i40e: Do not enable NAPI on q_vectors that have no rings
[linux/fpc-iii.git] / drivers / net / ethernet / intel / i40e / i40e_main.c
blobbecffd15c092609afb0c12be2a851d177f2d7674
1 /*******************************************************************************
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 - 2016 Intel Corporation.
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
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 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
31 /* Local includes */
32 #include "i40e.h"
33 #include "i40e_diag.h"
34 #include <net/udp_tunnel.h>
36 const char i40e_driver_name[] = "i40e";
37 static const char i40e_driver_string[] =
38 "Intel(R) Ethernet Connection XL710 Network Driver";
40 #define DRV_KERN "-k"
42 #define DRV_VERSION_MAJOR 1
43 #define DRV_VERSION_MINOR 6
44 #define DRV_VERSION_BUILD 16
45 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
46 __stringify(DRV_VERSION_MINOR) "." \
47 __stringify(DRV_VERSION_BUILD) DRV_KERN
48 const char i40e_driver_version_str[] = DRV_VERSION;
49 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
51 /* a bit of forward declarations */
52 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
53 static void i40e_handle_reset_warning(struct i40e_pf *pf);
54 static int i40e_add_vsi(struct i40e_vsi *vsi);
55 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
56 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
57 static int i40e_setup_misc_vector(struct i40e_pf *pf);
58 static void i40e_determine_queue_usage(struct i40e_pf *pf);
59 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
60 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
61 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
63 /* i40e_pci_tbl - PCI Device ID Table
65 * Last entry must be all 0s
67 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
68 * Class, Class Mask, private data (not used) }
70 static const struct pci_device_id i40e_pci_tbl[] = {
71 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
72 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
73 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
74 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
75 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
76 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
77 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
78 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
79 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
80 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
81 {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
82 {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
83 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
84 {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
85 {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
86 {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
87 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
88 {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
89 /* required last entry */
90 {0, }
92 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
94 #define I40E_MAX_VF_COUNT 128
95 static int debug = -1;
96 module_param(debug, int, 0);
97 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
99 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
100 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
101 MODULE_LICENSE("GPL");
102 MODULE_VERSION(DRV_VERSION);
104 static struct workqueue_struct *i40e_wq;
107 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
108 * @hw: pointer to the HW structure
109 * @mem: ptr to mem struct to fill out
110 * @size: size of memory requested
111 * @alignment: what to align the allocation to
113 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
114 u64 size, u32 alignment)
116 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
118 mem->size = ALIGN(size, alignment);
119 mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
120 &mem->pa, GFP_KERNEL);
121 if (!mem->va)
122 return -ENOMEM;
124 return 0;
128 * i40e_free_dma_mem_d - OS specific memory free for shared code
129 * @hw: pointer to the HW structure
130 * @mem: ptr to mem struct to free
132 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
134 struct i40e_pf *pf = (struct i40e_pf *)hw->back;
136 dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
137 mem->va = NULL;
138 mem->pa = 0;
139 mem->size = 0;
141 return 0;
145 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
146 * @hw: pointer to the HW structure
147 * @mem: ptr to mem struct to fill out
148 * @size: size of memory requested
150 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
151 u32 size)
153 mem->size = size;
154 mem->va = kzalloc(size, GFP_KERNEL);
156 if (!mem->va)
157 return -ENOMEM;
159 return 0;
163 * i40e_free_virt_mem_d - OS specific memory free for shared code
164 * @hw: pointer to the HW structure
165 * @mem: ptr to mem struct to free
167 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
169 /* it's ok to kfree a NULL pointer */
170 kfree(mem->va);
171 mem->va = NULL;
172 mem->size = 0;
174 return 0;
178 * i40e_get_lump - find a lump of free generic resource
179 * @pf: board private structure
180 * @pile: the pile of resource to search
181 * @needed: the number of items needed
182 * @id: an owner id to stick on the items assigned
184 * Returns the base item index of the lump, or negative for error
186 * The search_hint trick and lack of advanced fit-finding only work
187 * because we're highly likely to have all the same size lump requests.
188 * Linear search time and any fragmentation should be minimal.
190 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
191 u16 needed, u16 id)
193 int ret = -ENOMEM;
194 int i, j;
196 if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
197 dev_info(&pf->pdev->dev,
198 "param err: pile=%p needed=%d id=0x%04x\n",
199 pile, needed, id);
200 return -EINVAL;
203 /* start the linear search with an imperfect hint */
204 i = pile->search_hint;
205 while (i < pile->num_entries) {
206 /* skip already allocated entries */
207 if (pile->list[i] & I40E_PILE_VALID_BIT) {
208 i++;
209 continue;
212 /* do we have enough in this lump? */
213 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
214 if (pile->list[i+j] & I40E_PILE_VALID_BIT)
215 break;
218 if (j == needed) {
219 /* there was enough, so assign it to the requestor */
220 for (j = 0; j < needed; j++)
221 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
222 ret = i;
223 pile->search_hint = i + j;
224 break;
227 /* not enough, so skip over it and continue looking */
228 i += j;
231 return ret;
235 * i40e_put_lump - return a lump of generic resource
236 * @pile: the pile of resource to search
237 * @index: the base item index
238 * @id: the owner id of the items assigned
240 * Returns the count of items in the lump
242 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
244 int valid_id = (id | I40E_PILE_VALID_BIT);
245 int count = 0;
246 int i;
248 if (!pile || index >= pile->num_entries)
249 return -EINVAL;
251 for (i = index;
252 i < pile->num_entries && pile->list[i] == valid_id;
253 i++) {
254 pile->list[i] = 0;
255 count++;
258 if (count && index < pile->search_hint)
259 pile->search_hint = index;
261 return count;
265 * i40e_find_vsi_from_id - searches for the vsi with the given id
266 * @pf - the pf structure to search for the vsi
267 * @id - id of the vsi it is searching for
269 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
271 int i;
273 for (i = 0; i < pf->num_alloc_vsi; i++)
274 if (pf->vsi[i] && (pf->vsi[i]->id == id))
275 return pf->vsi[i];
277 return NULL;
281 * i40e_service_event_schedule - Schedule the service task to wake up
282 * @pf: board private structure
284 * If not already scheduled, this puts the task into the work queue
286 void i40e_service_event_schedule(struct i40e_pf *pf)
288 if (!test_bit(__I40E_DOWN, &pf->state) &&
289 !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
290 !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
291 queue_work(i40e_wq, &pf->service_task);
295 * i40e_tx_timeout - Respond to a Tx Hang
296 * @netdev: network interface device structure
298 * If any port has noticed a Tx timeout, it is likely that the whole
299 * device is munged, not just the one netdev port, so go for the full
300 * reset.
302 #ifdef I40E_FCOE
303 void i40e_tx_timeout(struct net_device *netdev)
304 #else
305 static void i40e_tx_timeout(struct net_device *netdev)
306 #endif
308 struct i40e_netdev_priv *np = netdev_priv(netdev);
309 struct i40e_vsi *vsi = np->vsi;
310 struct i40e_pf *pf = vsi->back;
311 struct i40e_ring *tx_ring = NULL;
312 unsigned int i, hung_queue = 0;
313 u32 head, val;
315 pf->tx_timeout_count++;
317 /* find the stopped queue the same way the stack does */
318 for (i = 0; i < netdev->num_tx_queues; i++) {
319 struct netdev_queue *q;
320 unsigned long trans_start;
322 q = netdev_get_tx_queue(netdev, i);
323 trans_start = q->trans_start;
324 if (netif_xmit_stopped(q) &&
325 time_after(jiffies,
326 (trans_start + netdev->watchdog_timeo))) {
327 hung_queue = i;
328 break;
332 if (i == netdev->num_tx_queues) {
333 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
334 } else {
335 /* now that we have an index, find the tx_ring struct */
336 for (i = 0; i < vsi->num_queue_pairs; i++) {
337 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
338 if (hung_queue ==
339 vsi->tx_rings[i]->queue_index) {
340 tx_ring = vsi->tx_rings[i];
341 break;
347 if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
348 pf->tx_timeout_recovery_level = 1; /* reset after some time */
349 else if (time_before(jiffies,
350 (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
351 return; /* don't do any new action before the next timeout */
353 if (tx_ring) {
354 head = i40e_get_head(tx_ring);
355 /* Read interrupt register */
356 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
357 val = rd32(&pf->hw,
358 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
359 tx_ring->vsi->base_vector - 1));
360 else
361 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
363 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",
364 vsi->seid, hung_queue, tx_ring->next_to_clean,
365 head, tx_ring->next_to_use,
366 readl(tx_ring->tail), val);
369 pf->tx_timeout_last_recovery = jiffies;
370 netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
371 pf->tx_timeout_recovery_level, hung_queue);
373 switch (pf->tx_timeout_recovery_level) {
374 case 1:
375 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
376 break;
377 case 2:
378 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
379 break;
380 case 3:
381 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
382 break;
383 default:
384 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
385 break;
388 i40e_service_event_schedule(pf);
389 pf->tx_timeout_recovery_level++;
393 * i40e_get_vsi_stats_struct - Get System Network Statistics
394 * @vsi: the VSI we care about
396 * Returns the address of the device statistics structure.
397 * The statistics are actually updated from the service task.
399 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
401 return &vsi->net_stats;
405 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
406 * @netdev: network interface device structure
408 * Returns the address of the device statistics structure.
409 * The statistics are actually updated from the service task.
411 #ifdef I40E_FCOE
412 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
413 struct net_device *netdev,
414 struct rtnl_link_stats64 *stats)
415 #else
416 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
417 struct net_device *netdev,
418 struct rtnl_link_stats64 *stats)
419 #endif
421 struct i40e_netdev_priv *np = netdev_priv(netdev);
422 struct i40e_ring *tx_ring, *rx_ring;
423 struct i40e_vsi *vsi = np->vsi;
424 struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
425 int i;
427 if (test_bit(__I40E_DOWN, &vsi->state))
428 return stats;
430 if (!vsi->tx_rings)
431 return stats;
433 rcu_read_lock();
434 for (i = 0; i < vsi->num_queue_pairs; i++) {
435 u64 bytes, packets;
436 unsigned int start;
438 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
439 if (!tx_ring)
440 continue;
442 do {
443 start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
444 packets = tx_ring->stats.packets;
445 bytes = tx_ring->stats.bytes;
446 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
448 stats->tx_packets += packets;
449 stats->tx_bytes += bytes;
450 rx_ring = &tx_ring[1];
452 do {
453 start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
454 packets = rx_ring->stats.packets;
455 bytes = rx_ring->stats.bytes;
456 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
458 stats->rx_packets += packets;
459 stats->rx_bytes += bytes;
461 rcu_read_unlock();
463 /* following stats updated by i40e_watchdog_subtask() */
464 stats->multicast = vsi_stats->multicast;
465 stats->tx_errors = vsi_stats->tx_errors;
466 stats->tx_dropped = vsi_stats->tx_dropped;
467 stats->rx_errors = vsi_stats->rx_errors;
468 stats->rx_dropped = vsi_stats->rx_dropped;
469 stats->rx_crc_errors = vsi_stats->rx_crc_errors;
470 stats->rx_length_errors = vsi_stats->rx_length_errors;
472 return stats;
476 * i40e_vsi_reset_stats - Resets all stats of the given vsi
477 * @vsi: the VSI to have its stats reset
479 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
481 struct rtnl_link_stats64 *ns;
482 int i;
484 if (!vsi)
485 return;
487 ns = i40e_get_vsi_stats_struct(vsi);
488 memset(ns, 0, sizeof(*ns));
489 memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
490 memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
491 memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
492 if (vsi->rx_rings && vsi->rx_rings[0]) {
493 for (i = 0; i < vsi->num_queue_pairs; i++) {
494 memset(&vsi->rx_rings[i]->stats, 0,
495 sizeof(vsi->rx_rings[i]->stats));
496 memset(&vsi->rx_rings[i]->rx_stats, 0,
497 sizeof(vsi->rx_rings[i]->rx_stats));
498 memset(&vsi->tx_rings[i]->stats, 0,
499 sizeof(vsi->tx_rings[i]->stats));
500 memset(&vsi->tx_rings[i]->tx_stats, 0,
501 sizeof(vsi->tx_rings[i]->tx_stats));
504 vsi->stat_offsets_loaded = false;
508 * i40e_pf_reset_stats - Reset all of the stats for the given PF
509 * @pf: the PF to be reset
511 void i40e_pf_reset_stats(struct i40e_pf *pf)
513 int i;
515 memset(&pf->stats, 0, sizeof(pf->stats));
516 memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
517 pf->stat_offsets_loaded = false;
519 for (i = 0; i < I40E_MAX_VEB; i++) {
520 if (pf->veb[i]) {
521 memset(&pf->veb[i]->stats, 0,
522 sizeof(pf->veb[i]->stats));
523 memset(&pf->veb[i]->stats_offsets, 0,
524 sizeof(pf->veb[i]->stats_offsets));
525 pf->veb[i]->stat_offsets_loaded = false;
528 pf->hw_csum_rx_error = 0;
532 * i40e_stat_update48 - read and update a 48 bit stat from the chip
533 * @hw: ptr to the hardware info
534 * @hireg: the high 32 bit reg to read
535 * @loreg: the low 32 bit reg to read
536 * @offset_loaded: has the initial offset been loaded yet
537 * @offset: ptr to current offset value
538 * @stat: ptr to the stat
540 * Since the device stats are not reset at PFReset, they likely will not
541 * be zeroed when the driver starts. We'll save the first values read
542 * and use them as offsets to be subtracted from the raw values in order
543 * to report stats that count from zero. In the process, we also manage
544 * the potential roll-over.
546 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
547 bool offset_loaded, u64 *offset, u64 *stat)
549 u64 new_data;
551 if (hw->device_id == I40E_DEV_ID_QEMU) {
552 new_data = rd32(hw, loreg);
553 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
554 } else {
555 new_data = rd64(hw, loreg);
557 if (!offset_loaded)
558 *offset = new_data;
559 if (likely(new_data >= *offset))
560 *stat = new_data - *offset;
561 else
562 *stat = (new_data + BIT_ULL(48)) - *offset;
563 *stat &= 0xFFFFFFFFFFFFULL;
567 * i40e_stat_update32 - read and update a 32 bit stat from the chip
568 * @hw: ptr to the hardware info
569 * @reg: the hw reg to read
570 * @offset_loaded: has the initial offset been loaded yet
571 * @offset: ptr to current offset value
572 * @stat: ptr to the stat
574 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
575 bool offset_loaded, u64 *offset, u64 *stat)
577 u32 new_data;
579 new_data = rd32(hw, reg);
580 if (!offset_loaded)
581 *offset = new_data;
582 if (likely(new_data >= *offset))
583 *stat = (u32)(new_data - *offset);
584 else
585 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
589 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
590 * @vsi: the VSI to be updated
592 void i40e_update_eth_stats(struct i40e_vsi *vsi)
594 int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
595 struct i40e_pf *pf = vsi->back;
596 struct i40e_hw *hw = &pf->hw;
597 struct i40e_eth_stats *oes;
598 struct i40e_eth_stats *es; /* device's eth stats */
600 es = &vsi->eth_stats;
601 oes = &vsi->eth_stats_offsets;
603 /* Gather up the stats that the hw collects */
604 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
605 vsi->stat_offsets_loaded,
606 &oes->tx_errors, &es->tx_errors);
607 i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
608 vsi->stat_offsets_loaded,
609 &oes->rx_discards, &es->rx_discards);
610 i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
611 vsi->stat_offsets_loaded,
612 &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
613 i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
614 vsi->stat_offsets_loaded,
615 &oes->tx_errors, &es->tx_errors);
617 i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
618 I40E_GLV_GORCL(stat_idx),
619 vsi->stat_offsets_loaded,
620 &oes->rx_bytes, &es->rx_bytes);
621 i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
622 I40E_GLV_UPRCL(stat_idx),
623 vsi->stat_offsets_loaded,
624 &oes->rx_unicast, &es->rx_unicast);
625 i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
626 I40E_GLV_MPRCL(stat_idx),
627 vsi->stat_offsets_loaded,
628 &oes->rx_multicast, &es->rx_multicast);
629 i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
630 I40E_GLV_BPRCL(stat_idx),
631 vsi->stat_offsets_loaded,
632 &oes->rx_broadcast, &es->rx_broadcast);
634 i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
635 I40E_GLV_GOTCL(stat_idx),
636 vsi->stat_offsets_loaded,
637 &oes->tx_bytes, &es->tx_bytes);
638 i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
639 I40E_GLV_UPTCL(stat_idx),
640 vsi->stat_offsets_loaded,
641 &oes->tx_unicast, &es->tx_unicast);
642 i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
643 I40E_GLV_MPTCL(stat_idx),
644 vsi->stat_offsets_loaded,
645 &oes->tx_multicast, &es->tx_multicast);
646 i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
647 I40E_GLV_BPTCL(stat_idx),
648 vsi->stat_offsets_loaded,
649 &oes->tx_broadcast, &es->tx_broadcast);
650 vsi->stat_offsets_loaded = true;
654 * i40e_update_veb_stats - Update Switch component statistics
655 * @veb: the VEB being updated
657 static void i40e_update_veb_stats(struct i40e_veb *veb)
659 struct i40e_pf *pf = veb->pf;
660 struct i40e_hw *hw = &pf->hw;
661 struct i40e_eth_stats *oes;
662 struct i40e_eth_stats *es; /* device's eth stats */
663 struct i40e_veb_tc_stats *veb_oes;
664 struct i40e_veb_tc_stats *veb_es;
665 int i, idx = 0;
667 idx = veb->stats_idx;
668 es = &veb->stats;
669 oes = &veb->stats_offsets;
670 veb_es = &veb->tc_stats;
671 veb_oes = &veb->tc_stats_offsets;
673 /* Gather up the stats that the hw collects */
674 i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
675 veb->stat_offsets_loaded,
676 &oes->tx_discards, &es->tx_discards);
677 if (hw->revision_id > 0)
678 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
679 veb->stat_offsets_loaded,
680 &oes->rx_unknown_protocol,
681 &es->rx_unknown_protocol);
682 i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
683 veb->stat_offsets_loaded,
684 &oes->rx_bytes, &es->rx_bytes);
685 i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
686 veb->stat_offsets_loaded,
687 &oes->rx_unicast, &es->rx_unicast);
688 i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
689 veb->stat_offsets_loaded,
690 &oes->rx_multicast, &es->rx_multicast);
691 i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
692 veb->stat_offsets_loaded,
693 &oes->rx_broadcast, &es->rx_broadcast);
695 i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
696 veb->stat_offsets_loaded,
697 &oes->tx_bytes, &es->tx_bytes);
698 i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
699 veb->stat_offsets_loaded,
700 &oes->tx_unicast, &es->tx_unicast);
701 i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
702 veb->stat_offsets_loaded,
703 &oes->tx_multicast, &es->tx_multicast);
704 i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
705 veb->stat_offsets_loaded,
706 &oes->tx_broadcast, &es->tx_broadcast);
707 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
708 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
709 I40E_GLVEBTC_RPCL(i, idx),
710 veb->stat_offsets_loaded,
711 &veb_oes->tc_rx_packets[i],
712 &veb_es->tc_rx_packets[i]);
713 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
714 I40E_GLVEBTC_RBCL(i, idx),
715 veb->stat_offsets_loaded,
716 &veb_oes->tc_rx_bytes[i],
717 &veb_es->tc_rx_bytes[i]);
718 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
719 I40E_GLVEBTC_TPCL(i, idx),
720 veb->stat_offsets_loaded,
721 &veb_oes->tc_tx_packets[i],
722 &veb_es->tc_tx_packets[i]);
723 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
724 I40E_GLVEBTC_TBCL(i, idx),
725 veb->stat_offsets_loaded,
726 &veb_oes->tc_tx_bytes[i],
727 &veb_es->tc_tx_bytes[i]);
729 veb->stat_offsets_loaded = true;
732 #ifdef I40E_FCOE
734 * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
735 * @vsi: the VSI that is capable of doing FCoE
737 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
739 struct i40e_pf *pf = vsi->back;
740 struct i40e_hw *hw = &pf->hw;
741 struct i40e_fcoe_stats *ofs;
742 struct i40e_fcoe_stats *fs; /* device's eth stats */
743 int idx;
745 if (vsi->type != I40E_VSI_FCOE)
746 return;
748 idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
749 fs = &vsi->fcoe_stats;
750 ofs = &vsi->fcoe_stats_offsets;
752 i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
753 vsi->fcoe_stat_offsets_loaded,
754 &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
755 i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
756 vsi->fcoe_stat_offsets_loaded,
757 &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
758 i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
759 vsi->fcoe_stat_offsets_loaded,
760 &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
761 i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
762 vsi->fcoe_stat_offsets_loaded,
763 &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
764 i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
765 vsi->fcoe_stat_offsets_loaded,
766 &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
767 i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
768 vsi->fcoe_stat_offsets_loaded,
769 &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
770 i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
771 vsi->fcoe_stat_offsets_loaded,
772 &ofs->fcoe_last_error, &fs->fcoe_last_error);
773 i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
774 vsi->fcoe_stat_offsets_loaded,
775 &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
777 vsi->fcoe_stat_offsets_loaded = true;
780 #endif
782 * i40e_update_vsi_stats - Update the vsi statistics counters.
783 * @vsi: the VSI to be updated
785 * There are a few instances where we store the same stat in a
786 * couple of different structs. This is partly because we have
787 * the netdev stats that need to be filled out, which is slightly
788 * different from the "eth_stats" defined by the chip and used in
789 * VF communications. We sort it out here.
791 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
793 struct i40e_pf *pf = vsi->back;
794 struct rtnl_link_stats64 *ons;
795 struct rtnl_link_stats64 *ns; /* netdev stats */
796 struct i40e_eth_stats *oes;
797 struct i40e_eth_stats *es; /* device's eth stats */
798 u32 tx_restart, tx_busy;
799 u64 tx_lost_interrupt;
800 struct i40e_ring *p;
801 u32 rx_page, rx_buf;
802 u64 bytes, packets;
803 unsigned int start;
804 u64 tx_linearize;
805 u64 tx_force_wb;
806 u64 rx_p, rx_b;
807 u64 tx_p, tx_b;
808 u16 q;
810 if (test_bit(__I40E_DOWN, &vsi->state) ||
811 test_bit(__I40E_CONFIG_BUSY, &pf->state))
812 return;
814 ns = i40e_get_vsi_stats_struct(vsi);
815 ons = &vsi->net_stats_offsets;
816 es = &vsi->eth_stats;
817 oes = &vsi->eth_stats_offsets;
819 /* Gather up the netdev and vsi stats that the driver collects
820 * on the fly during packet processing
822 rx_b = rx_p = 0;
823 tx_b = tx_p = 0;
824 tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
825 tx_lost_interrupt = 0;
826 rx_page = 0;
827 rx_buf = 0;
828 rcu_read_lock();
829 for (q = 0; q < vsi->num_queue_pairs; q++) {
830 /* locate Tx ring */
831 p = ACCESS_ONCE(vsi->tx_rings[q]);
833 do {
834 start = u64_stats_fetch_begin_irq(&p->syncp);
835 packets = p->stats.packets;
836 bytes = p->stats.bytes;
837 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
838 tx_b += bytes;
839 tx_p += packets;
840 tx_restart += p->tx_stats.restart_queue;
841 tx_busy += p->tx_stats.tx_busy;
842 tx_linearize += p->tx_stats.tx_linearize;
843 tx_force_wb += p->tx_stats.tx_force_wb;
844 tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
846 /* Rx queue is part of the same block as Tx queue */
847 p = &p[1];
848 do {
849 start = u64_stats_fetch_begin_irq(&p->syncp);
850 packets = p->stats.packets;
851 bytes = p->stats.bytes;
852 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
853 rx_b += bytes;
854 rx_p += packets;
855 rx_buf += p->rx_stats.alloc_buff_failed;
856 rx_page += p->rx_stats.alloc_page_failed;
858 rcu_read_unlock();
859 vsi->tx_restart = tx_restart;
860 vsi->tx_busy = tx_busy;
861 vsi->tx_linearize = tx_linearize;
862 vsi->tx_force_wb = tx_force_wb;
863 vsi->tx_lost_interrupt = tx_lost_interrupt;
864 vsi->rx_page_failed = rx_page;
865 vsi->rx_buf_failed = rx_buf;
867 ns->rx_packets = rx_p;
868 ns->rx_bytes = rx_b;
869 ns->tx_packets = tx_p;
870 ns->tx_bytes = tx_b;
872 /* update netdev stats from eth stats */
873 i40e_update_eth_stats(vsi);
874 ons->tx_errors = oes->tx_errors;
875 ns->tx_errors = es->tx_errors;
876 ons->multicast = oes->rx_multicast;
877 ns->multicast = es->rx_multicast;
878 ons->rx_dropped = oes->rx_discards;
879 ns->rx_dropped = es->rx_discards;
880 ons->tx_dropped = oes->tx_discards;
881 ns->tx_dropped = es->tx_discards;
883 /* pull in a couple PF stats if this is the main vsi */
884 if (vsi == pf->vsi[pf->lan_vsi]) {
885 ns->rx_crc_errors = pf->stats.crc_errors;
886 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
887 ns->rx_length_errors = pf->stats.rx_length_errors;
892 * i40e_update_pf_stats - Update the PF statistics counters.
893 * @pf: the PF to be updated
895 static void i40e_update_pf_stats(struct i40e_pf *pf)
897 struct i40e_hw_port_stats *osd = &pf->stats_offsets;
898 struct i40e_hw_port_stats *nsd = &pf->stats;
899 struct i40e_hw *hw = &pf->hw;
900 u32 val;
901 int i;
903 i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
904 I40E_GLPRT_GORCL(hw->port),
905 pf->stat_offsets_loaded,
906 &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
907 i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
908 I40E_GLPRT_GOTCL(hw->port),
909 pf->stat_offsets_loaded,
910 &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
911 i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
912 pf->stat_offsets_loaded,
913 &osd->eth.rx_discards,
914 &nsd->eth.rx_discards);
915 i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
916 I40E_GLPRT_UPRCL(hw->port),
917 pf->stat_offsets_loaded,
918 &osd->eth.rx_unicast,
919 &nsd->eth.rx_unicast);
920 i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
921 I40E_GLPRT_MPRCL(hw->port),
922 pf->stat_offsets_loaded,
923 &osd->eth.rx_multicast,
924 &nsd->eth.rx_multicast);
925 i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
926 I40E_GLPRT_BPRCL(hw->port),
927 pf->stat_offsets_loaded,
928 &osd->eth.rx_broadcast,
929 &nsd->eth.rx_broadcast);
930 i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
931 I40E_GLPRT_UPTCL(hw->port),
932 pf->stat_offsets_loaded,
933 &osd->eth.tx_unicast,
934 &nsd->eth.tx_unicast);
935 i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
936 I40E_GLPRT_MPTCL(hw->port),
937 pf->stat_offsets_loaded,
938 &osd->eth.tx_multicast,
939 &nsd->eth.tx_multicast);
940 i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
941 I40E_GLPRT_BPTCL(hw->port),
942 pf->stat_offsets_loaded,
943 &osd->eth.tx_broadcast,
944 &nsd->eth.tx_broadcast);
946 i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
947 pf->stat_offsets_loaded,
948 &osd->tx_dropped_link_down,
949 &nsd->tx_dropped_link_down);
951 i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
952 pf->stat_offsets_loaded,
953 &osd->crc_errors, &nsd->crc_errors);
955 i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
956 pf->stat_offsets_loaded,
957 &osd->illegal_bytes, &nsd->illegal_bytes);
959 i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
960 pf->stat_offsets_loaded,
961 &osd->mac_local_faults,
962 &nsd->mac_local_faults);
963 i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
964 pf->stat_offsets_loaded,
965 &osd->mac_remote_faults,
966 &nsd->mac_remote_faults);
968 i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
969 pf->stat_offsets_loaded,
970 &osd->rx_length_errors,
971 &nsd->rx_length_errors);
973 i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
974 pf->stat_offsets_loaded,
975 &osd->link_xon_rx, &nsd->link_xon_rx);
976 i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
977 pf->stat_offsets_loaded,
978 &osd->link_xon_tx, &nsd->link_xon_tx);
979 i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
980 pf->stat_offsets_loaded,
981 &osd->link_xoff_rx, &nsd->link_xoff_rx);
982 i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
983 pf->stat_offsets_loaded,
984 &osd->link_xoff_tx, &nsd->link_xoff_tx);
986 for (i = 0; i < 8; i++) {
987 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
988 pf->stat_offsets_loaded,
989 &osd->priority_xoff_rx[i],
990 &nsd->priority_xoff_rx[i]);
991 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
992 pf->stat_offsets_loaded,
993 &osd->priority_xon_rx[i],
994 &nsd->priority_xon_rx[i]);
995 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
996 pf->stat_offsets_loaded,
997 &osd->priority_xon_tx[i],
998 &nsd->priority_xon_tx[i]);
999 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1000 pf->stat_offsets_loaded,
1001 &osd->priority_xoff_tx[i],
1002 &nsd->priority_xoff_tx[i]);
1003 i40e_stat_update32(hw,
1004 I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1005 pf->stat_offsets_loaded,
1006 &osd->priority_xon_2_xoff[i],
1007 &nsd->priority_xon_2_xoff[i]);
1010 i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1011 I40E_GLPRT_PRC64L(hw->port),
1012 pf->stat_offsets_loaded,
1013 &osd->rx_size_64, &nsd->rx_size_64);
1014 i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1015 I40E_GLPRT_PRC127L(hw->port),
1016 pf->stat_offsets_loaded,
1017 &osd->rx_size_127, &nsd->rx_size_127);
1018 i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1019 I40E_GLPRT_PRC255L(hw->port),
1020 pf->stat_offsets_loaded,
1021 &osd->rx_size_255, &nsd->rx_size_255);
1022 i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1023 I40E_GLPRT_PRC511L(hw->port),
1024 pf->stat_offsets_loaded,
1025 &osd->rx_size_511, &nsd->rx_size_511);
1026 i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1027 I40E_GLPRT_PRC1023L(hw->port),
1028 pf->stat_offsets_loaded,
1029 &osd->rx_size_1023, &nsd->rx_size_1023);
1030 i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1031 I40E_GLPRT_PRC1522L(hw->port),
1032 pf->stat_offsets_loaded,
1033 &osd->rx_size_1522, &nsd->rx_size_1522);
1034 i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1035 I40E_GLPRT_PRC9522L(hw->port),
1036 pf->stat_offsets_loaded,
1037 &osd->rx_size_big, &nsd->rx_size_big);
1039 i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1040 I40E_GLPRT_PTC64L(hw->port),
1041 pf->stat_offsets_loaded,
1042 &osd->tx_size_64, &nsd->tx_size_64);
1043 i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1044 I40E_GLPRT_PTC127L(hw->port),
1045 pf->stat_offsets_loaded,
1046 &osd->tx_size_127, &nsd->tx_size_127);
1047 i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1048 I40E_GLPRT_PTC255L(hw->port),
1049 pf->stat_offsets_loaded,
1050 &osd->tx_size_255, &nsd->tx_size_255);
1051 i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1052 I40E_GLPRT_PTC511L(hw->port),
1053 pf->stat_offsets_loaded,
1054 &osd->tx_size_511, &nsd->tx_size_511);
1055 i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1056 I40E_GLPRT_PTC1023L(hw->port),
1057 pf->stat_offsets_loaded,
1058 &osd->tx_size_1023, &nsd->tx_size_1023);
1059 i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1060 I40E_GLPRT_PTC1522L(hw->port),
1061 pf->stat_offsets_loaded,
1062 &osd->tx_size_1522, &nsd->tx_size_1522);
1063 i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1064 I40E_GLPRT_PTC9522L(hw->port),
1065 pf->stat_offsets_loaded,
1066 &osd->tx_size_big, &nsd->tx_size_big);
1068 i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1069 pf->stat_offsets_loaded,
1070 &osd->rx_undersize, &nsd->rx_undersize);
1071 i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1072 pf->stat_offsets_loaded,
1073 &osd->rx_fragments, &nsd->rx_fragments);
1074 i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1075 pf->stat_offsets_loaded,
1076 &osd->rx_oversize, &nsd->rx_oversize);
1077 i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1078 pf->stat_offsets_loaded,
1079 &osd->rx_jabber, &nsd->rx_jabber);
1081 /* FDIR stats */
1082 i40e_stat_update32(hw,
1083 I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1084 pf->stat_offsets_loaded,
1085 &osd->fd_atr_match, &nsd->fd_atr_match);
1086 i40e_stat_update32(hw,
1087 I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1088 pf->stat_offsets_loaded,
1089 &osd->fd_sb_match, &nsd->fd_sb_match);
1090 i40e_stat_update32(hw,
1091 I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1092 pf->stat_offsets_loaded,
1093 &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1095 val = rd32(hw, I40E_PRTPM_EEE_STAT);
1096 nsd->tx_lpi_status =
1097 (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1098 I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1099 nsd->rx_lpi_status =
1100 (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1101 I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1102 i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1103 pf->stat_offsets_loaded,
1104 &osd->tx_lpi_count, &nsd->tx_lpi_count);
1105 i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1106 pf->stat_offsets_loaded,
1107 &osd->rx_lpi_count, &nsd->rx_lpi_count);
1109 if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1110 !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1111 nsd->fd_sb_status = true;
1112 else
1113 nsd->fd_sb_status = false;
1115 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1116 !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1117 nsd->fd_atr_status = true;
1118 else
1119 nsd->fd_atr_status = false;
1121 pf->stat_offsets_loaded = true;
1125 * i40e_update_stats - Update the various statistics counters.
1126 * @vsi: the VSI to be updated
1128 * Update the various stats for this VSI and its related entities.
1130 void i40e_update_stats(struct i40e_vsi *vsi)
1132 struct i40e_pf *pf = vsi->back;
1134 if (vsi == pf->vsi[pf->lan_vsi])
1135 i40e_update_pf_stats(pf);
1137 i40e_update_vsi_stats(vsi);
1138 #ifdef I40E_FCOE
1139 i40e_update_fcoe_stats(vsi);
1140 #endif
1144 * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1145 * @vsi: the VSI to be searched
1146 * @macaddr: the MAC address
1147 * @vlan: the vlan
1148 * @is_vf: make sure its a VF filter, else doesn't matter
1149 * @is_netdev: make sure its a netdev filter, else doesn't matter
1151 * Returns ptr to the filter object or NULL
1153 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1154 u8 *macaddr, s16 vlan,
1155 bool is_vf, bool is_netdev)
1157 struct i40e_mac_filter *f;
1159 if (!vsi || !macaddr)
1160 return NULL;
1162 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1163 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1164 (vlan == f->vlan) &&
1165 (!is_vf || f->is_vf) &&
1166 (!is_netdev || f->is_netdev))
1167 return f;
1169 return NULL;
1173 * i40e_find_mac - Find a mac addr in the macvlan filters list
1174 * @vsi: the VSI to be searched
1175 * @macaddr: the MAC address we are searching for
1176 * @is_vf: make sure its a VF filter, else doesn't matter
1177 * @is_netdev: make sure its a netdev filter, else doesn't matter
1179 * Returns the first filter with the provided MAC address or NULL if
1180 * MAC address was not found
1182 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1183 bool is_vf, bool is_netdev)
1185 struct i40e_mac_filter *f;
1187 if (!vsi || !macaddr)
1188 return NULL;
1190 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1191 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1192 (!is_vf || f->is_vf) &&
1193 (!is_netdev || f->is_netdev))
1194 return f;
1196 return NULL;
1200 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1201 * @vsi: the VSI to be searched
1203 * Returns true if VSI is in vlan mode or false otherwise
1205 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1207 struct i40e_mac_filter *f;
1209 /* Only -1 for all the filters denotes not in vlan mode
1210 * so we have to go through all the list in order to make sure
1212 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1213 if (f->vlan >= 0 || vsi->info.pvid)
1214 return true;
1217 return false;
1221 * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1222 * @vsi: the VSI to be searched
1223 * @macaddr: the mac address to be filtered
1224 * @is_vf: true if it is a VF
1225 * @is_netdev: true if it is a netdev
1227 * Goes through all the macvlan filters and adds a
1228 * macvlan filter for each unique vlan that already exists
1230 * Returns first filter found on success, else NULL
1232 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1233 bool is_vf, bool is_netdev)
1235 struct i40e_mac_filter *f;
1237 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1238 if (vsi->info.pvid)
1239 f->vlan = le16_to_cpu(vsi->info.pvid);
1240 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1241 is_vf, is_netdev)) {
1242 if (!i40e_add_filter(vsi, macaddr, f->vlan,
1243 is_vf, is_netdev))
1244 return NULL;
1248 return list_first_entry_or_null(&vsi->mac_filter_list,
1249 struct i40e_mac_filter, list);
1253 * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1254 * @vsi: the VSI to be searched
1255 * @macaddr: the mac address to be removed
1256 * @is_vf: true if it is a VF
1257 * @is_netdev: true if it is a netdev
1259 * Removes a given MAC address from a VSI, regardless of VLAN
1261 * Returns 0 for success, or error
1263 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1264 bool is_vf, bool is_netdev)
1266 struct i40e_mac_filter *f = NULL;
1267 int changed = 0;
1269 WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1270 "Missing mac_filter_list_lock\n");
1271 list_for_each_entry(f, &vsi->mac_filter_list, list) {
1272 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1273 (is_vf == f->is_vf) &&
1274 (is_netdev == f->is_netdev)) {
1275 f->counter--;
1276 changed = 1;
1277 if (f->counter == 0)
1278 f->state = I40E_FILTER_REMOVE;
1281 if (changed) {
1282 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1283 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1284 return 0;
1286 return -ENOENT;
1290 * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1291 * @vsi: the PF Main VSI - inappropriate for any other VSI
1292 * @macaddr: the MAC address
1294 * Remove whatever filter the firmware set up so the driver can manage
1295 * its own filtering intelligently.
1297 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1299 struct i40e_aqc_remove_macvlan_element_data element;
1300 struct i40e_pf *pf = vsi->back;
1302 /* Only appropriate for the PF main VSI */
1303 if (vsi->type != I40E_VSI_MAIN)
1304 return;
1306 memset(&element, 0, sizeof(element));
1307 ether_addr_copy(element.mac_addr, macaddr);
1308 element.vlan_tag = 0;
1309 /* Ignore error returns, some firmware does it this way... */
1310 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1311 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1313 memset(&element, 0, sizeof(element));
1314 ether_addr_copy(element.mac_addr, macaddr);
1315 element.vlan_tag = 0;
1316 /* ...and some firmware does it this way. */
1317 element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1318 I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1319 i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1323 * i40e_add_filter - Add a mac/vlan filter to the VSI
1324 * @vsi: the VSI to be searched
1325 * @macaddr: the MAC address
1326 * @vlan: the vlan
1327 * @is_vf: make sure its a VF filter, else doesn't matter
1328 * @is_netdev: make sure its a netdev filter, else doesn't matter
1330 * Returns ptr to the filter object or NULL when no memory available.
1332 * NOTE: This function is expected to be called with mac_filter_list_lock
1333 * being held.
1335 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1336 u8 *macaddr, s16 vlan,
1337 bool is_vf, bool is_netdev)
1339 struct i40e_mac_filter *f;
1340 int changed = false;
1342 if (!vsi || !macaddr)
1343 return NULL;
1345 /* Do not allow broadcast filter to be added since broadcast filter
1346 * is added as part of add VSI for any newly created VSI except
1347 * FDIR VSI
1349 if (is_broadcast_ether_addr(macaddr))
1350 return NULL;
1352 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1353 if (!f) {
1354 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1355 if (!f)
1356 goto add_filter_out;
1358 ether_addr_copy(f->macaddr, macaddr);
1359 f->vlan = vlan;
1360 /* If we're in overflow promisc mode, set the state directly
1361 * to failed, so we don't bother to try sending the filter
1362 * to the hardware.
1364 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))
1365 f->state = I40E_FILTER_FAILED;
1366 else
1367 f->state = I40E_FILTER_NEW;
1368 changed = true;
1369 INIT_LIST_HEAD(&f->list);
1370 list_add_tail(&f->list, &vsi->mac_filter_list);
1373 /* increment counter and add a new flag if needed */
1374 if (is_vf) {
1375 if (!f->is_vf) {
1376 f->is_vf = true;
1377 f->counter++;
1379 } else if (is_netdev) {
1380 if (!f->is_netdev) {
1381 f->is_netdev = true;
1382 f->counter++;
1384 } else {
1385 f->counter++;
1388 if (changed) {
1389 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1390 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1393 add_filter_out:
1394 return f;
1398 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1399 * @vsi: the VSI to be searched
1400 * @macaddr: the MAC address
1401 * @vlan: the vlan
1402 * @is_vf: make sure it's a VF filter, else doesn't matter
1403 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1405 * NOTE: This function is expected to be called with mac_filter_list_lock
1406 * being held.
1407 * ANOTHER NOTE: This function MUST be called from within the context of
1408 * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1409 * instead of list_for_each_entry().
1411 void i40e_del_filter(struct i40e_vsi *vsi,
1412 u8 *macaddr, s16 vlan,
1413 bool is_vf, bool is_netdev)
1415 struct i40e_mac_filter *f;
1417 if (!vsi || !macaddr)
1418 return;
1420 f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1421 if (!f || f->counter == 0)
1422 return;
1424 if (is_vf) {
1425 if (f->is_vf) {
1426 f->is_vf = false;
1427 f->counter--;
1429 } else if (is_netdev) {
1430 if (f->is_netdev) {
1431 f->is_netdev = false;
1432 f->counter--;
1434 } else {
1435 /* make sure we don't remove a filter in use by VF or netdev */
1436 int min_f = 0;
1438 min_f += (f->is_vf ? 1 : 0);
1439 min_f += (f->is_netdev ? 1 : 0);
1441 if (f->counter > min_f)
1442 f->counter--;
1445 /* counter == 0 tells sync_filters_subtask to
1446 * remove the filter from the firmware's list
1448 if (f->counter == 0) {
1449 if ((f->state == I40E_FILTER_FAILED) ||
1450 (f->state == I40E_FILTER_NEW)) {
1451 /* this one never got added by the FW. Just remove it,
1452 * no need to sync anything.
1454 list_del(&f->list);
1455 kfree(f);
1456 } else {
1457 f->state = I40E_FILTER_REMOVE;
1458 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1459 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1465 * i40e_set_mac - NDO callback to set mac address
1466 * @netdev: network interface device structure
1467 * @p: pointer to an address structure
1469 * Returns 0 on success, negative on failure
1471 #ifdef I40E_FCOE
1472 int i40e_set_mac(struct net_device *netdev, void *p)
1473 #else
1474 static int i40e_set_mac(struct net_device *netdev, void *p)
1475 #endif
1477 struct i40e_netdev_priv *np = netdev_priv(netdev);
1478 struct i40e_vsi *vsi = np->vsi;
1479 struct i40e_pf *pf = vsi->back;
1480 struct i40e_hw *hw = &pf->hw;
1481 struct sockaddr *addr = p;
1483 if (!is_valid_ether_addr(addr->sa_data))
1484 return -EADDRNOTAVAIL;
1486 if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1487 netdev_info(netdev, "already using mac address %pM\n",
1488 addr->sa_data);
1489 return 0;
1492 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1493 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1494 return -EADDRNOTAVAIL;
1496 if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1497 netdev_info(netdev, "returning to hw mac address %pM\n",
1498 hw->mac.addr);
1499 else
1500 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1502 spin_lock_bh(&vsi->mac_filter_list_lock);
1503 i40e_del_mac_all_vlan(vsi, netdev->dev_addr, false, true);
1504 i40e_put_mac_in_vlan(vsi, addr->sa_data, false, true);
1505 spin_unlock_bh(&vsi->mac_filter_list_lock);
1506 ether_addr_copy(netdev->dev_addr, addr->sa_data);
1507 if (vsi->type == I40E_VSI_MAIN) {
1508 i40e_status ret;
1510 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1511 I40E_AQC_WRITE_TYPE_LAA_WOL,
1512 addr->sa_data, NULL);
1513 if (ret)
1514 netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1515 i40e_stat_str(hw, ret),
1516 i40e_aq_str(hw, hw->aq.asq_last_status));
1519 /* schedule our worker thread which will take care of
1520 * applying the new filter changes
1522 i40e_service_event_schedule(vsi->back);
1523 return 0;
1527 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1528 * @vsi: the VSI being setup
1529 * @ctxt: VSI context structure
1530 * @enabled_tc: Enabled TCs bitmap
1531 * @is_add: True if called before Add VSI
1533 * Setup VSI queue mapping for enabled traffic classes.
1535 #ifdef I40E_FCOE
1536 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1537 struct i40e_vsi_context *ctxt,
1538 u8 enabled_tc,
1539 bool is_add)
1540 #else
1541 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1542 struct i40e_vsi_context *ctxt,
1543 u8 enabled_tc,
1544 bool is_add)
1545 #endif
1547 struct i40e_pf *pf = vsi->back;
1548 u16 sections = 0;
1549 u8 netdev_tc = 0;
1550 u16 numtc = 0;
1551 u16 qcount;
1552 u8 offset;
1553 u16 qmap;
1554 int i;
1555 u16 num_tc_qps = 0;
1557 sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1558 offset = 0;
1560 if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1561 /* Find numtc from enabled TC bitmap */
1562 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1563 if (enabled_tc & BIT(i)) /* TC is enabled */
1564 numtc++;
1566 if (!numtc) {
1567 dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1568 numtc = 1;
1570 } else {
1571 /* At least TC0 is enabled in case of non-DCB case */
1572 numtc = 1;
1575 vsi->tc_config.numtc = numtc;
1576 vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1577 /* Number of queues per enabled TC */
1578 qcount = vsi->alloc_queue_pairs;
1580 num_tc_qps = qcount / numtc;
1581 num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1583 /* Setup queue offset/count for all TCs for given VSI */
1584 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1585 /* See if the given TC is enabled for the given VSI */
1586 if (vsi->tc_config.enabled_tc & BIT(i)) {
1587 /* TC is enabled */
1588 int pow, num_qps;
1590 switch (vsi->type) {
1591 case I40E_VSI_MAIN:
1592 qcount = min_t(int, pf->alloc_rss_size,
1593 num_tc_qps);
1594 break;
1595 #ifdef I40E_FCOE
1596 case I40E_VSI_FCOE:
1597 qcount = num_tc_qps;
1598 break;
1599 #endif
1600 case I40E_VSI_FDIR:
1601 case I40E_VSI_SRIOV:
1602 case I40E_VSI_VMDQ2:
1603 default:
1604 qcount = num_tc_qps;
1605 WARN_ON(i != 0);
1606 break;
1608 vsi->tc_config.tc_info[i].qoffset = offset;
1609 vsi->tc_config.tc_info[i].qcount = qcount;
1611 /* find the next higher power-of-2 of num queue pairs */
1612 num_qps = qcount;
1613 pow = 0;
1614 while (num_qps && (BIT_ULL(pow) < qcount)) {
1615 pow++;
1616 num_qps >>= 1;
1619 vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1620 qmap =
1621 (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1622 (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1624 offset += qcount;
1625 } else {
1626 /* TC is not enabled so set the offset to
1627 * default queue and allocate one queue
1628 * for the given TC.
1630 vsi->tc_config.tc_info[i].qoffset = 0;
1631 vsi->tc_config.tc_info[i].qcount = 1;
1632 vsi->tc_config.tc_info[i].netdev_tc = 0;
1634 qmap = 0;
1636 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1639 /* Set actual Tx/Rx queue pairs */
1640 vsi->num_queue_pairs = offset;
1641 if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1642 if (vsi->req_queue_pairs > 0)
1643 vsi->num_queue_pairs = vsi->req_queue_pairs;
1644 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1645 vsi->num_queue_pairs = pf->num_lan_msix;
1648 /* Scheduler section valid can only be set for ADD VSI */
1649 if (is_add) {
1650 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1652 ctxt->info.up_enable_bits = enabled_tc;
1654 if (vsi->type == I40E_VSI_SRIOV) {
1655 ctxt->info.mapping_flags |=
1656 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1657 for (i = 0; i < vsi->num_queue_pairs; i++)
1658 ctxt->info.queue_mapping[i] =
1659 cpu_to_le16(vsi->base_queue + i);
1660 } else {
1661 ctxt->info.mapping_flags |=
1662 cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1663 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1665 ctxt->info.valid_sections |= cpu_to_le16(sections);
1669 * i40e_set_rx_mode - NDO callback to set the netdev filters
1670 * @netdev: network interface device structure
1672 #ifdef I40E_FCOE
1673 void i40e_set_rx_mode(struct net_device *netdev)
1674 #else
1675 static void i40e_set_rx_mode(struct net_device *netdev)
1676 #endif
1678 struct i40e_netdev_priv *np = netdev_priv(netdev);
1679 struct i40e_mac_filter *f, *ftmp;
1680 struct i40e_vsi *vsi = np->vsi;
1681 struct netdev_hw_addr *uca;
1682 struct netdev_hw_addr *mca;
1683 struct netdev_hw_addr *ha;
1685 spin_lock_bh(&vsi->mac_filter_list_lock);
1687 /* add addr if not already in the filter list */
1688 netdev_for_each_uc_addr(uca, netdev) {
1689 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1690 if (i40e_is_vsi_in_vlan(vsi))
1691 i40e_put_mac_in_vlan(vsi, uca->addr,
1692 false, true);
1693 else
1694 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1695 false, true);
1699 netdev_for_each_mc_addr(mca, netdev) {
1700 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1701 if (i40e_is_vsi_in_vlan(vsi))
1702 i40e_put_mac_in_vlan(vsi, mca->addr,
1703 false, true);
1704 else
1705 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1706 false, true);
1710 /* remove filter if not in netdev list */
1711 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1713 if (!f->is_netdev)
1714 continue;
1716 netdev_for_each_mc_addr(mca, netdev)
1717 if (ether_addr_equal(mca->addr, f->macaddr))
1718 goto bottom_of_search_loop;
1720 netdev_for_each_uc_addr(uca, netdev)
1721 if (ether_addr_equal(uca->addr, f->macaddr))
1722 goto bottom_of_search_loop;
1724 for_each_dev_addr(netdev, ha)
1725 if (ether_addr_equal(ha->addr, f->macaddr))
1726 goto bottom_of_search_loop;
1728 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1729 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1731 bottom_of_search_loop:
1732 continue;
1734 spin_unlock_bh(&vsi->mac_filter_list_lock);
1736 /* check for other flag changes */
1737 if (vsi->current_netdev_flags != vsi->netdev->flags) {
1738 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1739 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1742 /* schedule our worker thread which will take care of
1743 * applying the new filter changes
1745 i40e_service_event_schedule(vsi->back);
1749 * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1750 * @vsi: pointer to vsi struct
1751 * @from: Pointer to list which contains MAC filter entries - changes to
1752 * those entries needs to be undone.
1754 * MAC filter entries from list were slated to be removed from device.
1756 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1757 struct list_head *from)
1759 struct i40e_mac_filter *f, *ftmp;
1761 list_for_each_entry_safe(f, ftmp, from, list) {
1762 /* Move the element back into MAC filter list*/
1763 list_move_tail(&f->list, &vsi->mac_filter_list);
1768 * i40e_update_filter_state - Update filter state based on return data
1769 * from firmware
1770 * @count: Number of filters added
1771 * @add_list: return data from fw
1772 * @head: pointer to first filter in current batch
1773 * @aq_err: status from fw
1775 * MAC filter entries from list were slated to be added to device. Returns
1776 * number of successful filters. Note that 0 does NOT mean success!
1778 static int
1779 i40e_update_filter_state(int count,
1780 struct i40e_aqc_add_macvlan_element_data *add_list,
1781 struct i40e_mac_filter *add_head, int aq_err)
1783 int retval = 0;
1784 int i;
1787 if (!aq_err) {
1788 retval = count;
1789 /* Everything's good, mark all filters active. */
1790 for (i = 0; i < count ; i++) {
1791 add_head->state = I40E_FILTER_ACTIVE;
1792 add_head = list_next_entry(add_head, list);
1794 } else if (aq_err == I40E_AQ_RC_ENOSPC) {
1795 /* Device ran out of filter space. Check the return value
1796 * for each filter to see which ones are active.
1798 for (i = 0; i < count ; i++) {
1799 if (add_list[i].match_method ==
1800 I40E_AQC_MM_ERR_NO_RES) {
1801 add_head->state = I40E_FILTER_FAILED;
1802 } else {
1803 add_head->state = I40E_FILTER_ACTIVE;
1804 retval++;
1806 add_head = list_next_entry(add_head, list);
1808 } else {
1809 /* Some other horrible thing happened, fail all filters */
1810 retval = 0;
1811 for (i = 0; i < count ; i++) {
1812 add_head->state = I40E_FILTER_FAILED;
1813 add_head = list_next_entry(add_head, list);
1816 return retval;
1820 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1821 * @vsi: ptr to the VSI
1823 * Push any outstanding VSI filter changes through the AdminQ.
1825 * Returns 0 or error value
1827 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1829 struct i40e_mac_filter *f, *ftmp, *add_head = NULL;
1830 struct list_head tmp_add_list, tmp_del_list;
1831 struct i40e_hw *hw = &vsi->back->hw;
1832 bool promisc_changed = false;
1833 char vsi_name[16] = "PF";
1834 int filter_list_len = 0;
1835 u32 changed_flags = 0;
1836 i40e_status aq_ret = 0;
1837 int retval = 0;
1838 struct i40e_pf *pf;
1839 int num_add = 0;
1840 int num_del = 0;
1841 int aq_err = 0;
1842 u16 cmd_flags;
1843 int list_size;
1844 int fcnt;
1846 /* empty array typed pointers, kcalloc later */
1847 struct i40e_aqc_add_macvlan_element_data *add_list;
1848 struct i40e_aqc_remove_macvlan_element_data *del_list;
1850 while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1851 usleep_range(1000, 2000);
1852 pf = vsi->back;
1854 if (vsi->netdev) {
1855 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1856 vsi->current_netdev_flags = vsi->netdev->flags;
1859 INIT_LIST_HEAD(&tmp_add_list);
1860 INIT_LIST_HEAD(&tmp_del_list);
1862 if (vsi->type == I40E_VSI_SRIOV)
1863 snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
1864 else if (vsi->type != I40E_VSI_MAIN)
1865 snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
1867 if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1868 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1870 spin_lock_bh(&vsi->mac_filter_list_lock);
1871 /* Create a list of filters to delete. */
1872 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1873 if (f->state == I40E_FILTER_REMOVE) {
1874 WARN_ON(f->counter != 0);
1875 /* Move the element into temporary del_list */
1876 list_move_tail(&f->list, &tmp_del_list);
1877 vsi->active_filters--;
1879 if (f->state == I40E_FILTER_NEW) {
1880 WARN_ON(f->counter == 0);
1881 /* Move the element into temporary add_list */
1882 list_move_tail(&f->list, &tmp_add_list);
1885 spin_unlock_bh(&vsi->mac_filter_list_lock);
1888 /* Now process 'del_list' outside the lock */
1889 if (!list_empty(&tmp_del_list)) {
1890 filter_list_len = hw->aq.asq_buf_size /
1891 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1892 list_size = filter_list_len *
1893 sizeof(struct i40e_aqc_remove_macvlan_element_data);
1894 del_list = kzalloc(list_size, GFP_ATOMIC);
1895 if (!del_list) {
1896 /* Undo VSI's MAC filter entry element updates */
1897 spin_lock_bh(&vsi->mac_filter_list_lock);
1898 i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1899 spin_unlock_bh(&vsi->mac_filter_list_lock);
1900 retval = -ENOMEM;
1901 goto out;
1904 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1905 cmd_flags = 0;
1907 /* add to delete list */
1908 ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1909 if (f->vlan == I40E_VLAN_ANY) {
1910 del_list[num_del].vlan_tag = 0;
1911 cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1912 } else {
1913 del_list[num_del].vlan_tag =
1914 cpu_to_le16((u16)(f->vlan));
1917 cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1918 del_list[num_del].flags = cmd_flags;
1919 num_del++;
1921 /* flush a full buffer */
1922 if (num_del == filter_list_len) {
1923 aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid,
1924 del_list,
1925 num_del, NULL);
1926 aq_err = hw->aq.asq_last_status;
1927 num_del = 0;
1928 memset(del_list, 0, list_size);
1930 /* Explicitly ignore and do not report when
1931 * firmware returns ENOENT.
1933 if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1934 retval = -EIO;
1935 dev_info(&pf->pdev->dev,
1936 "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1937 vsi_name,
1938 i40e_stat_str(hw, aq_ret),
1939 i40e_aq_str(hw, aq_err));
1942 /* Release memory for MAC filter entries which were
1943 * synced up with HW.
1945 list_del(&f->list);
1946 kfree(f);
1949 if (num_del) {
1950 aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, del_list,
1951 num_del, NULL);
1952 aq_err = hw->aq.asq_last_status;
1953 num_del = 0;
1955 /* Explicitly ignore and do not report when firmware
1956 * returns ENOENT.
1958 if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1959 retval = -EIO;
1960 dev_info(&pf->pdev->dev,
1961 "ignoring delete macvlan error on %s, err %s aq_err %s\n",
1962 vsi_name,
1963 i40e_stat_str(hw, aq_ret),
1964 i40e_aq_str(hw, aq_err));
1968 kfree(del_list);
1969 del_list = NULL;
1972 if (!list_empty(&tmp_add_list)) {
1973 /* Do all the adds now. */
1974 filter_list_len = hw->aq.asq_buf_size /
1975 sizeof(struct i40e_aqc_add_macvlan_element_data);
1976 list_size = filter_list_len *
1977 sizeof(struct i40e_aqc_add_macvlan_element_data);
1978 add_list = kzalloc(list_size, GFP_ATOMIC);
1979 if (!add_list) {
1980 retval = -ENOMEM;
1981 goto out;
1983 num_add = 0;
1984 list_for_each_entry(f, &tmp_add_list, list) {
1985 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1986 &vsi->state)) {
1987 f->state = I40E_FILTER_FAILED;
1988 continue;
1990 /* add to add array */
1991 if (num_add == 0)
1992 add_head = f;
1993 cmd_flags = 0;
1994 ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
1995 if (f->vlan == I40E_VLAN_ANY) {
1996 add_list[num_add].vlan_tag = 0;
1997 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1998 } else {
1999 add_list[num_add].vlan_tag =
2000 cpu_to_le16((u16)(f->vlan));
2002 add_list[num_add].queue_number = 0;
2003 cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2004 add_list[num_add].flags = cpu_to_le16(cmd_flags);
2005 num_add++;
2007 /* flush a full buffer */
2008 if (num_add == filter_list_len) {
2009 aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2010 add_list, num_add,
2011 NULL);
2012 aq_err = hw->aq.asq_last_status;
2013 fcnt = i40e_update_filter_state(num_add,
2014 add_list,
2015 add_head,
2016 aq_ret);
2017 vsi->active_filters += fcnt;
2019 if (fcnt != num_add) {
2020 promisc_changed = true;
2021 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2022 &vsi->state);
2023 vsi->promisc_threshold =
2024 (vsi->active_filters * 3) / 4;
2025 dev_warn(&pf->pdev->dev,
2026 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2027 i40e_aq_str(hw, aq_err),
2028 vsi_name);
2030 memset(add_list, 0, list_size);
2031 num_add = 0;
2034 if (num_add) {
2035 aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2036 add_list, num_add, NULL);
2037 aq_err = hw->aq.asq_last_status;
2038 fcnt = i40e_update_filter_state(num_add, add_list,
2039 add_head, aq_ret);
2040 vsi->active_filters += fcnt;
2041 if (fcnt != num_add) {
2042 promisc_changed = true;
2043 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2044 &vsi->state);
2045 vsi->promisc_threshold =
2046 (vsi->active_filters * 3) / 4;
2047 dev_warn(&pf->pdev->dev,
2048 "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2049 i40e_aq_str(hw, aq_err), vsi_name);
2052 /* Now move all of the filters from the temp add list back to
2053 * the VSI's list.
2055 spin_lock_bh(&vsi->mac_filter_list_lock);
2056 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2057 list_move_tail(&f->list, &vsi->mac_filter_list);
2059 spin_unlock_bh(&vsi->mac_filter_list_lock);
2060 kfree(add_list);
2061 add_list = NULL;
2064 /* Check to see if we can drop out of overflow promiscuous mode. */
2065 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state) &&
2066 (vsi->active_filters < vsi->promisc_threshold)) {
2067 int failed_count = 0;
2068 /* See if we have any failed filters. We can't drop out of
2069 * promiscuous until these have all been deleted.
2071 spin_lock_bh(&vsi->mac_filter_list_lock);
2072 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2073 if (f->state == I40E_FILTER_FAILED)
2074 failed_count++;
2076 spin_unlock_bh(&vsi->mac_filter_list_lock);
2077 if (!failed_count) {
2078 dev_info(&pf->pdev->dev,
2079 "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2080 vsi_name);
2081 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2082 promisc_changed = true;
2083 vsi->promisc_threshold = 0;
2087 /* if the VF is not trusted do not do promisc */
2088 if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2089 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2090 goto out;
2093 /* check for changes in promiscuous modes */
2094 if (changed_flags & IFF_ALLMULTI) {
2095 bool cur_multipromisc;
2097 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2098 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2099 vsi->seid,
2100 cur_multipromisc,
2101 NULL);
2102 if (aq_ret) {
2103 retval = i40e_aq_rc_to_posix(aq_ret,
2104 hw->aq.asq_last_status);
2105 dev_info(&pf->pdev->dev,
2106 "set multi promisc failed on %s, err %s aq_err %s\n",
2107 vsi_name,
2108 i40e_stat_str(hw, aq_ret),
2109 i40e_aq_str(hw, hw->aq.asq_last_status));
2112 if ((changed_flags & IFF_PROMISC) ||
2113 (promisc_changed &&
2114 test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))) {
2115 bool cur_promisc;
2117 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2118 test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2119 &vsi->state));
2120 if ((vsi->type == I40E_VSI_MAIN) &&
2121 (pf->lan_veb != I40E_NO_VEB) &&
2122 !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2123 /* set defport ON for Main VSI instead of true promisc
2124 * this way we will get all unicast/multicast and VLAN
2125 * promisc behavior but will not get VF or VMDq traffic
2126 * replicated on the Main VSI.
2128 if (pf->cur_promisc != cur_promisc) {
2129 pf->cur_promisc = cur_promisc;
2130 if (cur_promisc)
2131 aq_ret =
2132 i40e_aq_set_default_vsi(hw,
2133 vsi->seid,
2134 NULL);
2135 else
2136 aq_ret =
2137 i40e_aq_clear_default_vsi(hw,
2138 vsi->seid,
2139 NULL);
2140 if (aq_ret) {
2141 retval = i40e_aq_rc_to_posix(aq_ret,
2142 hw->aq.asq_last_status);
2143 dev_info(&pf->pdev->dev,
2144 "Set default VSI failed on %s, err %s, aq_err %s\n",
2145 vsi_name,
2146 i40e_stat_str(hw, aq_ret),
2147 i40e_aq_str(hw,
2148 hw->aq.asq_last_status));
2151 } else {
2152 aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2154 vsi->seid,
2155 cur_promisc, NULL,
2156 true);
2157 if (aq_ret) {
2158 retval =
2159 i40e_aq_rc_to_posix(aq_ret,
2160 hw->aq.asq_last_status);
2161 dev_info(&pf->pdev->dev,
2162 "set unicast promisc failed on %s, err %s, aq_err %s\n",
2163 vsi_name,
2164 i40e_stat_str(hw, aq_ret),
2165 i40e_aq_str(hw,
2166 hw->aq.asq_last_status));
2168 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2170 vsi->seid,
2171 cur_promisc, NULL);
2172 if (aq_ret) {
2173 retval =
2174 i40e_aq_rc_to_posix(aq_ret,
2175 hw->aq.asq_last_status);
2176 dev_info(&pf->pdev->dev,
2177 "set multicast promisc failed on %s, err %s, aq_err %s\n",
2178 vsi_name,
2179 i40e_stat_str(hw, aq_ret),
2180 i40e_aq_str(hw,
2181 hw->aq.asq_last_status));
2184 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2185 vsi->seid,
2186 cur_promisc, NULL);
2187 if (aq_ret) {
2188 retval = i40e_aq_rc_to_posix(aq_ret,
2189 pf->hw.aq.asq_last_status);
2190 dev_info(&pf->pdev->dev,
2191 "set brdcast promisc failed, err %s, aq_err %s\n",
2192 i40e_stat_str(hw, aq_ret),
2193 i40e_aq_str(hw,
2194 hw->aq.asq_last_status));
2197 out:
2198 /* if something went wrong then set the changed flag so we try again */
2199 if (retval)
2200 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2202 clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2203 return retval;
2207 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2208 * @pf: board private structure
2210 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2212 int v;
2214 if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2215 return;
2216 pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2218 for (v = 0; v < pf->num_alloc_vsi; v++) {
2219 if (pf->vsi[v] &&
2220 (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2221 int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2223 if (ret) {
2224 /* come back and try again later */
2225 pf->flags |= I40E_FLAG_FILTER_SYNC;
2226 break;
2233 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2234 * @netdev: network interface device structure
2235 * @new_mtu: new value for maximum frame size
2237 * Returns 0 on success, negative on failure
2239 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2241 struct i40e_netdev_priv *np = netdev_priv(netdev);
2242 int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2243 struct i40e_vsi *vsi = np->vsi;
2245 /* MTU < 68 is an error and causes problems on some kernels */
2246 if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2247 return -EINVAL;
2249 netdev_info(netdev, "changing MTU from %d to %d\n",
2250 netdev->mtu, new_mtu);
2251 netdev->mtu = new_mtu;
2252 if (netif_running(netdev))
2253 i40e_vsi_reinit_locked(vsi);
2254 i40e_notify_client_of_l2_param_changes(vsi);
2255 return 0;
2259 * i40e_ioctl - Access the hwtstamp interface
2260 * @netdev: network interface device structure
2261 * @ifr: interface request data
2262 * @cmd: ioctl command
2264 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2266 struct i40e_netdev_priv *np = netdev_priv(netdev);
2267 struct i40e_pf *pf = np->vsi->back;
2269 switch (cmd) {
2270 case SIOCGHWTSTAMP:
2271 return i40e_ptp_get_ts_config(pf, ifr);
2272 case SIOCSHWTSTAMP:
2273 return i40e_ptp_set_ts_config(pf, ifr);
2274 default:
2275 return -EOPNOTSUPP;
2280 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2281 * @vsi: the vsi being adjusted
2283 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2285 struct i40e_vsi_context ctxt;
2286 i40e_status ret;
2288 if ((vsi->info.valid_sections &
2289 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2290 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2291 return; /* already enabled */
2293 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2294 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2295 I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2297 ctxt.seid = vsi->seid;
2298 ctxt.info = vsi->info;
2299 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2300 if (ret) {
2301 dev_info(&vsi->back->pdev->dev,
2302 "update vlan stripping failed, err %s aq_err %s\n",
2303 i40e_stat_str(&vsi->back->hw, ret),
2304 i40e_aq_str(&vsi->back->hw,
2305 vsi->back->hw.aq.asq_last_status));
2310 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2311 * @vsi: the vsi being adjusted
2313 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2315 struct i40e_vsi_context ctxt;
2316 i40e_status ret;
2318 if ((vsi->info.valid_sections &
2319 cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2320 ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2321 I40E_AQ_VSI_PVLAN_EMOD_MASK))
2322 return; /* already disabled */
2324 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2325 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2326 I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2328 ctxt.seid = vsi->seid;
2329 ctxt.info = vsi->info;
2330 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2331 if (ret) {
2332 dev_info(&vsi->back->pdev->dev,
2333 "update vlan stripping failed, err %s aq_err %s\n",
2334 i40e_stat_str(&vsi->back->hw, ret),
2335 i40e_aq_str(&vsi->back->hw,
2336 vsi->back->hw.aq.asq_last_status));
2341 * i40e_vlan_rx_register - Setup or shutdown vlan offload
2342 * @netdev: network interface to be adjusted
2343 * @features: netdev features to test if VLAN offload is enabled or not
2345 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2347 struct i40e_netdev_priv *np = netdev_priv(netdev);
2348 struct i40e_vsi *vsi = np->vsi;
2350 if (features & NETIF_F_HW_VLAN_CTAG_RX)
2351 i40e_vlan_stripping_enable(vsi);
2352 else
2353 i40e_vlan_stripping_disable(vsi);
2357 * i40e_vsi_add_vlan - Add vsi membership for given vlan
2358 * @vsi: the vsi being configured
2359 * @vid: vlan id to be added (0 = untagged only , -1 = any)
2361 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2363 struct i40e_mac_filter *f, *ftmp, *add_f;
2364 bool is_netdev, is_vf;
2366 is_vf = (vsi->type == I40E_VSI_SRIOV);
2367 is_netdev = !!(vsi->netdev);
2369 /* Locked once because all functions invoked below iterates list*/
2370 spin_lock_bh(&vsi->mac_filter_list_lock);
2372 if (is_netdev) {
2373 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2374 is_vf, is_netdev);
2375 if (!add_f) {
2376 dev_info(&vsi->back->pdev->dev,
2377 "Could not add vlan filter %d for %pM\n",
2378 vid, vsi->netdev->dev_addr);
2379 spin_unlock_bh(&vsi->mac_filter_list_lock);
2380 return -ENOMEM;
2384 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2385 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2386 if (!add_f) {
2387 dev_info(&vsi->back->pdev->dev,
2388 "Could not add vlan filter %d for %pM\n",
2389 vid, f->macaddr);
2390 spin_unlock_bh(&vsi->mac_filter_list_lock);
2391 return -ENOMEM;
2395 /* Now if we add a vlan tag, make sure to check if it is the first
2396 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2397 * with 0, so we now accept untagged and specified tagged traffic
2398 * (and not all tags along with untagged)
2400 if (vid > 0) {
2401 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2402 I40E_VLAN_ANY,
2403 is_vf, is_netdev)) {
2404 i40e_del_filter(vsi, vsi->netdev->dev_addr,
2405 I40E_VLAN_ANY, is_vf, is_netdev);
2406 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2407 is_vf, is_netdev);
2408 if (!add_f) {
2409 dev_info(&vsi->back->pdev->dev,
2410 "Could not add filter 0 for %pM\n",
2411 vsi->netdev->dev_addr);
2412 spin_unlock_bh(&vsi->mac_filter_list_lock);
2413 return -ENOMEM;
2418 /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2419 if (vid > 0 && !vsi->info.pvid) {
2420 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2421 if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2422 is_vf, is_netdev))
2423 continue;
2424 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2425 is_vf, is_netdev);
2426 add_f = i40e_add_filter(vsi, f->macaddr,
2427 0, is_vf, is_netdev);
2428 if (!add_f) {
2429 dev_info(&vsi->back->pdev->dev,
2430 "Could not add filter 0 for %pM\n",
2431 f->macaddr);
2432 spin_unlock_bh(&vsi->mac_filter_list_lock);
2433 return -ENOMEM;
2438 spin_unlock_bh(&vsi->mac_filter_list_lock);
2440 /* schedule our worker thread which will take care of
2441 * applying the new filter changes
2443 i40e_service_event_schedule(vsi->back);
2444 return 0;
2448 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2449 * @vsi: the vsi being configured
2450 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2452 * Return: 0 on success or negative otherwise
2454 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2456 struct net_device *netdev = vsi->netdev;
2457 struct i40e_mac_filter *f, *ftmp, *add_f;
2458 bool is_vf, is_netdev;
2459 int filter_count = 0;
2461 is_vf = (vsi->type == I40E_VSI_SRIOV);
2462 is_netdev = !!(netdev);
2464 /* Locked once because all functions invoked below iterates list */
2465 spin_lock_bh(&vsi->mac_filter_list_lock);
2467 if (is_netdev)
2468 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2470 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
2471 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2473 /* go through all the filters for this VSI and if there is only
2474 * vid == 0 it means there are no other filters, so vid 0 must
2475 * be replaced with -1. This signifies that we should from now
2476 * on accept any traffic (with any tag present, or untagged)
2478 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2479 if (is_netdev) {
2480 if (f->vlan &&
2481 ether_addr_equal(netdev->dev_addr, f->macaddr))
2482 filter_count++;
2485 if (f->vlan)
2486 filter_count++;
2489 if (!filter_count && is_netdev) {
2490 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2491 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2492 is_vf, is_netdev);
2493 if (!f) {
2494 dev_info(&vsi->back->pdev->dev,
2495 "Could not add filter %d for %pM\n",
2496 I40E_VLAN_ANY, netdev->dev_addr);
2497 spin_unlock_bh(&vsi->mac_filter_list_lock);
2498 return -ENOMEM;
2502 if (!filter_count) {
2503 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2504 i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2505 add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2506 is_vf, is_netdev);
2507 if (!add_f) {
2508 dev_info(&vsi->back->pdev->dev,
2509 "Could not add filter %d for %pM\n",
2510 I40E_VLAN_ANY, f->macaddr);
2511 spin_unlock_bh(&vsi->mac_filter_list_lock);
2512 return -ENOMEM;
2517 spin_unlock_bh(&vsi->mac_filter_list_lock);
2519 /* schedule our worker thread which will take care of
2520 * applying the new filter changes
2522 i40e_service_event_schedule(vsi->back);
2523 return 0;
2527 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2528 * @netdev: network interface to be adjusted
2529 * @vid: vlan id to be added
2531 * net_device_ops implementation for adding vlan ids
2533 #ifdef I40E_FCOE
2534 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2535 __always_unused __be16 proto, u16 vid)
2536 #else
2537 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2538 __always_unused __be16 proto, u16 vid)
2539 #endif
2541 struct i40e_netdev_priv *np = netdev_priv(netdev);
2542 struct i40e_vsi *vsi = np->vsi;
2543 int ret = 0;
2545 if (vid > 4095)
2546 return -EINVAL;
2548 /* If the network stack called us with vid = 0 then
2549 * it is asking to receive priority tagged packets with
2550 * vlan id 0. Our HW receives them by default when configured
2551 * to receive untagged packets so there is no need to add an
2552 * extra filter for vlan 0 tagged packets.
2554 if (vid)
2555 ret = i40e_vsi_add_vlan(vsi, vid);
2557 if (!ret && (vid < VLAN_N_VID))
2558 set_bit(vid, vsi->active_vlans);
2560 return ret;
2564 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2565 * @netdev: network interface to be adjusted
2566 * @vid: vlan id to be removed
2568 * net_device_ops implementation for removing vlan ids
2570 #ifdef I40E_FCOE
2571 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2572 __always_unused __be16 proto, u16 vid)
2573 #else
2574 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2575 __always_unused __be16 proto, u16 vid)
2576 #endif
2578 struct i40e_netdev_priv *np = netdev_priv(netdev);
2579 struct i40e_vsi *vsi = np->vsi;
2581 /* return code is ignored as there is nothing a user
2582 * can do about failure to remove and a log message was
2583 * already printed from the other function
2585 i40e_vsi_kill_vlan(vsi, vid);
2587 clear_bit(vid, vsi->active_vlans);
2589 return 0;
2593 * i40e_macaddr_init - explicitly write the mac address filters
2595 * @vsi: pointer to the vsi
2596 * @macaddr: the MAC address
2598 * This is needed when the macaddr has been obtained by other
2599 * means than the default, e.g., from Open Firmware or IDPROM.
2600 * Returns 0 on success, negative on failure
2602 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2604 int ret;
2605 struct i40e_aqc_add_macvlan_element_data element;
2607 ret = i40e_aq_mac_address_write(&vsi->back->hw,
2608 I40E_AQC_WRITE_TYPE_LAA_WOL,
2609 macaddr, NULL);
2610 if (ret) {
2611 dev_info(&vsi->back->pdev->dev,
2612 "Addr change for VSI failed: %d\n", ret);
2613 return -EADDRNOTAVAIL;
2616 memset(&element, 0, sizeof(element));
2617 ether_addr_copy(element.mac_addr, macaddr);
2618 element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2619 ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2620 if (ret) {
2621 dev_info(&vsi->back->pdev->dev,
2622 "add filter failed err %s aq_err %s\n",
2623 i40e_stat_str(&vsi->back->hw, ret),
2624 i40e_aq_str(&vsi->back->hw,
2625 vsi->back->hw.aq.asq_last_status));
2627 return ret;
2631 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2632 * @vsi: the vsi being brought back up
2634 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2636 u16 vid;
2638 if (!vsi->netdev)
2639 return;
2641 i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2643 for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2644 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2645 vid);
2649 * i40e_vsi_add_pvid - Add pvid for the VSI
2650 * @vsi: the vsi being adjusted
2651 * @vid: the vlan id to set as a PVID
2653 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2655 struct i40e_vsi_context ctxt;
2656 i40e_status ret;
2658 vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2659 vsi->info.pvid = cpu_to_le16(vid);
2660 vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2661 I40E_AQ_VSI_PVLAN_INSERT_PVID |
2662 I40E_AQ_VSI_PVLAN_EMOD_STR;
2664 ctxt.seid = vsi->seid;
2665 ctxt.info = vsi->info;
2666 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2667 if (ret) {
2668 dev_info(&vsi->back->pdev->dev,
2669 "add pvid failed, err %s aq_err %s\n",
2670 i40e_stat_str(&vsi->back->hw, ret),
2671 i40e_aq_str(&vsi->back->hw,
2672 vsi->back->hw.aq.asq_last_status));
2673 return -ENOENT;
2676 return 0;
2680 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2681 * @vsi: the vsi being adjusted
2683 * Just use the vlan_rx_register() service to put it back to normal
2685 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2687 i40e_vlan_stripping_disable(vsi);
2689 vsi->info.pvid = 0;
2693 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2694 * @vsi: ptr to the VSI
2696 * If this function returns with an error, then it's possible one or
2697 * more of the rings is populated (while the rest are not). It is the
2698 * callers duty to clean those orphaned rings.
2700 * Return 0 on success, negative on failure
2702 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2704 int i, err = 0;
2706 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2707 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2709 return err;
2713 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2714 * @vsi: ptr to the VSI
2716 * Free VSI's transmit software resources
2718 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2720 int i;
2722 if (!vsi->tx_rings)
2723 return;
2725 for (i = 0; i < vsi->num_queue_pairs; i++)
2726 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2727 i40e_free_tx_resources(vsi->tx_rings[i]);
2731 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2732 * @vsi: ptr to the VSI
2734 * If this function returns with an error, then it's possible one or
2735 * more of the rings is populated (while the rest are not). It is the
2736 * callers duty to clean those orphaned rings.
2738 * Return 0 on success, negative on failure
2740 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2742 int i, err = 0;
2744 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2745 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2746 #ifdef I40E_FCOE
2747 i40e_fcoe_setup_ddp_resources(vsi);
2748 #endif
2749 return err;
2753 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2754 * @vsi: ptr to the VSI
2756 * Free all receive software resources
2758 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2760 int i;
2762 if (!vsi->rx_rings)
2763 return;
2765 for (i = 0; i < vsi->num_queue_pairs; i++)
2766 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2767 i40e_free_rx_resources(vsi->rx_rings[i]);
2768 #ifdef I40E_FCOE
2769 i40e_fcoe_free_ddp_resources(vsi);
2770 #endif
2774 * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2775 * @ring: The Tx ring to configure
2777 * This enables/disables XPS for a given Tx descriptor ring
2778 * based on the TCs enabled for the VSI that ring belongs to.
2780 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2782 struct i40e_vsi *vsi = ring->vsi;
2783 cpumask_var_t mask;
2785 if (!ring->q_vector || !ring->netdev)
2786 return;
2788 /* Single TC mode enable XPS */
2789 if (vsi->tc_config.numtc <= 1) {
2790 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2791 netif_set_xps_queue(ring->netdev,
2792 &ring->q_vector->affinity_mask,
2793 ring->queue_index);
2794 } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2795 /* Disable XPS to allow selection based on TC */
2796 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2797 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2798 free_cpumask_var(mask);
2801 /* schedule our worker thread which will take care of
2802 * applying the new filter changes
2804 i40e_service_event_schedule(vsi->back);
2808 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2809 * @ring: The Tx ring to configure
2811 * Configure the Tx descriptor ring in the HMC context.
2813 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2815 struct i40e_vsi *vsi = ring->vsi;
2816 u16 pf_q = vsi->base_queue + ring->queue_index;
2817 struct i40e_hw *hw = &vsi->back->hw;
2818 struct i40e_hmc_obj_txq tx_ctx;
2819 i40e_status err = 0;
2820 u32 qtx_ctl = 0;
2822 /* some ATR related tx ring init */
2823 if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2824 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2825 ring->atr_count = 0;
2826 } else {
2827 ring->atr_sample_rate = 0;
2830 /* configure XPS */
2831 i40e_config_xps_tx_ring(ring);
2833 /* clear the context structure first */
2834 memset(&tx_ctx, 0, sizeof(tx_ctx));
2836 tx_ctx.new_context = 1;
2837 tx_ctx.base = (ring->dma / 128);
2838 tx_ctx.qlen = ring->count;
2839 tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2840 I40E_FLAG_FD_ATR_ENABLED));
2841 #ifdef I40E_FCOE
2842 tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2843 #endif
2844 tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2845 /* FDIR VSI tx ring can still use RS bit and writebacks */
2846 if (vsi->type != I40E_VSI_FDIR)
2847 tx_ctx.head_wb_ena = 1;
2848 tx_ctx.head_wb_addr = ring->dma +
2849 (ring->count * sizeof(struct i40e_tx_desc));
2851 /* As part of VSI creation/update, FW allocates certain
2852 * Tx arbitration queue sets for each TC enabled for
2853 * the VSI. The FW returns the handles to these queue
2854 * sets as part of the response buffer to Add VSI,
2855 * Update VSI, etc. AQ commands. It is expected that
2856 * these queue set handles be associated with the Tx
2857 * queues by the driver as part of the TX queue context
2858 * initialization. This has to be done regardless of
2859 * DCB as by default everything is mapped to TC0.
2861 tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2862 tx_ctx.rdylist_act = 0;
2864 /* clear the context in the HMC */
2865 err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2866 if (err) {
2867 dev_info(&vsi->back->pdev->dev,
2868 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2869 ring->queue_index, pf_q, err);
2870 return -ENOMEM;
2873 /* set the context in the HMC */
2874 err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2875 if (err) {
2876 dev_info(&vsi->back->pdev->dev,
2877 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2878 ring->queue_index, pf_q, err);
2879 return -ENOMEM;
2882 /* Now associate this queue with this PCI function */
2883 if (vsi->type == I40E_VSI_VMDQ2) {
2884 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2885 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2886 I40E_QTX_CTL_VFVM_INDX_MASK;
2887 } else {
2888 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2891 qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2892 I40E_QTX_CTL_PF_INDX_MASK);
2893 wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2894 i40e_flush(hw);
2896 /* cache tail off for easier writes later */
2897 ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2899 return 0;
2903 * i40e_configure_rx_ring - Configure a receive ring context
2904 * @ring: The Rx ring to configure
2906 * Configure the Rx descriptor ring in the HMC context.
2908 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2910 struct i40e_vsi *vsi = ring->vsi;
2911 u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2912 u16 pf_q = vsi->base_queue + ring->queue_index;
2913 struct i40e_hw *hw = &vsi->back->hw;
2914 struct i40e_hmc_obj_rxq rx_ctx;
2915 i40e_status err = 0;
2917 ring->state = 0;
2919 /* clear the context structure first */
2920 memset(&rx_ctx, 0, sizeof(rx_ctx));
2922 ring->rx_buf_len = vsi->rx_buf_len;
2924 rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2926 rx_ctx.base = (ring->dma / 128);
2927 rx_ctx.qlen = ring->count;
2929 /* use 32 byte descriptors */
2930 rx_ctx.dsize = 1;
2932 /* descriptor type is always zero
2933 * rx_ctx.dtype = 0;
2935 rx_ctx.hsplit_0 = 0;
2937 rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
2938 if (hw->revision_id == 0)
2939 rx_ctx.lrxqthresh = 0;
2940 else
2941 rx_ctx.lrxqthresh = 2;
2942 rx_ctx.crcstrip = 1;
2943 rx_ctx.l2tsel = 1;
2944 /* this controls whether VLAN is stripped from inner headers */
2945 rx_ctx.showiv = 0;
2946 #ifdef I40E_FCOE
2947 rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2948 #endif
2949 /* set the prefena field to 1 because the manual says to */
2950 rx_ctx.prefena = 1;
2952 /* clear the context in the HMC */
2953 err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2954 if (err) {
2955 dev_info(&vsi->back->pdev->dev,
2956 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2957 ring->queue_index, pf_q, err);
2958 return -ENOMEM;
2961 /* set the context in the HMC */
2962 err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2963 if (err) {
2964 dev_info(&vsi->back->pdev->dev,
2965 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2966 ring->queue_index, pf_q, err);
2967 return -ENOMEM;
2970 /* cache tail for quicker writes, and clear the reg before use */
2971 ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2972 writel(0, ring->tail);
2974 i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2976 return 0;
2980 * i40e_vsi_configure_tx - Configure the VSI for Tx
2981 * @vsi: VSI structure describing this set of rings and resources
2983 * Configure the Tx VSI for operation.
2985 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2987 int err = 0;
2988 u16 i;
2990 for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2991 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2993 return err;
2997 * i40e_vsi_configure_rx - Configure the VSI for Rx
2998 * @vsi: the VSI being configured
3000 * Configure the Rx VSI for operation.
3002 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3004 int err = 0;
3005 u16 i;
3007 if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
3008 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
3009 + ETH_FCS_LEN + VLAN_HLEN;
3010 else
3011 vsi->max_frame = I40E_RXBUFFER_2048;
3013 vsi->rx_buf_len = I40E_RXBUFFER_2048;
3015 #ifdef I40E_FCOE
3016 /* setup rx buffer for FCoE */
3017 if ((vsi->type == I40E_VSI_FCOE) &&
3018 (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
3019 vsi->rx_buf_len = I40E_RXBUFFER_3072;
3020 vsi->max_frame = I40E_RXBUFFER_3072;
3023 #endif /* I40E_FCOE */
3024 /* round up for the chip's needs */
3025 vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
3026 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3028 /* set up individual rings */
3029 for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3030 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3032 return err;
3036 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3037 * @vsi: ptr to the VSI
3039 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3041 struct i40e_ring *tx_ring, *rx_ring;
3042 u16 qoffset, qcount;
3043 int i, n;
3045 if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3046 /* Reset the TC information */
3047 for (i = 0; i < vsi->num_queue_pairs; i++) {
3048 rx_ring = vsi->rx_rings[i];
3049 tx_ring = vsi->tx_rings[i];
3050 rx_ring->dcb_tc = 0;
3051 tx_ring->dcb_tc = 0;
3055 for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3056 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3057 continue;
3059 qoffset = vsi->tc_config.tc_info[n].qoffset;
3060 qcount = vsi->tc_config.tc_info[n].qcount;
3061 for (i = qoffset; i < (qoffset + qcount); i++) {
3062 rx_ring = vsi->rx_rings[i];
3063 tx_ring = vsi->tx_rings[i];
3064 rx_ring->dcb_tc = n;
3065 tx_ring->dcb_tc = n;
3071 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3072 * @vsi: ptr to the VSI
3074 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3076 struct i40e_pf *pf = vsi->back;
3077 int err;
3079 if (vsi->netdev)
3080 i40e_set_rx_mode(vsi->netdev);
3082 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3083 err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3084 if (err) {
3085 dev_warn(&pf->pdev->dev,
3086 "could not set up macaddr; err %d\n", err);
3092 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3093 * @vsi: Pointer to the targeted VSI
3095 * This function replays the hlist on the hw where all the SB Flow Director
3096 * filters were saved.
3098 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3100 struct i40e_fdir_filter *filter;
3101 struct i40e_pf *pf = vsi->back;
3102 struct hlist_node *node;
3104 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3105 return;
3107 hlist_for_each_entry_safe(filter, node,
3108 &pf->fdir_filter_list, fdir_node) {
3109 i40e_add_del_fdir(vsi, filter, true);
3114 * i40e_vsi_configure - Set up the VSI for action
3115 * @vsi: the VSI being configured
3117 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3119 int err;
3121 i40e_set_vsi_rx_mode(vsi);
3122 i40e_restore_vlan(vsi);
3123 i40e_vsi_config_dcb_rings(vsi);
3124 err = i40e_vsi_configure_tx(vsi);
3125 if (!err)
3126 err = i40e_vsi_configure_rx(vsi);
3128 return err;
3132 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3133 * @vsi: the VSI being configured
3135 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3137 struct i40e_pf *pf = vsi->back;
3138 struct i40e_hw *hw = &pf->hw;
3139 u16 vector;
3140 int i, q;
3141 u32 qp;
3143 /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3144 * and PFINT_LNKLSTn registers, e.g.:
3145 * PFINT_ITRn[0..n-1] gets msix-1..msix-n (qpair interrupts)
3147 qp = vsi->base_queue;
3148 vector = vsi->base_vector;
3149 for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3150 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3152 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3153 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3154 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3155 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3156 q_vector->rx.itr);
3157 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3158 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3159 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3160 q_vector->tx.itr);
3161 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3162 INTRL_USEC_TO_REG(vsi->int_rate_limit));
3164 /* Linked list for the queuepairs assigned to this vector */
3165 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3166 for (q = 0; q < q_vector->num_ringpairs; q++) {
3167 u32 val;
3169 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3170 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3171 (vector << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3172 (qp << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3173 (I40E_QUEUE_TYPE_TX
3174 << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3176 wr32(hw, I40E_QINT_RQCTL(qp), val);
3178 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3179 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3180 (vector << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3181 ((qp+1) << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3182 (I40E_QUEUE_TYPE_RX
3183 << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3185 /* Terminate the linked list */
3186 if (q == (q_vector->num_ringpairs - 1))
3187 val |= (I40E_QUEUE_END_OF_LIST
3188 << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3190 wr32(hw, I40E_QINT_TQCTL(qp), val);
3191 qp++;
3195 i40e_flush(hw);
3199 * i40e_enable_misc_int_causes - enable the non-queue interrupts
3200 * @hw: ptr to the hardware info
3202 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3204 struct i40e_hw *hw = &pf->hw;
3205 u32 val;
3207 /* clear things first */
3208 wr32(hw, I40E_PFINT_ICR0_ENA, 0); /* disable all */
3209 rd32(hw, I40E_PFINT_ICR0); /* read to clear */
3211 val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK |
3212 I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK |
3213 I40E_PFINT_ICR0_ENA_GRST_MASK |
3214 I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3215 I40E_PFINT_ICR0_ENA_GPIO_MASK |
3216 I40E_PFINT_ICR0_ENA_HMC_ERR_MASK |
3217 I40E_PFINT_ICR0_ENA_VFLR_MASK |
3218 I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3220 if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3221 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3223 if (pf->flags & I40E_FLAG_PTP)
3224 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3226 wr32(hw, I40E_PFINT_ICR0_ENA, val);
3228 /* SW_ITR_IDX = 0, but don't change INTENA */
3229 wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3230 I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3232 /* OTHER_ITR_IDX = 0 */
3233 wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3237 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3238 * @vsi: the VSI being configured
3240 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3242 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3243 struct i40e_pf *pf = vsi->back;
3244 struct i40e_hw *hw = &pf->hw;
3245 u32 val;
3247 /* set the ITR configuration */
3248 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3249 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3250 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3251 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3252 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3253 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3254 wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3256 i40e_enable_misc_int_causes(pf);
3258 /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3259 wr32(hw, I40E_PFINT_LNKLST0, 0);
3261 /* Associate the queue pair to the vector and enable the queue int */
3262 val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3263 (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3264 (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3266 wr32(hw, I40E_QINT_RQCTL(0), val);
3268 val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3269 (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3270 (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3272 wr32(hw, I40E_QINT_TQCTL(0), val);
3273 i40e_flush(hw);
3277 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3278 * @pf: board private structure
3280 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3282 struct i40e_hw *hw = &pf->hw;
3284 wr32(hw, I40E_PFINT_DYN_CTL0,
3285 I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3286 i40e_flush(hw);
3290 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3291 * @pf: board private structure
3292 * @clearpba: true when all pending interrupt events should be cleared
3294 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3296 struct i40e_hw *hw = &pf->hw;
3297 u32 val;
3299 val = I40E_PFINT_DYN_CTL0_INTENA_MASK |
3300 (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3301 (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3303 wr32(hw, I40E_PFINT_DYN_CTL0, val);
3304 i40e_flush(hw);
3308 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3309 * @irq: interrupt number
3310 * @data: pointer to a q_vector
3312 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3314 struct i40e_q_vector *q_vector = data;
3316 if (!q_vector->tx.ring && !q_vector->rx.ring)
3317 return IRQ_HANDLED;
3319 napi_schedule_irqoff(&q_vector->napi);
3321 return IRQ_HANDLED;
3325 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3326 * @vsi: the VSI being configured
3327 * @basename: name for the vector
3329 * Allocates MSI-X vectors and requests interrupts from the kernel.
3331 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3333 int q_vectors = vsi->num_q_vectors;
3334 struct i40e_pf *pf = vsi->back;
3335 int base = vsi->base_vector;
3336 int rx_int_idx = 0;
3337 int tx_int_idx = 0;
3338 int vector, err;
3340 for (vector = 0; vector < q_vectors; vector++) {
3341 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3343 if (q_vector->tx.ring && q_vector->rx.ring) {
3344 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3345 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3346 tx_int_idx++;
3347 } else if (q_vector->rx.ring) {
3348 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3349 "%s-%s-%d", basename, "rx", rx_int_idx++);
3350 } else if (q_vector->tx.ring) {
3351 snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3352 "%s-%s-%d", basename, "tx", tx_int_idx++);
3353 } else {
3354 /* skip this unused q_vector */
3355 continue;
3357 err = request_irq(pf->msix_entries[base + vector].vector,
3358 vsi->irq_handler,
3360 q_vector->name,
3361 q_vector);
3362 if (err) {
3363 dev_info(&pf->pdev->dev,
3364 "MSIX request_irq failed, error: %d\n", err);
3365 goto free_queue_irqs;
3367 /* assign the mask for this irq */
3368 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3369 &q_vector->affinity_mask);
3372 vsi->irqs_ready = true;
3373 return 0;
3375 free_queue_irqs:
3376 while (vector) {
3377 vector--;
3378 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3379 NULL);
3380 free_irq(pf->msix_entries[base + vector].vector,
3381 &(vsi->q_vectors[vector]));
3383 return err;
3387 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3388 * @vsi: the VSI being un-configured
3390 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3392 struct i40e_pf *pf = vsi->back;
3393 struct i40e_hw *hw = &pf->hw;
3394 int base = vsi->base_vector;
3395 int i;
3397 for (i = 0; i < vsi->num_queue_pairs; i++) {
3398 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3399 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3402 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3403 for (i = vsi->base_vector;
3404 i < (vsi->num_q_vectors + vsi->base_vector); i++)
3405 wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3407 i40e_flush(hw);
3408 for (i = 0; i < vsi->num_q_vectors; i++)
3409 synchronize_irq(pf->msix_entries[i + base].vector);
3410 } else {
3411 /* Legacy and MSI mode - this stops all interrupt handling */
3412 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3413 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3414 i40e_flush(hw);
3415 synchronize_irq(pf->pdev->irq);
3420 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3421 * @vsi: the VSI being configured
3423 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3425 struct i40e_pf *pf = vsi->back;
3426 int i;
3428 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3429 for (i = 0; i < vsi->num_q_vectors; i++)
3430 i40e_irq_dynamic_enable(vsi, i);
3431 } else {
3432 i40e_irq_dynamic_enable_icr0(pf, true);
3435 i40e_flush(&pf->hw);
3436 return 0;
3440 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3441 * @pf: board private structure
3443 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3445 /* Disable ICR 0 */
3446 wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3447 i40e_flush(&pf->hw);
3451 * i40e_intr - MSI/Legacy and non-queue interrupt handler
3452 * @irq: interrupt number
3453 * @data: pointer to a q_vector
3455 * This is the handler used for all MSI/Legacy interrupts, and deals
3456 * with both queue and non-queue interrupts. This is also used in
3457 * MSIX mode to handle the non-queue interrupts.
3459 static irqreturn_t i40e_intr(int irq, void *data)
3461 struct i40e_pf *pf = (struct i40e_pf *)data;
3462 struct i40e_hw *hw = &pf->hw;
3463 irqreturn_t ret = IRQ_NONE;
3464 u32 icr0, icr0_remaining;
3465 u32 val, ena_mask;
3467 icr0 = rd32(hw, I40E_PFINT_ICR0);
3468 ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3470 /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3471 if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3472 goto enable_intr;
3474 /* if interrupt but no bits showing, must be SWINT */
3475 if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3476 (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3477 pf->sw_int_count++;
3479 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3480 (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3481 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3482 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3483 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3486 /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3487 if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3488 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3489 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3491 /* We do not have a way to disarm Queue causes while leaving
3492 * interrupt enabled for all other causes, ideally
3493 * interrupt should be disabled while we are in NAPI but
3494 * this is not a performance path and napi_schedule()
3495 * can deal with rescheduling.
3497 if (!test_bit(__I40E_DOWN, &pf->state))
3498 napi_schedule_irqoff(&q_vector->napi);
3501 if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3502 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3503 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3504 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3507 if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3508 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3509 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3512 if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3513 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3514 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3517 if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3518 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3519 set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3520 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3521 val = rd32(hw, I40E_GLGEN_RSTAT);
3522 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3523 >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3524 if (val == I40E_RESET_CORER) {
3525 pf->corer_count++;
3526 } else if (val == I40E_RESET_GLOBR) {
3527 pf->globr_count++;
3528 } else if (val == I40E_RESET_EMPR) {
3529 pf->empr_count++;
3530 set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3534 if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3535 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3536 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3537 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3538 rd32(hw, I40E_PFHMC_ERRORINFO),
3539 rd32(hw, I40E_PFHMC_ERRORDATA));
3542 if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3543 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3545 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3546 icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3547 i40e_ptp_tx_hwtstamp(pf);
3551 /* If a critical error is pending we have no choice but to reset the
3552 * device.
3553 * Report and mask out any remaining unexpected interrupts.
3555 icr0_remaining = icr0 & ena_mask;
3556 if (icr0_remaining) {
3557 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3558 icr0_remaining);
3559 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3560 (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3561 (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3562 dev_info(&pf->pdev->dev, "device will be reset\n");
3563 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3564 i40e_service_event_schedule(pf);
3566 ena_mask &= ~icr0_remaining;
3568 ret = IRQ_HANDLED;
3570 enable_intr:
3571 /* re-enable interrupt causes */
3572 wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3573 if (!test_bit(__I40E_DOWN, &pf->state)) {
3574 i40e_service_event_schedule(pf);
3575 i40e_irq_dynamic_enable_icr0(pf, false);
3578 return ret;
3582 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3583 * @tx_ring: tx ring to clean
3584 * @budget: how many cleans we're allowed
3586 * Returns true if there's any budget left (e.g. the clean is finished)
3588 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3590 struct i40e_vsi *vsi = tx_ring->vsi;
3591 u16 i = tx_ring->next_to_clean;
3592 struct i40e_tx_buffer *tx_buf;
3593 struct i40e_tx_desc *tx_desc;
3595 tx_buf = &tx_ring->tx_bi[i];
3596 tx_desc = I40E_TX_DESC(tx_ring, i);
3597 i -= tx_ring->count;
3599 do {
3600 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3602 /* if next_to_watch is not set then there is no work pending */
3603 if (!eop_desc)
3604 break;
3606 /* prevent any other reads prior to eop_desc */
3607 smp_rmb();
3609 /* if the descriptor isn't done, no work yet to do */
3610 if (!(eop_desc->cmd_type_offset_bsz &
3611 cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3612 break;
3614 /* clear next_to_watch to prevent false hangs */
3615 tx_buf->next_to_watch = NULL;
3617 tx_desc->buffer_addr = 0;
3618 tx_desc->cmd_type_offset_bsz = 0;
3619 /* move past filter desc */
3620 tx_buf++;
3621 tx_desc++;
3622 i++;
3623 if (unlikely(!i)) {
3624 i -= tx_ring->count;
3625 tx_buf = tx_ring->tx_bi;
3626 tx_desc = I40E_TX_DESC(tx_ring, 0);
3628 /* unmap skb header data */
3629 dma_unmap_single(tx_ring->dev,
3630 dma_unmap_addr(tx_buf, dma),
3631 dma_unmap_len(tx_buf, len),
3632 DMA_TO_DEVICE);
3633 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3634 kfree(tx_buf->raw_buf);
3636 tx_buf->raw_buf = NULL;
3637 tx_buf->tx_flags = 0;
3638 tx_buf->next_to_watch = NULL;
3639 dma_unmap_len_set(tx_buf, len, 0);
3640 tx_desc->buffer_addr = 0;
3641 tx_desc->cmd_type_offset_bsz = 0;
3643 /* move us past the eop_desc for start of next FD desc */
3644 tx_buf++;
3645 tx_desc++;
3646 i++;
3647 if (unlikely(!i)) {
3648 i -= tx_ring->count;
3649 tx_buf = tx_ring->tx_bi;
3650 tx_desc = I40E_TX_DESC(tx_ring, 0);
3653 /* update budget accounting */
3654 budget--;
3655 } while (likely(budget));
3657 i += tx_ring->count;
3658 tx_ring->next_to_clean = i;
3660 if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3661 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3663 return budget > 0;
3667 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3668 * @irq: interrupt number
3669 * @data: pointer to a q_vector
3671 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3673 struct i40e_q_vector *q_vector = data;
3674 struct i40e_vsi *vsi;
3676 if (!q_vector->tx.ring)
3677 return IRQ_HANDLED;
3679 vsi = q_vector->tx.ring->vsi;
3680 i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3682 return IRQ_HANDLED;
3686 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3687 * @vsi: the VSI being configured
3688 * @v_idx: vector index
3689 * @qp_idx: queue pair index
3691 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3693 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3694 struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3695 struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3697 tx_ring->q_vector = q_vector;
3698 tx_ring->next = q_vector->tx.ring;
3699 q_vector->tx.ring = tx_ring;
3700 q_vector->tx.count++;
3702 rx_ring->q_vector = q_vector;
3703 rx_ring->next = q_vector->rx.ring;
3704 q_vector->rx.ring = rx_ring;
3705 q_vector->rx.count++;
3709 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3710 * @vsi: the VSI being configured
3712 * This function maps descriptor rings to the queue-specific vectors
3713 * we were allotted through the MSI-X enabling code. Ideally, we'd have
3714 * one vector per queue pair, but on a constrained vector budget, we
3715 * group the queue pairs as "efficiently" as possible.
3717 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3719 int qp_remaining = vsi->num_queue_pairs;
3720 int q_vectors = vsi->num_q_vectors;
3721 int num_ringpairs;
3722 int v_start = 0;
3723 int qp_idx = 0;
3725 /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3726 * group them so there are multiple queues per vector.
3727 * It is also important to go through all the vectors available to be
3728 * sure that if we don't use all the vectors, that the remaining vectors
3729 * are cleared. This is especially important when decreasing the
3730 * number of queues in use.
3732 for (; v_start < q_vectors; v_start++) {
3733 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3735 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3737 q_vector->num_ringpairs = num_ringpairs;
3739 q_vector->rx.count = 0;
3740 q_vector->tx.count = 0;
3741 q_vector->rx.ring = NULL;
3742 q_vector->tx.ring = NULL;
3744 while (num_ringpairs--) {
3745 i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3746 qp_idx++;
3747 qp_remaining--;
3753 * i40e_vsi_request_irq - Request IRQ from the OS
3754 * @vsi: the VSI being configured
3755 * @basename: name for the vector
3757 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3759 struct i40e_pf *pf = vsi->back;
3760 int err;
3762 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3763 err = i40e_vsi_request_irq_msix(vsi, basename);
3764 else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3765 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3766 pf->int_name, pf);
3767 else
3768 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3769 pf->int_name, pf);
3771 if (err)
3772 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3774 return err;
3777 #ifdef CONFIG_NET_POLL_CONTROLLER
3779 * i40e_netpoll - A Polling 'interrupt' handler
3780 * @netdev: network interface device structure
3782 * This is used by netconsole to send skbs without having to re-enable
3783 * interrupts. It's not called while the normal interrupt routine is executing.
3785 #ifdef I40E_FCOE
3786 void i40e_netpoll(struct net_device *netdev)
3787 #else
3788 static void i40e_netpoll(struct net_device *netdev)
3789 #endif
3791 struct i40e_netdev_priv *np = netdev_priv(netdev);
3792 struct i40e_vsi *vsi = np->vsi;
3793 struct i40e_pf *pf = vsi->back;
3794 int i;
3796 /* if interface is down do nothing */
3797 if (test_bit(__I40E_DOWN, &vsi->state))
3798 return;
3800 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3801 for (i = 0; i < vsi->num_q_vectors; i++)
3802 i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3803 } else {
3804 i40e_intr(pf->pdev->irq, netdev);
3807 #endif
3810 * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3811 * @pf: the PF being configured
3812 * @pf_q: the PF queue
3813 * @enable: enable or disable state of the queue
3815 * This routine will wait for the given Tx queue of the PF to reach the
3816 * enabled or disabled state.
3817 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3818 * multiple retries; else will return 0 in case of success.
3820 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3822 int i;
3823 u32 tx_reg;
3825 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3826 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3827 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3828 break;
3830 usleep_range(10, 20);
3832 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3833 return -ETIMEDOUT;
3835 return 0;
3839 * i40e_vsi_control_tx - Start or stop a VSI's rings
3840 * @vsi: the VSI being configured
3841 * @enable: start or stop the rings
3843 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3845 struct i40e_pf *pf = vsi->back;
3846 struct i40e_hw *hw = &pf->hw;
3847 int i, j, pf_q, ret = 0;
3848 u32 tx_reg;
3850 pf_q = vsi->base_queue;
3851 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3853 /* warn the TX unit of coming changes */
3854 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3855 if (!enable)
3856 usleep_range(10, 20);
3858 for (j = 0; j < 50; j++) {
3859 tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3860 if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3861 ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3862 break;
3863 usleep_range(1000, 2000);
3865 /* Skip if the queue is already in the requested state */
3866 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3867 continue;
3869 /* turn on/off the queue */
3870 if (enable) {
3871 wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3872 tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3873 } else {
3874 tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3877 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3878 /* No waiting for the Tx queue to disable */
3879 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3880 continue;
3882 /* wait for the change to finish */
3883 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3884 if (ret) {
3885 dev_info(&pf->pdev->dev,
3886 "VSI seid %d Tx ring %d %sable timeout\n",
3887 vsi->seid, pf_q, (enable ? "en" : "dis"));
3888 break;
3892 if (hw->revision_id == 0)
3893 mdelay(50);
3894 return ret;
3898 * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3899 * @pf: the PF being configured
3900 * @pf_q: the PF queue
3901 * @enable: enable or disable state of the queue
3903 * This routine will wait for the given Rx queue of the PF to reach the
3904 * enabled or disabled state.
3905 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3906 * multiple retries; else will return 0 in case of success.
3908 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3910 int i;
3911 u32 rx_reg;
3913 for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3914 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3915 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3916 break;
3918 usleep_range(10, 20);
3920 if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3921 return -ETIMEDOUT;
3923 return 0;
3927 * i40e_vsi_control_rx - Start or stop a VSI's rings
3928 * @vsi: the VSI being configured
3929 * @enable: start or stop the rings
3931 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3933 struct i40e_pf *pf = vsi->back;
3934 struct i40e_hw *hw = &pf->hw;
3935 int i, j, pf_q, ret = 0;
3936 u32 rx_reg;
3938 pf_q = vsi->base_queue;
3939 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3940 for (j = 0; j < 50; j++) {
3941 rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3942 if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3943 ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3944 break;
3945 usleep_range(1000, 2000);
3948 /* Skip if the queue is already in the requested state */
3949 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3950 continue;
3952 /* turn on/off the queue */
3953 if (enable)
3954 rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3955 else
3956 rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3957 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3958 /* No waiting for the Tx queue to disable */
3959 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3960 continue;
3962 /* wait for the change to finish */
3963 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3964 if (ret) {
3965 dev_info(&pf->pdev->dev,
3966 "VSI seid %d Rx ring %d %sable timeout\n",
3967 vsi->seid, pf_q, (enable ? "en" : "dis"));
3968 break;
3972 return ret;
3976 * i40e_vsi_control_rings - Start or stop a VSI's rings
3977 * @vsi: the VSI being configured
3978 * @enable: start or stop the rings
3980 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3982 int ret = 0;
3984 /* do rx first for enable and last for disable */
3985 if (request) {
3986 ret = i40e_vsi_control_rx(vsi, request);
3987 if (ret)
3988 return ret;
3989 ret = i40e_vsi_control_tx(vsi, request);
3990 } else {
3991 /* Ignore return value, we need to shutdown whatever we can */
3992 i40e_vsi_control_tx(vsi, request);
3993 i40e_vsi_control_rx(vsi, request);
3996 return ret;
4000 * i40e_vsi_free_irq - Free the irq association with the OS
4001 * @vsi: the VSI being configured
4003 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4005 struct i40e_pf *pf = vsi->back;
4006 struct i40e_hw *hw = &pf->hw;
4007 int base = vsi->base_vector;
4008 u32 val, qp;
4009 int i;
4011 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4012 if (!vsi->q_vectors)
4013 return;
4015 if (!vsi->irqs_ready)
4016 return;
4018 vsi->irqs_ready = false;
4019 for (i = 0; i < vsi->num_q_vectors; i++) {
4020 u16 vector = i + base;
4022 /* free only the irqs that were actually requested */
4023 if (!vsi->q_vectors[i] ||
4024 !vsi->q_vectors[i]->num_ringpairs)
4025 continue;
4027 /* clear the affinity_mask in the IRQ descriptor */
4028 irq_set_affinity_hint(pf->msix_entries[vector].vector,
4029 NULL);
4030 synchronize_irq(pf->msix_entries[vector].vector);
4031 free_irq(pf->msix_entries[vector].vector,
4032 vsi->q_vectors[i]);
4034 /* Tear down the interrupt queue link list
4036 * We know that they come in pairs and always
4037 * the Rx first, then the Tx. To clear the
4038 * link list, stick the EOL value into the
4039 * next_q field of the registers.
4041 val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4042 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4043 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4044 val |= I40E_QUEUE_END_OF_LIST
4045 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4046 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4048 while (qp != I40E_QUEUE_END_OF_LIST) {
4049 u32 next;
4051 val = rd32(hw, I40E_QINT_RQCTL(qp));
4053 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4054 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4055 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4056 I40E_QINT_RQCTL_INTEVENT_MASK);
4058 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4059 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4061 wr32(hw, I40E_QINT_RQCTL(qp), val);
4063 val = rd32(hw, I40E_QINT_TQCTL(qp));
4065 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4066 >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4068 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4069 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4070 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4071 I40E_QINT_TQCTL_INTEVENT_MASK);
4073 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4074 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4076 wr32(hw, I40E_QINT_TQCTL(qp), val);
4077 qp = next;
4080 } else {
4081 free_irq(pf->pdev->irq, pf);
4083 val = rd32(hw, I40E_PFINT_LNKLST0);
4084 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4085 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4086 val |= I40E_QUEUE_END_OF_LIST
4087 << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4088 wr32(hw, I40E_PFINT_LNKLST0, val);
4090 val = rd32(hw, I40E_QINT_RQCTL(qp));
4091 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK |
4092 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4093 I40E_QINT_RQCTL_CAUSE_ENA_MASK |
4094 I40E_QINT_RQCTL_INTEVENT_MASK);
4096 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4097 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4099 wr32(hw, I40E_QINT_RQCTL(qp), val);
4101 val = rd32(hw, I40E_QINT_TQCTL(qp));
4103 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK |
4104 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4105 I40E_QINT_TQCTL_CAUSE_ENA_MASK |
4106 I40E_QINT_TQCTL_INTEVENT_MASK);
4108 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4109 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4111 wr32(hw, I40E_QINT_TQCTL(qp), val);
4116 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4117 * @vsi: the VSI being configured
4118 * @v_idx: Index of vector to be freed
4120 * This function frees the memory allocated to the q_vector. In addition if
4121 * NAPI is enabled it will delete any references to the NAPI struct prior
4122 * to freeing the q_vector.
4124 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4126 struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4127 struct i40e_ring *ring;
4129 if (!q_vector)
4130 return;
4132 /* disassociate q_vector from rings */
4133 i40e_for_each_ring(ring, q_vector->tx)
4134 ring->q_vector = NULL;
4136 i40e_for_each_ring(ring, q_vector->rx)
4137 ring->q_vector = NULL;
4139 /* only VSI w/ an associated netdev is set up w/ NAPI */
4140 if (vsi->netdev)
4141 netif_napi_del(&q_vector->napi);
4143 vsi->q_vectors[v_idx] = NULL;
4145 kfree_rcu(q_vector, rcu);
4149 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4150 * @vsi: the VSI being un-configured
4152 * This frees the memory allocated to the q_vectors and
4153 * deletes references to the NAPI struct.
4155 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4157 int v_idx;
4159 for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4160 i40e_free_q_vector(vsi, v_idx);
4164 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4165 * @pf: board private structure
4167 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4169 /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4170 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4171 pci_disable_msix(pf->pdev);
4172 kfree(pf->msix_entries);
4173 pf->msix_entries = NULL;
4174 kfree(pf->irq_pile);
4175 pf->irq_pile = NULL;
4176 } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4177 pci_disable_msi(pf->pdev);
4179 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4183 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4184 * @pf: board private structure
4186 * We go through and clear interrupt specific resources and reset the structure
4187 * to pre-load conditions
4189 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4191 int i;
4193 i40e_stop_misc_vector(pf);
4194 if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4195 synchronize_irq(pf->msix_entries[0].vector);
4196 free_irq(pf->msix_entries[0].vector, pf);
4199 i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4200 I40E_IWARP_IRQ_PILE_ID);
4202 i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4203 for (i = 0; i < pf->num_alloc_vsi; i++)
4204 if (pf->vsi[i])
4205 i40e_vsi_free_q_vectors(pf->vsi[i]);
4206 i40e_reset_interrupt_capability(pf);
4210 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4211 * @vsi: the VSI being configured
4213 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4215 int q_idx;
4217 if (!vsi->netdev)
4218 return;
4220 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4221 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4223 if (q_vector->rx.ring || q_vector->tx.ring)
4224 napi_enable(&q_vector->napi);
4229 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4230 * @vsi: the VSI being configured
4232 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4234 int q_idx;
4236 if (!vsi->netdev)
4237 return;
4239 for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++) {
4240 struct i40e_q_vector *q_vector = vsi->q_vectors[q_idx];
4242 if (q_vector->rx.ring || q_vector->tx.ring)
4243 napi_disable(&q_vector->napi);
4248 * i40e_vsi_close - Shut down a VSI
4249 * @vsi: the vsi to be quelled
4251 static void i40e_vsi_close(struct i40e_vsi *vsi)
4253 bool reset = false;
4255 if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4256 i40e_down(vsi);
4257 i40e_vsi_free_irq(vsi);
4258 i40e_vsi_free_tx_resources(vsi);
4259 i40e_vsi_free_rx_resources(vsi);
4260 vsi->current_netdev_flags = 0;
4261 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4262 reset = true;
4263 i40e_notify_client_of_netdev_close(vsi, reset);
4267 * i40e_quiesce_vsi - Pause a given VSI
4268 * @vsi: the VSI being paused
4270 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4272 if (test_bit(__I40E_DOWN, &vsi->state))
4273 return;
4275 /* No need to disable FCoE VSI when Tx suspended */
4276 if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4277 vsi->type == I40E_VSI_FCOE) {
4278 dev_dbg(&vsi->back->pdev->dev,
4279 "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4280 return;
4283 set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4284 if (vsi->netdev && netif_running(vsi->netdev))
4285 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4286 else
4287 i40e_vsi_close(vsi);
4291 * i40e_unquiesce_vsi - Resume a given VSI
4292 * @vsi: the VSI being resumed
4294 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4296 if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4297 return;
4299 clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4300 if (vsi->netdev && netif_running(vsi->netdev))
4301 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4302 else
4303 i40e_vsi_open(vsi); /* this clears the DOWN bit */
4307 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4308 * @pf: the PF
4310 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4312 int v;
4314 for (v = 0; v < pf->num_alloc_vsi; v++) {
4315 if (pf->vsi[v])
4316 i40e_quiesce_vsi(pf->vsi[v]);
4321 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4322 * @pf: the PF
4324 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4326 int v;
4328 for (v = 0; v < pf->num_alloc_vsi; v++) {
4329 if (pf->vsi[v])
4330 i40e_unquiesce_vsi(pf->vsi[v]);
4334 #ifdef CONFIG_I40E_DCB
4336 * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4337 * @vsi: the VSI being configured
4339 * This function waits for the given VSI's queues to be disabled.
4341 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4343 struct i40e_pf *pf = vsi->back;
4344 int i, pf_q, ret;
4346 pf_q = vsi->base_queue;
4347 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4348 /* Check and wait for the disable status of the queue */
4349 ret = i40e_pf_txq_wait(pf, pf_q, false);
4350 if (ret) {
4351 dev_info(&pf->pdev->dev,
4352 "VSI seid %d Tx ring %d disable timeout\n",
4353 vsi->seid, pf_q);
4354 return ret;
4358 pf_q = vsi->base_queue;
4359 for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4360 /* Check and wait for the disable status of the queue */
4361 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4362 if (ret) {
4363 dev_info(&pf->pdev->dev,
4364 "VSI seid %d Rx ring %d disable timeout\n",
4365 vsi->seid, pf_q);
4366 return ret;
4370 return 0;
4374 * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4375 * @pf: the PF
4377 * This function waits for the queues to be in disabled state for all the
4378 * VSIs that are managed by this PF.
4380 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4382 int v, ret = 0;
4384 for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4385 /* No need to wait for FCoE VSI queues */
4386 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4387 ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4388 if (ret)
4389 break;
4393 return ret;
4396 #endif
4399 * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4400 * @q_idx: TX queue number
4401 * @vsi: Pointer to VSI struct
4403 * This function checks specified queue for given VSI. Detects hung condition.
4404 * Sets hung bit since it is two step process. Before next run of service task
4405 * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4406 * hung condition remain unchanged and during subsequent run, this function
4407 * issues SW interrupt to recover from hung condition.
4409 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4411 struct i40e_ring *tx_ring = NULL;
4412 struct i40e_pf *pf;
4413 u32 head, val, tx_pending_hw;
4414 int i;
4416 pf = vsi->back;
4418 /* now that we have an index, find the tx_ring struct */
4419 for (i = 0; i < vsi->num_queue_pairs; i++) {
4420 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4421 if (q_idx == vsi->tx_rings[i]->queue_index) {
4422 tx_ring = vsi->tx_rings[i];
4423 break;
4428 if (!tx_ring)
4429 return;
4431 /* Read interrupt register */
4432 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4433 val = rd32(&pf->hw,
4434 I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4435 tx_ring->vsi->base_vector - 1));
4436 else
4437 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4439 head = i40e_get_head(tx_ring);
4441 tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4443 /* HW is done executing descriptors, updated HEAD write back,
4444 * but SW hasn't processed those descriptors. If interrupt is
4445 * not generated from this point ON, it could result into
4446 * dev_watchdog detecting timeout on those netdev_queue,
4447 * hence proactively trigger SW interrupt.
4449 if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4450 /* NAPI Poll didn't run and clear since it was set */
4451 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4452 &tx_ring->q_vector->hung_detected)) {
4453 netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4454 vsi->seid, q_idx, tx_pending_hw,
4455 tx_ring->next_to_clean, head,
4456 tx_ring->next_to_use,
4457 readl(tx_ring->tail));
4458 netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4459 vsi->seid, q_idx, val);
4460 i40e_force_wb(vsi, tx_ring->q_vector);
4461 } else {
4462 /* First Chance - detected possible hung */
4463 set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4464 &tx_ring->q_vector->hung_detected);
4468 /* This is the case where we have interrupts missing,
4469 * so the tx_pending in HW will most likely be 0, but we
4470 * will have tx_pending in SW since the WB happened but the
4471 * interrupt got lost.
4473 if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4474 (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4475 if (napi_reschedule(&tx_ring->q_vector->napi))
4476 tx_ring->tx_stats.tx_lost_interrupt++;
4481 * i40e_detect_recover_hung - Function to detect and recover hung_queues
4482 * @pf: pointer to PF struct
4484 * LAN VSI has netdev and netdev has TX queues. This function is to check
4485 * each of those TX queues if they are hung, trigger recovery by issuing
4486 * SW interrupt.
4488 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4490 struct net_device *netdev;
4491 struct i40e_vsi *vsi;
4492 int i;
4494 /* Only for LAN VSI */
4495 vsi = pf->vsi[pf->lan_vsi];
4497 if (!vsi)
4498 return;
4500 /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4501 if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4502 test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4503 return;
4505 /* Make sure type is MAIN VSI */
4506 if (vsi->type != I40E_VSI_MAIN)
4507 return;
4509 netdev = vsi->netdev;
4510 if (!netdev)
4511 return;
4513 /* Bail out if netif_carrier is not OK */
4514 if (!netif_carrier_ok(netdev))
4515 return;
4517 /* Go thru' TX queues for netdev */
4518 for (i = 0; i < netdev->num_tx_queues; i++) {
4519 struct netdev_queue *q;
4521 q = netdev_get_tx_queue(netdev, i);
4522 if (q)
4523 i40e_detect_recover_hung_queue(i, vsi);
4528 * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4529 * @pf: pointer to PF
4531 * Get TC map for ISCSI PF type that will include iSCSI TC
4532 * and LAN TC.
4534 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4536 struct i40e_dcb_app_priority_table app;
4537 struct i40e_hw *hw = &pf->hw;
4538 u8 enabled_tc = 1; /* TC0 is always enabled */
4539 u8 tc, i;
4540 /* Get the iSCSI APP TLV */
4541 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4543 for (i = 0; i < dcbcfg->numapps; i++) {
4544 app = dcbcfg->app[i];
4545 if (app.selector == I40E_APP_SEL_TCPIP &&
4546 app.protocolid == I40E_APP_PROTOID_ISCSI) {
4547 tc = dcbcfg->etscfg.prioritytable[app.priority];
4548 enabled_tc |= BIT(tc);
4549 break;
4553 return enabled_tc;
4557 * i40e_dcb_get_num_tc - Get the number of TCs from DCBx config
4558 * @dcbcfg: the corresponding DCBx configuration structure
4560 * Return the number of TCs from given DCBx configuration
4562 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4564 int i, tc_unused = 0;
4565 u8 num_tc = 0;
4566 u8 ret = 0;
4568 /* Scan the ETS Config Priority Table to find
4569 * traffic class enabled for a given priority
4570 * and create a bitmask of enabled TCs
4572 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4573 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4575 /* Now scan the bitmask to check for
4576 * contiguous TCs starting with TC0
4578 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4579 if (num_tc & BIT(i)) {
4580 if (!tc_unused) {
4581 ret++;
4582 } else {
4583 pr_err("Non-contiguous TC - Disabling DCB\n");
4584 return 1;
4586 } else {
4587 tc_unused = 1;
4591 /* There is always at least TC0 */
4592 if (!ret)
4593 ret = 1;
4595 return ret;
4599 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4600 * @dcbcfg: the corresponding DCBx configuration structure
4602 * Query the current DCB configuration and return the number of
4603 * traffic classes enabled from the given DCBX config
4605 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4607 u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4608 u8 enabled_tc = 1;
4609 u8 i;
4611 for (i = 0; i < num_tc; i++)
4612 enabled_tc |= BIT(i);
4614 return enabled_tc;
4618 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4619 * @pf: PF being queried
4621 * Return number of traffic classes enabled for the given PF
4623 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4625 struct i40e_hw *hw = &pf->hw;
4626 u8 i, enabled_tc = 1;
4627 u8 num_tc = 0;
4628 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4630 /* If DCB is not enabled then always in single TC */
4631 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4632 return 1;
4634 /* SFP mode will be enabled for all TCs on port */
4635 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4636 return i40e_dcb_get_num_tc(dcbcfg);
4638 /* MFP mode return count of enabled TCs for this PF */
4639 if (pf->hw.func_caps.iscsi)
4640 enabled_tc = i40e_get_iscsi_tc_map(pf);
4641 else
4642 return 1; /* Only TC0 */
4644 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4645 if (enabled_tc & BIT(i))
4646 num_tc++;
4648 return num_tc;
4652 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4653 * @pf: PF being queried
4655 * Return a bitmap for enabled traffic classes for this PF.
4657 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4659 /* If DCB is not enabled for this PF then just return default TC */
4660 if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4661 return I40E_DEFAULT_TRAFFIC_CLASS;
4663 /* SFP mode we want PF to be enabled for all TCs */
4664 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4665 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4667 /* MFP enabled and iSCSI PF type */
4668 if (pf->hw.func_caps.iscsi)
4669 return i40e_get_iscsi_tc_map(pf);
4670 else
4671 return I40E_DEFAULT_TRAFFIC_CLASS;
4675 * i40e_vsi_get_bw_info - Query VSI BW Information
4676 * @vsi: the VSI being queried
4678 * Returns 0 on success, negative value on failure
4680 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4682 struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4683 struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4684 struct i40e_pf *pf = vsi->back;
4685 struct i40e_hw *hw = &pf->hw;
4686 i40e_status ret;
4687 u32 tc_bw_max;
4688 int i;
4690 /* Get the VSI level BW configuration */
4691 ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4692 if (ret) {
4693 dev_info(&pf->pdev->dev,
4694 "couldn't get PF vsi bw config, err %s aq_err %s\n",
4695 i40e_stat_str(&pf->hw, ret),
4696 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4697 return -EINVAL;
4700 /* Get the VSI level BW configuration per TC */
4701 ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4702 NULL);
4703 if (ret) {
4704 dev_info(&pf->pdev->dev,
4705 "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4706 i40e_stat_str(&pf->hw, ret),
4707 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4708 return -EINVAL;
4711 if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4712 dev_info(&pf->pdev->dev,
4713 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4714 bw_config.tc_valid_bits,
4715 bw_ets_config.tc_valid_bits);
4716 /* Still continuing */
4719 vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4720 vsi->bw_max_quanta = bw_config.max_bw;
4721 tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4722 (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4723 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4724 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4725 vsi->bw_ets_limit_credits[i] =
4726 le16_to_cpu(bw_ets_config.credits[i]);
4727 /* 3 bits out of 4 for each TC */
4728 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4731 return 0;
4735 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4736 * @vsi: the VSI being configured
4737 * @enabled_tc: TC bitmap
4738 * @bw_credits: BW shared credits per TC
4740 * Returns 0 on success, negative value on failure
4742 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4743 u8 *bw_share)
4745 struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4746 i40e_status ret;
4747 int i;
4749 bw_data.tc_valid_bits = enabled_tc;
4750 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4751 bw_data.tc_bw_credits[i] = bw_share[i];
4753 ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4754 NULL);
4755 if (ret) {
4756 dev_info(&vsi->back->pdev->dev,
4757 "AQ command Config VSI BW allocation per TC failed = %d\n",
4758 vsi->back->hw.aq.asq_last_status);
4759 return -EINVAL;
4762 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4763 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4765 return 0;
4769 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4770 * @vsi: the VSI being configured
4771 * @enabled_tc: TC map to be enabled
4774 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4776 struct net_device *netdev = vsi->netdev;
4777 struct i40e_pf *pf = vsi->back;
4778 struct i40e_hw *hw = &pf->hw;
4779 u8 netdev_tc = 0;
4780 int i;
4781 struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4783 if (!netdev)
4784 return;
4786 if (!enabled_tc) {
4787 netdev_reset_tc(netdev);
4788 return;
4791 /* Set up actual enabled TCs on the VSI */
4792 if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4793 return;
4795 /* set per TC queues for the VSI */
4796 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4797 /* Only set TC queues for enabled tcs
4799 * e.g. For a VSI that has TC0 and TC3 enabled the
4800 * enabled_tc bitmap would be 0x00001001; the driver
4801 * will set the numtc for netdev as 2 that will be
4802 * referenced by the netdev layer as TC 0 and 1.
4804 if (vsi->tc_config.enabled_tc & BIT(i))
4805 netdev_set_tc_queue(netdev,
4806 vsi->tc_config.tc_info[i].netdev_tc,
4807 vsi->tc_config.tc_info[i].qcount,
4808 vsi->tc_config.tc_info[i].qoffset);
4811 /* Assign UP2TC map for the VSI */
4812 for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4813 /* Get the actual TC# for the UP */
4814 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4815 /* Get the mapped netdev TC# for the UP */
4816 netdev_tc = vsi->tc_config.tc_info[ets_tc].netdev_tc;
4817 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4822 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4823 * @vsi: the VSI being configured
4824 * @ctxt: the ctxt buffer returned from AQ VSI update param command
4826 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4827 struct i40e_vsi_context *ctxt)
4829 /* copy just the sections touched not the entire info
4830 * since not all sections are valid as returned by
4831 * update vsi params
4833 vsi->info.mapping_flags = ctxt->info.mapping_flags;
4834 memcpy(&vsi->info.queue_mapping,
4835 &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4836 memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4837 sizeof(vsi->info.tc_mapping));
4841 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4842 * @vsi: VSI to be configured
4843 * @enabled_tc: TC bitmap
4845 * This configures a particular VSI for TCs that are mapped to the
4846 * given TC bitmap. It uses default bandwidth share for TCs across
4847 * VSIs to configure TC for a particular VSI.
4849 * NOTE:
4850 * It is expected that the VSI queues have been quisced before calling
4851 * this function.
4853 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4855 u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4856 struct i40e_vsi_context ctxt;
4857 int ret = 0;
4858 int i;
4860 /* Check if enabled_tc is same as existing or new TCs */
4861 if (vsi->tc_config.enabled_tc == enabled_tc)
4862 return ret;
4864 /* Enable ETS TCs with equal BW Share for now across all VSIs */
4865 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4866 if (enabled_tc & BIT(i))
4867 bw_share[i] = 1;
4870 ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4871 if (ret) {
4872 dev_info(&vsi->back->pdev->dev,
4873 "Failed configuring TC map %d for VSI %d\n",
4874 enabled_tc, vsi->seid);
4875 goto out;
4878 /* Update Queue Pairs Mapping for currently enabled UPs */
4879 ctxt.seid = vsi->seid;
4880 ctxt.pf_num = vsi->back->hw.pf_id;
4881 ctxt.vf_num = 0;
4882 ctxt.uplink_seid = vsi->uplink_seid;
4883 ctxt.info = vsi->info;
4884 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4886 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4887 ctxt.info.valid_sections |=
4888 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4889 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4892 /* Update the VSI after updating the VSI queue-mapping information */
4893 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4894 if (ret) {
4895 dev_info(&vsi->back->pdev->dev,
4896 "Update vsi tc config failed, err %s aq_err %s\n",
4897 i40e_stat_str(&vsi->back->hw, ret),
4898 i40e_aq_str(&vsi->back->hw,
4899 vsi->back->hw.aq.asq_last_status));
4900 goto out;
4902 /* update the local VSI info with updated queue map */
4903 i40e_vsi_update_queue_map(vsi, &ctxt);
4904 vsi->info.valid_sections = 0;
4906 /* Update current VSI BW information */
4907 ret = i40e_vsi_get_bw_info(vsi);
4908 if (ret) {
4909 dev_info(&vsi->back->pdev->dev,
4910 "Failed updating vsi bw info, err %s aq_err %s\n",
4911 i40e_stat_str(&vsi->back->hw, ret),
4912 i40e_aq_str(&vsi->back->hw,
4913 vsi->back->hw.aq.asq_last_status));
4914 goto out;
4917 /* Update the netdev TC setup */
4918 i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4919 out:
4920 return ret;
4924 * i40e_veb_config_tc - Configure TCs for given VEB
4925 * @veb: given VEB
4926 * @enabled_tc: TC bitmap
4928 * Configures given TC bitmap for VEB (switching) element
4930 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4932 struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4933 struct i40e_pf *pf = veb->pf;
4934 int ret = 0;
4935 int i;
4937 /* No TCs or already enabled TCs just return */
4938 if (!enabled_tc || veb->enabled_tc == enabled_tc)
4939 return ret;
4941 bw_data.tc_valid_bits = enabled_tc;
4942 /* bw_data.absolute_credits is not set (relative) */
4944 /* Enable ETS TCs with equal BW Share for now */
4945 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4946 if (enabled_tc & BIT(i))
4947 bw_data.tc_bw_share_credits[i] = 1;
4950 ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4951 &bw_data, NULL);
4952 if (ret) {
4953 dev_info(&pf->pdev->dev,
4954 "VEB bw config failed, err %s aq_err %s\n",
4955 i40e_stat_str(&pf->hw, ret),
4956 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4957 goto out;
4960 /* Update the BW information */
4961 ret = i40e_veb_get_bw_info(veb);
4962 if (ret) {
4963 dev_info(&pf->pdev->dev,
4964 "Failed getting veb bw config, err %s aq_err %s\n",
4965 i40e_stat_str(&pf->hw, ret),
4966 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4969 out:
4970 return ret;
4973 #ifdef CONFIG_I40E_DCB
4975 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4976 * @pf: PF struct
4978 * Reconfigure VEB/VSIs on a given PF; it is assumed that
4979 * the caller would've quiesce all the VSIs before calling
4980 * this function
4982 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4984 u8 tc_map = 0;
4985 int ret;
4986 u8 v;
4988 /* Enable the TCs available on PF to all VEBs */
4989 tc_map = i40e_pf_get_tc_map(pf);
4990 for (v = 0; v < I40E_MAX_VEB; v++) {
4991 if (!pf->veb[v])
4992 continue;
4993 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4994 if (ret) {
4995 dev_info(&pf->pdev->dev,
4996 "Failed configuring TC for VEB seid=%d\n",
4997 pf->veb[v]->seid);
4998 /* Will try to configure as many components */
5002 /* Update each VSI */
5003 for (v = 0; v < pf->num_alloc_vsi; v++) {
5004 if (!pf->vsi[v])
5005 continue;
5007 /* - Enable all TCs for the LAN VSI
5008 #ifdef I40E_FCOE
5009 * - For FCoE VSI only enable the TC configured
5010 * as per the APP TLV
5011 #endif
5012 * - For all others keep them at TC0 for now
5014 if (v == pf->lan_vsi)
5015 tc_map = i40e_pf_get_tc_map(pf);
5016 else
5017 tc_map = I40E_DEFAULT_TRAFFIC_CLASS;
5018 #ifdef I40E_FCOE
5019 if (pf->vsi[v]->type == I40E_VSI_FCOE)
5020 tc_map = i40e_get_fcoe_tc_map(pf);
5021 #endif /* #ifdef I40E_FCOE */
5023 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5024 if (ret) {
5025 dev_info(&pf->pdev->dev,
5026 "Failed configuring TC for VSI seid=%d\n",
5027 pf->vsi[v]->seid);
5028 /* Will try to configure as many components */
5029 } else {
5030 /* Re-configure VSI vectors based on updated TC map */
5031 i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5032 if (pf->vsi[v]->netdev)
5033 i40e_dcbnl_set_all(pf->vsi[v]);
5039 * i40e_resume_port_tx - Resume port Tx
5040 * @pf: PF struct
5042 * Resume a port's Tx and issue a PF reset in case of failure to
5043 * resume.
5045 static int i40e_resume_port_tx(struct i40e_pf *pf)
5047 struct i40e_hw *hw = &pf->hw;
5048 int ret;
5050 ret = i40e_aq_resume_port_tx(hw, NULL);
5051 if (ret) {
5052 dev_info(&pf->pdev->dev,
5053 "Resume Port Tx failed, err %s aq_err %s\n",
5054 i40e_stat_str(&pf->hw, ret),
5055 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5056 /* Schedule PF reset to recover */
5057 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5058 i40e_service_event_schedule(pf);
5061 return ret;
5065 * i40e_init_pf_dcb - Initialize DCB configuration
5066 * @pf: PF being configured
5068 * Query the current DCB configuration and cache it
5069 * in the hardware structure
5071 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5073 struct i40e_hw *hw = &pf->hw;
5074 int err = 0;
5076 /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5077 if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5078 goto out;
5080 /* Get the initial DCB configuration */
5081 err = i40e_init_dcb(hw);
5082 if (!err) {
5083 /* Device/Function is not DCBX capable */
5084 if ((!hw->func_caps.dcb) ||
5085 (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5086 dev_info(&pf->pdev->dev,
5087 "DCBX offload is not supported or is disabled for this PF.\n");
5089 if (pf->flags & I40E_FLAG_MFP_ENABLED)
5090 goto out;
5092 } else {
5093 /* When status is not DISABLED then DCBX in FW */
5094 pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5095 DCB_CAP_DCBX_VER_IEEE;
5097 pf->flags |= I40E_FLAG_DCB_CAPABLE;
5098 /* Enable DCB tagging only when more than one TC
5099 * or explicitly disable if only one TC
5101 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5102 pf->flags |= I40E_FLAG_DCB_ENABLED;
5103 else
5104 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5105 dev_dbg(&pf->pdev->dev,
5106 "DCBX offload is supported for this PF.\n");
5108 } else {
5109 dev_info(&pf->pdev->dev,
5110 "Query for DCB configuration failed, err %s aq_err %s\n",
5111 i40e_stat_str(&pf->hw, err),
5112 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5115 out:
5116 return err;
5118 #endif /* CONFIG_I40E_DCB */
5119 #define SPEED_SIZE 14
5120 #define FC_SIZE 8
5122 * i40e_print_link_message - print link up or down
5123 * @vsi: the VSI for which link needs a message
5125 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5127 char *speed = "Unknown";
5128 char *fc = "Unknown";
5130 if (vsi->current_isup == isup)
5131 return;
5132 vsi->current_isup = isup;
5133 if (!isup) {
5134 netdev_info(vsi->netdev, "NIC Link is Down\n");
5135 return;
5138 /* Warn user if link speed on NPAR enabled partition is not at
5139 * least 10GB
5141 if (vsi->back->hw.func_caps.npar_enable &&
5142 (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5143 vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5144 netdev_warn(vsi->netdev,
5145 "The partition detected link speed that is less than 10Gbps\n");
5147 switch (vsi->back->hw.phy.link_info.link_speed) {
5148 case I40E_LINK_SPEED_40GB:
5149 speed = "40 G";
5150 break;
5151 case I40E_LINK_SPEED_20GB:
5152 speed = "20 G";
5153 break;
5154 case I40E_LINK_SPEED_10GB:
5155 speed = "10 G";
5156 break;
5157 case I40E_LINK_SPEED_1GB:
5158 speed = "1000 M";
5159 break;
5160 case I40E_LINK_SPEED_100MB:
5161 speed = "100 M";
5162 break;
5163 default:
5164 break;
5167 switch (vsi->back->hw.fc.current_mode) {
5168 case I40E_FC_FULL:
5169 fc = "RX/TX";
5170 break;
5171 case I40E_FC_TX_PAUSE:
5172 fc = "TX";
5173 break;
5174 case I40E_FC_RX_PAUSE:
5175 fc = "RX";
5176 break;
5177 default:
5178 fc = "None";
5179 break;
5182 netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5183 speed, fc);
5187 * i40e_up_complete - Finish the last steps of bringing up a connection
5188 * @vsi: the VSI being configured
5190 static int i40e_up_complete(struct i40e_vsi *vsi)
5192 struct i40e_pf *pf = vsi->back;
5193 int err;
5195 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5196 i40e_vsi_configure_msix(vsi);
5197 else
5198 i40e_configure_msi_and_legacy(vsi);
5200 /* start rings */
5201 err = i40e_vsi_control_rings(vsi, true);
5202 if (err)
5203 return err;
5205 clear_bit(__I40E_DOWN, &vsi->state);
5206 i40e_napi_enable_all(vsi);
5207 i40e_vsi_enable_irq(vsi);
5209 if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5210 (vsi->netdev)) {
5211 i40e_print_link_message(vsi, true);
5212 netif_tx_start_all_queues(vsi->netdev);
5213 netif_carrier_on(vsi->netdev);
5214 } else if (vsi->netdev) {
5215 i40e_print_link_message(vsi, false);
5216 /* need to check for qualified module here*/
5217 if ((pf->hw.phy.link_info.link_info &
5218 I40E_AQ_MEDIA_AVAILABLE) &&
5219 (!(pf->hw.phy.link_info.an_info &
5220 I40E_AQ_QUALIFIED_MODULE)))
5221 netdev_err(vsi->netdev,
5222 "the driver failed to link because an unqualified module was detected.");
5225 /* replay FDIR SB filters */
5226 if (vsi->type == I40E_VSI_FDIR) {
5227 /* reset fd counters */
5228 pf->fd_add_err = pf->fd_atr_cnt = 0;
5229 if (pf->fd_tcp_rule > 0) {
5230 pf->auto_disable_flags |= I40E_FLAG_FD_ATR_ENABLED;
5231 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5232 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5233 pf->fd_tcp_rule = 0;
5235 i40e_fdir_filter_restore(vsi);
5238 /* On the next run of the service_task, notify any clients of the new
5239 * opened netdev
5241 pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5242 i40e_service_event_schedule(pf);
5244 return 0;
5248 * i40e_vsi_reinit_locked - Reset the VSI
5249 * @vsi: the VSI being configured
5251 * Rebuild the ring structs after some configuration
5252 * has changed, e.g. MTU size.
5254 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5256 struct i40e_pf *pf = vsi->back;
5258 WARN_ON(in_interrupt());
5259 while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5260 usleep_range(1000, 2000);
5261 i40e_down(vsi);
5263 i40e_up(vsi);
5264 clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5268 * i40e_up - Bring the connection back up after being down
5269 * @vsi: the VSI being configured
5271 int i40e_up(struct i40e_vsi *vsi)
5273 int err;
5275 err = i40e_vsi_configure(vsi);
5276 if (!err)
5277 err = i40e_up_complete(vsi);
5279 return err;
5283 * i40e_down - Shutdown the connection processing
5284 * @vsi: the VSI being stopped
5286 void i40e_down(struct i40e_vsi *vsi)
5288 int i;
5290 /* It is assumed that the caller of this function
5291 * sets the vsi->state __I40E_DOWN bit.
5293 if (vsi->netdev) {
5294 netif_carrier_off(vsi->netdev);
5295 netif_tx_disable(vsi->netdev);
5297 i40e_vsi_disable_irq(vsi);
5298 i40e_vsi_control_rings(vsi, false);
5299 i40e_napi_disable_all(vsi);
5301 for (i = 0; i < vsi->num_queue_pairs; i++) {
5302 i40e_clean_tx_ring(vsi->tx_rings[i]);
5303 i40e_clean_rx_ring(vsi->rx_rings[i]);
5306 i40e_notify_client_of_netdev_close(vsi, false);
5311 * i40e_setup_tc - configure multiple traffic classes
5312 * @netdev: net device to configure
5313 * @tc: number of traffic classes to enable
5315 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5317 struct i40e_netdev_priv *np = netdev_priv(netdev);
5318 struct i40e_vsi *vsi = np->vsi;
5319 struct i40e_pf *pf = vsi->back;
5320 u8 enabled_tc = 0;
5321 int ret = -EINVAL;
5322 int i;
5324 /* Check if DCB enabled to continue */
5325 if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5326 netdev_info(netdev, "DCB is not enabled for adapter\n");
5327 goto exit;
5330 /* Check if MFP enabled */
5331 if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5332 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5333 goto exit;
5336 /* Check whether tc count is within enabled limit */
5337 if (tc > i40e_pf_get_num_tc(pf)) {
5338 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5339 goto exit;
5342 /* Generate TC map for number of tc requested */
5343 for (i = 0; i < tc; i++)
5344 enabled_tc |= BIT(i);
5346 /* Requesting same TC configuration as already enabled */
5347 if (enabled_tc == vsi->tc_config.enabled_tc)
5348 return 0;
5350 /* Quiesce VSI queues */
5351 i40e_quiesce_vsi(vsi);
5353 /* Configure VSI for enabled TCs */
5354 ret = i40e_vsi_config_tc(vsi, enabled_tc);
5355 if (ret) {
5356 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5357 vsi->seid);
5358 goto exit;
5361 /* Unquiesce VSI */
5362 i40e_unquiesce_vsi(vsi);
5364 exit:
5365 return ret;
5368 #ifdef I40E_FCOE
5369 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5370 struct tc_to_netdev *tc)
5371 #else
5372 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5373 struct tc_to_netdev *tc)
5374 #endif
5376 if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5377 return -EINVAL;
5378 return i40e_setup_tc(netdev, tc->tc);
5382 * i40e_open - Called when a network interface is made active
5383 * @netdev: network interface device structure
5385 * The open entry point is called when a network interface is made
5386 * active by the system (IFF_UP). At this point all resources needed
5387 * for transmit and receive operations are allocated, the interrupt
5388 * handler is registered with the OS, the netdev watchdog subtask is
5389 * enabled, and the stack is notified that the interface is ready.
5391 * Returns 0 on success, negative value on failure
5393 int i40e_open(struct net_device *netdev)
5395 struct i40e_netdev_priv *np = netdev_priv(netdev);
5396 struct i40e_vsi *vsi = np->vsi;
5397 struct i40e_pf *pf = vsi->back;
5398 int err;
5400 /* disallow open during test or if eeprom is broken */
5401 if (test_bit(__I40E_TESTING, &pf->state) ||
5402 test_bit(__I40E_BAD_EEPROM, &pf->state))
5403 return -EBUSY;
5405 netif_carrier_off(netdev);
5407 err = i40e_vsi_open(vsi);
5408 if (err)
5409 return err;
5411 /* configure global TSO hardware offload settings */
5412 wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5413 TCP_FLAG_FIN) >> 16);
5414 wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5415 TCP_FLAG_FIN |
5416 TCP_FLAG_CWR) >> 16);
5417 wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5419 udp_tunnel_get_rx_info(netdev);
5421 return 0;
5425 * i40e_vsi_open -
5426 * @vsi: the VSI to open
5428 * Finish initialization of the VSI.
5430 * Returns 0 on success, negative value on failure
5432 int i40e_vsi_open(struct i40e_vsi *vsi)
5434 struct i40e_pf *pf = vsi->back;
5435 char int_name[I40E_INT_NAME_STR_LEN];
5436 int err;
5438 /* allocate descriptors */
5439 err = i40e_vsi_setup_tx_resources(vsi);
5440 if (err)
5441 goto err_setup_tx;
5442 err = i40e_vsi_setup_rx_resources(vsi);
5443 if (err)
5444 goto err_setup_rx;
5446 err = i40e_vsi_configure(vsi);
5447 if (err)
5448 goto err_setup_rx;
5450 if (vsi->netdev) {
5451 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5452 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5453 err = i40e_vsi_request_irq(vsi, int_name);
5454 if (err)
5455 goto err_setup_rx;
5457 /* Notify the stack of the actual queue counts. */
5458 err = netif_set_real_num_tx_queues(vsi->netdev,
5459 vsi->num_queue_pairs);
5460 if (err)
5461 goto err_set_queues;
5463 err = netif_set_real_num_rx_queues(vsi->netdev,
5464 vsi->num_queue_pairs);
5465 if (err)
5466 goto err_set_queues;
5468 } else if (vsi->type == I40E_VSI_FDIR) {
5469 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5470 dev_driver_string(&pf->pdev->dev),
5471 dev_name(&pf->pdev->dev));
5472 err = i40e_vsi_request_irq(vsi, int_name);
5474 } else {
5475 err = -EINVAL;
5476 goto err_setup_rx;
5479 err = i40e_up_complete(vsi);
5480 if (err)
5481 goto err_up_complete;
5483 return 0;
5485 err_up_complete:
5486 i40e_down(vsi);
5487 err_set_queues:
5488 i40e_vsi_free_irq(vsi);
5489 err_setup_rx:
5490 i40e_vsi_free_rx_resources(vsi);
5491 err_setup_tx:
5492 i40e_vsi_free_tx_resources(vsi);
5493 if (vsi == pf->vsi[pf->lan_vsi])
5494 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5496 return err;
5500 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5501 * @pf: Pointer to PF
5503 * This function destroys the hlist where all the Flow Director
5504 * filters were saved.
5506 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5508 struct i40e_fdir_filter *filter;
5509 struct hlist_node *node2;
5511 hlist_for_each_entry_safe(filter, node2,
5512 &pf->fdir_filter_list, fdir_node) {
5513 hlist_del(&filter->fdir_node);
5514 kfree(filter);
5516 pf->fdir_pf_active_filters = 0;
5520 * i40e_close - Disables a network interface
5521 * @netdev: network interface device structure
5523 * The close entry point is called when an interface is de-activated
5524 * by the OS. The hardware is still under the driver's control, but
5525 * this netdev interface is disabled.
5527 * Returns 0, this is not allowed to fail
5529 int i40e_close(struct net_device *netdev)
5531 struct i40e_netdev_priv *np = netdev_priv(netdev);
5532 struct i40e_vsi *vsi = np->vsi;
5534 i40e_vsi_close(vsi);
5536 return 0;
5540 * i40e_do_reset - Start a PF or Core Reset sequence
5541 * @pf: board private structure
5542 * @reset_flags: which reset is requested
5544 * The essential difference in resets is that the PF Reset
5545 * doesn't clear the packet buffers, doesn't reset the PE
5546 * firmware, and doesn't bother the other PFs on the chip.
5548 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5550 u32 val;
5552 WARN_ON(in_interrupt());
5555 /* do the biggest reset indicated */
5556 if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5558 /* Request a Global Reset
5560 * This will start the chip's countdown to the actual full
5561 * chip reset event, and a warning interrupt to be sent
5562 * to all PFs, including the requestor. Our handler
5563 * for the warning interrupt will deal with the shutdown
5564 * and recovery of the switch setup.
5566 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5567 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5568 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5569 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5571 } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5573 /* Request a Core Reset
5575 * Same as Global Reset, except does *not* include the MAC/PHY
5577 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5578 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5579 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5580 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5581 i40e_flush(&pf->hw);
5583 } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5585 /* Request a PF Reset
5587 * Resets only the PF-specific registers
5589 * This goes directly to the tear-down and rebuild of
5590 * the switch, since we need to do all the recovery as
5591 * for the Core Reset.
5593 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5594 i40e_handle_reset_warning(pf);
5596 } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5597 int v;
5599 /* Find the VSI(s) that requested a re-init */
5600 dev_info(&pf->pdev->dev,
5601 "VSI reinit requested\n");
5602 for (v = 0; v < pf->num_alloc_vsi; v++) {
5603 struct i40e_vsi *vsi = pf->vsi[v];
5605 if (vsi != NULL &&
5606 test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5607 i40e_vsi_reinit_locked(pf->vsi[v]);
5608 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5611 } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5612 int v;
5614 /* Find the VSI(s) that needs to be brought down */
5615 dev_info(&pf->pdev->dev, "VSI down requested\n");
5616 for (v = 0; v < pf->num_alloc_vsi; v++) {
5617 struct i40e_vsi *vsi = pf->vsi[v];
5619 if (vsi != NULL &&
5620 test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5621 set_bit(__I40E_DOWN, &vsi->state);
5622 i40e_down(vsi);
5623 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5626 } else {
5627 dev_info(&pf->pdev->dev,
5628 "bad reset request 0x%08x\n", reset_flags);
5632 #ifdef CONFIG_I40E_DCB
5634 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5635 * @pf: board private structure
5636 * @old_cfg: current DCB config
5637 * @new_cfg: new DCB config
5639 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5640 struct i40e_dcbx_config *old_cfg,
5641 struct i40e_dcbx_config *new_cfg)
5643 bool need_reconfig = false;
5645 /* Check if ETS configuration has changed */
5646 if (memcmp(&new_cfg->etscfg,
5647 &old_cfg->etscfg,
5648 sizeof(new_cfg->etscfg))) {
5649 /* If Priority Table has changed reconfig is needed */
5650 if (memcmp(&new_cfg->etscfg.prioritytable,
5651 &old_cfg->etscfg.prioritytable,
5652 sizeof(new_cfg->etscfg.prioritytable))) {
5653 need_reconfig = true;
5654 dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5657 if (memcmp(&new_cfg->etscfg.tcbwtable,
5658 &old_cfg->etscfg.tcbwtable,
5659 sizeof(new_cfg->etscfg.tcbwtable)))
5660 dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5662 if (memcmp(&new_cfg->etscfg.tsatable,
5663 &old_cfg->etscfg.tsatable,
5664 sizeof(new_cfg->etscfg.tsatable)))
5665 dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5668 /* Check if PFC configuration has changed */
5669 if (memcmp(&new_cfg->pfc,
5670 &old_cfg->pfc,
5671 sizeof(new_cfg->pfc))) {
5672 need_reconfig = true;
5673 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5676 /* Check if APP Table has changed */
5677 if (memcmp(&new_cfg->app,
5678 &old_cfg->app,
5679 sizeof(new_cfg->app))) {
5680 need_reconfig = true;
5681 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5684 dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5685 return need_reconfig;
5689 * i40e_handle_lldp_event - Handle LLDP Change MIB event
5690 * @pf: board private structure
5691 * @e: event info posted on ARQ
5693 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5694 struct i40e_arq_event_info *e)
5696 struct i40e_aqc_lldp_get_mib *mib =
5697 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5698 struct i40e_hw *hw = &pf->hw;
5699 struct i40e_dcbx_config tmp_dcbx_cfg;
5700 bool need_reconfig = false;
5701 int ret = 0;
5702 u8 type;
5704 /* Not DCB capable or capability disabled */
5705 if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5706 return ret;
5708 /* Ignore if event is not for Nearest Bridge */
5709 type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5710 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5711 dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5712 if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5713 return ret;
5715 /* Check MIB Type and return if event for Remote MIB update */
5716 type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5717 dev_dbg(&pf->pdev->dev,
5718 "LLDP event mib type %s\n", type ? "remote" : "local");
5719 if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5720 /* Update the remote cached instance and return */
5721 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5722 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5723 &hw->remote_dcbx_config);
5724 goto exit;
5727 /* Store the old configuration */
5728 tmp_dcbx_cfg = hw->local_dcbx_config;
5730 /* Reset the old DCBx configuration data */
5731 memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5732 /* Get updated DCBX data from firmware */
5733 ret = i40e_get_dcb_config(&pf->hw);
5734 if (ret) {
5735 dev_info(&pf->pdev->dev,
5736 "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5737 i40e_stat_str(&pf->hw, ret),
5738 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5739 goto exit;
5742 /* No change detected in DCBX configs */
5743 if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5744 sizeof(tmp_dcbx_cfg))) {
5745 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5746 goto exit;
5749 need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5750 &hw->local_dcbx_config);
5752 i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5754 if (!need_reconfig)
5755 goto exit;
5757 /* Enable DCB tagging only when more than one TC */
5758 if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5759 pf->flags |= I40E_FLAG_DCB_ENABLED;
5760 else
5761 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5763 set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5764 /* Reconfiguration needed quiesce all VSIs */
5765 i40e_pf_quiesce_all_vsi(pf);
5767 /* Changes in configuration update VEB/VSI */
5768 i40e_dcb_reconfigure(pf);
5770 ret = i40e_resume_port_tx(pf);
5772 clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5773 /* In case of error no point in resuming VSIs */
5774 if (ret)
5775 goto exit;
5777 /* Wait for the PF's queues to be disabled */
5778 ret = i40e_pf_wait_queues_disabled(pf);
5779 if (ret) {
5780 /* Schedule PF reset to recover */
5781 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5782 i40e_service_event_schedule(pf);
5783 } else {
5784 i40e_pf_unquiesce_all_vsi(pf);
5785 /* Notify the client for the DCB changes */
5786 i40e_notify_client_of_l2_param_changes(pf->vsi[pf->lan_vsi]);
5789 exit:
5790 return ret;
5792 #endif /* CONFIG_I40E_DCB */
5795 * i40e_do_reset_safe - Protected reset path for userland calls.
5796 * @pf: board private structure
5797 * @reset_flags: which reset is requested
5800 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5802 rtnl_lock();
5803 i40e_do_reset(pf, reset_flags);
5804 rtnl_unlock();
5808 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5809 * @pf: board private structure
5810 * @e: event info posted on ARQ
5812 * Handler for LAN Queue Overflow Event generated by the firmware for PF
5813 * and VF queues
5815 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5816 struct i40e_arq_event_info *e)
5818 struct i40e_aqc_lan_overflow *data =
5819 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5820 u32 queue = le32_to_cpu(data->prtdcb_rupto);
5821 u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5822 struct i40e_hw *hw = &pf->hw;
5823 struct i40e_vf *vf;
5824 u16 vf_id;
5826 dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5827 queue, qtx_ctl);
5829 /* Queue belongs to VF, find the VF and issue VF reset */
5830 if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5831 >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5832 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5833 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5834 vf_id -= hw->func_caps.vf_base_id;
5835 vf = &pf->vf[vf_id];
5836 i40e_vc_notify_vf_reset(vf);
5837 /* Allow VF to process pending reset notification */
5838 msleep(20);
5839 i40e_reset_vf(vf, false);
5844 * i40e_service_event_complete - Finish up the service event
5845 * @pf: board private structure
5847 static void i40e_service_event_complete(struct i40e_pf *pf)
5849 WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5851 /* flush memory to make sure state is correct before next watchog */
5852 smp_mb__before_atomic();
5853 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5857 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5858 * @pf: board private structure
5860 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5862 u32 val, fcnt_prog;
5864 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5865 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5866 return fcnt_prog;
5870 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5871 * @pf: board private structure
5873 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5875 u32 val, fcnt_prog;
5877 val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5878 fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5879 ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5880 I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5881 return fcnt_prog;
5885 * i40e_get_global_fd_count - Get total FD filters programmed on device
5886 * @pf: board private structure
5888 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5890 u32 val, fcnt_prog;
5892 val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5893 fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5894 ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5895 I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5896 return fcnt_prog;
5900 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5901 * @pf: board private structure
5903 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5905 struct i40e_fdir_filter *filter;
5906 u32 fcnt_prog, fcnt_avail;
5907 struct hlist_node *node;
5909 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5910 return;
5912 /* Check if, FD SB or ATR was auto disabled and if there is enough room
5913 * to re-enable
5915 fcnt_prog = i40e_get_global_fd_count(pf);
5916 fcnt_avail = pf->fdir_pf_filter_count;
5917 if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5918 (pf->fd_add_err == 0) ||
5919 (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5920 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5921 (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5922 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5923 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5924 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5928 /* Wait for some more space to be available to turn on ATR. We also
5929 * must check that no existing ntuple rules for TCP are in effect
5931 if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5932 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5933 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED) &&
5934 (pf->fd_tcp_rule == 0)) {
5935 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5936 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5937 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table and there are no conflicting ntuple rules\n");
5941 /* if hw had a problem adding a filter, delete it */
5942 if (pf->fd_inv > 0) {
5943 hlist_for_each_entry_safe(filter, node,
5944 &pf->fdir_filter_list, fdir_node) {
5945 if (filter->fd_id == pf->fd_inv) {
5946 hlist_del(&filter->fdir_node);
5947 kfree(filter);
5948 pf->fdir_pf_active_filters--;
5954 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5955 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5957 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5958 * @pf: board private structure
5960 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5962 unsigned long min_flush_time;
5963 int flush_wait_retry = 50;
5964 bool disable_atr = false;
5965 int fd_room;
5966 int reg;
5968 if (!time_after(jiffies, pf->fd_flush_timestamp +
5969 (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5970 return;
5972 /* If the flush is happening too quick and we have mostly SB rules we
5973 * should not re-enable ATR for some time.
5975 min_flush_time = pf->fd_flush_timestamp +
5976 (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5977 fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5979 if (!(time_after(jiffies, min_flush_time)) &&
5980 (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5981 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5982 dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5983 disable_atr = true;
5986 pf->fd_flush_timestamp = jiffies;
5987 pf->auto_disable_flags |= I40E_FLAG_FD_ATR_ENABLED;
5988 /* flush all filters */
5989 wr32(&pf->hw, I40E_PFQF_CTL_1,
5990 I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5991 i40e_flush(&pf->hw);
5992 pf->fd_flush_cnt++;
5993 pf->fd_add_err = 0;
5994 do {
5995 /* Check FD flush status every 5-6msec */
5996 usleep_range(5000, 6000);
5997 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5998 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5999 break;
6000 } while (flush_wait_retry--);
6001 if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6002 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6003 } else {
6004 /* replay sideband filters */
6005 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6006 if (!disable_atr)
6007 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6008 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6009 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6010 dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6015 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6016 * @pf: board private structure
6018 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6020 return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6023 /* We can see up to 256 filter programming desc in transit if the filters are
6024 * being applied really fast; before we see the first
6025 * filter miss error on Rx queue 0. Accumulating enough error messages before
6026 * reacting will make sure we don't cause flush too often.
6028 #define I40E_MAX_FD_PROGRAM_ERROR 256
6031 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6032 * @pf: board private structure
6034 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6037 /* if interface is down do nothing */
6038 if (test_bit(__I40E_DOWN, &pf->state))
6039 return;
6041 if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6042 i40e_fdir_flush_and_replay(pf);
6044 i40e_fdir_check_and_reenable(pf);
6049 * i40e_vsi_link_event - notify VSI of a link event
6050 * @vsi: vsi to be notified
6051 * @link_up: link up or down
6053 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6055 if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6056 return;
6058 switch (vsi->type) {
6059 case I40E_VSI_MAIN:
6060 #ifdef I40E_FCOE
6061 case I40E_VSI_FCOE:
6062 #endif
6063 if (!vsi->netdev || !vsi->netdev_registered)
6064 break;
6066 if (link_up) {
6067 netif_carrier_on(vsi->netdev);
6068 netif_tx_wake_all_queues(vsi->netdev);
6069 } else {
6070 netif_carrier_off(vsi->netdev);
6071 netif_tx_stop_all_queues(vsi->netdev);
6073 break;
6075 case I40E_VSI_SRIOV:
6076 case I40E_VSI_VMDQ2:
6077 case I40E_VSI_CTRL:
6078 case I40E_VSI_IWARP:
6079 case I40E_VSI_MIRROR:
6080 default:
6081 /* there is no notification for other VSIs */
6082 break;
6087 * i40e_veb_link_event - notify elements on the veb of a link event
6088 * @veb: veb to be notified
6089 * @link_up: link up or down
6091 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6093 struct i40e_pf *pf;
6094 int i;
6096 if (!veb || !veb->pf)
6097 return;
6098 pf = veb->pf;
6100 /* depth first... */
6101 for (i = 0; i < I40E_MAX_VEB; i++)
6102 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6103 i40e_veb_link_event(pf->veb[i], link_up);
6105 /* ... now the local VSIs */
6106 for (i = 0; i < pf->num_alloc_vsi; i++)
6107 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6108 i40e_vsi_link_event(pf->vsi[i], link_up);
6112 * i40e_link_event - Update netif_carrier status
6113 * @pf: board private structure
6115 static void i40e_link_event(struct i40e_pf *pf)
6117 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6118 u8 new_link_speed, old_link_speed;
6119 i40e_status status;
6120 bool new_link, old_link;
6122 /* save off old link status information */
6123 pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6125 /* set this to force the get_link_status call to refresh state */
6126 pf->hw.phy.get_link_info = true;
6128 old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6130 status = i40e_get_link_status(&pf->hw, &new_link);
6131 if (status) {
6132 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6133 status);
6134 return;
6137 old_link_speed = pf->hw.phy.link_info_old.link_speed;
6138 new_link_speed = pf->hw.phy.link_info.link_speed;
6140 if (new_link == old_link &&
6141 new_link_speed == old_link_speed &&
6142 (test_bit(__I40E_DOWN, &vsi->state) ||
6143 new_link == netif_carrier_ok(vsi->netdev)))
6144 return;
6146 if (!test_bit(__I40E_DOWN, &vsi->state))
6147 i40e_print_link_message(vsi, new_link);
6149 /* Notify the base of the switch tree connected to
6150 * the link. Floating VEBs are not notified.
6152 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6153 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6154 else
6155 i40e_vsi_link_event(vsi, new_link);
6157 if (pf->vf)
6158 i40e_vc_notify_link_state(pf);
6160 if (pf->flags & I40E_FLAG_PTP)
6161 i40e_ptp_set_increment(pf);
6165 * i40e_watchdog_subtask - periodic checks not using event driven response
6166 * @pf: board private structure
6168 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6170 int i;
6172 /* if interface is down do nothing */
6173 if (test_bit(__I40E_DOWN, &pf->state) ||
6174 test_bit(__I40E_CONFIG_BUSY, &pf->state))
6175 return;
6177 /* make sure we don't do these things too often */
6178 if (time_before(jiffies, (pf->service_timer_previous +
6179 pf->service_timer_period)))
6180 return;
6181 pf->service_timer_previous = jiffies;
6183 if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6184 i40e_link_event(pf);
6186 /* Update the stats for active netdevs so the network stack
6187 * can look at updated numbers whenever it cares to
6189 for (i = 0; i < pf->num_alloc_vsi; i++)
6190 if (pf->vsi[i] && pf->vsi[i]->netdev)
6191 i40e_update_stats(pf->vsi[i]);
6193 if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6194 /* Update the stats for the active switching components */
6195 for (i = 0; i < I40E_MAX_VEB; i++)
6196 if (pf->veb[i])
6197 i40e_update_veb_stats(pf->veb[i]);
6200 i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6204 * i40e_reset_subtask - Set up for resetting the device and driver
6205 * @pf: board private structure
6207 static void i40e_reset_subtask(struct i40e_pf *pf)
6209 u32 reset_flags = 0;
6211 rtnl_lock();
6212 if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6213 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6214 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6216 if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6217 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6218 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6220 if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6221 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6222 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6224 if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6225 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6226 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6228 if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6229 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6230 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6233 /* If there's a recovery already waiting, it takes
6234 * precedence before starting a new reset sequence.
6236 if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6237 i40e_handle_reset_warning(pf);
6238 goto unlock;
6241 /* If we're already down or resetting, just bail */
6242 if (reset_flags &&
6243 !test_bit(__I40E_DOWN, &pf->state) &&
6244 !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6245 i40e_do_reset(pf, reset_flags);
6247 unlock:
6248 rtnl_unlock();
6252 * i40e_handle_link_event - Handle link event
6253 * @pf: board private structure
6254 * @e: event info posted on ARQ
6256 static void i40e_handle_link_event(struct i40e_pf *pf,
6257 struct i40e_arq_event_info *e)
6259 struct i40e_aqc_get_link_status *status =
6260 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6262 /* Do a new status request to re-enable LSE reporting
6263 * and load new status information into the hw struct
6264 * This completely ignores any state information
6265 * in the ARQ event info, instead choosing to always
6266 * issue the AQ update link status command.
6268 i40e_link_event(pf);
6270 /* check for unqualified module, if link is down */
6271 if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6272 (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6273 (!(status->link_info & I40E_AQ_LINK_UP)))
6274 dev_err(&pf->pdev->dev,
6275 "The driver failed to link because an unqualified module was detected.\n");
6279 * i40e_clean_adminq_subtask - Clean the AdminQ rings
6280 * @pf: board private structure
6282 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6284 struct i40e_arq_event_info event;
6285 struct i40e_hw *hw = &pf->hw;
6286 u16 pending, i = 0;
6287 i40e_status ret;
6288 u16 opcode;
6289 u32 oldval;
6290 u32 val;
6292 /* Do not run clean AQ when PF reset fails */
6293 if (test_bit(__I40E_RESET_FAILED, &pf->state))
6294 return;
6296 /* check for error indications */
6297 val = rd32(&pf->hw, pf->hw.aq.arq.len);
6298 oldval = val;
6299 if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6300 if (hw->debug_mask & I40E_DEBUG_AQ)
6301 dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6302 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6304 if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6305 if (hw->debug_mask & I40E_DEBUG_AQ)
6306 dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6307 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6308 pf->arq_overflows++;
6310 if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6311 if (hw->debug_mask & I40E_DEBUG_AQ)
6312 dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6313 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6315 if (oldval != val)
6316 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6318 val = rd32(&pf->hw, pf->hw.aq.asq.len);
6319 oldval = val;
6320 if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6321 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6322 dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6323 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6325 if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6326 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6327 dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6328 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6330 if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6331 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6332 dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6333 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6335 if (oldval != val)
6336 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6338 event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6339 event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6340 if (!event.msg_buf)
6341 return;
6343 do {
6344 ret = i40e_clean_arq_element(hw, &event, &pending);
6345 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6346 break;
6347 else if (ret) {
6348 dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6349 break;
6352 opcode = le16_to_cpu(event.desc.opcode);
6353 switch (opcode) {
6355 case i40e_aqc_opc_get_link_status:
6356 i40e_handle_link_event(pf, &event);
6357 break;
6358 case i40e_aqc_opc_send_msg_to_pf:
6359 ret = i40e_vc_process_vf_msg(pf,
6360 le16_to_cpu(event.desc.retval),
6361 le32_to_cpu(event.desc.cookie_high),
6362 le32_to_cpu(event.desc.cookie_low),
6363 event.msg_buf,
6364 event.msg_len);
6365 break;
6366 case i40e_aqc_opc_lldp_update_mib:
6367 dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6368 #ifdef CONFIG_I40E_DCB
6369 rtnl_lock();
6370 ret = i40e_handle_lldp_event(pf, &event);
6371 rtnl_unlock();
6372 #endif /* CONFIG_I40E_DCB */
6373 break;
6374 case i40e_aqc_opc_event_lan_overflow:
6375 dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6376 i40e_handle_lan_overflow_event(pf, &event);
6377 break;
6378 case i40e_aqc_opc_send_msg_to_peer:
6379 dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6380 break;
6381 case i40e_aqc_opc_nvm_erase:
6382 case i40e_aqc_opc_nvm_update:
6383 case i40e_aqc_opc_oem_post_update:
6384 i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6385 "ARQ NVM operation 0x%04x completed\n",
6386 opcode);
6387 break;
6388 default:
6389 dev_info(&pf->pdev->dev,
6390 "ARQ: Unknown event 0x%04x ignored\n",
6391 opcode);
6392 break;
6394 } while (pending && (i++ < pf->adminq_work_limit));
6396 clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6397 /* re-enable Admin queue interrupt cause */
6398 val = rd32(hw, I40E_PFINT_ICR0_ENA);
6399 val |= I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6400 wr32(hw, I40E_PFINT_ICR0_ENA, val);
6401 i40e_flush(hw);
6403 kfree(event.msg_buf);
6407 * i40e_verify_eeprom - make sure eeprom is good to use
6408 * @pf: board private structure
6410 static void i40e_verify_eeprom(struct i40e_pf *pf)
6412 int err;
6414 err = i40e_diag_eeprom_test(&pf->hw);
6415 if (err) {
6416 /* retry in case of garbage read */
6417 err = i40e_diag_eeprom_test(&pf->hw);
6418 if (err) {
6419 dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6420 err);
6421 set_bit(__I40E_BAD_EEPROM, &pf->state);
6425 if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6426 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6427 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6432 * i40e_enable_pf_switch_lb
6433 * @pf: pointer to the PF structure
6435 * enable switch loop back or die - no point in a return value
6437 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6439 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6440 struct i40e_vsi_context ctxt;
6441 int ret;
6443 ctxt.seid = pf->main_vsi_seid;
6444 ctxt.pf_num = pf->hw.pf_id;
6445 ctxt.vf_num = 0;
6446 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6447 if (ret) {
6448 dev_info(&pf->pdev->dev,
6449 "couldn't get PF vsi config, err %s aq_err %s\n",
6450 i40e_stat_str(&pf->hw, ret),
6451 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6452 return;
6454 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6455 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6456 ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6458 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6459 if (ret) {
6460 dev_info(&pf->pdev->dev,
6461 "update vsi switch failed, err %s aq_err %s\n",
6462 i40e_stat_str(&pf->hw, ret),
6463 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6468 * i40e_disable_pf_switch_lb
6469 * @pf: pointer to the PF structure
6471 * disable switch loop back or die - no point in a return value
6473 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6475 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6476 struct i40e_vsi_context ctxt;
6477 int ret;
6479 ctxt.seid = pf->main_vsi_seid;
6480 ctxt.pf_num = pf->hw.pf_id;
6481 ctxt.vf_num = 0;
6482 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6483 if (ret) {
6484 dev_info(&pf->pdev->dev,
6485 "couldn't get PF vsi config, err %s aq_err %s\n",
6486 i40e_stat_str(&pf->hw, ret),
6487 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6488 return;
6490 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6491 ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6492 ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6494 ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6495 if (ret) {
6496 dev_info(&pf->pdev->dev,
6497 "update vsi switch failed, err %s aq_err %s\n",
6498 i40e_stat_str(&pf->hw, ret),
6499 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6504 * i40e_config_bridge_mode - Configure the HW bridge mode
6505 * @veb: pointer to the bridge instance
6507 * Configure the loop back mode for the LAN VSI that is downlink to the
6508 * specified HW bridge instance. It is expected this function is called
6509 * when a new HW bridge is instantiated.
6511 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6513 struct i40e_pf *pf = veb->pf;
6515 if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6516 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6517 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6518 if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6519 i40e_disable_pf_switch_lb(pf);
6520 else
6521 i40e_enable_pf_switch_lb(pf);
6525 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6526 * @veb: pointer to the VEB instance
6528 * This is a recursive function that first builds the attached VSIs then
6529 * recurses in to build the next layer of VEB. We track the connections
6530 * through our own index numbers because the seid's from the HW could
6531 * change across the reset.
6533 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6535 struct i40e_vsi *ctl_vsi = NULL;
6536 struct i40e_pf *pf = veb->pf;
6537 int v, veb_idx;
6538 int ret;
6540 /* build VSI that owns this VEB, temporarily attached to base VEB */
6541 for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6542 if (pf->vsi[v] &&
6543 pf->vsi[v]->veb_idx == veb->idx &&
6544 pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6545 ctl_vsi = pf->vsi[v];
6546 break;
6549 if (!ctl_vsi) {
6550 dev_info(&pf->pdev->dev,
6551 "missing owner VSI for veb_idx %d\n", veb->idx);
6552 ret = -ENOENT;
6553 goto end_reconstitute;
6555 if (ctl_vsi != pf->vsi[pf->lan_vsi])
6556 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6557 ret = i40e_add_vsi(ctl_vsi);
6558 if (ret) {
6559 dev_info(&pf->pdev->dev,
6560 "rebuild of veb_idx %d owner VSI failed: %d\n",
6561 veb->idx, ret);
6562 goto end_reconstitute;
6564 i40e_vsi_reset_stats(ctl_vsi);
6566 /* create the VEB in the switch and move the VSI onto the VEB */
6567 ret = i40e_add_veb(veb, ctl_vsi);
6568 if (ret)
6569 goto end_reconstitute;
6571 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6572 veb->bridge_mode = BRIDGE_MODE_VEB;
6573 else
6574 veb->bridge_mode = BRIDGE_MODE_VEPA;
6575 i40e_config_bridge_mode(veb);
6577 /* create the remaining VSIs attached to this VEB */
6578 for (v = 0; v < pf->num_alloc_vsi; v++) {
6579 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6580 continue;
6582 if (pf->vsi[v]->veb_idx == veb->idx) {
6583 struct i40e_vsi *vsi = pf->vsi[v];
6585 vsi->uplink_seid = veb->seid;
6586 ret = i40e_add_vsi(vsi);
6587 if (ret) {
6588 dev_info(&pf->pdev->dev,
6589 "rebuild of vsi_idx %d failed: %d\n",
6590 v, ret);
6591 goto end_reconstitute;
6593 i40e_vsi_reset_stats(vsi);
6597 /* create any VEBs attached to this VEB - RECURSION */
6598 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6599 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6600 pf->veb[veb_idx]->uplink_seid = veb->seid;
6601 ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6602 if (ret)
6603 break;
6607 end_reconstitute:
6608 return ret;
6612 * i40e_get_capabilities - get info about the HW
6613 * @pf: the PF struct
6615 static int i40e_get_capabilities(struct i40e_pf *pf)
6617 struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6618 u16 data_size;
6619 int buf_len;
6620 int err;
6622 buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6623 do {
6624 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6625 if (!cap_buf)
6626 return -ENOMEM;
6628 /* this loads the data into the hw struct for us */
6629 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6630 &data_size,
6631 i40e_aqc_opc_list_func_capabilities,
6632 NULL);
6633 /* data loaded, buffer no longer needed */
6634 kfree(cap_buf);
6636 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6637 /* retry with a larger buffer */
6638 buf_len = data_size;
6639 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6640 dev_info(&pf->pdev->dev,
6641 "capability discovery failed, err %s aq_err %s\n",
6642 i40e_stat_str(&pf->hw, err),
6643 i40e_aq_str(&pf->hw,
6644 pf->hw.aq.asq_last_status));
6645 return -ENODEV;
6647 } while (err);
6649 if (pf->hw.debug_mask & I40E_DEBUG_USER)
6650 dev_info(&pf->pdev->dev,
6651 "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",
6652 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6653 pf->hw.func_caps.num_msix_vectors,
6654 pf->hw.func_caps.num_msix_vectors_vf,
6655 pf->hw.func_caps.fd_filters_guaranteed,
6656 pf->hw.func_caps.fd_filters_best_effort,
6657 pf->hw.func_caps.num_tx_qp,
6658 pf->hw.func_caps.num_vsis);
6660 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6661 + pf->hw.func_caps.num_vfs)
6662 if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6663 dev_info(&pf->pdev->dev,
6664 "got num_vsis %d, setting num_vsis to %d\n",
6665 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6666 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6669 return 0;
6672 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6675 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6676 * @pf: board private structure
6678 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6680 struct i40e_vsi *vsi;
6681 int i;
6683 /* quick workaround for an NVM issue that leaves a critical register
6684 * uninitialized
6686 if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6687 static const u32 hkey[] = {
6688 0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6689 0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6690 0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6691 0x95b3a76d};
6693 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6694 wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6697 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6698 return;
6700 /* find existing VSI and see if it needs configuring */
6701 vsi = NULL;
6702 for (i = 0; i < pf->num_alloc_vsi; i++) {
6703 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6704 vsi = pf->vsi[i];
6705 break;
6709 /* create a new VSI if none exists */
6710 if (!vsi) {
6711 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6712 pf->vsi[pf->lan_vsi]->seid, 0);
6713 if (!vsi) {
6714 dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6715 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6716 return;
6720 i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6724 * i40e_fdir_teardown - release the Flow Director resources
6725 * @pf: board private structure
6727 static void i40e_fdir_teardown(struct i40e_pf *pf)
6729 int i;
6731 i40e_fdir_filter_exit(pf);
6732 for (i = 0; i < pf->num_alloc_vsi; i++) {
6733 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6734 i40e_vsi_release(pf->vsi[i]);
6735 break;
6741 * i40e_prep_for_reset - prep for the core to reset
6742 * @pf: board private structure
6744 * Close up the VFs and other things in prep for PF Reset.
6746 static void i40e_prep_for_reset(struct i40e_pf *pf)
6748 struct i40e_hw *hw = &pf->hw;
6749 i40e_status ret = 0;
6750 u32 v;
6752 clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6753 if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6754 return;
6755 if (i40e_check_asq_alive(&pf->hw))
6756 i40e_vc_notify_reset(pf);
6758 dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6760 /* quiesce the VSIs and their queues that are not already DOWN */
6761 i40e_pf_quiesce_all_vsi(pf);
6763 for (v = 0; v < pf->num_alloc_vsi; v++) {
6764 if (pf->vsi[v])
6765 pf->vsi[v]->seid = 0;
6768 i40e_shutdown_adminq(&pf->hw);
6770 /* call shutdown HMC */
6771 if (hw->hmc.hmc_obj) {
6772 ret = i40e_shutdown_lan_hmc(hw);
6773 if (ret)
6774 dev_warn(&pf->pdev->dev,
6775 "shutdown_lan_hmc failed: %d\n", ret);
6780 * i40e_send_version - update firmware with driver version
6781 * @pf: PF struct
6783 static void i40e_send_version(struct i40e_pf *pf)
6785 struct i40e_driver_version dv;
6787 dv.major_version = DRV_VERSION_MAJOR;
6788 dv.minor_version = DRV_VERSION_MINOR;
6789 dv.build_version = DRV_VERSION_BUILD;
6790 dv.subbuild_version = 0;
6791 strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6792 i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6796 * i40e_reset_and_rebuild - reset and rebuild using a saved config
6797 * @pf: board private structure
6798 * @reinit: if the Main VSI needs to re-initialized.
6800 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6802 struct i40e_hw *hw = &pf->hw;
6803 u8 set_fc_aq_fail = 0;
6804 i40e_status ret;
6805 u32 val;
6806 u32 v;
6808 /* Now we wait for GRST to settle out.
6809 * We don't have to delete the VEBs or VSIs from the hw switch
6810 * because the reset will make them disappear.
6812 ret = i40e_pf_reset(hw);
6813 if (ret) {
6814 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6815 set_bit(__I40E_RESET_FAILED, &pf->state);
6816 goto clear_recovery;
6818 pf->pfr_count++;
6820 if (test_bit(__I40E_DOWN, &pf->state))
6821 goto clear_recovery;
6822 dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6824 /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6825 ret = i40e_init_adminq(&pf->hw);
6826 if (ret) {
6827 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6828 i40e_stat_str(&pf->hw, ret),
6829 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6830 goto clear_recovery;
6833 /* re-verify the eeprom if we just had an EMP reset */
6834 if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6835 i40e_verify_eeprom(pf);
6837 i40e_clear_pxe_mode(hw);
6838 ret = i40e_get_capabilities(pf);
6839 if (ret)
6840 goto end_core_reset;
6842 ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6843 hw->func_caps.num_rx_qp,
6844 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6845 if (ret) {
6846 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6847 goto end_core_reset;
6849 ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6850 if (ret) {
6851 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6852 goto end_core_reset;
6855 #ifdef CONFIG_I40E_DCB
6856 ret = i40e_init_pf_dcb(pf);
6857 if (ret) {
6858 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6859 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6860 /* Continue without DCB enabled */
6862 #endif /* CONFIG_I40E_DCB */
6863 #ifdef I40E_FCOE
6864 i40e_init_pf_fcoe(pf);
6866 #endif
6867 /* do basic switch setup */
6868 ret = i40e_setup_pf_switch(pf, reinit);
6869 if (ret)
6870 goto end_core_reset;
6872 /* The driver only wants link up/down and module qualification
6873 * reports from firmware. Note the negative logic.
6875 ret = i40e_aq_set_phy_int_mask(&pf->hw,
6876 ~(I40E_AQ_EVENT_LINK_UPDOWN |
6877 I40E_AQ_EVENT_MEDIA_NA |
6878 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6879 if (ret)
6880 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6881 i40e_stat_str(&pf->hw, ret),
6882 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6884 /* make sure our flow control settings are restored */
6885 ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6886 if (ret)
6887 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6888 i40e_stat_str(&pf->hw, ret),
6889 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6891 /* Rebuild the VSIs and VEBs that existed before reset.
6892 * They are still in our local switch element arrays, so only
6893 * need to rebuild the switch model in the HW.
6895 * If there were VEBs but the reconstitution failed, we'll try
6896 * try to recover minimal use by getting the basic PF VSI working.
6898 if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6899 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6900 /* find the one VEB connected to the MAC, and find orphans */
6901 for (v = 0; v < I40E_MAX_VEB; v++) {
6902 if (!pf->veb[v])
6903 continue;
6905 if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6906 pf->veb[v]->uplink_seid == 0) {
6907 ret = i40e_reconstitute_veb(pf->veb[v]);
6909 if (!ret)
6910 continue;
6912 /* If Main VEB failed, we're in deep doodoo,
6913 * so give up rebuilding the switch and set up
6914 * for minimal rebuild of PF VSI.
6915 * If orphan failed, we'll report the error
6916 * but try to keep going.
6918 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6919 dev_info(&pf->pdev->dev,
6920 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6921 ret);
6922 pf->vsi[pf->lan_vsi]->uplink_seid
6923 = pf->mac_seid;
6924 break;
6925 } else if (pf->veb[v]->uplink_seid == 0) {
6926 dev_info(&pf->pdev->dev,
6927 "rebuild of orphan VEB failed: %d\n",
6928 ret);
6934 if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6935 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6936 /* no VEB, so rebuild only the Main VSI */
6937 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6938 if (ret) {
6939 dev_info(&pf->pdev->dev,
6940 "rebuild of Main VSI failed: %d\n", ret);
6941 goto end_core_reset;
6945 /* Reconfigure hardware for allowing smaller MSS in the case
6946 * of TSO, so that we avoid the MDD being fired and causing
6947 * a reset in the case of small MSS+TSO.
6949 #define I40E_REG_MSS 0x000E64DC
6950 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6951 #define I40E_64BYTE_MSS 0x400000
6952 val = rd32(hw, I40E_REG_MSS);
6953 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6954 val &= ~I40E_REG_MSS_MIN_MASK;
6955 val |= I40E_64BYTE_MSS;
6956 wr32(hw, I40E_REG_MSS, val);
6959 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6960 msleep(75);
6961 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6962 if (ret)
6963 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6964 i40e_stat_str(&pf->hw, ret),
6965 i40e_aq_str(&pf->hw,
6966 pf->hw.aq.asq_last_status));
6968 /* reinit the misc interrupt */
6969 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6970 ret = i40e_setup_misc_vector(pf);
6972 /* Add a filter to drop all Flow control frames from any VSI from being
6973 * transmitted. By doing so we stop a malicious VF from sending out
6974 * PAUSE or PFC frames and potentially controlling traffic for other
6975 * PF/VF VSIs.
6976 * The FW can still send Flow control frames if enabled.
6978 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6979 pf->main_vsi_seid);
6981 /* restart the VSIs that were rebuilt and running before the reset */
6982 i40e_pf_unquiesce_all_vsi(pf);
6984 if (pf->num_alloc_vfs) {
6985 for (v = 0; v < pf->num_alloc_vfs; v++)
6986 i40e_reset_vf(&pf->vf[v], true);
6989 /* tell the firmware that we're starting */
6990 i40e_send_version(pf);
6992 end_core_reset:
6993 clear_bit(__I40E_RESET_FAILED, &pf->state);
6994 clear_recovery:
6995 clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6999 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7000 * @pf: board private structure
7002 * Close up the VFs and other things in prep for a Core Reset,
7003 * then get ready to rebuild the world.
7005 static void i40e_handle_reset_warning(struct i40e_pf *pf)
7007 i40e_prep_for_reset(pf);
7008 i40e_reset_and_rebuild(pf, false);
7012 * i40e_handle_mdd_event
7013 * @pf: pointer to the PF structure
7015 * Called from the MDD irq handler to identify possibly malicious vfs
7017 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7019 struct i40e_hw *hw = &pf->hw;
7020 bool mdd_detected = false;
7021 bool pf_mdd_detected = false;
7022 struct i40e_vf *vf;
7023 u32 reg;
7024 int i;
7026 if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7027 return;
7029 /* find what triggered the MDD event */
7030 reg = rd32(hw, I40E_GL_MDET_TX);
7031 if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7032 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7033 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7034 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7035 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7036 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7037 I40E_GL_MDET_TX_EVENT_SHIFT;
7038 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7039 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7040 pf->hw.func_caps.base_queue;
7041 if (netif_msg_tx_err(pf))
7042 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7043 event, queue, pf_num, vf_num);
7044 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7045 mdd_detected = true;
7047 reg = rd32(hw, I40E_GL_MDET_RX);
7048 if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7049 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7050 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7051 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7052 I40E_GL_MDET_RX_EVENT_SHIFT;
7053 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7054 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7055 pf->hw.func_caps.base_queue;
7056 if (netif_msg_rx_err(pf))
7057 dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7058 event, queue, func);
7059 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7060 mdd_detected = true;
7063 if (mdd_detected) {
7064 reg = rd32(hw, I40E_PF_MDET_TX);
7065 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7066 wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7067 dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7068 pf_mdd_detected = true;
7070 reg = rd32(hw, I40E_PF_MDET_RX);
7071 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7072 wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7073 dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7074 pf_mdd_detected = true;
7076 /* Queue belongs to the PF, initiate a reset */
7077 if (pf_mdd_detected) {
7078 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7079 i40e_service_event_schedule(pf);
7083 /* see if one of the VFs needs its hand slapped */
7084 for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7085 vf = &(pf->vf[i]);
7086 reg = rd32(hw, I40E_VP_MDET_TX(i));
7087 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7088 wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7089 vf->num_mdd_events++;
7090 dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7094 reg = rd32(hw, I40E_VP_MDET_RX(i));
7095 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7096 wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7097 vf->num_mdd_events++;
7098 dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7102 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7103 dev_info(&pf->pdev->dev,
7104 "Too many MDD events on VF %d, disabled\n", i);
7105 dev_info(&pf->pdev->dev,
7106 "Use PF Control I/F to re-enable the VF\n");
7107 set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7111 /* re-enable mdd interrupt cause */
7112 clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7113 reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7114 reg |= I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7115 wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7116 i40e_flush(hw);
7120 * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7121 * @pf: board private structure
7123 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7125 struct i40e_hw *hw = &pf->hw;
7126 i40e_status ret;
7127 __be16 port;
7128 int i;
7130 if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7131 return;
7133 pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7135 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7136 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7137 pf->pending_udp_bitmap &= ~BIT_ULL(i);
7138 port = pf->udp_ports[i].index;
7139 if (port)
7140 ret = i40e_aq_add_udp_tunnel(hw, port,
7141 pf->udp_ports[i].type,
7142 NULL, NULL);
7143 else
7144 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7146 if (ret) {
7147 dev_dbg(&pf->pdev->dev,
7148 "%s %s port %d, index %d failed, err %s aq_err %s\n",
7149 pf->udp_ports[i].type ? "vxlan" : "geneve",
7150 port ? "add" : "delete",
7151 ntohs(port), i,
7152 i40e_stat_str(&pf->hw, ret),
7153 i40e_aq_str(&pf->hw,
7154 pf->hw.aq.asq_last_status));
7155 pf->udp_ports[i].index = 0;
7162 * i40e_service_task - Run the driver's async subtasks
7163 * @work: pointer to work_struct containing our data
7165 static void i40e_service_task(struct work_struct *work)
7167 struct i40e_pf *pf = container_of(work,
7168 struct i40e_pf,
7169 service_task);
7170 unsigned long start_time = jiffies;
7172 /* don't bother with service tasks if a reset is in progress */
7173 if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7174 i40e_service_event_complete(pf);
7175 return;
7178 i40e_detect_recover_hung(pf);
7179 i40e_sync_filters_subtask(pf);
7180 i40e_reset_subtask(pf);
7181 i40e_handle_mdd_event(pf);
7182 i40e_vc_process_vflr_event(pf);
7183 i40e_watchdog_subtask(pf);
7184 i40e_fdir_reinit_subtask(pf);
7185 i40e_client_subtask(pf);
7186 i40e_sync_filters_subtask(pf);
7187 i40e_sync_udp_filters_subtask(pf);
7188 i40e_clean_adminq_subtask(pf);
7190 i40e_service_event_complete(pf);
7192 /* If the tasks have taken longer than one timer cycle or there
7193 * is more work to be done, reschedule the service task now
7194 * rather than wait for the timer to tick again.
7196 if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7197 test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state) ||
7198 test_bit(__I40E_MDD_EVENT_PENDING, &pf->state) ||
7199 test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7200 i40e_service_event_schedule(pf);
7204 * i40e_service_timer - timer callback
7205 * @data: pointer to PF struct
7207 static void i40e_service_timer(unsigned long data)
7209 struct i40e_pf *pf = (struct i40e_pf *)data;
7211 mod_timer(&pf->service_timer,
7212 round_jiffies(jiffies + pf->service_timer_period));
7213 i40e_service_event_schedule(pf);
7217 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7218 * @vsi: the VSI being configured
7220 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7222 struct i40e_pf *pf = vsi->back;
7224 switch (vsi->type) {
7225 case I40E_VSI_MAIN:
7226 vsi->alloc_queue_pairs = pf->num_lan_qps;
7227 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7228 I40E_REQ_DESCRIPTOR_MULTIPLE);
7229 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7230 vsi->num_q_vectors = pf->num_lan_msix;
7231 else
7232 vsi->num_q_vectors = 1;
7234 break;
7236 case I40E_VSI_FDIR:
7237 vsi->alloc_queue_pairs = 1;
7238 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7239 I40E_REQ_DESCRIPTOR_MULTIPLE);
7240 vsi->num_q_vectors = pf->num_fdsb_msix;
7241 break;
7243 case I40E_VSI_VMDQ2:
7244 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7245 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7246 I40E_REQ_DESCRIPTOR_MULTIPLE);
7247 vsi->num_q_vectors = pf->num_vmdq_msix;
7248 break;
7250 case I40E_VSI_SRIOV:
7251 vsi->alloc_queue_pairs = pf->num_vf_qps;
7252 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7253 I40E_REQ_DESCRIPTOR_MULTIPLE);
7254 break;
7256 #ifdef I40E_FCOE
7257 case I40E_VSI_FCOE:
7258 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7259 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7260 I40E_REQ_DESCRIPTOR_MULTIPLE);
7261 vsi->num_q_vectors = pf->num_fcoe_msix;
7262 break;
7264 #endif /* I40E_FCOE */
7265 default:
7266 WARN_ON(1);
7267 return -ENODATA;
7270 return 0;
7274 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7275 * @type: VSI pointer
7276 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7278 * On error: returns error code (negative)
7279 * On success: returns 0
7281 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7283 int size;
7284 int ret = 0;
7286 /* allocate memory for both Tx and Rx ring pointers */
7287 size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7288 vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7289 if (!vsi->tx_rings)
7290 return -ENOMEM;
7291 vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7293 if (alloc_qvectors) {
7294 /* allocate memory for q_vector pointers */
7295 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7296 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7297 if (!vsi->q_vectors) {
7298 ret = -ENOMEM;
7299 goto err_vectors;
7302 return ret;
7304 err_vectors:
7305 kfree(vsi->tx_rings);
7306 return ret;
7310 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7311 * @pf: board private structure
7312 * @type: type of VSI
7314 * On error: returns error code (negative)
7315 * On success: returns vsi index in PF (positive)
7317 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7319 int ret = -ENODEV;
7320 struct i40e_vsi *vsi;
7321 int vsi_idx;
7322 int i;
7324 /* Need to protect the allocation of the VSIs at the PF level */
7325 mutex_lock(&pf->switch_mutex);
7327 /* VSI list may be fragmented if VSI creation/destruction has
7328 * been happening. We can afford to do a quick scan to look
7329 * for any free VSIs in the list.
7331 * find next empty vsi slot, looping back around if necessary
7333 i = pf->next_vsi;
7334 while (i < pf->num_alloc_vsi && pf->vsi[i])
7335 i++;
7336 if (i >= pf->num_alloc_vsi) {
7337 i = 0;
7338 while (i < pf->next_vsi && pf->vsi[i])
7339 i++;
7342 if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7343 vsi_idx = i; /* Found one! */
7344 } else {
7345 ret = -ENODEV;
7346 goto unlock_pf; /* out of VSI slots! */
7348 pf->next_vsi = ++i;
7350 vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7351 if (!vsi) {
7352 ret = -ENOMEM;
7353 goto unlock_pf;
7355 vsi->type = type;
7356 vsi->back = pf;
7357 set_bit(__I40E_DOWN, &vsi->state);
7358 vsi->flags = 0;
7359 vsi->idx = vsi_idx;
7360 vsi->int_rate_limit = 0;
7361 vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7362 pf->rss_table_size : 64;
7363 vsi->netdev_registered = false;
7364 vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7365 INIT_LIST_HEAD(&vsi->mac_filter_list);
7366 vsi->irqs_ready = false;
7368 ret = i40e_set_num_rings_in_vsi(vsi);
7369 if (ret)
7370 goto err_rings;
7372 ret = i40e_vsi_alloc_arrays(vsi, true);
7373 if (ret)
7374 goto err_rings;
7376 /* Setup default MSIX irq handler for VSI */
7377 i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7379 /* Initialize VSI lock */
7380 spin_lock_init(&vsi->mac_filter_list_lock);
7381 pf->vsi[vsi_idx] = vsi;
7382 ret = vsi_idx;
7383 goto unlock_pf;
7385 err_rings:
7386 pf->next_vsi = i - 1;
7387 kfree(vsi);
7388 unlock_pf:
7389 mutex_unlock(&pf->switch_mutex);
7390 return ret;
7394 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7395 * @type: VSI pointer
7396 * @free_qvectors: a bool to specify if q_vectors need to be freed.
7398 * On error: returns error code (negative)
7399 * On success: returns 0
7401 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7403 /* free the ring and vector containers */
7404 if (free_qvectors) {
7405 kfree(vsi->q_vectors);
7406 vsi->q_vectors = NULL;
7408 kfree(vsi->tx_rings);
7409 vsi->tx_rings = NULL;
7410 vsi->rx_rings = NULL;
7414 * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7415 * and lookup table
7416 * @vsi: Pointer to VSI structure
7418 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7420 if (!vsi)
7421 return;
7423 kfree(vsi->rss_hkey_user);
7424 vsi->rss_hkey_user = NULL;
7426 kfree(vsi->rss_lut_user);
7427 vsi->rss_lut_user = NULL;
7431 * i40e_vsi_clear - Deallocate the VSI provided
7432 * @vsi: the VSI being un-configured
7434 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7436 struct i40e_pf *pf;
7438 if (!vsi)
7439 return 0;
7441 if (!vsi->back)
7442 goto free_vsi;
7443 pf = vsi->back;
7445 mutex_lock(&pf->switch_mutex);
7446 if (!pf->vsi[vsi->idx]) {
7447 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7448 vsi->idx, vsi->idx, vsi, vsi->type);
7449 goto unlock_vsi;
7452 if (pf->vsi[vsi->idx] != vsi) {
7453 dev_err(&pf->pdev->dev,
7454 "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7455 pf->vsi[vsi->idx]->idx,
7456 pf->vsi[vsi->idx],
7457 pf->vsi[vsi->idx]->type,
7458 vsi->idx, vsi, vsi->type);
7459 goto unlock_vsi;
7462 /* updates the PF for this cleared vsi */
7463 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7464 i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7466 i40e_vsi_free_arrays(vsi, true);
7467 i40e_clear_rss_config_user(vsi);
7469 pf->vsi[vsi->idx] = NULL;
7470 if (vsi->idx < pf->next_vsi)
7471 pf->next_vsi = vsi->idx;
7473 unlock_vsi:
7474 mutex_unlock(&pf->switch_mutex);
7475 free_vsi:
7476 kfree(vsi);
7478 return 0;
7482 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7483 * @vsi: the VSI being cleaned
7485 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7487 int i;
7489 if (vsi->tx_rings && vsi->tx_rings[0]) {
7490 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7491 kfree_rcu(vsi->tx_rings[i], rcu);
7492 vsi->tx_rings[i] = NULL;
7493 vsi->rx_rings[i] = NULL;
7499 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7500 * @vsi: the VSI being configured
7502 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7504 struct i40e_ring *tx_ring, *rx_ring;
7505 struct i40e_pf *pf = vsi->back;
7506 int i;
7508 /* Set basic values in the rings to be used later during open() */
7509 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7510 /* allocate space for both Tx and Rx in one shot */
7511 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7512 if (!tx_ring)
7513 goto err_out;
7515 tx_ring->queue_index = i;
7516 tx_ring->reg_idx = vsi->base_queue + i;
7517 tx_ring->ring_active = false;
7518 tx_ring->vsi = vsi;
7519 tx_ring->netdev = vsi->netdev;
7520 tx_ring->dev = &pf->pdev->dev;
7521 tx_ring->count = vsi->num_desc;
7522 tx_ring->size = 0;
7523 tx_ring->dcb_tc = 0;
7524 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7525 tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7526 tx_ring->tx_itr_setting = pf->tx_itr_default;
7527 vsi->tx_rings[i] = tx_ring;
7529 rx_ring = &tx_ring[1];
7530 rx_ring->queue_index = i;
7531 rx_ring->reg_idx = vsi->base_queue + i;
7532 rx_ring->ring_active = false;
7533 rx_ring->vsi = vsi;
7534 rx_ring->netdev = vsi->netdev;
7535 rx_ring->dev = &pf->pdev->dev;
7536 rx_ring->count = vsi->num_desc;
7537 rx_ring->size = 0;
7538 rx_ring->dcb_tc = 0;
7539 rx_ring->rx_itr_setting = pf->rx_itr_default;
7540 vsi->rx_rings[i] = rx_ring;
7543 return 0;
7545 err_out:
7546 i40e_vsi_clear_rings(vsi);
7547 return -ENOMEM;
7551 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7552 * @pf: board private structure
7553 * @vectors: the number of MSI-X vectors to request
7555 * Returns the number of vectors reserved, or error
7557 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7559 vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7560 I40E_MIN_MSIX, vectors);
7561 if (vectors < 0) {
7562 dev_info(&pf->pdev->dev,
7563 "MSI-X vector reservation failed: %d\n", vectors);
7564 vectors = 0;
7567 return vectors;
7571 * i40e_init_msix - Setup the MSIX capability
7572 * @pf: board private structure
7574 * Work with the OS to set up the MSIX vectors needed.
7576 * Returns the number of vectors reserved or negative on failure
7578 static int i40e_init_msix(struct i40e_pf *pf)
7580 struct i40e_hw *hw = &pf->hw;
7581 int vectors_left;
7582 int v_budget, i;
7583 int v_actual;
7584 int iwarp_requested = 0;
7586 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7587 return -ENODEV;
7589 /* The number of vectors we'll request will be comprised of:
7590 * - Add 1 for "other" cause for Admin Queue events, etc.
7591 * - The number of LAN queue pairs
7592 * - Queues being used for RSS.
7593 * We don't need as many as max_rss_size vectors.
7594 * use rss_size instead in the calculation since that
7595 * is governed by number of cpus in the system.
7596 * - assumes symmetric Tx/Rx pairing
7597 * - The number of VMDq pairs
7598 * - The CPU count within the NUMA node if iWARP is enabled
7599 #ifdef I40E_FCOE
7600 * - The number of FCOE qps.
7601 #endif
7602 * Once we count this up, try the request.
7604 * If we can't get what we want, we'll simplify to nearly nothing
7605 * and try again. If that still fails, we punt.
7607 vectors_left = hw->func_caps.num_msix_vectors;
7608 v_budget = 0;
7610 /* reserve one vector for miscellaneous handler */
7611 if (vectors_left) {
7612 v_budget++;
7613 vectors_left--;
7616 /* reserve vectors for the main PF traffic queues */
7617 pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7618 vectors_left -= pf->num_lan_msix;
7619 v_budget += pf->num_lan_msix;
7621 /* reserve one vector for sideband flow director */
7622 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7623 if (vectors_left) {
7624 pf->num_fdsb_msix = 1;
7625 v_budget++;
7626 vectors_left--;
7627 } else {
7628 pf->num_fdsb_msix = 0;
7632 #ifdef I40E_FCOE
7633 /* can we reserve enough for FCoE? */
7634 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7635 if (!vectors_left)
7636 pf->num_fcoe_msix = 0;
7637 else if (vectors_left >= pf->num_fcoe_qps)
7638 pf->num_fcoe_msix = pf->num_fcoe_qps;
7639 else
7640 pf->num_fcoe_msix = 1;
7641 v_budget += pf->num_fcoe_msix;
7642 vectors_left -= pf->num_fcoe_msix;
7645 #endif
7646 /* can we reserve enough for iWARP? */
7647 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7648 iwarp_requested = pf->num_iwarp_msix;
7650 if (!vectors_left)
7651 pf->num_iwarp_msix = 0;
7652 else if (vectors_left < pf->num_iwarp_msix)
7653 pf->num_iwarp_msix = 1;
7654 v_budget += pf->num_iwarp_msix;
7655 vectors_left -= pf->num_iwarp_msix;
7658 /* any vectors left over go for VMDq support */
7659 if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7660 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7661 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7663 if (!vectors_left) {
7664 pf->num_vmdq_msix = 0;
7665 pf->num_vmdq_qps = 0;
7666 } else {
7667 /* if we're short on vectors for what's desired, we limit
7668 * the queues per vmdq. If this is still more than are
7669 * available, the user will need to change the number of
7670 * queues/vectors used by the PF later with the ethtool
7671 * channels command
7673 if (vmdq_vecs < vmdq_vecs_wanted)
7674 pf->num_vmdq_qps = 1;
7675 pf->num_vmdq_msix = pf->num_vmdq_qps;
7677 v_budget += vmdq_vecs;
7678 vectors_left -= vmdq_vecs;
7682 pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7683 GFP_KERNEL);
7684 if (!pf->msix_entries)
7685 return -ENOMEM;
7687 for (i = 0; i < v_budget; i++)
7688 pf->msix_entries[i].entry = i;
7689 v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7691 if (v_actual < I40E_MIN_MSIX) {
7692 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7693 kfree(pf->msix_entries);
7694 pf->msix_entries = NULL;
7695 pci_disable_msix(pf->pdev);
7696 return -ENODEV;
7698 } else if (v_actual == I40E_MIN_MSIX) {
7699 /* Adjust for minimal MSIX use */
7700 pf->num_vmdq_vsis = 0;
7701 pf->num_vmdq_qps = 0;
7702 pf->num_lan_qps = 1;
7703 pf->num_lan_msix = 1;
7705 } else if (!vectors_left) {
7706 /* If we have limited resources, we will start with no vectors
7707 * for the special features and then allocate vectors to some
7708 * of these features based on the policy and at the end disable
7709 * the features that did not get any vectors.
7711 int vec;
7713 dev_info(&pf->pdev->dev,
7714 "MSI-X vector limit reached, attempting to redistribute vectors\n");
7715 /* reserve the misc vector */
7716 vec = v_actual - 1;
7718 /* Scale vector usage down */
7719 pf->num_vmdq_msix = 1; /* force VMDqs to only one vector */
7720 pf->num_vmdq_vsis = 1;
7721 pf->num_vmdq_qps = 1;
7722 #ifdef I40E_FCOE
7723 pf->num_fcoe_qps = 0;
7724 pf->num_fcoe_msix = 0;
7725 #endif
7727 /* partition out the remaining vectors */
7728 switch (vec) {
7729 case 2:
7730 pf->num_lan_msix = 1;
7731 break;
7732 case 3:
7733 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7734 pf->num_lan_msix = 1;
7735 pf->num_iwarp_msix = 1;
7736 } else {
7737 pf->num_lan_msix = 2;
7739 #ifdef I40E_FCOE
7740 /* give one vector to FCoE */
7741 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7742 pf->num_lan_msix = 1;
7743 pf->num_fcoe_msix = 1;
7745 #endif
7746 break;
7747 default:
7748 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7749 pf->num_iwarp_msix = min_t(int, (vec / 3),
7750 iwarp_requested);
7751 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7752 I40E_DEFAULT_NUM_VMDQ_VSI);
7753 } else {
7754 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7755 I40E_DEFAULT_NUM_VMDQ_VSI);
7757 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7758 pf->num_fdsb_msix = 1;
7759 vec--;
7761 pf->num_lan_msix = min_t(int,
7762 (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7763 pf->num_lan_msix);
7764 pf->num_lan_qps = pf->num_lan_msix;
7765 #ifdef I40E_FCOE
7766 /* give one vector to FCoE */
7767 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7768 pf->num_fcoe_msix = 1;
7769 vec--;
7771 #endif
7772 break;
7776 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
7777 (pf->num_fdsb_msix == 0)) {
7778 dev_info(&pf->pdev->dev, "Sideband Flowdir disabled, not enough MSI-X vectors\n");
7779 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7781 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7782 (pf->num_vmdq_msix == 0)) {
7783 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7784 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7787 if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7788 (pf->num_iwarp_msix == 0)) {
7789 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7790 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7792 #ifdef I40E_FCOE
7794 if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7795 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7796 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7798 #endif
7799 i40e_debug(&pf->hw, I40E_DEBUG_INIT,
7800 "MSI-X vector distribution: PF %d, VMDq %d, FDSB %d, iWARP %d\n",
7801 pf->num_lan_msix,
7802 pf->num_vmdq_msix * pf->num_vmdq_vsis,
7803 pf->num_fdsb_msix,
7804 pf->num_iwarp_msix);
7806 return v_actual;
7810 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7811 * @vsi: the VSI being configured
7812 * @v_idx: index of the vector in the vsi struct
7813 * @cpu: cpu to be used on affinity_mask
7815 * We allocate one q_vector. If allocation fails we return -ENOMEM.
7817 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7819 struct i40e_q_vector *q_vector;
7821 /* allocate q_vector */
7822 q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7823 if (!q_vector)
7824 return -ENOMEM;
7826 q_vector->vsi = vsi;
7827 q_vector->v_idx = v_idx;
7828 cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7830 if (vsi->netdev)
7831 netif_napi_add(vsi->netdev, &q_vector->napi,
7832 i40e_napi_poll, NAPI_POLL_WEIGHT);
7834 q_vector->rx.latency_range = I40E_LOW_LATENCY;
7835 q_vector->tx.latency_range = I40E_LOW_LATENCY;
7837 /* tie q_vector and vsi together */
7838 vsi->q_vectors[v_idx] = q_vector;
7840 return 0;
7844 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7845 * @vsi: the VSI being configured
7847 * We allocate one q_vector per queue interrupt. If allocation fails we
7848 * return -ENOMEM.
7850 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7852 struct i40e_pf *pf = vsi->back;
7853 int err, v_idx, num_q_vectors, current_cpu;
7855 /* if not MSIX, give the one vector only to the LAN VSI */
7856 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7857 num_q_vectors = vsi->num_q_vectors;
7858 else if (vsi == pf->vsi[pf->lan_vsi])
7859 num_q_vectors = 1;
7860 else
7861 return -EINVAL;
7863 current_cpu = cpumask_first(cpu_online_mask);
7865 for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7866 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7867 if (err)
7868 goto err_out;
7869 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7870 if (unlikely(current_cpu >= nr_cpu_ids))
7871 current_cpu = cpumask_first(cpu_online_mask);
7874 return 0;
7876 err_out:
7877 while (v_idx--)
7878 i40e_free_q_vector(vsi, v_idx);
7880 return err;
7884 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7885 * @pf: board private structure to initialize
7887 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7889 int vectors = 0;
7890 ssize_t size;
7892 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7893 vectors = i40e_init_msix(pf);
7894 if (vectors < 0) {
7895 pf->flags &= ~(I40E_FLAG_MSIX_ENABLED |
7896 I40E_FLAG_IWARP_ENABLED |
7897 #ifdef I40E_FCOE
7898 I40E_FLAG_FCOE_ENABLED |
7899 #endif
7900 I40E_FLAG_RSS_ENABLED |
7901 I40E_FLAG_DCB_CAPABLE |
7902 I40E_FLAG_DCB_ENABLED |
7903 I40E_FLAG_SRIOV_ENABLED |
7904 I40E_FLAG_FD_SB_ENABLED |
7905 I40E_FLAG_FD_ATR_ENABLED |
7906 I40E_FLAG_VMDQ_ENABLED);
7908 /* rework the queue expectations without MSIX */
7909 i40e_determine_queue_usage(pf);
7913 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7914 (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7915 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7916 vectors = pci_enable_msi(pf->pdev);
7917 if (vectors < 0) {
7918 dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7919 vectors);
7920 pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7922 vectors = 1; /* one MSI or Legacy vector */
7925 if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7926 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7928 /* set up vector assignment tracking */
7929 size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7930 pf->irq_pile = kzalloc(size, GFP_KERNEL);
7931 if (!pf->irq_pile) {
7932 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7933 return -ENOMEM;
7935 pf->irq_pile->num_entries = vectors;
7936 pf->irq_pile->search_hint = 0;
7938 /* track first vector for misc interrupts, ignore return */
7939 (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7941 return 0;
7945 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7946 * @pf: board private structure
7948 * This sets up the handler for MSIX 0, which is used to manage the
7949 * non-queue interrupts, e.g. AdminQ and errors. This is not used
7950 * when in MSI or Legacy interrupt mode.
7952 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7954 struct i40e_hw *hw = &pf->hw;
7955 int err = 0;
7957 /* Only request the irq if this is the first time through, and
7958 * not when we're rebuilding after a Reset
7960 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7961 err = request_irq(pf->msix_entries[0].vector,
7962 i40e_intr, 0, pf->int_name, pf);
7963 if (err) {
7964 dev_info(&pf->pdev->dev,
7965 "request_irq for %s failed: %d\n",
7966 pf->int_name, err);
7967 return -EFAULT;
7971 i40e_enable_misc_int_causes(pf);
7973 /* associate no queues to the misc vector */
7974 wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7975 wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7977 i40e_flush(hw);
7979 i40e_irq_dynamic_enable_icr0(pf, true);
7981 return err;
7985 * i40e_config_rss_aq - Prepare for RSS using AQ commands
7986 * @vsi: vsi structure
7987 * @seed: RSS hash seed
7989 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7990 u8 *lut, u16 lut_size)
7992 struct i40e_pf *pf = vsi->back;
7993 struct i40e_hw *hw = &pf->hw;
7994 int ret = 0;
7996 if (seed) {
7997 struct i40e_aqc_get_set_rss_key_data *seed_dw =
7998 (struct i40e_aqc_get_set_rss_key_data *)seed;
7999 ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8000 if (ret) {
8001 dev_info(&pf->pdev->dev,
8002 "Cannot set RSS key, err %s aq_err %s\n",
8003 i40e_stat_str(hw, ret),
8004 i40e_aq_str(hw, hw->aq.asq_last_status));
8005 return ret;
8008 if (lut) {
8009 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8011 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8012 if (ret) {
8013 dev_info(&pf->pdev->dev,
8014 "Cannot set RSS lut, err %s aq_err %s\n",
8015 i40e_stat_str(hw, ret),
8016 i40e_aq_str(hw, hw->aq.asq_last_status));
8017 return ret;
8020 return ret;
8024 * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8025 * @vsi: Pointer to vsi structure
8026 * @seed: Buffter to store the hash keys
8027 * @lut: Buffer to store the lookup table entries
8028 * @lut_size: Size of buffer to store the lookup table entries
8030 * Return 0 on success, negative on failure
8032 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8033 u8 *lut, u16 lut_size)
8035 struct i40e_pf *pf = vsi->back;
8036 struct i40e_hw *hw = &pf->hw;
8037 int ret = 0;
8039 if (seed) {
8040 ret = i40e_aq_get_rss_key(hw, vsi->id,
8041 (struct i40e_aqc_get_set_rss_key_data *)seed);
8042 if (ret) {
8043 dev_info(&pf->pdev->dev,
8044 "Cannot get RSS key, err %s aq_err %s\n",
8045 i40e_stat_str(&pf->hw, ret),
8046 i40e_aq_str(&pf->hw,
8047 pf->hw.aq.asq_last_status));
8048 return ret;
8052 if (lut) {
8053 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8055 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8056 if (ret) {
8057 dev_info(&pf->pdev->dev,
8058 "Cannot get RSS lut, err %s aq_err %s\n",
8059 i40e_stat_str(&pf->hw, ret),
8060 i40e_aq_str(&pf->hw,
8061 pf->hw.aq.asq_last_status));
8062 return ret;
8066 return ret;
8070 * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8071 * @vsi: VSI structure
8073 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8075 u8 seed[I40E_HKEY_ARRAY_SIZE];
8076 struct i40e_pf *pf = vsi->back;
8077 u8 *lut;
8078 int ret;
8080 if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8081 return 0;
8083 if (!vsi->rss_size)
8084 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8085 vsi->num_queue_pairs);
8086 if (!vsi->rss_size)
8087 return -EINVAL;
8089 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8090 if (!lut)
8091 return -ENOMEM;
8092 /* Use the user configured hash keys and lookup table if there is one,
8093 * otherwise use default
8095 if (vsi->rss_lut_user)
8096 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8097 else
8098 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8099 if (vsi->rss_hkey_user)
8100 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8101 else
8102 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8103 ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8104 kfree(lut);
8106 return ret;
8110 * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8111 * @vsi: Pointer to vsi structure
8112 * @seed: RSS hash seed
8113 * @lut: Lookup table
8114 * @lut_size: Lookup table size
8116 * Returns 0 on success, negative on failure
8118 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8119 const u8 *lut, u16 lut_size)
8121 struct i40e_pf *pf = vsi->back;
8122 struct i40e_hw *hw = &pf->hw;
8123 u16 vf_id = vsi->vf_id;
8124 u8 i;
8126 /* Fill out hash function seed */
8127 if (seed) {
8128 u32 *seed_dw = (u32 *)seed;
8130 if (vsi->type == I40E_VSI_MAIN) {
8131 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8132 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8133 seed_dw[i]);
8134 } else if (vsi->type == I40E_VSI_SRIOV) {
8135 for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8136 i40e_write_rx_ctl(hw,
8137 I40E_VFQF_HKEY1(i, vf_id),
8138 seed_dw[i]);
8139 } else {
8140 dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8144 if (lut) {
8145 u32 *lut_dw = (u32 *)lut;
8147 if (vsi->type == I40E_VSI_MAIN) {
8148 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8149 return -EINVAL;
8150 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8151 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8152 } else if (vsi->type == I40E_VSI_SRIOV) {
8153 if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8154 return -EINVAL;
8155 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8156 i40e_write_rx_ctl(hw,
8157 I40E_VFQF_HLUT1(i, vf_id),
8158 lut_dw[i]);
8159 } else {
8160 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8163 i40e_flush(hw);
8165 return 0;
8169 * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8170 * @vsi: Pointer to VSI structure
8171 * @seed: Buffer to store the keys
8172 * @lut: Buffer to store the lookup table entries
8173 * @lut_size: Size of buffer to store the lookup table entries
8175 * Returns 0 on success, negative on failure
8177 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8178 u8 *lut, u16 lut_size)
8180 struct i40e_pf *pf = vsi->back;
8181 struct i40e_hw *hw = &pf->hw;
8182 u16 i;
8184 if (seed) {
8185 u32 *seed_dw = (u32 *)seed;
8187 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8188 seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8190 if (lut) {
8191 u32 *lut_dw = (u32 *)lut;
8193 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8194 return -EINVAL;
8195 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8196 lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8199 return 0;
8203 * i40e_config_rss - Configure RSS keys and lut
8204 * @vsi: Pointer to VSI structure
8205 * @seed: RSS hash seed
8206 * @lut: Lookup table
8207 * @lut_size: Lookup table size
8209 * Returns 0 on success, negative on failure
8211 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8213 struct i40e_pf *pf = vsi->back;
8215 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8216 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8217 else
8218 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8222 * i40e_get_rss - Get RSS keys and lut
8223 * @vsi: Pointer to VSI structure
8224 * @seed: Buffer to store the keys
8225 * @lut: Buffer to store the lookup table entries
8226 * lut_size: Size of buffer to store the lookup table entries
8228 * Returns 0 on success, negative on failure
8230 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8232 struct i40e_pf *pf = vsi->back;
8234 if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8235 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8236 else
8237 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8241 * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8242 * @pf: Pointer to board private structure
8243 * @lut: Lookup table
8244 * @rss_table_size: Lookup table size
8245 * @rss_size: Range of queue number for hashing
8247 void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8248 u16 rss_table_size, u16 rss_size)
8250 u16 i;
8252 for (i = 0; i < rss_table_size; i++)
8253 lut[i] = i % rss_size;
8257 * i40e_pf_config_rss - Prepare for RSS if used
8258 * @pf: board private structure
8260 static int i40e_pf_config_rss(struct i40e_pf *pf)
8262 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8263 u8 seed[I40E_HKEY_ARRAY_SIZE];
8264 u8 *lut;
8265 struct i40e_hw *hw = &pf->hw;
8266 u32 reg_val;
8267 u64 hena;
8268 int ret;
8270 /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8271 hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8272 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8273 hena |= i40e_pf_get_default_rss_hena(pf);
8275 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8276 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8278 /* Determine the RSS table size based on the hardware capabilities */
8279 reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8280 reg_val = (pf->rss_table_size == 512) ?
8281 (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8282 (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8283 i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8285 /* Determine the RSS size of the VSI */
8286 if (!vsi->rss_size)
8287 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8288 vsi->num_queue_pairs);
8289 if (!vsi->rss_size)
8290 return -EINVAL;
8292 lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8293 if (!lut)
8294 return -ENOMEM;
8296 /* Use user configured lut if there is one, otherwise use default */
8297 if (vsi->rss_lut_user)
8298 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8299 else
8300 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8302 /* Use user configured hash key if there is one, otherwise
8303 * use default.
8305 if (vsi->rss_hkey_user)
8306 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8307 else
8308 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8309 ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8310 kfree(lut);
8312 return ret;
8316 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8317 * @pf: board private structure
8318 * @queue_count: the requested queue count for rss.
8320 * returns 0 if rss is not enabled, if enabled returns the final rss queue
8321 * count which may be different from the requested queue count.
8323 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8325 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8326 int new_rss_size;
8328 if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8329 return 0;
8331 new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8333 if (queue_count != vsi->num_queue_pairs) {
8334 vsi->req_queue_pairs = queue_count;
8335 i40e_prep_for_reset(pf);
8337 pf->alloc_rss_size = new_rss_size;
8339 i40e_reset_and_rebuild(pf, true);
8341 /* Discard the user configured hash keys and lut, if less
8342 * queues are enabled.
8344 if (queue_count < vsi->rss_size) {
8345 i40e_clear_rss_config_user(vsi);
8346 dev_dbg(&pf->pdev->dev,
8347 "discard user configured hash keys and lut\n");
8350 /* Reset vsi->rss_size, as number of enabled queues changed */
8351 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8352 vsi->num_queue_pairs);
8354 i40e_pf_config_rss(pf);
8356 dev_info(&pf->pdev->dev, "RSS count/HW max RSS count: %d/%d\n",
8357 pf->alloc_rss_size, pf->rss_size_max);
8358 return pf->alloc_rss_size;
8362 * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8363 * @pf: board private structure
8365 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8367 i40e_status status;
8368 bool min_valid, max_valid;
8369 u32 max_bw, min_bw;
8371 status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8372 &min_valid, &max_valid);
8374 if (!status) {
8375 if (min_valid)
8376 pf->npar_min_bw = min_bw;
8377 if (max_valid)
8378 pf->npar_max_bw = max_bw;
8381 return status;
8385 * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8386 * @pf: board private structure
8388 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8390 struct i40e_aqc_configure_partition_bw_data bw_data;
8391 i40e_status status;
8393 /* Set the valid bit for this PF */
8394 bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8395 bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8396 bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8398 /* Set the new bandwidths */
8399 status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8401 return status;
8405 * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8406 * @pf: board private structure
8408 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8410 /* Commit temporary BW setting to permanent NVM image */
8411 enum i40e_admin_queue_err last_aq_status;
8412 i40e_status ret;
8413 u16 nvm_word;
8415 if (pf->hw.partition_id != 1) {
8416 dev_info(&pf->pdev->dev,
8417 "Commit BW only works on partition 1! This is partition %d",
8418 pf->hw.partition_id);
8419 ret = I40E_NOT_SUPPORTED;
8420 goto bw_commit_out;
8423 /* Acquire NVM for read access */
8424 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8425 last_aq_status = pf->hw.aq.asq_last_status;
8426 if (ret) {
8427 dev_info(&pf->pdev->dev,
8428 "Cannot acquire NVM for read access, err %s aq_err %s\n",
8429 i40e_stat_str(&pf->hw, ret),
8430 i40e_aq_str(&pf->hw, last_aq_status));
8431 goto bw_commit_out;
8434 /* Read word 0x10 of NVM - SW compatibility word 1 */
8435 ret = i40e_aq_read_nvm(&pf->hw,
8436 I40E_SR_NVM_CONTROL_WORD,
8437 0x10, sizeof(nvm_word), &nvm_word,
8438 false, NULL);
8439 /* Save off last admin queue command status before releasing
8440 * the NVM
8442 last_aq_status = pf->hw.aq.asq_last_status;
8443 i40e_release_nvm(&pf->hw);
8444 if (ret) {
8445 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8446 i40e_stat_str(&pf->hw, ret),
8447 i40e_aq_str(&pf->hw, last_aq_status));
8448 goto bw_commit_out;
8451 /* Wait a bit for NVM release to complete */
8452 msleep(50);
8454 /* Acquire NVM for write access */
8455 ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8456 last_aq_status = pf->hw.aq.asq_last_status;
8457 if (ret) {
8458 dev_info(&pf->pdev->dev,
8459 "Cannot acquire NVM for write access, err %s aq_err %s\n",
8460 i40e_stat_str(&pf->hw, ret),
8461 i40e_aq_str(&pf->hw, last_aq_status));
8462 goto bw_commit_out;
8464 /* Write it back out unchanged to initiate update NVM,
8465 * which will force a write of the shadow (alt) RAM to
8466 * the NVM - thus storing the bandwidth values permanently.
8468 ret = i40e_aq_update_nvm(&pf->hw,
8469 I40E_SR_NVM_CONTROL_WORD,
8470 0x10, sizeof(nvm_word),
8471 &nvm_word, true, NULL);
8472 /* Save off last admin queue command status before releasing
8473 * the NVM
8475 last_aq_status = pf->hw.aq.asq_last_status;
8476 i40e_release_nvm(&pf->hw);
8477 if (ret)
8478 dev_info(&pf->pdev->dev,
8479 "BW settings NOT SAVED, err %s aq_err %s\n",
8480 i40e_stat_str(&pf->hw, ret),
8481 i40e_aq_str(&pf->hw, last_aq_status));
8482 bw_commit_out:
8484 return ret;
8488 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8489 * @pf: board private structure to initialize
8491 * i40e_sw_init initializes the Adapter private data structure.
8492 * Fields are initialized based on PCI device information and
8493 * OS network device settings (MTU size).
8495 static int i40e_sw_init(struct i40e_pf *pf)
8497 int err = 0;
8498 int size;
8500 pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8501 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8502 if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8503 if (I40E_DEBUG_USER & debug)
8504 pf->hw.debug_mask = debug;
8505 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8506 I40E_DEFAULT_MSG_ENABLE);
8509 /* Set default capability flags */
8510 pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8511 I40E_FLAG_MSI_ENABLED |
8512 I40E_FLAG_MSIX_ENABLED;
8514 /* Set default ITR */
8515 pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8516 pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8518 /* Depending on PF configurations, it is possible that the RSS
8519 * maximum might end up larger than the available queues
8521 pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8522 pf->alloc_rss_size = 1;
8523 pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8524 pf->rss_size_max = min_t(int, pf->rss_size_max,
8525 pf->hw.func_caps.num_tx_qp);
8526 if (pf->hw.func_caps.rss) {
8527 pf->flags |= I40E_FLAG_RSS_ENABLED;
8528 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8529 num_online_cpus());
8532 /* MFP mode enabled */
8533 if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8534 pf->flags |= I40E_FLAG_MFP_ENABLED;
8535 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8536 if (i40e_get_npar_bw_setting(pf))
8537 dev_warn(&pf->pdev->dev,
8538 "Could not get NPAR bw settings\n");
8539 else
8540 dev_info(&pf->pdev->dev,
8541 "Min BW = %8.8x, Max BW = %8.8x\n",
8542 pf->npar_min_bw, pf->npar_max_bw);
8545 /* FW/NVM is not yet fixed in this regard */
8546 if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8547 (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8548 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8549 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8550 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8551 pf->hw.num_partitions > 1)
8552 dev_info(&pf->pdev->dev,
8553 "Flow Director Sideband mode Disabled in MFP mode\n");
8554 else
8555 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8556 pf->fdir_pf_filter_count =
8557 pf->hw.func_caps.fd_filters_guaranteed;
8558 pf->hw.fdir_shared_filter_count =
8559 pf->hw.func_caps.fd_filters_best_effort;
8562 if (i40e_is_mac_710(&pf->hw) &&
8563 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8564 (pf->hw.aq.fw_maj_ver < 4))) {
8565 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8566 /* No DCB support for FW < v4.33 */
8567 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8570 /* Disable FW LLDP if FW < v4.3 */
8571 if (i40e_is_mac_710(&pf->hw) &&
8572 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8573 (pf->hw.aq.fw_maj_ver < 4)))
8574 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8576 /* Use the FW Set LLDP MIB API if FW > v4.40 */
8577 if (i40e_is_mac_710(&pf->hw) &&
8578 (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8579 (pf->hw.aq.fw_maj_ver >= 5)))
8580 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8582 if (pf->hw.func_caps.vmdq) {
8583 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8584 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8585 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8588 if (pf->hw.func_caps.iwarp) {
8589 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8590 /* IWARP needs one extra vector for CQP just like MISC.*/
8591 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8594 #ifdef I40E_FCOE
8595 i40e_init_pf_fcoe(pf);
8597 #endif /* I40E_FCOE */
8598 #ifdef CONFIG_PCI_IOV
8599 if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8600 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8601 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8602 pf->num_req_vfs = min_t(int,
8603 pf->hw.func_caps.num_vfs,
8604 I40E_MAX_VF_COUNT);
8606 #endif /* CONFIG_PCI_IOV */
8607 if (pf->hw.mac.type == I40E_MAC_X722) {
8608 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8609 I40E_FLAG_128_QP_RSS_CAPABLE |
8610 I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8611 I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8612 I40E_FLAG_WB_ON_ITR_CAPABLE |
8613 I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8614 I40E_FLAG_NO_PCI_LINK_CHECK |
8615 I40E_FLAG_USE_SET_LLDP_MIB |
8616 I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8617 } else if ((pf->hw.aq.api_maj_ver > 1) ||
8618 ((pf->hw.aq.api_maj_ver == 1) &&
8619 (pf->hw.aq.api_min_ver > 4))) {
8620 /* Supported in FW API version higher than 1.4 */
8621 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8622 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8623 } else {
8624 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8627 pf->eeprom_version = 0xDEAD;
8628 pf->lan_veb = I40E_NO_VEB;
8629 pf->lan_vsi = I40E_NO_VSI;
8631 /* By default FW has this off for performance reasons */
8632 pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8634 /* set up queue assignment tracking */
8635 size = sizeof(struct i40e_lump_tracking)
8636 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8637 pf->qp_pile = kzalloc(size, GFP_KERNEL);
8638 if (!pf->qp_pile) {
8639 err = -ENOMEM;
8640 goto sw_init_done;
8642 pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8643 pf->qp_pile->search_hint = 0;
8645 pf->tx_timeout_recovery_level = 1;
8647 mutex_init(&pf->switch_mutex);
8649 /* If NPAR is enabled nudge the Tx scheduler */
8650 if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8651 i40e_set_npar_bw_setting(pf);
8653 sw_init_done:
8654 return err;
8658 * i40e_set_ntuple - set the ntuple feature flag and take action
8659 * @pf: board private structure to initialize
8660 * @features: the feature set that the stack is suggesting
8662 * returns a bool to indicate if reset needs to happen
8664 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8666 bool need_reset = false;
8668 /* Check if Flow Director n-tuple support was enabled or disabled. If
8669 * the state changed, we need to reset.
8671 if (features & NETIF_F_NTUPLE) {
8672 /* Enable filters and mark for reset */
8673 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8674 need_reset = true;
8675 /* enable FD_SB only if there is MSI-X vector */
8676 if (pf->num_fdsb_msix > 0)
8677 pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8678 } else {
8679 /* turn off filters, mark for reset and clear SW filter list */
8680 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8681 need_reset = true;
8682 i40e_fdir_filter_exit(pf);
8684 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8685 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8686 /* reset fd counters */
8687 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8688 pf->fdir_pf_active_filters = 0;
8689 /* if ATR was auto disabled it can be re-enabled. */
8690 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8691 (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
8692 pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8693 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8694 dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8697 return need_reset;
8701 * i40e_clear_rss_lut - clear the rx hash lookup table
8702 * @vsi: the VSI being configured
8704 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
8706 struct i40e_pf *pf = vsi->back;
8707 struct i40e_hw *hw = &pf->hw;
8708 u16 vf_id = vsi->vf_id;
8709 u8 i;
8711 if (vsi->type == I40E_VSI_MAIN) {
8712 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8713 wr32(hw, I40E_PFQF_HLUT(i), 0);
8714 } else if (vsi->type == I40E_VSI_SRIOV) {
8715 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8716 i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
8717 } else {
8718 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8723 * i40e_set_features - set the netdev feature flags
8724 * @netdev: ptr to the netdev being adjusted
8725 * @features: the feature set that the stack is suggesting
8727 static int i40e_set_features(struct net_device *netdev,
8728 netdev_features_t features)
8730 struct i40e_netdev_priv *np = netdev_priv(netdev);
8731 struct i40e_vsi *vsi = np->vsi;
8732 struct i40e_pf *pf = vsi->back;
8733 bool need_reset;
8735 if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
8736 i40e_pf_config_rss(pf);
8737 else if (!(features & NETIF_F_RXHASH) &&
8738 netdev->features & NETIF_F_RXHASH)
8739 i40e_clear_rss_lut(vsi);
8741 if (features & NETIF_F_HW_VLAN_CTAG_RX)
8742 i40e_vlan_stripping_enable(vsi);
8743 else
8744 i40e_vlan_stripping_disable(vsi);
8746 need_reset = i40e_set_ntuple(pf, features);
8748 if (need_reset)
8749 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8751 return 0;
8755 * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8756 * @pf: board private structure
8757 * @port: The UDP port to look up
8759 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8761 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8763 u8 i;
8765 for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8766 if (pf->udp_ports[i].index == port)
8767 return i;
8770 return i;
8774 * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8775 * @netdev: This physical port's netdev
8776 * @ti: Tunnel endpoint information
8778 static void i40e_udp_tunnel_add(struct net_device *netdev,
8779 struct udp_tunnel_info *ti)
8781 struct i40e_netdev_priv *np = netdev_priv(netdev);
8782 struct i40e_vsi *vsi = np->vsi;
8783 struct i40e_pf *pf = vsi->back;
8784 __be16 port = ti->port;
8785 u8 next_idx;
8786 u8 idx;
8788 idx = i40e_get_udp_port_idx(pf, port);
8790 /* Check if port already exists */
8791 if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8792 netdev_info(netdev, "port %d already offloaded\n",
8793 ntohs(port));
8794 return;
8797 /* Now check if there is space to add the new port */
8798 next_idx = i40e_get_udp_port_idx(pf, 0);
8800 if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8801 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8802 ntohs(port));
8803 return;
8806 switch (ti->type) {
8807 case UDP_TUNNEL_TYPE_VXLAN:
8808 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8809 break;
8810 case UDP_TUNNEL_TYPE_GENEVE:
8811 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8812 return;
8813 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8814 break;
8815 default:
8816 return;
8819 /* New port: add it and mark its index in the bitmap */
8820 pf->udp_ports[next_idx].index = port;
8821 pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8822 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8826 * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
8827 * @netdev: This physical port's netdev
8828 * @ti: Tunnel endpoint information
8830 static void i40e_udp_tunnel_del(struct net_device *netdev,
8831 struct udp_tunnel_info *ti)
8833 struct i40e_netdev_priv *np = netdev_priv(netdev);
8834 struct i40e_vsi *vsi = np->vsi;
8835 struct i40e_pf *pf = vsi->back;
8836 __be16 port = ti->port;
8837 u8 idx;
8839 idx = i40e_get_udp_port_idx(pf, port);
8841 /* Check if port already exists */
8842 if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
8843 goto not_found;
8845 switch (ti->type) {
8846 case UDP_TUNNEL_TYPE_VXLAN:
8847 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
8848 goto not_found;
8849 break;
8850 case UDP_TUNNEL_TYPE_GENEVE:
8851 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
8852 goto not_found;
8853 break;
8854 default:
8855 goto not_found;
8858 /* if port exists, set it to 0 (mark for deletion)
8859 * and make it pending
8861 pf->udp_ports[idx].index = 0;
8862 pf->pending_udp_bitmap |= BIT_ULL(idx);
8863 pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8865 return;
8866 not_found:
8867 netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
8868 ntohs(port));
8871 static int i40e_get_phys_port_id(struct net_device *netdev,
8872 struct netdev_phys_item_id *ppid)
8874 struct i40e_netdev_priv *np = netdev_priv(netdev);
8875 struct i40e_pf *pf = np->vsi->back;
8876 struct i40e_hw *hw = &pf->hw;
8878 if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8879 return -EOPNOTSUPP;
8881 ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8882 memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8884 return 0;
8888 * i40e_ndo_fdb_add - add an entry to the hardware database
8889 * @ndm: the input from the stack
8890 * @tb: pointer to array of nladdr (unused)
8891 * @dev: the net device pointer
8892 * @addr: the MAC address entry being added
8893 * @flags: instructions from stack about fdb operation
8895 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8896 struct net_device *dev,
8897 const unsigned char *addr, u16 vid,
8898 u16 flags)
8900 struct i40e_netdev_priv *np = netdev_priv(dev);
8901 struct i40e_pf *pf = np->vsi->back;
8902 int err = 0;
8904 if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8905 return -EOPNOTSUPP;
8907 if (vid) {
8908 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8909 return -EINVAL;
8912 /* Hardware does not support aging addresses so if a
8913 * ndm_state is given only allow permanent addresses
8915 if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8916 netdev_info(dev, "FDB only supports static addresses\n");
8917 return -EINVAL;
8920 if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8921 err = dev_uc_add_excl(dev, addr);
8922 else if (is_multicast_ether_addr(addr))
8923 err = dev_mc_add_excl(dev, addr);
8924 else
8925 err = -EINVAL;
8927 /* Only return duplicate errors if NLM_F_EXCL is set */
8928 if (err == -EEXIST && !(flags & NLM_F_EXCL))
8929 err = 0;
8931 return err;
8935 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8936 * @dev: the netdev being configured
8937 * @nlh: RTNL message
8939 * Inserts a new hardware bridge if not already created and
8940 * enables the bridging mode requested (VEB or VEPA). If the
8941 * hardware bridge has already been inserted and the request
8942 * is to change the mode then that requires a PF reset to
8943 * allow rebuild of the components with required hardware
8944 * bridge mode enabled.
8946 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8947 struct nlmsghdr *nlh,
8948 u16 flags)
8950 struct i40e_netdev_priv *np = netdev_priv(dev);
8951 struct i40e_vsi *vsi = np->vsi;
8952 struct i40e_pf *pf = vsi->back;
8953 struct i40e_veb *veb = NULL;
8954 struct nlattr *attr, *br_spec;
8955 int i, rem;
8957 /* Only for PF VSI for now */
8958 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8959 return -EOPNOTSUPP;
8961 /* Find the HW bridge for PF VSI */
8962 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8963 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8964 veb = pf->veb[i];
8967 br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8969 nla_for_each_nested(attr, br_spec, rem) {
8970 __u16 mode;
8972 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8973 continue;
8975 mode = nla_get_u16(attr);
8976 if ((mode != BRIDGE_MODE_VEPA) &&
8977 (mode != BRIDGE_MODE_VEB))
8978 return -EINVAL;
8980 /* Insert a new HW bridge */
8981 if (!veb) {
8982 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8983 vsi->tc_config.enabled_tc);
8984 if (veb) {
8985 veb->bridge_mode = mode;
8986 i40e_config_bridge_mode(veb);
8987 } else {
8988 /* No Bridge HW offload available */
8989 return -ENOENT;
8991 break;
8992 } else if (mode != veb->bridge_mode) {
8993 /* Existing HW bridge but different mode needs reset */
8994 veb->bridge_mode = mode;
8995 /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8996 if (mode == BRIDGE_MODE_VEB)
8997 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8998 else
8999 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9000 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
9001 break;
9005 return 0;
9009 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9010 * @skb: skb buff
9011 * @pid: process id
9012 * @seq: RTNL message seq #
9013 * @dev: the netdev being configured
9014 * @filter_mask: unused
9015 * @nlflags: netlink flags passed in
9017 * Return the mode in which the hardware bridge is operating in
9018 * i.e VEB or VEPA.
9020 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9021 struct net_device *dev,
9022 u32 __always_unused filter_mask,
9023 int nlflags)
9025 struct i40e_netdev_priv *np = netdev_priv(dev);
9026 struct i40e_vsi *vsi = np->vsi;
9027 struct i40e_pf *pf = vsi->back;
9028 struct i40e_veb *veb = NULL;
9029 int i;
9031 /* Only for PF VSI for now */
9032 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9033 return -EOPNOTSUPP;
9035 /* Find the HW bridge for the PF VSI */
9036 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9037 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9038 veb = pf->veb[i];
9041 if (!veb)
9042 return 0;
9044 return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9045 0, 0, nlflags, filter_mask, NULL);
9048 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9049 * inner mac plus all inner ethertypes.
9051 #define I40E_MAX_TUNNEL_HDR_LEN 128
9053 * i40e_features_check - Validate encapsulated packet conforms to limits
9054 * @skb: skb buff
9055 * @dev: This physical port's netdev
9056 * @features: Offload features that the stack believes apply
9058 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9059 struct net_device *dev,
9060 netdev_features_t features)
9062 if (skb->encapsulation &&
9063 ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9064 I40E_MAX_TUNNEL_HDR_LEN))
9065 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9067 return features;
9070 static const struct net_device_ops i40e_netdev_ops = {
9071 .ndo_open = i40e_open,
9072 .ndo_stop = i40e_close,
9073 .ndo_start_xmit = i40e_lan_xmit_frame,
9074 .ndo_get_stats64 = i40e_get_netdev_stats_struct,
9075 .ndo_set_rx_mode = i40e_set_rx_mode,
9076 .ndo_validate_addr = eth_validate_addr,
9077 .ndo_set_mac_address = i40e_set_mac,
9078 .ndo_change_mtu = i40e_change_mtu,
9079 .ndo_do_ioctl = i40e_ioctl,
9080 .ndo_tx_timeout = i40e_tx_timeout,
9081 .ndo_vlan_rx_add_vid = i40e_vlan_rx_add_vid,
9082 .ndo_vlan_rx_kill_vid = i40e_vlan_rx_kill_vid,
9083 #ifdef CONFIG_NET_POLL_CONTROLLER
9084 .ndo_poll_controller = i40e_netpoll,
9085 #endif
9086 .ndo_setup_tc = __i40e_setup_tc,
9087 #ifdef I40E_FCOE
9088 .ndo_fcoe_enable = i40e_fcoe_enable,
9089 .ndo_fcoe_disable = i40e_fcoe_disable,
9090 #endif
9091 .ndo_set_features = i40e_set_features,
9092 .ndo_set_vf_mac = i40e_ndo_set_vf_mac,
9093 .ndo_set_vf_vlan = i40e_ndo_set_vf_port_vlan,
9094 .ndo_set_vf_rate = i40e_ndo_set_vf_bw,
9095 .ndo_get_vf_config = i40e_ndo_get_vf_config,
9096 .ndo_set_vf_link_state = i40e_ndo_set_vf_link_state,
9097 .ndo_set_vf_spoofchk = i40e_ndo_set_vf_spoofchk,
9098 .ndo_set_vf_trust = i40e_ndo_set_vf_trust,
9099 .ndo_udp_tunnel_add = i40e_udp_tunnel_add,
9100 .ndo_udp_tunnel_del = i40e_udp_tunnel_del,
9101 .ndo_get_phys_port_id = i40e_get_phys_port_id,
9102 .ndo_fdb_add = i40e_ndo_fdb_add,
9103 .ndo_features_check = i40e_features_check,
9104 .ndo_bridge_getlink = i40e_ndo_bridge_getlink,
9105 .ndo_bridge_setlink = i40e_ndo_bridge_setlink,
9109 * i40e_config_netdev - Setup the netdev flags
9110 * @vsi: the VSI being configured
9112 * Returns 0 on success, negative value on failure
9114 static int i40e_config_netdev(struct i40e_vsi *vsi)
9116 struct i40e_pf *pf = vsi->back;
9117 struct i40e_hw *hw = &pf->hw;
9118 struct i40e_netdev_priv *np;
9119 struct net_device *netdev;
9120 u8 mac_addr[ETH_ALEN];
9121 int etherdev_size;
9123 etherdev_size = sizeof(struct i40e_netdev_priv);
9124 netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9125 if (!netdev)
9126 return -ENOMEM;
9128 vsi->netdev = netdev;
9129 np = netdev_priv(netdev);
9130 np->vsi = vsi;
9132 netdev->hw_enc_features |= NETIF_F_SG |
9133 NETIF_F_IP_CSUM |
9134 NETIF_F_IPV6_CSUM |
9135 NETIF_F_HIGHDMA |
9136 NETIF_F_SOFT_FEATURES |
9137 NETIF_F_TSO |
9138 NETIF_F_TSO_ECN |
9139 NETIF_F_TSO6 |
9140 NETIF_F_GSO_GRE |
9141 NETIF_F_GSO_GRE_CSUM |
9142 NETIF_F_GSO_IPXIP4 |
9143 NETIF_F_GSO_IPXIP6 |
9144 NETIF_F_GSO_UDP_TUNNEL |
9145 NETIF_F_GSO_UDP_TUNNEL_CSUM |
9146 NETIF_F_GSO_PARTIAL |
9147 NETIF_F_SCTP_CRC |
9148 NETIF_F_RXHASH |
9149 NETIF_F_RXCSUM |
9152 if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9153 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9155 netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9157 /* record features VLANs can make use of */
9158 netdev->vlan_features |= netdev->hw_enc_features |
9159 NETIF_F_TSO_MANGLEID;
9161 if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9162 netdev->hw_features |= NETIF_F_NTUPLE;
9164 netdev->hw_features |= netdev->hw_enc_features |
9165 NETIF_F_HW_VLAN_CTAG_TX |
9166 NETIF_F_HW_VLAN_CTAG_RX;
9168 netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9169 netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9171 if (vsi->type == I40E_VSI_MAIN) {
9172 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9173 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9174 /* The following steps are necessary to prevent reception
9175 * of tagged packets - some older NVM configurations load a
9176 * default a MAC-VLAN filter that accepts any tagged packet
9177 * which must be replaced by a normal filter.
9179 i40e_rm_default_mac_filter(vsi, mac_addr);
9180 spin_lock_bh(&vsi->mac_filter_list_lock);
9181 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, true);
9182 spin_unlock_bh(&vsi->mac_filter_list_lock);
9183 } else {
9184 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9185 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9186 pf->vsi[pf->lan_vsi]->netdev->name);
9187 random_ether_addr(mac_addr);
9189 spin_lock_bh(&vsi->mac_filter_list_lock);
9190 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9191 spin_unlock_bh(&vsi->mac_filter_list_lock);
9194 ether_addr_copy(netdev->dev_addr, mac_addr);
9195 ether_addr_copy(netdev->perm_addr, mac_addr);
9197 netdev->priv_flags |= IFF_UNICAST_FLT;
9198 netdev->priv_flags |= IFF_SUPP_NOFCS;
9199 /* Setup netdev TC information */
9200 i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9202 netdev->netdev_ops = &i40e_netdev_ops;
9203 netdev->watchdog_timeo = 5 * HZ;
9204 i40e_set_ethtool_ops(netdev);
9205 #ifdef I40E_FCOE
9206 i40e_fcoe_config_netdev(netdev, vsi);
9207 #endif
9209 return 0;
9213 * i40e_vsi_delete - Delete a VSI from the switch
9214 * @vsi: the VSI being removed
9216 * Returns 0 on success, negative value on failure
9218 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9220 /* remove default VSI is not allowed */
9221 if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9222 return;
9224 i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9228 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9229 * @vsi: the VSI being queried
9231 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9233 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9235 struct i40e_veb *veb;
9236 struct i40e_pf *pf = vsi->back;
9238 /* Uplink is not a bridge so default to VEB */
9239 if (vsi->veb_idx == I40E_NO_VEB)
9240 return 1;
9242 veb = pf->veb[vsi->veb_idx];
9243 if (!veb) {
9244 dev_info(&pf->pdev->dev,
9245 "There is no veb associated with the bridge\n");
9246 return -ENOENT;
9249 /* Uplink is a bridge in VEPA mode */
9250 if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9251 return 0;
9252 } else {
9253 /* Uplink is a bridge in VEB mode */
9254 return 1;
9257 /* VEPA is now default bridge, so return 0 */
9258 return 0;
9262 * i40e_add_vsi - Add a VSI to the switch
9263 * @vsi: the VSI being configured
9265 * This initializes a VSI context depending on the VSI type to be added and
9266 * passes it down to the add_vsi aq command.
9268 static int i40e_add_vsi(struct i40e_vsi *vsi)
9270 int ret = -ENODEV;
9271 i40e_status aq_ret = 0;
9272 struct i40e_pf *pf = vsi->back;
9273 struct i40e_hw *hw = &pf->hw;
9274 struct i40e_vsi_context ctxt;
9275 struct i40e_mac_filter *f, *ftmp;
9277 u8 enabled_tc = 0x1; /* TC0 enabled */
9278 int f_count = 0;
9280 memset(&ctxt, 0, sizeof(ctxt));
9281 switch (vsi->type) {
9282 case I40E_VSI_MAIN:
9283 /* The PF's main VSI is already setup as part of the
9284 * device initialization, so we'll not bother with
9285 * the add_vsi call, but we will retrieve the current
9286 * VSI context.
9288 ctxt.seid = pf->main_vsi_seid;
9289 ctxt.pf_num = pf->hw.pf_id;
9290 ctxt.vf_num = 0;
9291 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9292 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9293 if (ret) {
9294 dev_info(&pf->pdev->dev,
9295 "couldn't get PF vsi config, err %s aq_err %s\n",
9296 i40e_stat_str(&pf->hw, ret),
9297 i40e_aq_str(&pf->hw,
9298 pf->hw.aq.asq_last_status));
9299 return -ENOENT;
9301 vsi->info = ctxt.info;
9302 vsi->info.valid_sections = 0;
9304 vsi->seid = ctxt.seid;
9305 vsi->id = ctxt.vsi_number;
9307 enabled_tc = i40e_pf_get_tc_map(pf);
9309 /* MFP mode setup queue map and update VSI */
9310 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9311 !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9312 memset(&ctxt, 0, sizeof(ctxt));
9313 ctxt.seid = pf->main_vsi_seid;
9314 ctxt.pf_num = pf->hw.pf_id;
9315 ctxt.vf_num = 0;
9316 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9317 ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9318 if (ret) {
9319 dev_info(&pf->pdev->dev,
9320 "update vsi failed, err %s aq_err %s\n",
9321 i40e_stat_str(&pf->hw, ret),
9322 i40e_aq_str(&pf->hw,
9323 pf->hw.aq.asq_last_status));
9324 ret = -ENOENT;
9325 goto err;
9327 /* update the local VSI info queue map */
9328 i40e_vsi_update_queue_map(vsi, &ctxt);
9329 vsi->info.valid_sections = 0;
9330 } else {
9331 /* Default/Main VSI is only enabled for TC0
9332 * reconfigure it to enable all TCs that are
9333 * available on the port in SFP mode.
9334 * For MFP case the iSCSI PF would use this
9335 * flow to enable LAN+iSCSI TC.
9337 ret = i40e_vsi_config_tc(vsi, enabled_tc);
9338 if (ret) {
9339 dev_info(&pf->pdev->dev,
9340 "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9341 enabled_tc,
9342 i40e_stat_str(&pf->hw, ret),
9343 i40e_aq_str(&pf->hw,
9344 pf->hw.aq.asq_last_status));
9345 ret = -ENOENT;
9348 break;
9350 case I40E_VSI_FDIR:
9351 ctxt.pf_num = hw->pf_id;
9352 ctxt.vf_num = 0;
9353 ctxt.uplink_seid = vsi->uplink_seid;
9354 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9355 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9356 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9357 (i40e_is_vsi_uplink_mode_veb(vsi))) {
9358 ctxt.info.valid_sections |=
9359 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9360 ctxt.info.switch_id =
9361 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9363 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9364 break;
9366 case I40E_VSI_VMDQ2:
9367 ctxt.pf_num = hw->pf_id;
9368 ctxt.vf_num = 0;
9369 ctxt.uplink_seid = vsi->uplink_seid;
9370 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9371 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9373 /* This VSI is connected to VEB so the switch_id
9374 * should be set to zero by default.
9376 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9377 ctxt.info.valid_sections |=
9378 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9379 ctxt.info.switch_id =
9380 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9383 /* Setup the VSI tx/rx queue map for TC0 only for now */
9384 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9385 break;
9387 case I40E_VSI_SRIOV:
9388 ctxt.pf_num = hw->pf_id;
9389 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9390 ctxt.uplink_seid = vsi->uplink_seid;
9391 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9392 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9394 /* This VSI is connected to VEB so the switch_id
9395 * should be set to zero by default.
9397 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9398 ctxt.info.valid_sections |=
9399 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9400 ctxt.info.switch_id =
9401 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9404 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9405 ctxt.info.valid_sections |=
9406 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9407 ctxt.info.queueing_opt_flags |=
9408 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9409 I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9412 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9413 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9414 if (pf->vf[vsi->vf_id].spoofchk) {
9415 ctxt.info.valid_sections |=
9416 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9417 ctxt.info.sec_flags |=
9418 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9419 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9421 /* Setup the VSI tx/rx queue map for TC0 only for now */
9422 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9423 break;
9425 #ifdef I40E_FCOE
9426 case I40E_VSI_FCOE:
9427 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9428 if (ret) {
9429 dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9430 return ret;
9432 break;
9434 #endif /* I40E_FCOE */
9435 case I40E_VSI_IWARP:
9436 /* send down message to iWARP */
9437 break;
9439 default:
9440 return -ENODEV;
9443 if (vsi->type != I40E_VSI_MAIN) {
9444 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9445 if (ret) {
9446 dev_info(&vsi->back->pdev->dev,
9447 "add vsi failed, err %s aq_err %s\n",
9448 i40e_stat_str(&pf->hw, ret),
9449 i40e_aq_str(&pf->hw,
9450 pf->hw.aq.asq_last_status));
9451 ret = -ENOENT;
9452 goto err;
9454 vsi->info = ctxt.info;
9455 vsi->info.valid_sections = 0;
9456 vsi->seid = ctxt.seid;
9457 vsi->id = ctxt.vsi_number;
9459 /* Except FDIR VSI, for all othet VSI set the broadcast filter */
9460 if (vsi->type != I40E_VSI_FDIR) {
9461 aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9462 if (aq_ret) {
9463 ret = i40e_aq_rc_to_posix(aq_ret,
9464 hw->aq.asq_last_status);
9465 dev_info(&pf->pdev->dev,
9466 "set brdcast promisc failed, err %s, aq_err %s\n",
9467 i40e_stat_str(hw, aq_ret),
9468 i40e_aq_str(hw, hw->aq.asq_last_status));
9472 vsi->active_filters = 0;
9473 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9474 spin_lock_bh(&vsi->mac_filter_list_lock);
9475 /* If macvlan filters already exist, force them to get loaded */
9476 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9477 f->state = I40E_FILTER_NEW;
9478 f_count++;
9480 spin_unlock_bh(&vsi->mac_filter_list_lock);
9482 if (f_count) {
9483 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9484 pf->flags |= I40E_FLAG_FILTER_SYNC;
9487 /* Update VSI BW information */
9488 ret = i40e_vsi_get_bw_info(vsi);
9489 if (ret) {
9490 dev_info(&pf->pdev->dev,
9491 "couldn't get vsi bw info, err %s aq_err %s\n",
9492 i40e_stat_str(&pf->hw, ret),
9493 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9494 /* VSI is already added so not tearing that up */
9495 ret = 0;
9498 err:
9499 return ret;
9503 * i40e_vsi_release - Delete a VSI and free its resources
9504 * @vsi: the VSI being removed
9506 * Returns 0 on success or < 0 on error
9508 int i40e_vsi_release(struct i40e_vsi *vsi)
9510 struct i40e_mac_filter *f, *ftmp;
9511 struct i40e_veb *veb = NULL;
9512 struct i40e_pf *pf;
9513 u16 uplink_seid;
9514 int i, n;
9516 pf = vsi->back;
9518 /* release of a VEB-owner or last VSI is not allowed */
9519 if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9520 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9521 vsi->seid, vsi->uplink_seid);
9522 return -ENODEV;
9524 if (vsi == pf->vsi[pf->lan_vsi] &&
9525 !test_bit(__I40E_DOWN, &pf->state)) {
9526 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9527 return -ENODEV;
9530 uplink_seid = vsi->uplink_seid;
9531 if (vsi->type != I40E_VSI_SRIOV) {
9532 if (vsi->netdev_registered) {
9533 vsi->netdev_registered = false;
9534 if (vsi->netdev) {
9535 /* results in a call to i40e_close() */
9536 unregister_netdev(vsi->netdev);
9538 } else {
9539 i40e_vsi_close(vsi);
9541 i40e_vsi_disable_irq(vsi);
9544 spin_lock_bh(&vsi->mac_filter_list_lock);
9545 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9546 i40e_del_filter(vsi, f->macaddr, f->vlan,
9547 f->is_vf, f->is_netdev);
9548 spin_unlock_bh(&vsi->mac_filter_list_lock);
9550 i40e_sync_vsi_filters(vsi);
9552 i40e_vsi_delete(vsi);
9553 i40e_vsi_free_q_vectors(vsi);
9554 if (vsi->netdev) {
9555 free_netdev(vsi->netdev);
9556 vsi->netdev = NULL;
9558 i40e_vsi_clear_rings(vsi);
9559 i40e_vsi_clear(vsi);
9561 /* If this was the last thing on the VEB, except for the
9562 * controlling VSI, remove the VEB, which puts the controlling
9563 * VSI onto the next level down in the switch.
9565 * Well, okay, there's one more exception here: don't remove
9566 * the orphan VEBs yet. We'll wait for an explicit remove request
9567 * from up the network stack.
9569 for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9570 if (pf->vsi[i] &&
9571 pf->vsi[i]->uplink_seid == uplink_seid &&
9572 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9573 n++; /* count the VSIs */
9576 for (i = 0; i < I40E_MAX_VEB; i++) {
9577 if (!pf->veb[i])
9578 continue;
9579 if (pf->veb[i]->uplink_seid == uplink_seid)
9580 n++; /* count the VEBs */
9581 if (pf->veb[i]->seid == uplink_seid)
9582 veb = pf->veb[i];
9584 if (n == 0 && veb && veb->uplink_seid != 0)
9585 i40e_veb_release(veb);
9587 return 0;
9591 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9592 * @vsi: ptr to the VSI
9594 * This should only be called after i40e_vsi_mem_alloc() which allocates the
9595 * corresponding SW VSI structure and initializes num_queue_pairs for the
9596 * newly allocated VSI.
9598 * Returns 0 on success or negative on failure
9600 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9602 int ret = -ENOENT;
9603 struct i40e_pf *pf = vsi->back;
9605 if (vsi->q_vectors[0]) {
9606 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9607 vsi->seid);
9608 return -EEXIST;
9611 if (vsi->base_vector) {
9612 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9613 vsi->seid, vsi->base_vector);
9614 return -EEXIST;
9617 ret = i40e_vsi_alloc_q_vectors(vsi);
9618 if (ret) {
9619 dev_info(&pf->pdev->dev,
9620 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9621 vsi->num_q_vectors, vsi->seid, ret);
9622 vsi->num_q_vectors = 0;
9623 goto vector_setup_out;
9626 /* In Legacy mode, we do not have to get any other vector since we
9627 * piggyback on the misc/ICR0 for queue interrupts.
9629 if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9630 return ret;
9631 if (vsi->num_q_vectors)
9632 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9633 vsi->num_q_vectors, vsi->idx);
9634 if (vsi->base_vector < 0) {
9635 dev_info(&pf->pdev->dev,
9636 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9637 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9638 i40e_vsi_free_q_vectors(vsi);
9639 ret = -ENOENT;
9640 goto vector_setup_out;
9643 vector_setup_out:
9644 return ret;
9648 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9649 * @vsi: pointer to the vsi.
9651 * This re-allocates a vsi's queue resources.
9653 * Returns pointer to the successfully allocated and configured VSI sw struct
9654 * on success, otherwise returns NULL on failure.
9656 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9658 struct i40e_pf *pf;
9659 u8 enabled_tc;
9660 int ret;
9662 if (!vsi)
9663 return NULL;
9665 pf = vsi->back;
9667 i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9668 i40e_vsi_clear_rings(vsi);
9670 i40e_vsi_free_arrays(vsi, false);
9671 i40e_set_num_rings_in_vsi(vsi);
9672 ret = i40e_vsi_alloc_arrays(vsi, false);
9673 if (ret)
9674 goto err_vsi;
9676 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9677 if (ret < 0) {
9678 dev_info(&pf->pdev->dev,
9679 "failed to get tracking for %d queues for VSI %d err %d\n",
9680 vsi->alloc_queue_pairs, vsi->seid, ret);
9681 goto err_vsi;
9683 vsi->base_queue = ret;
9685 /* Update the FW view of the VSI. Force a reset of TC and queue
9686 * layout configurations.
9688 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9689 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9690 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9691 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9692 if (vsi->type == I40E_VSI_MAIN)
9693 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9695 /* assign it some queues */
9696 ret = i40e_alloc_rings(vsi);
9697 if (ret)
9698 goto err_rings;
9700 /* map all of the rings to the q_vectors */
9701 i40e_vsi_map_rings_to_vectors(vsi);
9702 return vsi;
9704 err_rings:
9705 i40e_vsi_free_q_vectors(vsi);
9706 if (vsi->netdev_registered) {
9707 vsi->netdev_registered = false;
9708 unregister_netdev(vsi->netdev);
9709 free_netdev(vsi->netdev);
9710 vsi->netdev = NULL;
9712 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9713 err_vsi:
9714 i40e_vsi_clear(vsi);
9715 return NULL;
9719 * i40e_vsi_setup - Set up a VSI by a given type
9720 * @pf: board private structure
9721 * @type: VSI type
9722 * @uplink_seid: the switch element to link to
9723 * @param1: usage depends upon VSI type. For VF types, indicates VF id
9725 * This allocates the sw VSI structure and its queue resources, then add a VSI
9726 * to the identified VEB.
9728 * Returns pointer to the successfully allocated and configure VSI sw struct on
9729 * success, otherwise returns NULL on failure.
9731 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9732 u16 uplink_seid, u32 param1)
9734 struct i40e_vsi *vsi = NULL;
9735 struct i40e_veb *veb = NULL;
9736 int ret, i;
9737 int v_idx;
9739 /* The requested uplink_seid must be either
9740 * - the PF's port seid
9741 * no VEB is needed because this is the PF
9742 * or this is a Flow Director special case VSI
9743 * - seid of an existing VEB
9744 * - seid of a VSI that owns an existing VEB
9745 * - seid of a VSI that doesn't own a VEB
9746 * a new VEB is created and the VSI becomes the owner
9747 * - seid of the PF VSI, which is what creates the first VEB
9748 * this is a special case of the previous
9750 * Find which uplink_seid we were given and create a new VEB if needed
9752 for (i = 0; i < I40E_MAX_VEB; i++) {
9753 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9754 veb = pf->veb[i];
9755 break;
9759 if (!veb && uplink_seid != pf->mac_seid) {
9761 for (i = 0; i < pf->num_alloc_vsi; i++) {
9762 if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9763 vsi = pf->vsi[i];
9764 break;
9767 if (!vsi) {
9768 dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9769 uplink_seid);
9770 return NULL;
9773 if (vsi->uplink_seid == pf->mac_seid)
9774 veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9775 vsi->tc_config.enabled_tc);
9776 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9777 veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9778 vsi->tc_config.enabled_tc);
9779 if (veb) {
9780 if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9781 dev_info(&vsi->back->pdev->dev,
9782 "New VSI creation error, uplink seid of LAN VSI expected.\n");
9783 return NULL;
9785 /* We come up by default in VEPA mode if SRIOV is not
9786 * already enabled, in which case we can't force VEPA
9787 * mode.
9789 if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9790 veb->bridge_mode = BRIDGE_MODE_VEPA;
9791 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9793 i40e_config_bridge_mode(veb);
9795 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9796 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9797 veb = pf->veb[i];
9799 if (!veb) {
9800 dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9801 return NULL;
9804 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9805 uplink_seid = veb->seid;
9808 /* get vsi sw struct */
9809 v_idx = i40e_vsi_mem_alloc(pf, type);
9810 if (v_idx < 0)
9811 goto err_alloc;
9812 vsi = pf->vsi[v_idx];
9813 if (!vsi)
9814 goto err_alloc;
9815 vsi->type = type;
9816 vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9818 if (type == I40E_VSI_MAIN)
9819 pf->lan_vsi = v_idx;
9820 else if (type == I40E_VSI_SRIOV)
9821 vsi->vf_id = param1;
9822 /* assign it some queues */
9823 ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9824 vsi->idx);
9825 if (ret < 0) {
9826 dev_info(&pf->pdev->dev,
9827 "failed to get tracking for %d queues for VSI %d err=%d\n",
9828 vsi->alloc_queue_pairs, vsi->seid, ret);
9829 goto err_vsi;
9831 vsi->base_queue = ret;
9833 /* get a VSI from the hardware */
9834 vsi->uplink_seid = uplink_seid;
9835 ret = i40e_add_vsi(vsi);
9836 if (ret)
9837 goto err_vsi;
9839 switch (vsi->type) {
9840 /* setup the netdev if needed */
9841 case I40E_VSI_MAIN:
9842 /* Apply relevant filters if a platform-specific mac
9843 * address was selected.
9845 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9846 ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9847 if (ret) {
9848 dev_warn(&pf->pdev->dev,
9849 "could not set up macaddr; err %d\n",
9850 ret);
9853 case I40E_VSI_VMDQ2:
9854 case I40E_VSI_FCOE:
9855 ret = i40e_config_netdev(vsi);
9856 if (ret)
9857 goto err_netdev;
9858 ret = register_netdev(vsi->netdev);
9859 if (ret)
9860 goto err_netdev;
9861 vsi->netdev_registered = true;
9862 netif_carrier_off(vsi->netdev);
9863 #ifdef CONFIG_I40E_DCB
9864 /* Setup DCB netlink interface */
9865 i40e_dcbnl_setup(vsi);
9866 #endif /* CONFIG_I40E_DCB */
9867 /* fall through */
9869 case I40E_VSI_FDIR:
9870 /* set up vectors and rings if needed */
9871 ret = i40e_vsi_setup_vectors(vsi);
9872 if (ret)
9873 goto err_msix;
9875 ret = i40e_alloc_rings(vsi);
9876 if (ret)
9877 goto err_rings;
9879 /* map all of the rings to the q_vectors */
9880 i40e_vsi_map_rings_to_vectors(vsi);
9882 i40e_vsi_reset_stats(vsi);
9883 break;
9885 default:
9886 /* no netdev or rings for the other VSI types */
9887 break;
9890 if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9891 (vsi->type == I40E_VSI_VMDQ2)) {
9892 ret = i40e_vsi_config_rss(vsi);
9894 return vsi;
9896 err_rings:
9897 i40e_vsi_free_q_vectors(vsi);
9898 err_msix:
9899 if (vsi->netdev_registered) {
9900 vsi->netdev_registered = false;
9901 unregister_netdev(vsi->netdev);
9902 free_netdev(vsi->netdev);
9903 vsi->netdev = NULL;
9905 err_netdev:
9906 i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9907 err_vsi:
9908 i40e_vsi_clear(vsi);
9909 err_alloc:
9910 return NULL;
9914 * i40e_veb_get_bw_info - Query VEB BW information
9915 * @veb: the veb to query
9917 * Query the Tx scheduler BW configuration data for given VEB
9919 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9921 struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9922 struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9923 struct i40e_pf *pf = veb->pf;
9924 struct i40e_hw *hw = &pf->hw;
9925 u32 tc_bw_max;
9926 int ret = 0;
9927 int i;
9929 ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9930 &bw_data, NULL);
9931 if (ret) {
9932 dev_info(&pf->pdev->dev,
9933 "query veb bw config failed, err %s aq_err %s\n",
9934 i40e_stat_str(&pf->hw, ret),
9935 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9936 goto out;
9939 ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9940 &ets_data, NULL);
9941 if (ret) {
9942 dev_info(&pf->pdev->dev,
9943 "query veb bw ets config failed, err %s aq_err %s\n",
9944 i40e_stat_str(&pf->hw, ret),
9945 i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9946 goto out;
9949 veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9950 veb->bw_max_quanta = ets_data.tc_bw_max;
9951 veb->is_abs_credits = bw_data.absolute_credits_enable;
9952 veb->enabled_tc = ets_data.tc_valid_bits;
9953 tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9954 (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9955 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9956 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9957 veb->bw_tc_limit_credits[i] =
9958 le16_to_cpu(bw_data.tc_bw_limits[i]);
9959 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9962 out:
9963 return ret;
9967 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9968 * @pf: board private structure
9970 * On error: returns error code (negative)
9971 * On success: returns vsi index in PF (positive)
9973 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9975 int ret = -ENOENT;
9976 struct i40e_veb *veb;
9977 int i;
9979 /* Need to protect the allocation of switch elements at the PF level */
9980 mutex_lock(&pf->switch_mutex);
9982 /* VEB list may be fragmented if VEB creation/destruction has
9983 * been happening. We can afford to do a quick scan to look
9984 * for any free slots in the list.
9986 * find next empty veb slot, looping back around if necessary
9988 i = 0;
9989 while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9990 i++;
9991 if (i >= I40E_MAX_VEB) {
9992 ret = -ENOMEM;
9993 goto err_alloc_veb; /* out of VEB slots! */
9996 veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9997 if (!veb) {
9998 ret = -ENOMEM;
9999 goto err_alloc_veb;
10001 veb->pf = pf;
10002 veb->idx = i;
10003 veb->enabled_tc = 1;
10005 pf->veb[i] = veb;
10006 ret = i;
10007 err_alloc_veb:
10008 mutex_unlock(&pf->switch_mutex);
10009 return ret;
10013 * i40e_switch_branch_release - Delete a branch of the switch tree
10014 * @branch: where to start deleting
10016 * This uses recursion to find the tips of the branch to be
10017 * removed, deleting until we get back to and can delete this VEB.
10019 static void i40e_switch_branch_release(struct i40e_veb *branch)
10021 struct i40e_pf *pf = branch->pf;
10022 u16 branch_seid = branch->seid;
10023 u16 veb_idx = branch->idx;
10024 int i;
10026 /* release any VEBs on this VEB - RECURSION */
10027 for (i = 0; i < I40E_MAX_VEB; i++) {
10028 if (!pf->veb[i])
10029 continue;
10030 if (pf->veb[i]->uplink_seid == branch->seid)
10031 i40e_switch_branch_release(pf->veb[i]);
10034 /* Release the VSIs on this VEB, but not the owner VSI.
10036 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10037 * the VEB itself, so don't use (*branch) after this loop.
10039 for (i = 0; i < pf->num_alloc_vsi; i++) {
10040 if (!pf->vsi[i])
10041 continue;
10042 if (pf->vsi[i]->uplink_seid == branch_seid &&
10043 (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10044 i40e_vsi_release(pf->vsi[i]);
10048 /* There's one corner case where the VEB might not have been
10049 * removed, so double check it here and remove it if needed.
10050 * This case happens if the veb was created from the debugfs
10051 * commands and no VSIs were added to it.
10053 if (pf->veb[veb_idx])
10054 i40e_veb_release(pf->veb[veb_idx]);
10058 * i40e_veb_clear - remove veb struct
10059 * @veb: the veb to remove
10061 static void i40e_veb_clear(struct i40e_veb *veb)
10063 if (!veb)
10064 return;
10066 if (veb->pf) {
10067 struct i40e_pf *pf = veb->pf;
10069 mutex_lock(&pf->switch_mutex);
10070 if (pf->veb[veb->idx] == veb)
10071 pf->veb[veb->idx] = NULL;
10072 mutex_unlock(&pf->switch_mutex);
10075 kfree(veb);
10079 * i40e_veb_release - Delete a VEB and free its resources
10080 * @veb: the VEB being removed
10082 void i40e_veb_release(struct i40e_veb *veb)
10084 struct i40e_vsi *vsi = NULL;
10085 struct i40e_pf *pf;
10086 int i, n = 0;
10088 pf = veb->pf;
10090 /* find the remaining VSI and check for extras */
10091 for (i = 0; i < pf->num_alloc_vsi; i++) {
10092 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10093 n++;
10094 vsi = pf->vsi[i];
10097 if (n != 1) {
10098 dev_info(&pf->pdev->dev,
10099 "can't remove VEB %d with %d VSIs left\n",
10100 veb->seid, n);
10101 return;
10104 /* move the remaining VSI to uplink veb */
10105 vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10106 if (veb->uplink_seid) {
10107 vsi->uplink_seid = veb->uplink_seid;
10108 if (veb->uplink_seid == pf->mac_seid)
10109 vsi->veb_idx = I40E_NO_VEB;
10110 else
10111 vsi->veb_idx = veb->veb_idx;
10112 } else {
10113 /* floating VEB */
10114 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10115 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10118 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10119 i40e_veb_clear(veb);
10123 * i40e_add_veb - create the VEB in the switch
10124 * @veb: the VEB to be instantiated
10125 * @vsi: the controlling VSI
10127 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10129 struct i40e_pf *pf = veb->pf;
10130 bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10131 int ret;
10133 ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10134 veb->enabled_tc, false,
10135 &veb->seid, enable_stats, NULL);
10137 /* get a VEB from the hardware */
10138 if (ret) {
10139 dev_info(&pf->pdev->dev,
10140 "couldn't add VEB, err %s aq_err %s\n",
10141 i40e_stat_str(&pf->hw, ret),
10142 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10143 return -EPERM;
10146 /* get statistics counter */
10147 ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10148 &veb->stats_idx, NULL, NULL, NULL);
10149 if (ret) {
10150 dev_info(&pf->pdev->dev,
10151 "couldn't get VEB statistics idx, err %s aq_err %s\n",
10152 i40e_stat_str(&pf->hw, ret),
10153 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10154 return -EPERM;
10156 ret = i40e_veb_get_bw_info(veb);
10157 if (ret) {
10158 dev_info(&pf->pdev->dev,
10159 "couldn't get VEB bw info, err %s aq_err %s\n",
10160 i40e_stat_str(&pf->hw, ret),
10161 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10162 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10163 return -ENOENT;
10166 vsi->uplink_seid = veb->seid;
10167 vsi->veb_idx = veb->idx;
10168 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10170 return 0;
10174 * i40e_veb_setup - Set up a VEB
10175 * @pf: board private structure
10176 * @flags: VEB setup flags
10177 * @uplink_seid: the switch element to link to
10178 * @vsi_seid: the initial VSI seid
10179 * @enabled_tc: Enabled TC bit-map
10181 * This allocates the sw VEB structure and links it into the switch
10182 * It is possible and legal for this to be a duplicate of an already
10183 * existing VEB. It is also possible for both uplink and vsi seids
10184 * to be zero, in order to create a floating VEB.
10186 * Returns pointer to the successfully allocated VEB sw struct on
10187 * success, otherwise returns NULL on failure.
10189 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10190 u16 uplink_seid, u16 vsi_seid,
10191 u8 enabled_tc)
10193 struct i40e_veb *veb, *uplink_veb = NULL;
10194 int vsi_idx, veb_idx;
10195 int ret;
10197 /* if one seid is 0, the other must be 0 to create a floating relay */
10198 if ((uplink_seid == 0 || vsi_seid == 0) &&
10199 (uplink_seid + vsi_seid != 0)) {
10200 dev_info(&pf->pdev->dev,
10201 "one, not both seid's are 0: uplink=%d vsi=%d\n",
10202 uplink_seid, vsi_seid);
10203 return NULL;
10206 /* make sure there is such a vsi and uplink */
10207 for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10208 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10209 break;
10210 if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10211 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10212 vsi_seid);
10213 return NULL;
10216 if (uplink_seid && uplink_seid != pf->mac_seid) {
10217 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10218 if (pf->veb[veb_idx] &&
10219 pf->veb[veb_idx]->seid == uplink_seid) {
10220 uplink_veb = pf->veb[veb_idx];
10221 break;
10224 if (!uplink_veb) {
10225 dev_info(&pf->pdev->dev,
10226 "uplink seid %d not found\n", uplink_seid);
10227 return NULL;
10231 /* get veb sw struct */
10232 veb_idx = i40e_veb_mem_alloc(pf);
10233 if (veb_idx < 0)
10234 goto err_alloc;
10235 veb = pf->veb[veb_idx];
10236 veb->flags = flags;
10237 veb->uplink_seid = uplink_seid;
10238 veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10239 veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10241 /* create the VEB in the switch */
10242 ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10243 if (ret)
10244 goto err_veb;
10245 if (vsi_idx == pf->lan_vsi)
10246 pf->lan_veb = veb->idx;
10248 return veb;
10250 err_veb:
10251 i40e_veb_clear(veb);
10252 err_alloc:
10253 return NULL;
10257 * i40e_setup_pf_switch_element - set PF vars based on switch type
10258 * @pf: board private structure
10259 * @ele: element we are building info from
10260 * @num_reported: total number of elements
10261 * @printconfig: should we print the contents
10263 * helper function to assist in extracting a few useful SEID values.
10265 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10266 struct i40e_aqc_switch_config_element_resp *ele,
10267 u16 num_reported, bool printconfig)
10269 u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10270 u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10271 u8 element_type = ele->element_type;
10272 u16 seid = le16_to_cpu(ele->seid);
10274 if (printconfig)
10275 dev_info(&pf->pdev->dev,
10276 "type=%d seid=%d uplink=%d downlink=%d\n",
10277 element_type, seid, uplink_seid, downlink_seid);
10279 switch (element_type) {
10280 case I40E_SWITCH_ELEMENT_TYPE_MAC:
10281 pf->mac_seid = seid;
10282 break;
10283 case I40E_SWITCH_ELEMENT_TYPE_VEB:
10284 /* Main VEB? */
10285 if (uplink_seid != pf->mac_seid)
10286 break;
10287 if (pf->lan_veb == I40E_NO_VEB) {
10288 int v;
10290 /* find existing or else empty VEB */
10291 for (v = 0; v < I40E_MAX_VEB; v++) {
10292 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10293 pf->lan_veb = v;
10294 break;
10297 if (pf->lan_veb == I40E_NO_VEB) {
10298 v = i40e_veb_mem_alloc(pf);
10299 if (v < 0)
10300 break;
10301 pf->lan_veb = v;
10305 pf->veb[pf->lan_veb]->seid = seid;
10306 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10307 pf->veb[pf->lan_veb]->pf = pf;
10308 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10309 break;
10310 case I40E_SWITCH_ELEMENT_TYPE_VSI:
10311 if (num_reported != 1)
10312 break;
10313 /* This is immediately after a reset so we can assume this is
10314 * the PF's VSI
10316 pf->mac_seid = uplink_seid;
10317 pf->pf_seid = downlink_seid;
10318 pf->main_vsi_seid = seid;
10319 if (printconfig)
10320 dev_info(&pf->pdev->dev,
10321 "pf_seid=%d main_vsi_seid=%d\n",
10322 pf->pf_seid, pf->main_vsi_seid);
10323 break;
10324 case I40E_SWITCH_ELEMENT_TYPE_PF:
10325 case I40E_SWITCH_ELEMENT_TYPE_VF:
10326 case I40E_SWITCH_ELEMENT_TYPE_EMP:
10327 case I40E_SWITCH_ELEMENT_TYPE_BMC:
10328 case I40E_SWITCH_ELEMENT_TYPE_PE:
10329 case I40E_SWITCH_ELEMENT_TYPE_PA:
10330 /* ignore these for now */
10331 break;
10332 default:
10333 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10334 element_type, seid);
10335 break;
10340 * i40e_fetch_switch_configuration - Get switch config from firmware
10341 * @pf: board private structure
10342 * @printconfig: should we print the contents
10344 * Get the current switch configuration from the device and
10345 * extract a few useful SEID values.
10347 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10349 struct i40e_aqc_get_switch_config_resp *sw_config;
10350 u16 next_seid = 0;
10351 int ret = 0;
10352 u8 *aq_buf;
10353 int i;
10355 aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10356 if (!aq_buf)
10357 return -ENOMEM;
10359 sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10360 do {
10361 u16 num_reported, num_total;
10363 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10364 I40E_AQ_LARGE_BUF,
10365 &next_seid, NULL);
10366 if (ret) {
10367 dev_info(&pf->pdev->dev,
10368 "get switch config failed err %s aq_err %s\n",
10369 i40e_stat_str(&pf->hw, ret),
10370 i40e_aq_str(&pf->hw,
10371 pf->hw.aq.asq_last_status));
10372 kfree(aq_buf);
10373 return -ENOENT;
10376 num_reported = le16_to_cpu(sw_config->header.num_reported);
10377 num_total = le16_to_cpu(sw_config->header.num_total);
10379 if (printconfig)
10380 dev_info(&pf->pdev->dev,
10381 "header: %d reported %d total\n",
10382 num_reported, num_total);
10384 for (i = 0; i < num_reported; i++) {
10385 struct i40e_aqc_switch_config_element_resp *ele =
10386 &sw_config->element[i];
10388 i40e_setup_pf_switch_element(pf, ele, num_reported,
10389 printconfig);
10391 } while (next_seid != 0);
10393 kfree(aq_buf);
10394 return ret;
10398 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10399 * @pf: board private structure
10400 * @reinit: if the Main VSI needs to re-initialized.
10402 * Returns 0 on success, negative value on failure
10404 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10406 u16 flags = 0;
10407 int ret;
10409 /* find out what's out there already */
10410 ret = i40e_fetch_switch_configuration(pf, false);
10411 if (ret) {
10412 dev_info(&pf->pdev->dev,
10413 "couldn't fetch switch config, err %s aq_err %s\n",
10414 i40e_stat_str(&pf->hw, ret),
10415 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10416 return ret;
10418 i40e_pf_reset_stats(pf);
10420 /* set the switch config bit for the whole device to
10421 * support limited promisc or true promisc
10422 * when user requests promisc. The default is limited
10423 * promisc.
10426 if ((pf->hw.pf_id == 0) &&
10427 !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10428 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10430 if (pf->hw.pf_id == 0) {
10431 u16 valid_flags;
10433 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10434 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10435 NULL);
10436 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10437 dev_info(&pf->pdev->dev,
10438 "couldn't set switch config bits, err %s aq_err %s\n",
10439 i40e_stat_str(&pf->hw, ret),
10440 i40e_aq_str(&pf->hw,
10441 pf->hw.aq.asq_last_status));
10442 /* not a fatal problem, just keep going */
10446 /* first time setup */
10447 if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10448 struct i40e_vsi *vsi = NULL;
10449 u16 uplink_seid;
10451 /* Set up the PF VSI associated with the PF's main VSI
10452 * that is already in the HW switch
10454 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10455 uplink_seid = pf->veb[pf->lan_veb]->seid;
10456 else
10457 uplink_seid = pf->mac_seid;
10458 if (pf->lan_vsi == I40E_NO_VSI)
10459 vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10460 else if (reinit)
10461 vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10462 if (!vsi) {
10463 dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10464 i40e_fdir_teardown(pf);
10465 return -EAGAIN;
10467 } else {
10468 /* force a reset of TC and queue layout configurations */
10469 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10471 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10472 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10473 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10475 i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10477 i40e_fdir_sb_setup(pf);
10479 /* Setup static PF queue filter control settings */
10480 ret = i40e_setup_pf_filter_control(pf);
10481 if (ret) {
10482 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10483 ret);
10484 /* Failure here should not stop continuing other steps */
10487 /* enable RSS in the HW, even for only one queue, as the stack can use
10488 * the hash
10490 if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10491 i40e_pf_config_rss(pf);
10493 /* fill in link information and enable LSE reporting */
10494 i40e_update_link_info(&pf->hw);
10495 i40e_link_event(pf);
10497 /* Initialize user-specific link properties */
10498 pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10499 I40E_AQ_AN_COMPLETED) ? true : false);
10501 i40e_ptp_init(pf);
10503 return ret;
10507 * i40e_determine_queue_usage - Work out queue distribution
10508 * @pf: board private structure
10510 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10512 int queues_left;
10514 pf->num_lan_qps = 0;
10515 #ifdef I40E_FCOE
10516 pf->num_fcoe_qps = 0;
10517 #endif
10519 /* Find the max queues to be put into basic use. We'll always be
10520 * using TC0, whether or not DCB is running, and TC0 will get the
10521 * big RSS set.
10523 queues_left = pf->hw.func_caps.num_tx_qp;
10525 if ((queues_left == 1) ||
10526 !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10527 /* one qp for PF, no queues for anything else */
10528 queues_left = 0;
10529 pf->alloc_rss_size = pf->num_lan_qps = 1;
10531 /* make sure all the fancies are disabled */
10532 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10533 I40E_FLAG_IWARP_ENABLED |
10534 #ifdef I40E_FCOE
10535 I40E_FLAG_FCOE_ENABLED |
10536 #endif
10537 I40E_FLAG_FD_SB_ENABLED |
10538 I40E_FLAG_FD_ATR_ENABLED |
10539 I40E_FLAG_DCB_CAPABLE |
10540 I40E_FLAG_DCB_ENABLED |
10541 I40E_FLAG_SRIOV_ENABLED |
10542 I40E_FLAG_VMDQ_ENABLED);
10543 } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10544 I40E_FLAG_FD_SB_ENABLED |
10545 I40E_FLAG_FD_ATR_ENABLED |
10546 I40E_FLAG_DCB_CAPABLE))) {
10547 /* one qp for PF */
10548 pf->alloc_rss_size = pf->num_lan_qps = 1;
10549 queues_left -= pf->num_lan_qps;
10551 pf->flags &= ~(I40E_FLAG_RSS_ENABLED |
10552 I40E_FLAG_IWARP_ENABLED |
10553 #ifdef I40E_FCOE
10554 I40E_FLAG_FCOE_ENABLED |
10555 #endif
10556 I40E_FLAG_FD_SB_ENABLED |
10557 I40E_FLAG_FD_ATR_ENABLED |
10558 I40E_FLAG_DCB_ENABLED |
10559 I40E_FLAG_VMDQ_ENABLED);
10560 } else {
10561 /* Not enough queues for all TCs */
10562 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10563 (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10564 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10565 I40E_FLAG_DCB_ENABLED);
10566 dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10568 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10569 num_online_cpus());
10570 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10571 pf->hw.func_caps.num_tx_qp);
10573 queues_left -= pf->num_lan_qps;
10576 #ifdef I40E_FCOE
10577 if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10578 if (I40E_DEFAULT_FCOE <= queues_left) {
10579 pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10580 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10581 pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10582 } else {
10583 pf->num_fcoe_qps = 0;
10584 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10585 dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10588 queues_left -= pf->num_fcoe_qps;
10591 #endif
10592 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10593 if (queues_left > 1) {
10594 queues_left -= 1; /* save 1 queue for FD */
10595 } else {
10596 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10597 dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10601 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10602 pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10603 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10604 (queues_left / pf->num_vf_qps));
10605 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10608 if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10609 pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10610 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10611 (queues_left / pf->num_vmdq_qps));
10612 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10615 pf->queues_left = queues_left;
10616 dev_dbg(&pf->pdev->dev,
10617 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10618 pf->hw.func_caps.num_tx_qp,
10619 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10620 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10621 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10622 queues_left);
10623 #ifdef I40E_FCOE
10624 dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10625 #endif
10629 * i40e_setup_pf_filter_control - Setup PF static filter control
10630 * @pf: PF to be setup
10632 * i40e_setup_pf_filter_control sets up a PF's initial filter control
10633 * settings. If PE/FCoE are enabled then it will also set the per PF
10634 * based filter sizes required for them. It also enables Flow director,
10635 * ethertype and macvlan type filter settings for the pf.
10637 * Returns 0 on success, negative on failure
10639 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10641 struct i40e_filter_control_settings *settings = &pf->filter_settings;
10643 settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10645 /* Flow Director is enabled */
10646 if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10647 settings->enable_fdir = true;
10649 /* Ethtype and MACVLAN filters enabled for PF */
10650 settings->enable_ethtype = true;
10651 settings->enable_macvlan = true;
10653 if (i40e_set_filter_control(&pf->hw, settings))
10654 return -ENOENT;
10656 return 0;
10659 #define INFO_STRING_LEN 255
10660 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10661 static void i40e_print_features(struct i40e_pf *pf)
10663 struct i40e_hw *hw = &pf->hw;
10664 char *buf;
10665 int i;
10667 buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10668 if (!buf)
10669 return;
10671 i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10672 #ifdef CONFIG_PCI_IOV
10673 i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10674 #endif
10675 i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10676 pf->hw.func_caps.num_vsis,
10677 pf->vsi[pf->lan_vsi]->num_queue_pairs);
10678 if (pf->flags & I40E_FLAG_RSS_ENABLED)
10679 i += snprintf(&buf[i], REMAIN(i), " RSS");
10680 if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10681 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10682 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10683 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10684 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10686 if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10687 i += snprintf(&buf[i], REMAIN(i), " DCB");
10688 i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10689 i += snprintf(&buf[i], REMAIN(i), " Geneve");
10690 if (pf->flags & I40E_FLAG_PTP)
10691 i += snprintf(&buf[i], REMAIN(i), " PTP");
10692 #ifdef I40E_FCOE
10693 if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10694 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10695 #endif
10696 if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10697 i += snprintf(&buf[i], REMAIN(i), " VEB");
10698 else
10699 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10701 dev_info(&pf->pdev->dev, "%s\n", buf);
10702 kfree(buf);
10703 WARN_ON(i > INFO_STRING_LEN);
10707 * i40e_get_platform_mac_addr - get platform-specific MAC address
10709 * @pdev: PCI device information struct
10710 * @pf: board private structure
10712 * Look up the MAC address in Open Firmware on systems that support it,
10713 * and use IDPROM on SPARC if no OF address is found. On return, the
10714 * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10715 * has been selected.
10717 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10719 pf->flags &= ~I40E_FLAG_PF_MAC;
10720 if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10721 pf->flags |= I40E_FLAG_PF_MAC;
10725 * i40e_probe - Device initialization routine
10726 * @pdev: PCI device information struct
10727 * @ent: entry in i40e_pci_tbl
10729 * i40e_probe initializes a PF identified by a pci_dev structure.
10730 * The OS initialization, configuring of the PF private structure,
10731 * and a hardware reset occur.
10733 * Returns 0 on success, negative on failure
10735 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10737 struct i40e_aq_get_phy_abilities_resp abilities;
10738 struct i40e_pf *pf;
10739 struct i40e_hw *hw;
10740 static u16 pfs_found;
10741 u16 wol_nvm_bits;
10742 u16 link_status;
10743 int err;
10744 u32 val;
10745 u32 i;
10746 u8 set_fc_aq_fail;
10748 err = pci_enable_device_mem(pdev);
10749 if (err)
10750 return err;
10752 /* set up for high or low dma */
10753 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10754 if (err) {
10755 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10756 if (err) {
10757 dev_err(&pdev->dev,
10758 "DMA configuration failed: 0x%x\n", err);
10759 goto err_dma;
10763 /* set up pci connections */
10764 err = pci_request_mem_regions(pdev, i40e_driver_name);
10765 if (err) {
10766 dev_info(&pdev->dev,
10767 "pci_request_selected_regions failed %d\n", err);
10768 goto err_pci_reg;
10771 pci_enable_pcie_error_reporting(pdev);
10772 pci_set_master(pdev);
10774 /* Now that we have a PCI connection, we need to do the
10775 * low level device setup. This is primarily setting up
10776 * the Admin Queue structures and then querying for the
10777 * device's current profile information.
10779 pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10780 if (!pf) {
10781 err = -ENOMEM;
10782 goto err_pf_alloc;
10784 pf->next_vsi = 0;
10785 pf->pdev = pdev;
10786 set_bit(__I40E_DOWN, &pf->state);
10788 hw = &pf->hw;
10789 hw->back = pf;
10791 pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10792 I40E_MAX_CSR_SPACE);
10794 hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10795 if (!hw->hw_addr) {
10796 err = -EIO;
10797 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10798 (unsigned int)pci_resource_start(pdev, 0),
10799 pf->ioremap_len, err);
10800 goto err_ioremap;
10802 hw->vendor_id = pdev->vendor;
10803 hw->device_id = pdev->device;
10804 pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10805 hw->subsystem_vendor_id = pdev->subsystem_vendor;
10806 hw->subsystem_device_id = pdev->subsystem_device;
10807 hw->bus.device = PCI_SLOT(pdev->devfn);
10808 hw->bus.func = PCI_FUNC(pdev->devfn);
10809 pf->instance = pfs_found;
10811 /* set up the locks for the AQ, do this only once in probe
10812 * and destroy them only once in remove
10814 mutex_init(&hw->aq.asq_mutex);
10815 mutex_init(&hw->aq.arq_mutex);
10817 if (debug != -1) {
10818 pf->msg_enable = pf->hw.debug_mask;
10819 pf->msg_enable = debug;
10822 /* do a special CORER for clearing PXE mode once at init */
10823 if (hw->revision_id == 0 &&
10824 (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10825 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10826 i40e_flush(hw);
10827 msleep(200);
10828 pf->corer_count++;
10830 i40e_clear_pxe_mode(hw);
10833 /* Reset here to make sure all is clean and to define PF 'n' */
10834 i40e_clear_hw(hw);
10835 err = i40e_pf_reset(hw);
10836 if (err) {
10837 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10838 goto err_pf_reset;
10840 pf->pfr_count++;
10842 hw->aq.num_arq_entries = I40E_AQ_LEN;
10843 hw->aq.num_asq_entries = I40E_AQ_LEN;
10844 hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10845 hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10846 pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10848 snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10849 "%s-%s:misc",
10850 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10852 err = i40e_init_shared_code(hw);
10853 if (err) {
10854 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10855 err);
10856 goto err_pf_reset;
10859 /* set up a default setting for link flow control */
10860 pf->hw.fc.requested_mode = I40E_FC_NONE;
10862 err = i40e_init_adminq(hw);
10863 if (err) {
10864 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10865 dev_info(&pdev->dev,
10866 "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");
10867 else
10868 dev_info(&pdev->dev,
10869 "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10871 goto err_pf_reset;
10874 /* provide nvm, fw, api versions */
10875 dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10876 hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10877 hw->aq.api_maj_ver, hw->aq.api_min_ver,
10878 i40e_nvm_version_str(hw));
10880 if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10881 hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10882 dev_info(&pdev->dev,
10883 "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");
10884 else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10885 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10886 dev_info(&pdev->dev,
10887 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10889 i40e_verify_eeprom(pf);
10891 /* Rev 0 hardware was never productized */
10892 if (hw->revision_id < 1)
10893 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");
10895 i40e_clear_pxe_mode(hw);
10896 err = i40e_get_capabilities(pf);
10897 if (err)
10898 goto err_adminq_setup;
10900 err = i40e_sw_init(pf);
10901 if (err) {
10902 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10903 goto err_sw_init;
10906 err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10907 hw->func_caps.num_rx_qp,
10908 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10909 if (err) {
10910 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10911 goto err_init_lan_hmc;
10914 err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10915 if (err) {
10916 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10917 err = -ENOENT;
10918 goto err_configure_lan_hmc;
10921 /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10922 * Ignore error return codes because if it was already disabled via
10923 * hardware settings this will fail
10925 if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10926 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10927 i40e_aq_stop_lldp(hw, true, NULL);
10930 i40e_get_mac_addr(hw, hw->mac.addr);
10931 /* allow a platform config to override the HW addr */
10932 i40e_get_platform_mac_addr(pdev, pf);
10933 if (!is_valid_ether_addr(hw->mac.addr)) {
10934 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10935 err = -EIO;
10936 goto err_mac_addr;
10938 dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10939 ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10940 i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10941 if (is_valid_ether_addr(hw->mac.port_addr))
10942 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10943 #ifdef I40E_FCOE
10944 err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10945 if (err)
10946 dev_info(&pdev->dev,
10947 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10948 if (!is_valid_ether_addr(hw->mac.san_addr)) {
10949 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10950 hw->mac.san_addr);
10951 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10953 dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10954 #endif /* I40E_FCOE */
10956 pci_set_drvdata(pdev, pf);
10957 pci_save_state(pdev);
10958 #ifdef CONFIG_I40E_DCB
10959 err = i40e_init_pf_dcb(pf);
10960 if (err) {
10961 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10962 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE & I40E_FLAG_DCB_ENABLED);
10963 /* Continue without DCB enabled */
10965 #endif /* CONFIG_I40E_DCB */
10967 /* set up periodic task facility */
10968 setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10969 pf->service_timer_period = HZ;
10971 INIT_WORK(&pf->service_task, i40e_service_task);
10972 clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10973 pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10975 /* NVM bit on means WoL disabled for the port */
10976 i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10977 if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10978 pf->wol_en = false;
10979 else
10980 pf->wol_en = true;
10981 device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10983 /* set up the main switch operations */
10984 i40e_determine_queue_usage(pf);
10985 err = i40e_init_interrupt_scheme(pf);
10986 if (err)
10987 goto err_switch_setup;
10989 /* The number of VSIs reported by the FW is the minimum guaranteed
10990 * to us; HW supports far more and we share the remaining pool with
10991 * the other PFs. We allocate space for more than the guarantee with
10992 * the understanding that we might not get them all later.
10994 if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10995 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10996 else
10997 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10999 /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
11000 pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11001 GFP_KERNEL);
11002 if (!pf->vsi) {
11003 err = -ENOMEM;
11004 goto err_switch_setup;
11007 #ifdef CONFIG_PCI_IOV
11008 /* prep for VF support */
11009 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11010 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11011 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11012 if (pci_num_vf(pdev))
11013 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11015 #endif
11016 err = i40e_setup_pf_switch(pf, false);
11017 if (err) {
11018 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11019 goto err_vsis;
11022 /* Make sure flow control is set according to current settings */
11023 err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11024 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11025 dev_dbg(&pf->pdev->dev,
11026 "Set fc with err %s aq_err %s on get_phy_cap\n",
11027 i40e_stat_str(hw, err),
11028 i40e_aq_str(hw, hw->aq.asq_last_status));
11029 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11030 dev_dbg(&pf->pdev->dev,
11031 "Set fc with err %s aq_err %s on set_phy_config\n",
11032 i40e_stat_str(hw, err),
11033 i40e_aq_str(hw, hw->aq.asq_last_status));
11034 if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11035 dev_dbg(&pf->pdev->dev,
11036 "Set fc with err %s aq_err %s on get_link_info\n",
11037 i40e_stat_str(hw, err),
11038 i40e_aq_str(hw, hw->aq.asq_last_status));
11040 /* if FDIR VSI was set up, start it now */
11041 for (i = 0; i < pf->num_alloc_vsi; i++) {
11042 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11043 i40e_vsi_open(pf->vsi[i]);
11044 break;
11048 /* The driver only wants link up/down and module qualification
11049 * reports from firmware. Note the negative logic.
11051 err = i40e_aq_set_phy_int_mask(&pf->hw,
11052 ~(I40E_AQ_EVENT_LINK_UPDOWN |
11053 I40E_AQ_EVENT_MEDIA_NA |
11054 I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11055 if (err)
11056 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11057 i40e_stat_str(&pf->hw, err),
11058 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11060 /* Reconfigure hardware for allowing smaller MSS in the case
11061 * of TSO, so that we avoid the MDD being fired and causing
11062 * a reset in the case of small MSS+TSO.
11064 val = rd32(hw, I40E_REG_MSS);
11065 if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11066 val &= ~I40E_REG_MSS_MIN_MASK;
11067 val |= I40E_64BYTE_MSS;
11068 wr32(hw, I40E_REG_MSS, val);
11071 if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11072 msleep(75);
11073 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11074 if (err)
11075 dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11076 i40e_stat_str(&pf->hw, err),
11077 i40e_aq_str(&pf->hw,
11078 pf->hw.aq.asq_last_status));
11080 /* The main driver is (mostly) up and happy. We need to set this state
11081 * before setting up the misc vector or we get a race and the vector
11082 * ends up disabled forever.
11084 clear_bit(__I40E_DOWN, &pf->state);
11086 /* In case of MSIX we are going to setup the misc vector right here
11087 * to handle admin queue events etc. In case of legacy and MSI
11088 * the misc functionality and queue processing is combined in
11089 * the same vector and that gets setup at open.
11091 if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11092 err = i40e_setup_misc_vector(pf);
11093 if (err) {
11094 dev_info(&pdev->dev,
11095 "setup of misc vector failed: %d\n", err);
11096 goto err_vsis;
11100 #ifdef CONFIG_PCI_IOV
11101 /* prep for VF support */
11102 if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11103 (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11104 !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11105 /* disable link interrupts for VFs */
11106 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11107 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11108 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11109 i40e_flush(hw);
11111 if (pci_num_vf(pdev)) {
11112 dev_info(&pdev->dev,
11113 "Active VFs found, allocating resources.\n");
11114 err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11115 if (err)
11116 dev_info(&pdev->dev,
11117 "Error %d allocating resources for existing VFs\n",
11118 err);
11121 #endif /* CONFIG_PCI_IOV */
11123 if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11124 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11125 pf->num_iwarp_msix,
11126 I40E_IWARP_IRQ_PILE_ID);
11127 if (pf->iwarp_base_vector < 0) {
11128 dev_info(&pdev->dev,
11129 "failed to get tracking for %d vectors for IWARP err=%d\n",
11130 pf->num_iwarp_msix, pf->iwarp_base_vector);
11131 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11135 i40e_dbg_pf_init(pf);
11137 /* tell the firmware that we're starting */
11138 i40e_send_version(pf);
11140 /* since everything's happy, start the service_task timer */
11141 mod_timer(&pf->service_timer,
11142 round_jiffies(jiffies + pf->service_timer_period));
11144 /* add this PF to client device list and launch a client service task */
11145 err = i40e_lan_add_device(pf);
11146 if (err)
11147 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11148 err);
11150 #ifdef I40E_FCOE
11151 /* create FCoE interface */
11152 i40e_fcoe_vsi_setup(pf);
11154 #endif
11155 #define PCI_SPEED_SIZE 8
11156 #define PCI_WIDTH_SIZE 8
11157 /* Devices on the IOSF bus do not have this information
11158 * and will report PCI Gen 1 x 1 by default so don't bother
11159 * checking them.
11161 if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11162 char speed[PCI_SPEED_SIZE] = "Unknown";
11163 char width[PCI_WIDTH_SIZE] = "Unknown";
11165 /* Get the negotiated link width and speed from PCI config
11166 * space
11168 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11169 &link_status);
11171 i40e_set_pci_config_data(hw, link_status);
11173 switch (hw->bus.speed) {
11174 case i40e_bus_speed_8000:
11175 strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11176 case i40e_bus_speed_5000:
11177 strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11178 case i40e_bus_speed_2500:
11179 strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11180 default:
11181 break;
11183 switch (hw->bus.width) {
11184 case i40e_bus_width_pcie_x8:
11185 strncpy(width, "8", PCI_WIDTH_SIZE); break;
11186 case i40e_bus_width_pcie_x4:
11187 strncpy(width, "4", PCI_WIDTH_SIZE); break;
11188 case i40e_bus_width_pcie_x2:
11189 strncpy(width, "2", PCI_WIDTH_SIZE); break;
11190 case i40e_bus_width_pcie_x1:
11191 strncpy(width, "1", PCI_WIDTH_SIZE); break;
11192 default:
11193 break;
11196 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11197 speed, width);
11199 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11200 hw->bus.speed < i40e_bus_speed_8000) {
11201 dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11202 dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11206 /* get the requested speeds from the fw */
11207 err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11208 if (err)
11209 dev_dbg(&pf->pdev->dev, "get requested speeds ret = %s last_status = %s\n",
11210 i40e_stat_str(&pf->hw, err),
11211 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11212 pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11214 /* get the supported phy types from the fw */
11215 err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11216 if (err)
11217 dev_dbg(&pf->pdev->dev, "get supported phy types ret = %s last_status = %s\n",
11218 i40e_stat_str(&pf->hw, err),
11219 i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11220 pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11222 /* Add a filter to drop all Flow control frames from any VSI from being
11223 * transmitted. By doing so we stop a malicious VF from sending out
11224 * PAUSE or PFC frames and potentially controlling traffic for other
11225 * PF/VF VSIs.
11226 * The FW can still send Flow control frames if enabled.
11228 i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11229 pf->main_vsi_seid);
11231 if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11232 (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11233 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11235 /* print a string summarizing features */
11236 i40e_print_features(pf);
11238 return 0;
11240 /* Unwind what we've done if something failed in the setup */
11241 err_vsis:
11242 set_bit(__I40E_DOWN, &pf->state);
11243 i40e_clear_interrupt_scheme(pf);
11244 kfree(pf->vsi);
11245 err_switch_setup:
11246 i40e_reset_interrupt_capability(pf);
11247 del_timer_sync(&pf->service_timer);
11248 err_mac_addr:
11249 err_configure_lan_hmc:
11250 (void)i40e_shutdown_lan_hmc(hw);
11251 err_init_lan_hmc:
11252 kfree(pf->qp_pile);
11253 err_sw_init:
11254 err_adminq_setup:
11255 err_pf_reset:
11256 iounmap(hw->hw_addr);
11257 err_ioremap:
11258 kfree(pf);
11259 err_pf_alloc:
11260 pci_disable_pcie_error_reporting(pdev);
11261 pci_release_mem_regions(pdev);
11262 err_pci_reg:
11263 err_dma:
11264 pci_disable_device(pdev);
11265 return err;
11269 * i40e_remove - Device removal routine
11270 * @pdev: PCI device information struct
11272 * i40e_remove is called by the PCI subsystem to alert the driver
11273 * that is should release a PCI device. This could be caused by a
11274 * Hot-Plug event, or because the driver is going to be removed from
11275 * memory.
11277 static void i40e_remove(struct pci_dev *pdev)
11279 struct i40e_pf *pf = pci_get_drvdata(pdev);
11280 struct i40e_hw *hw = &pf->hw;
11281 i40e_status ret_code;
11282 int i;
11284 i40e_dbg_pf_exit(pf);
11286 i40e_ptp_stop(pf);
11288 /* Disable RSS in hw */
11289 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11290 i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11292 /* no more scheduling of any task */
11293 set_bit(__I40E_SUSPENDED, &pf->state);
11294 set_bit(__I40E_DOWN, &pf->state);
11295 if (pf->service_timer.data)
11296 del_timer_sync(&pf->service_timer);
11297 if (pf->service_task.func)
11298 cancel_work_sync(&pf->service_task);
11300 if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11301 i40e_free_vfs(pf);
11302 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11305 i40e_fdir_teardown(pf);
11307 /* If there is a switch structure or any orphans, remove them.
11308 * This will leave only the PF's VSI remaining.
11310 for (i = 0; i < I40E_MAX_VEB; i++) {
11311 if (!pf->veb[i])
11312 continue;
11314 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11315 pf->veb[i]->uplink_seid == 0)
11316 i40e_switch_branch_release(pf->veb[i]);
11319 /* Now we can shutdown the PF's VSI, just before we kill
11320 * adminq and hmc.
11322 if (pf->vsi[pf->lan_vsi])
11323 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11325 /* remove attached clients */
11326 ret_code = i40e_lan_del_device(pf);
11327 if (ret_code) {
11328 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11329 ret_code);
11332 /* shutdown and destroy the HMC */
11333 if (hw->hmc.hmc_obj) {
11334 ret_code = i40e_shutdown_lan_hmc(hw);
11335 if (ret_code)
11336 dev_warn(&pdev->dev,
11337 "Failed to destroy the HMC resources: %d\n",
11338 ret_code);
11341 /* shutdown the adminq */
11342 i40e_shutdown_adminq(hw);
11344 /* destroy the locks only once, here */
11345 mutex_destroy(&hw->aq.arq_mutex);
11346 mutex_destroy(&hw->aq.asq_mutex);
11348 /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11349 i40e_clear_interrupt_scheme(pf);
11350 for (i = 0; i < pf->num_alloc_vsi; i++) {
11351 if (pf->vsi[i]) {
11352 i40e_vsi_clear_rings(pf->vsi[i]);
11353 i40e_vsi_clear(pf->vsi[i]);
11354 pf->vsi[i] = NULL;
11358 for (i = 0; i < I40E_MAX_VEB; i++) {
11359 kfree(pf->veb[i]);
11360 pf->veb[i] = NULL;
11363 kfree(pf->qp_pile);
11364 kfree(pf->vsi);
11366 iounmap(hw->hw_addr);
11367 kfree(pf);
11368 pci_release_mem_regions(pdev);
11370 pci_disable_pcie_error_reporting(pdev);
11371 pci_disable_device(pdev);
11375 * i40e_pci_error_detected - warning that something funky happened in PCI land
11376 * @pdev: PCI device information struct
11378 * Called to warn that something happened and the error handling steps
11379 * are in progress. Allows the driver to quiesce things, be ready for
11380 * remediation.
11382 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11383 enum pci_channel_state error)
11385 struct i40e_pf *pf = pci_get_drvdata(pdev);
11387 dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11389 if (!pf) {
11390 dev_info(&pdev->dev,
11391 "Cannot recover - error happened during device probe\n");
11392 return PCI_ERS_RESULT_DISCONNECT;
11395 /* shutdown all operations */
11396 if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11397 rtnl_lock();
11398 i40e_prep_for_reset(pf);
11399 rtnl_unlock();
11402 /* Request a slot reset */
11403 return PCI_ERS_RESULT_NEED_RESET;
11407 * i40e_pci_error_slot_reset - a PCI slot reset just happened
11408 * @pdev: PCI device information struct
11410 * Called to find if the driver can work with the device now that
11411 * the pci slot has been reset. If a basic connection seems good
11412 * (registers are readable and have sane content) then return a
11413 * happy little PCI_ERS_RESULT_xxx.
11415 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11417 struct i40e_pf *pf = pci_get_drvdata(pdev);
11418 pci_ers_result_t result;
11419 int err;
11420 u32 reg;
11422 dev_dbg(&pdev->dev, "%s\n", __func__);
11423 if (pci_enable_device_mem(pdev)) {
11424 dev_info(&pdev->dev,
11425 "Cannot re-enable PCI device after reset.\n");
11426 result = PCI_ERS_RESULT_DISCONNECT;
11427 } else {
11428 pci_set_master(pdev);
11429 pci_restore_state(pdev);
11430 pci_save_state(pdev);
11431 pci_wake_from_d3(pdev, false);
11433 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11434 if (reg == 0)
11435 result = PCI_ERS_RESULT_RECOVERED;
11436 else
11437 result = PCI_ERS_RESULT_DISCONNECT;
11440 err = pci_cleanup_aer_uncorrect_error_status(pdev);
11441 if (err) {
11442 dev_info(&pdev->dev,
11443 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11444 err);
11445 /* non-fatal, continue */
11448 return result;
11452 * i40e_pci_error_resume - restart operations after PCI error recovery
11453 * @pdev: PCI device information struct
11455 * Called to allow the driver to bring things back up after PCI error
11456 * and/or reset recovery has finished.
11458 static void i40e_pci_error_resume(struct pci_dev *pdev)
11460 struct i40e_pf *pf = pci_get_drvdata(pdev);
11462 dev_dbg(&pdev->dev, "%s\n", __func__);
11463 if (test_bit(__I40E_SUSPENDED, &pf->state))
11464 return;
11466 rtnl_lock();
11467 i40e_handle_reset_warning(pf);
11468 rtnl_unlock();
11472 * i40e_shutdown - PCI callback for shutting down
11473 * @pdev: PCI device information struct
11475 static void i40e_shutdown(struct pci_dev *pdev)
11477 struct i40e_pf *pf = pci_get_drvdata(pdev);
11478 struct i40e_hw *hw = &pf->hw;
11480 set_bit(__I40E_SUSPENDED, &pf->state);
11481 set_bit(__I40E_DOWN, &pf->state);
11482 rtnl_lock();
11483 i40e_prep_for_reset(pf);
11484 rtnl_unlock();
11486 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11487 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11489 del_timer_sync(&pf->service_timer);
11490 cancel_work_sync(&pf->service_task);
11491 i40e_fdir_teardown(pf);
11493 rtnl_lock();
11494 i40e_prep_for_reset(pf);
11495 rtnl_unlock();
11497 wr32(hw, I40E_PFPM_APM,
11498 (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11499 wr32(hw, I40E_PFPM_WUFC,
11500 (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11502 i40e_clear_interrupt_scheme(pf);
11504 if (system_state == SYSTEM_POWER_OFF) {
11505 pci_wake_from_d3(pdev, pf->wol_en);
11506 pci_set_power_state(pdev, PCI_D3hot);
11510 #ifdef CONFIG_PM
11512 * i40e_suspend - PCI callback for moving to D3
11513 * @pdev: PCI device information struct
11515 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11517 struct i40e_pf *pf = pci_get_drvdata(pdev);
11518 struct i40e_hw *hw = &pf->hw;
11519 int retval = 0;
11521 set_bit(__I40E_SUSPENDED, &pf->state);
11522 set_bit(__I40E_DOWN, &pf->state);
11524 rtnl_lock();
11525 i40e_prep_for_reset(pf);
11526 rtnl_unlock();
11528 wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11529 wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11531 i40e_stop_misc_vector(pf);
11533 retval = pci_save_state(pdev);
11534 if (retval)
11535 return retval;
11537 pci_wake_from_d3(pdev, pf->wol_en);
11538 pci_set_power_state(pdev, PCI_D3hot);
11540 return retval;
11544 * i40e_resume - PCI callback for waking up from D3
11545 * @pdev: PCI device information struct
11547 static int i40e_resume(struct pci_dev *pdev)
11549 struct i40e_pf *pf = pci_get_drvdata(pdev);
11550 u32 err;
11552 pci_set_power_state(pdev, PCI_D0);
11553 pci_restore_state(pdev);
11554 /* pci_restore_state() clears dev->state_saves, so
11555 * call pci_save_state() again to restore it.
11557 pci_save_state(pdev);
11559 err = pci_enable_device_mem(pdev);
11560 if (err) {
11561 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11562 return err;
11564 pci_set_master(pdev);
11566 /* no wakeup events while running */
11567 pci_wake_from_d3(pdev, false);
11569 /* handling the reset will rebuild the device state */
11570 if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11571 clear_bit(__I40E_DOWN, &pf->state);
11572 rtnl_lock();
11573 i40e_reset_and_rebuild(pf, false);
11574 rtnl_unlock();
11577 return 0;
11580 #endif
11581 static const struct pci_error_handlers i40e_err_handler = {
11582 .error_detected = i40e_pci_error_detected,
11583 .slot_reset = i40e_pci_error_slot_reset,
11584 .resume = i40e_pci_error_resume,
11587 static struct pci_driver i40e_driver = {
11588 .name = i40e_driver_name,
11589 .id_table = i40e_pci_tbl,
11590 .probe = i40e_probe,
11591 .remove = i40e_remove,
11592 #ifdef CONFIG_PM
11593 .suspend = i40e_suspend,
11594 .resume = i40e_resume,
11595 #endif
11596 .shutdown = i40e_shutdown,
11597 .err_handler = &i40e_err_handler,
11598 .sriov_configure = i40e_pci_sriov_configure,
11602 * i40e_init_module - Driver registration routine
11604 * i40e_init_module is the first routine called when the driver is
11605 * loaded. All it does is register with the PCI subsystem.
11607 static int __init i40e_init_module(void)
11609 pr_info("%s: %s - version %s\n", i40e_driver_name,
11610 i40e_driver_string, i40e_driver_version_str);
11611 pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11613 /* we will see if single thread per module is enough for now,
11614 * it can't be any worse than using the system workqueue which
11615 * was already single threaded
11617 i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
11618 i40e_driver_name);
11619 if (!i40e_wq) {
11620 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11621 return -ENOMEM;
11624 i40e_dbg_init();
11625 return pci_register_driver(&i40e_driver);
11627 module_init(i40e_init_module);
11630 * i40e_exit_module - Driver exit cleanup routine
11632 * i40e_exit_module is called just before the driver is removed
11633 * from memory.
11635 static void __exit i40e_exit_module(void)
11637 pci_unregister_driver(&i40e_driver);
11638 destroy_workqueue(i40e_wq);
11639 i40e_dbg_exit();
11641 module_exit(i40e_exit_module);