto vector_setup_out;
6445 if (vsi->num_q_vectors)
6446 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
6447 vsi->num_q_vectors, vsi->idx);
6448 if (vsi->base_vector < 0) {
6449 dev_info(&pf->pdev->dev,
6450 "failed to get q tracking for VSI %d, err=%d\n",
6451 vsi->seid, vsi->base_vector);
6452 i40e_vsi_free_q_vectors(vsi);
6454 goto vector_setup_out;
6462 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
6463 * @vsi: pointer to the vsi.
6465 * This re-allocates a vsi's queue resources.
6467 * Returns pointer to the successfully allocated and configured VSI sw struct
6468 * on success, otherwise returns NULL on failure.
6470 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
6472 struct i40e_pf *pf = vsi->back;
6476 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
6477 i40e_vsi_clear_rings(vsi);
6479 i40e_vsi_free_arrays(vsi, false);
6480 i40e_set_num_rings_in_vsi(vsi);
6481 ret = i40e_vsi_alloc_arrays(vsi, false);
6485 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
6487 dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
6491 vsi->base_queue = ret;
6493 /* Update the FW view of the VSI. Force a reset of TC and queue
6494 * layout configurations.
6496 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
6497 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
6498 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
6499 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
6501 /* assign it some queues */
6502 ret = i40e_alloc_rings(vsi);
6506 /* map all of the rings to the q_vectors */
6507 i40e_vsi_map_rings_to_vectors(vsi);
6511 i40e_vsi_free_q_vectors(vsi);
6512 if (vsi->netdev_registered) {
6513 vsi->netdev_registered = false;
6514 unregister_netdev(vsi->netdev);
6515 free_netdev(vsi->netdev);
6518 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
6520 i40e_vsi_clear(vsi);
6525 * i40e_vsi_setup - Set up a VSI by a given type
6526 * @pf: board private structure
6528 * @uplink_seid: the switch element to link to
6529 * @param1: usage depends upon VSI type. For VF types, indicates VF id
6531 * This allocates the sw VSI structure and its queue resources, then add a VSI
6532 * to the identified VEB.
6534 * Returns pointer to the successfully allocated and configure VSI sw struct on
6535 * success, otherwise returns NULL on failure.
6537 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
6538 u16 uplink_seid, u32 param1)
6540 struct i40e_vsi *vsi = NULL;
6541 struct i40e_veb *veb = NULL;
6545 /* The requested uplink_seid must be either
6546 * - the PF's port seid
6547 * no VEB is needed because this is the PF
6548 * or this is a Flow Director special case VSI
6549 * - seid of an existing VEB
6550 * - seid of a VSI that owns an existing VEB
6551 * - seid of a VSI that doesn't own a VEB
6552 * a new VEB is created and the VSI becomes the owner
6553 * - seid of the PF VSI, which is what creates the first VEB
6554 * this is a special case of the previous
6556 * Find which uplink_seid we were given and create a new VEB if needed
6558 for (i = 0; i < I40E_MAX_VEB; i++) {
6559 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
6565 if (!veb && uplink_seid != pf->mac_seid) {
6567 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6568 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
6574 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
6579 if (vsi->uplink_seid == pf->mac_seid)
6580 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
6581 vsi->tc_config.enabled_tc);
6582 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
6583 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
6584 vsi->tc_config.enabled_tc);
6586 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
6587 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
6591 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
6595 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6596 uplink_seid = veb->seid;
6599 /* get vsi sw struct */
6600 v_idx = i40e_vsi_mem_alloc(pf, type);
6603 vsi = pf->vsi[v_idx];
6605 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
6607 if (type == I40E_VSI_MAIN)
6608 pf->lan_vsi = v_idx;
6609 else if (type == I40E_VSI_SRIOV)
6610 vsi->vf_id = param1;
6611 /* assign it some queues */
6612 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
6614 dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
6618 vsi->base_queue = ret;
6620 /* get a VSI from the hardware */
6621 vsi->uplink_seid = uplink_seid;
6622 ret = i40e_add_vsi(vsi);
6626 switch (vsi->type) {
6627 /* setup the netdev if needed */
6629 case I40E_VSI_VMDQ2:
6630 ret = i40e_config_netdev(vsi);
6633 ret = register_netdev(vsi->netdev);
6636 vsi->netdev_registered = true;
6637 netif_carrier_off(vsi->netdev);
6641 /* set up vectors and rings if needed */
6642 ret = i40e_vsi_setup_vectors(vsi);
6646 ret = i40e_alloc_rings(vsi);
6650 /* map all of the rings to the q_vectors */
6651 i40e_vsi_map_rings_to_vectors(vsi);
6653 i40e_vsi_reset_stats(vsi);
6657 /* no netdev or rings for the other VSI types */
6664 i40e_vsi_free_q_vectors(vsi);
6666 if (vsi->netdev_registered) {
6667 vsi->netdev_registered = false;
6668 unregister_netdev(vsi->netdev);
6669 free_netdev(vsi->netdev);
6673 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
6675 i40e_vsi_clear(vsi);
6681 * i40e_veb_get_bw_info - Query VEB BW information
6682 * @veb: the veb to query
6684 * Query the Tx scheduler BW configuration data for given VEB
6686 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
6688 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
6689 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
6690 struct i40e_pf *pf = veb->pf;
6691 struct i40e_hw *hw = &pf->hw;
6696 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
6699 dev_info(&pf->pdev->dev,
6700 "query veb bw config failed, aq_err=%d\n",
6701 hw->aq.asq_last_status);
6705 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
6708 dev_info(&pf->pdev->dev,
6709 "query veb bw ets config failed, aq_err=%d\n",
6710 hw->aq.asq_last_status);
6714 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
6715 veb->bw_max_quanta = ets_data.tc_bw_max;
6716 veb->is_abs_credits = bw_data.absolute_credits_enable;
6717 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
6718 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
6719 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6720 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
6721 veb->bw_tc_limit_credits[i] =
6722 le16_to_cpu(bw_data.tc_bw_limits[i]);
6723 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
6731 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
6732 * @pf: board private structure
6734 * On error: returns error code (negative)
6735 * On success: returns vsi index in PF (positive)
6737 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
6740 struct i40e_veb *veb;
6743 /* Need to protect the allocation of switch elements at the PF level */
6744 mutex_lock(&pf->switch_mutex);
6746 /* VEB list may be fragmented if VEB creation/destruction has
6747 * been happening. We can afford to do a quick scan to look
6748 * for any free slots in the list.
6750 * find next empty veb slot, looping back around if necessary
6753 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
6755 if (i >= I40E_MAX_VEB) {
6757 goto err_alloc_veb; /* out of VEB slots! */
6760 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
6767 veb->enabled_tc = 1;
6772 mutex_unlock(&pf->switch_mutex);
6777 * i40e_switch_branch_release - Delete a branch of the switch tree
6778 * @branch: where to start deleting
6780 * This uses recursion to find the tips of the branch to be
6781 * removed, deleting until we get back to and can delete this VEB.
6783 static void i40e_switch_branch_release(struct i40e_veb *branch)
6785 struct i40e_pf *pf = branch->pf;
6786 u16 branch_seid = branch->seid;
6787 u16 veb_idx = branch->idx;
6790 /* release any VEBs on this VEB - RECURSION */
6791 for (i = 0; i < I40E_MAX_VEB; i++) {
6794 if (pf->veb[i]->uplink_seid == branch->seid)
6795 i40e_switch_branch_release(pf->veb[i]);
6798 /* Release the VSIs on this VEB, but not the owner VSI.
6800 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
6801 * the VEB itself, so don't use (*branch) after this loop.
6803 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6806 if (pf->vsi[i]->uplink_seid == branch_seid &&
6807 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
6808 i40e_vsi_release(pf->vsi[i]);
6812 /* There's one corner case where the VEB might not have been
6813 * removed, so double check it here and remove it if needed.
6814 * This case happens if the veb was created from the debugfs
6815 * commands and no VSIs were added to it.
6817 if (pf->veb[veb_idx])
6818 i40e_veb_release(pf->veb[veb_idx]);
6822 * i40e_veb_clear - remove veb struct
6823 * @veb: the veb to remove
6825 static void i40e_veb_clear(struct i40e_veb *veb)
6831 struct i40e_pf *pf = veb->pf;
6833 mutex_lock(&pf->switch_mutex);
6834 if (pf->veb[veb->idx] == veb)
6835 pf->veb[veb->idx] = NULL;
6836 mutex_unlock(&pf->switch_mutex);
6843 * i40e_veb_release - Delete a VEB and free its resources
6844 * @veb: the VEB being removed
6846 void i40e_veb_release(struct i40e_veb *veb)
6848 struct i40e_vsi *vsi = NULL;
6854 /* find the remaining VSI and check for extras */
6855 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6856 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
6862 dev_info(&pf->pdev->dev,
6863 "can't remove VEB %d with %d VSIs left\n",
6868 /* move the remaining VSI to uplink veb */
6869 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
6870 if (veb->uplink_seid) {
6871 vsi->uplink_seid = veb->uplink_seid;
6872 if (veb->uplink_seid == pf->mac_seid)
6873 vsi->veb_idx = I40E_NO_VEB;
6875 vsi->veb_idx = veb->veb_idx;
6878 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6879 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
6882 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
6883 i40e_veb_clear(veb);
6889 * i40e_add_veb - create the VEB in the switch
6890 * @veb: the VEB to be instantiated
6891 * @vsi: the controlling VSI
6893 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
6895 bool is_default = false;
6896 bool is_cloud = false;
6899 /* get a VEB from the hardware */
6900 ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
6901 veb->enabled_tc, is_default,
6902 is_cloud, &veb->seid, NULL);
6904 dev_info(&veb->pf->pdev->dev,
6905 "couldn't add VEB, err %d, aq_err %d\n",
6906 ret, veb->pf->hw.aq.asq_last_status);
6910 /* get statistics counter */
6911 ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
6912 &veb->stats_idx, NULL, NULL, NULL);
6914 dev_info(&veb->pf->pdev->dev,
6915 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
6916 ret, veb->pf->hw.aq.asq_last_status);
6919 ret = i40e_veb_get_bw_info(veb);
6921 dev_info(&veb->pf->pdev->dev,
6922 "couldn't get VEB bw info, err %d, aq_err %d\n",
6923 ret, veb->pf->hw.aq.asq_last_status);
6924 i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
6928 vsi->uplink_seid = veb->seid;
6929 vsi->veb_idx = veb->idx;
6930 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6936 * i40e_veb_setup - Set up a VEB
6937 * @pf: board private structure
6938 * @flags: VEB setup flags
6939 * @uplink_seid: the switch element to link to
6940 * @vsi_seid: the initial VSI seid
6941 * @enabled_tc: Enabled TC bit-map
6943 * This allocates the sw VEB structure and links it into the switch
6944 * It is possible and legal for this to be a duplicate of an already
6945 * existing VEB. It is also possible for both uplink and vsi seids
6946 * to be zero, in order to create a floating VEB.
6948 * Returns pointer to the successfully allocated VEB sw struct on
6949 * success, otherwise returns NULL on failure.
6951 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
6952 u16 uplink_seid, u16 vsi_seid,
6955 struct i40e_veb *veb, *uplink_veb = NULL;
6956 int vsi_idx, veb_idx;
6959 /* if one seid is 0, the other must be 0 to create a floating relay */
6960 if ((uplink_seid == 0 || vsi_seid == 0) &&
6961 (uplink_seid + vsi_seid != 0)) {
6962 dev_info(&pf->pdev->dev,
6963 "one, not both seid's are 0: uplink=%d vsi=%d\n",
6964 uplink_seid, vsi_seid);
6968 /* make sure there is such a vsi and uplink */
6969 for (vsi_idx = 0; vsi_idx < pf->hw.func_caps.num_vsis; vsi_idx++)
6970 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
6972 if (vsi_idx >= pf->hw.func_caps.num_vsis && vsi_seid != 0) {
6973 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
6978 if (uplink_seid && uplink_seid != pf->mac_seid) {
6979 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6980 if (pf->veb[veb_idx] &&
6981 pf->veb[veb_idx]->seid == uplink_seid) {
6982 uplink_veb = pf->veb[veb_idx];
6987 dev_info(&pf->pdev->dev,
6988 "uplink seid %d not found\n", uplink_seid);
6993 /* get veb sw struct */
6994 veb_idx = i40e_veb_mem_alloc(pf);
6997 veb = pf->veb[veb_idx];
6999 veb->uplink_seid = uplink_seid;
7000 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
7001 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
7003 /* create the VEB in the switch */
7004 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
7011 i40e_veb_clear(veb);
7017 * i40e_setup_pf_switch_element - set pf vars based on switch type
7018 * @pf: board private structure
7019 * @ele: element we are building info from
7020 * @num_reported: total number of elements
7021 * @printconfig: should we print the contents
7023 * helper function to assist in extracting a few useful SEID values.
7025 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
7026 struct i40e_aqc_switch_config_element_resp *ele,
7027 u16 num_reported, bool printconfig)
7029 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
7030 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
7031 u8 element_type = ele->element_type;
7032 u16 seid = le16_to_cpu(ele->seid);
7035 dev_info(&pf->pdev->dev,
7036 "type=%d seid=%d uplink=%d downlink=%d\n",
7037 element_type, seid, uplink_seid, downlink_seid);
7039 switch (element_type) {
7040 case I40E_SWITCH_ELEMENT_TYPE_MAC:
7041 pf->mac_seid = seid;
7043 case I40E_SWITCH_ELEMENT_TYPE_VEB:
7045 if (uplink_seid != pf->mac_seid)
7047 if (pf->lan_veb == I40E_NO_VEB) {
7050 /* find existing or else empty VEB */
7051 for (v = 0; v < I40E_MAX_VEB; v++) {
7052 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
7057 if (pf->lan_veb == I40E_NO_VEB) {
7058 v = i40e_veb_mem_alloc(pf);
7065 pf->veb[pf->lan_veb]->seid = seid;
7066 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
7067 pf->veb[pf->lan_veb]->pf = pf;
7068 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
7070 case I40E_SWITCH_ELEMENT_TYPE_VSI:
7071 if (num_reported != 1)
7073 /* This is immediately after a reset so we can assume this is
7076 pf->mac_seid = uplink_seid;
7077 pf->pf_seid = downlink_seid;
7078 pf->main_vsi_seid = seid;
7080 dev_info(&pf->pdev->dev,
7081 "pf_seid=%d main_vsi_seid=%d\n",
7082 pf->pf_seid, pf->main_vsi_seid);
7084 case I40E_SWITCH_ELEMENT_TYPE_PF:
7085 case I40E_SWITCH_ELEMENT_TYPE_VF:
7086 case I40E_SWITCH_ELEMENT_TYPE_EMP:
7087 case I40E_SWITCH_ELEMENT_TYPE_BMC:
7088 case I40E_SWITCH_ELEMENT_TYPE_PE:
7089 case I40E_SWITCH_ELEMENT_TYPE_PA:
7090 /* ignore these for now */
7093 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
7094 element_type, seid);
7100 * i40e_fetch_switch_configuration - Get switch config from firmware
7101 * @pf: board private structure
7102 * @printconfig: should we print the contents
7104 * Get the current switch configuration from the device and
7105 * extract a few useful SEID values.
7107 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
7109 struct i40e_aqc_get_switch_config_resp *sw_config;
7115 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
7119 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
7121 u16 num_reported, num_total;
7123 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
7127 dev_info(&pf->pdev->dev,
7128 "get switch config failed %d aq_err=%x\n",
7129 ret, pf->hw.aq.asq_last_status);
7134 num_reported = le16_to_cpu(sw_config->header.num_reported);
7135 num_total = le16_to_cpu(sw_config->header.num_total);
7138 dev_info(&pf->pdev->dev,
7139 "header: %d reported %d total\n",
7140 num_reported, num_total);
7143 int sz = sizeof(*sw_config) * num_reported;
7145 kfree(pf->sw_config);
7146 pf->sw_config = kzalloc(sz, GFP_KERNEL);
7148 memcpy(pf->sw_config, sw_config, sz);
7151 for (i = 0; i < num_reported; i++) {
7152 struct i40e_aqc_switch_config_element_resp *ele =
7153 &sw_config->element[i];
7155 i40e_setup_pf_switch_element(pf, ele, num_reported,
7158 } while (next_seid != 0);
7165 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
7166 * @pf: board private structure
7167 * @reinit: if the Main VSI needs to re-initialized.
7169 * Returns 0 on success, negative value on failure
7171 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
7173 u32 rxfc = 0, txfc = 0, rxfc_reg;
7176 /* find out what's out there already */
7177 ret = i40e_fetch_switch_configuration(pf, false);
7179 dev_info(&pf->pdev->dev,
7180 "couldn't fetch switch config, err %d, aq_err %d\n",
7181 ret, pf->hw.aq.asq_last_status);
7184 i40e_pf_reset_stats(pf);
7186 /* fdir VSI must happen first to be sure it gets queue 0, but only
7187 * if there is enough room for the fdir VSI
7189 if (pf->num_lan_qps > 1)
7190 i40e_fdir_setup(pf);
7192 /* first time setup */
7193 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
7194 struct i40e_vsi *vsi = NULL;
7197 /* Set up the PF VSI associated with the PF's main VSI
7198 * that is already in the HW switch
7200 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
7201 uplink_seid = pf->veb[pf->lan_veb]->seid;
7203 uplink_seid = pf->mac_seid;
7204 if (pf->lan_vsi == I40E_NO_VSI)
7205 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
7207 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
7209 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
7210 i40e_fdir_teardown(pf);
7214 /* force a reset of TC and queue layout configurations */
7215 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
7216 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
7217 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
7218 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
7220 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
7222 /* Setup static PF queue filter control settings */
7223 ret = i40e_setup_pf_filter_control(pf);
7225 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
7227 /* Failure here should not stop continuing other steps */
7230 /* enable RSS in the HW, even for only one queue, as the stack can use
7233 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
7234 i40e_config_rss(pf);
7236 /* fill in link information and enable LSE reporting */
7237 i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
7238 i40e_link_event(pf);
7240 /* Initialize user-specific link properties */
7241 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
7242 I40E_AQ_AN_COMPLETED) ? true : false);
7243 /* requested_mode is set in probe or by ethtool */
7244 if (!pf->fc_autoneg_status)
7247 if ((pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX) &&
7248 (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX))
7249 pf->hw.fc.current_mode = I40E_FC_FULL;
7250 else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX)
7251 pf->hw.fc.current_mode = I40E_FC_TX_PAUSE;
7252 else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX)
7253 pf->hw.fc.current_mode = I40E_FC_RX_PAUSE;
7255 pf->hw.fc.current_mode = I40E_FC_NONE;
7257 /* sync the flow control settings with the auto-neg values */
7258 switch (pf->hw.fc.current_mode) {
7263 case I40E_FC_TX_PAUSE:
7267 case I40E_FC_RX_PAUSE:
7272 case I40E_FC_DEFAULT:
7279 /* no default case, we have to handle all possibilities here */
7282 wr32(&pf->hw, I40E_PRTDCB_FCCFG, txfc << I40E_PRTDCB_FCCFG_TFCE_SHIFT);
7284 rxfc_reg = rd32(&pf->hw, I40E_PRTDCB_MFLCN) &
7285 ~I40E_PRTDCB_MFLCN_RFCE_MASK;
7286 rxfc_reg |= (rxfc << I40E_PRTDCB_MFLCN_RFCE_SHIFT);
7288 wr32(&pf->hw, I40E_PRTDCB_MFLCN, rxfc_reg);
7293 /* disable L2 flow control, user can turn it on if they wish */
7294 wr32(&pf->hw, I40E_PRTDCB_FCCFG, 0);
7295 wr32(&pf->hw, I40E_PRTDCB_MFLCN, rd32(&pf->hw, I40E_PRTDCB_MFLCN) &
7296 ~I40E_PRTDCB_MFLCN_RFCE_MASK);
7303 * i40e_set_rss_size - helper to set rss_size
7304 * @pf: board private structure
7305 * @queues_left: how many queues
7307 static u16 i40e_set_rss_size(struct i40e_pf *pf, int queues_left)
7311 num_tc0 = min_t(int, queues_left, pf->rss_size_max);
7312 num_tc0 = min_t(int, num_tc0, num_online_cpus());
7313 num_tc0 = rounddown_pow_of_two(num_tc0);
7319 * i40e_determine_queue_usage - Work out queue distribution
7320 * @pf: board private structure
7322 static void i40e_determine_queue_usage(struct i40e_pf *pf)
7327 pf->num_lan_qps = 0;
7328 pf->num_tc_qps = rounddown_pow_of_two(pf->num_tc_qps);
7329 accum_tc_size = (I40E_MAX_TRAFFIC_CLASS - 1) * pf->num_tc_qps;
7331 /* Find the max queues to be put into basic use. We'll always be
7332 * using TC0, whether or not DCB is running, and TC0 will get the
7335 queues_left = pf->hw.func_caps.num_tx_qp;
7337 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) ||
7338 !(pf->flags & (I40E_FLAG_RSS_ENABLED |
7339 I40E_FLAG_FDIR_ENABLED | I40E_FLAG_DCB_ENABLED)) ||
7340 (queues_left == 1)) {
7342 /* one qp for PF, no queues for anything else */
7344 pf->rss_size = pf->num_lan_qps = 1;
7346 /* make sure all the fancies are disabled */
7347 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
7348 I40E_FLAG_FDIR_ENABLED |
7349 I40E_FLAG_FDIR_ATR_ENABLED |
7350 I40E_FLAG_DCB_ENABLED |
7351 I40E_FLAG_SRIOV_ENABLED |
7352 I40E_FLAG_VMDQ_ENABLED);
7354 } else if (pf->flags & I40E_FLAG_RSS_ENABLED &&
7355 !(pf->flags & I40E_FLAG_FDIR_ENABLED) &&
7356 !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
7358 pf->rss_size = i40e_set_rss_size(pf, queues_left);
7360 queues_left -= pf->rss_size;
7361 pf->num_lan_qps = pf->rss_size_max;
7363 } else if (pf->flags & I40E_FLAG_RSS_ENABLED &&
7364 !(pf->flags & I40E_FLAG_FDIR_ENABLED) &&
7365 (pf->flags & I40E_FLAG_DCB_ENABLED)) {
7367 /* save num_tc_qps queues for TCs 1 thru 7 and the rest
7368 * are set up for RSS in TC0
7370 queues_left -= accum_tc_size;
7372 pf->rss_size = i40e_set_rss_size(pf, queues_left);
7374 queues_left -= pf->rss_size;
7375 if (queues_left < 0) {
7376 dev_info(&pf->pdev->dev, "not enough queues for DCB\n");
7380 pf->num_lan_qps = pf->rss_size_max + accum_tc_size;
7382 } else if (pf->flags & I40E_FLAG_RSS_ENABLED &&
7383 (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
7384 !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
7386 queues_left -= 1; /* save 1 queue for FD */
7388 pf->rss_size = i40e_set_rss_size(pf, queues_left);
7390 queues_left -= pf->rss_size;
7391 if (queues_left < 0) {
7392 dev_info(&pf->pdev->dev, "not enough queues for Flow Director\n");
7396 pf->num_lan_qps = pf->rss_size_max;
7398 } else if (pf->flags & I40E_FLAG_RSS_ENABLED &&
7399 (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
7400 (pf->flags & I40E_FLAG_DCB_ENABLED)) {
7402 /* save 1 queue for TCs 1 thru 7,
7403 * 1 queue for flow director,
7404 * and the rest are set up for RSS in TC0
7407 queues_left -= accum_tc_size;
7409 pf->rss_size = i40e_set_rss_size(pf, queues_left);
7410 queues_left -= pf->rss_size;
7411 if (queues_left < 0) {
7412 dev_info(&pf->pdev->dev, "not enough queues for DCB and Flow Director\n");
7416 pf->num_lan_qps = pf->rss_size_max + accum_tc_size;
7419 dev_info(&pf->pdev->dev,
7420 "Invalid configuration, flags=0x%08llx\n", pf->flags);
7424 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
7425 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
7426 pf->num_req_vfs = min_t(int, pf->num_req_vfs, (queues_left /
7428 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
7431 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7432 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
7433 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
7434 (queues_left / pf->num_vmdq_qps));
7435 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
7438 pf->queues_left = queues_left;
7443 * i40e_setup_pf_filter_control - Setup PF static filter control
7444 * @pf: PF to be setup
7446 * i40e_setup_pf_filter_control sets up a pf's initial filter control
7447 * settings. If PE/FCoE are enabled then it will also set the per PF
7448 * based filter sizes required for them. It also enables Flow director,
7449 * ethertype and macvlan type filter settings for the pf.
7451 * Returns 0 on success, negative on failure
7453 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
7455 struct i40e_filter_control_settings *settings = &pf->filter_settings;
7457 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
7459 /* Flow Director is enabled */
7460 if (pf->flags & (I40E_FLAG_FDIR_ENABLED | I40E_FLAG_FDIR_ATR_ENABLED))
7461 settings->enable_fdir = true;
7463 /* Ethtype and MACVLAN filters enabled for PF */
7464 settings->enable_ethtype = true;
7465 settings->enable_macvlan = true;
7467 if (i40e_set_filter_control(&pf->hw, settings))
7474 * i40e_probe - Device initialization routine
7475 * @pdev: PCI device information struct
7476 * @ent: entry in i40e_pci_tbl
7478 * i40e_probe initializes a pf identified by a pci_dev structure.
7479 * The OS initialization, configuring of the pf private structure,
7480 * and a hardware reset occur.
7482 * Returns 0 on success, negative on failure
7484 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7486 struct i40e_driver_version dv;
7489 static u16 pfs_found;
7494 err = pci_enable_device_mem(pdev);
7498 /* set up for high or low dma */
7499 if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7500 /* coherent mask for the same size will always succeed if
7503 dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
7504 } else if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(32))) {
7505 dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(32));
7507 dev_err(&pdev->dev, "DMA configuration failed: %d\n", err);
7512 /* set up pci connections */
7513 err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7514 IORESOURCE_MEM), i40e_driver_name);
7516 dev_info(&pdev->dev,
7517 "pci_request_selected_regions failed %d\n", err);
7521 pci_enable_pcie_error_reporting(pdev);
7522 pci_set_master(pdev);
7524 /* Now that we have a PCI connection, we need to do the
7525 * low level device setup. This is primarily setting up
7526 * the Admin Queue structures and then querying for the
7527 * device's current profile information.
7529 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
7536 set_bit(__I40E_DOWN, &pf->state);
7540 hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7541 pci_resource_len(pdev, 0));
7544 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
7545 (unsigned int)pci_resource_start(pdev, 0),
7546 (unsigned int)pci_resource_len(pdev, 0), err);
7549 hw->vendor_id = pdev->vendor;
7550 hw->device_id = pdev->device;
7551 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
7552 hw->subsystem_vendor_id = pdev->subsystem_vendor;
7553 hw->subsystem_device_id = pdev->subsystem_device;
7554 hw->bus.device = PCI_SLOT(pdev->devfn);
7555 hw->bus.func = PCI_FUNC(pdev->devfn);
7556 pf->instance = pfs_found;
7558 /* do a special CORER for clearing PXE mode once at init */
7559 if (hw->revision_id == 0 &&
7560 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
7561 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
7566 i40e_clear_pxe_mode(hw);
7569 /* Reset here to make sure all is clean and to define PF 'n' */
7570 err = i40e_pf_reset(hw);
7572 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
7577 hw->aq.num_arq_entries = I40E_AQ_LEN;
7578 hw->aq.num_asq_entries = I40E_AQ_LEN;
7579 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
7580 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
7581 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
7582 snprintf(pf->misc_int_name, sizeof(pf->misc_int_name) - 1,
7584 dev_driver_string(&pf->pdev->dev), pf->hw.pf_id);
7586 err = i40e_init_shared_code(hw);
7588 dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
7592 /* set up a default setting for link flow control */
7593 pf->hw.fc.requested_mode = I40E_FC_NONE;
7595 err = i40e_init_adminq(hw);
7596 dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
7597 if (((hw->nvm.version & I40E_NVM_VERSION_HI_MASK)
7598 >> I40E_NVM_VERSION_HI_SHIFT) != I40E_CURRENT_NVM_VERSION_HI) {
7599 dev_info(&pdev->dev,
7600 "warning: NVM version not supported, supported version: %02x.%02x\n",
7601 I40E_CURRENT_NVM_VERSION_HI,
7602 I40E_CURRENT_NVM_VERSION_LO);
7605 dev_info(&pdev->dev,
7606 "init_adminq failed: %d expecting API %02x.%02x\n",
7608 I40E_FW_API_VERSION_MAJOR, I40E_FW_API_VERSION_MINOR);
7612 err = i40e_get_capabilities(pf);
7614 goto err_adminq_setup;
7616 err = i40e_sw_init(pf);
7618 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
7622 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7623 hw->func_caps.num_rx_qp,
7624 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
7626 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
7627 goto err_init_lan_hmc;
7630 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7632 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
7634 goto err_configure_lan_hmc;
7637 i40e_get_mac_addr(hw, hw->mac.addr);
7638 if (!is_valid_ether_addr(hw->mac.addr)) {
7639 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
7643 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
7644 memcpy(hw->mac.perm_addr, hw->mac.addr, ETH_ALEN);
7646 pci_set_drvdata(pdev, pf);
7647 pci_save_state(pdev);
7649 /* set up periodic task facility */
7650 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
7651 pf->service_timer_period = HZ;
7653 INIT_WORK(&pf->service_task, i40e_service_task);
7654 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
7655 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
7656 pf->link_check_timeout = jiffies;
7658 /* WoL defaults to disabled */
7660 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
7662 /* set up the main switch operations */
7663 i40e_determine_queue_usage(pf);
7664 i40e_init_interrupt_scheme(pf);
7666 /* Set up the *vsi struct based on the number of VSIs in the HW,
7667 * and set up our local tracking of the MAIN PF vsi.
7669 len = sizeof(struct i40e_vsi *) * pf->hw.func_caps.num_vsis;
7670 pf->vsi = kzalloc(len, GFP_KERNEL);
7673 goto err_switch_setup;
7676 err = i40e_setup_pf_switch(pf, false);
7678 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
7682 /* The main driver is (mostly) up and happy. We need to set this state
7683 * before setting up the misc vector or we get a race and the vector
7684 * ends up disabled forever.
7686 clear_bit(__I40E_DOWN, &pf->state);
7688 /* In case of MSIX we are going to setup the misc vector right here
7689 * to handle admin queue events etc. In case of legacy and MSI
7690 * the misc functionality and queue processing is combined in
7691 * the same vector and that gets setup at open.
7693 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7694 err = i40e_setup_misc_vector(pf);
7696 dev_info(&pdev->dev,
7697 "setup of misc vector failed: %d\n", err);
7702 /* prep for VF support */
7703 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
7704 (pf->flags & I40E_FLAG_MSIX_ENABLED)) {
7707 /* disable link interrupts for VFs */
7708 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
7709 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
7710 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
7716 i40e_dbg_pf_init(pf);
7718 /* tell the firmware that we're starting */
7719 dv.major_version = DRV_VERSION_MAJOR;
7720 dv.minor_version = DRV_VERSION_MINOR;
7721 dv.build_version = DRV_VERSION_BUILD;
7722 dv.subbuild_version = 0;
7723 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
7725 /* since everything's happy, start the service_task timer */
7726 mod_timer(&pf->service_timer,
7727 round_jiffies(jiffies + pf->service_timer_period));
7729 /* Get the negotiated link width and speed from PCI config space */
7730 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA, &link_status);
7732 i40e_set_pci_config_data(hw, link_status);
7734 dev_info(&pdev->dev, "PCI Express: %s %s\n",
7735 (hw->bus.speed == i40e_bus_speed_8000 ? "Speed 8.0GT/s" :
7736 hw->bus.speed == i40e_bus_speed_5000 ? "Speed 5.0GT/s" :
7737 hw->bus.speed == i40e_bus_speed_2500 ? "Speed 2.5GT/s" :
7739 (hw->bus.width == i40e_bus_width_pcie_x8 ? "Width x8" :
7740 hw->bus.width == i40e_bus_width_pcie_x4 ? "Width x4" :
7741 hw->bus.width == i40e_bus_width_pcie_x2 ? "Width x2" :
7742 hw->bus.width == i40e_bus_width_pcie_x1 ? "Width x1" :
7745 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
7746 hw->bus.speed < i40e_bus_speed_8000) {
7747 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
7748 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
7753 /* Unwind what we've done if something failed in the setup */
7755 set_bit(__I40E_DOWN, &pf->state);
7756 i40e_clear_interrupt_scheme(pf);
7759 i40e_reset_interrupt_capability(pf);
7760 del_timer_sync(&pf->service_timer);
7762 err_configure_lan_hmc:
7763 (void)i40e_shutdown_lan_hmc(hw);
7766 kfree(pf->irq_pile);
7769 (void)i40e_shutdown_adminq(hw);
7771 iounmap(hw->hw_addr);
7775 pci_disable_pcie_error_reporting(pdev);
7776 pci_release_selected_regions(pdev,
7777 pci_select_bars(pdev, IORESOURCE_MEM));
7780 pci_disable_device(pdev);
7785 * i40e_remove - Device removal routine
7786 * @pdev: PCI device information struct
7788 * i40e_remove is called by the PCI subsystem to alert the driver
7789 * that is should release a PCI device. This could be caused by a
7790 * Hot-Plug event, or because the driver is going to be removed from
7793 static void i40e_remove(struct pci_dev *pdev)
7795 struct i40e_pf *pf = pci_get_drvdata(pdev);
7796 i40e_status ret_code;
7800 i40e_dbg_pf_exit(pf);
7802 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
7804 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
7807 /* no more scheduling of any task */
7808 set_bit(__I40E_DOWN, &pf->state);
7809 del_timer_sync(&pf->service_timer);
7810 cancel_work_sync(&pf->service_task);
7812 i40e_fdir_teardown(pf);
7814 /* If there is a switch structure or any orphans, remove them.
7815 * This will leave only the PF's VSI remaining.
7817 for (i = 0; i < I40E_MAX_VEB; i++) {
7821 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
7822 pf->veb[i]->uplink_seid == 0)
7823 i40e_switch_branch_release(pf->veb[i]);
7826 /* Now we can shutdown the PF's VSI, just before we kill
7829 if (pf->vsi[pf->lan_vsi])
7830 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
7832 i40e_stop_misc_vector(pf);
7833 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7834 synchronize_irq(pf->msix_entries[0].vector);
7835 free_irq(pf->msix_entries[0].vector, pf);
7838 /* shutdown and destroy the HMC */
7839 ret_code = i40e_shutdown_lan_hmc(&pf->hw);
7841 dev_warn(&pdev->dev,
7842 "Failed to destroy the HMC resources: %d\n", ret_code);
7844 /* shutdown the adminq */
7845 ret_code = i40e_shutdown_adminq(&pf->hw);
7847 dev_warn(&pdev->dev,
7848 "Failed to destroy the Admin Queue resources: %d\n",
7851 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
7852 i40e_clear_interrupt_scheme(pf);
7853 for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7855 i40e_vsi_clear_rings(pf->vsi[i]);
7856 i40e_vsi_clear(pf->vsi[i]);
7861 for (i = 0; i < I40E_MAX_VEB; i++) {
7867 kfree(pf->irq_pile);
7868 kfree(pf->sw_config);
7871 /* force a PF reset to clean anything leftover */
7872 reg = rd32(&pf->hw, I40E_PFGEN_CTRL);
7873 wr32(&pf->hw, I40E_PFGEN_CTRL, (reg | I40E_PFGEN_CTRL_PFSWR_MASK));
7874 i40e_flush(&pf->hw);
7876 iounmap(pf->hw.hw_addr);
7878 pci_release_selected_regions(pdev,
7879 pci_select_bars(pdev, IORESOURCE_MEM));
7881 pci_disable_pcie_error_reporting(pdev);
7882 pci_disable_device(pdev);
7886 * i40e_pci_error_detected - warning that something funky happened in PCI land
7887 * @pdev: PCI device information struct
7889 * Called to warn that something happened and the error handling steps
7890 * are in progress. Allows the driver to quiesce things, be ready for
7893 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
7894 enum pci_channel_state error)
7896 struct i40e_pf *pf = pci_get_drvdata(pdev);
7898 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
7900 /* shutdown all operations */
7901 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
7903 i40e_prep_for_reset(pf);
7907 /* Request a slot reset */
7908 return PCI_ERS_RESULT_NEED_RESET;
7912 * i40e_pci_error_slot_reset - a PCI slot reset just happened
7913 * @pdev: PCI device information struct
7915 * Called to find if the driver can work with the device now that
7916 * the pci slot has been reset. If a basic connection seems good
7917 * (registers are readable and have sane content) then return a
7918 * happy little PCI_ERS_RESULT_xxx.
7920 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
7922 struct i40e_pf *pf = pci_get_drvdata(pdev);
7923 pci_ers_result_t result;
7927 dev_info(&pdev->dev, "%s\n", __func__);
7928 if (pci_enable_device_mem(pdev)) {
7929 dev_info(&pdev->dev,
7930 "Cannot re-enable PCI device after reset.\n");
7931 result = PCI_ERS_RESULT_DISCONNECT;
7933 pci_set_master(pdev);
7934 pci_restore_state(pdev);
7935 pci_save_state(pdev);
7936 pci_wake_from_d3(pdev, false);
7938 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
7940 result = PCI_ERS_RESULT_RECOVERED;
7942 result = PCI_ERS_RESULT_DISCONNECT;
7945 err = pci_cleanup_aer_uncorrect_error_status(pdev);
7947 dev_info(&pdev->dev,
7948 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7950 /* non-fatal, continue */
7957 * i40e_pci_error_resume - restart operations after PCI error recovery
7958 * @pdev: PCI device information struct
7960 * Called to allow the driver to bring things back up after PCI error
7961 * and/or reset recovery has finished.
7963 static void i40e_pci_error_resume(struct pci_dev *pdev)
7965 struct i40e_pf *pf = pci_get_drvdata(pdev);
7967 dev_info(&pdev->dev, "%s\n", __func__);
7968 if (test_bit(__I40E_SUSPENDED, &pf->state))
7972 i40e_handle_reset_warning(pf);
7977 * i40e_shutdown - PCI callback for shutting down
7978 * @pdev: PCI device information struct
7980 static void i40e_shutdown(struct pci_dev *pdev)
7982 struct i40e_pf *pf = pci_get_drvdata(pdev);
7983 struct i40e_hw *hw = &pf->hw;
7985 set_bit(__I40E_SUSPENDED, &pf->state);
7986 set_bit(__I40E_DOWN, &pf->state);
7988 i40e_prep_for_reset(pf);
7991 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
7992 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
7994 if (system_state == SYSTEM_POWER_OFF) {
7995 pci_wake_from_d3(pdev, pf->wol_en);
7996 pci_set_power_state(pdev, PCI_D3hot);
8002 * i40e_suspend - PCI callback for moving to D3
8003 * @pdev: PCI device information struct
8005 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
8007 struct i40e_pf *pf = pci_get_drvdata(pdev);
8008 struct i40e_hw *hw = &pf->hw;
8010 set_bit(__I40E_SUSPENDED, &pf->state);
8011 set_bit(__I40E_DOWN, &pf->state);
8013 i40e_prep_for_reset(pf);
8016 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
8017 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
8019 pci_wake_from_d3(pdev, pf->wol_en);
8020 pci_set_power_state(pdev, PCI_D3hot);
8026 * i40e_resume - PCI callback for waking up from D3
8027 * @pdev: PCI device information struct
8029 static int i40e_resume(struct pci_dev *pdev)
8031 struct i40e_pf *pf = pci_get_drvdata(pdev);
8034 pci_set_power_state(pdev, PCI_D0);
8035 pci_restore_state(pdev);
8036 /* pci_restore_state() clears dev->state_saves, so
8037 * call pci_save_state() again to restore it.
8039 pci_save_state(pdev);
8041 err = pci_enable_device_mem(pdev);
8044 "%s: Cannot enable PCI device from suspend\n",
8048 pci_set_master(pdev);
8050 /* no wakeup events while running */
8051 pci_wake_from_d3(pdev, false);
8053 /* handling the reset will rebuild the device state */
8054 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
8055 clear_bit(__I40E_DOWN, &pf->state);
8057 i40e_reset_and_rebuild(pf, false);
8065 static const struct pci_error_handlers i40e_err_handler = {
8066 .error_detected = i40e_pci_error_detected,
8067 .slot_reset = i40e_pci_error_slot_reset,
8068 .resume = i40e_pci_error_resume,
8071 static struct pci_driver i40e_driver = {
8072 .name = i40e_driver_name,
8073 .id_table = i40e_pci_tbl,
8074 .probe = i40e_probe,
8075 .remove = i40e_remove,
8077 .suspend = i40e_suspend,
8078 .resume = i40e_resume,
8080 .shutdown = i40e_shutdown,
8081 .err_handler = &i40e_err_handler,
8082 .sriov_configure = i40e_pci_sriov_configure,
8086 * i40e_init_module - Driver registration routine
8088 * i40e_init_module is the first routine called when the driver is
8089 * loaded. All it does is register with the PCI subsystem.
8091 static int __init i40e_init_module(void)
8093 pr_info("%s: %s - version %s\n", i40e_driver_name,
8094 i40e_driver_string, i40e_driver_version_str);
8095 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
8097 return pci_register_driver(&i40e_driver);
8099 module_init(i40e_init_module);
8102 * i40e_exit_module - Driver exit cleanup routine
8104 * i40e_exit_module is called just before the driver is removed
8107 static void __exit i40e_exit_module(void)
8109 pci_unregister_driver(&i40e_driver);
8112 module_exit(i40e_exit_module);