i40e/i40evf: Add a workaround to drop all flow control frames
[linux-2.6/btrfs-unstable.git] / drivers / net / ethernet / intel / i40e / i40e_main.c
blobd05254c741d0ae5492da44b629c7169702a4c9b0
1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 - 2015 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
13 * more details.
15 * You should have received a copy of the GNU General Public License along
16 * with this program. If not, see <http://www.gnu.org/licenses/>.
18 * The full GNU General Public License is included in this distribution in
19 * the file called "COPYING".
21 * Contact Information:
22 * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23 * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25 ******************************************************************************/
27 /* Local includes */
28 #include "i40e.h"
29 #include "i40e_diag.h"
30 #ifdef CONFIG_I40E_VXLAN
31 #include <net/vxlan.h>
32 #endif
34 const char i40e_driver_name[] = "i40e";
35 static const char i40e_driver_string[] =
36 "Intel(R) Ethernet Connection XL710 Network Driver";
38 #define DRV_KERN "-k"
40 #define DRV_VERSION_MAJOR 1
41 #define DRV_VERSION_MINOR 3
42 #define DRV_VERSION_BUILD 38
43 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
44 __stringify(DRV_VERSION_MINOR) "." \
45 __stringify(DRV_VERSION_BUILD) DRV_KERN
46 const char i40e_driver_version_str[] = DRV_VERSION;
47 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
49 /* a bit of forward declarations */
50 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
51 static void i40e_handle_reset_warning(struct i40e_pf *pf);
52 static int i40e_add_vsi(struct i40e_vsi *vsi);
53 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
54 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
55 static int i40e_setup_misc_vector(struct i40e_pf *pf);
56 static void i40e_determine_queue_usage(struct i40e_pf *pf);
57 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
58 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
59 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
61 /* i40e_pci_tbl - PCI Device ID Table
63 * Last entry must be all 0s
65 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
66 * Class, Class Mask, private data (not used) }
68 static const struct pci_device_id i40e_pci_tbl[] = {
69 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
70 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
71 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_A), 0},
72 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
73 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
74 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
75 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
76 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
77 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
78 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
79 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
80 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
81 {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
82 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
83 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
84 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
85 /* required last entry */
86 {0, }
88 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
90 #define I40E_MAX_VF_COUNT 128
91 static int debug = -1;
92 module_param(debug, int, 0);
93 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
95 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
96 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
97 MODULE_LICENSE("GPL");
98 MODULE_VERSION(DRV_VERSION);
101 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
102 * @hw: pointer to the HW structure
103 * @mem: ptr to mem struct to fill out
104 * @size: size of memory requested
105 * @alignment: what to align the allocation to
107 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
108 u64 size, u32 alignment)
110 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
112 mem->size = ALIGN(size, alignment);
113 mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
114 &mem->pa, GFP_KERNEL);
115 if (!mem->va)
116 return -ENOMEM;
118 return 0;
122 * i40e_free_dma_mem_d - OS specific memory free for shared code
123 * @hw: pointer to the HW structure
124 * @mem: ptr to mem struct to free
126 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
128 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
130 dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
131 mem->va = NULL;
132 mem->pa = 0;
133 mem->size = 0;
135 return 0;
139 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
140 * @hw: pointer to the HW structure
141 * @mem: ptr to mem struct to fill out
142 * @size: size of memory requested
144 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
145 u32 size)
147 mem->size = size;
148 mem->va = kzalloc(size, GFP_KERNEL);
150 if (!mem->va)
151 return -ENOMEM;
153 return 0;
157 * i40e_free_virt_mem_d - OS specific memory free for shared code
158 * @hw: pointer to the HW structure
159 * @mem: ptr to mem struct to free
161 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
163 /* it's ok to kfree a NULL pointer */
164 kfree(mem->va);
165 mem->va = NULL;
166 mem->size = 0;
168 return 0;
172 * i40e_get_lump - find a lump of free generic resource
173 * @pf: board private structure
174 * @pile: the pile of resource to search
175 * @needed: the number of items needed
176 * @id: an owner id to stick on the items assigned
178 * Returns the base item index of the lump, or negative for error
180 * The search_hint trick and lack of advanced fit-finding only work
181 * because we're highly likely to have all the same size lump requests.
182 * Linear search time and any fragmentation should be minimal.
184 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
185 u16 needed, u16 id)
187 int ret = -ENOMEM;
188 int i, j;
190 if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
191 dev_info(&pf->pdev->dev,
192 "param err: pile=%p needed=%d id=0x%04x\n",
193 pile, needed, id);
194 return -EINVAL;
197 /* start the linear search with an imperfect hint */
198 i = pile->search_hint;
199 while (i < pile->num_entries) {
200 /* skip already allocated entries */
201 if (pile->list[i] & I40E_PILE_VALID_BIT) {
202 i++;
203 continue;
206 /* do we have enough in this lump? */
207 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
208 if (pile->list[i+j] & I40E_PILE_VALID_BIT)
209 break;
212 if (j == needed) {
213 /* there was enough, so assign it to the requestor */
214 for (j = 0; j < needed; j++)
215 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
216 ret = i;
217 pile->search_hint = i + j;
218 break;
221 /* not enough, so skip over it and continue looking */
222 i += j;
225 return ret;
229 * i40e_put_lump - return a lump of generic resource
230 * @pile: the pile of resource to search
231 * @index: the base item index
232 * @id: the owner id of the items assigned
234 * Returns the count of items in the lump
236 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
238 int valid_id = (id | I40E_PILE_VALID_BIT);
239 int count = 0;
240 int i;
242 if (!pile || index >= pile->num_entries)
243 return -EINVAL;
245 for (i = index;
246 i < pile->num_entries && pile->list[i] == valid_id;
247 i++) {
248 pile->list[i] = 0;
249 count++;
252 if (count && index < pile->search_hint)
253 pile->search_hint = index;
255 return count;
259 * i40e_find_vsi_from_id - searches for the vsi with the given id
260 * @pf - the pf structure to search for the vsi
261 * @id - id of the vsi it is searching for
263 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
265 int i;
267 for (i = 0; i < pf->num_alloc_vsi; i++)
268 if (pf->vsi[i] && (pf->vsi[i]->id == id))
269 return pf->vsi[i];
271 return NULL;
275 * i40e_service_event_schedule - Schedule the service task to wake up
276 * @pf: board private structure
278 * If not already scheduled, this puts the task into the work queue
280 static void i40e_service_event_schedule(struct i40e_pf *pf)
282 if (!test_bit(__I40E_DOWN, &pf->state) &&
283 !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
284 !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
285 schedule_work(&pf->service_task);
289 * i40e_tx_timeout - Respond to a Tx Hang
290 * @netdev: network interface device structure
292 * If any port has noticed a Tx timeout, it is likely that the whole
293 * device is munged, not just the one netdev port, so go for the full
294 * reset.
296 #ifdef I40E_FCOE
297 void i40e_tx_timeout(struct net_device *netdev)
298 #else
299 static void i40e_tx_timeout(struct net_device *netdev)
300 #endif
302 struct i40e_netdev_priv *np = netdev_priv(netdev);
303 struct i40e_vsi *vsi = np->vsi;
304 struct i40e_pf *pf = vsi->back;
305 struct i40e_ring *tx_ring = NULL;
306 unsigned int i, hung_queue = 0;
307 u32 head, val;
309 pf->tx_timeout_count++;
311 /* find the stopped queue the same way the stack does */
312 for (i = 0; i < netdev->num_tx_queues; i++) {
313 struct netdev_queue *q;
314 unsigned long trans_start;
316 q = netdev_get_tx_queue(netdev, i);
317 trans_start = q->trans_start ? : netdev->trans_start;
318 if (netif_xmit_stopped(q) &&
319 time_after(jiffies,
320 (trans_start + netdev->watchdog_timeo))) {
321 hung_queue = i;
322 break;
326 if (i == netdev->num_tx_queues) {
327 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
328 } else {
329 /* now that we have an index, find the tx_ring struct */
330 for (i = 0; i < vsi->num_queue_pairs; i++) {
331 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
332 if (hung_queue ==
333 vsi->tx_rings[i]->queue_index) {
334 tx_ring = vsi->tx_rings[i];
335 break;
341 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
342 pf->tx_timeout_recovery_level = 1; /* reset after some time */
343 else if (time_before(jiffies,
344 (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
345 return; /* don't do any new action before the next timeout */
347 if (tx_ring) {
348 head = i40e_get_head(tx_ring);
349 /* Read interrupt register */
350 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
351 val = rd32(&pf->hw,
352 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
353 tx_ring->vsi->base_vector - 1));
354 else
355 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
357 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
358 vsi->seid, hung_queue, tx_ring->next_to_clean,
359 head, tx_ring->next_to_use,
360 readl(tx_ring->tail), val);
363 pf->tx_timeout_last_recovery = jiffies;
364 netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
365 pf->tx_timeout_recovery_level, hung_queue);
367 switch (pf->tx_timeout_recovery_level) {
368 case 1:
369 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
370 break;
371 case 2:
372 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
373 break;
374 case 3:
375 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
376 break;
377 default:
378 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
379 break;
382 i40e_service_event_schedule(pf);
383 pf->tx_timeout_recovery_level++;
387 * i40e_release_rx_desc - Store the new tail and head values
388 * @rx_ring: ring to bump
389 * @val: new head index
391 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
393 rx_ring->next_to_use = val;
395 /* Force memory writes to complete before letting h/w
396 * know there are new descriptors to fetch. (Only
397 * applicable for weak-ordered memory model archs,
398 * such as IA-64).
400 wmb();
401 writel(val, rx_ring->tail);
405 * i40e_get_vsi_stats_struct - Get System Network Statistics
406 * @vsi: the VSI we care about
408 * Returns the address of the device statistics structure.
409 * The statistics are actually updated from the service task.
411 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
413 return &vsi->net_stats;
417 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
418 * @netdev: network interface device structure
420 * Returns the address of the device statistics structure.
421 * The statistics are actually updated from the service task.
423 #ifdef I40E_FCOE
424 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
425 struct net_device *netdev,
426 struct rtnl_link_stats64 *stats)
427 #else
428 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
429 struct net_device *netdev,
430 struct rtnl_link_stats64 *stats)
431 #endif
433 struct i40e_netdev_priv *np = netdev_priv(netdev);
434 struct i40e_ring *tx_ring, *rx_ring;
435 struct i40e_vsi *vsi = np->vsi;
436 struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
437 int i;
439 if (test_bit(__I40E_DOWN, &vsi->state))
440 return stats;
442 if (!vsi->tx_rings)
443 return stats;
445 rcu_read_lock();
446 for (i = 0; i < vsi->num_queue_pairs; i++) {
447 u64 bytes, packets;
448 unsigned int start;
450 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
451 if (!tx_ring)
452 continue;
454 do {
455 start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
456 packets = tx_ring->stats.packets;
457 bytes = tx_ring->stats.bytes;
458 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
460 stats->tx_packets += packets;
461 stats->tx_bytes += bytes;
462 rx_ring = &tx_ring[1];
464 do {
465 start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
466 packets = rx_ring->stats.packets;
467 bytes = rx_ring->stats.bytes;
468 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
470 stats->rx_packets += packets;
471 stats->rx_bytes += bytes;
473 rcu_read_unlock();
475 /* following stats updated by i40e_watchdog_subtask() */
476 stats->multicast = vsi_stats->multicast;
477 stats->tx_errors = vsi_stats->tx_errors;
478 stats->tx_dropped = vsi_stats->tx_dropped;
479 stats->rx_errors = vsi_stats->rx_errors;
480 stats->rx_dropped = vsi_stats->rx_dropped;
481 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
482 stats->rx_length_errors = vsi_stats->rx_length_errors;
484 return stats;
488 * i40e_vsi_reset_stats - Resets all stats of the given vsi
489 * @vsi: the VSI to have its stats reset
491 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
493 struct rtnl_link_stats64 *ns;
494 int i;
496 if (!vsi)
497 return;
499 ns = i40e_get_vsi_stats_struct(vsi);
500 memset(ns, 0, sizeof(*ns));
501 memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
502 memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
503 memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
504 if (vsi->rx_rings && vsi->rx_rings[0]) {
505 for (i = 0; i < vsi->num_queue_pairs; i++) {
506 memset(&vsi->rx_rings[i]->stats, 0,
507 sizeof(vsi->rx_rings[i]->stats));
508 memset(&vsi->rx_rings[i]->rx_stats, 0,
509 sizeof(vsi->rx_rings[i]->rx_stats));
510 memset(&vsi->tx_rings[i]->stats, 0,
511 sizeof(vsi->tx_rings[i]->stats));
512 memset(&vsi->tx_rings[i]->tx_stats, 0,
513 sizeof(vsi->tx_rings[i]->tx_stats));
516 vsi->stat_offsets_loaded = false;
520 * i40e_pf_reset_stats - Reset all of the stats for the given PF
521 * @pf: the PF to be reset
523 void i40e_pf_reset_stats(struct i40e_pf *pf)
525 int i;
527 memset(&pf->stats, 0, sizeof(pf->stats));
528 memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
529 pf->stat_offsets_loaded = false;
531 for (i = 0; i < I40E_MAX_VEB; i++) {
532 if (pf->veb[i]) {
533 memset(&pf->veb[i]->stats, 0,
534 sizeof(pf->veb[i]->stats));
535 memset(&pf->veb[i]->stats_offsets, 0,
536 sizeof(pf->veb[i]->stats_offsets));
537 pf->veb[i]->stat_offsets_loaded = false;
543 * i40e_stat_update48 - read and update a 48 bit stat from the chip
544 * @hw: ptr to the hardware info
545 * @hireg: the high 32 bit reg to read
546 * @loreg: the low 32 bit reg to read
547 * @offset_loaded: has the initial offset been loaded yet
548 * @offset: ptr to current offset value
549 * @stat: ptr to the stat
551 * Since the device stats are not reset at PFReset, they likely will not
552 * be zeroed when the driver starts. We'll save the first values read
553 * and use them as offsets to be subtracted from the raw values in order
554 * to report stats that count from zero. In the process, we also manage
555 * the potential roll-over.
557 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
558 bool offset_loaded, u64 *offset, u64 *stat)
560 u64 new_data;
562 if (hw->device_id == I40E_DEV_ID_QEMU) {
563 new_data = rd32(hw, loreg);
564 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
565 } else {
566 new_data = rd64(hw, loreg);
568 if (!offset_loaded)
569 *offset = new_data;
570 if (likely(new_data >= *offset))
571 *stat = new_data - *offset;
572 else
573 *stat = (new_data + BIT_ULL(48)) - *offset;
574 *stat &= 0xFFFFFFFFFFFFULL;
578 * i40e_stat_update32 - read and update a 32 bit stat from the chip
579 * @hw: ptr to the hardware info
580 * @reg: the hw reg to read
581 * @offset_loaded: has the initial offset been loaded yet
582 * @offset: ptr to current offset value
583 * @stat: ptr to the stat
585 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
586 bool offset_loaded, u64 *offset, u64 *stat)
588 u32 new_data;
590 new_data = rd32(hw, reg);
591 if (!offset_loaded)
592 *offset = new_data;
593 if (likely(new_data >= *offset))
594 *stat = (u32)(new_data - *offset);
595 else
596 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
600 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
601 * @vsi: the VSI to be updated
603 void i40e_update_eth_stats(struct i40e_vsi *vsi)
605 int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
606 struct i40e_pf *pf = vsi->back;
607 struct i40e_hw *hw = &pf->hw;
608 struct i40e_eth_stats *oes;
609 struct i40e_eth_stats *es; /* device's eth stats */
611 es = &vsi->eth_stats;
612 oes = &vsi->eth_stats_offsets;
614 /* Gather up the stats that the hw collects */
615 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
616 vsi->stat_offsets_loaded,
617 &oes->tx_errors, &es->tx_errors);
618 i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
619 vsi->stat_offsets_loaded,
620 &oes->rx_discards, &es->rx_discards);
621 i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
622 vsi->stat_offsets_loaded,
623 &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
624 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
625 vsi->stat_offsets_loaded,
626 &oes->tx_errors, &es->tx_errors);
628 i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
629 I40E_GLV_GORCL(stat_idx),
630 vsi->stat_offsets_loaded,
631 &oes->rx_bytes, &es->rx_bytes);
632 i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
633 I40E_GLV_UPRCL(stat_idx),
634 vsi->stat_offsets_loaded,
635 &oes->rx_unicast, &es->rx_unicast);
636 i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
637 I40E_GLV_MPRCL(stat_idx),
638 vsi->stat_offsets_loaded,
639 &oes->rx_multicast, &es->rx_multicast);
640 i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
641 I40E_GLV_BPRCL(stat_idx),
642 vsi->stat_offsets_loaded,
643 &oes->rx_broadcast, &es->rx_broadcast);
645 i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
646 I40E_GLV_GOTCL(stat_idx),
647 vsi->stat_offsets_loaded,
648 &oes->tx_bytes, &es->tx_bytes);
649 i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
650 I40E_GLV_UPTCL(stat_idx),
651 vsi->stat_offsets_loaded,
652 &oes->tx_unicast, &es->tx_unicast);
653 i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
654 I40E_GLV_MPTCL(stat_idx),
655 vsi->stat_offsets_loaded,
656 &oes->tx_multicast, &es->tx_multicast);
657 i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
658 I40E_GLV_BPTCL(stat_idx),
659 vsi->stat_offsets_loaded,
660 &oes->tx_broadcast, &es->tx_broadcast);
661 vsi->stat_offsets_loaded = true;
665 * i40e_update_veb_stats - Update Switch component statistics
666 * @veb: the VEB being updated
668 static void i40e_update_veb_stats(struct i40e_veb *veb)
670 struct i40e_pf *pf = veb->pf;
671 struct i40e_hw *hw = &pf->hw;
672 struct i40e_eth_stats *oes;
673 struct i40e_eth_stats *es; /* device's eth stats */
674 struct i40e_veb_tc_stats *veb_oes;
675 struct i40e_veb_tc_stats *veb_es;
676 int i, idx = 0;
678 idx = veb->stats_idx;
679 es = &veb->stats;
680 oes = &veb->stats_offsets;
681 veb_es = &veb->tc_stats;
682 veb_oes = &veb->tc_stats_offsets;
684 /* Gather up the stats that the hw collects */
685 i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
686 veb->stat_offsets_loaded,
687 &oes->tx_discards, &es->tx_discards);
688 if (hw->revision_id > 0)
689 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
690 veb->stat_offsets_loaded,
691 &oes->rx_unknown_protocol,
692 &es->rx_unknown_protocol);
693 i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
694 veb->stat_offsets_loaded,
695 &oes->rx_bytes, &es->rx_bytes);
696 i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
697 veb->stat_offsets_loaded,
698 &oes->rx_unicast, &es->rx_unicast);
699 i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
700 veb->stat_offsets_loaded,
701 &oes->rx_multicast, &es->rx_multicast);
702 i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
703 veb->stat_offsets_loaded,
704 &oes->rx_broadcast, &es->rx_broadcast);
706 i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
707 veb->stat_offsets_loaded,
708 &oes->tx_bytes, &es->tx_bytes);
709 i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
710 veb->stat_offsets_loaded,
711 &oes->tx_unicast, &es->tx_unicast);
712 i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
713 veb->stat_offsets_loaded,
714 &oes->tx_multicast, &es->tx_multicast);
715 i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
716 veb->stat_offsets_loaded,
717 &oes->tx_broadcast, &es->tx_broadcast);
718 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
719 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
720 I40E_GLVEBTC_RPCL(i, idx),
721 veb->stat_offsets_loaded,
722 &veb_oes->tc_rx_packets[i],
723 &veb_es->tc_rx_packets[i]);
724 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
725 I40E_GLVEBTC_RBCL(i, idx),
726 veb->stat_offsets_loaded,
727 &veb_oes->tc_rx_bytes[i],
728 &veb_es->tc_rx_bytes[i]);
729 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
730 I40E_GLVEBTC_TPCL(i, idx),
731 veb->stat_offsets_loaded,
732 &veb_oes->tc_tx_packets[i],
733 &veb_es->tc_tx_packets[i]);
734 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
735 I40E_GLVEBTC_TBCL(i, idx),
736 veb->stat_offsets_loaded,
737 &veb_oes->tc_tx_bytes[i],
738 &veb_es->tc_tx_bytes[i]);
740 veb->stat_offsets_loaded = true;
743 #ifdef I40E_FCOE
745 * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
746 * @vsi: the VSI that is capable of doing FCoE
748 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
750 struct i40e_pf *pf = vsi->back;
751 struct i40e_hw *hw = &pf->hw;
752 struct i40e_fcoe_stats *ofs;
753 struct i40e_fcoe_stats *fs; /* device's eth stats */
754 int idx;
756 if (vsi->type != I40E_VSI_FCOE)
757 return;
759 idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
760 fs = &vsi->fcoe_stats;
761 ofs = &vsi->fcoe_stats_offsets;
763 i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
764 vsi->fcoe_stat_offsets_loaded,
765 &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
766 i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
767 vsi->fcoe_stat_offsets_loaded,
768 &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
769 i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
770 vsi->fcoe_stat_offsets_loaded,
771 &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
772 i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
773 vsi->fcoe_stat_offsets_loaded,
774 &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
775 i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
776 vsi->fcoe_stat_offsets_loaded,
777 &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
778 i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
779 vsi->fcoe_stat_offsets_loaded,
780 &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
781 i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
782 vsi->fcoe_stat_offsets_loaded,
783 &ofs->fcoe_last_error, &fs->fcoe_last_error);
784 i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
785 vsi->fcoe_stat_offsets_loaded,
786 &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
788 vsi->fcoe_stat_offsets_loaded = true;
791 #endif
793 * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
794 * @pf: the corresponding PF
796 * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
798 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
800 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
801 struct i40e_hw_port_stats *nsd = &pf->stats;
802 struct i40e_hw *hw = &pf->hw;
803 u64 xoff = 0;
805 if ((hw->fc.current_mode != I40E_FC_FULL) &&
806 (hw->fc.current_mode != I40E_FC_RX_PAUSE))
807 return;
809 xoff = nsd->link_xoff_rx;
810 i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
811 pf->stat_offsets_loaded,
812 &osd->link_xoff_rx, &nsd->link_xoff_rx);
814 /* No new LFC xoff rx */
815 if (!(nsd->link_xoff_rx - xoff))
816 return;
821 * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
822 * @pf: the corresponding PF
824 * Update the Rx XOFF counter (PAUSE frames) in PFC mode
826 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
828 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
829 struct i40e_hw_port_stats *nsd = &pf->stats;
830 bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
831 struct i40e_dcbx_config *dcb_cfg;
832 struct i40e_hw *hw = &pf->hw;
833 u16 i;
834 u8 tc;
836 dcb_cfg = &hw->local_dcbx_config;
838 /* Collect Link XOFF stats when PFC is disabled */
839 if (!dcb_cfg->pfc.pfcenable) {
840 i40e_update_link_xoff_rx(pf);
841 return;
844 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
845 u64 prio_xoff = nsd->priority_xoff_rx[i];
847 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
848 pf->stat_offsets_loaded,
849 &osd->priority_xoff_rx[i],
850 &nsd->priority_xoff_rx[i]);
852 /* No new PFC xoff rx */
853 if (!(nsd->priority_xoff_rx[i] - prio_xoff))
854 continue;
855 /* Get the TC for given priority */
856 tc = dcb_cfg->etscfg.prioritytable[i];
857 xoff[tc] = true;
862 * i40e_update_vsi_stats - Update the vsi statistics counters.
863 * @vsi: the VSI to be updated
865 * There are a few instances where we store the same stat in a
866 * couple of different structs. This is partly because we have
867 * the netdev stats that need to be filled out, which is slightly
868 * different from the "eth_stats" defined by the chip and used in
869 * VF communications. We sort it out here.
871 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
873 struct i40e_pf *pf = vsi->back;
874 struct rtnl_link_stats64 *ons;
875 struct rtnl_link_stats64 *ns; /* netdev stats */
876 struct i40e_eth_stats *oes;
877 struct i40e_eth_stats *es; /* device's eth stats */
878 u32 tx_restart, tx_busy;
879 struct i40e_ring *p;
880 u32 rx_page, rx_buf;
881 u64 bytes, packets;
882 unsigned int start;
883 u64 tx_linearize;
884 u64 rx_p, rx_b;
885 u64 tx_p, tx_b;
886 u16 q;
888 if (test_bit(__I40E_DOWN, &vsi->state) ||
889 test_bit(__I40E_CONFIG_BUSY, &pf->state))
890 return;
892 ns = i40e_get_vsi_stats_struct(vsi);
893 ons = &vsi->net_stats_offsets;
894 es = &vsi->eth_stats;
895 oes = &vsi->eth_stats_offsets;
897 /* Gather up the netdev and vsi stats that the driver collects
898 * on the fly during packet processing
900 rx_b = rx_p = 0;
901 tx_b = tx_p = 0;
902 tx_restart = tx_busy = tx_linearize = 0;
903 rx_page = 0;
904 rx_buf = 0;
905 rcu_read_lock();
906 for (q = 0; q < vsi->num_queue_pairs; q++) {
907 /* locate Tx ring */
908 p = ACCESS_ONCE(vsi->tx_rings[q]);
910 do {
911 start = u64_stats_fetch_begin_irq(&p->syncp);
912 packets = p->stats.packets;
913 bytes = p->stats.bytes;
914 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
915 tx_b += bytes;
916 tx_p += packets;
917 tx_restart += p->tx_stats.restart_queue;
918 tx_busy += p->tx_stats.tx_busy;
919 tx_linearize += p->tx_stats.tx_linearize;
921 /* Rx queue is part of the same block as Tx queue */
922 p = &p[1];
923 do {
924 start = u64_stats_fetch_begin_irq(&p->syncp);
925 packets = p->stats.packets;
926 bytes = p->stats.bytes;
927 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
928 rx_b += bytes;
929 rx_p += packets;
930 rx_buf += p->rx_stats.alloc_buff_failed;
931 rx_page += p->rx_stats.alloc_page_failed;
933 rcu_read_unlock();
934 vsi->tx_restart = tx_restart;
935 vsi->tx_busy = tx_busy;
936 vsi->tx_linearize = tx_linearize;
937 vsi->rx_page_failed = rx_page;
938 vsi->rx_buf_failed = rx_buf;
940 ns->rx_packets = rx_p;
941 ns->rx_bytes = rx_b;
942 ns->tx_packets = tx_p;
943 ns->tx_bytes = tx_b;
945 /* update netdev stats from eth stats */
946 i40e_update_eth_stats(vsi);
947 ons->tx_errors = oes->tx_errors;
948 ns->tx_errors = es->tx_errors;
949 ons->multicast = oes->rx_multicast;
950 ns->multicast = es->rx_multicast;
951 ons->rx_dropped = oes->rx_discards;
952 ns->rx_dropped = es->rx_discards;
953 ons->tx_dropped = oes->tx_discards;
954 ns->tx_dropped = es->tx_discards;
956 /* pull in a couple PF stats if this is the main vsi */
957 if (vsi == pf->vsi[pf->lan_vsi]) {
958 ns->rx_crc_errors = pf->stats.crc_errors;
959 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
960 ns->rx_length_errors = pf->stats.rx_length_errors;
965 * i40e_update_pf_stats - Update the PF statistics counters.
966 * @pf: the PF to be updated
968 static void i40e_update_pf_stats(struct i40e_pf *pf)
970 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
971 struct i40e_hw_port_stats *nsd = &pf->stats;
972 struct i40e_hw *hw = &pf->hw;
973 u32 val;
974 int i;
976 i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
977 I40E_GLPRT_GORCL(hw->port),
978 pf->stat_offsets_loaded,
979 &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
980 i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
981 I40E_GLPRT_GOTCL(hw->port),
982 pf->stat_offsets_loaded,
983 &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
984 i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
985 pf->stat_offsets_loaded,
986 &osd->eth.rx_discards,
987 &nsd->eth.rx_discards);
988 i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
989 I40E_GLPRT_UPRCL(hw->port),
990 pf->stat_offsets_loaded,
991 &osd->eth.rx_unicast,
992 &nsd->eth.rx_unicast);
993 i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
994 I40E_GLPRT_MPRCL(hw->port),
995 pf->stat_offsets_loaded,
996 &osd->eth.rx_multicast,
997 &nsd->eth.rx_multicast);
998 i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
999 I40E_GLPRT_BPRCL(hw->port),
1000 pf->stat_offsets_loaded,
1001 &osd->eth.rx_broadcast,
1002 &nsd->eth.rx_broadcast);
1003 i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
1004 I40E_GLPRT_UPTCL(hw->port),
1005 pf->stat_offsets_loaded,
1006 &osd->eth.tx_unicast,
1007 &nsd->eth.tx_unicast);
1008 i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
1009 I40E_GLPRT_MPTCL(hw->port),
1010 pf->stat_offsets_loaded,
1011 &osd->eth.tx_multicast,
1012 &nsd->eth.tx_multicast);
1013 i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
1014 I40E_GLPRT_BPTCL(hw->port),
1015 pf->stat_offsets_loaded,
1016 &osd->eth.tx_broadcast,
1017 &nsd->eth.tx_broadcast);
1019 i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
1020 pf->stat_offsets_loaded,
1021 &osd->tx_dropped_link_down,
1022 &nsd->tx_dropped_link_down);
1024 i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
1025 pf->stat_offsets_loaded,
1026 &osd->crc_errors, &nsd->crc_errors);
1028 i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
1029 pf->stat_offsets_loaded,
1030 &osd->illegal_bytes, &nsd->illegal_bytes);
1032 i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
1033 pf->stat_offsets_loaded,
1034 &osd->mac_local_faults,
1035 &nsd->mac_local_faults);
1036 i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
1037 pf->stat_offsets_loaded,
1038 &osd->mac_remote_faults,
1039 &nsd->mac_remote_faults);
1041 i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
1042 pf->stat_offsets_loaded,
1043 &osd->rx_length_errors,
1044 &nsd->rx_length_errors);
1046 i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
1047 pf->stat_offsets_loaded,
1048 &osd->link_xon_rx, &nsd->link_xon_rx);
1049 i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
1050 pf->stat_offsets_loaded,
1051 &osd->link_xon_tx, &nsd->link_xon_tx);
1052 i40e_update_prio_xoff_rx(pf); /* handles I40E_GLPRT_LXOFFRXC */
1053 i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
1054 pf->stat_offsets_loaded,
1055 &osd->link_xoff_tx, &nsd->link_xoff_tx);
1057 for (i = 0; i < 8; i++) {
1058 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1059 pf->stat_offsets_loaded,
1060 &osd->priority_xon_rx[i],
1061 &nsd->priority_xon_rx[i]);
1062 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1063 pf->stat_offsets_loaded,
1064 &osd->priority_xon_tx[i],
1065 &nsd->priority_xon_tx[i]);
1066 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1067 pf->stat_offsets_loaded,
1068 &osd->priority_xoff_tx[i],
1069 &nsd->priority_xoff_tx[i]);
1070 i40e_stat_update32(hw,
1071 I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1072 pf->stat_offsets_loaded,
1073 &osd->priority_xon_2_xoff[i],
1074 &nsd->priority_xon_2_xoff[i]);
1077 i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1078 I40E_GLPRT_PRC64L(hw->port),
1079 pf->stat_offsets_loaded,
1080 &osd->rx_size_64, &nsd->rx_size_64);
1081 i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1082 I40E_GLPRT_PRC127L(hw->port),
1083 pf->stat_offsets_loaded,
1084 &osd->rx_size_127, &nsd->rx_size_127);
1085 i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1086 I40E_GLPRT_PRC255L(hw->port),
1087 pf->stat_offsets_loaded,
1088 &osd->rx_size_255, &nsd->rx_size_255);
1089 i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1090 I40E_GLPRT_PRC511L(hw->port),
1091 pf->stat_offsets_loaded,
1092 &osd->rx_size_511, &nsd->rx_size_511);
1093 i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1094 I40E_GLPRT_PRC1023L(hw->port),
1095 pf->stat_offsets_loaded,
1096 &osd->rx_size_1023, &nsd->rx_size_1023);
1097 i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1098 I40E_GLPRT_PRC1522L(hw->port),
1099 pf->stat_offsets_loaded,
1100 &osd->rx_size_1522, &nsd->rx_size_1522);
1101 i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1102 I40E_GLPRT_PRC9522L(hw->port),
1103 pf->stat_offsets_loaded,
1104 &osd->rx_size_big, &nsd->rx_size_big);
1106 i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1107 I40E_GLPRT_PTC64L(hw->port),
1108 pf->stat_offsets_loaded,
1109 &osd->tx_size_64, &nsd->tx_size_64);
1110 i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1111 I40E_GLPRT_PTC127L(hw->port),
1112 pf->stat_offsets_loaded,
1113 &osd->tx_size_127, &nsd->tx_size_127);
1114 i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1115 I40E_GLPRT_PTC255L(hw->port),
1116 pf->stat_offsets_loaded,
1117 &osd->tx_size_255, &nsd->tx_size_255);
1118 i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1119 I40E_GLPRT_PTC511L(hw->port),
1120 pf->stat_offsets_loaded,
1121 &osd->tx_size_511, &nsd->tx_size_511);
1122 i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1123 I40E_GLPRT_PTC1023L(hw->port),
1124 pf->stat_offsets_loaded,
1125 &osd->tx_size_1023, &nsd->tx_size_1023);
1126 i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1127 I40E_GLPRT_PTC1522L(hw->port),
1128 pf->stat_offsets_loaded,
1129 &osd->tx_size_1522, &nsd->tx_size_1522);
1130 i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1131 I40E_GLPRT_PTC9522L(hw->port),
1132 pf->stat_offsets_loaded,
1133 &osd->tx_size_big, &nsd->tx_size_big);
1135 i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1136 pf->stat_offsets_loaded,
1137 &osd->rx_undersize, &nsd->rx_undersize);
1138 i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1139 pf->stat_offsets_loaded,
1140 &osd->rx_fragments, &nsd->rx_fragments);
1141 i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1142 pf->stat_offsets_loaded,
1143 &osd->rx_oversize, &nsd->rx_oversize);
1144 i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1145 pf->stat_offsets_loaded,
1146 &osd->rx_jabber, &nsd->rx_jabber);
1148 /* FDIR stats */
1149 i40e_stat_update32(hw,
1150 I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1151 pf->stat_offsets_loaded,
1152 &osd->fd_atr_match, &nsd->fd_atr_match);
1153 i40e_stat_update32(hw,
1154 I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1155 pf->stat_offsets_loaded,
1156 &osd->fd_sb_match, &nsd->fd_sb_match);
1157 i40e_stat_update32(hw,
1158 I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1159 pf->stat_offsets_loaded,
1160 &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1162 val = rd32(hw, I40E_PRTPM_EEE_STAT);
1163 nsd->tx_lpi_status =
1164 (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1165 I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1166 nsd->rx_lpi_status =
1167 (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1168 I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1169 i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1170 pf->stat_offsets_loaded,
1171 &osd->tx_lpi_count, &nsd->tx_lpi_count);
1172 i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1173 pf->stat_offsets_loaded,
1174 &osd->rx_lpi_count, &nsd->rx_lpi_count);
1176 if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1177 !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1178 nsd->fd_sb_status = true;
1179 else
1180 nsd->fd_sb_status = false;
1182 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1183 !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1184 nsd->fd_atr_status = true;
1185 else
1186 nsd->fd_atr_status = false;
1188 pf->stat_offsets_loaded = true;
1192 * i40e_update_stats - Update the various statistics counters.
1193 * @vsi: the VSI to be updated
1195 * Update the various stats for this VSI and its related entities.
1197 void i40e_update_stats(struct i40e_vsi *vsi)
1199 struct i40e_pf *pf = vsi->back;
1201 if (vsi == pf->vsi[pf->lan_vsi])
1202 i40e_update_pf_stats(pf);
1204 i40e_update_vsi_stats(vsi);
1205 #ifdef I40E_FCOE
1206 i40e_update_fcoe_stats(vsi);
1207 #endif
1211 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1212 * @vsi: the VSI to be searched
1213 * @macaddr: the MAC address
1214 * @vlan: the vlan
1215 * @is_vf: make sure its a VF filter, else doesn't matter
1216 * @is_netdev: make sure its a netdev filter, else doesn't matter
1218 * Returns ptr to the filter object or NULL
1220 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1221 u8 *macaddr, s16 vlan,
1222 bool is_vf, bool is_netdev)
1224 struct i40e_mac_filter *f;
1226 if (!vsi || !macaddr)
1227 return NULL;
1229 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1230 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1231 (vlan == f->vlan) &&
1232 (!is_vf || f->is_vf) &&
1233 (!is_netdev || f->is_netdev))
1234 return f;
1236 return NULL;
1240 * i40e_find_mac - Find a mac addr in the macvlan filters list
1241 * @vsi: the VSI to be searched
1242 * @macaddr: the MAC address we are searching for
1243 * @is_vf: make sure its a VF filter, else doesn't matter
1244 * @is_netdev: make sure its a netdev filter, else doesn't matter
1246 * Returns the first filter with the provided MAC address or NULL if
1247 * MAC address was not found
1249 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1250 bool is_vf, bool is_netdev)
1252 struct i40e_mac_filter *f;
1254 if (!vsi || !macaddr)
1255 return NULL;
1257 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1258 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1259 (!is_vf || f->is_vf) &&
1260 (!is_netdev || f->is_netdev))
1261 return f;
1263 return NULL;
1267 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1268 * @vsi: the VSI to be searched
1270 * Returns true if VSI is in vlan mode or false otherwise
1272 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1274 struct i40e_mac_filter *f;
1276 /* Only -1 for all the filters denotes not in vlan mode
1277 * so we have to go through all the list in order to make sure
1279 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1280 if (f->vlan >= 0 || vsi->info.pvid)
1281 return true;
1284 return false;
1288 * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1289 * @vsi: the VSI to be searched
1290 * @macaddr: the mac address to be filtered
1291 * @is_vf: true if it is a VF
1292 * @is_netdev: true if it is a netdev
1294 * Goes through all the macvlan filters and adds a
1295 * macvlan filter for each unique vlan that already exists
1297 * Returns first filter found on success, else NULL
1299 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1300 bool is_vf, bool is_netdev)
1302 struct i40e_mac_filter *f;
1304 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1305 if (vsi->info.pvid)
1306 f->vlan = le16_to_cpu(vsi->info.pvid);
1307 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1308 is_vf, is_netdev)) {
1309 if (!i40e_add_filter(vsi, macaddr, f->vlan,
1310 is_vf, is_netdev))
1311 return NULL;
1315 return list_first_entry_or_null(&vsi->mac_filter_list,
1316 struct i40e_mac_filter, list);
1320 * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1321 * @vsi: the PF Main VSI - inappropriate for any other VSI
1322 * @macaddr: the MAC address
1324 * Some older firmware configurations set up a default promiscuous VLAN
1325 * filter that needs to be removed.
1327 static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1329 struct i40e_aqc_remove_macvlan_element_data element;
1330 struct i40e_pf *pf = vsi->back;
1331 i40e_status ret;
1333 /* Only appropriate for the PF main VSI */
1334 if (vsi->type != I40E_VSI_MAIN)
1335 return -EINVAL;
1337 memset(&element, 0, sizeof(element));
1338 ether_addr_copy(element.mac_addr, macaddr);
1339 element.vlan_tag = 0;
1340 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1341 I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1342 ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1343 if (ret)
1344 return -ENOENT;
1346 return 0;
1350 * i40e_add_filter - Add a mac/vlan filter to the VSI
1351 * @vsi: the VSI to be searched
1352 * @macaddr: the MAC address
1353 * @vlan: the vlan
1354 * @is_vf: make sure its a VF filter, else doesn't matter
1355 * @is_netdev: make sure its a netdev filter, else doesn't matter
1357 * Returns ptr to the filter object or NULL when no memory available.
1359 * NOTE: This function is expected to be called with mac_filter_list_lock
1360 * being held.
1362 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1363 u8 *macaddr, s16 vlan,
1364 bool is_vf, bool is_netdev)
1366 struct i40e_mac_filter *f;
1368 if (!vsi || !macaddr)
1369 return NULL;
1371 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1372 if (!f) {
1373 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1374 if (!f)
1375 goto add_filter_out;
1377 ether_addr_copy(f->macaddr, macaddr);
1378 f->vlan = vlan;
1379 f->changed = true;
1381 INIT_LIST_HEAD(&f->list);
1382 list_add(&f->list, &vsi->mac_filter_list);
1385 /* increment counter and add a new flag if needed */
1386 if (is_vf) {
1387 if (!f->is_vf) {
1388 f->is_vf = true;
1389 f->counter++;
1391 } else if (is_netdev) {
1392 if (!f->is_netdev) {
1393 f->is_netdev = true;
1394 f->counter++;
1396 } else {
1397 f->counter++;
1400 /* changed tells sync_filters_subtask to
1401 * push the filter down to the firmware
1403 if (f->changed) {
1404 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1405 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1408 add_filter_out:
1409 return f;
1413 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1414 * @vsi: the VSI to be searched
1415 * @macaddr: the MAC address
1416 * @vlan: the vlan
1417 * @is_vf: make sure it's a VF filter, else doesn't matter
1418 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1420 * NOTE: This function is expected to be called with mac_filter_list_lock
1421 * being held.
1423 void i40e_del_filter(struct i40e_vsi *vsi,
1424 u8 *macaddr, s16 vlan,
1425 bool is_vf, bool is_netdev)
1427 struct i40e_mac_filter *f;
1429 if (!vsi || !macaddr)
1430 return;
1432 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1433 if (!f || f->counter == 0)
1434 return;
1436 if (is_vf) {
1437 if (f->is_vf) {
1438 f->is_vf = false;
1439 f->counter--;
1441 } else if (is_netdev) {
1442 if (f->is_netdev) {
1443 f->is_netdev = false;
1444 f->counter--;
1446 } else {
1447 /* make sure we don't remove a filter in use by VF or netdev */
1448 int min_f = 0;
1450 min_f += (f->is_vf ? 1 : 0);
1451 min_f += (f->is_netdev ? 1 : 0);
1453 if (f->counter > min_f)
1454 f->counter--;
1457 /* counter == 0 tells sync_filters_subtask to
1458 * remove the filter from the firmware's list
1460 if (f->counter == 0) {
1461 f->changed = true;
1462 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1463 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1468 * i40e_set_mac - NDO callback to set mac address
1469 * @netdev: network interface device structure
1470 * @p: pointer to an address structure
1472 * Returns 0 on success, negative on failure
1474 #ifdef I40E_FCOE
1475 int i40e_set_mac(struct net_device *netdev, void *p)
1476 #else
1477 static int i40e_set_mac(struct net_device *netdev, void *p)
1478 #endif
1480 struct i40e_netdev_priv *np = netdev_priv(netdev);
1481 struct i40e_vsi *vsi = np->vsi;
1482 struct i40e_pf *pf = vsi->back;
1483 struct i40e_hw *hw = &pf->hw;
1484 struct sockaddr *addr = p;
1485 struct i40e_mac_filter *f;
1487 if (!is_valid_ether_addr(addr->sa_data))
1488 return -EADDRNOTAVAIL;
1490 if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1491 netdev_info(netdev, "already using mac address %pM\n",
1492 addr->sa_data);
1493 return 0;
1496 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1497 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1498 return -EADDRNOTAVAIL;
1500 if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1501 netdev_info(netdev, "returning to hw mac address %pM\n",
1502 hw->mac.addr);
1503 else
1504 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1506 if (vsi->type == I40E_VSI_MAIN) {
1507 i40e_status ret;
1509 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1510 I40E_AQC_WRITE_TYPE_LAA_WOL,
1511 addr->sa_data, NULL);
1512 if (ret) {
1513 netdev_info(netdev,
1514 "Addr change for Main VSI failed: %d\n",
1515 ret);
1516 return -EADDRNOTAVAIL;
1520 if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1521 struct i40e_aqc_remove_macvlan_element_data element;
1523 memset(&element, 0, sizeof(element));
1524 ether_addr_copy(element.mac_addr, netdev->dev_addr);
1525 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1526 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1527 } else {
1528 spin_lock_bh(&vsi->mac_filter_list_lock);
1529 i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1530 false, false);
1531 spin_unlock_bh(&vsi->mac_filter_list_lock);
1534 if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1535 struct i40e_aqc_add_macvlan_element_data element;
1537 memset(&element, 0, sizeof(element));
1538 ether_addr_copy(element.mac_addr, hw->mac.addr);
1539 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1540 i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1541 } else {
1542 spin_lock_bh(&vsi->mac_filter_list_lock);
1543 f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1544 false, false);
1545 if (f)
1546 f->is_laa = true;
1547 spin_unlock_bh(&vsi->mac_filter_list_lock);
1550 i40e_sync_vsi_filters(vsi, false);
1551 ether_addr_copy(netdev->dev_addr, addr->sa_data);
1553 return 0;
1557 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1558 * @vsi: the VSI being setup
1559 * @ctxt: VSI context structure
1560 * @enabled_tc: Enabled TCs bitmap
1561 * @is_add: True if called before Add VSI
1563 * Setup VSI queue mapping for enabled traffic classes.
1565 #ifdef I40E_FCOE
1566 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1567 struct i40e_vsi_context *ctxt,
1568 u8 enabled_tc,
1569 bool is_add)
1570 #else
1571 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1572 struct i40e_vsi_context *ctxt,
1573 u8 enabled_tc,
1574 bool is_add)
1575 #endif
1577 struct i40e_pf *pf = vsi->back;
1578 u16 sections = 0;
1579 u8 netdev_tc = 0;
1580 u16 numtc = 0;
1581 u16 qcount;
1582 u8 offset;
1583 u16 qmap;
1584 int i;
1585 u16 num_tc_qps = 0;
1587 sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1588 offset = 0;
1590 if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1591 /* Find numtc from enabled TC bitmap */
1592 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1593 if (enabled_tc & BIT_ULL(i)) /* TC is enabled */
1594 numtc++;
1596 if (!numtc) {
1597 dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1598 numtc = 1;
1600 } else {
1601 /* At least TC0 is enabled in case of non-DCB case */
1602 numtc = 1;
1605 vsi->tc_config.numtc = numtc;
1606 vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1607 /* Number of queues per enabled TC */
1608 /* In MFP case we can have a much lower count of MSIx
1609 * vectors available and so we need to lower the used
1610 * q count.
1612 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1613 qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1614 else
1615 qcount = vsi->alloc_queue_pairs;
1616 num_tc_qps = qcount / numtc;
1617 num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1619 /* Setup queue offset/count for all TCs for given VSI */
1620 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1621 /* See if the given TC is enabled for the given VSI */
1622 if (vsi->tc_config.enabled_tc & BIT_ULL(i)) {
1623 /* TC is enabled */
1624 int pow, num_qps;
1626 switch (vsi->type) {
1627 case I40E_VSI_MAIN:
1628 qcount = min_t(int, pf->rss_size, num_tc_qps);
1629 break;
1630 #ifdef I40E_FCOE
1631 case I40E_VSI_FCOE:
1632 qcount = num_tc_qps;
1633 break;
1634 #endif
1635 case I40E_VSI_FDIR:
1636 case I40E_VSI_SRIOV:
1637 case I40E_VSI_VMDQ2:
1638 default:
1639 qcount = num_tc_qps;
1640 WARN_ON(i != 0);
1641 break;
1643 vsi->tc_config.tc_info[i].qoffset = offset;
1644 vsi->tc_config.tc_info[i].qcount = qcount;
1646 /* find the next higher power-of-2 of num queue pairs */
1647 num_qps = qcount;
1648 pow = 0;
1649 while (num_qps && (BIT_ULL(pow) < qcount)) {
1650 pow++;
1651 num_qps >>= 1;
1654 vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1655 qmap =
1656 (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1657 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1659 offset += qcount;
1660 } else {
1661 /* TC is not enabled so set the offset to
1662 * default queue and allocate one queue
1663 * for the given TC.
1665 vsi->tc_config.tc_info[i].qoffset = 0;
1666 vsi->tc_config.tc_info[i].qcount = 1;
1667 vsi->tc_config.tc_info[i].netdev_tc = 0;
1669 qmap = 0;
1671 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1674 /* Set actual Tx/Rx queue pairs */
1675 vsi->num_queue_pairs = offset;
1676 if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1677 if (vsi->req_queue_pairs > 0)
1678 vsi->num_queue_pairs = vsi->req_queue_pairs;
1679 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1680 vsi->num_queue_pairs = pf->num_lan_msix;
1683 /* Scheduler section valid can only be set for ADD VSI */
1684 if (is_add) {
1685 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1687 ctxt->info.up_enable_bits = enabled_tc;
1689 if (vsi->type == I40E_VSI_SRIOV) {
1690 ctxt->info.mapping_flags |=
1691 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1692 for (i = 0; i < vsi->num_queue_pairs; i++)
1693 ctxt->info.queue_mapping[i] =
1694 cpu_to_le16(vsi->base_queue + i);
1695 } else {
1696 ctxt->info.mapping_flags |=
1697 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1698 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1700 ctxt->info.valid_sections |= cpu_to_le16(sections);
1704 * i40e_set_rx_mode - NDO callback to set the netdev filters
1705 * @netdev: network interface device structure
1707 #ifdef I40E_FCOE
1708 void i40e_set_rx_mode(struct net_device *netdev)
1709 #else
1710 static void i40e_set_rx_mode(struct net_device *netdev)
1711 #endif
1713 struct i40e_netdev_priv *np = netdev_priv(netdev);
1714 struct i40e_mac_filter *f, *ftmp;
1715 struct i40e_vsi *vsi = np->vsi;
1716 struct netdev_hw_addr *uca;
1717 struct netdev_hw_addr *mca;
1718 struct netdev_hw_addr *ha;
1720 spin_lock_bh(&vsi->mac_filter_list_lock);
1722 /* add addr if not already in the filter list */
1723 netdev_for_each_uc_addr(uca, netdev) {
1724 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1725 if (i40e_is_vsi_in_vlan(vsi))
1726 i40e_put_mac_in_vlan(vsi, uca->addr,
1727 false, true);
1728 else
1729 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1730 false, true);
1734 netdev_for_each_mc_addr(mca, netdev) {
1735 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1736 if (i40e_is_vsi_in_vlan(vsi))
1737 i40e_put_mac_in_vlan(vsi, mca->addr,
1738 false, true);
1739 else
1740 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1741 false, true);
1745 /* remove filter if not in netdev list */
1746 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1748 if (!f->is_netdev)
1749 continue;
1751 netdev_for_each_mc_addr(mca, netdev)
1752 if (ether_addr_equal(mca->addr, f->macaddr))
1753 goto bottom_of_search_loop;
1755 netdev_for_each_uc_addr(uca, netdev)
1756 if (ether_addr_equal(uca->addr, f->macaddr))
1757 goto bottom_of_search_loop;
1759 for_each_dev_addr(netdev, ha)
1760 if (ether_addr_equal(ha->addr, f->macaddr))
1761 goto bottom_of_search_loop;
1763 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1764 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1766 bottom_of_search_loop:
1767 continue;
1769 spin_unlock_bh(&vsi->mac_filter_list_lock);
1771 /* check for other flag changes */
1772 if (vsi->current_netdev_flags != vsi->netdev->flags) {
1773 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1774 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1779 * i40e_mac_filter_entry_clone - Clones a MAC filter entry
1780 * @src: source MAC filter entry to be clones
1782 * Returns the pointer to newly cloned MAC filter entry or NULL
1783 * in case of error
1785 static struct i40e_mac_filter *i40e_mac_filter_entry_clone(
1786 struct i40e_mac_filter *src)
1788 struct i40e_mac_filter *f;
1790 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1791 if (!f)
1792 return NULL;
1793 *f = *src;
1795 INIT_LIST_HEAD(&f->list);
1797 return f;
1801 * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1802 * @vsi: pointer to vsi struct
1803 * @from: Pointer to list which contains MAC filter entries - changes to
1804 * those entries needs to be undone.
1806 * MAC filter entries from list were slated to be removed from device.
1808 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1809 struct list_head *from)
1811 struct i40e_mac_filter *f, *ftmp;
1813 list_for_each_entry_safe(f, ftmp, from, list) {
1814 f->changed = true;
1815 /* Move the element back into MAC filter list*/
1816 list_move_tail(&f->list, &vsi->mac_filter_list);
1821 * i40e_undo_add_filter_entries - Undo the changes made to MAC filter entries
1822 * @vsi: pointer to vsi struct
1824 * MAC filter entries from list were slated to be added from device.
1826 static void i40e_undo_add_filter_entries(struct i40e_vsi *vsi)
1828 struct i40e_mac_filter *f, *ftmp;
1830 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1831 if (!f->changed && f->counter)
1832 f->changed = true;
1837 * i40e_cleanup_add_list - Deletes the element from add list and release
1838 * memory
1839 * @add_list: Pointer to list which contains MAC filter entries
1841 static void i40e_cleanup_add_list(struct list_head *add_list)
1843 struct i40e_mac_filter *f, *ftmp;
1845 list_for_each_entry_safe(f, ftmp, add_list, list) {
1846 list_del(&f->list);
1847 kfree(f);
1852 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1853 * @vsi: ptr to the VSI
1854 * @grab_rtnl: whether RTNL needs to be grabbed
1856 * Push any outstanding VSI filter changes through the AdminQ.
1858 * Returns 0 or error value
1860 int i40e_sync_vsi_filters(struct i40e_vsi *vsi, bool grab_rtnl)
1862 struct list_head tmp_del_list, tmp_add_list;
1863 struct i40e_mac_filter *f, *ftmp, *fclone;
1864 bool promisc_forced_on = false;
1865 bool add_happened = false;
1866 int filter_list_len = 0;
1867 u32 changed_flags = 0;
1868 bool err_cond = false;
1869 i40e_status ret = 0;
1870 struct i40e_pf *pf;
1871 int num_add = 0;
1872 int num_del = 0;
1873 int aq_err = 0;
1874 u16 cmd_flags;
1876 /* empty array typed pointers, kcalloc later */
1877 struct i40e_aqc_add_macvlan_element_data *add_list;
1878 struct i40e_aqc_remove_macvlan_element_data *del_list;
1880 while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1881 usleep_range(1000, 2000);
1882 pf = vsi->back;
1884 if (vsi->netdev) {
1885 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1886 vsi->current_netdev_flags = vsi->netdev->flags;
1889 INIT_LIST_HEAD(&tmp_del_list);
1890 INIT_LIST_HEAD(&tmp_add_list);
1892 if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1893 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1895 spin_lock_bh(&vsi->mac_filter_list_lock);
1896 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1897 if (!f->changed)
1898 continue;
1900 if (f->counter != 0)
1901 continue;
1902 f->changed = false;
1904 /* Move the element into temporary del_list */
1905 list_move_tail(&f->list, &tmp_del_list);
1908 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1909 if (!f->changed)
1910 continue;
1912 if (f->counter == 0)
1913 continue;
1914 f->changed = false;
1916 /* Clone MAC filter entry and add into temporary list */
1917 fclone = i40e_mac_filter_entry_clone(f);
1918 if (!fclone) {
1919 err_cond = true;
1920 break;
1922 list_add_tail(&fclone->list, &tmp_add_list);
1925 /* if failed to clone MAC filter entry - undo */
1926 if (err_cond) {
1927 i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1928 i40e_undo_add_filter_entries(vsi);
1930 spin_unlock_bh(&vsi->mac_filter_list_lock);
1932 if (err_cond)
1933 i40e_cleanup_add_list(&tmp_add_list);
1936 /* Now process 'del_list' outside the lock */
1937 if (!list_empty(&tmp_del_list)) {
1938 filter_list_len = pf->hw.aq.asq_buf_size /
1939 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1940 del_list = kcalloc(filter_list_len,
1941 sizeof(struct i40e_aqc_remove_macvlan_element_data),
1942 GFP_KERNEL);
1943 if (!del_list) {
1944 i40e_cleanup_add_list(&tmp_add_list);
1946 /* Undo VSI's MAC filter entry element updates */
1947 spin_lock_bh(&vsi->mac_filter_list_lock);
1948 i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1949 i40e_undo_add_filter_entries(vsi);
1950 spin_unlock_bh(&vsi->mac_filter_list_lock);
1951 return -ENOMEM;
1954 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1955 cmd_flags = 0;
1957 /* add to delete list */
1958 ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1959 del_list[num_del].vlan_tag =
1960 cpu_to_le16((u16)(f->vlan ==
1961 I40E_VLAN_ANY ? 0 : f->vlan));
1963 cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1964 del_list[num_del].flags = cmd_flags;
1965 num_del++;
1967 /* flush a full buffer */
1968 if (num_del == filter_list_len) {
1969 ret = i40e_aq_remove_macvlan(&pf->hw,
1970 vsi->seid, del_list, num_del,
1971 NULL);
1972 aq_err = pf->hw.aq.asq_last_status;
1973 num_del = 0;
1974 memset(del_list, 0, sizeof(*del_list));
1976 if (ret && aq_err != I40E_AQ_RC_ENOENT)
1977 dev_err(&pf->pdev->dev,
1978 "ignoring delete macvlan error, err %s, aq_err %s while flushing a full buffer\n",
1979 i40e_stat_str(&pf->hw, ret),
1980 i40e_aq_str(&pf->hw, aq_err));
1982 /* Release memory for MAC filter entries which were
1983 * synced up with HW.
1985 list_del(&f->list);
1986 kfree(f);
1989 if (num_del) {
1990 ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1991 del_list, num_del, NULL);
1992 aq_err = pf->hw.aq.asq_last_status;
1993 num_del = 0;
1995 if (ret && aq_err != I40E_AQ_RC_ENOENT)
1996 dev_info(&pf->pdev->dev,
1997 "ignoring delete macvlan error, err %s aq_err %s\n",
1998 i40e_stat_str(&pf->hw, ret),
1999 i40e_aq_str(&pf->hw, aq_err));
2002 kfree(del_list);
2003 del_list = NULL;
2006 if (!list_empty(&tmp_add_list)) {
2008 /* do all the adds now */
2009 filter_list_len = pf->hw.aq.asq_buf_size /
2010 sizeof(struct i40e_aqc_add_macvlan_element_data),
2011 add_list = kcalloc(filter_list_len,
2012 sizeof(struct i40e_aqc_add_macvlan_element_data),
2013 GFP_KERNEL);
2014 if (!add_list) {
2015 /* Purge element from temporary lists */
2016 i40e_cleanup_add_list(&tmp_add_list);
2018 /* Undo add filter entries from VSI MAC filter list */
2019 spin_lock_bh(&vsi->mac_filter_list_lock);
2020 i40e_undo_add_filter_entries(vsi);
2021 spin_unlock_bh(&vsi->mac_filter_list_lock);
2022 return -ENOMEM;
2025 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2027 add_happened = true;
2028 cmd_flags = 0;
2030 /* add to add array */
2031 ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
2032 add_list[num_add].vlan_tag =
2033 cpu_to_le16(
2034 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
2035 add_list[num_add].queue_number = 0;
2037 cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2038 add_list[num_add].flags = cpu_to_le16(cmd_flags);
2039 num_add++;
2041 /* flush a full buffer */
2042 if (num_add == filter_list_len) {
2043 ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2044 add_list, num_add,
2045 NULL);
2046 aq_err = pf->hw.aq.asq_last_status;
2047 num_add = 0;
2049 if (ret)
2050 break;
2051 memset(add_list, 0, sizeof(*add_list));
2053 /* Entries from tmp_add_list were cloned from MAC
2054 * filter list, hence clean those cloned entries
2056 list_del(&f->list);
2057 kfree(f);
2060 if (num_add) {
2061 ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
2062 add_list, num_add, NULL);
2063 aq_err = pf->hw.aq.asq_last_status;
2064 num_add = 0;
2066 kfree(add_list);
2067 add_list = NULL;
2069 if (add_happened && ret && aq_err != I40E_AQ_RC_EINVAL) {
2070 dev_info(&pf->pdev->dev,
2071 "add filter failed, err %s aq_err %s\n",
2072 i40e_stat_str(&pf->hw, ret),
2073 i40e_aq_str(&pf->hw, aq_err));
2074 if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
2075 !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2076 &vsi->state)) {
2077 promisc_forced_on = true;
2078 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2079 &vsi->state);
2080 dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
2085 /* check for changes in promiscuous modes */
2086 if (changed_flags & IFF_ALLMULTI) {
2087 bool cur_multipromisc;
2089 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2090 ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2091 vsi->seid,
2092 cur_multipromisc,
2093 NULL);
2094 if (ret)
2095 dev_info(&pf->pdev->dev,
2096 "set multi promisc failed, err %s aq_err %s\n",
2097 i40e_stat_str(&pf->hw, ret),
2098 i40e_aq_str(&pf->hw,
2099 pf->hw.aq.asq_last_status));
2101 if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
2102 bool cur_promisc;
2104 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2105 test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2106 &vsi->state));
2107 if (vsi->type == I40E_VSI_MAIN && pf->lan_veb != I40E_NO_VEB) {
2108 /* set defport ON for Main VSI instead of true promisc
2109 * this way we will get all unicast/multicast and VLAN
2110 * promisc behavior but will not get VF or VMDq traffic
2111 * replicated on the Main VSI.
2113 if (pf->cur_promisc != cur_promisc) {
2114 pf->cur_promisc = cur_promisc;
2115 if (grab_rtnl)
2116 i40e_do_reset_safe(pf,
2117 BIT(__I40E_PF_RESET_REQUESTED));
2118 else
2119 i40e_do_reset(pf,
2120 BIT(__I40E_PF_RESET_REQUESTED));
2122 } else {
2123 ret = i40e_aq_set_vsi_unicast_promiscuous(
2124 &vsi->back->hw,
2125 vsi->seid,
2126 cur_promisc, NULL);
2127 if (ret)
2128 dev_info(&pf->pdev->dev,
2129 "set unicast promisc failed, err %d, aq_err %d\n",
2130 ret, pf->hw.aq.asq_last_status);
2131 ret = i40e_aq_set_vsi_multicast_promiscuous(
2132 &vsi->back->hw,
2133 vsi->seid,
2134 cur_promisc, NULL);
2135 if (ret)
2136 dev_info(&pf->pdev->dev,
2137 "set multicast promisc failed, err %d, aq_err %d\n",
2138 ret, pf->hw.aq.asq_last_status);
2140 ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2141 vsi->seid,
2142 cur_promisc, NULL);
2143 if (ret)
2144 dev_info(&pf->pdev->dev,
2145 "set brdcast promisc failed, err %s, aq_err %s\n",
2146 i40e_stat_str(&pf->hw, ret),
2147 i40e_aq_str(&pf->hw,
2148 pf->hw.aq.asq_last_status));
2151 clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2152 return 0;
2156 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2157 * @pf: board private structure
2159 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2161 int v;
2163 if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2164 return;
2165 pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2167 for (v = 0; v < pf->num_alloc_vsi; v++) {
2168 if (pf->vsi[v] &&
2169 (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
2170 i40e_sync_vsi_filters(pf->vsi[v], true);
2175 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2176 * @netdev: network interface device structure
2177 * @new_mtu: new value for maximum frame size
2179 * Returns 0 on success, negative on failure
2181 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2183 struct i40e_netdev_priv *np = netdev_priv(netdev);
2184 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2185 struct i40e_vsi *vsi = np->vsi;
2187 /* MTU < 68 is an error and causes problems on some kernels */
2188 if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2189 return -EINVAL;
2191 netdev_info(netdev, "changing MTU from %d to %d\n",
2192 netdev->mtu, new_mtu);
2193 netdev->mtu = new_mtu;
2194 if (netif_running(netdev))
2195 i40e_vsi_reinit_locked(vsi);
2197 return 0;
2201 * i40e_ioctl - Access the hwtstamp interface
2202 * @netdev: network interface device structure
2203 * @ifr: interface request data
2204 * @cmd: ioctl command
2206 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2208 struct i40e_netdev_priv *np = netdev_priv(netdev);
2209 struct i40e_pf *pf = np->vsi->back;
2211 switch (cmd) {
2212 case SIOCGHWTSTAMP:
2213 return i40e_ptp_get_ts_config(pf, ifr);
2214 case SIOCSHWTSTAMP:
2215 return i40e_ptp_set_ts_config(pf, ifr);
2216 default:
2217 return -EOPNOTSUPP;
2222 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2223 * @vsi: the vsi being adjusted
2225 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2227 struct i40e_vsi_context ctxt;
2228 i40e_status ret;
2230 if ((vsi->info.valid_sections &
2231 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2232 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2233 return; /* already enabled */
2235 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2236 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2237 I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2239 ctxt.seid = vsi->seid;
2240 ctxt.info = vsi->info;
2241 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2242 if (ret) {
2243 dev_info(&vsi->back->pdev->dev,
2244 "update vlan stripping failed, err %s aq_err %s\n",
2245 i40e_stat_str(&vsi->back->hw, ret),
2246 i40e_aq_str(&vsi->back->hw,
2247 vsi->back->hw.aq.asq_last_status));
2252 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2253 * @vsi: the vsi being adjusted
2255 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2257 struct i40e_vsi_context ctxt;
2258 i40e_status ret;
2260 if ((vsi->info.valid_sections &
2261 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2262 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2263 I40E_AQ_VSI_PVLAN_EMOD_MASK))
2264 return; /* already disabled */
2266 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2267 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2268 I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2270 ctxt.seid = vsi->seid;
2271 ctxt.info = vsi->info;
2272 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2273 if (ret) {
2274 dev_info(&vsi->back->pdev->dev,
2275 "update vlan stripping failed, err %s aq_err %s\n",
2276 i40e_stat_str(&vsi->back->hw, ret),
2277 i40e_aq_str(&vsi->back->hw,
2278 vsi->back->hw.aq.asq_last_status));
2283 * i40e_vlan_rx_register - Setup or shutdown vlan offload
2284 * @netdev: network interface to be adjusted
2285 * @features: netdev features to test if VLAN offload is enabled or not
2287 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2289 struct i40e_netdev_priv *np = netdev_priv(netdev);
2290 struct i40e_vsi *vsi = np->vsi;
2292 if (features & NETIF_F_HW_VLAN_CTAG_RX)
2293 i40e_vlan_stripping_enable(vsi);
2294 else
2295 i40e_vlan_stripping_disable(vsi);
2299 * i40e_vsi_add_vlan - Add vsi membership for given vlan
2300 * @vsi: the vsi being configured
2301 * @vid: vlan id to be added (0 = untagged only , -1 = any)
2303 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2305 struct i40e_mac_filter *f, *add_f;
2306 bool is_netdev, is_vf;
2308 is_vf = (vsi->type == I40E_VSI_SRIOV);
2309 is_netdev = !!(vsi->netdev);
2311 /* Locked once because all functions invoked below iterates list*/
2312 spin_lock_bh(&vsi->mac_filter_list_lock);
2314 if (is_netdev) {
2315 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2316 is_vf, is_netdev);
2317 if (!add_f) {
2318 dev_info(&vsi->back->pdev->dev,
2319 "Could not add vlan filter %d for %pM\n",
2320 vid, vsi->netdev->dev_addr);
2321 spin_unlock_bh(&vsi->mac_filter_list_lock);
2322 return -ENOMEM;
2326 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2327 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2328 if (!add_f) {
2329 dev_info(&vsi->back->pdev->dev,
2330 "Could not add vlan filter %d for %pM\n",
2331 vid, f->macaddr);
2332 spin_unlock_bh(&vsi->mac_filter_list_lock);
2333 return -ENOMEM;
2337 /* Now if we add a vlan tag, make sure to check if it is the first
2338 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2339 * with 0, so we now accept untagged and specified tagged traffic
2340 * (and not any taged and untagged)
2342 if (vid > 0) {
2343 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2344 I40E_VLAN_ANY,
2345 is_vf, is_netdev)) {
2346 i40e_del_filter(vsi, vsi->netdev->dev_addr,
2347 I40E_VLAN_ANY, is_vf, is_netdev);
2348 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2349 is_vf, is_netdev);
2350 if (!add_f) {
2351 dev_info(&vsi->back->pdev->dev,
2352 "Could not add filter 0 for %pM\n",
2353 vsi->netdev->dev_addr);
2354 spin_unlock_bh(&vsi->mac_filter_list_lock);
2355 return -ENOMEM;
2360 /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2361 if (vid > 0 && !vsi->info.pvid) {
2362 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2363 if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2364 is_vf, is_netdev))
2365 continue;
2366 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2367 is_vf, is_netdev);
2368 add_f = i40e_add_filter(vsi, f->macaddr,
2369 0, is_vf, is_netdev);
2370 if (!add_f) {
2371 dev_info(&vsi->back->pdev->dev,
2372 "Could not add filter 0 for %pM\n",
2373 f->macaddr);
2374 spin_unlock_bh(&vsi->mac_filter_list_lock);
2375 return -ENOMEM;
2380 /* Make sure to release before sync_vsi_filter because that
2381 * function will lock/unlock as necessary
2383 spin_unlock_bh(&vsi->mac_filter_list_lock);
2385 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2386 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2387 return 0;
2389 return i40e_sync_vsi_filters(vsi, false);
2393 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2394 * @vsi: the vsi being configured
2395 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2397 * Return: 0 on success or negative otherwise
2399 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2401 struct net_device *netdev = vsi->netdev;
2402 struct i40e_mac_filter *f, *add_f;
2403 bool is_vf, is_netdev;
2404 int filter_count = 0;
2406 is_vf = (vsi->type == I40E_VSI_SRIOV);
2407 is_netdev = !!(netdev);
2409 /* Locked once because all functions invoked below iterates list */
2410 spin_lock_bh(&vsi->mac_filter_list_lock);
2412 if (is_netdev)
2413 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2415 list_for_each_entry(f, &vsi->mac_filter_list, list)
2416 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2418 /* go through all the filters for this VSI and if there is only
2419 * vid == 0 it means there are no other filters, so vid 0 must
2420 * be replaced with -1. This signifies that we should from now
2421 * on accept any traffic (with any tag present, or untagged)
2423 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2424 if (is_netdev) {
2425 if (f->vlan &&
2426 ether_addr_equal(netdev->dev_addr, f->macaddr))
2427 filter_count++;
2430 if (f->vlan)
2431 filter_count++;
2434 if (!filter_count && is_netdev) {
2435 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2436 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2437 is_vf, is_netdev);
2438 if (!f) {
2439 dev_info(&vsi->back->pdev->dev,
2440 "Could not add filter %d for %pM\n",
2441 I40E_VLAN_ANY, netdev->dev_addr);
2442 spin_unlock_bh(&vsi->mac_filter_list_lock);
2443 return -ENOMEM;
2447 if (!filter_count) {
2448 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2449 i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2450 add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2451 is_vf, is_netdev);
2452 if (!add_f) {
2453 dev_info(&vsi->back->pdev->dev,
2454 "Could not add filter %d for %pM\n",
2455 I40E_VLAN_ANY, f->macaddr);
2456 spin_unlock_bh(&vsi->mac_filter_list_lock);
2457 return -ENOMEM;
2462 /* Make sure to release before sync_vsi_filter because that
2463 * function with lock/unlock as necessary
2465 spin_unlock_bh(&vsi->mac_filter_list_lock);
2467 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2468 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2469 return 0;
2471 return i40e_sync_vsi_filters(vsi, false);
2475 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2476 * @netdev: network interface to be adjusted
2477 * @vid: vlan id to be added
2479 * net_device_ops implementation for adding vlan ids
2481 #ifdef I40E_FCOE
2482 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2483 __always_unused __be16 proto, u16 vid)
2484 #else
2485 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2486 __always_unused __be16 proto, u16 vid)
2487 #endif
2489 struct i40e_netdev_priv *np = netdev_priv(netdev);
2490 struct i40e_vsi *vsi = np->vsi;
2491 int ret = 0;
2493 if (vid > 4095)
2494 return -EINVAL;
2496 netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2498 /* If the network stack called us with vid = 0 then
2499 * it is asking to receive priority tagged packets with
2500 * vlan id 0. Our HW receives them by default when configured
2501 * to receive untagged packets so there is no need to add an
2502 * extra filter for vlan 0 tagged packets.
2504 if (vid)
2505 ret = i40e_vsi_add_vlan(vsi, vid);
2507 if (!ret && (vid < VLAN_N_VID))
2508 set_bit(vid, vsi->active_vlans);
2510 return ret;
2514 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2515 * @netdev: network interface to be adjusted
2516 * @vid: vlan id to be removed
2518 * net_device_ops implementation for removing vlan ids
2520 #ifdef I40E_FCOE
2521 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2522 __always_unused __be16 proto, u16 vid)
2523 #else
2524 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2525 __always_unused __be16 proto, u16 vid)
2526 #endif
2528 struct i40e_netdev_priv *np = netdev_priv(netdev);
2529 struct i40e_vsi *vsi = np->vsi;
2531 netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2533 /* return code is ignored as there is nothing a user
2534 * can do about failure to remove and a log message was
2535 * already printed from the other function
2537 i40e_vsi_kill_vlan(vsi, vid);
2539 clear_bit(vid, vsi->active_vlans);
2541 return 0;
2545 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2546 * @vsi: the vsi being brought back up
2548 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2550 u16 vid;
2552 if (!vsi->netdev)
2553 return;
2555 i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2557 for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2558 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2559 vid);
2563 * i40e_vsi_add_pvid - Add pvid for the VSI
2564 * @vsi: the vsi being adjusted
2565 * @vid: the vlan id to set as a PVID
2567 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2569 struct i40e_vsi_context ctxt;
2570 i40e_status ret;
2572 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2573 vsi->info.pvid = cpu_to_le16(vid);
2574 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2575 I40E_AQ_VSI_PVLAN_INSERT_PVID |
2576 I40E_AQ_VSI_PVLAN_EMOD_STR;
2578 ctxt.seid = vsi->seid;
2579 ctxt.info = vsi->info;
2580 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2581 if (ret) {
2582 dev_info(&vsi->back->pdev->dev,
2583 "add pvid failed, err %s aq_err %s\n",
2584 i40e_stat_str(&vsi->back->hw, ret),
2585 i40e_aq_str(&vsi->back->hw,
2586 vsi->back->hw.aq.asq_last_status));
2587 return -ENOENT;
2590 return 0;
2594 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2595 * @vsi: the vsi being adjusted
2597 * Just use the vlan_rx_register() service to put it back to normal
2599 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2601 i40e_vlan_stripping_disable(vsi);
2603 vsi->info.pvid = 0;
2607 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2608 * @vsi: ptr to the VSI
2610 * If this function returns with an error, then it's possible one or
2611 * more of the rings is populated (while the rest are not). It is the
2612 * callers duty to clean those orphaned rings.
2614 * Return 0 on success, negative on failure
2616 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2618 int i, err = 0;
2620 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2621 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2623 return err;
2627 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2628 * @vsi: ptr to the VSI
2630 * Free VSI's transmit software resources
2632 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2634 int i;
2636 if (!vsi->tx_rings)
2637 return;
2639 for (i = 0; i < vsi->num_queue_pairs; i++)
2640 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2641 i40e_free_tx_resources(vsi->tx_rings[i]);
2645 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2646 * @vsi: ptr to the VSI
2648 * If this function returns with an error, then it's possible one or
2649 * more of the rings is populated (while the rest are not). It is the
2650 * callers duty to clean those orphaned rings.
2652 * Return 0 on success, negative on failure
2654 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2656 int i, err = 0;
2658 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2659 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2660 #ifdef I40E_FCOE
2661 i40e_fcoe_setup_ddp_resources(vsi);
2662 #endif
2663 return err;
2667 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2668 * @vsi: ptr to the VSI
2670 * Free all receive software resources
2672 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2674 int i;
2676 if (!vsi->rx_rings)
2677 return;
2679 for (i = 0; i < vsi->num_queue_pairs; i++)
2680 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2681 i40e_free_rx_resources(vsi->rx_rings[i]);
2682 #ifdef I40E_FCOE
2683 i40e_fcoe_free_ddp_resources(vsi);
2684 #endif
2688 * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2689 * @ring: The Tx ring to configure
2691 * This enables/disables XPS for a given Tx descriptor ring
2692 * based on the TCs enabled for the VSI that ring belongs to.
2694 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2696 struct i40e_vsi *vsi = ring->vsi;
2697 cpumask_var_t mask;
2699 if (!ring->q_vector || !ring->netdev)
2700 return;
2702 /* Single TC mode enable XPS */
2703 if (vsi->tc_config.numtc <= 1) {
2704 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2705 netif_set_xps_queue(ring->netdev,
2706 &ring->q_vector->affinity_mask,
2707 ring->queue_index);
2708 } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2709 /* Disable XPS to allow selection based on TC */
2710 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2711 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2712 free_cpumask_var(mask);
2717 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2718 * @ring: The Tx ring to configure
2720 * Configure the Tx descriptor ring in the HMC context.
2722 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2724 struct i40e_vsi *vsi = ring->vsi;
2725 u16 pf_q = vsi->base_queue + ring->queue_index;
2726 struct i40e_hw *hw = &vsi->back->hw;
2727 struct i40e_hmc_obj_txq tx_ctx;
2728 i40e_status err = 0;
2729 u32 qtx_ctl = 0;
2731 /* some ATR related tx ring init */
2732 if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2733 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2734 ring->atr_count = 0;
2735 } else {
2736 ring->atr_sample_rate = 0;
2739 /* configure XPS */
2740 i40e_config_xps_tx_ring(ring);
2742 /* clear the context structure first */
2743 memset(&tx_ctx, 0, sizeof(tx_ctx));
2745 tx_ctx.new_context = 1;
2746 tx_ctx.base = (ring->dma / 128);
2747 tx_ctx.qlen = ring->count;
2748 tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2749 I40E_FLAG_FD_ATR_ENABLED));
2750 #ifdef I40E_FCOE
2751 tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2752 #endif
2753 tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2754 /* FDIR VSI tx ring can still use RS bit and writebacks */
2755 if (vsi->type != I40E_VSI_FDIR)
2756 tx_ctx.head_wb_ena = 1;
2757 tx_ctx.head_wb_addr = ring->dma +
2758 (ring->count * sizeof(struct i40e_tx_desc));
2760 /* As part of VSI creation/update, FW allocates certain
2761 * Tx arbitration queue sets for each TC enabled for
2762 * the VSI. The FW returns the handles to these queue
2763 * sets as part of the response buffer to Add VSI,
2764 * Update VSI, etc. AQ commands. It is expected that
2765 * these queue set handles be associated with the Tx
2766 * queues by the driver as part of the TX queue context
2767 * initialization. This has to be done regardless of
2768 * DCB as by default everything is mapped to TC0.
2770 tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2771 tx_ctx.rdylist_act = 0;
2773 /* clear the context in the HMC */
2774 err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2775 if (err) {
2776 dev_info(&vsi->back->pdev->dev,
2777 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2778 ring->queue_index, pf_q, err);
2779 return -ENOMEM;
2782 /* set the context in the HMC */
2783 err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2784 if (err) {
2785 dev_info(&vsi->back->pdev->dev,
2786 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2787 ring->queue_index, pf_q, err);
2788 return -ENOMEM;
2791 /* Now associate this queue with this PCI function */
2792 if (vsi->type == I40E_VSI_VMDQ2) {
2793 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2794 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2795 I40E_QTX_CTL_VFVM_INDX_MASK;
2796 } else {
2797 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2800 qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2801 I40E_QTX_CTL_PF_INDX_MASK);
2802 wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2803 i40e_flush(hw);
2805 /* cache tail off for easier writes later */
2806 ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2808 return 0;
2812 * i40e_configure_rx_ring - Configure a receive ring context
2813 * @ring: The Rx ring to configure
2815 * Configure the Rx descriptor ring in the HMC context.
2817 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2819 struct i40e_vsi *vsi = ring->vsi;
2820 u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2821 u16 pf_q = vsi->base_queue + ring->queue_index;
2822 struct i40e_hw *hw = &vsi->back->hw;
2823 struct i40e_hmc_obj_rxq rx_ctx;
2824 i40e_status err = 0;
2826 ring->state = 0;
2828 /* clear the context structure first */
2829 memset(&rx_ctx, 0, sizeof(rx_ctx));
2831 ring->rx_buf_len = vsi->rx_buf_len;
2832 ring->rx_hdr_len = vsi->rx_hdr_len;
2834 rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2835 rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2837 rx_ctx.base = (ring->dma / 128);
2838 rx_ctx.qlen = ring->count;
2840 if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2841 set_ring_16byte_desc_enabled(ring);
2842 rx_ctx.dsize = 0;
2843 } else {
2844 rx_ctx.dsize = 1;
2847 rx_ctx.dtype = vsi->dtype;
2848 if (vsi->dtype) {
2849 set_ring_ps_enabled(ring);
2850 rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2 |
2851 I40E_RX_SPLIT_IP |
2852 I40E_RX_SPLIT_TCP_UDP |
2853 I40E_RX_SPLIT_SCTP;
2854 } else {
2855 rx_ctx.hsplit_0 = 0;
2858 rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2859 (chain_len * ring->rx_buf_len));
2860 if (hw->revision_id == 0)
2861 rx_ctx.lrxqthresh = 0;
2862 else
2863 rx_ctx.lrxqthresh = 2;
2864 rx_ctx.crcstrip = 1;
2865 rx_ctx.l2tsel = 1;
2866 /* this controls whether VLAN is stripped from inner headers */
2867 rx_ctx.showiv = 0;
2868 #ifdef I40E_FCOE
2869 rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2870 #endif
2871 /* set the prefena field to 1 because the manual says to */
2872 rx_ctx.prefena = 1;
2874 /* clear the context in the HMC */
2875 err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2876 if (err) {
2877 dev_info(&vsi->back->pdev->dev,
2878 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2879 ring->queue_index, pf_q, err);
2880 return -ENOMEM;
2883 /* set the context in the HMC */
2884 err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2885 if (err) {
2886 dev_info(&vsi->back->pdev->dev,
2887 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2888 ring->queue_index, pf_q, err);
2889 return -ENOMEM;
2892 /* cache tail for quicker writes, and clear the reg before use */
2893 ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2894 writel(0, ring->tail);
2896 if (ring_is_ps_enabled(ring)) {
2897 i40e_alloc_rx_headers(ring);
2898 i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2899 } else {
2900 i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2903 return 0;
2907 * i40e_vsi_configure_tx - Configure the VSI for Tx
2908 * @vsi: VSI structure describing this set of rings and resources
2910 * Configure the Tx VSI for operation.
2912 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2914 int err = 0;
2915 u16 i;
2917 for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2918 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2920 return err;
2924 * i40e_vsi_configure_rx - Configure the VSI for Rx
2925 * @vsi: the VSI being configured
2927 * Configure the Rx VSI for operation.
2929 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2931 int err = 0;
2932 u16 i;
2934 if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2935 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2936 + ETH_FCS_LEN + VLAN_HLEN;
2937 else
2938 vsi->max_frame = I40E_RXBUFFER_2048;
2940 /* figure out correct receive buffer length */
2941 switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2942 I40E_FLAG_RX_PS_ENABLED)) {
2943 case I40E_FLAG_RX_1BUF_ENABLED:
2944 vsi->rx_hdr_len = 0;
2945 vsi->rx_buf_len = vsi->max_frame;
2946 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2947 break;
2948 case I40E_FLAG_RX_PS_ENABLED:
2949 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2950 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2951 vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2952 break;
2953 default:
2954 vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2955 vsi->rx_buf_len = I40E_RXBUFFER_2048;
2956 vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2957 break;
2960 #ifdef I40E_FCOE
2961 /* setup rx buffer for FCoE */
2962 if ((vsi->type == I40E_VSI_FCOE) &&
2963 (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2964 vsi->rx_hdr_len = 0;
2965 vsi->rx_buf_len = I40E_RXBUFFER_3072;
2966 vsi->max_frame = I40E_RXBUFFER_3072;
2967 vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2970 #endif /* I40E_FCOE */
2971 /* round up for the chip's needs */
2972 vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2973 BIT_ULL(I40E_RXQ_CTX_HBUFF_SHIFT));
2974 vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2975 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
2977 /* set up individual rings */
2978 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2979 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2981 return err;
2985 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2986 * @vsi: ptr to the VSI
2988 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2990 struct i40e_ring *tx_ring, *rx_ring;
2991 u16 qoffset, qcount;
2992 int i, n;
2994 if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
2995 /* Reset the TC information */
2996 for (i = 0; i < vsi->num_queue_pairs; i++) {
2997 rx_ring = vsi->rx_rings[i];
2998 tx_ring = vsi->tx_rings[i];
2999 rx_ring->dcb_tc = 0;
3000 tx_ring->dcb_tc = 0;
3004 for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3005 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3006 continue;
3008 qoffset = vsi->tc_config.tc_info[n].qoffset;
3009 qcount = vsi->tc_config.tc_info[n].qcount;
3010 for (i = qoffset; i < (qoffset + qcount); i++) {
3011 rx_ring = vsi->rx_rings[i];
3012 tx_ring = vsi->tx_rings[i];
3013 rx_ring->dcb_tc = n;
3014 tx_ring->dcb_tc = n;
3020 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3021 * @vsi: ptr to the VSI
3023 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3025 if (vsi->netdev)
3026 i40e_set_rx_mode(vsi->netdev);
3030 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3031 * @vsi: Pointer to the targeted VSI
3033 * This function replays the hlist on the hw where all the SB Flow Director
3034 * filters were saved.
3036 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3038 struct i40e_fdir_filter *filter;
3039 struct i40e_pf *pf = vsi->back;
3040 struct hlist_node *node;
3042 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3043 return;
3045 hlist_for_each_entry_safe(filter, node,
3046 &pf->fdir_filter_list, fdir_node) {
3047 i40e_add_del_fdir(vsi, filter, true);
3052 * i40e_vsi_configure - Set up the VSI for action
3053 * @vsi: the VSI being configured
3055 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3057 int err;
3059 i40e_set_vsi_rx_mode(vsi);
3060 i40e_restore_vlan(vsi);
3061 i40e_vsi_config_dcb_rings(vsi);
3062 err = i40e_vsi_configure_tx(vsi);
3063 if (!err)
3064 err = i40e_vsi_configure_rx(vsi);
3066 return err;
3070 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3071 * @vsi: the VSI being configured
3073 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3075 struct i40e_pf *pf = vsi->back;
3076 struct i40e_hw *hw = &pf->hw;
3077 u16 vector;
3078 int i, q;
3079 u32 qp;
3081 /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3082 * and PFINT_LNKLSTn registers, e.g.:
3083 * PFINT_ITRn[0..n-1] gets msix-1..msix-n (qpair interrupts)
3085 qp = vsi->base_queue;
3086 vector = vsi->base_vector;
3087 for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3088 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3090 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3091 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3092 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3093 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3094 q_vector->rx.itr);
3095 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3096 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3097 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3098 q_vector->tx.itr);
3099 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3100 INTRL_USEC_TO_REG(vsi->int_rate_limit));
3102 /* Linked list for the queuepairs assigned to this vector */
3103 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3104 for (q = 0; q < q_vector->num_ringpairs; q++) {
3105 u32 val;
3107 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3108 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3109 (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3110 (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3111 (I40E_QUEUE_TYPE_TX
3112 << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3114 wr32(hw, I40E_QINT_RQCTL(qp), val);
3116 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3117 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3118 (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3119 ((qp+1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3120 (I40E_QUEUE_TYPE_RX
3121 << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3123 /* Terminate the linked list */
3124 if (q == (q_vector->num_ringpairs - 1))
3125 val |= (I40E_QUEUE_END_OF_LIST
3126 << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3128 wr32(hw, I40E_QINT_TQCTL(qp), val);
3129 qp++;
3133 i40e_flush(hw);
3137 * i40e_enable_misc_int_causes - enable the non-queue interrupts
3138 * @hw: ptr to the hardware info
3140 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3142 struct i40e_hw *hw = &pf->hw;
3143 u32 val;
3145 /* clear things first */
3146 wr32(hw, I40E_PFINT_ICR0_ENA, 0); /* disable all */
3147 rd32(hw, I40E_PFINT_ICR0); /* read to clear */
3149 val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK |
3150 I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK |
3151 I40E_PFINT_ICR0_ENA_GRST_MASK |
3152 I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3153 I40E_PFINT_ICR0_ENA_GPIO_MASK |
3154 I40E_PFINT_ICR0_ENA_HMC_ERR_MASK |
3155 I40E_PFINT_ICR0_ENA_VFLR_MASK |
3156 I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3158 if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3159 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3161 if (pf->flags & I40E_FLAG_PTP)
3162 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3164 wr32(hw, I40E_PFINT_ICR0_ENA, val);
3166 /* SW_ITR_IDX = 0, but don't change INTENA */
3167 wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3168 I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3170 /* OTHER_ITR_IDX = 0 */
3171 wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3175 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3176 * @vsi: the VSI being configured
3178 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3180 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3181 struct i40e_pf *pf = vsi->back;
3182 struct i40e_hw *hw = &pf->hw;
3183 u32 val;
3185 /* set the ITR configuration */
3186 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3187 q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
3188 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3189 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3190 q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
3191 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3192 wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3194 i40e_enable_misc_int_causes(pf);
3196 /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3197 wr32(hw, I40E_PFINT_LNKLST0, 0);
3199 /* Associate the queue pair to the vector and enable the queue int */
3200 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3201 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3202 (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3204 wr32(hw, I40E_QINT_RQCTL(0), val);
3206 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3207 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3208 (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3210 wr32(hw, I40E_QINT_TQCTL(0), val);
3211 i40e_flush(hw);
3215 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3216 * @pf: board private structure
3218 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3220 struct i40e_hw *hw = &pf->hw;
3222 wr32(hw, I40E_PFINT_DYN_CTL0,
3223 I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3224 i40e_flush(hw);
3228 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3229 * @pf: board private structure
3231 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
3233 struct i40e_hw *hw = &pf->hw;
3234 u32 val;
3236 val = I40E_PFINT_DYN_CTL0_INTENA_MASK |
3237 I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
3238 (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3240 wr32(hw, I40E_PFINT_DYN_CTL0, val);
3241 i40e_flush(hw);
3245 * i40e_irq_dynamic_disable - Disable default interrupt generation settings
3246 * @vsi: pointer to a vsi
3247 * @vector: disable a particular Hw Interrupt vector
3249 void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
3251 struct i40e_pf *pf = vsi->back;
3252 struct i40e_hw *hw = &pf->hw;
3253 u32 val;
3255 val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
3256 wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
3257 i40e_flush(hw);
3261 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3262 * @irq: interrupt number
3263 * @data: pointer to a q_vector
3265 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3267 struct i40e_q_vector *q_vector = data;
3269 if (!q_vector->tx.ring && !q_vector->rx.ring)
3270 return IRQ_HANDLED;
3272 napi_schedule_irqoff(&q_vector->napi);
3274 return IRQ_HANDLED;
3278 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3279 * @vsi: the VSI being configured
3280 * @basename: name for the vector
3282 * Allocates MSI-X vectors and requests interrupts from the kernel.
3284 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3286 int q_vectors = vsi->num_q_vectors;
3287 struct i40e_pf *pf = vsi->back;
3288 int base = vsi->base_vector;
3289 int rx_int_idx = 0;
3290 int tx_int_idx = 0;
3291 int vector, err;
3293 for (vector = 0; vector < q_vectors; vector++) {
3294 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3296 if (q_vector->tx.ring && q_vector->rx.ring) {
3297 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3298 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3299 tx_int_idx++;
3300 } else if (q_vector->rx.ring) {
3301 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3302 "%s-%s-%d", basename, "rx", rx_int_idx++);
3303 } else if (q_vector->tx.ring) {
3304 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3305 "%s-%s-%d", basename, "tx", tx_int_idx++);
3306 } else {
3307 /* skip this unused q_vector */
3308 continue;
3310 err = request_irq(pf->msix_entries[base + vector].vector,
3311 vsi->irq_handler,
3313 q_vector->name,
3314 q_vector);
3315 if (err) {
3316 dev_info(&pf->pdev->dev,
3317 "MSIX request_irq failed, error: %d\n", err);
3318 goto free_queue_irqs;
3320 /* assign the mask for this irq */
3321 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3322 &q_vector->affinity_mask);
3325 vsi->irqs_ready = true;
3326 return 0;
3328 free_queue_irqs:
3329 while (vector) {
3330 vector--;
3331 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3332 NULL);
3333 free_irq(pf->msix_entries[base + vector].vector,
3334 &(vsi->q_vectors[vector]));
3336 return err;
3340 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3341 * @vsi: the VSI being un-configured
3343 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3345 struct i40e_pf *pf = vsi->back;
3346 struct i40e_hw *hw = &pf->hw;
3347 int base = vsi->base_vector;
3348 int i;
3350 for (i = 0; i < vsi->num_queue_pairs; i++) {
3351 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3352 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3355 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3356 for (i = vsi->base_vector;
3357 i < (vsi->num_q_vectors + vsi->base_vector); i++)
3358 wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3360 i40e_flush(hw);
3361 for (i = 0; i < vsi->num_q_vectors; i++)
3362 synchronize_irq(pf->msix_entries[i + base].vector);
3363 } else {
3364 /* Legacy and MSI mode - this stops all interrupt handling */
3365 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3366 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3367 i40e_flush(hw);
3368 synchronize_irq(pf->pdev->irq);
3373 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3374 * @vsi: the VSI being configured
3376 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3378 struct i40e_pf *pf = vsi->back;
3379 int i;
3381 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3382 for (i = 0; i < vsi->num_q_vectors; i++)
3383 i40e_irq_dynamic_enable(vsi, i);
3384 } else {
3385 i40e_irq_dynamic_enable_icr0(pf);
3388 i40e_flush(&pf->hw);
3389 return 0;
3393 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3394 * @pf: board private structure
3396 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3398 /* Disable ICR 0 */
3399 wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3400 i40e_flush(&pf->hw);
3404 * i40e_intr - MSI/Legacy and non-queue interrupt handler
3405 * @irq: interrupt number
3406 * @data: pointer to a q_vector
3408 * This is the handler used for all MSI/Legacy interrupts, and deals
3409 * with both queue and non-queue interrupts. This is also used in
3410 * MSIX mode to handle the non-queue interrupts.
3412 static irqreturn_t i40e_intr(int irq, void *data)
3414 struct i40e_pf *pf = (struct i40e_pf *)data;
3415 struct i40e_hw *hw = &pf->hw;
3416 irqreturn_t ret = IRQ_NONE;
3417 u32 icr0, icr0_remaining;
3418 u32 val, ena_mask;
3420 icr0 = rd32(hw, I40E_PFINT_ICR0);
3421 ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3423 /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3424 if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3425 goto enable_intr;
3427 /* if interrupt but no bits showing, must be SWINT */
3428 if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3429 (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3430 pf->sw_int_count++;
3432 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3433 (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3434 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3435 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3436 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3439 /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3440 if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3441 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3442 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3444 /* temporarily disable queue cause for NAPI processing */
3445 u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3447 qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3448 wr32(hw, I40E_QINT_RQCTL(0), qval);
3450 qval = rd32(hw, I40E_QINT_TQCTL(0));
3451 qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3452 wr32(hw, I40E_QINT_TQCTL(0), qval);
3454 if (!test_bit(__I40E_DOWN, &pf->state))
3455 napi_schedule_irqoff(&q_vector->napi);
3458 if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3459 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3460 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3463 if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3464 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3465 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3468 if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3469 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3470 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3473 if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3474 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3475 set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3476 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3477 val = rd32(hw, I40E_GLGEN_RSTAT);
3478 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3479 >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3480 if (val == I40E_RESET_CORER) {
3481 pf->corer_count++;
3482 } else if (val == I40E_RESET_GLOBR) {
3483 pf->globr_count++;
3484 } else if (val == I40E_RESET_EMPR) {
3485 pf->empr_count++;
3486 set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3490 if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3491 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3492 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3493 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3494 rd32(hw, I40E_PFHMC_ERRORINFO),
3495 rd32(hw, I40E_PFHMC_ERRORDATA));
3498 if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3499 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3501 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3502 icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3503 i40e_ptp_tx_hwtstamp(pf);
3507 /* If a critical error is pending we have no choice but to reset the
3508 * device.
3509 * Report and mask out any remaining unexpected interrupts.
3511 icr0_remaining = icr0 & ena_mask;
3512 if (icr0_remaining) {
3513 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3514 icr0_remaining);
3515 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3516 (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3517 (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3518 dev_info(&pf->pdev->dev, "device will be reset\n");
3519 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3520 i40e_service_event_schedule(pf);
3522 ena_mask &= ~icr0_remaining;
3524 ret = IRQ_HANDLED;
3526 enable_intr:
3527 /* re-enable interrupt causes */
3528 wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3529 if (!test_bit(__I40E_DOWN, &pf->state)) {
3530 i40e_service_event_schedule(pf);
3531 i40e_irq_dynamic_enable_icr0(pf);
3534 return ret;
3538 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3539 * @tx_ring: tx ring to clean
3540 * @budget: how many cleans we're allowed
3542 * Returns true if there's any budget left (e.g. the clean is finished)
3544 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3546 struct i40e_vsi *vsi = tx_ring->vsi;
3547 u16 i = tx_ring->next_to_clean;
3548 struct i40e_tx_buffer *tx_buf;
3549 struct i40e_tx_desc *tx_desc;
3551 tx_buf = &tx_ring->tx_bi[i];
3552 tx_desc = I40E_TX_DESC(tx_ring, i);
3553 i -= tx_ring->count;
3555 do {
3556 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3558 /* if next_to_watch is not set then there is no work pending */
3559 if (!eop_desc)
3560 break;
3562 /* prevent any other reads prior to eop_desc */
3563 read_barrier_depends();
3565 /* if the descriptor isn't done, no work yet to do */
3566 if (!(eop_desc->cmd_type_offset_bsz &
3567 cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3568 break;
3570 /* clear next_to_watch to prevent false hangs */
3571 tx_buf->next_to_watch = NULL;
3573 tx_desc->buffer_addr = 0;
3574 tx_desc->cmd_type_offset_bsz = 0;
3575 /* move past filter desc */
3576 tx_buf++;
3577 tx_desc++;
3578 i++;
3579 if (unlikely(!i)) {
3580 i -= tx_ring->count;
3581 tx_buf = tx_ring->tx_bi;
3582 tx_desc = I40E_TX_DESC(tx_ring, 0);
3584 /* unmap skb header data */
3585 dma_unmap_single(tx_ring->dev,
3586 dma_unmap_addr(tx_buf, dma),
3587 dma_unmap_len(tx_buf, len),
3588 DMA_TO_DEVICE);
3589 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3590 kfree(tx_buf->raw_buf);
3592 tx_buf->raw_buf = NULL;
3593 tx_buf->tx_flags = 0;
3594 tx_buf->next_to_watch = NULL;
3595 dma_unmap_len_set(tx_buf, len, 0);
3596 tx_desc->buffer_addr = 0;
3597 tx_desc->cmd_type_offset_bsz = 0;
3599 /* move us past the eop_desc for start of next FD desc */
3600 tx_buf++;
3601 tx_desc++;
3602 i++;
3603 if (unlikely(!i)) {
3604 i -= tx_ring->count;
3605 tx_buf = tx_ring->tx_bi;
3606 tx_desc = I40E_TX_DESC(tx_ring, 0);
3609 /* update budget accounting */
3610 budget--;
3611 } while (likely(budget));
3613 i += tx_ring->count;
3614 tx_ring->next_to_clean = i;
3616 if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3617 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3619 return budget > 0;
3623 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3624 * @irq: interrupt number
3625 * @data: pointer to a q_vector
3627 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3629 struct i40e_q_vector *q_vector = data;
3630 struct i40e_vsi *vsi;
3632 if (!q_vector->tx.ring)
3633 return IRQ_HANDLED;
3635 vsi = q_vector->tx.ring->vsi;
3636 i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3638 return IRQ_HANDLED;
3642 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3643 * @vsi: the VSI being configured
3644 * @v_idx: vector index
3645 * @qp_idx: queue pair index
3647 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3649 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3650 struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3651 struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3653 tx_ring->q_vector = q_vector;
3654 tx_ring->next = q_vector->tx.ring;
3655 q_vector->tx.ring = tx_ring;
3656 q_vector->tx.count++;
3658 rx_ring->q_vector = q_vector;
3659 rx_ring->next = q_vector->rx.ring;
3660 q_vector->rx.ring = rx_ring;
3661 q_vector->rx.count++;
3665 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3666 * @vsi: the VSI being configured
3668 * This function maps descriptor rings to the queue-specific vectors
3669 * we were allotted through the MSI-X enabling code. Ideally, we'd have
3670 * one vector per queue pair, but on a constrained vector budget, we
3671 * group the queue pairs as "efficiently" as possible.
3673 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3675 int qp_remaining = vsi->num_queue_pairs;
3676 int q_vectors = vsi->num_q_vectors;
3677 int num_ringpairs;
3678 int v_start = 0;
3679 int qp_idx = 0;
3681 /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3682 * group them so there are multiple queues per vector.
3683 * It is also important to go through all the vectors available to be
3684 * sure that if we don't use all the vectors, that the remaining vectors
3685 * are cleared. This is especially important when decreasing the
3686 * number of queues in use.
3688 for (; v_start < q_vectors; v_start++) {
3689 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3691 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3693 q_vector->num_ringpairs = num_ringpairs;
3695 q_vector->rx.count = 0;
3696 q_vector->tx.count = 0;
3697 q_vector->rx.ring = NULL;
3698 q_vector->tx.ring = NULL;
3700 while (num_ringpairs--) {
3701 i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3702 qp_idx++;
3703 qp_remaining--;
3709 * i40e_vsi_request_irq - Request IRQ from the OS
3710 * @vsi: the VSI being configured
3711 * @basename: name for the vector
3713 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3715 struct i40e_pf *pf = vsi->back;
3716 int err;
3718 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3719 err = i40e_vsi_request_irq_msix(vsi, basename);
3720 else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3721 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3722 pf->int_name, pf);
3723 else
3724 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3725 pf->int_name, pf);
3727 if (err)
3728 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3730 return err;
3733 #ifdef CONFIG_NET_POLL_CONTROLLER
3735 * i40e_netpoll - A Polling 'interrupt'handler
3736 * @netdev: network interface device structure
3738 * This is used by netconsole to send skbs without having to re-enable
3739 * interrupts. It's not called while the normal interrupt routine is executing.
3741 #ifdef I40E_FCOE
3742 void i40e_netpoll(struct net_device *netdev)
3743 #else
3744 static void i40e_netpoll(struct net_device *netdev)
3745 #endif
3747 struct i40e_netdev_priv *np = netdev_priv(netdev);
3748 struct i40e_vsi *vsi = np->vsi;
3749 struct i40e_pf *pf = vsi->back;
3750 int i;
3752 /* if interface is down do nothing */
3753 if (test_bit(__I40E_DOWN, &vsi->state))
3754 return;
3756 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3757 for (i = 0; i < vsi->num_q_vectors; i++)
3758 i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3759 } else {
3760 i40e_intr(pf->pdev->irq, netdev);
3763 #endif
3766 * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3767 * @pf: the PF being configured
3768 * @pf_q: the PF queue
3769 * @enable: enable or disable state of the queue
3771 * This routine will wait for the given Tx queue of the PF to reach the
3772 * enabled or disabled state.
3773 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3774 * multiple retries; else will return 0 in case of success.
3776 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3778 int i;
3779 u32 tx_reg;
3781 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3782 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3783 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3784 break;
3786 usleep_range(10, 20);
3788 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3789 return -ETIMEDOUT;
3791 return 0;
3795 * i40e_vsi_control_tx - Start or stop a VSI's rings
3796 * @vsi: the VSI being configured
3797 * @enable: start or stop the rings
3799 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3801 struct i40e_pf *pf = vsi->back;
3802 struct i40e_hw *hw = &pf->hw;
3803 int i, j, pf_q, ret = 0;
3804 u32 tx_reg;
3806 pf_q = vsi->base_queue;
3807 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3809 /* warn the TX unit of coming changes */
3810 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3811 if (!enable)
3812 usleep_range(10, 20);
3814 for (j = 0; j < 50; j++) {
3815 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3816 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3817 ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3818 break;
3819 usleep_range(1000, 2000);
3821 /* Skip if the queue is already in the requested state */
3822 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3823 continue;
3825 /* turn on/off the queue */
3826 if (enable) {
3827 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3828 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3829 } else {
3830 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3833 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3834 /* No waiting for the Tx queue to disable */
3835 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3836 continue;
3838 /* wait for the change to finish */
3839 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3840 if (ret) {
3841 dev_info(&pf->pdev->dev,
3842 "VSI seid %d Tx ring %d %sable timeout\n",
3843 vsi->seid, pf_q, (enable ? "en" : "dis"));
3844 break;
3848 if (hw->revision_id == 0)
3849 mdelay(50);
3850 return ret;
3854 * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3855 * @pf: the PF being configured
3856 * @pf_q: the PF queue
3857 * @enable: enable or disable state of the queue
3859 * This routine will wait for the given Rx queue of the PF to reach the
3860 * enabled or disabled state.
3861 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3862 * multiple retries; else will return 0 in case of success.
3864 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3866 int i;
3867 u32 rx_reg;
3869 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3870 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3871 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3872 break;
3874 usleep_range(10, 20);
3876 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3877 return -ETIMEDOUT;
3879 return 0;
3883 * i40e_vsi_control_rx - Start or stop a VSI's rings
3884 * @vsi: the VSI being configured
3885 * @enable: start or stop the rings
3887 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3889 struct i40e_pf *pf = vsi->back;
3890 struct i40e_hw *hw = &pf->hw;
3891 int i, j, pf_q, ret = 0;
3892 u32 rx_reg;
3894 pf_q = vsi->base_queue;
3895 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3896 for (j = 0; j < 50; j++) {
3897 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3898 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3899 ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3900 break;
3901 usleep_range(1000, 2000);
3904 /* Skip if the queue is already in the requested state */
3905 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3906 continue;
3908 /* turn on/off the queue */
3909 if (enable)
3910 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3911 else
3912 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3913 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3915 /* wait for the change to finish */
3916 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3917 if (ret) {
3918 dev_info(&pf->pdev->dev,
3919 "VSI seid %d Rx ring %d %sable timeout\n",
3920 vsi->seid, pf_q, (enable ? "en" : "dis"));
3921 break;
3925 return ret;
3929 * i40e_vsi_control_rings - Start or stop a VSI's rings
3930 * @vsi: the VSI being configured
3931 * @enable: start or stop the rings
3933 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3935 int ret = 0;
3937 /* do rx first for enable and last for disable */
3938 if (request) {
3939 ret = i40e_vsi_control_rx(vsi, request);
3940 if (ret)
3941 return ret;
3942 ret = i40e_vsi_control_tx(vsi, request);
3943 } else {
3944 /* Ignore return value, we need to shutdown whatever we can */
3945 i40e_vsi_control_tx(vsi, request);
3946 i40e_vsi_control_rx(vsi, request);
3949 return ret;
3953 * i40e_vsi_free_irq - Free the irq association with the OS
3954 * @vsi: the VSI being configured
3956 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3958 struct i40e_pf *pf = vsi->back;
3959 struct i40e_hw *hw = &pf->hw;
3960 int base = vsi->base_vector;
3961 u32 val, qp;
3962 int i;
3964 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3965 if (!vsi->q_vectors)
3966 return;
3968 if (!vsi->irqs_ready)
3969 return;
3971 vsi->irqs_ready = false;
3972 for (i = 0; i < vsi->num_q_vectors; i++) {
3973 u16 vector = i + base;
3975 /* free only the irqs that were actually requested */
3976 if (!vsi->q_vectors[i] ||
3977 !vsi->q_vectors[i]->num_ringpairs)
3978 continue;
3980 /* clear the affinity_mask in the IRQ descriptor */
3981 irq_set_affinity_hint(pf->msix_entries[vector].vector,
3982 NULL);
3983 free_irq(pf->msix_entries[vector].vector,
3984 vsi->q_vectors[i]);
3986 /* Tear down the interrupt queue link list
3988 * We know that they come in pairs and always
3989 * the Rx first, then the Tx. To clear the
3990 * link list, stick the EOL value into the
3991 * next_q field of the registers.
3993 val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3994 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3995 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3996 val |= I40E_QUEUE_END_OF_LIST
3997 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3998 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4000 while (qp != I40E_QUEUE_END_OF_LIST) {
4001 u32 next;
4003 val = rd32(hw, I40E_QINT_RQCTL(qp));
4005 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4006 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4007 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4008 I40E_QINT_RQCTL_INTEVENT_MASK);
4010 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4011 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4013 wr32(hw, I40E_QINT_RQCTL(qp), val);
4015 val = rd32(hw, I40E_QINT_TQCTL(qp));
4017 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4018 >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4020 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4021 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4022 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4023 I40E_QINT_TQCTL_INTEVENT_MASK);
4025 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4026 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4028 wr32(hw, I40E_QINT_TQCTL(qp), val);
4029 qp = next;
4032 } else {
4033 free_irq(pf->pdev->irq, pf);
4035 val = rd32(hw, I40E_PFINT_LNKLST0);
4036 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4037 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4038 val |= I40E_QUEUE_END_OF_LIST
4039 << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4040 wr32(hw, I40E_PFINT_LNKLST0, val);
4042 val = rd32(hw, I40E_QINT_RQCTL(qp));
4043 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4044 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4045 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4046 I40E_QINT_RQCTL_INTEVENT_MASK);
4048 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4049 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4051 wr32(hw, I40E_QINT_RQCTL(qp), val);
4053 val = rd32(hw, I40E_QINT_TQCTL(qp));
4055 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4056 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4057 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4058 I40E_QINT_TQCTL_INTEVENT_MASK);
4060 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4061 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4063 wr32(hw, I40E_QINT_TQCTL(qp), val);
4068 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4069 * @vsi: the VSI being configured
4070 * @v_idx: Index of vector to be freed
4072 * This function frees the memory allocated to the q_vector. In addition if
4073 * NAPI is enabled it will delete any references to the NAPI struct prior
4074 * to freeing the q_vector.
4076 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4078 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4079 struct i40e_ring *ring;
4081 if (!q_vector)
4082 return;
4084 /* disassociate q_vector from rings */
4085 i40e_for_each_ring(ring, q_vector->tx)
4086 ring->q_vector = NULL;
4088 i40e_for_each_ring(ring, q_vector->rx)
4089 ring->q_vector = NULL;
4091 /* only VSI w/ an associated netdev is set up w/ NAPI */
4092 if (vsi->netdev)
4093 netif_napi_del(&q_vector->napi);
4095 vsi->q_vectors[v_idx] = NULL;
4097 kfree_rcu(q_vector, rcu);
4101 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4102 * @vsi: the VSI being un-configured
4104 * This frees the memory allocated to the q_vectors and
4105 * deletes references to the NAPI struct.
4107 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4109 int v_idx;
4111 for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4112 i40e_free_q_vector(vsi, v_idx);
4116 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4117 * @pf: board private structure
4119 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4121 /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4122 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4123 pci_disable_msix(pf->pdev);
4124 kfree(pf->msix_entries);
4125 pf->msix_entries = NULL;
4126 kfree(pf->irq_pile);
4127 pf->irq_pile = NULL;
4128 } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4129 pci_disable_msi(pf->pdev);
4131 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4135 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4136 * @pf: board private structure
4138 * We go through and clear interrupt specific resources and reset the structure
4139 * to pre-load conditions
4141 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4143 int i;
4145 i40e_stop_misc_vector(pf);
4146 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4147 synchronize_irq(pf->msix_entries[0].vector);
4148 free_irq(pf->msix_entries[0].vector, pf);
4151 i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4152 for (i = 0; i < pf->num_alloc_vsi; i++)
4153 if (pf->vsi[i])
4154 i40e_vsi_free_q_vectors(pf->vsi[i]);
4155 i40e_reset_interrupt_capability(pf);
4159 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4160 * @vsi: the VSI being configured
4162 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4164 int q_idx;
4166 if (!vsi->netdev)
4167 return;
4169 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4170 napi_enable(&vsi->q_vectors[q_idx]->napi);
4174 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4175 * @vsi: the VSI being configured
4177 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4179 int q_idx;
4181 if (!vsi->netdev)
4182 return;
4184 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4185 napi_disable(&vsi->q_vectors[q_idx]->napi);
4189 * i40e_vsi_close - Shut down a VSI
4190 * @vsi: the vsi to be quelled
4192 static void i40e_vsi_close(struct i40e_vsi *vsi)
4194 if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4195 i40e_down(vsi);
4196 i40e_vsi_free_irq(vsi);
4197 i40e_vsi_free_tx_resources(vsi);
4198 i40e_vsi_free_rx_resources(vsi);
4199 vsi->current_netdev_flags = 0;
4203 * i40e_quiesce_vsi - Pause a given VSI
4204 * @vsi: the VSI being paused
4206 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4208 if (test_bit(__I40E_DOWN, &vsi->state))
4209 return;
4211 /* No need to disable FCoE VSI when Tx suspended */
4212 if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4213 vsi->type == I40E_VSI_FCOE) {
4214 dev_dbg(&vsi->back->pdev->dev,
4215 "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4216 return;
4219 set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4220 if (vsi->netdev && netif_running(vsi->netdev))
4221 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4222 else
4223 i40e_vsi_close(vsi);
4227 * i40e_unquiesce_vsi - Resume a given VSI
4228 * @vsi: the VSI being resumed
4230 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4232 if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4233 return;
4235 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4236 if (vsi->netdev && netif_running(vsi->netdev))
4237 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4238 else
4239 i40e_vsi_open(vsi); /* this clears the DOWN bit */
4243 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4244 * @pf: the PF
4246 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4248 int v;
4250 for (v = 0; v < pf->num_alloc_vsi; v++) {
4251 if (pf->vsi[v])
4252 i40e_quiesce_vsi(pf->vsi[v]);
4257 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4258 * @pf: the PF
4260 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4262 int v;
4264 for (v = 0; v < pf->num_alloc_vsi; v++) {
4265 if (pf->vsi[v])
4266 i40e_unquiesce_vsi(pf->vsi[v]);
4270 #ifdef CONFIG_I40E_DCB
4272 * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4273 * @vsi: the VSI being configured
4275 * This function waits for the given VSI's Tx queues to be disabled.
4277 static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4279 struct i40e_pf *pf = vsi->back;
4280 int i, pf_q, ret;
4282 pf_q = vsi->base_queue;
4283 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4284 /* Check and wait for the disable status of the queue */
4285 ret = i40e_pf_txq_wait(pf, pf_q, false);
4286 if (ret) {
4287 dev_info(&pf->pdev->dev,
4288 "VSI seid %d Tx ring %d disable timeout\n",
4289 vsi->seid, pf_q);
4290 return ret;
4294 return 0;
4298 * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4299 * @pf: the PF
4301 * This function waits for the Tx queues to be in disabled state for all the
4302 * VSIs that are managed by this PF.
4304 static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4306 int v, ret = 0;
4308 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4309 /* No need to wait for FCoE VSI queues */
4310 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4311 ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4312 if (ret)
4313 break;
4317 return ret;
4320 #endif
4323 * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4324 * @q_idx: TX queue number
4325 * @vsi: Pointer to VSI struct
4327 * This function checks specified queue for given VSI. Detects hung condition.
4328 * Sets hung bit since it is two step process. Before next run of service task
4329 * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4330 * hung condition remain unchanged and during subsequent run, this function
4331 * issues SW interrupt to recover from hung condition.
4333 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4335 struct i40e_ring *tx_ring = NULL;
4336 struct i40e_pf *pf;
4337 u32 head, val, tx_pending;
4338 int i;
4340 pf = vsi->back;
4342 /* now that we have an index, find the tx_ring struct */
4343 for (i = 0; i < vsi->num_queue_pairs; i++) {
4344 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4345 if (q_idx == vsi->tx_rings[i]->queue_index) {
4346 tx_ring = vsi->tx_rings[i];
4347 break;
4352 if (!tx_ring)
4353 return;
4355 /* Read interrupt register */
4356 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4357 val = rd32(&pf->hw,
4358 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4359 tx_ring->vsi->base_vector - 1));
4360 else
4361 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4363 head = i40e_get_head(tx_ring);
4365 tx_pending = i40e_get_tx_pending(tx_ring);
4367 /* Interrupts are disabled and TX pending is non-zero,
4368 * trigger the SW interrupt (don't wait). Worst case
4369 * there will be one extra interrupt which may result
4370 * into not cleaning any queues because queues are cleaned.
4372 if (tx_pending && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)))
4373 i40e_force_wb(vsi, tx_ring->q_vector);
4377 * i40e_detect_recover_hung - Function to detect and recover hung_queues
4378 * @pf: pointer to PF struct
4380 * LAN VSI has netdev and netdev has TX queues. This function is to check
4381 * each of those TX queues if they are hung, trigger recovery by issuing
4382 * SW interrupt.
4384 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4386 struct net_device *netdev;
4387 struct i40e_vsi *vsi;
4388 int i;
4390 /* Only for LAN VSI */
4391 vsi = pf->vsi[pf->lan_vsi];
4393 if (!vsi)
4394 return;
4396 /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4397 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4398 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4399 return;
4401 /* Make sure type is MAIN VSI */
4402 if (vsi->type != I40E_VSI_MAIN)
4403 return;
4405 netdev = vsi->netdev;
4406 if (!netdev)
4407 return;
4409 /* Bail out if netif_carrier is not OK */
4410 if (!netif_carrier_ok(netdev))
4411 return;
4413 /* Go thru' TX queues for netdev */
4414 for (i = 0; i < netdev->num_tx_queues; i++) {
4415 struct netdev_queue *q;
4417 q = netdev_get_tx_queue(netdev, i);
4418 if (q)
4419 i40e_detect_recover_hung_queue(i, vsi);
4424 * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4425 * @pf: pointer to PF
4427 * Get TC map for ISCSI PF type that will include iSCSI TC
4428 * and LAN TC.
4430 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4432 struct i40e_dcb_app_priority_table app;
4433 struct i40e_hw *hw = &pf->hw;
4434 u8 enabled_tc = 1; /* TC0 is always enabled */
4435 u8 tc, i;
4436 /* Get the iSCSI APP TLV */
4437 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4439 for (i = 0; i < dcbcfg->numapps; i++) {
4440 app = dcbcfg->app[i];
4441 if (app.selector == I40E_APP_SEL_TCPIP &&
4442 app.protocolid == I40E_APP_PROTOID_ISCSI) {
4443 tc = dcbcfg->etscfg.prioritytable[app.priority];
4444 enabled_tc |= BIT_ULL(tc);
4445 break;
4449 return enabled_tc;
4453 * i40e_dcb_get_num_tc - Get the number of TCs from DCBx config
4454 * @dcbcfg: the corresponding DCBx configuration structure
4456 * Return the number of TCs from given DCBx configuration
4458 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4460 u8 num_tc = 0;
4461 int i;
4463 /* Scan the ETS Config Priority Table to find
4464 * traffic class enabled for a given priority
4465 * and use the traffic class index to get the
4466 * number of traffic classes enabled
4468 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4469 if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4470 num_tc = dcbcfg->etscfg.prioritytable[i];
4473 /* Traffic class index starts from zero so
4474 * increment to return the actual count
4476 return num_tc + 1;
4480 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4481 * @dcbcfg: the corresponding DCBx configuration structure
4483 * Query the current DCB configuration and return the number of
4484 * traffic classes enabled from the given DCBX config
4486 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4488 u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4489 u8 enabled_tc = 1;
4490 u8 i;
4492 for (i = 0; i < num_tc; i++)
4493 enabled_tc |= BIT(i);
4495 return enabled_tc;
4499 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4500 * @pf: PF being queried
4502 * Return number of traffic classes enabled for the given PF
4504 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4506 struct i40e_hw *hw = &pf->hw;
4507 u8 i, enabled_tc;
4508 u8 num_tc = 0;
4509 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4511 /* If DCB is not enabled then always in single TC */
4512 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4513 return 1;
4515 /* SFP mode will be enabled for all TCs on port */
4516 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4517 return i40e_dcb_get_num_tc(dcbcfg);
4519 /* MFP mode return count of enabled TCs for this PF */
4520 if (pf->hw.func_caps.iscsi)
4521 enabled_tc = i40e_get_iscsi_tc_map(pf);
4522 else
4523 return 1; /* Only TC0 */
4525 /* At least have TC0 */
4526 enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4527 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4528 if (enabled_tc & BIT_ULL(i))
4529 num_tc++;
4531 return num_tc;
4535 * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4536 * @pf: PF being queried
4538 * Return a bitmap for first enabled traffic class for this PF.
4540 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4542 u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4543 u8 i = 0;
4545 if (!enabled_tc)
4546 return 0x1; /* TC0 */
4548 /* Find the first enabled TC */
4549 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4550 if (enabled_tc & BIT_ULL(i))
4551 break;
4554 return BIT(i);
4558 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4559 * @pf: PF being queried
4561 * Return a bitmap for enabled traffic classes for this PF.
4563 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4565 /* If DCB is not enabled for this PF then just return default TC */
4566 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4567 return i40e_pf_get_default_tc(pf);
4569 /* SFP mode we want PF to be enabled for all TCs */
4570 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4571 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4573 /* MFP enabled and iSCSI PF type */
4574 if (pf->hw.func_caps.iscsi)
4575 return i40e_get_iscsi_tc_map(pf);
4576 else
4577 return i40e_pf_get_default_tc(pf);
4581 * i40e_vsi_get_bw_info - Query VSI BW Information
4582 * @vsi: the VSI being queried
4584 * Returns 0 on success, negative value on failure
4586 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4588 struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4589 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4590 struct i40e_pf *pf = vsi->back;
4591 struct i40e_hw *hw = &pf->hw;
4592 i40e_status ret;
4593 u32 tc_bw_max;
4594 int i;
4596 /* Get the VSI level BW configuration */
4597 ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4598 if (ret) {
4599 dev_info(&pf->pdev->dev,
4600 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4601 i40e_stat_str(&pf->hw, ret),
4602 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4603 return -EINVAL;
4606 /* Get the VSI level BW configuration per TC */
4607 ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4608 NULL);
4609 if (ret) {
4610 dev_info(&pf->pdev->dev,
4611 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4612 i40e_stat_str(&pf->hw, ret),
4613 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4614 return -EINVAL;
4617 if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4618 dev_info(&pf->pdev->dev,
4619 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4620 bw_config.tc_valid_bits,
4621 bw_ets_config.tc_valid_bits);
4622 /* Still continuing */
4625 vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4626 vsi->bw_max_quanta = bw_config.max_bw;
4627 tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4628 (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4629 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4630 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4631 vsi->bw_ets_limit_credits[i] =
4632 le16_to_cpu(bw_ets_config.credits[i]);
4633 /* 3 bits out of 4 for each TC */
4634 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4637 return 0;
4641 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4642 * @vsi: the VSI being configured
4643 * @enabled_tc: TC bitmap
4644 * @bw_credits: BW shared credits per TC
4646 * Returns 0 on success, negative value on failure
4648 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4649 u8 *bw_share)
4651 struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4652 i40e_status ret;
4653 int i;
4655 bw_data.tc_valid_bits = enabled_tc;
4656 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4657 bw_data.tc_bw_credits[i] = bw_share[i];
4659 ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4660 NULL);
4661 if (ret) {
4662 dev_info(&vsi->back->pdev->dev,
4663 "AQ command Config VSI BW allocation per TC failed = %d\n",
4664 vsi->back->hw.aq.asq_last_status);
4665 return -EINVAL;
4668 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4669 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4671 return 0;
4675 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4676 * @vsi: the VSI being configured
4677 * @enabled_tc: TC map to be enabled
4680 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4682 struct net_device *netdev = vsi->netdev;
4683 struct i40e_pf *pf = vsi->back;
4684 struct i40e_hw *hw = &pf->hw;
4685 u8 netdev_tc = 0;
4686 int i;
4687 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4689 if (!netdev)
4690 return;
4692 if (!enabled_tc) {
4693 netdev_reset_tc(netdev);
4694 return;
4697 /* Set up actual enabled TCs on the VSI */
4698 if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4699 return;
4701 /* set per TC queues for the VSI */
4702 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4703 /* Only set TC queues for enabled tcs
4705 * e.g. For a VSI that has TC0 and TC3 enabled the
4706 * enabled_tc bitmap would be 0x00001001; the driver
4707 * will set the numtc for netdev as 2 that will be
4708 * referenced by the netdev layer as TC 0 and 1.
4710 if (vsi->tc_config.enabled_tc & BIT_ULL(i))
4711 netdev_set_tc_queue(netdev,
4712 vsi->tc_config.tc_info[i].netdev_tc,
4713 vsi->tc_config.tc_info[i].qcount,
4714 vsi->tc_config.tc_info[i].qoffset);
4717 /* Assign UP2TC map for the VSI */
4718 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4719 /* Get the actual TC# for the UP */
4720 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4721 /* Get the mapped netdev TC# for the UP */
4722 netdev_tc = vsi->tc_config.tc_info[ets_tc].netdev_tc;
4723 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4728 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4729 * @vsi: the VSI being configured
4730 * @ctxt: the ctxt buffer returned from AQ VSI update param command
4732 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4733 struct i40e_vsi_context *ctxt)
4735 /* copy just the sections touched not the entire info
4736 * since not all sections are valid as returned by
4737 * update vsi params
4739 vsi->info.mapping_flags = ctxt->info.mapping_flags;
4740 memcpy(&vsi->info.queue_mapping,
4741 &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4742 memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4743 sizeof(vsi->info.tc_mapping));
4747 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4748 * @vsi: VSI to be configured
4749 * @enabled_tc: TC bitmap
4751 * This configures a particular VSI for TCs that are mapped to the
4752 * given TC bitmap. It uses default bandwidth share for TCs across
4753 * VSIs to configure TC for a particular VSI.
4755 * NOTE:
4756 * It is expected that the VSI queues have been quisced before calling
4757 * this function.
4759 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4761 u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4762 struct i40e_vsi_context ctxt;
4763 int ret = 0;
4764 int i;
4766 /* Check if enabled_tc is same as existing or new TCs */
4767 if (vsi->tc_config.enabled_tc == enabled_tc)
4768 return ret;
4770 /* Enable ETS TCs with equal BW Share for now across all VSIs */
4771 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4772 if (enabled_tc & BIT_ULL(i))
4773 bw_share[i] = 1;
4776 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4777 if (ret) {
4778 dev_info(&vsi->back->pdev->dev,
4779 "Failed configuring TC map %d for VSI %d\n",
4780 enabled_tc, vsi->seid);
4781 goto out;
4784 /* Update Queue Pairs Mapping for currently enabled UPs */
4785 ctxt.seid = vsi->seid;
4786 ctxt.pf_num = vsi->back->hw.pf_id;
4787 ctxt.vf_num = 0;
4788 ctxt.uplink_seid = vsi->uplink_seid;
4789 ctxt.info = vsi->info;
4790 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4792 /* Update the VSI after updating the VSI queue-mapping information */
4793 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4794 if (ret) {
4795 dev_info(&vsi->back->pdev->dev,
4796 "Update vsi tc config failed, err %s aq_err %s\n",
4797 i40e_stat_str(&vsi->back->hw, ret),
4798 i40e_aq_str(&vsi->back->hw,
4799 vsi->back->hw.aq.asq_last_status));
4800 goto out;
4802 /* update the local VSI info with updated queue map */
4803 i40e_vsi_update_queue_map(vsi, &ctxt);
4804 vsi->info.valid_sections = 0;
4806 /* Update current VSI BW information */
4807 ret = i40e_vsi_get_bw_info(vsi);
4808 if (ret) {
4809 dev_info(&vsi->back->pdev->dev,
4810 "Failed updating vsi bw info, err %s aq_err %s\n",
4811 i40e_stat_str(&vsi->back->hw, ret),
4812 i40e_aq_str(&vsi->back->hw,
4813 vsi->back->hw.aq.asq_last_status));
4814 goto out;
4817 /* Update the netdev TC setup */
4818 i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4819 out:
4820 return ret;
4824 * i40e_veb_config_tc - Configure TCs for given VEB
4825 * @veb: given VEB
4826 * @enabled_tc: TC bitmap
4828 * Configures given TC bitmap for VEB (switching) element
4830 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4832 struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4833 struct i40e_pf *pf = veb->pf;
4834 int ret = 0;
4835 int i;
4837 /* No TCs or already enabled TCs just return */
4838 if (!enabled_tc || veb->enabled_tc == enabled_tc)
4839 return ret;
4841 bw_data.tc_valid_bits = enabled_tc;
4842 /* bw_data.absolute_credits is not set (relative) */
4844 /* Enable ETS TCs with equal BW Share for now */
4845 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4846 if (enabled_tc & BIT_ULL(i))
4847 bw_data.tc_bw_share_credits[i] = 1;
4850 ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4851 &bw_data, NULL);
4852 if (ret) {
4853 dev_info(&pf->pdev->dev,
4854 "VEB bw config failed, err %s aq_err %s\n",
4855 i40e_stat_str(&pf->hw, ret),
4856 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4857 goto out;
4860 /* Update the BW information */
4861 ret = i40e_veb_get_bw_info(veb);
4862 if (ret) {
4863 dev_info(&pf->pdev->dev,
4864 "Failed getting veb bw config, err %s aq_err %s\n",
4865 i40e_stat_str(&pf->hw, ret),
4866 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4869 out:
4870 return ret;
4873 #ifdef CONFIG_I40E_DCB
4875 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4876 * @pf: PF struct
4878 * Reconfigure VEB/VSIs on a given PF; it is assumed that
4879 * the caller would've quiesce all the VSIs before calling
4880 * this function
4882 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4884 u8 tc_map = 0;
4885 int ret;
4886 u8 v;
4888 /* Enable the TCs available on PF to all VEBs */
4889 tc_map = i40e_pf_get_tc_map(pf);
4890 for (v = 0; v < I40E_MAX_VEB; v++) {
4891 if (!pf->veb[v])
4892 continue;
4893 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4894 if (ret) {
4895 dev_info(&pf->pdev->dev,
4896 "Failed configuring TC for VEB seid=%d\n",
4897 pf->veb[v]->seid);
4898 /* Will try to configure as many components */
4902 /* Update each VSI */
4903 for (v = 0; v < pf->num_alloc_vsi; v++) {
4904 if (!pf->vsi[v])
4905 continue;
4907 /* - Enable all TCs for the LAN VSI
4908 #ifdef I40E_FCOE
4909 * - For FCoE VSI only enable the TC configured
4910 * as per the APP TLV
4911 #endif
4912 * - For all others keep them at TC0 for now
4914 if (v == pf->lan_vsi)
4915 tc_map = i40e_pf_get_tc_map(pf);
4916 else
4917 tc_map = i40e_pf_get_default_tc(pf);
4918 #ifdef I40E_FCOE
4919 if (pf->vsi[v]->type == I40E_VSI_FCOE)
4920 tc_map = i40e_get_fcoe_tc_map(pf);
4921 #endif /* #ifdef I40E_FCOE */
4923 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4924 if (ret) {
4925 dev_info(&pf->pdev->dev,
4926 "Failed configuring TC for VSI seid=%d\n",
4927 pf->vsi[v]->seid);
4928 /* Will try to configure as many components */
4929 } else {
4930 /* Re-configure VSI vectors based on updated TC map */
4931 i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4932 if (pf->vsi[v]->netdev)
4933 i40e_dcbnl_set_all(pf->vsi[v]);
4939 * i40e_resume_port_tx - Resume port Tx
4940 * @pf: PF struct
4942 * Resume a port's Tx and issue a PF reset in case of failure to
4943 * resume.
4945 static int i40e_resume_port_tx(struct i40e_pf *pf)
4947 struct i40e_hw *hw = &pf->hw;
4948 int ret;
4950 ret = i40e_aq_resume_port_tx(hw, NULL);
4951 if (ret) {
4952 dev_info(&pf->pdev->dev,
4953 "Resume Port Tx failed, err %s aq_err %s\n",
4954 i40e_stat_str(&pf->hw, ret),
4955 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4956 /* Schedule PF reset to recover */
4957 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4958 i40e_service_event_schedule(pf);
4961 return ret;
4965 * i40e_init_pf_dcb - Initialize DCB configuration
4966 * @pf: PF being configured
4968 * Query the current DCB configuration and cache it
4969 * in the hardware structure
4971 static int i40e_init_pf_dcb(struct i40e_pf *pf)
4973 struct i40e_hw *hw = &pf->hw;
4974 int err = 0;
4976 /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
4977 if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
4978 (pf->hw.aq.fw_maj_ver < 4))
4979 goto out;
4981 /* Get the initial DCB configuration */
4982 err = i40e_init_dcb(hw);
4983 if (!err) {
4984 /* Device/Function is not DCBX capable */
4985 if ((!hw->func_caps.dcb) ||
4986 (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
4987 dev_info(&pf->pdev->dev,
4988 "DCBX offload is not supported or is disabled for this PF.\n");
4990 if (pf->flags & I40E_FLAG_MFP_ENABLED)
4991 goto out;
4993 } else {
4994 /* When status is not DISABLED then DCBX in FW */
4995 pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
4996 DCB_CAP_DCBX_VER_IEEE;
4998 pf->flags |= I40E_FLAG_DCB_CAPABLE;
4999 /* Enable DCB tagging only when more than one TC */
5000 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5001 pf->flags |= I40E_FLAG_DCB_ENABLED;
5002 dev_dbg(&pf->pdev->dev,
5003 "DCBX offload is supported for this PF.\n");
5005 } else {
5006 dev_info(&pf->pdev->dev,
5007 "Query for DCB configuration failed, err %s aq_err %s\n",
5008 i40e_stat_str(&pf->hw, err),
5009 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5012 out:
5013 return err;
5015 #endif /* CONFIG_I40E_DCB */
5016 #define SPEED_SIZE 14
5017 #define FC_SIZE 8
5019 * i40e_print_link_message - print link up or down
5020 * @vsi: the VSI for which link needs a message
5022 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5024 char *speed = "Unknown";
5025 char *fc = "Unknown";
5027 if (vsi->current_isup == isup)
5028 return;
5029 vsi->current_isup = isup;
5030 if (!isup) {
5031 netdev_info(vsi->netdev, "NIC Link is Down\n");
5032 return;
5035 /* Warn user if link speed on NPAR enabled partition is not at
5036 * least 10GB
5038 if (vsi->back->hw.func_caps.npar_enable &&
5039 (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5040 vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5041 netdev_warn(vsi->netdev,
5042 "The partition detected link speed that is less than 10Gbps\n");
5044 switch (vsi->back->hw.phy.link_info.link_speed) {
5045 case I40E_LINK_SPEED_40GB:
5046 speed = "40 G";
5047 break;
5048 case I40E_LINK_SPEED_20GB:
5049 speed = "20 G";
5050 break;
5051 case I40E_LINK_SPEED_10GB:
5052 speed = "10 G";
5053 break;
5054 case I40E_LINK_SPEED_1GB:
5055 speed = "1000 M";
5056 break;
5057 case I40E_LINK_SPEED_100MB:
5058 speed = "100 M";
5059 break;
5060 default:
5061 break;
5064 switch (vsi->back->hw.fc.current_mode) {
5065 case I40E_FC_FULL:
5066 fc = "RX/TX";
5067 break;
5068 case I40E_FC_TX_PAUSE:
5069 fc = "TX";
5070 break;
5071 case I40E_FC_RX_PAUSE:
5072 fc = "RX";
5073 break;
5074 default:
5075 fc = "None";
5076 break;
5079 netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5080 speed, fc);
5084 * i40e_up_complete - Finish the last steps of bringing up a connection
5085 * @vsi: the VSI being configured
5087 static int i40e_up_complete(struct i40e_vsi *vsi)
5089 struct i40e_pf *pf = vsi->back;
5090 int err;
5092 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5093 i40e_vsi_configure_msix(vsi);
5094 else
5095 i40e_configure_msi_and_legacy(vsi);
5097 /* start rings */
5098 err = i40e_vsi_control_rings(vsi, true);
5099 if (err)
5100 return err;
5102 clear_bit(__I40E_DOWN, &vsi->state);
5103 i40e_napi_enable_all(vsi);
5104 i40e_vsi_enable_irq(vsi);
5106 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5107 (vsi->netdev)) {
5108 i40e_print_link_message(vsi, true);
5109 netif_tx_start_all_queues(vsi->netdev);
5110 netif_carrier_on(vsi->netdev);
5111 } else if (vsi->netdev) {
5112 i40e_print_link_message(vsi, false);
5113 /* need to check for qualified module here*/
5114 if ((pf->hw.phy.link_info.link_info &
5115 I40E_AQ_MEDIA_AVAILABLE) &&
5116 (!(pf->hw.phy.link_info.an_info &
5117 I40E_AQ_QUALIFIED_MODULE)))
5118 netdev_err(vsi->netdev,
5119 "the driver failed to link because an unqualified module was detected.");
5122 /* replay FDIR SB filters */
5123 if (vsi->type == I40E_VSI_FDIR) {
5124 /* reset fd counters */
5125 pf->fd_add_err = pf->fd_atr_cnt = 0;
5126 if (pf->fd_tcp_rule > 0) {
5127 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5128 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5129 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5130 pf->fd_tcp_rule = 0;
5132 i40e_fdir_filter_restore(vsi);
5134 i40e_service_event_schedule(pf);
5136 return 0;
5140 * i40e_vsi_reinit_locked - Reset the VSI
5141 * @vsi: the VSI being configured
5143 * Rebuild the ring structs after some configuration
5144 * has changed, e.g. MTU size.
5146 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5148 struct i40e_pf *pf = vsi->back;
5150 WARN_ON(in_interrupt());
5151 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5152 usleep_range(1000, 2000);
5153 i40e_down(vsi);
5155 /* Give a VF some time to respond to the reset. The
5156 * two second wait is based upon the watchdog cycle in
5157 * the VF driver.
5159 if (vsi->type == I40E_VSI_SRIOV)
5160 msleep(2000);
5161 i40e_up(vsi);
5162 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5166 * i40e_up - Bring the connection back up after being down
5167 * @vsi: the VSI being configured
5169 int i40e_up(struct i40e_vsi *vsi)
5171 int err;
5173 err = i40e_vsi_configure(vsi);
5174 if (!err)
5175 err = i40e_up_complete(vsi);
5177 return err;
5181 * i40e_down - Shutdown the connection processing
5182 * @vsi: the VSI being stopped
5184 void i40e_down(struct i40e_vsi *vsi)
5186 int i;
5188 /* It is assumed that the caller of this function
5189 * sets the vsi->state __I40E_DOWN bit.
5191 if (vsi->netdev) {
5192 netif_carrier_off(vsi->netdev);
5193 netif_tx_disable(vsi->netdev);
5195 i40e_vsi_disable_irq(vsi);
5196 i40e_vsi_control_rings(vsi, false);
5197 i40e_napi_disable_all(vsi);
5199 for (i = 0; i < vsi->num_queue_pairs; i++) {
5200 i40e_clean_tx_ring(vsi->tx_rings[i]);
5201 i40e_clean_rx_ring(vsi->rx_rings[i]);
5206 * i40e_setup_tc - configure multiple traffic classes
5207 * @netdev: net device to configure
5208 * @tc: number of traffic classes to enable
5210 #ifdef I40E_FCOE
5211 int i40e_setup_tc(struct net_device *netdev, u8 tc)
5212 #else
5213 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5214 #endif
5216 struct i40e_netdev_priv *np = netdev_priv(netdev);
5217 struct i40e_vsi *vsi = np->vsi;
5218 struct i40e_pf *pf = vsi->back;
5219 u8 enabled_tc = 0;
5220 int ret = -EINVAL;
5221 int i;
5223 /* Check if DCB enabled to continue */
5224 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5225 netdev_info(netdev, "DCB is not enabled for adapter\n");
5226 goto exit;
5229 /* Check if MFP enabled */
5230 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5231 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5232 goto exit;
5235 /* Check whether tc count is within enabled limit */
5236 if (tc > i40e_pf_get_num_tc(pf)) {
5237 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5238 goto exit;
5241 /* Generate TC map for number of tc requested */
5242 for (i = 0; i < tc; i++)
5243 enabled_tc |= BIT_ULL(i);
5245 /* Requesting same TC configuration as already enabled */
5246 if (enabled_tc == vsi->tc_config.enabled_tc)
5247 return 0;
5249 /* Quiesce VSI queues */
5250 i40e_quiesce_vsi(vsi);
5252 /* Configure VSI for enabled TCs */
5253 ret = i40e_vsi_config_tc(vsi, enabled_tc);
5254 if (ret) {
5255 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5256 vsi->seid);
5257 goto exit;
5260 /* Unquiesce VSI */
5261 i40e_unquiesce_vsi(vsi);
5263 exit:
5264 return ret;
5268 * i40e_open - Called when a network interface is made active
5269 * @netdev: network interface device structure
5271 * The open entry point is called when a network interface is made
5272 * active by the system (IFF_UP). At this point all resources needed
5273 * for transmit and receive operations are allocated, the interrupt
5274 * handler is registered with the OS, the netdev watchdog subtask is
5275 * enabled, and the stack is notified that the interface is ready.
5277 * Returns 0 on success, negative value on failure
5279 int i40e_open(struct net_device *netdev)
5281 struct i40e_netdev_priv *np = netdev_priv(netdev);
5282 struct i40e_vsi *vsi = np->vsi;
5283 struct i40e_pf *pf = vsi->back;
5284 int err;
5286 /* disallow open during test or if eeprom is broken */
5287 if (test_bit(__I40E_TESTING, &pf->state) ||
5288 test_bit(__I40E_BAD_EEPROM, &pf->state))
5289 return -EBUSY;
5291 netif_carrier_off(netdev);
5293 err = i40e_vsi_open(vsi);
5294 if (err)
5295 return err;
5297 /* configure global TSO hardware offload settings */
5298 wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5299 TCP_FLAG_FIN) >> 16);
5300 wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5301 TCP_FLAG_FIN |
5302 TCP_FLAG_CWR) >> 16);
5303 wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5305 #ifdef CONFIG_I40E_VXLAN
5306 vxlan_get_rx_port(netdev);
5307 #endif
5309 return 0;
5313 * i40e_vsi_open -
5314 * @vsi: the VSI to open
5316 * Finish initialization of the VSI.
5318 * Returns 0 on success, negative value on failure
5320 int i40e_vsi_open(struct i40e_vsi *vsi)
5322 struct i40e_pf *pf = vsi->back;
5323 char int_name[I40E_INT_NAME_STR_LEN];
5324 int err;
5326 /* allocate descriptors */
5327 err = i40e_vsi_setup_tx_resources(vsi);
5328 if (err)
5329 goto err_setup_tx;
5330 err = i40e_vsi_setup_rx_resources(vsi);
5331 if (err)
5332 goto err_setup_rx;
5334 err = i40e_vsi_configure(vsi);
5335 if (err)
5336 goto err_setup_rx;
5338 if (vsi->netdev) {
5339 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5340 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5341 err = i40e_vsi_request_irq(vsi, int_name);
5342 if (err)
5343 goto err_setup_rx;
5345 /* Notify the stack of the actual queue counts. */
5346 err = netif_set_real_num_tx_queues(vsi->netdev,
5347 vsi->num_queue_pairs);
5348 if (err)
5349 goto err_set_queues;
5351 err = netif_set_real_num_rx_queues(vsi->netdev,
5352 vsi->num_queue_pairs);
5353 if (err)
5354 goto err_set_queues;
5356 } else if (vsi->type == I40E_VSI_FDIR) {
5357 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5358 dev_driver_string(&pf->pdev->dev),
5359 dev_name(&pf->pdev->dev));
5360 err = i40e_vsi_request_irq(vsi, int_name);
5362 } else {
5363 err = -EINVAL;
5364 goto err_setup_rx;
5367 err = i40e_up_complete(vsi);
5368 if (err)
5369 goto err_up_complete;
5371 return 0;
5373 err_up_complete:
5374 i40e_down(vsi);
5375 err_set_queues:
5376 i40e_vsi_free_irq(vsi);
5377 err_setup_rx:
5378 i40e_vsi_free_rx_resources(vsi);
5379 err_setup_tx:
5380 i40e_vsi_free_tx_resources(vsi);
5381 if (vsi == pf->vsi[pf->lan_vsi])
5382 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5384 return err;
5388 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5389 * @pf: Pointer to PF
5391 * This function destroys the hlist where all the Flow Director
5392 * filters were saved.
5394 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5396 struct i40e_fdir_filter *filter;
5397 struct hlist_node *node2;
5399 hlist_for_each_entry_safe(filter, node2,
5400 &pf->fdir_filter_list, fdir_node) {
5401 hlist_del(&filter->fdir_node);
5402 kfree(filter);
5404 pf->fdir_pf_active_filters = 0;
5408 * i40e_close - Disables a network interface
5409 * @netdev: network interface device structure
5411 * The close entry point is called when an interface is de-activated
5412 * by the OS. The hardware is still under the driver's control, but
5413 * this netdev interface is disabled.
5415 * Returns 0, this is not allowed to fail
5417 #ifdef I40E_FCOE
5418 int i40e_close(struct net_device *netdev)
5419 #else
5420 static int i40e_close(struct net_device *netdev)
5421 #endif
5423 struct i40e_netdev_priv *np = netdev_priv(netdev);
5424 struct i40e_vsi *vsi = np->vsi;
5426 i40e_vsi_close(vsi);
5428 return 0;
5432 * i40e_do_reset - Start a PF or Core Reset sequence
5433 * @pf: board private structure
5434 * @reset_flags: which reset is requested
5436 * The essential difference in resets is that the PF Reset
5437 * doesn't clear the packet buffers, doesn't reset the PE
5438 * firmware, and doesn't bother the other PFs on the chip.
5440 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5442 u32 val;
5444 WARN_ON(in_interrupt());
5446 if (i40e_check_asq_alive(&pf->hw))
5447 i40e_vc_notify_reset(pf);
5449 /* do the biggest reset indicated */
5450 if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5452 /* Request a Global Reset
5454 * This will start the chip's countdown to the actual full
5455 * chip reset event, and a warning interrupt to be sent
5456 * to all PFs, including the requestor. Our handler
5457 * for the warning interrupt will deal with the shutdown
5458 * and recovery of the switch setup.
5460 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5461 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5462 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5463 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5465 } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5467 /* Request a Core Reset
5469 * Same as Global Reset, except does *not* include the MAC/PHY
5471 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5472 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5473 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5474 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5475 i40e_flush(&pf->hw);
5477 } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5479 /* Request a PF Reset
5481 * Resets only the PF-specific registers
5483 * This goes directly to the tear-down and rebuild of
5484 * the switch, since we need to do all the recovery as
5485 * for the Core Reset.
5487 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5488 i40e_handle_reset_warning(pf);
5490 } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5491 int v;
5493 /* Find the VSI(s) that requested a re-init */
5494 dev_info(&pf->pdev->dev,
5495 "VSI reinit requested\n");
5496 for (v = 0; v < pf->num_alloc_vsi; v++) {
5497 struct i40e_vsi *vsi = pf->vsi[v];
5499 if (vsi != NULL &&
5500 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5501 i40e_vsi_reinit_locked(pf->vsi[v]);
5502 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5505 } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5506 int v;
5508 /* Find the VSI(s) that needs to be brought down */
5509 dev_info(&pf->pdev->dev, "VSI down requested\n");
5510 for (v = 0; v < pf->num_alloc_vsi; v++) {
5511 struct i40e_vsi *vsi = pf->vsi[v];
5513 if (vsi != NULL &&
5514 test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5515 set_bit(__I40E_DOWN, &vsi->state);
5516 i40e_down(vsi);
5517 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5520 } else {
5521 dev_info(&pf->pdev->dev,
5522 "bad reset request 0x%08x\n", reset_flags);
5526 #ifdef CONFIG_I40E_DCB
5528 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5529 * @pf: board private structure
5530 * @old_cfg: current DCB config
5531 * @new_cfg: new DCB config
5533 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5534 struct i40e_dcbx_config *old_cfg,
5535 struct i40e_dcbx_config *new_cfg)
5537 bool need_reconfig = false;
5539 /* Check if ETS configuration has changed */
5540 if (memcmp(&new_cfg->etscfg,
5541 &old_cfg->etscfg,
5542 sizeof(new_cfg->etscfg))) {
5543 /* If Priority Table has changed reconfig is needed */
5544 if (memcmp(&new_cfg->etscfg.prioritytable,
5545 &old_cfg->etscfg.prioritytable,
5546 sizeof(new_cfg->etscfg.prioritytable))) {
5547 need_reconfig = true;
5548 dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5551 if (memcmp(&new_cfg->etscfg.tcbwtable,
5552 &old_cfg->etscfg.tcbwtable,
5553 sizeof(new_cfg->etscfg.tcbwtable)))
5554 dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5556 if (memcmp(&new_cfg->etscfg.tsatable,
5557 &old_cfg->etscfg.tsatable,
5558 sizeof(new_cfg->etscfg.tsatable)))
5559 dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5562 /* Check if PFC configuration has changed */
5563 if (memcmp(&new_cfg->pfc,
5564 &old_cfg->pfc,
5565 sizeof(new_cfg->pfc))) {
5566 need_reconfig = true;
5567 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5570 /* Check if APP Table has changed */
5571 if (memcmp(&new_cfg->app,
5572 &old_cfg->app,
5573 sizeof(new_cfg->app))) {
5574 need_reconfig = true;
5575 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5578 dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5579 return need_reconfig;
5583 * i40e_handle_lldp_event - Handle LLDP Change MIB event
5584 * @pf: board private structure
5585 * @e: event info posted on ARQ
5587 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5588 struct i40e_arq_event_info *e)
5590 struct i40e_aqc_lldp_get_mib *mib =
5591 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5592 struct i40e_hw *hw = &pf->hw;
5593 struct i40e_dcbx_config tmp_dcbx_cfg;
5594 bool need_reconfig = false;
5595 int ret = 0;
5596 u8 type;
5598 /* Not DCB capable or capability disabled */
5599 if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5600 return ret;
5602 /* Ignore if event is not for Nearest Bridge */
5603 type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5604 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5605 dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5606 if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5607 return ret;
5609 /* Check MIB Type and return if event for Remote MIB update */
5610 type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5611 dev_dbg(&pf->pdev->dev,
5612 "LLDP event mib type %s\n", type ? "remote" : "local");
5613 if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5614 /* Update the remote cached instance and return */
5615 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5616 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5617 &hw->remote_dcbx_config);
5618 goto exit;
5621 /* Store the old configuration */
5622 tmp_dcbx_cfg = hw->local_dcbx_config;
5624 /* Reset the old DCBx configuration data */
5625 memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5626 /* Get updated DCBX data from firmware */
5627 ret = i40e_get_dcb_config(&pf->hw);
5628 if (ret) {
5629 dev_info(&pf->pdev->dev,
5630 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5631 i40e_stat_str(&pf->hw, ret),
5632 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5633 goto exit;
5636 /* No change detected in DCBX configs */
5637 if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5638 sizeof(tmp_dcbx_cfg))) {
5639 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5640 goto exit;
5643 need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5644 &hw->local_dcbx_config);
5646 i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5648 if (!need_reconfig)
5649 goto exit;
5651 /* Enable DCB tagging only when more than one TC */
5652 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5653 pf->flags |= I40E_FLAG_DCB_ENABLED;
5654 else
5655 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5657 set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5658 /* Reconfiguration needed quiesce all VSIs */
5659 i40e_pf_quiesce_all_vsi(pf);
5661 /* Changes in configuration update VEB/VSI */
5662 i40e_dcb_reconfigure(pf);
5664 ret = i40e_resume_port_tx(pf);
5666 clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5667 /* In case of error no point in resuming VSIs */
5668 if (ret)
5669 goto exit;
5671 /* Wait for the PF's Tx queues to be disabled */
5672 ret = i40e_pf_wait_txq_disabled(pf);
5673 if (ret) {
5674 /* Schedule PF reset to recover */
5675 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5676 i40e_service_event_schedule(pf);
5677 } else {
5678 i40e_pf_unquiesce_all_vsi(pf);
5681 exit:
5682 return ret;
5684 #endif /* CONFIG_I40E_DCB */
5687 * i40e_do_reset_safe - Protected reset path for userland calls.
5688 * @pf: board private structure
5689 * @reset_flags: which reset is requested
5692 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5694 rtnl_lock();
5695 i40e_do_reset(pf, reset_flags);
5696 rtnl_unlock();
5700 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5701 * @pf: board private structure
5702 * @e: event info posted on ARQ
5704 * Handler for LAN Queue Overflow Event generated by the firmware for PF
5705 * and VF queues
5707 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5708 struct i40e_arq_event_info *e)
5710 struct i40e_aqc_lan_overflow *data =
5711 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5712 u32 queue = le32_to_cpu(data->prtdcb_rupto);
5713 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5714 struct i40e_hw *hw = &pf->hw;
5715 struct i40e_vf *vf;
5716 u16 vf_id;
5718 dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5719 queue, qtx_ctl);
5721 /* Queue belongs to VF, find the VF and issue VF reset */
5722 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5723 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5724 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5725 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5726 vf_id -= hw->func_caps.vf_base_id;
5727 vf = &pf->vf[vf_id];
5728 i40e_vc_notify_vf_reset(vf);
5729 /* Allow VF to process pending reset notification */
5730 msleep(20);
5731 i40e_reset_vf(vf, false);
5736 * i40e_service_event_complete - Finish up the service event
5737 * @pf: board private structure
5739 static void i40e_service_event_complete(struct i40e_pf *pf)
5741 BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5743 /* flush memory to make sure state is correct before next watchog */
5744 smp_mb__before_atomic();
5745 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5749 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5750 * @pf: board private structure
5752 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5754 u32 val, fcnt_prog;
5756 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5757 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5758 return fcnt_prog;
5762 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5763 * @pf: board private structure
5765 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5767 u32 val, fcnt_prog;
5769 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5770 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5771 ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5772 I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5773 return fcnt_prog;
5777 * i40e_get_global_fd_count - Get total FD filters programmed on device
5778 * @pf: board private structure
5780 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5782 u32 val, fcnt_prog;
5784 val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5785 fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5786 ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5787 I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5788 return fcnt_prog;
5792 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5793 * @pf: board private structure
5795 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5797 struct i40e_fdir_filter *filter;
5798 u32 fcnt_prog, fcnt_avail;
5799 struct hlist_node *node;
5801 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5802 return;
5804 /* Check if, FD SB or ATR was auto disabled and if there is enough room
5805 * to re-enable
5807 fcnt_prog = i40e_get_global_fd_count(pf);
5808 fcnt_avail = pf->fdir_pf_filter_count;
5809 if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5810 (pf->fd_add_err == 0) ||
5811 (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5812 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5813 (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5814 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5815 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5816 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5819 /* Wait for some more space to be available to turn on ATR */
5820 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5821 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5822 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5823 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5824 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5825 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5829 /* if hw had a problem adding a filter, delete it */
5830 if (pf->fd_inv > 0) {
5831 hlist_for_each_entry_safe(filter, node,
5832 &pf->fdir_filter_list, fdir_node) {
5833 if (filter->fd_id == pf->fd_inv) {
5834 hlist_del(&filter->fdir_node);
5835 kfree(filter);
5836 pf->fdir_pf_active_filters--;
5842 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5843 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5845 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5846 * @pf: board private structure
5848 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5850 unsigned long min_flush_time;
5851 int flush_wait_retry = 50;
5852 bool disable_atr = false;
5853 int fd_room;
5854 int reg;
5856 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5857 return;
5859 if (!time_after(jiffies, pf->fd_flush_timestamp +
5860 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5861 return;
5863 /* If the flush is happening too quick and we have mostly SB rules we
5864 * should not re-enable ATR for some time.
5866 min_flush_time = pf->fd_flush_timestamp +
5867 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5868 fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5870 if (!(time_after(jiffies, min_flush_time)) &&
5871 (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5872 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5873 dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5874 disable_atr = true;
5877 pf->fd_flush_timestamp = jiffies;
5878 pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5879 /* flush all filters */
5880 wr32(&pf->hw, I40E_PFQF_CTL_1,
5881 I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5882 i40e_flush(&pf->hw);
5883 pf->fd_flush_cnt++;
5884 pf->fd_add_err = 0;
5885 do {
5886 /* Check FD flush status every 5-6msec */
5887 usleep_range(5000, 6000);
5888 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5889 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5890 break;
5891 } while (flush_wait_retry--);
5892 if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5893 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5894 } else {
5895 /* replay sideband filters */
5896 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5897 if (!disable_atr)
5898 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5899 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5900 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5901 dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5907 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5908 * @pf: board private structure
5910 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5912 return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5915 /* We can see up to 256 filter programming desc in transit if the filters are
5916 * being applied really fast; before we see the first
5917 * filter miss error on Rx queue 0. Accumulating enough error messages before
5918 * reacting will make sure we don't cause flush too often.
5920 #define I40E_MAX_FD_PROGRAM_ERROR 256
5923 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5924 * @pf: board private structure
5926 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5929 /* if interface is down do nothing */
5930 if (test_bit(__I40E_DOWN, &pf->state))
5931 return;
5933 if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5934 return;
5936 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5937 i40e_fdir_flush_and_replay(pf);
5939 i40e_fdir_check_and_reenable(pf);
5944 * i40e_vsi_link_event - notify VSI of a link event
5945 * @vsi: vsi to be notified
5946 * @link_up: link up or down
5948 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5950 if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5951 return;
5953 switch (vsi->type) {
5954 case I40E_VSI_MAIN:
5955 #ifdef I40E_FCOE
5956 case I40E_VSI_FCOE:
5957 #endif
5958 if (!vsi->netdev || !vsi->netdev_registered)
5959 break;
5961 if (link_up) {
5962 netif_carrier_on(vsi->netdev);
5963 netif_tx_wake_all_queues(vsi->netdev);
5964 } else {
5965 netif_carrier_off(vsi->netdev);
5966 netif_tx_stop_all_queues(vsi->netdev);
5968 break;
5970 case I40E_VSI_SRIOV:
5971 case I40E_VSI_VMDQ2:
5972 case I40E_VSI_CTRL:
5973 case I40E_VSI_MIRROR:
5974 default:
5975 /* there is no notification for other VSIs */
5976 break;
5981 * i40e_veb_link_event - notify elements on the veb of a link event
5982 * @veb: veb to be notified
5983 * @link_up: link up or down
5985 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
5987 struct i40e_pf *pf;
5988 int i;
5990 if (!veb || !veb->pf)
5991 return;
5992 pf = veb->pf;
5994 /* depth first... */
5995 for (i = 0; i < I40E_MAX_VEB; i++)
5996 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
5997 i40e_veb_link_event(pf->veb[i], link_up);
5999 /* ... now the local VSIs */
6000 for (i = 0; i < pf->num_alloc_vsi; i++)
6001 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6002 i40e_vsi_link_event(pf->vsi[i], link_up);
6006 * i40e_link_event - Update netif_carrier status
6007 * @pf: board private structure
6009 static void i40e_link_event(struct i40e_pf *pf)
6011 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6012 u8 new_link_speed, old_link_speed;
6013 i40e_status status;
6014 bool new_link, old_link;
6016 /* set this to force the get_link_status call to refresh state */
6017 pf->hw.phy.get_link_info = true;
6019 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6021 status = i40e_get_link_status(&pf->hw, &new_link);
6022 if (status) {
6023 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6024 status);
6025 return;
6028 old_link_speed = pf->hw.phy.link_info_old.link_speed;
6029 new_link_speed = pf->hw.phy.link_info.link_speed;
6031 if (new_link == old_link &&
6032 new_link_speed == old_link_speed &&
6033 (test_bit(__I40E_DOWN, &vsi->state) ||
6034 new_link == netif_carrier_ok(vsi->netdev)))
6035 return;
6037 if (!test_bit(__I40E_DOWN, &vsi->state))
6038 i40e_print_link_message(vsi, new_link);
6040 /* Notify the base of the switch tree connected to
6041 * the link. Floating VEBs are not notified.
6043 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6044 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6045 else
6046 i40e_vsi_link_event(vsi, new_link);
6048 if (pf->vf)
6049 i40e_vc_notify_link_state(pf);
6051 if (pf->flags & I40E_FLAG_PTP)
6052 i40e_ptp_set_increment(pf);
6056 * i40e_watchdog_subtask - periodic checks not using event driven response
6057 * @pf: board private structure
6059 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6061 int i;
6063 /* if interface is down do nothing */
6064 if (test_bit(__I40E_DOWN, &pf->state) ||
6065 test_bit(__I40E_CONFIG_BUSY, &pf->state))
6066 return;
6068 /* make sure we don't do these things too often */
6069 if (time_before(jiffies, (pf->service_timer_previous +
6070 pf->service_timer_period)))
6071 return;
6072 pf->service_timer_previous = jiffies;
6074 if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6075 i40e_link_event(pf);
6077 /* Update the stats for active netdevs so the network stack
6078 * can look at updated numbers whenever it cares to
6080 for (i = 0; i < pf->num_alloc_vsi; i++)
6081 if (pf->vsi[i] && pf->vsi[i]->netdev)
6082 i40e_update_stats(pf->vsi[i]);
6084 if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6085 /* Update the stats for the active switching components */
6086 for (i = 0; i < I40E_MAX_VEB; i++)
6087 if (pf->veb[i])
6088 i40e_update_veb_stats(pf->veb[i]);
6091 i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6095 * i40e_reset_subtask - Set up for resetting the device and driver
6096 * @pf: board private structure
6098 static void i40e_reset_subtask(struct i40e_pf *pf)
6100 u32 reset_flags = 0;
6102 rtnl_lock();
6103 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6104 reset_flags |= BIT_ULL(__I40E_REINIT_REQUESTED);
6105 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6107 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6108 reset_flags |= BIT_ULL(__I40E_PF_RESET_REQUESTED);
6109 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6111 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6112 reset_flags |= BIT_ULL(__I40E_CORE_RESET_REQUESTED);
6113 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6115 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6116 reset_flags |= BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED);
6117 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6119 if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6120 reset_flags |= BIT_ULL(__I40E_DOWN_REQUESTED);
6121 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6124 /* If there's a recovery already waiting, it takes
6125 * precedence before starting a new reset sequence.
6127 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6128 i40e_handle_reset_warning(pf);
6129 goto unlock;
6132 /* If we're already down or resetting, just bail */
6133 if (reset_flags &&
6134 !test_bit(__I40E_DOWN, &pf->state) &&
6135 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6136 i40e_do_reset(pf, reset_flags);
6138 unlock:
6139 rtnl_unlock();
6143 * i40e_handle_link_event - Handle link event
6144 * @pf: board private structure
6145 * @e: event info posted on ARQ
6147 static void i40e_handle_link_event(struct i40e_pf *pf,
6148 struct i40e_arq_event_info *e)
6150 struct i40e_hw *hw = &pf->hw;
6151 struct i40e_aqc_get_link_status *status =
6152 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6154 /* save off old link status information */
6155 hw->phy.link_info_old = hw->phy.link_info;
6157 /* Do a new status request to re-enable LSE reporting
6158 * and load new status information into the hw struct
6159 * This completely ignores any state information
6160 * in the ARQ event info, instead choosing to always
6161 * issue the AQ update link status command.
6163 i40e_link_event(pf);
6165 /* check for unqualified module, if link is down */
6166 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6167 (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6168 (!(status->link_info & I40E_AQ_LINK_UP)))
6169 dev_err(&pf->pdev->dev,
6170 "The driver failed to link because an unqualified module was detected.\n");
6174 * i40e_clean_adminq_subtask - Clean the AdminQ rings
6175 * @pf: board private structure
6177 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6179 struct i40e_arq_event_info event;
6180 struct i40e_hw *hw = &pf->hw;
6181 u16 pending, i = 0;
6182 i40e_status ret;
6183 u16 opcode;
6184 u32 oldval;
6185 u32 val;
6187 /* Do not run clean AQ when PF reset fails */
6188 if (test_bit(__I40E_RESET_FAILED, &pf->state))
6189 return;
6191 /* check for error indications */
6192 val = rd32(&pf->hw, pf->hw.aq.arq.len);
6193 oldval = val;
6194 if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6195 dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6196 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6198 if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6199 dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6200 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6202 if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6203 dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6204 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6206 if (oldval != val)
6207 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6209 val = rd32(&pf->hw, pf->hw.aq.asq.len);
6210 oldval = val;
6211 if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6212 dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6213 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6215 if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6216 dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6217 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6219 if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6220 dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6221 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6223 if (oldval != val)
6224 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6226 event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6227 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6228 if (!event.msg_buf)
6229 return;
6231 do {
6232 ret = i40e_clean_arq_element(hw, &event, &pending);
6233 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6234 break;
6235 else if (ret) {
6236 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6237 break;
6240 opcode = le16_to_cpu(event.desc.opcode);
6241 switch (opcode) {
6243 case i40e_aqc_opc_get_link_status:
6244 i40e_handle_link_event(pf, &event);
6245 break;
6246 case i40e_aqc_opc_send_msg_to_pf:
6247 ret = i40e_vc_process_vf_msg(pf,
6248 le16_to_cpu(event.desc.retval),
6249 le32_to_cpu(event.desc.cookie_high),
6250 le32_to_cpu(event.desc.cookie_low),
6251 event.msg_buf,
6252 event.msg_len);
6253 break;
6254 case i40e_aqc_opc_lldp_update_mib:
6255 dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6256 #ifdef CONFIG_I40E_DCB
6257 rtnl_lock();
6258 ret = i40e_handle_lldp_event(pf, &event);
6259 rtnl_unlock();
6260 #endif /* CONFIG_I40E_DCB */
6261 break;
6262 case i40e_aqc_opc_event_lan_overflow:
6263 dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6264 i40e_handle_lan_overflow_event(pf, &event);
6265 break;
6266 case i40e_aqc_opc_send_msg_to_peer:
6267 dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6268 break;
6269 case i40e_aqc_opc_nvm_erase:
6270 case i40e_aqc_opc_nvm_update:
6271 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
6272 break;
6273 default:
6274 dev_info(&pf->pdev->dev,
6275 "ARQ Error: Unknown event 0x%04x received\n",
6276 opcode);
6277 break;
6279 } while (pending && (i++ < pf->adminq_work_limit));
6281 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6282 /* re-enable Admin queue interrupt cause */
6283 val = rd32(hw, I40E_PFINT_ICR0_ENA);
6284 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6285 wr32(hw, I40E_PFINT_ICR0_ENA, val);
6286 i40e_flush(hw);
6288 kfree(event.msg_buf);
6292 * i40e_verify_eeprom - make sure eeprom is good to use
6293 * @pf: board private structure
6295 static void i40e_verify_eeprom(struct i40e_pf *pf)
6297 int err;
6299 err = i40e_diag_eeprom_test(&pf->hw);
6300 if (err) {
6301 /* retry in case of garbage read */
6302 err = i40e_diag_eeprom_test(&pf->hw);
6303 if (err) {
6304 dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6305 err);
6306 set_bit(__I40E_BAD_EEPROM, &pf->state);
6310 if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6311 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6312 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6317 * i40e_enable_pf_switch_lb
6318 * @pf: pointer to the PF structure
6320 * enable switch loop back or die - no point in a return value
6322 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6324 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6325 struct i40e_vsi_context ctxt;
6326 int ret;
6328 ctxt.seid = pf->main_vsi_seid;
6329 ctxt.pf_num = pf->hw.pf_id;
6330 ctxt.vf_num = 0;
6331 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6332 if (ret) {
6333 dev_info(&pf->pdev->dev,
6334 "couldn't get PF vsi config, err %s aq_err %s\n",
6335 i40e_stat_str(&pf->hw, ret),
6336 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6337 return;
6339 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6340 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6341 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6343 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6344 if (ret) {
6345 dev_info(&pf->pdev->dev,
6346 "update vsi switch failed, err %s aq_err %s\n",
6347 i40e_stat_str(&pf->hw, ret),
6348 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6353 * i40e_disable_pf_switch_lb
6354 * @pf: pointer to the PF structure
6356 * disable switch loop back or die - no point in a return value
6358 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6360 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6361 struct i40e_vsi_context ctxt;
6362 int ret;
6364 ctxt.seid = pf->main_vsi_seid;
6365 ctxt.pf_num = pf->hw.pf_id;
6366 ctxt.vf_num = 0;
6367 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6368 if (ret) {
6369 dev_info(&pf->pdev->dev,
6370 "couldn't get PF vsi config, err %s aq_err %s\n",
6371 i40e_stat_str(&pf->hw, ret),
6372 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6373 return;
6375 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6376 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6377 ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6379 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6380 if (ret) {
6381 dev_info(&pf->pdev->dev,
6382 "update vsi switch failed, err %s aq_err %s\n",
6383 i40e_stat_str(&pf->hw, ret),
6384 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6389 * i40e_config_bridge_mode - Configure the HW bridge mode
6390 * @veb: pointer to the bridge instance
6392 * Configure the loop back mode for the LAN VSI that is downlink to the
6393 * specified HW bridge instance. It is expected this function is called
6394 * when a new HW bridge is instantiated.
6396 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6398 struct i40e_pf *pf = veb->pf;
6400 if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6401 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6402 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6403 if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6404 i40e_disable_pf_switch_lb(pf);
6405 else
6406 i40e_enable_pf_switch_lb(pf);
6410 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6411 * @veb: pointer to the VEB instance
6413 * This is a recursive function that first builds the attached VSIs then
6414 * recurses in to build the next layer of VEB. We track the connections
6415 * through our own index numbers because the seid's from the HW could
6416 * change across the reset.
6418 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6420 struct i40e_vsi *ctl_vsi = NULL;
6421 struct i40e_pf *pf = veb->pf;
6422 int v, veb_idx;
6423 int ret;
6425 /* build VSI that owns this VEB, temporarily attached to base VEB */
6426 for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6427 if (pf->vsi[v] &&
6428 pf->vsi[v]->veb_idx == veb->idx &&
6429 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6430 ctl_vsi = pf->vsi[v];
6431 break;
6434 if (!ctl_vsi) {
6435 dev_info(&pf->pdev->dev,
6436 "missing owner VSI for veb_idx %d\n", veb->idx);
6437 ret = -ENOENT;
6438 goto end_reconstitute;
6440 if (ctl_vsi != pf->vsi[pf->lan_vsi])
6441 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6442 ret = i40e_add_vsi(ctl_vsi);
6443 if (ret) {
6444 dev_info(&pf->pdev->dev,
6445 "rebuild of veb_idx %d owner VSI failed: %d\n",
6446 veb->idx, ret);
6447 goto end_reconstitute;
6449 i40e_vsi_reset_stats(ctl_vsi);
6451 /* create the VEB in the switch and move the VSI onto the VEB */
6452 ret = i40e_add_veb(veb, ctl_vsi);
6453 if (ret)
6454 goto end_reconstitute;
6456 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6457 veb->bridge_mode = BRIDGE_MODE_VEB;
6458 else
6459 veb->bridge_mode = BRIDGE_MODE_VEPA;
6460 i40e_config_bridge_mode(veb);
6462 /* create the remaining VSIs attached to this VEB */
6463 for (v = 0; v < pf->num_alloc_vsi; v++) {
6464 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6465 continue;
6467 if (pf->vsi[v]->veb_idx == veb->idx) {
6468 struct i40e_vsi *vsi = pf->vsi[v];
6470 vsi->uplink_seid = veb->seid;
6471 ret = i40e_add_vsi(vsi);
6472 if (ret) {
6473 dev_info(&pf->pdev->dev,
6474 "rebuild of vsi_idx %d failed: %d\n",
6475 v, ret);
6476 goto end_reconstitute;
6478 i40e_vsi_reset_stats(vsi);
6482 /* create any VEBs attached to this VEB - RECURSION */
6483 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6484 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6485 pf->veb[veb_idx]->uplink_seid = veb->seid;
6486 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6487 if (ret)
6488 break;
6492 end_reconstitute:
6493 return ret;
6497 * i40e_get_capabilities - get info about the HW
6498 * @pf: the PF struct
6500 static int i40e_get_capabilities(struct i40e_pf *pf)
6502 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6503 u16 data_size;
6504 int buf_len;
6505 int err;
6507 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6508 do {
6509 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6510 if (!cap_buf)
6511 return -ENOMEM;
6513 /* this loads the data into the hw struct for us */
6514 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6515 &data_size,
6516 i40e_aqc_opc_list_func_capabilities,
6517 NULL);
6518 /* data loaded, buffer no longer needed */
6519 kfree(cap_buf);
6521 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6522 /* retry with a larger buffer */
6523 buf_len = data_size;
6524 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6525 dev_info(&pf->pdev->dev,
6526 "capability discovery failed, err %s aq_err %s\n",
6527 i40e_stat_str(&pf->hw, err),
6528 i40e_aq_str(&pf->hw,
6529 pf->hw.aq.asq_last_status));
6530 return -ENODEV;
6532 } while (err);
6534 if (pf->hw.debug_mask & I40E_DEBUG_USER)
6535 dev_info(&pf->pdev->dev,
6536 "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",
6537 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6538 pf->hw.func_caps.num_msix_vectors,
6539 pf->hw.func_caps.num_msix_vectors_vf,
6540 pf->hw.func_caps.fd_filters_guaranteed,
6541 pf->hw.func_caps.fd_filters_best_effort,
6542 pf->hw.func_caps.num_tx_qp,
6543 pf->hw.func_caps.num_vsis);
6545 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6546 + pf->hw.func_caps.num_vfs)
6547 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6548 dev_info(&pf->pdev->dev,
6549 "got num_vsis %d, setting num_vsis to %d\n",
6550 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6551 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6554 return 0;
6557 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6560 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6561 * @pf: board private structure
6563 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6565 struct i40e_vsi *vsi;
6566 int i;
6568 /* quick workaround for an NVM issue that leaves a critical register
6569 * uninitialized
6571 if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6572 static const u32 hkey[] = {
6573 0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6574 0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6575 0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6576 0x95b3a76d};
6578 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6579 wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6582 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6583 return;
6585 /* find existing VSI and see if it needs configuring */
6586 vsi = NULL;
6587 for (i = 0; i < pf->num_alloc_vsi; i++) {
6588 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6589 vsi = pf->vsi[i];
6590 break;
6594 /* create a new VSI if none exists */
6595 if (!vsi) {
6596 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6597 pf->vsi[pf->lan_vsi]->seid, 0);
6598 if (!vsi) {
6599 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6600 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6601 return;
6605 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6609 * i40e_fdir_teardown - release the Flow Director resources
6610 * @pf: board private structure
6612 static void i40e_fdir_teardown(struct i40e_pf *pf)
6614 int i;
6616 i40e_fdir_filter_exit(pf);
6617 for (i = 0; i < pf->num_alloc_vsi; i++) {
6618 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6619 i40e_vsi_release(pf->vsi[i]);
6620 break;
6626 * i40e_prep_for_reset - prep for the core to reset
6627 * @pf: board private structure
6629 * Close up the VFs and other things in prep for PF Reset.
6631 static void i40e_prep_for_reset(struct i40e_pf *pf)
6633 struct i40e_hw *hw = &pf->hw;
6634 i40e_status ret = 0;
6635 u32 v;
6637 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6638 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6639 return;
6641 dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6643 /* quiesce the VSIs and their queues that are not already DOWN */
6644 i40e_pf_quiesce_all_vsi(pf);
6646 for (v = 0; v < pf->num_alloc_vsi; v++) {
6647 if (pf->vsi[v])
6648 pf->vsi[v]->seid = 0;
6651 i40e_shutdown_adminq(&pf->hw);
6653 /* call shutdown HMC */
6654 if (hw->hmc.hmc_obj) {
6655 ret = i40e_shutdown_lan_hmc(hw);
6656 if (ret)
6657 dev_warn(&pf->pdev->dev,
6658 "shutdown_lan_hmc failed: %d\n", ret);
6663 * i40e_send_version - update firmware with driver version
6664 * @pf: PF struct
6666 static void i40e_send_version(struct i40e_pf *pf)
6668 struct i40e_driver_version dv;
6670 dv.major_version = DRV_VERSION_MAJOR;
6671 dv.minor_version = DRV_VERSION_MINOR;
6672 dv.build_version = DRV_VERSION_BUILD;
6673 dv.subbuild_version = 0;
6674 strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6675 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6679 * i40e_reset_and_rebuild - reset and rebuild using a saved config
6680 * @pf: board private structure
6681 * @reinit: if the Main VSI needs to re-initialized.
6683 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6685 struct i40e_hw *hw = &pf->hw;
6686 u8 set_fc_aq_fail = 0;
6687 i40e_status ret;
6688 u32 v;
6690 /* Now we wait for GRST to settle out.
6691 * We don't have to delete the VEBs or VSIs from the hw switch
6692 * because the reset will make them disappear.
6694 ret = i40e_pf_reset(hw);
6695 if (ret) {
6696 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6697 set_bit(__I40E_RESET_FAILED, &pf->state);
6698 goto clear_recovery;
6700 pf->pfr_count++;
6702 if (test_bit(__I40E_DOWN, &pf->state))
6703 goto clear_recovery;
6704 dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6706 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6707 ret = i40e_init_adminq(&pf->hw);
6708 if (ret) {
6709 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6710 i40e_stat_str(&pf->hw, ret),
6711 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6712 goto clear_recovery;
6715 /* re-verify the eeprom if we just had an EMP reset */
6716 if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6717 i40e_verify_eeprom(pf);
6719 i40e_clear_pxe_mode(hw);
6720 ret = i40e_get_capabilities(pf);
6721 if (ret)
6722 goto end_core_reset;
6724 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6725 hw->func_caps.num_rx_qp,
6726 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6727 if (ret) {
6728 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6729 goto end_core_reset;
6731 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6732 if (ret) {
6733 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6734 goto end_core_reset;
6737 #ifdef CONFIG_I40E_DCB
6738 ret = i40e_init_pf_dcb(pf);
6739 if (ret) {
6740 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6741 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6742 /* Continue without DCB enabled */
6744 #endif /* CONFIG_I40E_DCB */
6745 #ifdef I40E_FCOE
6746 i40e_init_pf_fcoe(pf);
6748 #endif
6749 /* do basic switch setup */
6750 ret = i40e_setup_pf_switch(pf, reinit);
6751 if (ret)
6752 goto end_core_reset;
6754 /* driver is only interested in link up/down and module qualification
6755 * reports from firmware
6757 ret = i40e_aq_set_phy_int_mask(&pf->hw,
6758 I40E_AQ_EVENT_LINK_UPDOWN |
6759 I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6760 if (ret)
6761 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6762 i40e_stat_str(&pf->hw, ret),
6763 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6765 /* make sure our flow control settings are restored */
6766 ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6767 if (ret)
6768 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6769 i40e_stat_str(&pf->hw, ret),
6770 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6772 /* Rebuild the VSIs and VEBs that existed before reset.
6773 * They are still in our local switch element arrays, so only
6774 * need to rebuild the switch model in the HW.
6776 * If there were VEBs but the reconstitution failed, we'll try
6777 * try to recover minimal use by getting the basic PF VSI working.
6779 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6780 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6781 /* find the one VEB connected to the MAC, and find orphans */
6782 for (v = 0; v < I40E_MAX_VEB; v++) {
6783 if (!pf->veb[v])
6784 continue;
6786 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6787 pf->veb[v]->uplink_seid == 0) {
6788 ret = i40e_reconstitute_veb(pf->veb[v]);
6790 if (!ret)
6791 continue;
6793 /* If Main VEB failed, we're in deep doodoo,
6794 * so give up rebuilding the switch and set up
6795 * for minimal rebuild of PF VSI.
6796 * If orphan failed, we'll report the error
6797 * but try to keep going.
6799 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6800 dev_info(&pf->pdev->dev,
6801 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6802 ret);
6803 pf->vsi[pf->lan_vsi]->uplink_seid
6804 = pf->mac_seid;
6805 break;
6806 } else if (pf->veb[v]->uplink_seid == 0) {
6807 dev_info(&pf->pdev->dev,
6808 "rebuild of orphan VEB failed: %d\n",
6809 ret);
6815 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6816 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6817 /* no VEB, so rebuild only the Main VSI */
6818 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6819 if (ret) {
6820 dev_info(&pf->pdev->dev,
6821 "rebuild of Main VSI failed: %d\n", ret);
6822 goto end_core_reset;
6826 if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6827 (pf->hw.aq.fw_maj_ver < 4)) {
6828 msleep(75);
6829 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6830 if (ret)
6831 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6832 i40e_stat_str(&pf->hw, ret),
6833 i40e_aq_str(&pf->hw,
6834 pf->hw.aq.asq_last_status));
6836 /* reinit the misc interrupt */
6837 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6838 ret = i40e_setup_misc_vector(pf);
6840 /* Add a filter to drop all Flow control frames from any VSI from being
6841 * transmitted. By doing so we stop a malicious VF from sending out
6842 * PAUSE or PFC frames and potentially controlling traffic for other
6843 * PF/VF VSIs.
6844 * The FW can still send Flow control frames if enabled.
6846 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6847 pf->main_vsi_seid);
6849 /* restart the VSIs that were rebuilt and running before the reset */
6850 i40e_pf_unquiesce_all_vsi(pf);
6852 if (pf->num_alloc_vfs) {
6853 for (v = 0; v < pf->num_alloc_vfs; v++)
6854 i40e_reset_vf(&pf->vf[v], true);
6857 /* tell the firmware that we're starting */
6858 i40e_send_version(pf);
6860 end_core_reset:
6861 clear_bit(__I40E_RESET_FAILED, &pf->state);
6862 clear_recovery:
6863 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6867 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6868 * @pf: board private structure
6870 * Close up the VFs and other things in prep for a Core Reset,
6871 * then get ready to rebuild the world.
6873 static void i40e_handle_reset_warning(struct i40e_pf *pf)
6875 i40e_prep_for_reset(pf);
6876 i40e_reset_and_rebuild(pf, false);
6880 * i40e_handle_mdd_event
6881 * @pf: pointer to the PF structure
6883 * Called from the MDD irq handler to identify possibly malicious vfs
6885 static void i40e_handle_mdd_event(struct i40e_pf *pf)
6887 struct i40e_hw *hw = &pf->hw;
6888 bool mdd_detected = false;
6889 bool pf_mdd_detected = false;
6890 struct i40e_vf *vf;
6891 u32 reg;
6892 int i;
6894 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6895 return;
6897 /* find what triggered the MDD event */
6898 reg = rd32(hw, I40E_GL_MDET_TX);
6899 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6900 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6901 I40E_GL_MDET_TX_PF_NUM_SHIFT;
6902 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6903 I40E_GL_MDET_TX_VF_NUM_SHIFT;
6904 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6905 I40E_GL_MDET_TX_EVENT_SHIFT;
6906 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6907 I40E_GL_MDET_TX_QUEUE_SHIFT) -
6908 pf->hw.func_caps.base_queue;
6909 if (netif_msg_tx_err(pf))
6910 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6911 event, queue, pf_num, vf_num);
6912 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6913 mdd_detected = true;
6915 reg = rd32(hw, I40E_GL_MDET_RX);
6916 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6917 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6918 I40E_GL_MDET_RX_FUNCTION_SHIFT;
6919 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6920 I40E_GL_MDET_RX_EVENT_SHIFT;
6921 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6922 I40E_GL_MDET_RX_QUEUE_SHIFT) -
6923 pf->hw.func_caps.base_queue;
6924 if (netif_msg_rx_err(pf))
6925 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6926 event, queue, func);
6927 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6928 mdd_detected = true;
6931 if (mdd_detected) {
6932 reg = rd32(hw, I40E_PF_MDET_TX);
6933 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
6934 wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
6935 dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
6936 pf_mdd_detected = true;
6938 reg = rd32(hw, I40E_PF_MDET_RX);
6939 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
6940 wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
6941 dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
6942 pf_mdd_detected = true;
6944 /* Queue belongs to the PF, initiate a reset */
6945 if (pf_mdd_detected) {
6946 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6947 i40e_service_event_schedule(pf);
6951 /* see if one of the VFs needs its hand slapped */
6952 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
6953 vf = &(pf->vf[i]);
6954 reg = rd32(hw, I40E_VP_MDET_TX(i));
6955 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
6956 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
6957 vf->num_mdd_events++;
6958 dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
6962 reg = rd32(hw, I40E_VP_MDET_RX(i));
6963 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
6964 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
6965 vf->num_mdd_events++;
6966 dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
6970 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
6971 dev_info(&pf->pdev->dev,
6972 "Too many MDD events on VF %d, disabled\n", i);
6973 dev_info(&pf->pdev->dev,
6974 "Use PF Control I/F to re-enable the VF\n");
6975 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
6979 /* re-enable mdd interrupt cause */
6980 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
6981 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
6982 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
6983 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
6984 i40e_flush(hw);
6987 #ifdef CONFIG_I40E_VXLAN
6989 * i40e_sync_vxlan_filters_subtask - Sync the VSI filter list with HW
6990 * @pf: board private structure
6992 static void i40e_sync_vxlan_filters_subtask(struct i40e_pf *pf)
6994 struct i40e_hw *hw = &pf->hw;
6995 i40e_status ret;
6996 __be16 port;
6997 int i;
6999 if (!(pf->flags & I40E_FLAG_VXLAN_FILTER_SYNC))
7000 return;
7002 pf->flags &= ~I40E_FLAG_VXLAN_FILTER_SYNC;
7004 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7005 if (pf->pending_vxlan_bitmap & BIT_ULL(i)) {
7006 pf->pending_vxlan_bitmap &= ~BIT_ULL(i);
7007 port = pf->vxlan_ports[i];
7008 if (port)
7009 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7010 I40E_AQC_TUNNEL_TYPE_VXLAN,
7011 NULL, NULL);
7012 else
7013 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7015 if (ret) {
7016 dev_info(&pf->pdev->dev,
7017 "%s vxlan port %d, index %d failed, err %s aq_err %s\n",
7018 port ? "add" : "delete",
7019 ntohs(port), i,
7020 i40e_stat_str(&pf->hw, ret),
7021 i40e_aq_str(&pf->hw,
7022 pf->hw.aq.asq_last_status));
7023 pf->vxlan_ports[i] = 0;
7029 #endif
7031 * i40e_service_task - Run the driver's async subtasks
7032 * @work: pointer to work_struct containing our data
7034 static void i40e_service_task(struct work_struct *work)
7036 struct i40e_pf *pf = container_of(work,
7037 struct i40e_pf,
7038 service_task);
7039 unsigned long start_time = jiffies;
7041 /* don't bother with service tasks if a reset is in progress */
7042 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7043 i40e_service_event_complete(pf);
7044 return;
7047 i40e_detect_recover_hung(pf);
7048 i40e_reset_subtask(pf);
7049 i40e_handle_mdd_event(pf);
7050 i40e_vc_process_vflr_event(pf);
7051 i40e_watchdog_subtask(pf);
7052 i40e_fdir_reinit_subtask(pf);
7053 i40e_sync_filters_subtask(pf);
7054 #ifdef CONFIG_I40E_VXLAN
7055 i40e_sync_vxlan_filters_subtask(pf);
7056 #endif
7057 i40e_clean_adminq_subtask(pf);
7059 i40e_service_event_complete(pf);
7061 /* If the tasks have taken longer than one timer cycle or there
7062 * is more work to be done, reschedule the service task now
7063 * rather than wait for the timer to tick again.
7065 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7066 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
7067 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
7068 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7069 i40e_service_event_schedule(pf);
7073 * i40e_service_timer - timer callback
7074 * @data: pointer to PF struct
7076 static void i40e_service_timer(unsigned long data)
7078 struct i40e_pf *pf = (struct i40e_pf *)data;
7080 mod_timer(&pf->service_timer,
7081 round_jiffies(jiffies + pf->service_timer_period));
7082 i40e_service_event_schedule(pf);
7086 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7087 * @vsi: the VSI being configured
7089 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7091 struct i40e_pf *pf = vsi->back;
7093 switch (vsi->type) {
7094 case I40E_VSI_MAIN:
7095 vsi->alloc_queue_pairs = pf->num_lan_qps;
7096 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7097 I40E_REQ_DESCRIPTOR_MULTIPLE);
7098 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7099 vsi->num_q_vectors = pf->num_lan_msix;
7100 else
7101 vsi->num_q_vectors = 1;
7103 break;
7105 case I40E_VSI_FDIR:
7106 vsi->alloc_queue_pairs = 1;
7107 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7108 I40E_REQ_DESCRIPTOR_MULTIPLE);
7109 vsi->num_q_vectors = 1;
7110 break;
7112 case I40E_VSI_VMDQ2:
7113 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7114 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7115 I40E_REQ_DESCRIPTOR_MULTIPLE);
7116 vsi->num_q_vectors = pf->num_vmdq_msix;
7117 break;
7119 case I40E_VSI_SRIOV:
7120 vsi->alloc_queue_pairs = pf->num_vf_qps;
7121 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7122 I40E_REQ_DESCRIPTOR_MULTIPLE);
7123 break;
7125 #ifdef I40E_FCOE
7126 case I40E_VSI_FCOE:
7127 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7128 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7129 I40E_REQ_DESCRIPTOR_MULTIPLE);
7130 vsi->num_q_vectors = pf->num_fcoe_msix;
7131 break;
7133 #endif /* I40E_FCOE */
7134 default:
7135 WARN_ON(1);
7136 return -ENODATA;
7139 return 0;
7143 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7144 * @type: VSI pointer
7145 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7147 * On error: returns error code (negative)
7148 * On success: returns 0
7150 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7152 int size;
7153 int ret = 0;
7155 /* allocate memory for both Tx and Rx ring pointers */
7156 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7157 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7158 if (!vsi->tx_rings)
7159 return -ENOMEM;
7160 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7162 if (alloc_qvectors) {
7163 /* allocate memory for q_vector pointers */
7164 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7165 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7166 if (!vsi->q_vectors) {
7167 ret = -ENOMEM;
7168 goto err_vectors;
7171 return ret;
7173 err_vectors:
7174 kfree(vsi->tx_rings);
7175 return ret;
7179 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7180 * @pf: board private structure
7181 * @type: type of VSI
7183 * On error: returns error code (negative)
7184 * On success: returns vsi index in PF (positive)
7186 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7188 int ret = -ENODEV;
7189 struct i40e_vsi *vsi;
7190 int vsi_idx;
7191 int i;
7193 /* Need to protect the allocation of the VSIs at the PF level */
7194 mutex_lock(&pf->switch_mutex);
7196 /* VSI list may be fragmented if VSI creation/destruction has
7197 * been happening. We can afford to do a quick scan to look
7198 * for any free VSIs in the list.
7200 * find next empty vsi slot, looping back around if necessary
7202 i = pf->next_vsi;
7203 while (i < pf->num_alloc_vsi && pf->vsi[i])
7204 i++;
7205 if (i >= pf->num_alloc_vsi) {
7206 i = 0;
7207 while (i < pf->next_vsi && pf->vsi[i])
7208 i++;
7211 if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7212 vsi_idx = i; /* Found one! */
7213 } else {
7214 ret = -ENODEV;
7215 goto unlock_pf; /* out of VSI slots! */
7217 pf->next_vsi = ++i;
7219 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7220 if (!vsi) {
7221 ret = -ENOMEM;
7222 goto unlock_pf;
7224 vsi->type = type;
7225 vsi->back = pf;
7226 set_bit(__I40E_DOWN, &vsi->state);
7227 vsi->flags = 0;
7228 vsi->idx = vsi_idx;
7229 vsi->rx_itr_setting = pf->rx_itr_default;
7230 vsi->tx_itr_setting = pf->tx_itr_default;
7231 vsi->int_rate_limit = 0;
7232 vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7233 pf->rss_table_size : 64;
7234 vsi->netdev_registered = false;
7235 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7236 INIT_LIST_HEAD(&vsi->mac_filter_list);
7237 vsi->irqs_ready = false;
7239 ret = i40e_set_num_rings_in_vsi(vsi);
7240 if (ret)
7241 goto err_rings;
7243 ret = i40e_vsi_alloc_arrays(vsi, true);
7244 if (ret)
7245 goto err_rings;
7247 /* Setup default MSIX irq handler for VSI */
7248 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7250 /* Initialize VSI lock */
7251 spin_lock_init(&vsi->mac_filter_list_lock);
7252 pf->vsi[vsi_idx] = vsi;
7253 ret = vsi_idx;
7254 goto unlock_pf;
7256 err_rings:
7257 pf->next_vsi = i - 1;
7258 kfree(vsi);
7259 unlock_pf:
7260 mutex_unlock(&pf->switch_mutex);
7261 return ret;
7265 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7266 * @type: VSI pointer
7267 * @free_qvectors: a bool to specify if q_vectors need to be freed.
7269 * On error: returns error code (negative)
7270 * On success: returns 0
7272 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7274 /* free the ring and vector containers */
7275 if (free_qvectors) {
7276 kfree(vsi->q_vectors);
7277 vsi->q_vectors = NULL;
7279 kfree(vsi->tx_rings);
7280 vsi->tx_rings = NULL;
7281 vsi->rx_rings = NULL;
7285 * i40e_vsi_clear - Deallocate the VSI provided
7286 * @vsi: the VSI being un-configured
7288 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7290 struct i40e_pf *pf;
7292 if (!vsi)
7293 return 0;
7295 if (!vsi->back)
7296 goto free_vsi;
7297 pf = vsi->back;
7299 mutex_lock(&pf->switch_mutex);
7300 if (!pf->vsi[vsi->idx]) {
7301 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7302 vsi->idx, vsi->idx, vsi, vsi->type);
7303 goto unlock_vsi;
7306 if (pf->vsi[vsi->idx] != vsi) {
7307 dev_err(&pf->pdev->dev,
7308 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7309 pf->vsi[vsi->idx]->idx,
7310 pf->vsi[vsi->idx],
7311 pf->vsi[vsi->idx]->type,
7312 vsi->idx, vsi, vsi->type);
7313 goto unlock_vsi;
7316 /* updates the PF for this cleared vsi */
7317 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7318 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7320 i40e_vsi_free_arrays(vsi, true);
7322 pf->vsi[vsi->idx] = NULL;
7323 if (vsi->idx < pf->next_vsi)
7324 pf->next_vsi = vsi->idx;
7326 unlock_vsi:
7327 mutex_unlock(&pf->switch_mutex);
7328 free_vsi:
7329 kfree(vsi);
7331 return 0;
7335 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7336 * @vsi: the VSI being cleaned
7338 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7340 int i;
7342 if (vsi->tx_rings && vsi->tx_rings[0]) {
7343 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7344 kfree_rcu(vsi->tx_rings[i], rcu);
7345 vsi->tx_rings[i] = NULL;
7346 vsi->rx_rings[i] = NULL;
7352 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7353 * @vsi: the VSI being configured
7355 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7357 struct i40e_ring *tx_ring, *rx_ring;
7358 struct i40e_pf *pf = vsi->back;
7359 int i;
7361 /* Set basic values in the rings to be used later during open() */
7362 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7363 /* allocate space for both Tx and Rx in one shot */
7364 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7365 if (!tx_ring)
7366 goto err_out;
7368 tx_ring->queue_index = i;
7369 tx_ring->reg_idx = vsi->base_queue + i;
7370 tx_ring->ring_active = false;
7371 tx_ring->vsi = vsi;
7372 tx_ring->netdev = vsi->netdev;
7373 tx_ring->dev = &pf->pdev->dev;
7374 tx_ring->count = vsi->num_desc;
7375 tx_ring->size = 0;
7376 tx_ring->dcb_tc = 0;
7377 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7378 tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7379 if (vsi->back->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE)
7380 tx_ring->flags |= I40E_TXR_FLAGS_OUTER_UDP_CSUM;
7381 vsi->tx_rings[i] = tx_ring;
7383 rx_ring = &tx_ring[1];
7384 rx_ring->queue_index = i;
7385 rx_ring->reg_idx = vsi->base_queue + i;
7386 rx_ring->ring_active = false;
7387 rx_ring->vsi = vsi;
7388 rx_ring->netdev = vsi->netdev;
7389 rx_ring->dev = &pf->pdev->dev;
7390 rx_ring->count = vsi->num_desc;
7391 rx_ring->size = 0;
7392 rx_ring->dcb_tc = 0;
7393 if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7394 set_ring_16byte_desc_enabled(rx_ring);
7395 else
7396 clear_ring_16byte_desc_enabled(rx_ring);
7397 vsi->rx_rings[i] = rx_ring;
7400 return 0;
7402 err_out:
7403 i40e_vsi_clear_rings(vsi);
7404 return -ENOMEM;
7408 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7409 * @pf: board private structure
7410 * @vectors: the number of MSI-X vectors to request
7412 * Returns the number of vectors reserved, or error
7414 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7416 vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7417 I40E_MIN_MSIX, vectors);
7418 if (vectors < 0) {
7419 dev_info(&pf->pdev->dev,
7420 "MSI-X vector reservation failed: %d\n", vectors);
7421 vectors = 0;
7424 return vectors;
7428 * i40e_init_msix - Setup the MSIX capability
7429 * @pf: board private structure
7431 * Work with the OS to set up the MSIX vectors needed.
7433 * Returns the number of vectors reserved or negative on failure
7435 static int i40e_init_msix(struct i40e_pf *pf)
7437 struct i40e_hw *hw = &pf->hw;
7438 int vectors_left;
7439 int v_budget, i;
7440 int v_actual;
7442 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7443 return -ENODEV;
7445 /* The number of vectors we'll request will be comprised of:
7446 * - Add 1 for "other" cause for Admin Queue events, etc.
7447 * - The number of LAN queue pairs
7448 * - Queues being used for RSS.
7449 * We don't need as many as max_rss_size vectors.
7450 * use rss_size instead in the calculation since that
7451 * is governed by number of cpus in the system.
7452 * - assumes symmetric Tx/Rx pairing
7453 * - The number of VMDq pairs
7454 #ifdef I40E_FCOE
7455 * - The number of FCOE qps.
7456 #endif
7457 * Once we count this up, try the request.
7459 * If we can't get what we want, we'll simplify to nearly nothing
7460 * and try again. If that still fails, we punt.
7462 vectors_left = hw->func_caps.num_msix_vectors;
7463 v_budget = 0;
7465 /* reserve one vector for miscellaneous handler */
7466 if (vectors_left) {
7467 v_budget++;
7468 vectors_left--;
7471 /* reserve vectors for the main PF traffic queues */
7472 pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7473 vectors_left -= pf->num_lan_msix;
7474 v_budget += pf->num_lan_msix;
7476 /* reserve one vector for sideband flow director */
7477 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7478 if (vectors_left) {
7479 v_budget++;
7480 vectors_left--;
7481 } else {
7482 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7486 #ifdef I40E_FCOE
7487 /* can we reserve enough for FCoE? */
7488 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7489 if (!vectors_left)
7490 pf->num_fcoe_msix = 0;
7491 else if (vectors_left >= pf->num_fcoe_qps)
7492 pf->num_fcoe_msix = pf->num_fcoe_qps;
7493 else
7494 pf->num_fcoe_msix = 1;
7495 v_budget += pf->num_fcoe_msix;
7496 vectors_left -= pf->num_fcoe_msix;
7499 #endif
7500 /* any vectors left over go for VMDq support */
7501 if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7502 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7503 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7505 /* if we're short on vectors for what's desired, we limit
7506 * the queues per vmdq. If this is still more than are
7507 * available, the user will need to change the number of
7508 * queues/vectors used by the PF later with the ethtool
7509 * channels command
7511 if (vmdq_vecs < vmdq_vecs_wanted)
7512 pf->num_vmdq_qps = 1;
7513 pf->num_vmdq_msix = pf->num_vmdq_qps;
7515 v_budget += vmdq_vecs;
7516 vectors_left -= vmdq_vecs;
7519 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7520 GFP_KERNEL);
7521 if (!pf->msix_entries)
7522 return -ENOMEM;
7524 for (i = 0; i < v_budget; i++)
7525 pf->msix_entries[i].entry = i;
7526 v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7528 if (v_actual != v_budget) {
7529 /* If we have limited resources, we will start with no vectors
7530 * for the special features and then allocate vectors to some
7531 * of these features based on the policy and at the end disable
7532 * the features that did not get any vectors.
7534 #ifdef I40E_FCOE
7535 pf->num_fcoe_qps = 0;
7536 pf->num_fcoe_msix = 0;
7537 #endif
7538 pf->num_vmdq_msix = 0;
7541 if (v_actual < I40E_MIN_MSIX) {
7542 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7543 kfree(pf->msix_entries);
7544 pf->msix_entries = NULL;
7545 return -ENODEV;
7547 } else if (v_actual == I40E_MIN_MSIX) {
7548 /* Adjust for minimal MSIX use */
7549 pf->num_vmdq_vsis = 0;
7550 pf->num_vmdq_qps = 0;
7551 pf->num_lan_qps = 1;
7552 pf->num_lan_msix = 1;
7554 } else if (v_actual != v_budget) {
7555 int vec;
7557 /* reserve the misc vector */
7558 vec = v_actual - 1;
7560 /* Scale vector usage down */
7561 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
7562 pf->num_vmdq_vsis = 1;
7563 pf->num_vmdq_qps = 1;
7564 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7566 /* partition out the remaining vectors */
7567 switch (vec) {
7568 case 2:
7569 pf->num_lan_msix = 1;
7570 break;
7571 case 3:
7572 #ifdef I40E_FCOE
7573 /* give one vector to FCoE */
7574 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7575 pf->num_lan_msix = 1;
7576 pf->num_fcoe_msix = 1;
7578 #else
7579 pf->num_lan_msix = 2;
7580 #endif
7581 break;
7582 default:
7583 #ifdef I40E_FCOE
7584 /* give one vector to FCoE */
7585 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7586 pf->num_fcoe_msix = 1;
7587 vec--;
7589 #endif
7590 /* give the rest to the PF */
7591 pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
7592 break;
7596 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7597 (pf->num_vmdq_msix == 0)) {
7598 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7599 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7601 #ifdef I40E_FCOE
7603 if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7604 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7605 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7607 #endif
7608 return v_actual;
7612 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7613 * @vsi: the VSI being configured
7614 * @v_idx: index of the vector in the vsi struct
7616 * We allocate one q_vector. If allocation fails we return -ENOMEM.
7618 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7620 struct i40e_q_vector *q_vector;
7622 /* allocate q_vector */
7623 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7624 if (!q_vector)
7625 return -ENOMEM;
7627 q_vector->vsi = vsi;
7628 q_vector->v_idx = v_idx;
7629 cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7630 if (vsi->netdev)
7631 netif_napi_add(vsi->netdev, &q_vector->napi,
7632 i40e_napi_poll, NAPI_POLL_WEIGHT);
7634 q_vector->rx.latency_range = I40E_LOW_LATENCY;
7635 q_vector->tx.latency_range = I40E_LOW_LATENCY;
7637 /* tie q_vector and vsi together */
7638 vsi->q_vectors[v_idx] = q_vector;
7640 return 0;
7644 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7645 * @vsi: the VSI being configured
7647 * We allocate one q_vector per queue interrupt. If allocation fails we
7648 * return -ENOMEM.
7650 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7652 struct i40e_pf *pf = vsi->back;
7653 int v_idx, num_q_vectors;
7654 int err;
7656 /* if not MSIX, give the one vector only to the LAN VSI */
7657 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7658 num_q_vectors = vsi->num_q_vectors;
7659 else if (vsi == pf->vsi[pf->lan_vsi])
7660 num_q_vectors = 1;
7661 else
7662 return -EINVAL;
7664 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7665 err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7666 if (err)
7667 goto err_out;
7670 return 0;
7672 err_out:
7673 while (v_idx--)
7674 i40e_free_q_vector(vsi, v_idx);
7676 return err;
7680 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7681 * @pf: board private structure to initialize
7683 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7685 int vectors = 0;
7686 ssize_t size;
7688 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7689 vectors = i40e_init_msix(pf);
7690 if (vectors < 0) {
7691 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
7692 #ifdef I40E_FCOE
7693 I40E_FLAG_FCOE_ENABLED |
7694 #endif
7695 I40E_FLAG_RSS_ENABLED |
7696 I40E_FLAG_DCB_CAPABLE |
7697 I40E_FLAG_SRIOV_ENABLED |
7698 I40E_FLAG_FD_SB_ENABLED |
7699 I40E_FLAG_FD_ATR_ENABLED |
7700 I40E_FLAG_VMDQ_ENABLED);
7702 /* rework the queue expectations without MSIX */
7703 i40e_determine_queue_usage(pf);
7707 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7708 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7709 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7710 vectors = pci_enable_msi(pf->pdev);
7711 if (vectors < 0) {
7712 dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7713 vectors);
7714 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7716 vectors = 1; /* one MSI or Legacy vector */
7719 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7720 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7722 /* set up vector assignment tracking */
7723 size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7724 pf->irq_pile = kzalloc(size, GFP_KERNEL);
7725 if (!pf->irq_pile) {
7726 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7727 return -ENOMEM;
7729 pf->irq_pile->num_entries = vectors;
7730 pf->irq_pile->search_hint = 0;
7732 /* track first vector for misc interrupts, ignore return */
7733 (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7735 return 0;
7739 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7740 * @pf: board private structure
7742 * This sets up the handler for MSIX 0, which is used to manage the
7743 * non-queue interrupts, e.g. AdminQ and errors. This is not used
7744 * when in MSI or Legacy interrupt mode.
7746 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7748 struct i40e_hw *hw = &pf->hw;
7749 int err = 0;
7751 /* Only request the irq if this is the first time through, and
7752 * not when we're rebuilding after a Reset
7754 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7755 err = request_irq(pf->msix_entries[0].vector,
7756 i40e_intr, 0, pf->int_name, pf);
7757 if (err) {
7758 dev_info(&pf->pdev->dev,
7759 "request_irq for %s failed: %d\n",
7760 pf->int_name, err);
7761 return -EFAULT;
7765 i40e_enable_misc_int_causes(pf);
7767 /* associate no queues to the misc vector */
7768 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7769 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7771 i40e_flush(hw);
7773 i40e_irq_dynamic_enable_icr0(pf);
7775 return err;
7779 * i40e_config_rss_aq - Prepare for RSS using AQ commands
7780 * @vsi: vsi structure
7781 * @seed: RSS hash seed
7783 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed)
7785 struct i40e_aqc_get_set_rss_key_data rss_key;
7786 struct i40e_pf *pf = vsi->back;
7787 struct i40e_hw *hw = &pf->hw;
7788 bool pf_lut = false;
7789 u8 *rss_lut;
7790 int ret, i;
7792 memset(&rss_key, 0, sizeof(rss_key));
7793 memcpy(&rss_key, seed, sizeof(rss_key));
7795 rss_lut = kzalloc(pf->rss_table_size, GFP_KERNEL);
7796 if (!rss_lut)
7797 return -ENOMEM;
7799 /* Populate the LUT with max no. of queues in round robin fashion */
7800 for (i = 0; i < vsi->rss_table_size; i++)
7801 rss_lut[i] = i % vsi->rss_size;
7803 ret = i40e_aq_set_rss_key(hw, vsi->id, &rss_key);
7804 if (ret) {
7805 dev_info(&pf->pdev->dev,
7806 "Cannot set RSS key, err %s aq_err %s\n",
7807 i40e_stat_str(&pf->hw, ret),
7808 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7809 goto config_rss_aq_out;
7812 if (vsi->type == I40E_VSI_MAIN)
7813 pf_lut = true;
7815 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, rss_lut,
7816 vsi->rss_table_size);
7817 if (ret)
7818 dev_info(&pf->pdev->dev,
7819 "Cannot set RSS lut, err %s aq_err %s\n",
7820 i40e_stat_str(&pf->hw, ret),
7821 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
7823 config_rss_aq_out:
7824 kfree(rss_lut);
7825 return ret;
7829 * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
7830 * @vsi: VSI structure
7832 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
7834 u8 seed[I40E_HKEY_ARRAY_SIZE];
7835 struct i40e_pf *pf = vsi->back;
7837 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7838 vsi->rss_size = min_t(int, pf->rss_size, vsi->num_queue_pairs);
7840 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
7841 return i40e_config_rss_aq(vsi, seed);
7843 return 0;
7847 * i40e_config_rss_reg - Prepare for RSS if used
7848 * @pf: board private structure
7849 * @seed: RSS hash seed
7851 static int i40e_config_rss_reg(struct i40e_pf *pf, const u8 *seed)
7853 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7854 struct i40e_hw *hw = &pf->hw;
7855 u32 *seed_dw = (u32 *)seed;
7856 u32 current_queue = 0;
7857 u32 lut = 0;
7858 int i, j;
7860 /* Fill out hash function seed */
7861 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7862 wr32(hw, I40E_PFQF_HKEY(i), seed_dw[i]);
7864 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++) {
7865 lut = 0;
7866 for (j = 0; j < 4; j++) {
7867 if (current_queue == vsi->rss_size)
7868 current_queue = 0;
7869 lut |= ((current_queue) << (8 * j));
7870 current_queue++;
7872 wr32(&pf->hw, I40E_PFQF_HLUT(i), lut);
7874 i40e_flush(hw);
7876 return 0;
7880 * i40e_config_rss - Prepare for RSS if used
7881 * @pf: board private structure
7883 static int i40e_config_rss(struct i40e_pf *pf)
7885 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7886 u8 seed[I40E_HKEY_ARRAY_SIZE];
7887 struct i40e_hw *hw = &pf->hw;
7888 u32 reg_val;
7889 u64 hena;
7891 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
7893 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
7894 hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
7895 ((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
7896 hena |= i40e_pf_get_default_rss_hena(pf);
7898 wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
7899 wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
7901 vsi->rss_size = min_t(int, pf->rss_size, vsi->num_queue_pairs);
7903 /* Determine the RSS table size based on the hardware capabilities */
7904 reg_val = rd32(hw, I40E_PFQF_CTL_0);
7905 reg_val = (pf->rss_table_size == 512) ?
7906 (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
7907 (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
7908 wr32(hw, I40E_PFQF_CTL_0, reg_val);
7910 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
7911 return i40e_config_rss_aq(pf->vsi[pf->lan_vsi], seed);
7912 else
7913 return i40e_config_rss_reg(pf, seed);
7917 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
7918 * @pf: board private structure
7919 * @queue_count: the requested queue count for rss.
7921 * returns 0 if rss is not enabled, if enabled returns the final rss queue
7922 * count which may be different from the requested queue count.
7924 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
7926 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7927 int new_rss_size;
7929 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
7930 return 0;
7932 new_rss_size = min_t(int, queue_count, pf->rss_size_max);
7934 if (queue_count != vsi->num_queue_pairs) {
7935 vsi->req_queue_pairs = queue_count;
7936 i40e_prep_for_reset(pf);
7938 pf->rss_size = new_rss_size;
7940 i40e_reset_and_rebuild(pf, true);
7941 i40e_config_rss(pf);
7943 dev_info(&pf->pdev->dev, "RSS count: %d\n", pf->rss_size);
7944 return pf->rss_size;
7948 * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
7949 * @pf: board private structure
7951 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
7953 i40e_status status;
7954 bool min_valid, max_valid;
7955 u32 max_bw, min_bw;
7957 status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
7958 &min_valid, &max_valid);
7960 if (!status) {
7961 if (min_valid)
7962 pf->npar_min_bw = min_bw;
7963 if (max_valid)
7964 pf->npar_max_bw = max_bw;
7967 return status;
7971 * i40e_set_npar_bw_setting - Set BW settings for this PF partition
7972 * @pf: board private structure
7974 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
7976 struct i40e_aqc_configure_partition_bw_data bw_data;
7977 i40e_status status;
7979 /* Set the valid bit for this PF */
7980 bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
7981 bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
7982 bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
7984 /* Set the new bandwidths */
7985 status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
7987 return status;
7991 * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
7992 * @pf: board private structure
7994 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
7996 /* Commit temporary BW setting to permanent NVM image */
7997 enum i40e_admin_queue_err last_aq_status;
7998 i40e_status ret;
7999 u16 nvm_word;
8001 if (pf->hw.partition_id != 1) {
8002 dev_info(&pf->pdev->dev,
8003 "Commit BW only works on partition 1! This is partition %d",
8004 pf->hw.partition_id);
8005 ret = I40E_NOT_SUPPORTED;
8006 goto bw_commit_out;
8009 /* Acquire NVM for read access */
8010 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8011 last_aq_status = pf->hw.aq.asq_last_status;
8012 if (ret) {
8013 dev_info(&pf->pdev->dev,
8014 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8015 i40e_stat_str(&pf->hw, ret),
8016 i40e_aq_str(&pf->hw, last_aq_status));
8017 goto bw_commit_out;
8020 /* Read word 0x10 of NVM - SW compatibility word 1 */
8021 ret = i40e_aq_read_nvm(&pf->hw,
8022 I40E_SR_NVM_CONTROL_WORD,
8023 0x10, sizeof(nvm_word), &nvm_word,
8024 false, NULL);
8025 /* Save off last admin queue command status before releasing
8026 * the NVM
8028 last_aq_status = pf->hw.aq.asq_last_status;
8029 i40e_release_nvm(&pf->hw);
8030 if (ret) {
8031 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8032 i40e_stat_str(&pf->hw, ret),
8033 i40e_aq_str(&pf->hw, last_aq_status));
8034 goto bw_commit_out;
8037 /* Wait a bit for NVM release to complete */
8038 msleep(50);
8040 /* Acquire NVM for write access */
8041 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8042 last_aq_status = pf->hw.aq.asq_last_status;
8043 if (ret) {
8044 dev_info(&pf->pdev->dev,
8045 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8046 i40e_stat_str(&pf->hw, ret),
8047 i40e_aq_str(&pf->hw, last_aq_status));
8048 goto bw_commit_out;
8050 /* Write it back out unchanged to initiate update NVM,
8051 * which will force a write of the shadow (alt) RAM to
8052 * the NVM - thus storing the bandwidth values permanently.
8054 ret = i40e_aq_update_nvm(&pf->hw,
8055 I40E_SR_NVM_CONTROL_WORD,
8056 0x10, sizeof(nvm_word),
8057 &nvm_word, true, NULL);
8058 /* Save off last admin queue command status before releasing
8059 * the NVM
8061 last_aq_status = pf->hw.aq.asq_last_status;
8062 i40e_release_nvm(&pf->hw);
8063 if (ret)
8064 dev_info(&pf->pdev->dev,
8065 "BW settings NOT SAVED, err %s aq_err %s\n",
8066 i40e_stat_str(&pf->hw, ret),
8067 i40e_aq_str(&pf->hw, last_aq_status));
8068 bw_commit_out:
8070 return ret;
8074 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8075 * @pf: board private structure to initialize
8077 * i40e_sw_init initializes the Adapter private data structure.
8078 * Fields are initialized based on PCI device information and
8079 * OS network device settings (MTU size).
8081 static int i40e_sw_init(struct i40e_pf *pf)
8083 int err = 0;
8084 int size;
8086 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8087 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8088 pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
8089 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8090 if (I40E_DEBUG_USER & debug)
8091 pf->hw.debug_mask = debug;
8092 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8093 I40E_DEFAULT_MSG_ENABLE);
8096 /* Set default capability flags */
8097 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8098 I40E_FLAG_MSI_ENABLED |
8099 I40E_FLAG_LINK_POLLING_ENABLED |
8100 I40E_FLAG_MSIX_ENABLED;
8102 if (iommu_present(&pci_bus_type))
8103 pf->flags |= I40E_FLAG_RX_PS_ENABLED;
8104 else
8105 pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
8107 /* Set default ITR */
8108 pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8109 pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8111 /* Depending on PF configurations, it is possible that the RSS
8112 * maximum might end up larger than the available queues
8114 pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8115 pf->rss_size = 1;
8116 pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8117 pf->rss_size_max = min_t(int, pf->rss_size_max,
8118 pf->hw.func_caps.num_tx_qp);
8119 if (pf->hw.func_caps.rss) {
8120 pf->flags |= I40E_FLAG_RSS_ENABLED;
8121 pf->rss_size = min_t(int, pf->rss_size_max, num_online_cpus());
8124 /* MFP mode enabled */
8125 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8126 pf->flags |= I40E_FLAG_MFP_ENABLED;
8127 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8128 if (i40e_get_npar_bw_setting(pf))
8129 dev_warn(&pf->pdev->dev,
8130 "Could not get NPAR bw settings\n");
8131 else
8132 dev_info(&pf->pdev->dev,
8133 "Min BW = %8.8x, Max BW = %8.8x\n",
8134 pf->npar_min_bw, pf->npar_max_bw);
8137 /* FW/NVM is not yet fixed in this regard */
8138 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8139 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8140 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8141 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8142 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8143 pf->hw.num_partitions > 1)
8144 dev_info(&pf->pdev->dev,
8145 "Flow Director Sideband mode Disabled in MFP mode\n");
8146 else
8147 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8148 pf->fdir_pf_filter_count =
8149 pf->hw.func_caps.fd_filters_guaranteed;
8150 pf->hw.fdir_shared_filter_count =
8151 pf->hw.func_caps.fd_filters_best_effort;
8154 if (pf->hw.func_caps.vmdq) {
8155 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8156 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8159 #ifdef I40E_FCOE
8160 i40e_init_pf_fcoe(pf);
8162 #endif /* I40E_FCOE */
8163 #ifdef CONFIG_PCI_IOV
8164 if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8165 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8166 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8167 pf->num_req_vfs = min_t(int,
8168 pf->hw.func_caps.num_vfs,
8169 I40E_MAX_VF_COUNT);
8171 #endif /* CONFIG_PCI_IOV */
8172 if (pf->hw.mac.type == I40E_MAC_X722) {
8173 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8174 I40E_FLAG_128_QP_RSS_CAPABLE |
8175 I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8176 I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8177 I40E_FLAG_WB_ON_ITR_CAPABLE |
8178 I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE;
8180 pf->eeprom_version = 0xDEAD;
8181 pf->lan_veb = I40E_NO_VEB;
8182 pf->lan_vsi = I40E_NO_VSI;
8184 /* By default FW has this off for performance reasons */
8185 pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8187 /* set up queue assignment tracking */
8188 size = sizeof(struct i40e_lump_tracking)
8189 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8190 pf->qp_pile = kzalloc(size, GFP_KERNEL);
8191 if (!pf->qp_pile) {
8192 err = -ENOMEM;
8193 goto sw_init_done;
8195 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8196 pf->qp_pile->search_hint = 0;
8198 pf->tx_timeout_recovery_level = 1;
8200 mutex_init(&pf->switch_mutex);
8202 /* If NPAR is enabled nudge the Tx scheduler */
8203 if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8204 i40e_set_npar_bw_setting(pf);
8206 sw_init_done:
8207 return err;
8211 * i40e_set_ntuple - set the ntuple feature flag and take action
8212 * @pf: board private structure to initialize
8213 * @features: the feature set that the stack is suggesting
8215 * returns a bool to indicate if reset needs to happen
8217 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8219 bool need_reset = false;
8221 /* Check if Flow Director n-tuple support was enabled or disabled. If
8222 * the state changed, we need to reset.
8224 if (features & NETIF_F_NTUPLE) {
8225 /* Enable filters and mark for reset */
8226 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8227 need_reset = true;
8228 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8229 } else {
8230 /* turn off filters, mark for reset and clear SW filter list */
8231 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8232 need_reset = true;
8233 i40e_fdir_filter_exit(pf);
8235 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8236 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8237 /* reset fd counters */
8238 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8239 pf->fdir_pf_active_filters = 0;
8240 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8241 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8242 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8243 /* if ATR was auto disabled it can be re-enabled. */
8244 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8245 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8246 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8248 return need_reset;
8252 * i40e_set_features - set the netdev feature flags
8253 * @netdev: ptr to the netdev being adjusted
8254 * @features: the feature set that the stack is suggesting
8256 static int i40e_set_features(struct net_device *netdev,
8257 netdev_features_t features)
8259 struct i40e_netdev_priv *np = netdev_priv(netdev);
8260 struct i40e_vsi *vsi = np->vsi;
8261 struct i40e_pf *pf = vsi->back;
8262 bool need_reset;
8264 if (features & NETIF_F_HW_VLAN_CTAG_RX)
8265 i40e_vlan_stripping_enable(vsi);
8266 else
8267 i40e_vlan_stripping_disable(vsi);
8269 need_reset = i40e_set_ntuple(pf, features);
8271 if (need_reset)
8272 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8274 return 0;
8277 #ifdef CONFIG_I40E_VXLAN
8279 * i40e_get_vxlan_port_idx - Lookup a possibly offloaded for Rx UDP port
8280 * @pf: board private structure
8281 * @port: The UDP port to look up
8283 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8285 static u8 i40e_get_vxlan_port_idx(struct i40e_pf *pf, __be16 port)
8287 u8 i;
8289 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8290 if (pf->vxlan_ports[i] == port)
8291 return i;
8294 return i;
8298 * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
8299 * @netdev: This physical port's netdev
8300 * @sa_family: Socket Family that VXLAN is notifying us about
8301 * @port: New UDP port number that VXLAN started listening to
8303 static void i40e_add_vxlan_port(struct net_device *netdev,
8304 sa_family_t sa_family, __be16 port)
8306 struct i40e_netdev_priv *np = netdev_priv(netdev);
8307 struct i40e_vsi *vsi = np->vsi;
8308 struct i40e_pf *pf = vsi->back;
8309 u8 next_idx;
8310 u8 idx;
8312 if (sa_family == AF_INET6)
8313 return;
8315 idx = i40e_get_vxlan_port_idx(pf, port);
8317 /* Check if port already exists */
8318 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8319 netdev_info(netdev, "vxlan port %d already offloaded\n",
8320 ntohs(port));
8321 return;
8324 /* Now check if there is space to add the new port */
8325 next_idx = i40e_get_vxlan_port_idx(pf, 0);
8327 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8328 netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
8329 ntohs(port));
8330 return;
8333 /* New port: add it and mark its index in the bitmap */
8334 pf->vxlan_ports[next_idx] = port;
8335 pf->pending_vxlan_bitmap |= BIT_ULL(next_idx);
8336 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
8340 * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
8341 * @netdev: This physical port's netdev
8342 * @sa_family: Socket Family that VXLAN is notifying us about
8343 * @port: UDP port number that VXLAN stopped listening to
8345 static void i40e_del_vxlan_port(struct net_device *netdev,
8346 sa_family_t sa_family, __be16 port)
8348 struct i40e_netdev_priv *np = netdev_priv(netdev);
8349 struct i40e_vsi *vsi = np->vsi;
8350 struct i40e_pf *pf = vsi->back;
8351 u8 idx;
8353 if (sa_family == AF_INET6)
8354 return;
8356 idx = i40e_get_vxlan_port_idx(pf, port);
8358 /* Check if port already exists */
8359 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8360 /* if port exists, set it to 0 (mark for deletion)
8361 * and make it pending
8363 pf->vxlan_ports[idx] = 0;
8364 pf->pending_vxlan_bitmap |= BIT_ULL(idx);
8365 pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
8366 } else {
8367 netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
8368 ntohs(port));
8372 #endif
8373 static int i40e_get_phys_port_id(struct net_device *netdev,
8374 struct netdev_phys_item_id *ppid)
8376 struct i40e_netdev_priv *np = netdev_priv(netdev);
8377 struct i40e_pf *pf = np->vsi->back;
8378 struct i40e_hw *hw = &pf->hw;
8380 if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8381 return -EOPNOTSUPP;
8383 ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8384 memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8386 return 0;
8390 * i40e_ndo_fdb_add - add an entry to the hardware database
8391 * @ndm: the input from the stack
8392 * @tb: pointer to array of nladdr (unused)
8393 * @dev: the net device pointer
8394 * @addr: the MAC address entry being added
8395 * @flags: instructions from stack about fdb operation
8397 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8398 struct net_device *dev,
8399 const unsigned char *addr, u16 vid,
8400 u16 flags)
8402 struct i40e_netdev_priv *np = netdev_priv(dev);
8403 struct i40e_pf *pf = np->vsi->back;
8404 int err = 0;
8406 if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8407 return -EOPNOTSUPP;
8409 if (vid) {
8410 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8411 return -EINVAL;
8414 /* Hardware does not support aging addresses so if a
8415 * ndm_state is given only allow permanent addresses
8417 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8418 netdev_info(dev, "FDB only supports static addresses\n");
8419 return -EINVAL;
8422 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8423 err = dev_uc_add_excl(dev, addr);
8424 else if (is_multicast_ether_addr(addr))
8425 err = dev_mc_add_excl(dev, addr);
8426 else
8427 err = -EINVAL;
8429 /* Only return duplicate errors if NLM_F_EXCL is set */
8430 if (err == -EEXIST && !(flags & NLM_F_EXCL))
8431 err = 0;
8433 return err;
8437 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8438 * @dev: the netdev being configured
8439 * @nlh: RTNL message
8441 * Inserts a new hardware bridge if not already created and
8442 * enables the bridging mode requested (VEB or VEPA). If the
8443 * hardware bridge has already been inserted and the request
8444 * is to change the mode then that requires a PF reset to
8445 * allow rebuild of the components with required hardware
8446 * bridge mode enabled.
8448 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8449 struct nlmsghdr *nlh,
8450 u16 flags)
8452 struct i40e_netdev_priv *np = netdev_priv(dev);
8453 struct i40e_vsi *vsi = np->vsi;
8454 struct i40e_pf *pf = vsi->back;
8455 struct i40e_veb *veb = NULL;
8456 struct nlattr *attr, *br_spec;
8457 int i, rem;
8459 /* Only for PF VSI for now */
8460 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8461 return -EOPNOTSUPP;
8463 /* Find the HW bridge for PF VSI */
8464 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8465 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8466 veb = pf->veb[i];
8469 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8471 nla_for_each_nested(attr, br_spec, rem) {
8472 __u16 mode;
8474 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8475 continue;
8477 mode = nla_get_u16(attr);
8478 if ((mode != BRIDGE_MODE_VEPA) &&
8479 (mode != BRIDGE_MODE_VEB))
8480 return -EINVAL;
8482 /* Insert a new HW bridge */
8483 if (!veb) {
8484 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8485 vsi->tc_config.enabled_tc);
8486 if (veb) {
8487 veb->bridge_mode = mode;
8488 i40e_config_bridge_mode(veb);
8489 } else {
8490 /* No Bridge HW offload available */
8491 return -ENOENT;
8493 break;
8494 } else if (mode != veb->bridge_mode) {
8495 /* Existing HW bridge but different mode needs reset */
8496 veb->bridge_mode = mode;
8497 /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8498 if (mode == BRIDGE_MODE_VEB)
8499 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8500 else
8501 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8502 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8503 break;
8507 return 0;
8511 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8512 * @skb: skb buff
8513 * @pid: process id
8514 * @seq: RTNL message seq #
8515 * @dev: the netdev being configured
8516 * @filter_mask: unused
8517 * @nlflags: netlink flags passed in
8519 * Return the mode in which the hardware bridge is operating in
8520 * i.e VEB or VEPA.
8522 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8523 struct net_device *dev,
8524 u32 __always_unused filter_mask,
8525 int nlflags)
8527 struct i40e_netdev_priv *np = netdev_priv(dev);
8528 struct i40e_vsi *vsi = np->vsi;
8529 struct i40e_pf *pf = vsi->back;
8530 struct i40e_veb *veb = NULL;
8531 int i;
8533 /* Only for PF VSI for now */
8534 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8535 return -EOPNOTSUPP;
8537 /* Find the HW bridge for the PF VSI */
8538 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8539 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8540 veb = pf->veb[i];
8543 if (!veb)
8544 return 0;
8546 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8547 nlflags, 0, 0, filter_mask, NULL);
8550 #define I40E_MAX_TUNNEL_HDR_LEN 80
8552 * i40e_features_check - Validate encapsulated packet conforms to limits
8553 * @skb: skb buff
8554 * @dev: This physical port's netdev
8555 * @features: Offload features that the stack believes apply
8557 static netdev_features_t i40e_features_check(struct sk_buff *skb,
8558 struct net_device *dev,
8559 netdev_features_t features)
8561 if (skb->encapsulation &&
8562 (skb_inner_mac_header(skb) - skb_transport_header(skb) >
8563 I40E_MAX_TUNNEL_HDR_LEN))
8564 return features & ~(NETIF_F_ALL_CSUM | NETIF_F_GSO_MASK);
8566 return features;
8569 static const struct net_device_ops i40e_netdev_ops = {
8570 .ndo_open = i40e_open,
8571 .ndo_stop = i40e_close,
8572 .ndo_start_xmit = i40e_lan_xmit_frame,
8573 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
8574 .ndo_set_rx_mode = i40e_set_rx_mode,
8575 .ndo_validate_addr = eth_validate_addr,
8576 .ndo_set_mac_address = i40e_set_mac,
8577 .ndo_change_mtu = i40e_change_mtu,
8578 .ndo_do_ioctl = i40e_ioctl,
8579 .ndo_tx_timeout = i40e_tx_timeout,
8580 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
8581 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
8582 #ifdef CONFIG_NET_POLL_CONTROLLER
8583 .ndo_poll_controller = i40e_netpoll,
8584 #endif
8585 .ndo_setup_tc = i40e_setup_tc,
8586 #ifdef I40E_FCOE
8587 .ndo_fcoe_enable = i40e_fcoe_enable,
8588 .ndo_fcoe_disable = i40e_fcoe_disable,
8589 #endif
8590 .ndo_set_features = i40e_set_features,
8591 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
8592 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
8593 .ndo_set_vf_rate = i40e_ndo_set_vf_bw,
8594 .ndo_get_vf_config = i40e_ndo_get_vf_config,
8595 .ndo_set_vf_link_state = i40e_ndo_set_vf_link_state,
8596 .ndo_set_vf_spoofchk = i40e_ndo_set_vf_spoofchk,
8597 #ifdef CONFIG_I40E_VXLAN
8598 .ndo_add_vxlan_port = i40e_add_vxlan_port,
8599 .ndo_del_vxlan_port = i40e_del_vxlan_port,
8600 #endif
8601 .ndo_get_phys_port_id = i40e_get_phys_port_id,
8602 .ndo_fdb_add = i40e_ndo_fdb_add,
8603 .ndo_features_check = i40e_features_check,
8604 .ndo_bridge_getlink = i40e_ndo_bridge_getlink,
8605 .ndo_bridge_setlink = i40e_ndo_bridge_setlink,
8609 * i40e_config_netdev - Setup the netdev flags
8610 * @vsi: the VSI being configured
8612 * Returns 0 on success, negative value on failure
8614 static int i40e_config_netdev(struct i40e_vsi *vsi)
8616 u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8617 struct i40e_pf *pf = vsi->back;
8618 struct i40e_hw *hw = &pf->hw;
8619 struct i40e_netdev_priv *np;
8620 struct net_device *netdev;
8621 u8 mac_addr[ETH_ALEN];
8622 int etherdev_size;
8624 etherdev_size = sizeof(struct i40e_netdev_priv);
8625 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
8626 if (!netdev)
8627 return -ENOMEM;
8629 vsi->netdev = netdev;
8630 np = netdev_priv(netdev);
8631 np->vsi = vsi;
8633 netdev->hw_enc_features |= NETIF_F_IP_CSUM |
8634 NETIF_F_GSO_UDP_TUNNEL |
8635 NETIF_F_GSO_GRE |
8636 NETIF_F_TSO;
8638 netdev->features = NETIF_F_SG |
8639 NETIF_F_IP_CSUM |
8640 NETIF_F_SCTP_CSUM |
8641 NETIF_F_HIGHDMA |
8642 NETIF_F_GSO_UDP_TUNNEL |
8643 NETIF_F_GSO_GRE |
8644 NETIF_F_HW_VLAN_CTAG_TX |
8645 NETIF_F_HW_VLAN_CTAG_RX |
8646 NETIF_F_HW_VLAN_CTAG_FILTER |
8647 NETIF_F_IPV6_CSUM |
8648 NETIF_F_TSO |
8649 NETIF_F_TSO_ECN |
8650 NETIF_F_TSO6 |
8651 NETIF_F_RXCSUM |
8652 NETIF_F_RXHASH |
8655 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
8656 netdev->features |= NETIF_F_NTUPLE;
8658 /* copy netdev features into list of user selectable features */
8659 netdev->hw_features |= netdev->features;
8661 if (vsi->type == I40E_VSI_MAIN) {
8662 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
8663 ether_addr_copy(mac_addr, hw->mac.perm_addr);
8664 /* The following steps are necessary to prevent reception
8665 * of tagged packets - some older NVM configurations load a
8666 * default a MAC-VLAN filter that accepts any tagged packet
8667 * which must be replaced by a normal filter.
8669 if (!i40e_rm_default_mac_filter(vsi, mac_addr)) {
8670 spin_lock_bh(&vsi->mac_filter_list_lock);
8671 i40e_add_filter(vsi, mac_addr,
8672 I40E_VLAN_ANY, false, true);
8673 spin_unlock_bh(&vsi->mac_filter_list_lock);
8675 } else {
8676 /* relate the VSI_VMDQ name to the VSI_MAIN name */
8677 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
8678 pf->vsi[pf->lan_vsi]->netdev->name);
8679 random_ether_addr(mac_addr);
8681 spin_lock_bh(&vsi->mac_filter_list_lock);
8682 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
8683 spin_unlock_bh(&vsi->mac_filter_list_lock);
8686 spin_lock_bh(&vsi->mac_filter_list_lock);
8687 i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
8688 spin_unlock_bh(&vsi->mac_filter_list_lock);
8690 ether_addr_copy(netdev->dev_addr, mac_addr);
8691 ether_addr_copy(netdev->perm_addr, mac_addr);
8692 /* vlan gets same features (except vlan offload)
8693 * after any tweaks for specific VSI types
8695 netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
8696 NETIF_F_HW_VLAN_CTAG_RX |
8697 NETIF_F_HW_VLAN_CTAG_FILTER);
8698 netdev->priv_flags |= IFF_UNICAST_FLT;
8699 netdev->priv_flags |= IFF_SUPP_NOFCS;
8700 /* Setup netdev TC information */
8701 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
8703 netdev->netdev_ops = &i40e_netdev_ops;
8704 netdev->watchdog_timeo = 5 * HZ;
8705 i40e_set_ethtool_ops(netdev);
8706 #ifdef I40E_FCOE
8707 i40e_fcoe_config_netdev(netdev, vsi);
8708 #endif
8710 return 0;
8714 * i40e_vsi_delete - Delete a VSI from the switch
8715 * @vsi: the VSI being removed
8717 * Returns 0 on success, negative value on failure
8719 static void i40e_vsi_delete(struct i40e_vsi *vsi)
8721 /* remove default VSI is not allowed */
8722 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
8723 return;
8725 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
8729 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
8730 * @vsi: the VSI being queried
8732 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
8734 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
8736 struct i40e_veb *veb;
8737 struct i40e_pf *pf = vsi->back;
8739 /* Uplink is not a bridge so default to VEB */
8740 if (vsi->veb_idx == I40E_NO_VEB)
8741 return 1;
8743 veb = pf->veb[vsi->veb_idx];
8744 /* Uplink is a bridge in VEPA mode */
8745 if (veb && (veb->bridge_mode & BRIDGE_MODE_VEPA))
8746 return 0;
8748 /* Uplink is a bridge in VEB mode */
8749 return 1;
8753 * i40e_add_vsi - Add a VSI to the switch
8754 * @vsi: the VSI being configured
8756 * This initializes a VSI context depending on the VSI type to be added and
8757 * passes it down to the add_vsi aq command.
8759 static int i40e_add_vsi(struct i40e_vsi *vsi)
8761 int ret = -ENODEV;
8762 u8 laa_macaddr[ETH_ALEN];
8763 bool found_laa_mac_filter = false;
8764 struct i40e_pf *pf = vsi->back;
8765 struct i40e_hw *hw = &pf->hw;
8766 struct i40e_vsi_context ctxt;
8767 struct i40e_mac_filter *f, *ftmp;
8769 u8 enabled_tc = 0x1; /* TC0 enabled */
8770 int f_count = 0;
8772 memset(&ctxt, 0, sizeof(ctxt));
8773 switch (vsi->type) {
8774 case I40E_VSI_MAIN:
8775 /* The PF's main VSI is already setup as part of the
8776 * device initialization, so we'll not bother with
8777 * the add_vsi call, but we will retrieve the current
8778 * VSI context.
8780 ctxt.seid = pf->main_vsi_seid;
8781 ctxt.pf_num = pf->hw.pf_id;
8782 ctxt.vf_num = 0;
8783 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
8784 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8785 if (ret) {
8786 dev_info(&pf->pdev->dev,
8787 "couldn't get PF vsi config, err %s aq_err %s\n",
8788 i40e_stat_str(&pf->hw, ret),
8789 i40e_aq_str(&pf->hw,
8790 pf->hw.aq.asq_last_status));
8791 return -ENOENT;
8793 vsi->info = ctxt.info;
8794 vsi->info.valid_sections = 0;
8796 vsi->seid = ctxt.seid;
8797 vsi->id = ctxt.vsi_number;
8799 enabled_tc = i40e_pf_get_tc_map(pf);
8801 /* MFP mode setup queue map and update VSI */
8802 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
8803 !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
8804 memset(&ctxt, 0, sizeof(ctxt));
8805 ctxt.seid = pf->main_vsi_seid;
8806 ctxt.pf_num = pf->hw.pf_id;
8807 ctxt.vf_num = 0;
8808 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
8809 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
8810 if (ret) {
8811 dev_info(&pf->pdev->dev,
8812 "update vsi failed, err %s aq_err %s\n",
8813 i40e_stat_str(&pf->hw, ret),
8814 i40e_aq_str(&pf->hw,
8815 pf->hw.aq.asq_last_status));
8816 ret = -ENOENT;
8817 goto err;
8819 /* update the local VSI info queue map */
8820 i40e_vsi_update_queue_map(vsi, &ctxt);
8821 vsi->info.valid_sections = 0;
8822 } else {
8823 /* Default/Main VSI is only enabled for TC0
8824 * reconfigure it to enable all TCs that are
8825 * available on the port in SFP mode.
8826 * For MFP case the iSCSI PF would use this
8827 * flow to enable LAN+iSCSI TC.
8829 ret = i40e_vsi_config_tc(vsi, enabled_tc);
8830 if (ret) {
8831 dev_info(&pf->pdev->dev,
8832 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
8833 enabled_tc,
8834 i40e_stat_str(&pf->hw, ret),
8835 i40e_aq_str(&pf->hw,
8836 pf->hw.aq.asq_last_status));
8837 ret = -ENOENT;
8840 break;
8842 case I40E_VSI_FDIR:
8843 ctxt.pf_num = hw->pf_id;
8844 ctxt.vf_num = 0;
8845 ctxt.uplink_seid = vsi->uplink_seid;
8846 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8847 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8848 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
8849 (i40e_is_vsi_uplink_mode_veb(vsi))) {
8850 ctxt.info.valid_sections |=
8851 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8852 ctxt.info.switch_id =
8853 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8855 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8856 break;
8858 case I40E_VSI_VMDQ2:
8859 ctxt.pf_num = hw->pf_id;
8860 ctxt.vf_num = 0;
8861 ctxt.uplink_seid = vsi->uplink_seid;
8862 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8863 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
8865 /* This VSI is connected to VEB so the switch_id
8866 * should be set to zero by default.
8868 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8869 ctxt.info.valid_sections |=
8870 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8871 ctxt.info.switch_id =
8872 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8875 /* Setup the VSI tx/rx queue map for TC0 only for now */
8876 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8877 break;
8879 case I40E_VSI_SRIOV:
8880 ctxt.pf_num = hw->pf_id;
8881 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
8882 ctxt.uplink_seid = vsi->uplink_seid;
8883 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8884 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
8886 /* This VSI is connected to VEB so the switch_id
8887 * should be set to zero by default.
8889 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8890 ctxt.info.valid_sections |=
8891 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8892 ctxt.info.switch_id =
8893 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8896 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
8897 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
8898 if (pf->vf[vsi->vf_id].spoofchk) {
8899 ctxt.info.valid_sections |=
8900 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
8901 ctxt.info.sec_flags |=
8902 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
8903 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
8905 /* Setup the VSI tx/rx queue map for TC0 only for now */
8906 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8907 break;
8909 #ifdef I40E_FCOE
8910 case I40E_VSI_FCOE:
8911 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
8912 if (ret) {
8913 dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
8914 return ret;
8916 break;
8918 #endif /* I40E_FCOE */
8919 default:
8920 return -ENODEV;
8923 if (vsi->type != I40E_VSI_MAIN) {
8924 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
8925 if (ret) {
8926 dev_info(&vsi->back->pdev->dev,
8927 "add vsi failed, err %s aq_err %s\n",
8928 i40e_stat_str(&pf->hw, ret),
8929 i40e_aq_str(&pf->hw,
8930 pf->hw.aq.asq_last_status));
8931 ret = -ENOENT;
8932 goto err;
8934 vsi->info = ctxt.info;
8935 vsi->info.valid_sections = 0;
8936 vsi->seid = ctxt.seid;
8937 vsi->id = ctxt.vsi_number;
8940 spin_lock_bh(&vsi->mac_filter_list_lock);
8941 /* If macvlan filters already exist, force them to get loaded */
8942 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
8943 f->changed = true;
8944 f_count++;
8946 /* Expected to have only one MAC filter entry for LAA in list */
8947 if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
8948 ether_addr_copy(laa_macaddr, f->macaddr);
8949 found_laa_mac_filter = true;
8952 spin_unlock_bh(&vsi->mac_filter_list_lock);
8954 if (found_laa_mac_filter) {
8955 struct i40e_aqc_remove_macvlan_element_data element;
8957 memset(&element, 0, sizeof(element));
8958 ether_addr_copy(element.mac_addr, laa_macaddr);
8959 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
8960 ret = i40e_aq_remove_macvlan(hw, vsi->seid,
8961 &element, 1, NULL);
8962 if (ret) {
8963 /* some older FW has a different default */
8964 element.flags |=
8965 I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
8966 i40e_aq_remove_macvlan(hw, vsi->seid,
8967 &element, 1, NULL);
8970 i40e_aq_mac_address_write(hw,
8971 I40E_AQC_WRITE_TYPE_LAA_WOL,
8972 laa_macaddr, NULL);
8975 if (f_count) {
8976 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
8977 pf->flags |= I40E_FLAG_FILTER_SYNC;
8980 /* Update VSI BW information */
8981 ret = i40e_vsi_get_bw_info(vsi);
8982 if (ret) {
8983 dev_info(&pf->pdev->dev,
8984 "couldn't get vsi bw info, err %s aq_err %s\n",
8985 i40e_stat_str(&pf->hw, ret),
8986 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
8987 /* VSI is already added so not tearing that up */
8988 ret = 0;
8991 err:
8992 return ret;
8996 * i40e_vsi_release - Delete a VSI and free its resources
8997 * @vsi: the VSI being removed
8999 * Returns 0 on success or < 0 on error
9001 int i40e_vsi_release(struct i40e_vsi *vsi)
9003 struct i40e_mac_filter *f, *ftmp;
9004 struct i40e_veb *veb = NULL;
9005 struct i40e_pf *pf;
9006 u16 uplink_seid;
9007 int i, n;
9009 pf = vsi->back;
9011 /* release of a VEB-owner or last VSI is not allowed */
9012 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9013 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9014 vsi->seid, vsi->uplink_seid);
9015 return -ENODEV;
9017 if (vsi == pf->vsi[pf->lan_vsi] &&
9018 !test_bit(__I40E_DOWN, &pf->state)) {
9019 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9020 return -ENODEV;
9023 uplink_seid = vsi->uplink_seid;
9024 if (vsi->type != I40E_VSI_SRIOV) {
9025 if (vsi->netdev_registered) {
9026 vsi->netdev_registered = false;
9027 if (vsi->netdev) {
9028 /* results in a call to i40e_close() */
9029 unregister_netdev(vsi->netdev);
9031 } else {
9032 i40e_vsi_close(vsi);
9034 i40e_vsi_disable_irq(vsi);
9037 spin_lock_bh(&vsi->mac_filter_list_lock);
9038 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9039 i40e_del_filter(vsi, f->macaddr, f->vlan,
9040 f->is_vf, f->is_netdev);
9041 spin_unlock_bh(&vsi->mac_filter_list_lock);
9043 i40e_sync_vsi_filters(vsi, false);
9045 i40e_vsi_delete(vsi);
9046 i40e_vsi_free_q_vectors(vsi);
9047 if (vsi->netdev) {
9048 free_netdev(vsi->netdev);
9049 vsi->netdev = NULL;
9051 i40e_vsi_clear_rings(vsi);
9052 i40e_vsi_clear(vsi);
9054 /* If this was the last thing on the VEB, except for the
9055 * controlling VSI, remove the VEB, which puts the controlling
9056 * VSI onto the next level down in the switch.
9058 * Well, okay, there's one more exception here: don't remove
9059 * the orphan VEBs yet. We'll wait for an explicit remove request
9060 * from up the network stack.
9062 for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9063 if (pf->vsi[i] &&
9064 pf->vsi[i]->uplink_seid == uplink_seid &&
9065 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9066 n++; /* count the VSIs */
9069 for (i = 0; i < I40E_MAX_VEB; i++) {
9070 if (!pf->veb[i])
9071 continue;
9072 if (pf->veb[i]->uplink_seid == uplink_seid)
9073 n++; /* count the VEBs */
9074 if (pf->veb[i]->seid == uplink_seid)
9075 veb = pf->veb[i];
9077 if (n == 0 && veb && veb->uplink_seid != 0)
9078 i40e_veb_release(veb);
9080 return 0;
9084 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9085 * @vsi: ptr to the VSI
9087 * This should only be called after i40e_vsi_mem_alloc() which allocates the
9088 * corresponding SW VSI structure and initializes num_queue_pairs for the
9089 * newly allocated VSI.
9091 * Returns 0 on success or negative on failure
9093 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9095 int ret = -ENOENT;
9096 struct i40e_pf *pf = vsi->back;
9098 if (vsi->q_vectors[0]) {
9099 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9100 vsi->seid);
9101 return -EEXIST;
9104 if (vsi->base_vector) {
9105 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9106 vsi->seid, vsi->base_vector);
9107 return -EEXIST;
9110 ret = i40e_vsi_alloc_q_vectors(vsi);
9111 if (ret) {
9112 dev_info(&pf->pdev->dev,
9113 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9114 vsi->num_q_vectors, vsi->seid, ret);
9115 vsi->num_q_vectors = 0;
9116 goto vector_setup_out;
9119 /* In Legacy mode, we do not have to get any other vector since we
9120 * piggyback on the misc/ICR0 for queue interrupts.
9122 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9123 return ret;
9124 if (vsi->num_q_vectors)
9125 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9126 vsi->num_q_vectors, vsi->idx);
9127 if (vsi->base_vector < 0) {
9128 dev_info(&pf->pdev->dev,
9129 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9130 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9131 i40e_vsi_free_q_vectors(vsi);
9132 ret = -ENOENT;
9133 goto vector_setup_out;
9136 vector_setup_out:
9137 return ret;
9141 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9142 * @vsi: pointer to the vsi.
9144 * This re-allocates a vsi's queue resources.
9146 * Returns pointer to the successfully allocated and configured VSI sw struct
9147 * on success, otherwise returns NULL on failure.
9149 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9151 struct i40e_pf *pf = vsi->back;
9152 u8 enabled_tc;
9153 int ret;
9155 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9156 i40e_vsi_clear_rings(vsi);
9158 i40e_vsi_free_arrays(vsi, false);
9159 i40e_set_num_rings_in_vsi(vsi);
9160 ret = i40e_vsi_alloc_arrays(vsi, false);
9161 if (ret)
9162 goto err_vsi;
9164 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9165 if (ret < 0) {
9166 dev_info(&pf->pdev->dev,
9167 "failed to get tracking for %d queues for VSI %d err %d\n",
9168 vsi->alloc_queue_pairs, vsi->seid, ret);
9169 goto err_vsi;
9171 vsi->base_queue = ret;
9173 /* Update the FW view of the VSI. Force a reset of TC and queue
9174 * layout configurations.
9176 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9177 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9178 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9179 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9181 /* assign it some queues */
9182 ret = i40e_alloc_rings(vsi);
9183 if (ret)
9184 goto err_rings;
9186 /* map all of the rings to the q_vectors */
9187 i40e_vsi_map_rings_to_vectors(vsi);
9188 return vsi;
9190 err_rings:
9191 i40e_vsi_free_q_vectors(vsi);
9192 if (vsi->netdev_registered) {
9193 vsi->netdev_registered = false;
9194 unregister_netdev(vsi->netdev);
9195 free_netdev(vsi->netdev);
9196 vsi->netdev = NULL;
9198 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9199 err_vsi:
9200 i40e_vsi_clear(vsi);
9201 return NULL;
9205 * i40e_vsi_setup - Set up a VSI by a given type
9206 * @pf: board private structure
9207 * @type: VSI type
9208 * @uplink_seid: the switch element to link to
9209 * @param1: usage depends upon VSI type. For VF types, indicates VF id
9211 * This allocates the sw VSI structure and its queue resources, then add a VSI
9212 * to the identified VEB.
9214 * Returns pointer to the successfully allocated and configure VSI sw struct on
9215 * success, otherwise returns NULL on failure.
9217 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9218 u16 uplink_seid, u32 param1)
9220 struct i40e_vsi *vsi = NULL;
9221 struct i40e_veb *veb = NULL;
9222 int ret, i;
9223 int v_idx;
9225 /* The requested uplink_seid must be either
9226 * - the PF's port seid
9227 * no VEB is needed because this is the PF
9228 * or this is a Flow Director special case VSI
9229 * - seid of an existing VEB
9230 * - seid of a VSI that owns an existing VEB
9231 * - seid of a VSI that doesn't own a VEB
9232 * a new VEB is created and the VSI becomes the owner
9233 * - seid of the PF VSI, which is what creates the first VEB
9234 * this is a special case of the previous
9236 * Find which uplink_seid we were given and create a new VEB if needed
9238 for (i = 0; i < I40E_MAX_VEB; i++) {
9239 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9240 veb = pf->veb[i];
9241 break;
9245 if (!veb && uplink_seid != pf->mac_seid) {
9247 for (i = 0; i < pf->num_alloc_vsi; i++) {
9248 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9249 vsi = pf->vsi[i];
9250 break;
9253 if (!vsi) {
9254 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9255 uplink_seid);
9256 return NULL;
9259 if (vsi->uplink_seid == pf->mac_seid)
9260 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9261 vsi->tc_config.enabled_tc);
9262 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9263 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9264 vsi->tc_config.enabled_tc);
9265 if (veb) {
9266 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9267 dev_info(&vsi->back->pdev->dev,
9268 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9269 return NULL;
9271 /* We come up by default in VEPA mode if SRIOV is not
9272 * already enabled, in which case we can't force VEPA
9273 * mode.
9275 if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9276 veb->bridge_mode = BRIDGE_MODE_VEPA;
9277 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9279 i40e_config_bridge_mode(veb);
9281 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9282 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9283 veb = pf->veb[i];
9285 if (!veb) {
9286 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9287 return NULL;
9290 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9291 uplink_seid = veb->seid;
9294 /* get vsi sw struct */
9295 v_idx = i40e_vsi_mem_alloc(pf, type);
9296 if (v_idx < 0)
9297 goto err_alloc;
9298 vsi = pf->vsi[v_idx];
9299 if (!vsi)
9300 goto err_alloc;
9301 vsi->type = type;
9302 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9304 if (type == I40E_VSI_MAIN)
9305 pf->lan_vsi = v_idx;
9306 else if (type == I40E_VSI_SRIOV)
9307 vsi->vf_id = param1;
9308 /* assign it some queues */
9309 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9310 vsi->idx);
9311 if (ret < 0) {
9312 dev_info(&pf->pdev->dev,
9313 "failed to get tracking for %d queues for VSI %d err=%d\n",
9314 vsi->alloc_queue_pairs, vsi->seid, ret);
9315 goto err_vsi;
9317 vsi->base_queue = ret;
9319 /* get a VSI from the hardware */
9320 vsi->uplink_seid = uplink_seid;
9321 ret = i40e_add_vsi(vsi);
9322 if (ret)
9323 goto err_vsi;
9325 switch (vsi->type) {
9326 /* setup the netdev if needed */
9327 case I40E_VSI_MAIN:
9328 case I40E_VSI_VMDQ2:
9329 case I40E_VSI_FCOE:
9330 ret = i40e_config_netdev(vsi);
9331 if (ret)
9332 goto err_netdev;
9333 ret = register_netdev(vsi->netdev);
9334 if (ret)
9335 goto err_netdev;
9336 vsi->netdev_registered = true;
9337 netif_carrier_off(vsi->netdev);
9338 #ifdef CONFIG_I40E_DCB
9339 /* Setup DCB netlink interface */
9340 i40e_dcbnl_setup(vsi);
9341 #endif /* CONFIG_I40E_DCB */
9342 /* fall through */
9344 case I40E_VSI_FDIR:
9345 /* set up vectors and rings if needed */
9346 ret = i40e_vsi_setup_vectors(vsi);
9347 if (ret)
9348 goto err_msix;
9350 ret = i40e_alloc_rings(vsi);
9351 if (ret)
9352 goto err_rings;
9354 /* map all of the rings to the q_vectors */
9355 i40e_vsi_map_rings_to_vectors(vsi);
9357 i40e_vsi_reset_stats(vsi);
9358 break;
9360 default:
9361 /* no netdev or rings for the other VSI types */
9362 break;
9365 if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9366 (vsi->type == I40E_VSI_VMDQ2)) {
9367 ret = i40e_vsi_config_rss(vsi);
9369 return vsi;
9371 err_rings:
9372 i40e_vsi_free_q_vectors(vsi);
9373 err_msix:
9374 if (vsi->netdev_registered) {
9375 vsi->netdev_registered = false;
9376 unregister_netdev(vsi->netdev);
9377 free_netdev(vsi->netdev);
9378 vsi->netdev = NULL;
9380 err_netdev:
9381 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9382 err_vsi:
9383 i40e_vsi_clear(vsi);
9384 err_alloc:
9385 return NULL;
9389 * i40e_veb_get_bw_info - Query VEB BW information
9390 * @veb: the veb to query
9392 * Query the Tx scheduler BW configuration data for given VEB
9394 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9396 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9397 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9398 struct i40e_pf *pf = veb->pf;
9399 struct i40e_hw *hw = &pf->hw;
9400 u32 tc_bw_max;
9401 int ret = 0;
9402 int i;
9404 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9405 &bw_data, NULL);
9406 if (ret) {
9407 dev_info(&pf->pdev->dev,
9408 "query veb bw config failed, err %s aq_err %s\n",
9409 i40e_stat_str(&pf->hw, ret),
9410 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9411 goto out;
9414 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9415 &ets_data, NULL);
9416 if (ret) {
9417 dev_info(&pf->pdev->dev,
9418 "query veb bw ets config failed, err %s aq_err %s\n",
9419 i40e_stat_str(&pf->hw, ret),
9420 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9421 goto out;
9424 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9425 veb->bw_max_quanta = ets_data.tc_bw_max;
9426 veb->is_abs_credits = bw_data.absolute_credits_enable;
9427 veb->enabled_tc = ets_data.tc_valid_bits;
9428 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9429 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9430 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9431 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9432 veb->bw_tc_limit_credits[i] =
9433 le16_to_cpu(bw_data.tc_bw_limits[i]);
9434 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9437 out:
9438 return ret;
9442 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9443 * @pf: board private structure
9445 * On error: returns error code (negative)
9446 * On success: returns vsi index in PF (positive)
9448 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9450 int ret = -ENOENT;
9451 struct i40e_veb *veb;
9452 int i;
9454 /* Need to protect the allocation of switch elements at the PF level */
9455 mutex_lock(&pf->switch_mutex);
9457 /* VEB list may be fragmented if VEB creation/destruction has
9458 * been happening. We can afford to do a quick scan to look
9459 * for any free slots in the list.
9461 * find next empty veb slot, looping back around if necessary
9463 i = 0;
9464 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9465 i++;
9466 if (i >= I40E_MAX_VEB) {
9467 ret = -ENOMEM;
9468 goto err_alloc_veb; /* out of VEB slots! */
9471 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9472 if (!veb) {
9473 ret = -ENOMEM;
9474 goto err_alloc_veb;
9476 veb->pf = pf;
9477 veb->idx = i;
9478 veb->enabled_tc = 1;
9480 pf->veb[i] = veb;
9481 ret = i;
9482 err_alloc_veb:
9483 mutex_unlock(&pf->switch_mutex);
9484 return ret;
9488 * i40e_switch_branch_release - Delete a branch of the switch tree
9489 * @branch: where to start deleting
9491 * This uses recursion to find the tips of the branch to be
9492 * removed, deleting until we get back to and can delete this VEB.
9494 static void i40e_switch_branch_release(struct i40e_veb *branch)
9496 struct i40e_pf *pf = branch->pf;
9497 u16 branch_seid = branch->seid;
9498 u16 veb_idx = branch->idx;
9499 int i;
9501 /* release any VEBs on this VEB - RECURSION */
9502 for (i = 0; i < I40E_MAX_VEB; i++) {
9503 if (!pf->veb[i])
9504 continue;
9505 if (pf->veb[i]->uplink_seid == branch->seid)
9506 i40e_switch_branch_release(pf->veb[i]);
9509 /* Release the VSIs on this VEB, but not the owner VSI.
9511 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
9512 * the VEB itself, so don't use (*branch) after this loop.
9514 for (i = 0; i < pf->num_alloc_vsi; i++) {
9515 if (!pf->vsi[i])
9516 continue;
9517 if (pf->vsi[i]->uplink_seid == branch_seid &&
9518 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9519 i40e_vsi_release(pf->vsi[i]);
9523 /* There's one corner case where the VEB might not have been
9524 * removed, so double check it here and remove it if needed.
9525 * This case happens if the veb was created from the debugfs
9526 * commands and no VSIs were added to it.
9528 if (pf->veb[veb_idx])
9529 i40e_veb_release(pf->veb[veb_idx]);
9533 * i40e_veb_clear - remove veb struct
9534 * @veb: the veb to remove
9536 static void i40e_veb_clear(struct i40e_veb *veb)
9538 if (!veb)
9539 return;
9541 if (veb->pf) {
9542 struct i40e_pf *pf = veb->pf;
9544 mutex_lock(&pf->switch_mutex);
9545 if (pf->veb[veb->idx] == veb)
9546 pf->veb[veb->idx] = NULL;
9547 mutex_unlock(&pf->switch_mutex);
9550 kfree(veb);
9554 * i40e_veb_release - Delete a VEB and free its resources
9555 * @veb: the VEB being removed
9557 void i40e_veb_release(struct i40e_veb *veb)
9559 struct i40e_vsi *vsi = NULL;
9560 struct i40e_pf *pf;
9561 int i, n = 0;
9563 pf = veb->pf;
9565 /* find the remaining VSI and check for extras */
9566 for (i = 0; i < pf->num_alloc_vsi; i++) {
9567 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
9568 n++;
9569 vsi = pf->vsi[i];
9572 if (n != 1) {
9573 dev_info(&pf->pdev->dev,
9574 "can't remove VEB %d with %d VSIs left\n",
9575 veb->seid, n);
9576 return;
9579 /* move the remaining VSI to uplink veb */
9580 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
9581 if (veb->uplink_seid) {
9582 vsi->uplink_seid = veb->uplink_seid;
9583 if (veb->uplink_seid == pf->mac_seid)
9584 vsi->veb_idx = I40E_NO_VEB;
9585 else
9586 vsi->veb_idx = veb->veb_idx;
9587 } else {
9588 /* floating VEB */
9589 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9590 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
9593 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9594 i40e_veb_clear(veb);
9598 * i40e_add_veb - create the VEB in the switch
9599 * @veb: the VEB to be instantiated
9600 * @vsi: the controlling VSI
9602 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
9604 struct i40e_pf *pf = veb->pf;
9605 bool is_default = veb->pf->cur_promisc;
9606 bool is_cloud = false;
9607 int ret;
9609 /* get a VEB from the hardware */
9610 ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
9611 veb->enabled_tc, is_default,
9612 is_cloud, &veb->seid, NULL);
9613 if (ret) {
9614 dev_info(&pf->pdev->dev,
9615 "couldn't add VEB, err %s aq_err %s\n",
9616 i40e_stat_str(&pf->hw, ret),
9617 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9618 return -EPERM;
9621 /* get statistics counter */
9622 ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
9623 &veb->stats_idx, NULL, NULL, NULL);
9624 if (ret) {
9625 dev_info(&pf->pdev->dev,
9626 "couldn't get VEB statistics idx, err %s aq_err %s\n",
9627 i40e_stat_str(&pf->hw, ret),
9628 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9629 return -EPERM;
9631 ret = i40e_veb_get_bw_info(veb);
9632 if (ret) {
9633 dev_info(&pf->pdev->dev,
9634 "couldn't get VEB bw info, err %s aq_err %s\n",
9635 i40e_stat_str(&pf->hw, ret),
9636 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9637 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9638 return -ENOENT;
9641 vsi->uplink_seid = veb->seid;
9642 vsi->veb_idx = veb->idx;
9643 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9645 return 0;
9649 * i40e_veb_setup - Set up a VEB
9650 * @pf: board private structure
9651 * @flags: VEB setup flags
9652 * @uplink_seid: the switch element to link to
9653 * @vsi_seid: the initial VSI seid
9654 * @enabled_tc: Enabled TC bit-map
9656 * This allocates the sw VEB structure and links it into the switch
9657 * It is possible and legal for this to be a duplicate of an already
9658 * existing VEB. It is also possible for both uplink and vsi seids
9659 * to be zero, in order to create a floating VEB.
9661 * Returns pointer to the successfully allocated VEB sw struct on
9662 * success, otherwise returns NULL on failure.
9664 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
9665 u16 uplink_seid, u16 vsi_seid,
9666 u8 enabled_tc)
9668 struct i40e_veb *veb, *uplink_veb = NULL;
9669 int vsi_idx, veb_idx;
9670 int ret;
9672 /* if one seid is 0, the other must be 0 to create a floating relay */
9673 if ((uplink_seid == 0 || vsi_seid == 0) &&
9674 (uplink_seid + vsi_seid != 0)) {
9675 dev_info(&pf->pdev->dev,
9676 "one, not both seid's are 0: uplink=%d vsi=%d\n",
9677 uplink_seid, vsi_seid);
9678 return NULL;
9681 /* make sure there is such a vsi and uplink */
9682 for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
9683 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
9684 break;
9685 if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
9686 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
9687 vsi_seid);
9688 return NULL;
9691 if (uplink_seid && uplink_seid != pf->mac_seid) {
9692 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9693 if (pf->veb[veb_idx] &&
9694 pf->veb[veb_idx]->seid == uplink_seid) {
9695 uplink_veb = pf->veb[veb_idx];
9696 break;
9699 if (!uplink_veb) {
9700 dev_info(&pf->pdev->dev,
9701 "uplink seid %d not found\n", uplink_seid);
9702 return NULL;
9706 /* get veb sw struct */
9707 veb_idx = i40e_veb_mem_alloc(pf);
9708 if (veb_idx < 0)
9709 goto err_alloc;
9710 veb = pf->veb[veb_idx];
9711 veb->flags = flags;
9712 veb->uplink_seid = uplink_seid;
9713 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
9714 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
9716 /* create the VEB in the switch */
9717 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
9718 if (ret)
9719 goto err_veb;
9720 if (vsi_idx == pf->lan_vsi)
9721 pf->lan_veb = veb->idx;
9723 return veb;
9725 err_veb:
9726 i40e_veb_clear(veb);
9727 err_alloc:
9728 return NULL;
9732 * i40e_setup_pf_switch_element - set PF vars based on switch type
9733 * @pf: board private structure
9734 * @ele: element we are building info from
9735 * @num_reported: total number of elements
9736 * @printconfig: should we print the contents
9738 * helper function to assist in extracting a few useful SEID values.
9740 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
9741 struct i40e_aqc_switch_config_element_resp *ele,
9742 u16 num_reported, bool printconfig)
9744 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
9745 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
9746 u8 element_type = ele->element_type;
9747 u16 seid = le16_to_cpu(ele->seid);
9749 if (printconfig)
9750 dev_info(&pf->pdev->dev,
9751 "type=%d seid=%d uplink=%d downlink=%d\n",
9752 element_type, seid, uplink_seid, downlink_seid);
9754 switch (element_type) {
9755 case I40E_SWITCH_ELEMENT_TYPE_MAC:
9756 pf->mac_seid = seid;
9757 break;
9758 case I40E_SWITCH_ELEMENT_TYPE_VEB:
9759 /* Main VEB? */
9760 if (uplink_seid != pf->mac_seid)
9761 break;
9762 if (pf->lan_veb == I40E_NO_VEB) {
9763 int v;
9765 /* find existing or else empty VEB */
9766 for (v = 0; v < I40E_MAX_VEB; v++) {
9767 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
9768 pf->lan_veb = v;
9769 break;
9772 if (pf->lan_veb == I40E_NO_VEB) {
9773 v = i40e_veb_mem_alloc(pf);
9774 if (v < 0)
9775 break;
9776 pf->lan_veb = v;
9780 pf->veb[pf->lan_veb]->seid = seid;
9781 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
9782 pf->veb[pf->lan_veb]->pf = pf;
9783 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
9784 break;
9785 case I40E_SWITCH_ELEMENT_TYPE_VSI:
9786 if (num_reported != 1)
9787 break;
9788 /* This is immediately after a reset so we can assume this is
9789 * the PF's VSI
9791 pf->mac_seid = uplink_seid;
9792 pf->pf_seid = downlink_seid;
9793 pf->main_vsi_seid = seid;
9794 if (printconfig)
9795 dev_info(&pf->pdev->dev,
9796 "pf_seid=%d main_vsi_seid=%d\n",
9797 pf->pf_seid, pf->main_vsi_seid);
9798 break;
9799 case I40E_SWITCH_ELEMENT_TYPE_PF:
9800 case I40E_SWITCH_ELEMENT_TYPE_VF:
9801 case I40E_SWITCH_ELEMENT_TYPE_EMP:
9802 case I40E_SWITCH_ELEMENT_TYPE_BMC:
9803 case I40E_SWITCH_ELEMENT_TYPE_PE:
9804 case I40E_SWITCH_ELEMENT_TYPE_PA:
9805 /* ignore these for now */
9806 break;
9807 default:
9808 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
9809 element_type, seid);
9810 break;
9815 * i40e_fetch_switch_configuration - Get switch config from firmware
9816 * @pf: board private structure
9817 * @printconfig: should we print the contents
9819 * Get the current switch configuration from the device and
9820 * extract a few useful SEID values.
9822 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
9824 struct i40e_aqc_get_switch_config_resp *sw_config;
9825 u16 next_seid = 0;
9826 int ret = 0;
9827 u8 *aq_buf;
9828 int i;
9830 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
9831 if (!aq_buf)
9832 return -ENOMEM;
9834 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
9835 do {
9836 u16 num_reported, num_total;
9838 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
9839 I40E_AQ_LARGE_BUF,
9840 &next_seid, NULL);
9841 if (ret) {
9842 dev_info(&pf->pdev->dev,
9843 "get switch config failed err %s aq_err %s\n",
9844 i40e_stat_str(&pf->hw, ret),
9845 i40e_aq_str(&pf->hw,
9846 pf->hw.aq.asq_last_status));
9847 kfree(aq_buf);
9848 return -ENOENT;
9851 num_reported = le16_to_cpu(sw_config->header.num_reported);
9852 num_total = le16_to_cpu(sw_config->header.num_total);
9854 if (printconfig)
9855 dev_info(&pf->pdev->dev,
9856 "header: %d reported %d total\n",
9857 num_reported, num_total);
9859 for (i = 0; i < num_reported; i++) {
9860 struct i40e_aqc_switch_config_element_resp *ele =
9861 &sw_config->element[i];
9863 i40e_setup_pf_switch_element(pf, ele, num_reported,
9864 printconfig);
9866 } while (next_seid != 0);
9868 kfree(aq_buf);
9869 return ret;
9873 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
9874 * @pf: board private structure
9875 * @reinit: if the Main VSI needs to re-initialized.
9877 * Returns 0 on success, negative value on failure
9879 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
9881 int ret;
9883 /* find out what's out there already */
9884 ret = i40e_fetch_switch_configuration(pf, false);
9885 if (ret) {
9886 dev_info(&pf->pdev->dev,
9887 "couldn't fetch switch config, err %s aq_err %s\n",
9888 i40e_stat_str(&pf->hw, ret),
9889 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9890 return ret;
9892 i40e_pf_reset_stats(pf);
9894 /* first time setup */
9895 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
9896 struct i40e_vsi *vsi = NULL;
9897 u16 uplink_seid;
9899 /* Set up the PF VSI associated with the PF's main VSI
9900 * that is already in the HW switch
9902 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
9903 uplink_seid = pf->veb[pf->lan_veb]->seid;
9904 else
9905 uplink_seid = pf->mac_seid;
9906 if (pf->lan_vsi == I40E_NO_VSI)
9907 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
9908 else if (reinit)
9909 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
9910 if (!vsi) {
9911 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
9912 i40e_fdir_teardown(pf);
9913 return -EAGAIN;
9915 } else {
9916 /* force a reset of TC and queue layout configurations */
9917 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9919 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9920 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9921 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9923 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
9925 i40e_fdir_sb_setup(pf);
9927 /* Setup static PF queue filter control settings */
9928 ret = i40e_setup_pf_filter_control(pf);
9929 if (ret) {
9930 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
9931 ret);
9932 /* Failure here should not stop continuing other steps */
9935 /* enable RSS in the HW, even for only one queue, as the stack can use
9936 * the hash
9938 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
9939 i40e_config_rss(pf);
9941 /* fill in link information and enable LSE reporting */
9942 i40e_update_link_info(&pf->hw);
9943 i40e_link_event(pf);
9945 /* Initialize user-specific link properties */
9946 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
9947 I40E_AQ_AN_COMPLETED) ? true : false);
9949 i40e_ptp_init(pf);
9951 return ret;
9955 * i40e_determine_queue_usage - Work out queue distribution
9956 * @pf: board private structure
9958 static void i40e_determine_queue_usage(struct i40e_pf *pf)
9960 int queues_left;
9962 pf->num_lan_qps = 0;
9963 #ifdef I40E_FCOE
9964 pf->num_fcoe_qps = 0;
9965 #endif
9967 /* Find the max queues to be put into basic use. We'll always be
9968 * using TC0, whether or not DCB is running, and TC0 will get the
9969 * big RSS set.
9971 queues_left = pf->hw.func_caps.num_tx_qp;
9973 if ((queues_left == 1) ||
9974 !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
9975 /* one qp for PF, no queues for anything else */
9976 queues_left = 0;
9977 pf->rss_size = pf->num_lan_qps = 1;
9979 /* make sure all the fancies are disabled */
9980 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
9981 #ifdef I40E_FCOE
9982 I40E_FLAG_FCOE_ENABLED |
9983 #endif
9984 I40E_FLAG_FD_SB_ENABLED |
9985 I40E_FLAG_FD_ATR_ENABLED |
9986 I40E_FLAG_DCB_CAPABLE |
9987 I40E_FLAG_SRIOV_ENABLED |
9988 I40E_FLAG_VMDQ_ENABLED);
9989 } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
9990 I40E_FLAG_FD_SB_ENABLED |
9991 I40E_FLAG_FD_ATR_ENABLED |
9992 I40E_FLAG_DCB_CAPABLE))) {
9993 /* one qp for PF */
9994 pf->rss_size = pf->num_lan_qps = 1;
9995 queues_left -= pf->num_lan_qps;
9997 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
9998 #ifdef I40E_FCOE
9999 I40E_FLAG_FCOE_ENABLED |
10000 #endif
10001 I40E_FLAG_FD_SB_ENABLED |
10002 I40E_FLAG_FD_ATR_ENABLED |
10003 I40E_FLAG_DCB_ENABLED |
10004 I40E_FLAG_VMDQ_ENABLED);
10005 } else {
10006 /* Not enough queues for all TCs */
10007 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10008 (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10009 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10010 dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10012 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10013 num_online_cpus());
10014 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10015 pf->hw.func_caps.num_tx_qp);
10017 queues_left -= pf->num_lan_qps;
10020 #ifdef I40E_FCOE
10021 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10022 if (I40E_DEFAULT_FCOE <= queues_left) {
10023 pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10024 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10025 pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10026 } else {
10027 pf->num_fcoe_qps = 0;
10028 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10029 dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10032 queues_left -= pf->num_fcoe_qps;
10035 #endif
10036 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10037 if (queues_left > 1) {
10038 queues_left -= 1; /* save 1 queue for FD */
10039 } else {
10040 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10041 dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10045 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10046 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10047 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10048 (queues_left / pf->num_vf_qps));
10049 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10052 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10053 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10054 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10055 (queues_left / pf->num_vmdq_qps));
10056 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10059 pf->queues_left = queues_left;
10060 dev_dbg(&pf->pdev->dev,
10061 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10062 pf->hw.func_caps.num_tx_qp,
10063 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10064 pf->num_lan_qps, pf->rss_size, pf->num_req_vfs, pf->num_vf_qps,
10065 pf->num_vmdq_vsis, pf->num_vmdq_qps, queues_left);
10066 #ifdef I40E_FCOE
10067 dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10068 #endif
10072 * i40e_setup_pf_filter_control - Setup PF static filter control
10073 * @pf: PF to be setup
10075 * i40e_setup_pf_filter_control sets up a PF's initial filter control
10076 * settings. If PE/FCoE are enabled then it will also set the per PF
10077 * based filter sizes required for them. It also enables Flow director,
10078 * ethertype and macvlan type filter settings for the pf.
10080 * Returns 0 on success, negative on failure
10082 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10084 struct i40e_filter_control_settings *settings = &pf->filter_settings;
10086 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10088 /* Flow Director is enabled */
10089 if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10090 settings->enable_fdir = true;
10092 /* Ethtype and MACVLAN filters enabled for PF */
10093 settings->enable_ethtype = true;
10094 settings->enable_macvlan = true;
10096 if (i40e_set_filter_control(&pf->hw, settings))
10097 return -ENOENT;
10099 return 0;
10102 #define INFO_STRING_LEN 255
10103 static void i40e_print_features(struct i40e_pf *pf)
10105 struct i40e_hw *hw = &pf->hw;
10106 char *buf, *string;
10108 string = kzalloc(INFO_STRING_LEN, GFP_KERNEL);
10109 if (!string) {
10110 dev_err(&pf->pdev->dev, "Features string allocation failed\n");
10111 return;
10114 buf = string;
10116 buf += sprintf(string, "Features: PF-id[%d] ", hw->pf_id);
10117 #ifdef CONFIG_PCI_IOV
10118 buf += sprintf(buf, "VFs: %d ", pf->num_req_vfs);
10119 #endif
10120 buf += sprintf(buf, "VSIs: %d QP: %d RX: %s ",
10121 pf->hw.func_caps.num_vsis,
10122 pf->vsi[pf->lan_vsi]->num_queue_pairs,
10123 pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
10125 if (pf->flags & I40E_FLAG_RSS_ENABLED)
10126 buf += sprintf(buf, "RSS ");
10127 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10128 buf += sprintf(buf, "FD_ATR ");
10129 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10130 buf += sprintf(buf, "FD_SB ");
10131 buf += sprintf(buf, "NTUPLE ");
10133 if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10134 buf += sprintf(buf, "DCB ");
10135 #if IS_ENABLED(CONFIG_VXLAN)
10136 buf += sprintf(buf, "VxLAN ");
10137 #endif
10138 if (pf->flags & I40E_FLAG_PTP)
10139 buf += sprintf(buf, "PTP ");
10140 #ifdef I40E_FCOE
10141 if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10142 buf += sprintf(buf, "FCOE ");
10143 #endif
10144 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10145 buf += sprintf(buf, "VEB ");
10146 else
10147 buf += sprintf(buf, "VEPA ");
10149 BUG_ON(buf > (string + INFO_STRING_LEN));
10150 dev_info(&pf->pdev->dev, "%s\n", string);
10151 kfree(string);
10155 * i40e_probe - Device initialization routine
10156 * @pdev: PCI device information struct
10157 * @ent: entry in i40e_pci_tbl
10159 * i40e_probe initializes a PF identified by a pci_dev structure.
10160 * The OS initialization, configuring of the PF private structure,
10161 * and a hardware reset occur.
10163 * Returns 0 on success, negative on failure
10165 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10167 struct i40e_aq_get_phy_abilities_resp abilities;
10168 struct i40e_pf *pf;
10169 struct i40e_hw *hw;
10170 static u16 pfs_found;
10171 u16 wol_nvm_bits;
10172 u16 link_status;
10173 int err;
10174 u32 len;
10175 u32 i;
10177 err = pci_enable_device_mem(pdev);
10178 if (err)
10179 return err;
10181 /* set up for high or low dma */
10182 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10183 if (err) {
10184 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10185 if (err) {
10186 dev_err(&pdev->dev,
10187 "DMA configuration failed: 0x%x\n", err);
10188 goto err_dma;
10192 /* set up pci connections */
10193 err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
10194 IORESOURCE_MEM), i40e_driver_name);
10195 if (err) {
10196 dev_info(&pdev->dev,
10197 "pci_request_selected_regions failed %d\n", err);
10198 goto err_pci_reg;
10201 pci_enable_pcie_error_reporting(pdev);
10202 pci_set_master(pdev);
10204 /* Now that we have a PCI connection, we need to do the
10205 * low level device setup. This is primarily setting up
10206 * the Admin Queue structures and then querying for the
10207 * device's current profile information.
10209 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10210 if (!pf) {
10211 err = -ENOMEM;
10212 goto err_pf_alloc;
10214 pf->next_vsi = 0;
10215 pf->pdev = pdev;
10216 set_bit(__I40E_DOWN, &pf->state);
10218 hw = &pf->hw;
10219 hw->back = pf;
10221 pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10222 I40E_MAX_CSR_SPACE);
10224 hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10225 if (!hw->hw_addr) {
10226 err = -EIO;
10227 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10228 (unsigned int)pci_resource_start(pdev, 0),
10229 pf->ioremap_len, err);
10230 goto err_ioremap;
10232 hw->vendor_id = pdev->vendor;
10233 hw->device_id = pdev->device;
10234 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10235 hw->subsystem_vendor_id = pdev->subsystem_vendor;
10236 hw->subsystem_device_id = pdev->subsystem_device;
10237 hw->bus.device = PCI_SLOT(pdev->devfn);
10238 hw->bus.func = PCI_FUNC(pdev->devfn);
10239 pf->instance = pfs_found;
10241 if (debug != -1) {
10242 pf->msg_enable = pf->hw.debug_mask;
10243 pf->msg_enable = debug;
10246 /* do a special CORER for clearing PXE mode once at init */
10247 if (hw->revision_id == 0 &&
10248 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10249 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10250 i40e_flush(hw);
10251 msleep(200);
10252 pf->corer_count++;
10254 i40e_clear_pxe_mode(hw);
10257 /* Reset here to make sure all is clean and to define PF 'n' */
10258 i40e_clear_hw(hw);
10259 err = i40e_pf_reset(hw);
10260 if (err) {
10261 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10262 goto err_pf_reset;
10264 pf->pfr_count++;
10266 hw->aq.num_arq_entries = I40E_AQ_LEN;
10267 hw->aq.num_asq_entries = I40E_AQ_LEN;
10268 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10269 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10270 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10272 snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10273 "%s-%s:misc",
10274 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10276 err = i40e_init_shared_code(hw);
10277 if (err) {
10278 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10279 err);
10280 goto err_pf_reset;
10283 /* set up a default setting for link flow control */
10284 pf->hw.fc.requested_mode = I40E_FC_NONE;
10286 err = i40e_init_adminq(hw);
10288 /* provide nvm, fw, api versions */
10289 dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10290 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10291 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10292 i40e_nvm_version_str(hw));
10294 if (err) {
10295 dev_info(&pdev->dev,
10296 "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
10297 goto err_pf_reset;
10300 if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10301 hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10302 dev_info(&pdev->dev,
10303 "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
10304 else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10305 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10306 dev_info(&pdev->dev,
10307 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10309 i40e_verify_eeprom(pf);
10311 /* Rev 0 hardware was never productized */
10312 if (hw->revision_id < 1)
10313 dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
10315 i40e_clear_pxe_mode(hw);
10316 err = i40e_get_capabilities(pf);
10317 if (err)
10318 goto err_adminq_setup;
10320 err = i40e_sw_init(pf);
10321 if (err) {
10322 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10323 goto err_sw_init;
10326 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10327 hw->func_caps.num_rx_qp,
10328 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10329 if (err) {
10330 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10331 goto err_init_lan_hmc;
10334 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10335 if (err) {
10336 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10337 err = -ENOENT;
10338 goto err_configure_lan_hmc;
10341 /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10342 * Ignore error return codes because if it was already disabled via
10343 * hardware settings this will fail
10345 if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
10346 (pf->hw.aq.fw_maj_ver < 4)) {
10347 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10348 i40e_aq_stop_lldp(hw, true, NULL);
10351 i40e_get_mac_addr(hw, hw->mac.addr);
10352 if (!is_valid_ether_addr(hw->mac.addr)) {
10353 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10354 err = -EIO;
10355 goto err_mac_addr;
10357 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10358 ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10359 i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10360 if (is_valid_ether_addr(hw->mac.port_addr))
10361 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10362 #ifdef I40E_FCOE
10363 err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10364 if (err)
10365 dev_info(&pdev->dev,
10366 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10367 if (!is_valid_ether_addr(hw->mac.san_addr)) {
10368 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10369 hw->mac.san_addr);
10370 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10372 dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10373 #endif /* I40E_FCOE */
10375 pci_set_drvdata(pdev, pf);
10376 pci_save_state(pdev);
10377 #ifdef CONFIG_I40E_DCB
10378 err = i40e_init_pf_dcb(pf);
10379 if (err) {
10380 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10381 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
10382 /* Continue without DCB enabled */
10384 #endif /* CONFIG_I40E_DCB */
10386 /* set up periodic task facility */
10387 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10388 pf->service_timer_period = HZ;
10390 INIT_WORK(&pf->service_task, i40e_service_task);
10391 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10392 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10394 /* NVM bit on means WoL disabled for the port */
10395 i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10396 if ((1 << hw->port) & wol_nvm_bits || hw->partition_id != 1)
10397 pf->wol_en = false;
10398 else
10399 pf->wol_en = true;
10400 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10402 /* set up the main switch operations */
10403 i40e_determine_queue_usage(pf);
10404 err = i40e_init_interrupt_scheme(pf);
10405 if (err)
10406 goto err_switch_setup;
10408 /* The number of VSIs reported by the FW is the minimum guaranteed
10409 * to us; HW supports far more and we share the remaining pool with
10410 * the other PFs. We allocate space for more than the guarantee with
10411 * the understanding that we might not get them all later.
10413 if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10414 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10415 else
10416 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10418 /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10419 len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
10420 pf->vsi = kzalloc(len, GFP_KERNEL);
10421 if (!pf->vsi) {
10422 err = -ENOMEM;
10423 goto err_switch_setup;
10426 #ifdef CONFIG_PCI_IOV
10427 /* prep for VF support */
10428 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10429 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10430 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10431 if (pci_num_vf(pdev))
10432 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
10434 #endif
10435 err = i40e_setup_pf_switch(pf, false);
10436 if (err) {
10437 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
10438 goto err_vsis;
10440 /* if FDIR VSI was set up, start it now */
10441 for (i = 0; i < pf->num_alloc_vsi; i++) {
10442 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
10443 i40e_vsi_open(pf->vsi[i]);
10444 break;
10448 /* driver is only interested in link up/down and module qualification
10449 * reports from firmware
10451 err = i40e_aq_set_phy_int_mask(&pf->hw,
10452 I40E_AQ_EVENT_LINK_UPDOWN |
10453 I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
10454 if (err)
10455 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
10456 i40e_stat_str(&pf->hw, err),
10457 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10459 if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
10460 (pf->hw.aq.fw_maj_ver < 4)) {
10461 msleep(75);
10462 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
10463 if (err)
10464 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
10465 i40e_stat_str(&pf->hw, err),
10466 i40e_aq_str(&pf->hw,
10467 pf->hw.aq.asq_last_status));
10469 /* The main driver is (mostly) up and happy. We need to set this state
10470 * before setting up the misc vector or we get a race and the vector
10471 * ends up disabled forever.
10473 clear_bit(__I40E_DOWN, &pf->state);
10475 /* In case of MSIX we are going to setup the misc vector right here
10476 * to handle admin queue events etc. In case of legacy and MSI
10477 * the misc functionality and queue processing is combined in
10478 * the same vector and that gets setup at open.
10480 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
10481 err = i40e_setup_misc_vector(pf);
10482 if (err) {
10483 dev_info(&pdev->dev,
10484 "setup of misc vector failed: %d\n", err);
10485 goto err_vsis;
10489 #ifdef CONFIG_PCI_IOV
10490 /* prep for VF support */
10491 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10492 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
10493 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
10494 u32 val;
10496 /* disable link interrupts for VFs */
10497 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
10498 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
10499 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
10500 i40e_flush(hw);
10502 if (pci_num_vf(pdev)) {
10503 dev_info(&pdev->dev,
10504 "Active VFs found, allocating resources.\n");
10505 err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
10506 if (err)
10507 dev_info(&pdev->dev,
10508 "Error %d allocating resources for existing VFs\n",
10509 err);
10512 #endif /* CONFIG_PCI_IOV */
10514 pfs_found++;
10516 i40e_dbg_pf_init(pf);
10518 /* tell the firmware that we're starting */
10519 i40e_send_version(pf);
10521 /* since everything's happy, start the service_task timer */
10522 mod_timer(&pf->service_timer,
10523 round_jiffies(jiffies + pf->service_timer_period));
10525 #ifdef I40E_FCOE
10526 /* create FCoE interface */
10527 i40e_fcoe_vsi_setup(pf);
10529 #endif
10530 #define PCI_SPEED_SIZE 8
10531 #define PCI_WIDTH_SIZE 8
10532 /* Devices on the IOSF bus do not have this information
10533 * and will report PCI Gen 1 x 1 by default so don't bother
10534 * checking them.
10536 if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
10537 char speed[PCI_SPEED_SIZE] = "Unknown";
10538 char width[PCI_WIDTH_SIZE] = "Unknown";
10540 /* Get the negotiated link width and speed from PCI config
10541 * space
10543 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
10544 &link_status);
10546 i40e_set_pci_config_data(hw, link_status);
10548 switch (hw->bus.speed) {
10549 case i40e_bus_speed_8000:
10550 strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
10551 case i40e_bus_speed_5000:
10552 strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
10553 case i40e_bus_speed_2500:
10554 strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
10555 default:
10556 break;
10558 switch (hw->bus.width) {
10559 case i40e_bus_width_pcie_x8:
10560 strncpy(width, "8", PCI_WIDTH_SIZE); break;
10561 case i40e_bus_width_pcie_x4:
10562 strncpy(width, "4", PCI_WIDTH_SIZE); break;
10563 case i40e_bus_width_pcie_x2:
10564 strncpy(width, "2", PCI_WIDTH_SIZE); break;
10565 case i40e_bus_width_pcie_x1:
10566 strncpy(width, "1", PCI_WIDTH_SIZE); break;
10567 default:
10568 break;
10571 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
10572 speed, width);
10574 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
10575 hw->bus.speed < i40e_bus_speed_8000) {
10576 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
10577 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
10581 /* get the requested speeds from the fw */
10582 err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
10583 if (err)
10584 dev_dbg(&pf->pdev->dev, "get requested speeds ret = %s last_status = %s\n",
10585 i40e_stat_str(&pf->hw, err),
10586 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10587 pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
10589 /* get the supported phy types from the fw */
10590 err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
10591 if (err)
10592 dev_dbg(&pf->pdev->dev, "get supported phy types ret = %s last_status = %s\n",
10593 i40e_stat_str(&pf->hw, err),
10594 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10595 pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
10597 /* Add a filter to drop all Flow control frames from any VSI from being
10598 * transmitted. By doing so we stop a malicious VF from sending out
10599 * PAUSE or PFC frames and potentially controlling traffic for other
10600 * PF/VF VSIs.
10601 * The FW can still send Flow control frames if enabled.
10603 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
10604 pf->main_vsi_seid);
10606 /* print a string summarizing features */
10607 i40e_print_features(pf);
10609 return 0;
10611 /* Unwind what we've done if something failed in the setup */
10612 err_vsis:
10613 set_bit(__I40E_DOWN, &pf->state);
10614 i40e_clear_interrupt_scheme(pf);
10615 kfree(pf->vsi);
10616 err_switch_setup:
10617 i40e_reset_interrupt_capability(pf);
10618 del_timer_sync(&pf->service_timer);
10619 err_mac_addr:
10620 err_configure_lan_hmc:
10621 (void)i40e_shutdown_lan_hmc(hw);
10622 err_init_lan_hmc:
10623 kfree(pf->qp_pile);
10624 err_sw_init:
10625 err_adminq_setup:
10626 (void)i40e_shutdown_adminq(hw);
10627 err_pf_reset:
10628 iounmap(hw->hw_addr);
10629 err_ioremap:
10630 kfree(pf);
10631 err_pf_alloc:
10632 pci_disable_pcie_error_reporting(pdev);
10633 pci_release_selected_regions(pdev,
10634 pci_select_bars(pdev, IORESOURCE_MEM));
10635 err_pci_reg:
10636 err_dma:
10637 pci_disable_device(pdev);
10638 return err;
10642 * i40e_remove - Device removal routine
10643 * @pdev: PCI device information struct
10645 * i40e_remove is called by the PCI subsystem to alert the driver
10646 * that is should release a PCI device. This could be caused by a
10647 * Hot-Plug event, or because the driver is going to be removed from
10648 * memory.
10650 static void i40e_remove(struct pci_dev *pdev)
10652 struct i40e_pf *pf = pci_get_drvdata(pdev);
10653 struct i40e_hw *hw = &pf->hw;
10654 i40e_status ret_code;
10655 int i;
10657 i40e_dbg_pf_exit(pf);
10659 i40e_ptp_stop(pf);
10661 /* Disable RSS in hw */
10662 wr32(hw, I40E_PFQF_HENA(0), 0);
10663 wr32(hw, I40E_PFQF_HENA(1), 0);
10665 /* no more scheduling of any task */
10666 set_bit(__I40E_DOWN, &pf->state);
10667 del_timer_sync(&pf->service_timer);
10668 cancel_work_sync(&pf->service_task);
10669 i40e_fdir_teardown(pf);
10671 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
10672 i40e_free_vfs(pf);
10673 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
10676 i40e_fdir_teardown(pf);
10678 /* If there is a switch structure or any orphans, remove them.
10679 * This will leave only the PF's VSI remaining.
10681 for (i = 0; i < I40E_MAX_VEB; i++) {
10682 if (!pf->veb[i])
10683 continue;
10685 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
10686 pf->veb[i]->uplink_seid == 0)
10687 i40e_switch_branch_release(pf->veb[i]);
10690 /* Now we can shutdown the PF's VSI, just before we kill
10691 * adminq and hmc.
10693 if (pf->vsi[pf->lan_vsi])
10694 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
10696 /* shutdown and destroy the HMC */
10697 if (pf->hw.hmc.hmc_obj) {
10698 ret_code = i40e_shutdown_lan_hmc(&pf->hw);
10699 if (ret_code)
10700 dev_warn(&pdev->dev,
10701 "Failed to destroy the HMC resources: %d\n",
10702 ret_code);
10705 /* shutdown the adminq */
10706 ret_code = i40e_shutdown_adminq(&pf->hw);
10707 if (ret_code)
10708 dev_warn(&pdev->dev,
10709 "Failed to destroy the Admin Queue resources: %d\n",
10710 ret_code);
10712 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
10713 i40e_clear_interrupt_scheme(pf);
10714 for (i = 0; i < pf->num_alloc_vsi; i++) {
10715 if (pf->vsi[i]) {
10716 i40e_vsi_clear_rings(pf->vsi[i]);
10717 i40e_vsi_clear(pf->vsi[i]);
10718 pf->vsi[i] = NULL;
10722 for (i = 0; i < I40E_MAX_VEB; i++) {
10723 kfree(pf->veb[i]);
10724 pf->veb[i] = NULL;
10727 kfree(pf->qp_pile);
10728 kfree(pf->vsi);
10730 iounmap(pf->hw.hw_addr);
10731 kfree(pf);
10732 pci_release_selected_regions(pdev,
10733 pci_select_bars(pdev, IORESOURCE_MEM));
10735 pci_disable_pcie_error_reporting(pdev);
10736 pci_disable_device(pdev);
10740 * i40e_pci_error_detected - warning that something funky happened in PCI land
10741 * @pdev: PCI device information struct
10743 * Called to warn that something happened and the error handling steps
10744 * are in progress. Allows the driver to quiesce things, be ready for
10745 * remediation.
10747 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
10748 enum pci_channel_state error)
10750 struct i40e_pf *pf = pci_get_drvdata(pdev);
10752 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
10754 /* shutdown all operations */
10755 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
10756 rtnl_lock();
10757 i40e_prep_for_reset(pf);
10758 rtnl_unlock();
10761 /* Request a slot reset */
10762 return PCI_ERS_RESULT_NEED_RESET;
10766 * i40e_pci_error_slot_reset - a PCI slot reset just happened
10767 * @pdev: PCI device information struct
10769 * Called to find if the driver can work with the device now that
10770 * the pci slot has been reset. If a basic connection seems good
10771 * (registers are readable and have sane content) then return a
10772 * happy little PCI_ERS_RESULT_xxx.
10774 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
10776 struct i40e_pf *pf = pci_get_drvdata(pdev);
10777 pci_ers_result_t result;
10778 int err;
10779 u32 reg;
10781 dev_dbg(&pdev->dev, "%s\n", __func__);
10782 if (pci_enable_device_mem(pdev)) {
10783 dev_info(&pdev->dev,
10784 "Cannot re-enable PCI device after reset.\n");
10785 result = PCI_ERS_RESULT_DISCONNECT;
10786 } else {
10787 pci_set_master(pdev);
10788 pci_restore_state(pdev);
10789 pci_save_state(pdev);
10790 pci_wake_from_d3(pdev, false);
10792 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
10793 if (reg == 0)
10794 result = PCI_ERS_RESULT_RECOVERED;
10795 else
10796 result = PCI_ERS_RESULT_DISCONNECT;
10799 err = pci_cleanup_aer_uncorrect_error_status(pdev);
10800 if (err) {
10801 dev_info(&pdev->dev,
10802 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
10803 err);
10804 /* non-fatal, continue */
10807 return result;
10811 * i40e_pci_error_resume - restart operations after PCI error recovery
10812 * @pdev: PCI device information struct
10814 * Called to allow the driver to bring things back up after PCI error
10815 * and/or reset recovery has finished.
10817 static void i40e_pci_error_resume(struct pci_dev *pdev)
10819 struct i40e_pf *pf = pci_get_drvdata(pdev);
10821 dev_dbg(&pdev->dev, "%s\n", __func__);
10822 if (test_bit(__I40E_SUSPENDED, &pf->state))
10823 return;
10825 rtnl_lock();
10826 i40e_handle_reset_warning(pf);
10827 rtnl_unlock();
10831 * i40e_shutdown - PCI callback for shutting down
10832 * @pdev: PCI device information struct
10834 static void i40e_shutdown(struct pci_dev *pdev)
10836 struct i40e_pf *pf = pci_get_drvdata(pdev);
10837 struct i40e_hw *hw = &pf->hw;
10839 set_bit(__I40E_SUSPENDED, &pf->state);
10840 set_bit(__I40E_DOWN, &pf->state);
10841 rtnl_lock();
10842 i40e_prep_for_reset(pf);
10843 rtnl_unlock();
10845 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10846 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10848 del_timer_sync(&pf->service_timer);
10849 cancel_work_sync(&pf->service_task);
10850 i40e_fdir_teardown(pf);
10852 rtnl_lock();
10853 i40e_prep_for_reset(pf);
10854 rtnl_unlock();
10856 wr32(hw, I40E_PFPM_APM,
10857 (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10858 wr32(hw, I40E_PFPM_WUFC,
10859 (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10861 i40e_clear_interrupt_scheme(pf);
10863 if (system_state == SYSTEM_POWER_OFF) {
10864 pci_wake_from_d3(pdev, pf->wol_en);
10865 pci_set_power_state(pdev, PCI_D3hot);
10869 #ifdef CONFIG_PM
10871 * i40e_suspend - PCI callback for moving to D3
10872 * @pdev: PCI device information struct
10874 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
10876 struct i40e_pf *pf = pci_get_drvdata(pdev);
10877 struct i40e_hw *hw = &pf->hw;
10879 set_bit(__I40E_SUSPENDED, &pf->state);
10880 set_bit(__I40E_DOWN, &pf->state);
10882 rtnl_lock();
10883 i40e_prep_for_reset(pf);
10884 rtnl_unlock();
10886 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10887 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10889 pci_wake_from_d3(pdev, pf->wol_en);
10890 pci_set_power_state(pdev, PCI_D3hot);
10892 return 0;
10896 * i40e_resume - PCI callback for waking up from D3
10897 * @pdev: PCI device information struct
10899 static int i40e_resume(struct pci_dev *pdev)
10901 struct i40e_pf *pf = pci_get_drvdata(pdev);
10902 u32 err;
10904 pci_set_power_state(pdev, PCI_D0);
10905 pci_restore_state(pdev);
10906 /* pci_restore_state() clears dev->state_saves, so
10907 * call pci_save_state() again to restore it.
10909 pci_save_state(pdev);
10911 err = pci_enable_device_mem(pdev);
10912 if (err) {
10913 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
10914 return err;
10916 pci_set_master(pdev);
10918 /* no wakeup events while running */
10919 pci_wake_from_d3(pdev, false);
10921 /* handling the reset will rebuild the device state */
10922 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
10923 clear_bit(__I40E_DOWN, &pf->state);
10924 rtnl_lock();
10925 i40e_reset_and_rebuild(pf, false);
10926 rtnl_unlock();
10929 return 0;
10932 #endif
10933 static const struct pci_error_handlers i40e_err_handler = {
10934 .error_detected = i40e_pci_error_detected,
10935 .slot_reset = i40e_pci_error_slot_reset,
10936 .resume = i40e_pci_error_resume,
10939 static struct pci_driver i40e_driver = {
10940 .name = i40e_driver_name,
10941 .id_table = i40e_pci_tbl,
10942 .probe = i40e_probe,
10943 .remove = i40e_remove,
10944 #ifdef CONFIG_PM
10945 .suspend = i40e_suspend,
10946 .resume = i40e_resume,
10947 #endif
10948 .shutdown = i40e_shutdown,
10949 .err_handler = &i40e_err_handler,
10950 .sriov_configure = i40e_pci_sriov_configure,
10954 * i40e_init_module - Driver registration routine
10956 * i40e_init_module is the first routine called when the driver is
10957 * loaded. All it does is register with the PCI subsystem.
10959 static int __init i40e_init_module(void)
10961 pr_info("%s: %s - version %s\n", i40e_driver_name,
10962 i40e_driver_string, i40e_driver_version_str);
10963 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
10965 i40e_dbg_init();
10966 return pci_register_driver(&i40e_driver);
10968 module_init(i40e_init_module);
10971 * i40e_exit_module - Driver exit cleanup routine
10973 * i40e_exit_module is called just before the driver is removed
10974 * from memory.
10976 static void __exit i40e_exit_module(void)
10978 pci_unregister_driver(&i40e_driver);
10979 i40e_dbg_exit();
10981 module_exit(i40e_exit_module);