1/*******************************************************************************
2 *
3 * Intel Ethernet Controller XL710 Family Linux Driver
4 * Copyright(c) 2013 - 2015 Intel Corporation.
5 *
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.
9 *
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.
14 *
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/>.
17 *
18 * The full GNU General Public License is included in this distribution in
19 * the file called "COPYING".
20 *
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
24 *
25 ******************************************************************************/
26
27/* Local includes */
28#include "i40e.h"
29#include "i40e_diag.h"
30#ifdef CONFIG_I40E_VXLAN
31#include <net/vxlan.h>
32#endif
33
34const char i40e_driver_name[] = "i40e";
35static const char i40e_driver_string[] =
36			"Intel(R) Ethernet Connection XL710 Network Driver";
37
38#define DRV_KERN "-k"
39
40#define DRV_VERSION_MAJOR 1
41#define DRV_VERSION_MINOR 3
42#define DRV_VERSION_BUILD 2
43#define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
44	     __stringify(DRV_VERSION_MINOR) "." \
45	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
46const char i40e_driver_version_str[] = DRV_VERSION;
47static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
48
49/* a bit of forward declarations */
50static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
51static void i40e_handle_reset_warning(struct i40e_pf *pf);
52static int i40e_add_vsi(struct i40e_vsi *vsi);
53static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
54static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
55static int i40e_setup_misc_vector(struct i40e_pf *pf);
56static void i40e_determine_queue_usage(struct i40e_pf *pf);
57static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
58static void i40e_fdir_sb_setup(struct i40e_pf *pf);
59static int i40e_veb_get_bw_info(struct i40e_veb *veb);
60
61/* i40e_pci_tbl - PCI Device ID Table
62 *
63 * Last entry must be all 0s
64 *
65 * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
66 *   Class, Class Mask, private data (not used) }
67 */
68static const struct pci_device_id i40e_pci_tbl[] = {
69	{PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
70	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
71	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_A), 0},
72	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
73	{PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
74	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
75	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
76	{PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
77	{PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
78	{PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
79	/* required last entry */
80	{0, }
81};
82MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
83
84#define I40E_MAX_VF_COUNT 128
85static int debug = -1;
86module_param(debug, int, 0);
87MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
88
89MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
90MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
91MODULE_LICENSE("GPL");
92MODULE_VERSION(DRV_VERSION);
93
94/**
95 * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
96 * @hw:   pointer to the HW structure
97 * @mem:  ptr to mem struct to fill out
98 * @size: size of memory requested
99 * @alignment: what to align the allocation to
100 **/
101int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
102			    u64 size, u32 alignment)
103{
104	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
105
106	mem->size = ALIGN(size, alignment);
107	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
108				      &mem->pa, GFP_KERNEL);
109	if (!mem->va)
110		return -ENOMEM;
111
112	return 0;
113}
114
115/**
116 * i40e_free_dma_mem_d - OS specific memory free for shared code
117 * @hw:   pointer to the HW structure
118 * @mem:  ptr to mem struct to free
119 **/
120int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
121{
122	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
123
124	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
125	mem->va = NULL;
126	mem->pa = 0;
127	mem->size = 0;
128
129	return 0;
130}
131
132/**
133 * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
134 * @hw:   pointer to the HW structure
135 * @mem:  ptr to mem struct to fill out
136 * @size: size of memory requested
137 **/
138int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
139			     u32 size)
140{
141	mem->size = size;
142	mem->va = kzalloc(size, GFP_KERNEL);
143
144	if (!mem->va)
145		return -ENOMEM;
146
147	return 0;
148}
149
150/**
151 * i40e_free_virt_mem_d - OS specific memory free for shared code
152 * @hw:   pointer to the HW structure
153 * @mem:  ptr to mem struct to free
154 **/
155int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
156{
157	/* it's ok to kfree a NULL pointer */
158	kfree(mem->va);
159	mem->va = NULL;
160	mem->size = 0;
161
162	return 0;
163}
164
165/**
166 * i40e_get_lump - find a lump of free generic resource
167 * @pf: board private structure
168 * @pile: the pile of resource to search
169 * @needed: the number of items needed
170 * @id: an owner id to stick on the items assigned
171 *
172 * Returns the base item index of the lump, or negative for error
173 *
174 * The search_hint trick and lack of advanced fit-finding only work
175 * because we're highly likely to have all the same size lump requests.
176 * Linear search time and any fragmentation should be minimal.
177 **/
178static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
179			 u16 needed, u16 id)
180{
181	int ret = -ENOMEM;
182	int i, j;
183
184	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
185		dev_info(&pf->pdev->dev,
186			 "param err: pile=%p needed=%d id=0x%04x\n",
187			 pile, needed, id);
188		return -EINVAL;
189	}
190
191	/* start the linear search with an imperfect hint */
192	i = pile->search_hint;
193	while (i < pile->num_entries) {
194		/* skip already allocated entries */
195		if (pile->list[i] & I40E_PILE_VALID_BIT) {
196			i++;
197			continue;
198		}
199
200		/* do we have enough in this lump? */
201		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
202			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
203				break;
204		}
205
206		if (j == needed) {
207			/* there was enough, so assign it to the requestor */
208			for (j = 0; j < needed; j++)
209				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
210			ret = i;
211			pile->search_hint = i + j;
212			break;
213		} else {
214			/* not enough, so skip over it and continue looking */
215			i += j;
216		}
217	}
218
219	return ret;
220}
221
222/**
223 * i40e_put_lump - return a lump of generic resource
224 * @pile: the pile of resource to search
225 * @index: the base item index
226 * @id: the owner id of the items assigned
227 *
228 * Returns the count of items in the lump
229 **/
230static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
231{
232	int valid_id = (id | I40E_PILE_VALID_BIT);
233	int count = 0;
234	int i;
235
236	if (!pile || index >= pile->num_entries)
237		return -EINVAL;
238
239	for (i = index;
240	     i < pile->num_entries && pile->list[i] == valid_id;
241	     i++) {
242		pile->list[i] = 0;
243		count++;
244	}
245
246	if (count && index < pile->search_hint)
247		pile->search_hint = index;
248
249	return count;
250}
251
252/**
253 * i40e_find_vsi_from_id - searches for the vsi with the given id
254 * @pf - the pf structure to search for the vsi
255 * @id - id of the vsi it is searching for
256 **/
257struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
258{
259	int i;
260
261	for (i = 0; i < pf->num_alloc_vsi; i++)
262		if (pf->vsi[i] && (pf->vsi[i]->id == id))
263			return pf->vsi[i];
264
265	return NULL;
266}
267
268/**
269 * i40e_service_event_schedule - Schedule the service task to wake up
270 * @pf: board private structure
271 *
272 * If not already scheduled, this puts the task into the work queue
273 **/
274static void i40e_service_event_schedule(struct i40e_pf *pf)
275{
276	if (!test_bit(__I40E_DOWN, &pf->state) &&
277	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
278	    !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
279		schedule_work(&pf->service_task);
280}
281
282/**
283 * i40e_tx_timeout - Respond to a Tx Hang
284 * @netdev: network interface device structure
285 *
286 * If any port has noticed a Tx timeout, it is likely that the whole
287 * device is munged, not just the one netdev port, so go for the full
288 * reset.
289 **/
290#ifdef I40E_FCOE
291void i40e_tx_timeout(struct net_device *netdev)
292#else
293static void i40e_tx_timeout(struct net_device *netdev)
294#endif
295{
296	struct i40e_netdev_priv *np = netdev_priv(netdev);
297	struct i40e_vsi *vsi = np->vsi;
298	struct i40e_pf *pf = vsi->back;
299
300	pf->tx_timeout_count++;
301
302	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
303		pf->tx_timeout_recovery_level = 1;
304	pf->tx_timeout_last_recovery = jiffies;
305	netdev_info(netdev, "tx_timeout recovery level %d\n",
306		    pf->tx_timeout_recovery_level);
307
308	switch (pf->tx_timeout_recovery_level) {
309	case 0:
310		/* disable and re-enable queues for the VSI */
311		if (in_interrupt()) {
312			set_bit(__I40E_REINIT_REQUESTED, &pf->state);
313			set_bit(__I40E_REINIT_REQUESTED, &vsi->state);
314		} else {
315			i40e_vsi_reinit_locked(vsi);
316		}
317		break;
318	case 1:
319		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
320		break;
321	case 2:
322		set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
323		break;
324	case 3:
325		set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
326		break;
327	default:
328		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
329		set_bit(__I40E_DOWN_REQUESTED, &pf->state);
330		set_bit(__I40E_DOWN_REQUESTED, &vsi->state);
331		break;
332	}
333	i40e_service_event_schedule(pf);
334	pf->tx_timeout_recovery_level++;
335}
336
337/**
338 * i40e_release_rx_desc - Store the new tail and head values
339 * @rx_ring: ring to bump
340 * @val: new head index
341 **/
342static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
343{
344	rx_ring->next_to_use = val;
345
346	/* Force memory writes to complete before letting h/w
347	 * know there are new descriptors to fetch.  (Only
348	 * applicable for weak-ordered memory model archs,
349	 * such as IA-64).
350	 */
351	wmb();
352	writel(val, rx_ring->tail);
353}
354
355/**
356 * i40e_get_vsi_stats_struct - Get System Network Statistics
357 * @vsi: the VSI we care about
358 *
359 * Returns the address of the device statistics structure.
360 * The statistics are actually updated from the service task.
361 **/
362struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
363{
364	return &vsi->net_stats;
365}
366
367/**
368 * i40e_get_netdev_stats_struct - Get statistics for netdev interface
369 * @netdev: network interface device structure
370 *
371 * Returns the address of the device statistics structure.
372 * The statistics are actually updated from the service task.
373 **/
374#ifdef I40E_FCOE
375struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
376					     struct net_device *netdev,
377					     struct rtnl_link_stats64 *stats)
378#else
379static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
380					     struct net_device *netdev,
381					     struct rtnl_link_stats64 *stats)
382#endif
383{
384	struct i40e_netdev_priv *np = netdev_priv(netdev);
385	struct i40e_ring *tx_ring, *rx_ring;
386	struct i40e_vsi *vsi = np->vsi;
387	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
388	int i;
389
390	if (test_bit(__I40E_DOWN, &vsi->state))
391		return stats;
392
393	if (!vsi->tx_rings)
394		return stats;
395
396	rcu_read_lock();
397	for (i = 0; i < vsi->num_queue_pairs; i++) {
398		u64 bytes, packets;
399		unsigned int start;
400
401		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
402		if (!tx_ring)
403			continue;
404
405		do {
406			start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
407			packets = tx_ring->stats.packets;
408			bytes   = tx_ring->stats.bytes;
409		} while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
410
411		stats->tx_packets += packets;
412		stats->tx_bytes   += bytes;
413		rx_ring = &tx_ring[1];
414
415		do {
416			start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
417			packets = rx_ring->stats.packets;
418			bytes   = rx_ring->stats.bytes;
419		} while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
420
421		stats->rx_packets += packets;
422		stats->rx_bytes   += bytes;
423	}
424	rcu_read_unlock();
425
426	/* following stats updated by i40e_watchdog_subtask() */
427	stats->multicast	= vsi_stats->multicast;
428	stats->tx_errors	= vsi_stats->tx_errors;
429	stats->tx_dropped	= vsi_stats->tx_dropped;
430	stats->rx_errors	= vsi_stats->rx_errors;
431	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
432	stats->rx_length_errors	= vsi_stats->rx_length_errors;
433
434	return stats;
435}
436
437/**
438 * i40e_vsi_reset_stats - Resets all stats of the given vsi
439 * @vsi: the VSI to have its stats reset
440 **/
441void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
442{
443	struct rtnl_link_stats64 *ns;
444	int i;
445
446	if (!vsi)
447		return;
448
449	ns = i40e_get_vsi_stats_struct(vsi);
450	memset(ns, 0, sizeof(*ns));
451	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
452	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
453	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
454	if (vsi->rx_rings && vsi->rx_rings[0]) {
455		for (i = 0; i < vsi->num_queue_pairs; i++) {
456			memset(&vsi->rx_rings[i]->stats, 0 ,
457			       sizeof(vsi->rx_rings[i]->stats));
458			memset(&vsi->rx_rings[i]->rx_stats, 0 ,
459			       sizeof(vsi->rx_rings[i]->rx_stats));
460			memset(&vsi->tx_rings[i]->stats, 0 ,
461			       sizeof(vsi->tx_rings[i]->stats));
462			memset(&vsi->tx_rings[i]->tx_stats, 0,
463			       sizeof(vsi->tx_rings[i]->tx_stats));
464		}
465	}
466	vsi->stat_offsets_loaded = false;
467}
468
469/**
470 * i40e_pf_reset_stats - Reset all of the stats for the given PF
471 * @pf: the PF to be reset
472 **/
473void i40e_pf_reset_stats(struct i40e_pf *pf)
474{
475	int i;
476
477	memset(&pf->stats, 0, sizeof(pf->stats));
478	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
479	pf->stat_offsets_loaded = false;
480
481	for (i = 0; i < I40E_MAX_VEB; i++) {
482		if (pf->veb[i]) {
483			memset(&pf->veb[i]->stats, 0,
484			       sizeof(pf->veb[i]->stats));
485			memset(&pf->veb[i]->stats_offsets, 0,
486			       sizeof(pf->veb[i]->stats_offsets));
487			pf->veb[i]->stat_offsets_loaded = false;
488		}
489	}
490}
491
492/**
493 * i40e_stat_update48 - read and update a 48 bit stat from the chip
494 * @hw: ptr to the hardware info
495 * @hireg: the high 32 bit reg to read
496 * @loreg: the low 32 bit reg to read
497 * @offset_loaded: has the initial offset been loaded yet
498 * @offset: ptr to current offset value
499 * @stat: ptr to the stat
500 *
501 * Since the device stats are not reset at PFReset, they likely will not
502 * be zeroed when the driver starts.  We'll save the first values read
503 * and use them as offsets to be subtracted from the raw values in order
504 * to report stats that count from zero.  In the process, we also manage
505 * the potential roll-over.
506 **/
507static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
508			       bool offset_loaded, u64 *offset, u64 *stat)
509{
510	u64 new_data;
511
512	if (hw->device_id == I40E_DEV_ID_QEMU) {
513		new_data = rd32(hw, loreg);
514		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
515	} else {
516		new_data = rd64(hw, loreg);
517	}
518	if (!offset_loaded)
519		*offset = new_data;
520	if (likely(new_data >= *offset))
521		*stat = new_data - *offset;
522	else
523		*stat = (new_data + ((u64)1 << 48)) - *offset;
524	*stat &= 0xFFFFFFFFFFFFULL;
525}
526
527/**
528 * i40e_stat_update32 - read and update a 32 bit stat from the chip
529 * @hw: ptr to the hardware info
530 * @reg: the hw reg to read
531 * @offset_loaded: has the initial offset been loaded yet
532 * @offset: ptr to current offset value
533 * @stat: ptr to the stat
534 **/
535static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
536			       bool offset_loaded, u64 *offset, u64 *stat)
537{
538	u32 new_data;
539
540	new_data = rd32(hw, reg);
541	if (!offset_loaded)
542		*offset = new_data;
543	if (likely(new_data >= *offset))
544		*stat = (u32)(new_data - *offset);
545	else
546		*stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
547}
548
549/**
550 * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
551 * @vsi: the VSI to be updated
552 **/
553void i40e_update_eth_stats(struct i40e_vsi *vsi)
554{
555	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
556	struct i40e_pf *pf = vsi->back;
557	struct i40e_hw *hw = &pf->hw;
558	struct i40e_eth_stats *oes;
559	struct i40e_eth_stats *es;     /* device's eth stats */
560
561	es = &vsi->eth_stats;
562	oes = &vsi->eth_stats_offsets;
563
564	/* Gather up the stats that the hw collects */
565	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
566			   vsi->stat_offsets_loaded,
567			   &oes->tx_errors, &es->tx_errors);
568	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
569			   vsi->stat_offsets_loaded,
570			   &oes->rx_discards, &es->rx_discards);
571	i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
572			   vsi->stat_offsets_loaded,
573			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
574	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
575			   vsi->stat_offsets_loaded,
576			   &oes->tx_errors, &es->tx_errors);
577
578	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
579			   I40E_GLV_GORCL(stat_idx),
580			   vsi->stat_offsets_loaded,
581			   &oes->rx_bytes, &es->rx_bytes);
582	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
583			   I40E_GLV_UPRCL(stat_idx),
584			   vsi->stat_offsets_loaded,
585			   &oes->rx_unicast, &es->rx_unicast);
586	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
587			   I40E_GLV_MPRCL(stat_idx),
588			   vsi->stat_offsets_loaded,
589			   &oes->rx_multicast, &es->rx_multicast);
590	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
591			   I40E_GLV_BPRCL(stat_idx),
592			   vsi->stat_offsets_loaded,
593			   &oes->rx_broadcast, &es->rx_broadcast);
594
595	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
596			   I40E_GLV_GOTCL(stat_idx),
597			   vsi->stat_offsets_loaded,
598			   &oes->tx_bytes, &es->tx_bytes);
599	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
600			   I40E_GLV_UPTCL(stat_idx),
601			   vsi->stat_offsets_loaded,
602			   &oes->tx_unicast, &es->tx_unicast);
603	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
604			   I40E_GLV_MPTCL(stat_idx),
605			   vsi->stat_offsets_loaded,
606			   &oes->tx_multicast, &es->tx_multicast);
607	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
608			   I40E_GLV_BPTCL(stat_idx),
609			   vsi->stat_offsets_loaded,
610			   &oes->tx_broadcast, &es->tx_broadcast);
611	vsi->stat_offsets_loaded = true;
612}
613
614/**
615 * i40e_update_veb_stats - Update Switch component statistics
616 * @veb: the VEB being updated
617 **/
618static void i40e_update_veb_stats(struct i40e_veb *veb)
619{
620	struct i40e_pf *pf = veb->pf;
621	struct i40e_hw *hw = &pf->hw;
622	struct i40e_eth_stats *oes;
623	struct i40e_eth_stats *es;     /* device's eth stats */
624	int idx = 0;
625
626	idx = veb->stats_idx;
627	es = &veb->stats;
628	oes = &veb->stats_offsets;
629
630	/* Gather up the stats that the hw collects */
631	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
632			   veb->stat_offsets_loaded,
633			   &oes->tx_discards, &es->tx_discards);
634	if (hw->revision_id > 0)
635		i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
636				   veb->stat_offsets_loaded,
637				   &oes->rx_unknown_protocol,
638				   &es->rx_unknown_protocol);
639	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
640			   veb->stat_offsets_loaded,
641			   &oes->rx_bytes, &es->rx_bytes);
642	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
643			   veb->stat_offsets_loaded,
644			   &oes->rx_unicast, &es->rx_unicast);
645	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
646			   veb->stat_offsets_loaded,
647			   &oes->rx_multicast, &es->rx_multicast);
648	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
649			   veb->stat_offsets_loaded,
650			   &oes->rx_broadcast, &es->rx_broadcast);
651
652	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
653			   veb->stat_offsets_loaded,
654			   &oes->tx_bytes, &es->tx_bytes);
655	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
656			   veb->stat_offsets_loaded,
657			   &oes->tx_unicast, &es->tx_unicast);
658	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
659			   veb->stat_offsets_loaded,
660			   &oes->tx_multicast, &es->tx_multicast);
661	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
662			   veb->stat_offsets_loaded,
663			   &oes->tx_broadcast, &es->tx_broadcast);
664	veb->stat_offsets_loaded = true;
665}
666
667#ifdef I40E_FCOE
668/**
669 * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
670 * @vsi: the VSI that is capable of doing FCoE
671 **/
672static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
673{
674	struct i40e_pf *pf = vsi->back;
675	struct i40e_hw *hw = &pf->hw;
676	struct i40e_fcoe_stats *ofs;
677	struct i40e_fcoe_stats *fs;     /* device's eth stats */
678	int idx;
679
680	if (vsi->type != I40E_VSI_FCOE)
681		return;
682
683	idx = (pf->pf_seid - I40E_BASE_PF_SEID) + I40E_FCOE_PF_STAT_OFFSET;
684	fs = &vsi->fcoe_stats;
685	ofs = &vsi->fcoe_stats_offsets;
686
687	i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
688			   vsi->fcoe_stat_offsets_loaded,
689			   &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
690	i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
691			   vsi->fcoe_stat_offsets_loaded,
692			   &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
693	i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
694			   vsi->fcoe_stat_offsets_loaded,
695			   &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
696	i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
697			   vsi->fcoe_stat_offsets_loaded,
698			   &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
699	i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
700			   vsi->fcoe_stat_offsets_loaded,
701			   &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
702	i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
703			   vsi->fcoe_stat_offsets_loaded,
704			   &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
705	i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
706			   vsi->fcoe_stat_offsets_loaded,
707			   &ofs->fcoe_last_error, &fs->fcoe_last_error);
708	i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
709			   vsi->fcoe_stat_offsets_loaded,
710			   &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
711
712	vsi->fcoe_stat_offsets_loaded = true;
713}
714
715#endif
716/**
717 * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
718 * @pf: the corresponding PF
719 *
720 * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
721 **/
722static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
723{
724	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
725	struct i40e_hw_port_stats *nsd = &pf->stats;
726	struct i40e_hw *hw = &pf->hw;
727	u64 xoff = 0;
728	u16 i, v;
729
730	if ((hw->fc.current_mode != I40E_FC_FULL) &&
731	    (hw->fc.current_mode != I40E_FC_RX_PAUSE))
732		return;
733
734	xoff = nsd->link_xoff_rx;
735	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
736			   pf->stat_offsets_loaded,
737			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
738
739	/* No new LFC xoff rx */
740	if (!(nsd->link_xoff_rx - xoff))
741		return;
742
743	/* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
744	for (v = 0; v < pf->num_alloc_vsi; v++) {
745		struct i40e_vsi *vsi = pf->vsi[v];
746
747		if (!vsi || !vsi->tx_rings[0])
748			continue;
749
750		for (i = 0; i < vsi->num_queue_pairs; i++) {
751			struct i40e_ring *ring = vsi->tx_rings[i];
752			clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
753		}
754	}
755}
756
757/**
758 * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
759 * @pf: the corresponding PF
760 *
761 * Update the Rx XOFF counter (PAUSE frames) in PFC mode
762 **/
763static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
764{
765	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
766	struct i40e_hw_port_stats *nsd = &pf->stats;
767	bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
768	struct i40e_dcbx_config *dcb_cfg;
769	struct i40e_hw *hw = &pf->hw;
770	u16 i, v;
771	u8 tc;
772
773	dcb_cfg = &hw->local_dcbx_config;
774
775	/* See if DCB enabled with PFC TC */
776	if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
777	    !(dcb_cfg->pfc.pfcenable)) {
778		i40e_update_link_xoff_rx(pf);
779		return;
780	}
781
782	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
783		u64 prio_xoff = nsd->priority_xoff_rx[i];
784		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
785				   pf->stat_offsets_loaded,
786				   &osd->priority_xoff_rx[i],
787				   &nsd->priority_xoff_rx[i]);
788
789		/* No new PFC xoff rx */
790		if (!(nsd->priority_xoff_rx[i] - prio_xoff))
791			continue;
792		/* Get the TC for given priority */
793		tc = dcb_cfg->etscfg.prioritytable[i];
794		xoff[tc] = true;
795	}
796
797	/* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
798	for (v = 0; v < pf->num_alloc_vsi; v++) {
799		struct i40e_vsi *vsi = pf->vsi[v];
800
801		if (!vsi || !vsi->tx_rings[0])
802			continue;
803
804		for (i = 0; i < vsi->num_queue_pairs; i++) {
805			struct i40e_ring *ring = vsi->tx_rings[i];
806
807			tc = ring->dcb_tc;
808			if (xoff[tc])
809				clear_bit(__I40E_HANG_CHECK_ARMED,
810					  &ring->state);
811		}
812	}
813}
814
815/**
816 * i40e_update_vsi_stats - Update the vsi statistics counters.
817 * @vsi: the VSI to be updated
818 *
819 * There are a few instances where we store the same stat in a
820 * couple of different structs.  This is partly because we have
821 * the netdev stats that need to be filled out, which is slightly
822 * different from the "eth_stats" defined by the chip and used in
823 * VF communications.  We sort it out here.
824 **/
825static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
826{
827	struct i40e_pf *pf = vsi->back;
828	struct rtnl_link_stats64 *ons;
829	struct rtnl_link_stats64 *ns;   /* netdev stats */
830	struct i40e_eth_stats *oes;
831	struct i40e_eth_stats *es;     /* device's eth stats */
832	u32 tx_restart, tx_busy;
833	struct i40e_ring *p;
834	u32 rx_page, rx_buf;
835	u64 bytes, packets;
836	unsigned int start;
837	u64 rx_p, rx_b;
838	u64 tx_p, tx_b;
839	u16 q;
840
841	if (test_bit(__I40E_DOWN, &vsi->state) ||
842	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
843		return;
844
845	ns = i40e_get_vsi_stats_struct(vsi);
846	ons = &vsi->net_stats_offsets;
847	es = &vsi->eth_stats;
848	oes = &vsi->eth_stats_offsets;
849
850	/* Gather up the netdev and vsi stats that the driver collects
851	 * on the fly during packet processing
852	 */
853	rx_b = rx_p = 0;
854	tx_b = tx_p = 0;
855	tx_restart = tx_busy = 0;
856	rx_page = 0;
857	rx_buf = 0;
858	rcu_read_lock();
859	for (q = 0; q < vsi->num_queue_pairs; q++) {
860		/* locate Tx ring */
861		p = ACCESS_ONCE(vsi->tx_rings[q]);
862
863		do {
864			start = u64_stats_fetch_begin_irq(&p->syncp);
865			packets = p->stats.packets;
866			bytes = p->stats.bytes;
867		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
868		tx_b += bytes;
869		tx_p += packets;
870		tx_restart += p->tx_stats.restart_queue;
871		tx_busy += p->tx_stats.tx_busy;
872
873		/* Rx queue is part of the same block as Tx queue */
874		p = &p[1];
875		do {
876			start = u64_stats_fetch_begin_irq(&p->syncp);
877			packets = p->stats.packets;
878			bytes = p->stats.bytes;
879		} while (u64_stats_fetch_retry_irq(&p->syncp, start));
880		rx_b += bytes;
881		rx_p += packets;
882		rx_buf += p->rx_stats.alloc_buff_failed;
883		rx_page += p->rx_stats.alloc_page_failed;
884	}
885	rcu_read_unlock();
886	vsi->tx_restart = tx_restart;
887	vsi->tx_busy = tx_busy;
888	vsi->rx_page_failed = rx_page;
889	vsi->rx_buf_failed = rx_buf;
890
891	ns->rx_packets = rx_p;
892	ns->rx_bytes = rx_b;
893	ns->tx_packets = tx_p;
894	ns->tx_bytes = tx_b;
895
896	/* update netdev stats from eth stats */
897	i40e_update_eth_stats(vsi);
898	ons->tx_errors = oes->tx_errors;
899	ns->tx_errors = es->tx_errors;
900	ons->multicast = oes->rx_multicast;
901	ns->multicast = es->rx_multicast;
902	ons->rx_dropped = oes->rx_discards;
903	ns->rx_dropped = es->rx_discards;
904	ons->tx_dropped = oes->tx_discards;
905	ns->tx_dropped = es->tx_discards;
906
907	/* pull in a couple PF stats if this is the main vsi */
908	if (vsi == pf->vsi[pf->lan_vsi]) {
909		ns->rx_crc_errors = pf->stats.crc_errors;
910		ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
911		ns->rx_length_errors = pf->stats.rx_length_errors;
912	}
913}
914
915/**
916 * i40e_update_pf_stats - Update the PF statistics counters.
917 * @pf: the PF to be updated
918 **/
919static void i40e_update_pf_stats(struct i40e_pf *pf)
920{
921	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
922	struct i40e_hw_port_stats *nsd = &pf->stats;
923	struct i40e_hw *hw = &pf->hw;
924	u32 val;
925	int i;
926
927	i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
928			   I40E_GLPRT_GORCL(hw->port),
929			   pf->stat_offsets_loaded,
930			   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
931	i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
932			   I40E_GLPRT_GOTCL(hw->port),
933			   pf->stat_offsets_loaded,
934			   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
935	i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
936			   pf->stat_offsets_loaded,
937			   &osd->eth.rx_discards,
938			   &nsd->eth.rx_discards);
939	i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
940			   I40E_GLPRT_UPRCL(hw->port),
941			   pf->stat_offsets_loaded,
942			   &osd->eth.rx_unicast,
943			   &nsd->eth.rx_unicast);
944	i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
945			   I40E_GLPRT_MPRCL(hw->port),
946			   pf->stat_offsets_loaded,
947			   &osd->eth.rx_multicast,
948			   &nsd->eth.rx_multicast);
949	i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
950			   I40E_GLPRT_BPRCL(hw->port),
951			   pf->stat_offsets_loaded,
952			   &osd->eth.rx_broadcast,
953			   &nsd->eth.rx_broadcast);
954	i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
955			   I40E_GLPRT_UPTCL(hw->port),
956			   pf->stat_offsets_loaded,
957			   &osd->eth.tx_unicast,
958			   &nsd->eth.tx_unicast);
959	i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
960			   I40E_GLPRT_MPTCL(hw->port),
961			   pf->stat_offsets_loaded,
962			   &osd->eth.tx_multicast,
963			   &nsd->eth.tx_multicast);
964	i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
965			   I40E_GLPRT_BPTCL(hw->port),
966			   pf->stat_offsets_loaded,
967			   &osd->eth.tx_broadcast,
968			   &nsd->eth.tx_broadcast);
969
970	i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
971			   pf->stat_offsets_loaded,
972			   &osd->tx_dropped_link_down,
973			   &nsd->tx_dropped_link_down);
974
975	i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
976			   pf->stat_offsets_loaded,
977			   &osd->crc_errors, &nsd->crc_errors);
978
979	i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
980			   pf->stat_offsets_loaded,
981			   &osd->illegal_bytes, &nsd->illegal_bytes);
982
983	i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
984			   pf->stat_offsets_loaded,
985			   &osd->mac_local_faults,
986			   &nsd->mac_local_faults);
987	i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
988			   pf->stat_offsets_loaded,
989			   &osd->mac_remote_faults,
990			   &nsd->mac_remote_faults);
991
992	i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
993			   pf->stat_offsets_loaded,
994			   &osd->rx_length_errors,
995			   &nsd->rx_length_errors);
996
997	i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
998			   pf->stat_offsets_loaded,
999			   &osd->link_xon_rx, &nsd->link_xon_rx);
1000	i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
1001			   pf->stat_offsets_loaded,
1002			   &osd->link_xon_tx, &nsd->link_xon_tx);
1003	i40e_update_prio_xoff_rx(pf);  /* handles I40E_GLPRT_LXOFFRXC */
1004	i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
1005			   pf->stat_offsets_loaded,
1006			   &osd->link_xoff_tx, &nsd->link_xoff_tx);
1007
1008	for (i = 0; i < 8; i++) {
1009		i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
1010				   pf->stat_offsets_loaded,
1011				   &osd->priority_xon_rx[i],
1012				   &nsd->priority_xon_rx[i]);
1013		i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
1014				   pf->stat_offsets_loaded,
1015				   &osd->priority_xon_tx[i],
1016				   &nsd->priority_xon_tx[i]);
1017		i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1018				   pf->stat_offsets_loaded,
1019				   &osd->priority_xoff_tx[i],
1020				   &nsd->priority_xoff_tx[i]);
1021		i40e_stat_update32(hw,
1022				   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1023				   pf->stat_offsets_loaded,
1024				   &osd->priority_xon_2_xoff[i],
1025				   &nsd->priority_xon_2_xoff[i]);
1026	}
1027
1028	i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1029			   I40E_GLPRT_PRC64L(hw->port),
1030			   pf->stat_offsets_loaded,
1031			   &osd->rx_size_64, &nsd->rx_size_64);
1032	i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1033			   I40E_GLPRT_PRC127L(hw->port),
1034			   pf->stat_offsets_loaded,
1035			   &osd->rx_size_127, &nsd->rx_size_127);
1036	i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1037			   I40E_GLPRT_PRC255L(hw->port),
1038			   pf->stat_offsets_loaded,
1039			   &osd->rx_size_255, &nsd->rx_size_255);
1040	i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1041			   I40E_GLPRT_PRC511L(hw->port),
1042			   pf->stat_offsets_loaded,
1043			   &osd->rx_size_511, &nsd->rx_size_511);
1044	i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1045			   I40E_GLPRT_PRC1023L(hw->port),
1046			   pf->stat_offsets_loaded,
1047			   &osd->rx_size_1023, &nsd->rx_size_1023);
1048	i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1049			   I40E_GLPRT_PRC1522L(hw->port),
1050			   pf->stat_offsets_loaded,
1051			   &osd->rx_size_1522, &nsd->rx_size_1522);
1052	i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1053			   I40E_GLPRT_PRC9522L(hw->port),
1054			   pf->stat_offsets_loaded,
1055			   &osd->rx_size_big, &nsd->rx_size_big);
1056
1057	i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1058			   I40E_GLPRT_PTC64L(hw->port),
1059			   pf->stat_offsets_loaded,
1060			   &osd->tx_size_64, &nsd->tx_size_64);
1061	i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1062			   I40E_GLPRT_PTC127L(hw->port),
1063			   pf->stat_offsets_loaded,
1064			   &osd->tx_size_127, &nsd->tx_size_127);
1065	i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1066			   I40E_GLPRT_PTC255L(hw->port),
1067			   pf->stat_offsets_loaded,
1068			   &osd->tx_size_255, &nsd->tx_size_255);
1069	i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1070			   I40E_GLPRT_PTC511L(hw->port),
1071			   pf->stat_offsets_loaded,
1072			   &osd->tx_size_511, &nsd->tx_size_511);
1073	i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1074			   I40E_GLPRT_PTC1023L(hw->port),
1075			   pf->stat_offsets_loaded,
1076			   &osd->tx_size_1023, &nsd->tx_size_1023);
1077	i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1078			   I40E_GLPRT_PTC1522L(hw->port),
1079			   pf->stat_offsets_loaded,
1080			   &osd->tx_size_1522, &nsd->tx_size_1522);
1081	i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1082			   I40E_GLPRT_PTC9522L(hw->port),
1083			   pf->stat_offsets_loaded,
1084			   &osd->tx_size_big, &nsd->tx_size_big);
1085
1086	i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1087			   pf->stat_offsets_loaded,
1088			   &osd->rx_undersize, &nsd->rx_undersize);
1089	i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1090			   pf->stat_offsets_loaded,
1091			   &osd->rx_fragments, &nsd->rx_fragments);
1092	i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1093			   pf->stat_offsets_loaded,
1094			   &osd->rx_oversize, &nsd->rx_oversize);
1095	i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1096			   pf->stat_offsets_loaded,
1097			   &osd->rx_jabber, &nsd->rx_jabber);
1098
1099	/* FDIR stats */
1100	i40e_stat_update32(hw, I40E_GLQF_PCNT(pf->fd_atr_cnt_idx),
1101			   pf->stat_offsets_loaded,
1102			   &osd->fd_atr_match, &nsd->fd_atr_match);
1103	i40e_stat_update32(hw, I40E_GLQF_PCNT(pf->fd_sb_cnt_idx),
1104			   pf->stat_offsets_loaded,
1105			   &osd->fd_sb_match, &nsd->fd_sb_match);
1106
1107	val = rd32(hw, I40E_PRTPM_EEE_STAT);
1108	nsd->tx_lpi_status =
1109		       (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1110			I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1111	nsd->rx_lpi_status =
1112		       (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1113			I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1114	i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1115			   pf->stat_offsets_loaded,
1116			   &osd->tx_lpi_count, &nsd->tx_lpi_count);
1117	i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1118			   pf->stat_offsets_loaded,
1119			   &osd->rx_lpi_count, &nsd->rx_lpi_count);
1120
1121	pf->stat_offsets_loaded = true;
1122}
1123
1124/**
1125 * i40e_update_stats - Update the various statistics counters.
1126 * @vsi: the VSI to be updated
1127 *
1128 * Update the various stats for this VSI and its related entities.
1129 **/
1130void i40e_update_stats(struct i40e_vsi *vsi)
1131{
1132	struct i40e_pf *pf = vsi->back;
1133
1134	if (vsi == pf->vsi[pf->lan_vsi])
1135		i40e_update_pf_stats(pf);
1136
1137	i40e_update_vsi_stats(vsi);
1138#ifdef I40E_FCOE
1139	i40e_update_fcoe_stats(vsi);
1140#endif
1141}
1142
1143/**
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
1150 *
1151 * Returns ptr to the filter object or NULL
1152 **/
1153static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1154						u8 *macaddr, s16 vlan,
1155						bool is_vf, bool is_netdev)
1156{
1157	struct i40e_mac_filter *f;
1158
1159	if (!vsi || !macaddr)
1160		return NULL;
1161
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;
1168	}
1169	return NULL;
1170}
1171
1172/**
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
1178 *
1179 * Returns the first filter with the provided MAC address or NULL if
1180 * MAC address was not found
1181 **/
1182struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1183				      bool is_vf, bool is_netdev)
1184{
1185	struct i40e_mac_filter *f;
1186
1187	if (!vsi || !macaddr)
1188		return NULL;
1189
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;
1195	}
1196	return NULL;
1197}
1198
1199/**
1200 * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1201 * @vsi: the VSI to be searched
1202 *
1203 * Returns true if VSI is in vlan mode or false otherwise
1204 **/
1205bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1206{
1207	struct i40e_mac_filter *f;
1208
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
1211	 */
1212	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1213		if (f->vlan >= 0)
1214			return true;
1215	}
1216
1217	return false;
1218}
1219
1220/**
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
1226 *
1227 * Goes through all the macvlan filters and adds a
1228 * macvlan filter for each unique vlan that already exists
1229 *
1230 * Returns first filter found on success, else NULL
1231 **/
1232struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1233					     bool is_vf, bool is_netdev)
1234{
1235	struct i40e_mac_filter *f;
1236
1237	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1238		if (!i40e_find_filter(vsi, macaddr, f->vlan,
1239				      is_vf, is_netdev)) {
1240			if (!i40e_add_filter(vsi, macaddr, f->vlan,
1241					     is_vf, is_netdev))
1242				return NULL;
1243		}
1244	}
1245
1246	return list_first_entry_or_null(&vsi->mac_filter_list,
1247					struct i40e_mac_filter, list);
1248}
1249
1250/**
1251 * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1252 * @vsi: the PF Main VSI - inappropriate for any other VSI
1253 * @macaddr: the MAC address
1254 *
1255 * Some older firmware configurations set up a default promiscuous VLAN
1256 * filter that needs to be removed.
1257 **/
1258static int i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1259{
1260	struct i40e_aqc_remove_macvlan_element_data element;
1261	struct i40e_pf *pf = vsi->back;
1262	i40e_status aq_ret;
1263
1264	/* Only appropriate for the PF main VSI */
1265	if (vsi->type != I40E_VSI_MAIN)
1266		return -EINVAL;
1267
1268	memset(&element, 0, sizeof(element));
1269	ether_addr_copy(element.mac_addr, macaddr);
1270	element.vlan_tag = 0;
1271	element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1272			I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1273	aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1274	if (aq_ret)
1275		return -ENOENT;
1276
1277	return 0;
1278}
1279
1280/**
1281 * i40e_add_filter - Add a mac/vlan filter to the VSI
1282 * @vsi: the VSI to be searched
1283 * @macaddr: the MAC address
1284 * @vlan: the vlan
1285 * @is_vf: make sure its a VF filter, else doesn't matter
1286 * @is_netdev: make sure its a netdev filter, else doesn't matter
1287 *
1288 * Returns ptr to the filter object or NULL when no memory available.
1289 **/
1290struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1291					u8 *macaddr, s16 vlan,
1292					bool is_vf, bool is_netdev)
1293{
1294	struct i40e_mac_filter *f;
1295
1296	if (!vsi || !macaddr)
1297		return NULL;
1298
1299	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1300	if (!f) {
1301		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1302		if (!f)
1303			goto add_filter_out;
1304
1305		ether_addr_copy(f->macaddr, macaddr);
1306		f->vlan = vlan;
1307		f->changed = true;
1308
1309		INIT_LIST_HEAD(&f->list);
1310		list_add(&f->list, &vsi->mac_filter_list);
1311	}
1312
1313	/* increment counter and add a new flag if needed */
1314	if (is_vf) {
1315		if (!f->is_vf) {
1316			f->is_vf = true;
1317			f->counter++;
1318		}
1319	} else if (is_netdev) {
1320		if (!f->is_netdev) {
1321			f->is_netdev = true;
1322			f->counter++;
1323		}
1324	} else {
1325		f->counter++;
1326	}
1327
1328	/* changed tells sync_filters_subtask to
1329	 * push the filter down to the firmware
1330	 */
1331	if (f->changed) {
1332		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1333		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1334	}
1335
1336add_filter_out:
1337	return f;
1338}
1339
1340/**
1341 * i40e_del_filter - Remove a mac/vlan filter from the VSI
1342 * @vsi: the VSI to be searched
1343 * @macaddr: the MAC address
1344 * @vlan: the vlan
1345 * @is_vf: make sure it's a VF filter, else doesn't matter
1346 * @is_netdev: make sure it's a netdev filter, else doesn't matter
1347 **/
1348void i40e_del_filter(struct i40e_vsi *vsi,
1349		     u8 *macaddr, s16 vlan,
1350		     bool is_vf, bool is_netdev)
1351{
1352	struct i40e_mac_filter *f;
1353
1354	if (!vsi || !macaddr)
1355		return;
1356
1357	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1358	if (!f || f->counter == 0)
1359		return;
1360
1361	if (is_vf) {
1362		if (f->is_vf) {
1363			f->is_vf = false;
1364			f->counter--;
1365		}
1366	} else if (is_netdev) {
1367		if (f->is_netdev) {
1368			f->is_netdev = false;
1369			f->counter--;
1370		}
1371	} else {
1372		/* make sure we don't remove a filter in use by VF or netdev */
1373		int min_f = 0;
1374		min_f += (f->is_vf ? 1 : 0);
1375		min_f += (f->is_netdev ? 1 : 0);
1376
1377		if (f->counter > min_f)
1378			f->counter--;
1379	}
1380
1381	/* counter == 0 tells sync_filters_subtask to
1382	 * remove the filter from the firmware's list
1383	 */
1384	if (f->counter == 0) {
1385		f->changed = true;
1386		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1387		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1388	}
1389}
1390
1391/**
1392 * i40e_set_mac - NDO callback to set mac address
1393 * @netdev: network interface device structure
1394 * @p: pointer to an address structure
1395 *
1396 * Returns 0 on success, negative on failure
1397 **/
1398#ifdef I40E_FCOE
1399int i40e_set_mac(struct net_device *netdev, void *p)
1400#else
1401static int i40e_set_mac(struct net_device *netdev, void *p)
1402#endif
1403{
1404	struct i40e_netdev_priv *np = netdev_priv(netdev);
1405	struct i40e_vsi *vsi = np->vsi;
1406	struct i40e_pf *pf = vsi->back;
1407	struct i40e_hw *hw = &pf->hw;
1408	struct sockaddr *addr = p;
1409	struct i40e_mac_filter *f;
1410
1411	if (!is_valid_ether_addr(addr->sa_data))
1412		return -EADDRNOTAVAIL;
1413
1414	if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1415		netdev_info(netdev, "already using mac address %pM\n",
1416			    addr->sa_data);
1417		return 0;
1418	}
1419
1420	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1421	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1422		return -EADDRNOTAVAIL;
1423
1424	if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1425		netdev_info(netdev, "returning to hw mac address %pM\n",
1426			    hw->mac.addr);
1427	else
1428		netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1429
1430	if (vsi->type == I40E_VSI_MAIN) {
1431		i40e_status ret;
1432		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1433						I40E_AQC_WRITE_TYPE_LAA_WOL,
1434						addr->sa_data, NULL);
1435		if (ret) {
1436			netdev_info(netdev,
1437				    "Addr change for Main VSI failed: %d\n",
1438				    ret);
1439			return -EADDRNOTAVAIL;
1440		}
1441	}
1442
1443	if (ether_addr_equal(netdev->dev_addr, hw->mac.addr)) {
1444		struct i40e_aqc_remove_macvlan_element_data element;
1445
1446		memset(&element, 0, sizeof(element));
1447		ether_addr_copy(element.mac_addr, netdev->dev_addr);
1448		element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1449		i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1450	} else {
1451		i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1452				false, false);
1453	}
1454
1455	if (ether_addr_equal(addr->sa_data, hw->mac.addr)) {
1456		struct i40e_aqc_add_macvlan_element_data element;
1457
1458		memset(&element, 0, sizeof(element));
1459		ether_addr_copy(element.mac_addr, hw->mac.addr);
1460		element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
1461		i40e_aq_add_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1462	} else {
1463		f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY,
1464				    false, false);
1465		if (f)
1466			f->is_laa = true;
1467	}
1468
1469	i40e_sync_vsi_filters(vsi);
1470	ether_addr_copy(netdev->dev_addr, addr->sa_data);
1471
1472	return 0;
1473}
1474
1475/**
1476 * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1477 * @vsi: the VSI being setup
1478 * @ctxt: VSI context structure
1479 * @enabled_tc: Enabled TCs bitmap
1480 * @is_add: True if called before Add VSI
1481 *
1482 * Setup VSI queue mapping for enabled traffic classes.
1483 **/
1484#ifdef I40E_FCOE
1485void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1486			      struct i40e_vsi_context *ctxt,
1487			      u8 enabled_tc,
1488			      bool is_add)
1489#else
1490static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1491				     struct i40e_vsi_context *ctxt,
1492				     u8 enabled_tc,
1493				     bool is_add)
1494#endif
1495{
1496	struct i40e_pf *pf = vsi->back;
1497	u16 sections = 0;
1498	u8 netdev_tc = 0;
1499	u16 numtc = 0;
1500	u16 qcount;
1501	u8 offset;
1502	u16 qmap;
1503	int i;
1504	u16 num_tc_qps = 0;
1505
1506	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1507	offset = 0;
1508
1509	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1510		/* Find numtc from enabled TC bitmap */
1511		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1512			if (enabled_tc & (1 << i)) /* TC is enabled */
1513				numtc++;
1514		}
1515		if (!numtc) {
1516			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1517			numtc = 1;
1518		}
1519	} else {
1520		/* At least TC0 is enabled in case of non-DCB case */
1521		numtc = 1;
1522	}
1523
1524	vsi->tc_config.numtc = numtc;
1525	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1526	/* Number of queues per enabled TC */
1527	/* In MFP case we can have a much lower count of MSIx
1528	 * vectors available and so we need to lower the used
1529	 * q count.
1530	 */
1531	qcount = min_t(int, vsi->alloc_queue_pairs, pf->num_lan_msix);
1532	num_tc_qps = qcount / numtc;
1533	num_tc_qps = min_t(int, num_tc_qps, I40E_MAX_QUEUES_PER_TC);
1534
1535	/* Setup queue offset/count for all TCs for given VSI */
1536	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1537		/* See if the given TC is enabled for the given VSI */
1538		if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1539			int pow, num_qps;
1540
1541			switch (vsi->type) {
1542			case I40E_VSI_MAIN:
1543				qcount = min_t(int, pf->rss_size, num_tc_qps);
1544				break;
1545#ifdef I40E_FCOE
1546			case I40E_VSI_FCOE:
1547				qcount = num_tc_qps;
1548				break;
1549#endif
1550			case I40E_VSI_FDIR:
1551			case I40E_VSI_SRIOV:
1552			case I40E_VSI_VMDQ2:
1553			default:
1554				qcount = num_tc_qps;
1555				WARN_ON(i != 0);
1556				break;
1557			}
1558			vsi->tc_config.tc_info[i].qoffset = offset;
1559			vsi->tc_config.tc_info[i].qcount = qcount;
1560
1561			/* find the next higher power-of-2 of num queue pairs */
1562			num_qps = qcount;
1563			pow = 0;
1564			while (num_qps && ((1 << pow) < qcount)) {
1565				pow++;
1566				num_qps >>= 1;
1567			}
1568
1569			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1570			qmap =
1571			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1572			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1573
1574			offset += qcount;
1575		} else {
1576			/* TC is not enabled so set the offset to
1577			 * default queue and allocate one queue
1578			 * for the given TC.
1579			 */
1580			vsi->tc_config.tc_info[i].qoffset = 0;
1581			vsi->tc_config.tc_info[i].qcount = 1;
1582			vsi->tc_config.tc_info[i].netdev_tc = 0;
1583
1584			qmap = 0;
1585		}
1586		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1587	}
1588
1589	/* Set actual Tx/Rx queue pairs */
1590	vsi->num_queue_pairs = offset;
1591	if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1592		if (vsi->req_queue_pairs > 0)
1593			vsi->num_queue_pairs = vsi->req_queue_pairs;
1594		else
1595			vsi->num_queue_pairs = pf->num_lan_msix;
1596	}
1597
1598	/* Scheduler section valid can only be set for ADD VSI */
1599	if (is_add) {
1600		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1601
1602		ctxt->info.up_enable_bits = enabled_tc;
1603	}
1604	if (vsi->type == I40E_VSI_SRIOV) {
1605		ctxt->info.mapping_flags |=
1606				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1607		for (i = 0; i < vsi->num_queue_pairs; i++)
1608			ctxt->info.queue_mapping[i] =
1609					       cpu_to_le16(vsi->base_queue + i);
1610	} else {
1611		ctxt->info.mapping_flags |=
1612					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1613		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1614	}
1615	ctxt->info.valid_sections |= cpu_to_le16(sections);
1616}
1617
1618/**
1619 * i40e_set_rx_mode - NDO callback to set the netdev filters
1620 * @netdev: network interface device structure
1621 **/
1622#ifdef I40E_FCOE
1623void i40e_set_rx_mode(struct net_device *netdev)
1624#else
1625static void i40e_set_rx_mode(struct net_device *netdev)
1626#endif
1627{
1628	struct i40e_netdev_priv *np = netdev_priv(netdev);
1629	struct i40e_mac_filter *f, *ftmp;
1630	struct i40e_vsi *vsi = np->vsi;
1631	struct netdev_hw_addr *uca;
1632	struct netdev_hw_addr *mca;
1633	struct netdev_hw_addr *ha;
1634
1635	/* add addr if not already in the filter list */
1636	netdev_for_each_uc_addr(uca, netdev) {
1637		if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1638			if (i40e_is_vsi_in_vlan(vsi))
1639				i40e_put_mac_in_vlan(vsi, uca->addr,
1640						     false, true);
1641			else
1642				i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1643						false, true);
1644		}
1645	}
1646
1647	netdev_for_each_mc_addr(mca, netdev) {
1648		if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1649			if (i40e_is_vsi_in_vlan(vsi))
1650				i40e_put_mac_in_vlan(vsi, mca->addr,
1651						     false, true);
1652			else
1653				i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1654						false, true);
1655		}
1656	}
1657
1658	/* remove filter if not in netdev list */
1659	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1660		bool found = false;
1661
1662		if (!f->is_netdev)
1663			continue;
1664
1665		if (is_multicast_ether_addr(f->macaddr)) {
1666			netdev_for_each_mc_addr(mca, netdev) {
1667				if (ether_addr_equal(mca->addr, f->macaddr)) {
1668					found = true;
1669					break;
1670				}
1671			}
1672		} else {
1673			netdev_for_each_uc_addr(uca, netdev) {
1674				if (ether_addr_equal(uca->addr, f->macaddr)) {
1675					found = true;
1676					break;
1677				}
1678			}
1679
1680			for_each_dev_addr(netdev, ha) {
1681				if (ether_addr_equal(ha->addr, f->macaddr)) {
1682					found = true;
1683					break;
1684				}
1685			}
1686		}
1687		if (!found)
1688			i40e_del_filter(
1689			   vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1690	}
1691
1692	/* check for other flag changes */
1693	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1694		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1695		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1696	}
1697}
1698
1699/**
1700 * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1701 * @vsi: ptr to the VSI
1702 *
1703 * Push any outstanding VSI filter changes through the AdminQ.
1704 *
1705 * Returns 0 or error value
1706 **/
1707int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1708{
1709	struct i40e_mac_filter *f, *ftmp;
1710	bool promisc_forced_on = false;
1711	bool add_happened = false;
1712	int filter_list_len = 0;
1713	u32 changed_flags = 0;
1714	i40e_status aq_ret = 0;
1715	struct i40e_pf *pf;
1716	int num_add = 0;
1717	int num_del = 0;
1718	u16 cmd_flags;
1719
1720	/* empty array typed pointers, kcalloc later */
1721	struct i40e_aqc_add_macvlan_element_data *add_list;
1722	struct i40e_aqc_remove_macvlan_element_data *del_list;
1723
1724	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1725		usleep_range(1000, 2000);
1726	pf = vsi->back;
1727
1728	if (vsi->netdev) {
1729		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1730		vsi->current_netdev_flags = vsi->netdev->flags;
1731	}
1732
1733	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1734		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1735
1736		filter_list_len = pf->hw.aq.asq_buf_size /
1737			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1738		del_list = kcalloc(filter_list_len,
1739			    sizeof(struct i40e_aqc_remove_macvlan_element_data),
1740			    GFP_KERNEL);
1741		if (!del_list)
1742			return -ENOMEM;
1743
1744		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1745			if (!f->changed)
1746				continue;
1747
1748			if (f->counter != 0)
1749				continue;
1750			f->changed = false;
1751			cmd_flags = 0;
1752
1753			/* add to delete list */
1754			ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1755			del_list[num_del].vlan_tag =
1756				cpu_to_le16((u16)(f->vlan ==
1757					    I40E_VLAN_ANY ? 0 : f->vlan));
1758
1759			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1760			del_list[num_del].flags = cmd_flags;
1761			num_del++;
1762
1763			/* unlink from filter list */
1764			list_del(&f->list);
1765			kfree(f);
1766
1767			/* flush a full buffer */
1768			if (num_del == filter_list_len) {
1769				aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1770					    vsi->seid, del_list, num_del,
1771					    NULL);
1772				num_del = 0;
1773				memset(del_list, 0, sizeof(*del_list));
1774
1775				if (aq_ret &&
1776				    pf->hw.aq.asq_last_status !=
1777							      I40E_AQ_RC_ENOENT)
1778					dev_info(&pf->pdev->dev,
1779						 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1780						 aq_ret,
1781						 pf->hw.aq.asq_last_status);
1782			}
1783		}
1784		if (num_del) {
1785			aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1786						     del_list, num_del, NULL);
1787			num_del = 0;
1788
1789			if (aq_ret &&
1790			    pf->hw.aq.asq_last_status != I40E_AQ_RC_ENOENT)
1791				dev_info(&pf->pdev->dev,
1792					 "ignoring delete macvlan error, err %d, aq_err %d\n",
1793					 aq_ret, pf->hw.aq.asq_last_status);
1794		}
1795
1796		kfree(del_list);
1797		del_list = NULL;
1798
1799		/* do all the adds now */
1800		filter_list_len = pf->hw.aq.asq_buf_size /
1801			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1802		add_list = kcalloc(filter_list_len,
1803			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1804			       GFP_KERNEL);
1805		if (!add_list)
1806			return -ENOMEM;
1807
1808		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1809			if (!f->changed)
1810				continue;
1811
1812			if (f->counter == 0)
1813				continue;
1814			f->changed = false;
1815			add_happened = true;
1816			cmd_flags = 0;
1817
1818			/* add to add array */
1819			ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
1820			add_list[num_add].vlan_tag =
1821				cpu_to_le16(
1822				 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1823			add_list[num_add].queue_number = 0;
1824
1825			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1826			add_list[num_add].flags = cpu_to_le16(cmd_flags);
1827			num_add++;
1828
1829			/* flush a full buffer */
1830			if (num_add == filter_list_len) {
1831				aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1832							     add_list, num_add,
1833							     NULL);
1834				num_add = 0;
1835
1836				if (aq_ret)
1837					break;
1838				memset(add_list, 0, sizeof(*add_list));
1839			}
1840		}
1841		if (num_add) {
1842			aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1843						     add_list, num_add, NULL);
1844			num_add = 0;
1845		}
1846		kfree(add_list);
1847		add_list = NULL;
1848
1849		if (add_happened && aq_ret &&
1850		    pf->hw.aq.asq_last_status != I40E_AQ_RC_EINVAL) {
1851			dev_info(&pf->pdev->dev,
1852				 "add filter failed, err %d, aq_err %d\n",
1853				 aq_ret, pf->hw.aq.asq_last_status);
1854			if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1855			    !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1856				      &vsi->state)) {
1857				promisc_forced_on = true;
1858				set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1859					&vsi->state);
1860				dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1861			}
1862		}
1863	}
1864
1865	/* check for changes in promiscuous modes */
1866	if (changed_flags & IFF_ALLMULTI) {
1867		bool cur_multipromisc;
1868		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1869		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1870							       vsi->seid,
1871							       cur_multipromisc,
1872							       NULL);
1873		if (aq_ret)
1874			dev_info(&pf->pdev->dev,
1875				 "set multi promisc failed, err %d, aq_err %d\n",
1876				 aq_ret, pf->hw.aq.asq_last_status);
1877	}
1878	if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1879		bool cur_promisc;
1880		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1881			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1882					&vsi->state));
1883		aq_ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1884							     vsi->seid,
1885							     cur_promisc, NULL);
1886		if (aq_ret)
1887			dev_info(&pf->pdev->dev,
1888				 "set uni promisc failed, err %d, aq_err %d\n",
1889				 aq_ret, pf->hw.aq.asq_last_status);
1890		aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
1891						   vsi->seid,
1892						   cur_promisc, NULL);
1893		if (aq_ret)
1894			dev_info(&pf->pdev->dev,
1895				 "set brdcast promisc failed, err %d, aq_err %d\n",
1896				 aq_ret, pf->hw.aq.asq_last_status);
1897	}
1898
1899	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1900	return 0;
1901}
1902
1903/**
1904 * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1905 * @pf: board private structure
1906 **/
1907static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1908{
1909	int v;
1910
1911	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1912		return;
1913	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1914
1915	for (v = 0; v < pf->num_alloc_vsi; v++) {
1916		if (pf->vsi[v] &&
1917		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1918			i40e_sync_vsi_filters(pf->vsi[v]);
1919	}
1920}
1921
1922/**
1923 * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1924 * @netdev: network interface device structure
1925 * @new_mtu: new value for maximum frame size
1926 *
1927 * Returns 0 on success, negative on failure
1928 **/
1929static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1930{
1931	struct i40e_netdev_priv *np = netdev_priv(netdev);
1932	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
1933	struct i40e_vsi *vsi = np->vsi;
1934
1935	/* MTU < 68 is an error and causes problems on some kernels */
1936	if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1937		return -EINVAL;
1938
1939	netdev_info(netdev, "changing MTU from %d to %d\n",
1940		    netdev->mtu, new_mtu);
1941	netdev->mtu = new_mtu;
1942	if (netif_running(netdev))
1943		i40e_vsi_reinit_locked(vsi);
1944
1945	return 0;
1946}
1947
1948/**
1949 * i40e_ioctl - Access the hwtstamp interface
1950 * @netdev: network interface device structure
1951 * @ifr: interface request data
1952 * @cmd: ioctl command
1953 **/
1954int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
1955{
1956	struct i40e_netdev_priv *np = netdev_priv(netdev);
1957	struct i40e_pf *pf = np->vsi->back;
1958
1959	switch (cmd) {
1960	case SIOCGHWTSTAMP:
1961		return i40e_ptp_get_ts_config(pf, ifr);
1962	case SIOCSHWTSTAMP:
1963		return i40e_ptp_set_ts_config(pf, ifr);
1964	default:
1965		return -EOPNOTSUPP;
1966	}
1967}
1968
1969/**
1970 * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1971 * @vsi: the vsi being adjusted
1972 **/
1973void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1974{
1975	struct i40e_vsi_context ctxt;
1976	i40e_status ret;
1977
1978	if ((vsi->info.valid_sections &
1979	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1980	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1981		return;  /* already enabled */
1982
1983	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1984	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1985				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1986
1987	ctxt.seid = vsi->seid;
1988	ctxt.info = vsi->info;
1989	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1990	if (ret) {
1991		dev_info(&vsi->back->pdev->dev,
1992			 "%s: update vsi failed, aq_err=%d\n",
1993			 __func__, vsi->back->hw.aq.asq_last_status);
1994	}
1995}
1996
1997/**
1998 * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1999 * @vsi: the vsi being adjusted
2000 **/
2001void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2002{
2003	struct i40e_vsi_context ctxt;
2004	i40e_status ret;
2005
2006	if ((vsi->info.valid_sections &
2007	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2008	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2009	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
2010		return;  /* already disabled */
2011
2012	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2013	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2014				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2015
2016	ctxt.seid = vsi->seid;
2017	ctxt.info = vsi->info;
2018	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2019	if (ret) {
2020		dev_info(&vsi->back->pdev->dev,
2021			 "%s: update vsi failed, aq_err=%d\n",
2022			 __func__, vsi->back->hw.aq.asq_last_status);
2023	}
2024}
2025
2026/**
2027 * i40e_vlan_rx_register - Setup or shutdown vlan offload
2028 * @netdev: network interface to be adjusted
2029 * @features: netdev features to test if VLAN offload is enabled or not
2030 **/
2031static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2032{
2033	struct i40e_netdev_priv *np = netdev_priv(netdev);
2034	struct i40e_vsi *vsi = np->vsi;
2035
2036	if (features & NETIF_F_HW_VLAN_CTAG_RX)
2037		i40e_vlan_stripping_enable(vsi);
2038	else
2039		i40e_vlan_stripping_disable(vsi);
2040}
2041
2042/**
2043 * i40e_vsi_add_vlan - Add vsi membership for given vlan
2044 * @vsi: the vsi being configured
2045 * @vid: vlan id to be added (0 = untagged only , -1 = any)
2046 **/
2047int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2048{
2049	struct i40e_mac_filter *f, *add_f;
2050	bool is_netdev, is_vf;
2051
2052	is_vf = (vsi->type == I40E_VSI_SRIOV);
2053	is_netdev = !!(vsi->netdev);
2054
2055	if (is_netdev) {
2056		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2057					is_vf, is_netdev);
2058		if (!add_f) {
2059			dev_info(&vsi->back->pdev->dev,
2060				 "Could not add vlan filter %d for %pM\n",
2061				 vid, vsi->netdev->dev_addr);
2062			return -ENOMEM;
2063		}
2064	}
2065
2066	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2067		add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2068		if (!add_f) {
2069			dev_info(&vsi->back->pdev->dev,
2070				 "Could not add vlan filter %d for %pM\n",
2071				 vid, f->macaddr);
2072			return -ENOMEM;
2073		}
2074	}
2075
2076	/* Now if we add a vlan tag, make sure to check if it is the first
2077	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2078	 * with 0, so we now accept untagged and specified tagged traffic
2079	 * (and not any taged and untagged)
2080	 */
2081	if (vid > 0) {
2082		if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2083						  I40E_VLAN_ANY,
2084						  is_vf, is_netdev)) {
2085			i40e_del_filter(vsi, vsi->netdev->dev_addr,
2086					I40E_VLAN_ANY, is_vf, is_netdev);
2087			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2088						is_vf, is_netdev);
2089			if (!add_f) {
2090				dev_info(&vsi->back->pdev->dev,
2091					 "Could not add filter 0 for %pM\n",
2092					 vsi->netdev->dev_addr);
2093				return -ENOMEM;
2094			}
2095		}
2096	}
2097
2098	/* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2099	if (vid > 0 && !vsi->info.pvid) {
2100		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2101			if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2102					     is_vf, is_netdev)) {
2103				i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2104						is_vf, is_netdev);
2105				add_f = i40e_add_filter(vsi, f->macaddr,
2106							0, is_vf, is_netdev);
2107				if (!add_f) {
2108					dev_info(&vsi->back->pdev->dev,
2109						 "Could not add filter 0 for %pM\n",
2110						 f->macaddr);
2111					return -ENOMEM;
2112				}
2113			}
2114		}
2115	}
2116
2117	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2118	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2119		return 0;
2120
2121	return i40e_sync_vsi_filters(vsi);
2122}
2123
2124/**
2125 * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2126 * @vsi: the vsi being configured
2127 * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2128 *
2129 * Return: 0 on success or negative otherwise
2130 **/
2131int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2132{
2133	struct net_device *netdev = vsi->netdev;
2134	struct i40e_mac_filter *f, *add_f;
2135	bool is_vf, is_netdev;
2136	int filter_count = 0;
2137
2138	is_vf = (vsi->type == I40E_VSI_SRIOV);
2139	is_netdev = !!(netdev);
2140
2141	if (is_netdev)
2142		i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2143
2144	list_for_each_entry(f, &vsi->mac_filter_list, list)
2145		i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2146
2147	/* go through all the filters for this VSI and if there is only
2148	 * vid == 0 it means there are no other filters, so vid 0 must
2149	 * be replaced with -1. This signifies that we should from now
2150	 * on accept any traffic (with any tag present, or untagged)
2151	 */
2152	list_for_each_entry(f, &vsi->mac_filter_list, list) {
2153		if (is_netdev) {
2154			if (f->vlan &&
2155			    ether_addr_equal(netdev->dev_addr, f->macaddr))
2156				filter_count++;
2157		}
2158
2159		if (f->vlan)
2160			filter_count++;
2161	}
2162
2163	if (!filter_count && is_netdev) {
2164		i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2165		f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2166				    is_vf, is_netdev);
2167		if (!f) {
2168			dev_info(&vsi->back->pdev->dev,
2169				 "Could not add filter %d for %pM\n",
2170				 I40E_VLAN_ANY, netdev->dev_addr);
2171			return -ENOMEM;
2172		}
2173	}
2174
2175	if (!filter_count) {
2176		list_for_each_entry(f, &vsi->mac_filter_list, list) {
2177			i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2178			add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2179					    is_vf, is_netdev);
2180			if (!add_f) {
2181				dev_info(&vsi->back->pdev->dev,
2182					 "Could not add filter %d for %pM\n",
2183					 I40E_VLAN_ANY, f->macaddr);
2184				return -ENOMEM;
2185			}
2186		}
2187	}
2188
2189	if (test_bit(__I40E_DOWN, &vsi->back->state) ||
2190	    test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
2191		return 0;
2192
2193	return i40e_sync_vsi_filters(vsi);
2194}
2195
2196/**
2197 * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2198 * @netdev: network interface to be adjusted
2199 * @vid: vlan id to be added
2200 *
2201 * net_device_ops implementation for adding vlan ids
2202 **/
2203#ifdef I40E_FCOE
2204int i40e_vlan_rx_add_vid(struct net_device *netdev,
2205			 __always_unused __be16 proto, u16 vid)
2206#else
2207static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2208				__always_unused __be16 proto, u16 vid)
2209#endif
2210{
2211	struct i40e_netdev_priv *np = netdev_priv(netdev);
2212	struct i40e_vsi *vsi = np->vsi;
2213	int ret = 0;
2214
2215	if (vid > 4095)
2216		return -EINVAL;
2217
2218	netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
2219
2220	/* If the network stack called us with vid = 0 then
2221	 * it is asking to receive priority tagged packets with
2222	 * vlan id 0.  Our HW receives them by default when configured
2223	 * to receive untagged packets so there is no need to add an
2224	 * extra filter for vlan 0 tagged packets.
2225	 */
2226	if (vid)
2227		ret = i40e_vsi_add_vlan(vsi, vid);
2228
2229	if (!ret && (vid < VLAN_N_VID))
2230		set_bit(vid, vsi->active_vlans);
2231
2232	return ret;
2233}
2234
2235/**
2236 * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2237 * @netdev: network interface to be adjusted
2238 * @vid: vlan id to be removed
2239 *
2240 * net_device_ops implementation for removing vlan ids
2241 **/
2242#ifdef I40E_FCOE
2243int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2244			  __always_unused __be16 proto, u16 vid)
2245#else
2246static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2247				 __always_unused __be16 proto, u16 vid)
2248#endif
2249{
2250	struct i40e_netdev_priv *np = netdev_priv(netdev);
2251	struct i40e_vsi *vsi = np->vsi;
2252
2253	netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
2254
2255	/* return code is ignored as there is nothing a user
2256	 * can do about failure to remove and a log message was
2257	 * already printed from the other function
2258	 */
2259	i40e_vsi_kill_vlan(vsi, vid);
2260
2261	clear_bit(vid, vsi->active_vlans);
2262
2263	return 0;
2264}
2265
2266/**
2267 * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2268 * @vsi: the vsi being brought back up
2269 **/
2270static void i40e_restore_vlan(struct i40e_vsi *vsi)
2271{
2272	u16 vid;
2273
2274	if (!vsi->netdev)
2275		return;
2276
2277	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2278
2279	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2280		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2281				     vid);
2282}
2283
2284/**
2285 * i40e_vsi_add_pvid - Add pvid for the VSI
2286 * @vsi: the vsi being adjusted
2287 * @vid: the vlan id to set as a PVID
2288 **/
2289int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2290{
2291	struct i40e_vsi_context ctxt;
2292	i40e_status aq_ret;
2293
2294	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2295	vsi->info.pvid = cpu_to_le16(vid);
2296	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2297				    I40E_AQ_VSI_PVLAN_INSERT_PVID |
2298				    I40E_AQ_VSI_PVLAN_EMOD_STR;
2299
2300	ctxt.seid = vsi->seid;
2301	ctxt.info = vsi->info;
2302	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2303	if (aq_ret) {
2304		dev_info(&vsi->back->pdev->dev,
2305			 "%s: update vsi failed, aq_err=%d\n",
2306			 __func__, vsi->back->hw.aq.asq_last_status);
2307		return -ENOENT;
2308	}
2309
2310	return 0;
2311}
2312
2313/**
2314 * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2315 * @vsi: the vsi being adjusted
2316 *
2317 * Just use the vlan_rx_register() service to put it back to normal
2318 **/
2319void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2320{
2321	i40e_vlan_stripping_disable(vsi);
2322
2323	vsi->info.pvid = 0;
2324}
2325
2326/**
2327 * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2328 * @vsi: ptr to the VSI
2329 *
2330 * If this function returns with an error, then it's possible one or
2331 * more of the rings is populated (while the rest are not).  It is the
2332 * callers duty to clean those orphaned rings.
2333 *
2334 * Return 0 on success, negative on failure
2335 **/
2336static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2337{
2338	int i, err = 0;
2339
2340	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2341		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2342
2343	return err;
2344}
2345
2346/**
2347 * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2348 * @vsi: ptr to the VSI
2349 *
2350 * Free VSI's transmit software resources
2351 **/
2352static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2353{
2354	int i;
2355
2356	if (!vsi->tx_rings)
2357		return;
2358
2359	for (i = 0; i < vsi->num_queue_pairs; i++)
2360		if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2361			i40e_free_tx_resources(vsi->tx_rings[i]);
2362}
2363
2364/**
2365 * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2366 * @vsi: ptr to the VSI
2367 *
2368 * If this function returns with an error, then it's possible one or
2369 * more of the rings is populated (while the rest are not).  It is the
2370 * callers duty to clean those orphaned rings.
2371 *
2372 * Return 0 on success, negative on failure
2373 **/
2374static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2375{
2376	int i, err = 0;
2377
2378	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2379		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2380#ifdef I40E_FCOE
2381	i40e_fcoe_setup_ddp_resources(vsi);
2382#endif
2383	return err;
2384}
2385
2386/**
2387 * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2388 * @vsi: ptr to the VSI
2389 *
2390 * Free all receive software resources
2391 **/
2392static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2393{
2394	int i;
2395
2396	if (!vsi->rx_rings)
2397		return;
2398
2399	for (i = 0; i < vsi->num_queue_pairs; i++)
2400		if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2401			i40e_free_rx_resources(vsi->rx_rings[i]);
2402#ifdef I40E_FCOE
2403	i40e_fcoe_free_ddp_resources(vsi);
2404#endif
2405}
2406
2407/**
2408 * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2409 * @ring: The Tx ring to configure
2410 *
2411 * This enables/disables XPS for a given Tx descriptor ring
2412 * based on the TCs enabled for the VSI that ring belongs to.
2413 **/
2414static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2415{
2416	struct i40e_vsi *vsi = ring->vsi;
2417	cpumask_var_t mask;
2418
2419	if (!ring->q_vector || !ring->netdev)
2420		return;
2421
2422	/* Single TC mode enable XPS */
2423	if (vsi->tc_config.numtc <= 1) {
2424		if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2425			netif_set_xps_queue(ring->netdev,
2426					    &ring->q_vector->affinity_mask,
2427					    ring->queue_index);
2428	} else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2429		/* Disable XPS to allow selection based on TC */
2430		bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2431		netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2432		free_cpumask_var(mask);
2433	}
2434}
2435
2436/**
2437 * i40e_configure_tx_ring - Configure a transmit ring context and rest
2438 * @ring: The Tx ring to configure
2439 *
2440 * Configure the Tx descriptor ring in the HMC context.
2441 **/
2442static int i40e_configure_tx_ring(struct i40e_ring *ring)
2443{
2444	struct i40e_vsi *vsi = ring->vsi;
2445	u16 pf_q = vsi->base_queue + ring->queue_index;
2446	struct i40e_hw *hw = &vsi->back->hw;
2447	struct i40e_hmc_obj_txq tx_ctx;
2448	i40e_status err = 0;
2449	u32 qtx_ctl = 0;
2450
2451	/* some ATR related tx ring init */
2452	if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2453		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2454		ring->atr_count = 0;
2455	} else {
2456		ring->atr_sample_rate = 0;
2457	}
2458
2459	/* configure XPS */
2460	i40e_config_xps_tx_ring(ring);
2461
2462	/* clear the context structure first */
2463	memset(&tx_ctx, 0, sizeof(tx_ctx));
2464
2465	tx_ctx.new_context = 1;
2466	tx_ctx.base = (ring->dma / 128);
2467	tx_ctx.qlen = ring->count;
2468	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2469					       I40E_FLAG_FD_ATR_ENABLED));
2470#ifdef I40E_FCOE
2471	tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2472#endif
2473	tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2474	/* FDIR VSI tx ring can still use RS bit and writebacks */
2475	if (vsi->type != I40E_VSI_FDIR)
2476		tx_ctx.head_wb_ena = 1;
2477	tx_ctx.head_wb_addr = ring->dma +
2478			      (ring->count * sizeof(struct i40e_tx_desc));
2479
2480	/* As part of VSI creation/update, FW allocates certain
2481	 * Tx arbitration queue sets for each TC enabled for
2482	 * the VSI. The FW returns the handles to these queue
2483	 * sets as part of the response buffer to Add VSI,
2484	 * Update VSI, etc. AQ commands. It is expected that
2485	 * these queue set handles be associated with the Tx
2486	 * queues by the driver as part of the TX queue context
2487	 * initialization. This has to be done regardless of
2488	 * DCB as by default everything is mapped to TC0.
2489	 */
2490	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2491	tx_ctx.rdylist_act = 0;
2492
2493	/* clear the context in the HMC */
2494	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2495	if (err) {
2496		dev_info(&vsi->back->pdev->dev,
2497			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2498			 ring->queue_index, pf_q, err);
2499		return -ENOMEM;
2500	}
2501
2502	/* set the context in the HMC */
2503	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2504	if (err) {
2505		dev_info(&vsi->back->pdev->dev,
2506			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2507			 ring->queue_index, pf_q, err);
2508		return -ENOMEM;
2509	}
2510
2511	/* Now associate this queue with this PCI function */
2512	if (vsi->type == I40E_VSI_VMDQ2) {
2513		qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2514		qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2515			   I40E_QTX_CTL_VFVM_INDX_MASK;
2516	} else {
2517		qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2518	}
2519
2520	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2521		    I40E_QTX_CTL_PF_INDX_MASK);
2522	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2523	i40e_flush(hw);
2524
2525	clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2526
2527	/* cache tail off for easier writes later */
2528	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2529
2530	return 0;
2531}
2532
2533/**
2534 * i40e_configure_rx_ring - Configure a receive ring context
2535 * @ring: The Rx ring to configure
2536 *
2537 * Configure the Rx descriptor ring in the HMC context.
2538 **/
2539static int i40e_configure_rx_ring(struct i40e_ring *ring)
2540{
2541	struct i40e_vsi *vsi = ring->vsi;
2542	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2543	u16 pf_q = vsi->base_queue + ring->queue_index;
2544	struct i40e_hw *hw = &vsi->back->hw;
2545	struct i40e_hmc_obj_rxq rx_ctx;
2546	i40e_status err = 0;
2547
2548	ring->state = 0;
2549
2550	/* clear the context structure first */
2551	memset(&rx_ctx, 0, sizeof(rx_ctx));
2552
2553	ring->rx_buf_len = vsi->rx_buf_len;
2554	ring->rx_hdr_len = vsi->rx_hdr_len;
2555
2556	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2557	rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2558
2559	rx_ctx.base = (ring->dma / 128);
2560	rx_ctx.qlen = ring->count;
2561
2562	if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2563		set_ring_16byte_desc_enabled(ring);
2564		rx_ctx.dsize = 0;
2565	} else {
2566		rx_ctx.dsize = 1;
2567	}
2568
2569	rx_ctx.dtype = vsi->dtype;
2570	if (vsi->dtype) {
2571		set_ring_ps_enabled(ring);
2572		rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2573				  I40E_RX_SPLIT_IP      |
2574				  I40E_RX_SPLIT_TCP_UDP |
2575				  I40E_RX_SPLIT_SCTP;
2576	} else {
2577		rx_ctx.hsplit_0 = 0;
2578	}
2579
2580	rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2581				  (chain_len * ring->rx_buf_len));
2582	if (hw->revision_id == 0)
2583		rx_ctx.lrxqthresh = 0;
2584	else
2585		rx_ctx.lrxqthresh = 2;
2586	rx_ctx.crcstrip = 1;
2587	rx_ctx.l2tsel = 1;
2588	rx_ctx.showiv = 1;
2589#ifdef I40E_FCOE
2590	rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2591#endif
2592	/* set the prefena field to 1 because the manual says to */
2593	rx_ctx.prefena = 1;
2594
2595	/* clear the context in the HMC */
2596	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2597	if (err) {
2598		dev_info(&vsi->back->pdev->dev,
2599			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2600			 ring->queue_index, pf_q, err);
2601		return -ENOMEM;
2602	}
2603
2604	/* set the context in the HMC */
2605	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2606	if (err) {
2607		dev_info(&vsi->back->pdev->dev,
2608			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2609			 ring->queue_index, pf_q, err);
2610		return -ENOMEM;
2611	}
2612
2613	/* cache tail for quicker writes, and clear the reg before use */
2614	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2615	writel(0, ring->tail);
2616
2617	if (ring_is_ps_enabled(ring)) {
2618		i40e_alloc_rx_headers(ring);
2619		i40e_alloc_rx_buffers_ps(ring, I40E_DESC_UNUSED(ring));
2620	} else {
2621		i40e_alloc_rx_buffers_1buf(ring, I40E_DESC_UNUSED(ring));
2622	}
2623
2624	return 0;
2625}
2626
2627/**
2628 * i40e_vsi_configure_tx - Configure the VSI for Tx
2629 * @vsi: VSI structure describing this set of rings and resources
2630 *
2631 * Configure the Tx VSI for operation.
2632 **/
2633static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2634{
2635	int err = 0;
2636	u16 i;
2637
2638	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2639		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2640
2641	return err;
2642}
2643
2644/**
2645 * i40e_vsi_configure_rx - Configure the VSI for Rx
2646 * @vsi: the VSI being configured
2647 *
2648 * Configure the Rx VSI for operation.
2649 **/
2650static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2651{
2652	int err = 0;
2653	u16 i;
2654
2655	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2656		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2657			       + ETH_FCS_LEN + VLAN_HLEN;
2658	else
2659		vsi->max_frame = I40E_RXBUFFER_2048;
2660
2661	/* figure out correct receive buffer length */
2662	switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2663				    I40E_FLAG_RX_PS_ENABLED)) {
2664	case I40E_FLAG_RX_1BUF_ENABLED:
2665		vsi->rx_hdr_len = 0;
2666		vsi->rx_buf_len = vsi->max_frame;
2667		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2668		break;
2669	case I40E_FLAG_RX_PS_ENABLED:
2670		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2671		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2672		vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2673		break;
2674	default:
2675		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2676		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2677		vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2678		break;
2679	}
2680
2681#ifdef I40E_FCOE
2682	/* setup rx buffer for FCoE */
2683	if ((vsi->type == I40E_VSI_FCOE) &&
2684	    (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
2685		vsi->rx_hdr_len = 0;
2686		vsi->rx_buf_len = I40E_RXBUFFER_3072;
2687		vsi->max_frame = I40E_RXBUFFER_3072;
2688		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2689	}
2690
2691#endif /* I40E_FCOE */
2692	/* round up for the chip's needs */
2693	vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2694				(1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2695	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2696				(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2697
2698	/* set up individual rings */
2699	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2700		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2701
2702	return err;
2703}
2704
2705/**
2706 * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2707 * @vsi: ptr to the VSI
2708 **/
2709static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2710{
2711	struct i40e_ring *tx_ring, *rx_ring;
2712	u16 qoffset, qcount;
2713	int i, n;
2714
2715	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
2716		/* Reset the TC information */
2717		for (i = 0; i < vsi->num_queue_pairs; i++) {
2718			rx_ring = vsi->rx_rings[i];
2719			tx_ring = vsi->tx_rings[i];
2720			rx_ring->dcb_tc = 0;
2721			tx_ring->dcb_tc = 0;
2722		}
2723	}
2724
2725	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2726		if (!(vsi->tc_config.enabled_tc & (1 << n)))
2727			continue;
2728
2729		qoffset = vsi->tc_config.tc_info[n].qoffset;
2730		qcount = vsi->tc_config.tc_info[n].qcount;
2731		for (i = qoffset; i < (qoffset + qcount); i++) {
2732			rx_ring = vsi->rx_rings[i];
2733			tx_ring = vsi->tx_rings[i];
2734			rx_ring->dcb_tc = n;
2735			tx_ring->dcb_tc = n;
2736		}
2737	}
2738}
2739
2740/**
2741 * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2742 * @vsi: ptr to the VSI
2743 **/
2744static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2745{
2746	if (vsi->netdev)
2747		i40e_set_rx_mode(vsi->netdev);
2748}
2749
2750/**
2751 * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
2752 * @vsi: Pointer to the targeted VSI
2753 *
2754 * This function replays the hlist on the hw where all the SB Flow Director
2755 * filters were saved.
2756 **/
2757static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
2758{
2759	struct i40e_fdir_filter *filter;
2760	struct i40e_pf *pf = vsi->back;
2761	struct hlist_node *node;
2762
2763	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
2764		return;
2765
2766	hlist_for_each_entry_safe(filter, node,
2767				  &pf->fdir_filter_list, fdir_node) {
2768		i40e_add_del_fdir(vsi, filter, true);
2769	}
2770}
2771
2772/**
2773 * i40e_vsi_configure - Set up the VSI for action
2774 * @vsi: the VSI being configured
2775 **/
2776static int i40e_vsi_configure(struct i40e_vsi *vsi)
2777{
2778	int err;
2779
2780	i40e_set_vsi_rx_mode(vsi);
2781	i40e_restore_vlan(vsi);
2782	i40e_vsi_config_dcb_rings(vsi);
2783	err = i40e_vsi_configure_tx(vsi);
2784	if (!err)
2785		err = i40e_vsi_configure_rx(vsi);
2786
2787	return err;
2788}
2789
2790/**
2791 * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2792 * @vsi: the VSI being configured
2793 **/
2794static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2795{
2796	struct i40e_pf *pf = vsi->back;
2797	struct i40e_q_vector *q_vector;
2798	struct i40e_hw *hw = &pf->hw;
2799	u16 vector;
2800	int i, q;
2801	u32 val;
2802	u32 qp;
2803
2804	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
2805	 * and PFINT_LNKLSTn registers, e.g.:
2806	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
2807	 */
2808	qp = vsi->base_queue;
2809	vector = vsi->base_vector;
2810	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
2811		q_vector = vsi->q_vectors[i];
2812		q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2813		q_vector->rx.latency_range = I40E_LOW_LATENCY;
2814		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2815		     q_vector->rx.itr);
2816		q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2817		q_vector->tx.latency_range = I40E_LOW_LATENCY;
2818		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2819		     q_vector->tx.itr);
2820
2821		/* Linked list for the queuepairs assigned to this vector */
2822		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2823		for (q = 0; q < q_vector->num_ringpairs; q++) {
2824			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2825			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
2826			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2827			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2828			      (I40E_QUEUE_TYPE_TX
2829				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2830
2831			wr32(hw, I40E_QINT_RQCTL(qp), val);
2832
2833			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2834			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
2835			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2836			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2837			      (I40E_QUEUE_TYPE_RX
2838				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2839
2840			/* Terminate the linked list */
2841			if (q == (q_vector->num_ringpairs - 1))
2842				val |= (I40E_QUEUE_END_OF_LIST
2843					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2844
2845			wr32(hw, I40E_QINT_TQCTL(qp), val);
2846			qp++;
2847		}
2848	}
2849
2850	i40e_flush(hw);
2851}
2852
2853/**
2854 * i40e_enable_misc_int_causes - enable the non-queue interrupts
2855 * @hw: ptr to the hardware info
2856 **/
2857static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
2858{
2859	struct i40e_hw *hw = &pf->hw;
2860	u32 val;
2861
2862	/* clear things first */
2863	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
2864	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
2865
2866	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
2867	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
2868	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
2869	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2870	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
2871	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
2872	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
2873	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2874
2875	if (pf->flags & I40E_FLAG_PTP)
2876		val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
2877
2878	wr32(hw, I40E_PFINT_ICR0_ENA, val);
2879
2880	/* SW_ITR_IDX = 0, but don't change INTENA */
2881	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
2882					I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
2883
2884	/* OTHER_ITR_IDX = 0 */
2885	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2886}
2887
2888/**
2889 * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2890 * @vsi: the VSI being configured
2891 **/
2892static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2893{
2894	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
2895	struct i40e_pf *pf = vsi->back;
2896	struct i40e_hw *hw = &pf->hw;
2897	u32 val;
2898
2899	/* set the ITR configuration */
2900	q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2901	q_vector->rx.latency_range = I40E_LOW_LATENCY;
2902	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2903	q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2904	q_vector->tx.latency_range = I40E_LOW_LATENCY;
2905	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2906
2907	i40e_enable_misc_int_causes(pf);
2908
2909	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2910	wr32(hw, I40E_PFINT_LNKLST0, 0);
2911
2912	/* Associate the queue pair to the vector and enable the queue int */
2913	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
2914	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2915	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2916
2917	wr32(hw, I40E_QINT_RQCTL(0), val);
2918
2919	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
2920	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2921	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2922
2923	wr32(hw, I40E_QINT_TQCTL(0), val);
2924	i40e_flush(hw);
2925}
2926
2927/**
2928 * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
2929 * @pf: board private structure
2930 **/
2931void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
2932{
2933	struct i40e_hw *hw = &pf->hw;
2934
2935	wr32(hw, I40E_PFINT_DYN_CTL0,
2936	     I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2937	i40e_flush(hw);
2938}
2939
2940/**
2941 * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2942 * @pf: board private structure
2943 **/
2944void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2945{
2946	struct i40e_hw *hw = &pf->hw;
2947	u32 val;
2948
2949	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
2950	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2951	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2952
2953	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2954	i40e_flush(hw);
2955}
2956
2957/**
2958 * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2959 * @vsi: pointer to a vsi
2960 * @vector: enable a particular Hw Interrupt vector
2961 **/
2962void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2963{
2964	struct i40e_pf *pf = vsi->back;
2965	struct i40e_hw *hw = &pf->hw;
2966	u32 val;
2967
2968	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2969	      I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2970	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2971	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2972	/* skip the flush */
2973}
2974
2975/**
2976 * i40e_irq_dynamic_disable - Disable default interrupt generation settings
2977 * @vsi: pointer to a vsi
2978 * @vector: disable a particular Hw Interrupt vector
2979 **/
2980void i40e_irq_dynamic_disable(struct i40e_vsi *vsi, int vector)
2981{
2982	struct i40e_pf *pf = vsi->back;
2983	struct i40e_hw *hw = &pf->hw;
2984	u32 val;
2985
2986	val = I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT;
2987	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2988	i40e_flush(hw);
2989}
2990
2991/**
2992 * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2993 * @irq: interrupt number
2994 * @data: pointer to a q_vector
2995 **/
2996static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2997{
2998	struct i40e_q_vector *q_vector = data;
2999
3000	if (!q_vector->tx.ring && !q_vector->rx.ring)
3001		return IRQ_HANDLED;
3002
3003	napi_schedule(&q_vector->napi);
3004
3005	return IRQ_HANDLED;
3006}
3007
3008/**
3009 * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3010 * @vsi: the VSI being configured
3011 * @basename: name for the vector
3012 *
3013 * Allocates MSI-X vectors and requests interrupts from the kernel.
3014 **/
3015static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3016{
3017	int q_vectors = vsi->num_q_vectors;
3018	struct i40e_pf *pf = vsi->back;
3019	int base = vsi->base_vector;
3020	int rx_int_idx = 0;
3021	int tx_int_idx = 0;
3022	int vector, err;
3023
3024	for (vector = 0; vector < q_vectors; vector++) {
3025		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3026
3027		if (q_vector->tx.ring && q_vector->rx.ring) {
3028			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3029				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3030			tx_int_idx++;
3031		} else if (q_vector->rx.ring) {
3032			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3033				 "%s-%s-%d", basename, "rx", rx_int_idx++);
3034		} else if (q_vector->tx.ring) {
3035			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3036				 "%s-%s-%d", basename, "tx", tx_int_idx++);
3037		} else {
3038			/* skip this unused q_vector */
3039			continue;
3040		}
3041		err = request_irq(pf->msix_entries[base + vector].vector,
3042				  vsi->irq_handler,
3043				  0,
3044				  q_vector->name,
3045				  q_vector);
3046		if (err) {
3047			dev_info(&pf->pdev->dev,
3048				 "%s: request_irq failed, error: %d\n",
3049				 __func__, err);
3050			goto free_queue_irqs;
3051		}
3052		/* assign the mask for this irq */
3053		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3054				      &q_vector->affinity_mask);
3055	}
3056
3057	vsi->irqs_ready = true;
3058	return 0;
3059
3060free_queue_irqs:
3061	while (vector) {
3062		vector--;
3063		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3064				      NULL);
3065		free_irq(pf->msix_entries[base + vector].vector,
3066			 &(vsi->q_vectors[vector]));
3067	}
3068	return err;
3069}
3070
3071/**
3072 * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3073 * @vsi: the VSI being un-configured
3074 **/
3075static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3076{
3077	struct i40e_pf *pf = vsi->back;
3078	struct i40e_hw *hw = &pf->hw;
3079	int base = vsi->base_vector;
3080	int i;
3081
3082	for (i = 0; i < vsi->num_queue_pairs; i++) {
3083		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3084		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3085	}
3086
3087	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3088		for (i = vsi->base_vector;
3089		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3090			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3091
3092		i40e_flush(hw);
3093		for (i = 0; i < vsi->num_q_vectors; i++)
3094			synchronize_irq(pf->msix_entries[i + base].vector);
3095	} else {
3096		/* Legacy and MSI mode - this stops all interrupt handling */
3097		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3098		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3099		i40e_flush(hw);
3100		synchronize_irq(pf->pdev->irq);
3101	}
3102}
3103
3104/**
3105 * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3106 * @vsi: the VSI being configured
3107 **/
3108static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3109{
3110	struct i40e_pf *pf = vsi->back;
3111	int i;
3112
3113	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3114		for (i = vsi->base_vector;
3115		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
3116			i40e_irq_dynamic_enable(vsi, i);
3117	} else {
3118		i40e_irq_dynamic_enable_icr0(pf);
3119	}
3120
3121	i40e_flush(&pf->hw);
3122	return 0;
3123}
3124
3125/**
3126 * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3127 * @pf: board private structure
3128 **/
3129static void i40e_stop_misc_vector(struct i40e_pf *pf)
3130{
3131	/* Disable ICR 0 */
3132	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3133	i40e_flush(&pf->hw);
3134}
3135
3136/**
3137 * i40e_intr - MSI/Legacy and non-queue interrupt handler
3138 * @irq: interrupt number
3139 * @data: pointer to a q_vector
3140 *
3141 * This is the handler used for all MSI/Legacy interrupts, and deals
3142 * with both queue and non-queue interrupts.  This is also used in
3143 * MSIX mode to handle the non-queue interrupts.
3144 **/
3145static irqreturn_t i40e_intr(int irq, void *data)
3146{
3147	struct i40e_pf *pf = (struct i40e_pf *)data;
3148	struct i40e_hw *hw = &pf->hw;
3149	irqreturn_t ret = IRQ_NONE;
3150	u32 icr0, icr0_remaining;
3151	u32 val, ena_mask;
3152
3153	icr0 = rd32(hw, I40E_PFINT_ICR0);
3154	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3155
3156	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
3157	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3158		goto enable_intr;
3159
3160	/* if interrupt but no bits showing, must be SWINT */
3161	if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3162	    (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3163		pf->sw_int_count++;
3164
3165	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3166	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3167
3168		/* temporarily disable queue cause for NAPI processing */
3169		u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
3170		qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
3171		wr32(hw, I40E_QINT_RQCTL(0), qval);
3172
3173		qval = rd32(hw, I40E_QINT_TQCTL(0));
3174		qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
3175		wr32(hw, I40E_QINT_TQCTL(0), qval);
3176
3177		if (!test_bit(__I40E_DOWN, &pf->state))
3178			napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0]->napi);
3179	}
3180
3181	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3182		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3183		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3184	}
3185
3186	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3187		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3188		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3189	}
3190
3191	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3192		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3193		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3194	}
3195
3196	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3197		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3198			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3199		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3200		val = rd32(hw, I40E_GLGEN_RSTAT);
3201		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3202		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3203		if (val == I40E_RESET_CORER) {
3204			pf->corer_count++;
3205		} else if (val == I40E_RESET_GLOBR) {
3206			pf->globr_count++;
3207		} else if (val == I40E_RESET_EMPR) {
3208			pf->empr_count++;
3209			set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3210		}
3211	}
3212
3213	if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3214		icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3215		dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3216		dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3217			 rd32(hw, I40E_PFHMC_ERRORINFO),
3218			 rd32(hw, I40E_PFHMC_ERRORDATA));
3219	}
3220
3221	if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3222		u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3223
3224		if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3225			icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3226			i40e_ptp_tx_hwtstamp(pf);
3227		}
3228	}
3229
3230	/* If a critical error is pending we have no choice but to reset the
3231	 * device.
3232	 * Report and mask out any remaining unexpected interrupts.
3233	 */
3234	icr0_remaining = icr0 & ena_mask;
3235	if (icr0_remaining) {
3236		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3237			 icr0_remaining);
3238		if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3239		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3240		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3241			dev_info(&pf->pdev->dev, "device will be reset\n");
3242			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3243			i40e_service_event_schedule(pf);
3244		}
3245		ena_mask &= ~icr0_remaining;
3246	}
3247	ret = IRQ_HANDLED;
3248
3249enable_intr:
3250	/* re-enable interrupt causes */
3251	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3252	if (!test_bit(__I40E_DOWN, &pf->state)) {
3253		i40e_service_event_schedule(pf);
3254		i40e_irq_dynamic_enable_icr0(pf);
3255	}
3256
3257	return ret;
3258}
3259
3260/**
3261 * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3262 * @tx_ring:  tx ring to clean
3263 * @budget:   how many cleans we're allowed
3264 *
3265 * Returns true if there's any budget left (e.g. the clean is finished)
3266 **/
3267static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3268{
3269	struct i40e_vsi *vsi = tx_ring->vsi;
3270	u16 i = tx_ring->next_to_clean;
3271	struct i40e_tx_buffer *tx_buf;
3272	struct i40e_tx_desc *tx_desc;
3273
3274	tx_buf = &tx_ring->tx_bi[i];
3275	tx_desc = I40E_TX_DESC(tx_ring, i);
3276	i -= tx_ring->count;
3277
3278	do {
3279		struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3280
3281		/* if next_to_watch is not set then there is no work pending */
3282		if (!eop_desc)
3283			break;
3284
3285		/* prevent any other reads prior to eop_desc */
3286		read_barrier_depends();
3287
3288		/* if the descriptor isn't done, no work yet to do */
3289		if (!(eop_desc->cmd_type_offset_bsz &
3290		      cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3291			break;
3292
3293		/* clear next_to_watch to prevent false hangs */
3294		tx_buf->next_to_watch = NULL;
3295
3296		tx_desc->buffer_addr = 0;
3297		tx_desc->cmd_type_offset_bsz = 0;
3298		/* move past filter desc */
3299		tx_buf++;
3300		tx_desc++;
3301		i++;
3302		if (unlikely(!i)) {
3303			i -= tx_ring->count;
3304			tx_buf = tx_ring->tx_bi;
3305			tx_desc = I40E_TX_DESC(tx_ring, 0);
3306		}
3307		/* unmap skb header data */
3308		dma_unmap_single(tx_ring->dev,
3309				 dma_unmap_addr(tx_buf, dma),
3310				 dma_unmap_len(tx_buf, len),
3311				 DMA_TO_DEVICE);
3312		if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3313			kfree(tx_buf->raw_buf);
3314
3315		tx_buf->raw_buf = NULL;
3316		tx_buf->tx_flags = 0;
3317		tx_buf->next_to_watch = NULL;
3318		dma_unmap_len_set(tx_buf, len, 0);
3319		tx_desc->buffer_addr = 0;
3320		tx_desc->cmd_type_offset_bsz = 0;
3321
3322		/* move us past the eop_desc for start of next FD desc */
3323		tx_buf++;
3324		tx_desc++;
3325		i++;
3326		if (unlikely(!i)) {
3327			i -= tx_ring->count;
3328			tx_buf = tx_ring->tx_bi;
3329			tx_desc = I40E_TX_DESC(tx_ring, 0);
3330		}
3331
3332		/* update budget accounting */
3333		budget--;
3334	} while (likely(budget));
3335
3336	i += tx_ring->count;
3337	tx_ring->next_to_clean = i;
3338
3339	if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED) {
3340		i40e_irq_dynamic_enable(vsi,
3341				tx_ring->q_vector->v_idx + vsi->base_vector);
3342	}
3343	return budget > 0;
3344}
3345
3346/**
3347 * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3348 * @irq: interrupt number
3349 * @data: pointer to a q_vector
3350 **/
3351static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3352{
3353	struct i40e_q_vector *q_vector = data;
3354	struct i40e_vsi *vsi;
3355
3356	if (!q_vector->tx.ring)
3357		return IRQ_HANDLED;
3358
3359	vsi = q_vector->tx.ring->vsi;
3360	i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3361
3362	return IRQ_HANDLED;
3363}
3364
3365/**
3366 * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3367 * @vsi: the VSI being configured
3368 * @v_idx: vector index
3369 * @qp_idx: queue pair index
3370 **/
3371static void map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3372{
3373	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3374	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3375	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3376
3377	tx_ring->q_vector = q_vector;
3378	tx_ring->next = q_vector->tx.ring;
3379	q_vector->tx.ring = tx_ring;
3380	q_vector->tx.count++;
3381
3382	rx_ring->q_vector = q_vector;
3383	rx_ring->next = q_vector->rx.ring;
3384	q_vector->rx.ring = rx_ring;
3385	q_vector->rx.count++;
3386}
3387
3388/**
3389 * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3390 * @vsi: the VSI being configured
3391 *
3392 * This function maps descriptor rings to the queue-specific vectors
3393 * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3394 * one vector per queue pair, but on a constrained vector budget, we
3395 * group the queue pairs as "efficiently" as possible.
3396 **/
3397static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3398{
3399	int qp_remaining = vsi->num_queue_pairs;
3400	int q_vectors = vsi->num_q_vectors;
3401	int num_ringpairs;
3402	int v_start = 0;
3403	int qp_idx = 0;
3404
3405	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3406	 * group them so there are multiple queues per vector.
3407	 * It is also important to go through all the vectors available to be
3408	 * sure that if we don't use all the vectors, that the remaining vectors
3409	 * are cleared. This is especially important when decreasing the
3410	 * number of queues in use.
3411	 */
3412	for (; v_start < q_vectors; v_start++) {
3413		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3414
3415		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3416
3417		q_vector->num_ringpairs = num_ringpairs;
3418
3419		q_vector->rx.count = 0;
3420		q_vector->tx.count = 0;
3421		q_vector->rx.ring = NULL;
3422		q_vector->tx.ring = NULL;
3423
3424		while (num_ringpairs--) {
3425			map_vector_to_qp(vsi, v_start, qp_idx);
3426			qp_idx++;
3427			qp_remaining--;
3428		}
3429	}
3430}
3431
3432/**
3433 * i40e_vsi_request_irq - Request IRQ from the OS
3434 * @vsi: the VSI being configured
3435 * @basename: name for the vector
3436 **/
3437static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3438{
3439	struct i40e_pf *pf = vsi->back;
3440	int err;
3441
3442	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3443		err = i40e_vsi_request_irq_msix(vsi, basename);
3444	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3445		err = request_irq(pf->pdev->irq, i40e_intr, 0,
3446				  pf->int_name, pf);
3447	else
3448		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3449				  pf->int_name, pf);
3450
3451	if (err)
3452		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3453
3454	return err;
3455}
3456
3457#ifdef CONFIG_NET_POLL_CONTROLLER
3458/**
3459 * i40e_netpoll - A Polling 'interrupt'handler
3460 * @netdev: network interface device structure
3461 *
3462 * This is used by netconsole to send skbs without having to re-enable
3463 * interrupts.  It's not called while the normal interrupt routine is executing.
3464 **/
3465#ifdef I40E_FCOE
3466void i40e_netpoll(struct net_device *netdev)
3467#else
3468static void i40e_netpoll(struct net_device *netdev)
3469#endif
3470{
3471	struct i40e_netdev_priv *np = netdev_priv(netdev);
3472	struct i40e_vsi *vsi = np->vsi;
3473	struct i40e_pf *pf = vsi->back;
3474	int i;
3475
3476	/* if interface is down do nothing */
3477	if (test_bit(__I40E_DOWN, &vsi->state))
3478		return;
3479
3480	pf->flags |= I40E_FLAG_IN_NETPOLL;
3481	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3482		for (i = 0; i < vsi->num_q_vectors; i++)
3483			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3484	} else {
3485		i40e_intr(pf->pdev->irq, netdev);
3486	}
3487	pf->flags &= ~I40E_FLAG_IN_NETPOLL;
3488}
3489#endif
3490
3491/**
3492 * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3493 * @pf: the PF being configured
3494 * @pf_q: the PF queue
3495 * @enable: enable or disable state of the queue
3496 *
3497 * This routine will wait for the given Tx queue of the PF to reach the
3498 * enabled or disabled state.
3499 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3500 * multiple retries; else will return 0 in case of success.
3501 **/
3502static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3503{
3504	int i;
3505	u32 tx_reg;
3506
3507	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3508		tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3509		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3510			break;
3511
3512		usleep_range(10, 20);
3513	}
3514	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3515		return -ETIMEDOUT;
3516
3517	return 0;
3518}
3519
3520/**
3521 * i40e_vsi_control_tx - Start or stop a VSI's rings
3522 * @vsi: the VSI being configured
3523 * @enable: start or stop the rings
3524 **/
3525static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3526{
3527	struct i40e_pf *pf = vsi->back;
3528	struct i40e_hw *hw = &pf->hw;
3529	int i, j, pf_q, ret = 0;
3530	u32 tx_reg;
3531
3532	pf_q = vsi->base_queue;
3533	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3534
3535		/* warn the TX unit of coming changes */
3536		i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3537		if (!enable)
3538			usleep_range(10, 20);
3539
3540		for (j = 0; j < 50; j++) {
3541			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3542			if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3543			    ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3544				break;
3545			usleep_range(1000, 2000);
3546		}
3547		/* Skip if the queue is already in the requested state */
3548		if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3549			continue;
3550
3551		/* turn on/off the queue */
3552		if (enable) {
3553			wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3554			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3555		} else {
3556			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3557		}
3558
3559		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3560		/* No waiting for the Tx queue to disable */
3561		if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3562			continue;
3563
3564		/* wait for the change to finish */
3565		ret = i40e_pf_txq_wait(pf, pf_q, enable);
3566		if (ret) {
3567			dev_info(&pf->pdev->dev,
3568				 "%s: VSI seid %d Tx ring %d %sable timeout\n",
3569				 __func__, vsi->seid, pf_q,
3570				 (enable ? "en" : "dis"));
3571			break;
3572		}
3573	}
3574
3575	if (hw->revision_id == 0)
3576		mdelay(50);
3577	return ret;
3578}
3579
3580/**
3581 * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3582 * @pf: the PF being configured
3583 * @pf_q: the PF queue
3584 * @enable: enable or disable state of the queue
3585 *
3586 * This routine will wait for the given Rx queue of the PF to reach the
3587 * enabled or disabled state.
3588 * Returns -ETIMEDOUT in case of failing to reach the requested state after
3589 * multiple retries; else will return 0 in case of success.
3590 **/
3591static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3592{
3593	int i;
3594	u32 rx_reg;
3595
3596	for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3597		rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3598		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3599			break;
3600
3601		usleep_range(10, 20);
3602	}
3603	if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3604		return -ETIMEDOUT;
3605
3606	return 0;
3607}
3608
3609/**
3610 * i40e_vsi_control_rx - Start or stop a VSI's rings
3611 * @vsi: the VSI being configured
3612 * @enable: start or stop the rings
3613 **/
3614static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3615{
3616	struct i40e_pf *pf = vsi->back;
3617	struct i40e_hw *hw = &pf->hw;
3618	int i, j, pf_q, ret = 0;
3619	u32 rx_reg;
3620
3621	pf_q = vsi->base_queue;
3622	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3623		for (j = 0; j < 50; j++) {
3624			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3625			if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3626			    ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3627				break;
3628			usleep_range(1000, 2000);
3629		}
3630
3631		/* Skip if the queue is already in the requested state */
3632		if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3633			continue;
3634
3635		/* turn on/off the queue */
3636		if (enable)
3637			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3638		else
3639			rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3640		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3641
3642		/* wait for the change to finish */
3643		ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3644		if (ret) {
3645			dev_info(&pf->pdev->dev,
3646				 "%s: VSI seid %d Rx ring %d %sable timeout\n",
3647				 __func__, vsi->seid, pf_q,
3648				 (enable ? "en" : "dis"));
3649			break;
3650		}
3651	}
3652
3653	return ret;
3654}
3655
3656/**
3657 * i40e_vsi_control_rings - Start or stop a VSI's rings
3658 * @vsi: the VSI being configured
3659 * @enable: start or stop the rings
3660 **/
3661int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3662{
3663	int ret = 0;
3664
3665	/* do rx first for enable and last for disable */
3666	if (request) {
3667		ret = i40e_vsi_control_rx(vsi, request);
3668		if (ret)
3669			return ret;
3670		ret = i40e_vsi_control_tx(vsi, request);
3671	} else {
3672		/* Ignore return value, we need to shutdown whatever we can */
3673		i40e_vsi_control_tx(vsi, request);
3674		i40e_vsi_control_rx(vsi, request);
3675	}
3676
3677	return ret;
3678}
3679
3680/**
3681 * i40e_vsi_free_irq - Free the irq association with the OS
3682 * @vsi: the VSI being configured
3683 **/
3684static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3685{
3686	struct i40e_pf *pf = vsi->back;
3687	struct i40e_hw *hw = &pf->hw;
3688	int base = vsi->base_vector;
3689	u32 val, qp;
3690	int i;
3691
3692	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3693		if (!vsi->q_vectors)
3694			return;
3695
3696		if (!vsi->irqs_ready)
3697			return;
3698
3699		vsi->irqs_ready = false;
3700		for (i = 0; i < vsi->num_q_vectors; i++) {
3701			u16 vector = i + base;
3702
3703			/* free only the irqs that were actually requested */
3704			if (!vsi->q_vectors[i] ||
3705			    !vsi->q_vectors[i]->num_ringpairs)
3706				continue;
3707
3708			/* clear the affinity_mask in the IRQ descriptor */
3709			irq_set_affinity_hint(pf->msix_entries[vector].vector,
3710					      NULL);
3711			free_irq(pf->msix_entries[vector].vector,
3712				 vsi->q_vectors[i]);
3713
3714			/* Tear down the interrupt queue link list
3715			 *
3716			 * We know that they come in pairs and always
3717			 * the Rx first, then the Tx.  To clear the
3718			 * link list, stick the EOL value into the
3719			 * next_q field of the registers.
3720			 */
3721			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3722			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3723				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3724			val |= I40E_QUEUE_END_OF_LIST
3725				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3726			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3727
3728			while (qp != I40E_QUEUE_END_OF_LIST) {
3729				u32 next;
3730
3731				val = rd32(hw, I40E_QINT_RQCTL(qp));
3732
3733				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3734					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3735					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3736					 I40E_QINT_RQCTL_INTEVENT_MASK);
3737
3738				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3739					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3740
3741				wr32(hw, I40E_QINT_RQCTL(qp), val);
3742
3743				val = rd32(hw, I40E_QINT_TQCTL(qp));
3744
3745				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3746					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3747
3748				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3749					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3750					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3751					 I40E_QINT_TQCTL_INTEVENT_MASK);
3752
3753				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3754					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3755
3756				wr32(hw, I40E_QINT_TQCTL(qp), val);
3757				qp = next;
3758			}
3759		}
3760	} else {
3761		free_irq(pf->pdev->irq, pf);
3762
3763		val = rd32(hw, I40E_PFINT_LNKLST0);
3764		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3765			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3766		val |= I40E_QUEUE_END_OF_LIST
3767			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3768		wr32(hw, I40E_PFINT_LNKLST0, val);
3769
3770		val = rd32(hw, I40E_QINT_RQCTL(qp));
3771		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3772			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3773			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3774			 I40E_QINT_RQCTL_INTEVENT_MASK);
3775
3776		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3777			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3778
3779		wr32(hw, I40E_QINT_RQCTL(qp), val);
3780
3781		val = rd32(hw, I40E_QINT_TQCTL(qp));
3782
3783		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3784			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3785			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3786			 I40E_QINT_TQCTL_INTEVENT_MASK);
3787
3788		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3789			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3790
3791		wr32(hw, I40E_QINT_TQCTL(qp), val);
3792	}
3793}
3794
3795/**
3796 * i40e_free_q_vector - Free memory allocated for specific interrupt vector
3797 * @vsi: the VSI being configured
3798 * @v_idx: Index of vector to be freed
3799 *
3800 * This function frees the memory allocated to the q_vector.  In addition if
3801 * NAPI is enabled it will delete any references to the NAPI struct prior
3802 * to freeing the q_vector.
3803 **/
3804static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
3805{
3806	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3807	struct i40e_ring *ring;
3808
3809	if (!q_vector)
3810		return;
3811
3812	/* disassociate q_vector from rings */
3813	i40e_for_each_ring(ring, q_vector->tx)
3814		ring->q_vector = NULL;
3815
3816	i40e_for_each_ring(ring, q_vector->rx)
3817		ring->q_vector = NULL;
3818
3819	/* only VSI w/ an associated netdev is set up w/ NAPI */
3820	if (vsi->netdev)
3821		netif_napi_del(&q_vector->napi);
3822
3823	vsi->q_vectors[v_idx] = NULL;
3824
3825	kfree_rcu(q_vector, rcu);
3826}
3827
3828/**
3829 * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3830 * @vsi: the VSI being un-configured
3831 *
3832 * This frees the memory allocated to the q_vectors and
3833 * deletes references to the NAPI struct.
3834 **/
3835static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3836{
3837	int v_idx;
3838
3839	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
3840		i40e_free_q_vector(vsi, v_idx);
3841}
3842
3843/**
3844 * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3845 * @pf: board private structure
3846 **/
3847static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3848{
3849	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3850	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3851		pci_disable_msix(pf->pdev);
3852		kfree(pf->msix_entries);
3853		pf->msix_entries = NULL;
3854		kfree(pf->irq_pile);
3855		pf->irq_pile = NULL;
3856	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3857		pci_disable_msi(pf->pdev);
3858	}
3859	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3860}
3861
3862/**
3863 * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3864 * @pf: board private structure
3865 *
3866 * We go through and clear interrupt specific resources and reset the structure
3867 * to pre-load conditions
3868 **/
3869static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3870{
3871	int i;
3872
3873	i40e_stop_misc_vector(pf);
3874	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3875		synchronize_irq(pf->msix_entries[0].vector);
3876		free_irq(pf->msix_entries[0].vector, pf);
3877	}
3878
3879	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3880	for (i = 0; i < pf->num_alloc_vsi; i++)
3881		if (pf->vsi[i])
3882			i40e_vsi_free_q_vectors(pf->vsi[i]);
3883	i40e_reset_interrupt_capability(pf);
3884}
3885
3886/**
3887 * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3888 * @vsi: the VSI being configured
3889 **/
3890static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3891{
3892	int q_idx;
3893
3894	if (!vsi->netdev)
3895		return;
3896
3897	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3898		napi_enable(&vsi->q_vectors[q_idx]->napi);
3899}
3900
3901/**
3902 * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3903 * @vsi: the VSI being configured
3904 **/
3905static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3906{
3907	int q_idx;
3908
3909	if (!vsi->netdev)
3910		return;
3911
3912	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3913		napi_disable(&vsi->q_vectors[q_idx]->napi);
3914}
3915
3916/**
3917 * i40e_vsi_close - Shut down a VSI
3918 * @vsi: the vsi to be quelled
3919 **/
3920static void i40e_vsi_close(struct i40e_vsi *vsi)
3921{
3922	if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
3923		i40e_down(vsi);
3924	i40e_vsi_free_irq(vsi);
3925	i40e_vsi_free_tx_resources(vsi);
3926	i40e_vsi_free_rx_resources(vsi);
3927}
3928
3929/**
3930 * i40e_quiesce_vsi - Pause a given VSI
3931 * @vsi: the VSI being paused
3932 **/
3933static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3934{
3935	if (test_bit(__I40E_DOWN, &vsi->state))
3936		return;
3937
3938	/* No need to disable FCoE VSI when Tx suspended */
3939	if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
3940	    vsi->type == I40E_VSI_FCOE) {
3941		dev_dbg(&vsi->back->pdev->dev,
3942			"%s: VSI seid %d skipping FCoE VSI disable\n",
3943			 __func__, vsi->seid);
3944		return;
3945	}
3946
3947	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3948	if (vsi->netdev && netif_running(vsi->netdev)) {
3949		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3950	} else {
3951		i40e_vsi_close(vsi);
3952	}
3953}
3954
3955/**
3956 * i40e_unquiesce_vsi - Resume a given VSI
3957 * @vsi: the VSI being resumed
3958 **/
3959static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3960{
3961	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3962		return;
3963
3964	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3965	if (vsi->netdev && netif_running(vsi->netdev))
3966		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3967	else
3968		i40e_vsi_open(vsi);   /* this clears the DOWN bit */
3969}
3970
3971/**
3972 * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3973 * @pf: the PF
3974 **/
3975static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3976{
3977	int v;
3978
3979	for (v = 0; v < pf->num_alloc_vsi; v++) {
3980		if (pf->vsi[v])
3981			i40e_quiesce_vsi(pf->vsi[v]);
3982	}
3983}
3984
3985/**
3986 * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3987 * @pf: the PF
3988 **/
3989static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3990{
3991	int v;
3992
3993	for (v = 0; v < pf->num_alloc_vsi; v++) {
3994		if (pf->vsi[v])
3995			i40e_unquiesce_vsi(pf->vsi[v]);
3996	}
3997}
3998
3999#ifdef CONFIG_I40E_DCB
4000/**
4001 * i40e_vsi_wait_txq_disabled - Wait for VSI's queues to be disabled
4002 * @vsi: the VSI being configured
4003 *
4004 * This function waits for the given VSI's Tx queues to be disabled.
4005 **/
4006static int i40e_vsi_wait_txq_disabled(struct i40e_vsi *vsi)
4007{
4008	struct i40e_pf *pf = vsi->back;
4009	int i, pf_q, ret;
4010
4011	pf_q = vsi->base_queue;
4012	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4013		/* Check and wait for the disable status of the queue */
4014		ret = i40e_pf_txq_wait(pf, pf_q, false);
4015		if (ret) {
4016			dev_info(&pf->pdev->dev,
4017				 "%s: VSI seid %d Tx ring %d disable timeout\n",
4018				 __func__, vsi->seid, pf_q);
4019			return ret;
4020		}
4021	}
4022
4023	return 0;
4024}
4025
4026/**
4027 * i40e_pf_wait_txq_disabled - Wait for all queues of PF VSIs to be disabled
4028 * @pf: the PF
4029 *
4030 * This function waits for the Tx queues to be in disabled state for all the
4031 * VSIs that are managed by this PF.
4032 **/
4033static int i40e_pf_wait_txq_disabled(struct i40e_pf *pf)
4034{
4035	int v, ret = 0;
4036
4037	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4038		/* No need to wait for FCoE VSI queues */
4039		if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4040			ret = i40e_vsi_wait_txq_disabled(pf->vsi[v]);
4041			if (ret)
4042				break;
4043		}
4044	}
4045
4046	return ret;
4047}
4048
4049#endif
4050/**
4051 * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4052 * @pf: pointer to PF
4053 *
4054 * Get TC map for ISCSI PF type that will include iSCSI TC
4055 * and LAN TC.
4056 **/
4057static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4058{
4059	struct i40e_dcb_app_priority_table app;
4060	struct i40e_hw *hw = &pf->hw;
4061	u8 enabled_tc = 1; /* TC0 is always enabled */
4062	u8 tc, i;
4063	/* Get the iSCSI APP TLV */
4064	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4065
4066	for (i = 0; i < dcbcfg->numapps; i++) {
4067		app = dcbcfg->app[i];
4068		if (app.selector == I40E_APP_SEL_TCPIP &&
4069		    app.protocolid == I40E_APP_PROTOID_ISCSI) {
4070			tc = dcbcfg->etscfg.prioritytable[app.priority];
4071			enabled_tc |= (1 << tc);
4072			break;
4073		}
4074	}
4075
4076	return enabled_tc;
4077}
4078
4079/**
4080 * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4081 * @dcbcfg: the corresponding DCBx configuration structure
4082 *
4083 * Return the number of TCs from given DCBx configuration
4084 **/
4085static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4086{
4087	u8 num_tc = 0;
4088	int i;
4089
4090	/* Scan the ETS Config Priority Table to find
4091	 * traffic class enabled for a given priority
4092	 * and use the traffic class index to get the
4093	 * number of traffic classes enabled
4094	 */
4095	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4096		if (dcbcfg->etscfg.prioritytable[i] > num_tc)
4097			num_tc = dcbcfg->etscfg.prioritytable[i];
4098	}
4099
4100	/* Traffic class index starts from zero so
4101	 * increment to return the actual count
4102	 */
4103	return num_tc + 1;
4104}
4105
4106/**
4107 * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4108 * @dcbcfg: the corresponding DCBx configuration structure
4109 *
4110 * Query the current DCB configuration and return the number of
4111 * traffic classes enabled from the given DCBX config
4112 **/
4113static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4114{
4115	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4116	u8 enabled_tc = 1;
4117	u8 i;
4118
4119	for (i = 0; i < num_tc; i++)
4120		enabled_tc |= 1 << i;
4121
4122	return enabled_tc;
4123}
4124
4125/**
4126 * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4127 * @pf: PF being queried
4128 *
4129 * Return number of traffic classes enabled for the given PF
4130 **/
4131static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4132{
4133	struct i40e_hw *hw = &pf->hw;
4134	u8 i, enabled_tc;
4135	u8 num_tc = 0;
4136	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4137
4138	/* If DCB is not enabled then always in single TC */
4139	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4140		return 1;
4141
4142	/* SFP mode will be enabled for all TCs on port */
4143	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4144		return i40e_dcb_get_num_tc(dcbcfg);
4145
4146	/* MFP mode return count of enabled TCs for this PF */
4147	if (pf->hw.func_caps.iscsi)
4148		enabled_tc =  i40e_get_iscsi_tc_map(pf);
4149	else
4150		return 1; /* Only TC0 */
4151
4152	/* At least have TC0 */
4153	enabled_tc = (enabled_tc ? enabled_tc : 0x1);
4154	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4155		if (enabled_tc & (1 << i))
4156			num_tc++;
4157	}
4158	return num_tc;
4159}
4160
4161/**
4162 * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4163 * @pf: PF being queried
4164 *
4165 * Return a bitmap for first enabled traffic class for this PF.
4166 **/
4167static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4168{
4169	u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4170	u8 i = 0;
4171
4172	if (!enabled_tc)
4173		return 0x1; /* TC0 */
4174
4175	/* Find the first enabled TC */
4176	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4177		if (enabled_tc & (1 << i))
4178			break;
4179	}
4180
4181	return 1 << i;
4182}
4183
4184/**
4185 * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4186 * @pf: PF being queried
4187 *
4188 * Return a bitmap for enabled traffic classes for this PF.
4189 **/
4190static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4191{
4192	/* If DCB is not enabled for this PF then just return default TC */
4193	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4194		return i40e_pf_get_default_tc(pf);
4195
4196	/* SFP mode we want PF to be enabled for all TCs */
4197	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4198		return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4199
4200	/* MFP enabled and iSCSI PF type */
4201	if (pf->hw.func_caps.iscsi)
4202		return i40e_get_iscsi_tc_map(pf);
4203	else
4204		return i40e_pf_get_default_tc(pf);
4205}
4206
4207/**
4208 * i40e_vsi_get_bw_info - Query VSI BW Information
4209 * @vsi: the VSI being queried
4210 *
4211 * Returns 0 on success, negative value on failure
4212 **/
4213static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4214{
4215	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4216	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4217	struct i40e_pf *pf = vsi->back;
4218	struct i40e_hw *hw = &pf->hw;
4219	i40e_status aq_ret;
4220	u32 tc_bw_max;
4221	int i;
4222
4223	/* Get the VSI level BW configuration */
4224	aq_ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4225	if (aq_ret) {
4226		dev_info(&pf->pdev->dev,
4227			 "couldn't get PF vsi bw config, err %d, aq_err %d\n",
4228			 aq_ret, pf->hw.aq.asq_last_status);
4229		return -EINVAL;
4230	}
4231
4232	/* Get the VSI level BW configuration per TC */
4233	aq_ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4234						  NULL);
4235	if (aq_ret) {
4236		dev_info(&pf->pdev->dev,
4237			 "couldn't get PF vsi ets bw config, err %d, aq_err %d\n",
4238			 aq_ret, pf->hw.aq.asq_last_status);
4239		return -EINVAL;
4240	}
4241
4242	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4243		dev_info(&pf->pdev->dev,
4244			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4245			 bw_config.tc_valid_bits,
4246			 bw_ets_config.tc_valid_bits);
4247		/* Still continuing */
4248	}
4249
4250	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4251	vsi->bw_max_quanta = bw_config.max_bw;
4252	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4253		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4254	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4255		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4256		vsi->bw_ets_limit_credits[i] =
4257					le16_to_cpu(bw_ets_config.credits[i]);
4258		/* 3 bits out of 4 for each TC */
4259		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4260	}
4261
4262	return 0;
4263}
4264
4265/**
4266 * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4267 * @vsi: the VSI being configured
4268 * @enabled_tc: TC bitmap
4269 * @bw_credits: BW shared credits per TC
4270 *
4271 * Returns 0 on success, negative value on failure
4272 **/
4273static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4274				       u8 *bw_share)
4275{
4276	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4277	i40e_status aq_ret;
4278	int i;
4279
4280	bw_data.tc_valid_bits = enabled_tc;
4281	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4282		bw_data.tc_bw_credits[i] = bw_share[i];
4283
4284	aq_ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4285					  NULL);
4286	if (aq_ret) {
4287		dev_info(&vsi->back->pdev->dev,
4288			 "AQ command Config VSI BW allocation per TC failed = %d\n",
4289			 vsi->back->hw.aq.asq_last_status);
4290		return -EINVAL;
4291	}
4292
4293	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4294		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4295
4296	return 0;
4297}
4298
4299/**
4300 * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4301 * @vsi: the VSI being configured
4302 * @enabled_tc: TC map to be enabled
4303 *
4304 **/
4305static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4306{
4307	struct net_device *netdev = vsi->netdev;
4308	struct i40e_pf *pf = vsi->back;
4309	struct i40e_hw *hw = &pf->hw;
4310	u8 netdev_tc = 0;
4311	int i;
4312	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4313
4314	if (!netdev)
4315		return;
4316
4317	if (!enabled_tc) {
4318		netdev_reset_tc(netdev);
4319		return;
4320	}
4321
4322	/* Set up actual enabled TCs on the VSI */
4323	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4324		return;
4325
4326	/* set per TC queues for the VSI */
4327	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4328		/* Only set TC queues for enabled tcs
4329		 *
4330		 * e.g. For a VSI that has TC0 and TC3 enabled the
4331		 * enabled_tc bitmap would be 0x00001001; the driver
4332		 * will set the numtc for netdev as 2 that will be
4333		 * referenced by the netdev layer as TC 0 and 1.
4334		 */
4335		if (vsi->tc_config.enabled_tc & (1 << i))
4336			netdev_set_tc_queue(netdev,
4337					vsi->tc_config.tc_info[i].netdev_tc,
4338					vsi->tc_config.tc_info[i].qcount,
4339					vsi->tc_config.tc_info[i].qoffset);
4340	}
4341
4342	/* Assign UP2TC map for the VSI */
4343	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4344		/* Get the actual TC# for the UP */
4345		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4346		/* Get the mapped netdev TC# for the UP */
4347		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4348		netdev_set_prio_tc_map(netdev, i, netdev_tc);
4349	}
4350}
4351
4352/**
4353 * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4354 * @vsi: the VSI being configured
4355 * @ctxt: the ctxt buffer returned from AQ VSI update param command
4356 **/
4357static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4358				      struct i40e_vsi_context *ctxt)
4359{
4360	/* copy just the sections touched not the entire info
4361	 * since not all sections are valid as returned by
4362	 * update vsi params
4363	 */
4364	vsi->info.mapping_flags = ctxt->info.mapping_flags;
4365	memcpy(&vsi->info.queue_mapping,
4366	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4367	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4368	       sizeof(vsi->info.tc_mapping));
4369}
4370
4371/**
4372 * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4373 * @vsi: VSI to be configured
4374 * @enabled_tc: TC bitmap
4375 *
4376 * This configures a particular VSI for TCs that are mapped to the
4377 * given TC bitmap. It uses default bandwidth share for TCs across
4378 * VSIs to configure TC for a particular VSI.
4379 *
4380 * NOTE:
4381 * It is expected that the VSI queues have been quisced before calling
4382 * this function.
4383 **/
4384static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4385{
4386	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4387	struct i40e_vsi_context ctxt;
4388	int ret = 0;
4389	int i;
4390
4391	/* Check if enabled_tc is same as existing or new TCs */
4392	if (vsi->tc_config.enabled_tc == enabled_tc)
4393		return ret;
4394
4395	/* Enable ETS TCs with equal BW Share for now across all VSIs */
4396	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4397		if (enabled_tc & (1 << i))
4398			bw_share[i] = 1;
4399	}
4400
4401	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4402	if (ret) {
4403		dev_info(&vsi->back->pdev->dev,
4404			 "Failed configuring TC map %d for VSI %d\n",
4405			 enabled_tc, vsi->seid);
4406		goto out;
4407	}
4408
4409	/* Update Queue Pairs Mapping for currently enabled UPs */
4410	ctxt.seid = vsi->seid;
4411	ctxt.pf_num = vsi->back->hw.pf_id;
4412	ctxt.vf_num = 0;
4413	ctxt.uplink_seid = vsi->uplink_seid;
4414	ctxt.info = vsi->info;
4415	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4416
4417	/* Update the VSI after updating the VSI queue-mapping information */
4418	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4419	if (ret) {
4420		dev_info(&vsi->back->pdev->dev,
4421			 "update vsi failed, aq_err=%d\n",
4422			 vsi->back->hw.aq.asq_last_status);
4423		goto out;
4424	}
4425	/* update the local VSI info with updated queue map */
4426	i40e_vsi_update_queue_map(vsi, &ctxt);
4427	vsi->info.valid_sections = 0;
4428
4429	/* Update current VSI BW information */
4430	ret = i40e_vsi_get_bw_info(vsi);
4431	if (ret) {
4432		dev_info(&vsi->back->pdev->dev,
4433			 "Failed updating vsi bw info, aq_err=%d\n",
4434			 vsi->back->hw.aq.asq_last_status);
4435		goto out;
4436	}
4437
4438	/* Update the netdev TC setup */
4439	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4440out:
4441	return ret;
4442}
4443
4444/**
4445 * i40e_veb_config_tc - Configure TCs for given VEB
4446 * @veb: given VEB
4447 * @enabled_tc: TC bitmap
4448 *
4449 * Configures given TC bitmap for VEB (switching) element
4450 **/
4451int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4452{
4453	struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4454	struct i40e_pf *pf = veb->pf;
4455	int ret = 0;
4456	int i;
4457
4458	/* No TCs or already enabled TCs just return */
4459	if (!enabled_tc || veb->enabled_tc == enabled_tc)
4460		return ret;
4461
4462	bw_data.tc_valid_bits = enabled_tc;
4463	/* bw_data.absolute_credits is not set (relative) */
4464
4465	/* Enable ETS TCs with equal BW Share for now */
4466	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4467		if (enabled_tc & (1 << i))
4468			bw_data.tc_bw_share_credits[i] = 1;
4469	}
4470
4471	ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4472						   &bw_data, NULL);
4473	if (ret) {
4474		dev_info(&pf->pdev->dev,
4475			 "veb bw config failed, aq_err=%d\n",
4476			 pf->hw.aq.asq_last_status);
4477		goto out;
4478	}
4479
4480	/* Update the BW information */
4481	ret = i40e_veb_get_bw_info(veb);
4482	if (ret) {
4483		dev_info(&pf->pdev->dev,
4484			 "Failed getting veb bw config, aq_err=%d\n",
4485			 pf->hw.aq.asq_last_status);
4486	}
4487
4488out:
4489	return ret;
4490}
4491
4492#ifdef CONFIG_I40E_DCB
4493/**
4494 * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4495 * @pf: PF struct
4496 *
4497 * Reconfigure VEB/VSIs on a given PF; it is assumed that
4498 * the caller would've quiesce all the VSIs before calling
4499 * this function
4500 **/
4501static void i40e_dcb_reconfigure(struct i40e_pf *pf)
4502{
4503	u8 tc_map = 0;
4504	int ret;
4505	u8 v;
4506
4507	/* Enable the TCs available on PF to all VEBs */
4508	tc_map = i40e_pf_get_tc_map(pf);
4509	for (v = 0; v < I40E_MAX_VEB; v++) {
4510		if (!pf->veb[v])
4511			continue;
4512		ret = i40e_veb_config_tc(pf->veb[v], tc_map);
4513		if (ret) {
4514			dev_info(&pf->pdev->dev,
4515				 "Failed configuring TC for VEB seid=%d\n",
4516				 pf->veb[v]->seid);
4517			/* Will try to configure as many components */
4518		}
4519	}
4520
4521	/* Update each VSI */
4522	for (v = 0; v < pf->num_alloc_vsi; v++) {
4523		if (!pf->vsi[v])
4524			continue;
4525
4526		/* - Enable all TCs for the LAN VSI
4527#ifdef I40E_FCOE
4528		 * - For FCoE VSI only enable the TC configured
4529		 *   as per the APP TLV
4530#endif
4531		 * - For all others keep them at TC0 for now
4532		 */
4533		if (v == pf->lan_vsi)
4534			tc_map = i40e_pf_get_tc_map(pf);
4535		else
4536			tc_map = i40e_pf_get_default_tc(pf);
4537#ifdef I40E_FCOE
4538		if (pf->vsi[v]->type == I40E_VSI_FCOE)
4539			tc_map = i40e_get_fcoe_tc_map(pf);
4540#endif /* #ifdef I40E_FCOE */
4541
4542		ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
4543		if (ret) {
4544			dev_info(&pf->pdev->dev,
4545				 "Failed configuring TC for VSI seid=%d\n",
4546				 pf->vsi[v]->seid);
4547			/* Will try to configure as many components */
4548		} else {
4549			/* Re-configure VSI vectors based on updated TC map */
4550			i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
4551			if (pf->vsi[v]->netdev)
4552				i40e_dcbnl_set_all(pf->vsi[v]);
4553		}
4554	}
4555}
4556
4557/**
4558 * i40e_resume_port_tx - Resume port Tx
4559 * @pf: PF struct
4560 *
4561 * Resume a port's Tx and issue a PF reset in case of failure to
4562 * resume.
4563 **/
4564static int i40e_resume_port_tx(struct i40e_pf *pf)
4565{
4566	struct i40e_hw *hw = &pf->hw;
4567	int ret;
4568
4569	ret = i40e_aq_resume_port_tx(hw, NULL);
4570	if (ret) {
4571		dev_info(&pf->pdev->dev,
4572			 "AQ command Resume Port Tx failed = %d\n",
4573			  pf->hw.aq.asq_last_status);
4574		/* Schedule PF reset to recover */
4575		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4576		i40e_service_event_schedule(pf);
4577	}
4578
4579	return ret;
4580}
4581
4582/**
4583 * i40e_init_pf_dcb - Initialize DCB configuration
4584 * @pf: PF being configured
4585 *
4586 * Query the current DCB configuration and cache it
4587 * in the hardware structure
4588 **/
4589static int i40e_init_pf_dcb(struct i40e_pf *pf)
4590{
4591	struct i40e_hw *hw = &pf->hw;
4592	int err = 0;
4593
4594	/* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
4595	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
4596	    (pf->hw.aq.fw_maj_ver < 4))
4597		goto out;
4598
4599	/* Get the initial DCB configuration */
4600	err = i40e_init_dcb(hw);
4601	if (!err) {
4602		/* Device/Function is not DCBX capable */
4603		if ((!hw->func_caps.dcb) ||
4604		    (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
4605			dev_info(&pf->pdev->dev,
4606				 "DCBX offload is not supported or is disabled for this PF.\n");
4607
4608			if (pf->flags & I40E_FLAG_MFP_ENABLED)
4609				goto out;
4610
4611		} else {
4612			/* When status is not DISABLED then DCBX in FW */
4613			pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
4614				       DCB_CAP_DCBX_VER_IEEE;
4615
4616			pf->flags |= I40E_FLAG_DCB_CAPABLE;
4617			/* Enable DCB tagging only when more than one TC */
4618			if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
4619				pf->flags |= I40E_FLAG_DCB_ENABLED;
4620			dev_dbg(&pf->pdev->dev,
4621				"DCBX offload is supported for this PF.\n");
4622		}
4623	} else {
4624		dev_info(&pf->pdev->dev,
4625			 "AQ Querying DCB configuration failed: aq_err %d\n",
4626			 pf->hw.aq.asq_last_status);
4627	}
4628
4629out:
4630	return err;
4631}
4632#endif /* CONFIG_I40E_DCB */
4633#define SPEED_SIZE 14
4634#define FC_SIZE 8
4635/**
4636 * i40e_print_link_message - print link up or down
4637 * @vsi: the VSI for which link needs a message
4638 */
4639static void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
4640{
4641	char speed[SPEED_SIZE] = "Unknown";
4642	char fc[FC_SIZE] = "RX/TX";
4643
4644	if (!isup) {
4645		netdev_info(vsi->netdev, "NIC Link is Down\n");
4646		return;
4647	}
4648
4649	/* Warn user if link speed on NPAR enabled partition is not at
4650	 * least 10GB
4651	 */
4652	if (vsi->back->hw.func_caps.npar_enable &&
4653	    (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
4654	     vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
4655		netdev_warn(vsi->netdev,
4656			    "The partition detected link speed that is less than 10Gbps\n");
4657
4658	switch (vsi->back->hw.phy.link_info.link_speed) {
4659	case I40E_LINK_SPEED_40GB:
4660		strlcpy(speed, "40 Gbps", SPEED_SIZE);
4661		break;
4662	case I40E_LINK_SPEED_20GB:
4663		strncpy(speed, "20 Gbps", SPEED_SIZE);
4664		break;
4665	case I40E_LINK_SPEED_10GB:
4666		strlcpy(speed, "10 Gbps", SPEED_SIZE);
4667		break;
4668	case I40E_LINK_SPEED_1GB:
4669		strlcpy(speed, "1000 Mbps", SPEED_SIZE);
4670		break;
4671	case I40E_LINK_SPEED_100MB:
4672		strncpy(speed, "100 Mbps", SPEED_SIZE);
4673		break;
4674	default:
4675		break;
4676	}
4677
4678	switch (vsi->back->hw.fc.current_mode) {
4679	case I40E_FC_FULL:
4680		strlcpy(fc, "RX/TX", FC_SIZE);
4681		break;
4682	case I40E_FC_TX_PAUSE:
4683		strlcpy(fc, "TX", FC_SIZE);
4684		break;
4685	case I40E_FC_RX_PAUSE:
4686		strlcpy(fc, "RX", FC_SIZE);
4687		break;
4688	default:
4689		strlcpy(fc, "None", FC_SIZE);
4690		break;
4691	}
4692
4693	netdev_info(vsi->netdev, "NIC Link is Up %s Full Duplex, Flow Control: %s\n",
4694		    speed, fc);
4695}
4696
4697/**
4698 * i40e_up_complete - Finish the last steps of bringing up a connection
4699 * @vsi: the VSI being configured
4700 **/
4701static int i40e_up_complete(struct i40e_vsi *vsi)
4702{
4703	struct i40e_pf *pf = vsi->back;
4704	int err;
4705
4706	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4707		i40e_vsi_configure_msix(vsi);
4708	else
4709		i40e_configure_msi_and_legacy(vsi);
4710
4711	/* start rings */
4712	err = i40e_vsi_control_rings(vsi, true);
4713	if (err)
4714		return err;
4715
4716	clear_bit(__I40E_DOWN, &vsi->state);
4717	i40e_napi_enable_all(vsi);
4718	i40e_vsi_enable_irq(vsi);
4719
4720	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
4721	    (vsi->netdev)) {
4722		i40e_print_link_message(vsi, true);
4723		netif_tx_start_all_queues(vsi->netdev);
4724		netif_carrier_on(vsi->netdev);
4725	} else if (vsi->netdev) {
4726		i40e_print_link_message(vsi, false);
4727		/* need to check for qualified module here*/
4728		if ((pf->hw.phy.link_info.link_info &
4729			I40E_AQ_MEDIA_AVAILABLE) &&
4730		    (!(pf->hw.phy.link_info.an_info &
4731			I40E_AQ_QUALIFIED_MODULE)))
4732			netdev_err(vsi->netdev,
4733				   "the driver failed to link because an unqualified module was detected.");
4734	}
4735
4736	/* replay FDIR SB filters */
4737	if (vsi->type == I40E_VSI_FDIR) {
4738		/* reset fd counters */
4739		pf->fd_add_err = pf->fd_atr_cnt = 0;
4740		if (pf->fd_tcp_rule > 0) {
4741			pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
4742			dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
4743			pf->fd_tcp_rule = 0;
4744		}
4745		i40e_fdir_filter_restore(vsi);
4746	}
4747	i40e_service_event_schedule(pf);
4748
4749	return 0;
4750}
4751
4752/**
4753 * i40e_vsi_reinit_locked - Reset the VSI
4754 * @vsi: the VSI being configured
4755 *
4756 * Rebuild the ring structs after some configuration
4757 * has changed, e.g. MTU size.
4758 **/
4759static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
4760{
4761	struct i40e_pf *pf = vsi->back;
4762
4763	WARN_ON(in_interrupt());
4764	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
4765		usleep_range(1000, 2000);
4766	i40e_down(vsi);
4767
4768	/* Give a VF some time to respond to the reset.  The
4769	 * two second wait is based upon the watchdog cycle in
4770	 * the VF driver.
4771	 */
4772	if (vsi->type == I40E_VSI_SRIOV)
4773		msleep(2000);
4774	i40e_up(vsi);
4775	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
4776}
4777
4778/**
4779 * i40e_up - Bring the connection back up after being down
4780 * @vsi: the VSI being configured
4781 **/
4782int i40e_up(struct i40e_vsi *vsi)
4783{
4784	int err;
4785
4786	err = i40e_vsi_configure(vsi);
4787	if (!err)
4788		err = i40e_up_complete(vsi);
4789
4790	return err;
4791}
4792
4793/**
4794 * i40e_down - Shutdown the connection processing
4795 * @vsi: the VSI being stopped
4796 **/
4797void i40e_down(struct i40e_vsi *vsi)
4798{
4799	int i;
4800
4801	/* It is assumed that the caller of this function
4802	 * sets the vsi->state __I40E_DOWN bit.
4803	 */
4804	if (vsi->netdev) {
4805		netif_carrier_off(vsi->netdev);
4806		netif_tx_disable(vsi->netdev);
4807	}
4808	i40e_vsi_disable_irq(vsi);
4809	i40e_vsi_control_rings(vsi, false);
4810	i40e_napi_disable_all(vsi);
4811
4812	for (i = 0; i < vsi->num_queue_pairs; i++) {
4813		i40e_clean_tx_ring(vsi->tx_rings[i]);
4814		i40e_clean_rx_ring(vsi->rx_rings[i]);
4815	}
4816}
4817
4818/**
4819 * i40e_setup_tc - configure multiple traffic classes
4820 * @netdev: net device to configure
4821 * @tc: number of traffic classes to enable
4822 **/
4823#ifdef I40E_FCOE
4824int i40e_setup_tc(struct net_device *netdev, u8 tc)
4825#else
4826static int i40e_setup_tc(struct net_device *netdev, u8 tc)
4827#endif
4828{
4829	struct i40e_netdev_priv *np = netdev_priv(netdev);
4830	struct i40e_vsi *vsi = np->vsi;
4831	struct i40e_pf *pf = vsi->back;
4832	u8 enabled_tc = 0;
4833	int ret = -EINVAL;
4834	int i;
4835
4836	/* Check if DCB enabled to continue */
4837	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
4838		netdev_info(netdev, "DCB is not enabled for adapter\n");
4839		goto exit;
4840	}
4841
4842	/* Check if MFP enabled */
4843	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
4844		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
4845		goto exit;
4846	}
4847
4848	/* Check whether tc count is within enabled limit */
4849	if (tc > i40e_pf_get_num_tc(pf)) {
4850		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
4851		goto exit;
4852	}
4853
4854	/* Generate TC map for number of tc requested */
4855	for (i = 0; i < tc; i++)
4856		enabled_tc |= (1 << i);
4857
4858	/* Requesting same TC configuration as already enabled */
4859	if (enabled_tc == vsi->tc_config.enabled_tc)
4860		return 0;
4861
4862	/* Quiesce VSI queues */
4863	i40e_quiesce_vsi(vsi);
4864
4865	/* Configure VSI for enabled TCs */
4866	ret = i40e_vsi_config_tc(vsi, enabled_tc);
4867	if (ret) {
4868		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
4869			    vsi->seid);
4870		goto exit;
4871	}
4872
4873	/* Unquiesce VSI */
4874	i40e_unquiesce_vsi(vsi);
4875
4876exit:
4877	return ret;
4878}
4879
4880/**
4881 * i40e_open - Called when a network interface is made active
4882 * @netdev: network interface device structure
4883 *
4884 * The open entry point is called when a network interface is made
4885 * active by the system (IFF_UP).  At this point all resources needed
4886 * for transmit and receive operations are allocated, the interrupt
4887 * handler is registered with the OS, the netdev watchdog subtask is
4888 * enabled, and the stack is notified that the interface is ready.
4889 *
4890 * Returns 0 on success, negative value on failure
4891 **/
4892int i40e_open(struct net_device *netdev)
4893{
4894	struct i40e_netdev_priv *np = netdev_priv(netdev);
4895	struct i40e_vsi *vsi = np->vsi;
4896	struct i40e_pf *pf = vsi->back;
4897	int err;
4898
4899	/* disallow open during test or if eeprom is broken */
4900	if (test_bit(__I40E_TESTING, &pf->state) ||
4901	    test_bit(__I40E_BAD_EEPROM, &pf->state))
4902		return -EBUSY;
4903
4904	netif_carrier_off(netdev);
4905
4906	err = i40e_vsi_open(vsi);
4907	if (err)
4908		return err;
4909
4910	/* configure global TSO hardware offload settings */
4911	wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
4912						       TCP_FLAG_FIN) >> 16);
4913	wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
4914						       TCP_FLAG_FIN |
4915						       TCP_FLAG_CWR) >> 16);
4916	wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
4917
4918#ifdef CONFIG_I40E_VXLAN
4919	vxlan_get_rx_port(netdev);
4920#endif
4921
4922	return 0;
4923}
4924
4925/**
4926 * i40e_vsi_open -
4927 * @vsi: the VSI to open
4928 *
4929 * Finish initialization of the VSI.
4930 *
4931 * Returns 0 on success, negative value on failure
4932 **/
4933int i40e_vsi_open(struct i40e_vsi *vsi)
4934{
4935	struct i40e_pf *pf = vsi->back;
4936	char int_name[I40E_INT_NAME_STR_LEN];
4937	int err;
4938
4939	/* allocate descriptors */
4940	err = i40e_vsi_setup_tx_resources(vsi);
4941	if (err)
4942		goto err_setup_tx;
4943	err = i40e_vsi_setup_rx_resources(vsi);
4944	if (err)
4945		goto err_setup_rx;
4946
4947	err = i40e_vsi_configure(vsi);
4948	if (err)
4949		goto err_setup_rx;
4950
4951	if (vsi->netdev) {
4952		snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
4953			 dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
4954		err = i40e_vsi_request_irq(vsi, int_name);
4955		if (err)
4956			goto err_setup_rx;
4957
4958		/* Notify the stack of the actual queue counts. */
4959		err = netif_set_real_num_tx_queues(vsi->netdev,
4960						   vsi->num_queue_pairs);
4961		if (err)
4962			goto err_set_queues;
4963
4964		err = netif_set_real_num_rx_queues(vsi->netdev,
4965						   vsi->num_queue_pairs);
4966		if (err)
4967			goto err_set_queues;
4968
4969	} else if (vsi->type == I40E_VSI_FDIR) {
4970		snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
4971			 dev_driver_string(&pf->pdev->dev),
4972			 dev_name(&pf->pdev->dev));
4973		err = i40e_vsi_request_irq(vsi, int_name);
4974
4975	} else {
4976		err = -EINVAL;
4977		goto err_setup_rx;
4978	}
4979
4980	err = i40e_up_complete(vsi);
4981	if (err)
4982		goto err_up_complete;
4983
4984	return 0;
4985
4986err_up_complete:
4987	i40e_down(vsi);
4988err_set_queues:
4989	i40e_vsi_free_irq(vsi);
4990err_setup_rx:
4991	i40e_vsi_free_rx_resources(vsi);
4992err_setup_tx:
4993	i40e_vsi_free_tx_resources(vsi);
4994	if (vsi == pf->vsi[pf->lan_vsi])
4995		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
4996
4997	return err;
4998}
4999
5000/**
5001 * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5002 * @pf: Pointer to PF
5003 *
5004 * This function destroys the hlist where all the Flow Director
5005 * filters were saved.
5006 **/
5007static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5008{
5009	struct i40e_fdir_filter *filter;
5010	struct hlist_node *node2;
5011
5012	hlist_for_each_entry_safe(filter, node2,
5013				  &pf->fdir_filter_list, fdir_node) {
5014		hlist_del(&filter->fdir_node);
5015		kfree(filter);
5016	}
5017	pf->fdir_pf_active_filters = 0;
5018}
5019
5020/**
5021 * i40e_close - Disables a network interface
5022 * @netdev: network interface device structure
5023 *
5024 * The close entry point is called when an interface is de-activated
5025 * by the OS.  The hardware is still under the driver's control, but
5026 * this netdev interface is disabled.
5027 *
5028 * Returns 0, this is not allowed to fail
5029 **/
5030#ifdef I40E_FCOE
5031int i40e_close(struct net_device *netdev)
5032#else
5033static int i40e_close(struct net_device *netdev)
5034#endif
5035{
5036	struct i40e_netdev_priv *np = netdev_priv(netdev);
5037	struct i40e_vsi *vsi = np->vsi;
5038
5039	i40e_vsi_close(vsi);
5040
5041	return 0;
5042}
5043
5044/**
5045 * i40e_do_reset - Start a PF or Core Reset sequence
5046 * @pf: board private structure
5047 * @reset_flags: which reset is requested
5048 *
5049 * The essential difference in resets is that the PF Reset
5050 * doesn't clear the packet buffers, doesn't reset the PE
5051 * firmware, and doesn't bother the other PFs on the chip.
5052 **/
5053void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5054{
5055	u32 val;
5056
5057	WARN_ON(in_interrupt());
5058
5059	if (i40e_check_asq_alive(&pf->hw))
5060		i40e_vc_notify_reset(pf);
5061
5062	/* do the biggest reset indicated */
5063	if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
5064
5065		/* Request a Global Reset
5066		 *
5067		 * This will start the chip's countdown to the actual full
5068		 * chip reset event, and a warning interrupt to be sent
5069		 * to all PFs, including the requestor.  Our handler
5070		 * for the warning interrupt will deal with the shutdown
5071		 * and recovery of the switch setup.
5072		 */
5073		dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5074		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5075		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5076		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5077
5078	} else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
5079
5080		/* Request a Core Reset
5081		 *
5082		 * Same as Global Reset, except does *not* include the MAC/PHY
5083		 */
5084		dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5085		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5086		val |= I40E_GLGEN_RTRIG_CORER_MASK;
5087		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5088		i40e_flush(&pf->hw);
5089
5090	} else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
5091
5092		/* Request a PF Reset
5093		 *
5094		 * Resets only the PF-specific registers
5095		 *
5096		 * This goes directly to the tear-down and rebuild of
5097		 * the switch, since we need to do all the recovery as
5098		 * for the Core Reset.
5099		 */
5100		dev_dbg(&pf->pdev->dev, "PFR requested\n");
5101		i40e_handle_reset_warning(pf);
5102
5103	} else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
5104		int v;
5105
5106		/* Find the VSI(s) that requested a re-init */
5107		dev_info(&pf->pdev->dev,
5108			 "VSI reinit requested\n");
5109		for (v = 0; v < pf->num_alloc_vsi; v++) {
5110			struct i40e_vsi *vsi = pf->vsi[v];
5111			if (vsi != NULL &&
5112			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5113				i40e_vsi_reinit_locked(pf->vsi[v]);
5114				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5115			}
5116		}
5117
5118		/* no further action needed, so return now */
5119		return;
5120	} else if (reset_flags & (1 << __I40E_DOWN_REQUESTED)) {
5121		int v;
5122
5123		/* Find the VSI(s) that needs to be brought down */
5124		dev_info(&pf->pdev->dev, "VSI down requested\n");
5125		for (v = 0; v < pf->num_alloc_vsi; v++) {
5126			struct i40e_vsi *vsi = pf->vsi[v];
5127			if (vsi != NULL &&
5128			    test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5129				set_bit(__I40E_DOWN, &vsi->state);
5130				i40e_down(vsi);
5131				clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5132			}
5133		}
5134
5135		/* no further action needed, so return now */
5136		return;
5137	} else {
5138		dev_info(&pf->pdev->dev,
5139			 "bad reset request 0x%08x\n", reset_flags);
5140		return;
5141	}
5142}
5143
5144#ifdef CONFIG_I40E_DCB
5145/**
5146 * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5147 * @pf: board private structure
5148 * @old_cfg: current DCB config
5149 * @new_cfg: new DCB config
5150 **/
5151bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5152			    struct i40e_dcbx_config *old_cfg,
5153			    struct i40e_dcbx_config *new_cfg)
5154{
5155	bool need_reconfig = false;
5156
5157	/* Check if ETS configuration has changed */
5158	if (memcmp(&new_cfg->etscfg,
5159		   &old_cfg->etscfg,
5160		   sizeof(new_cfg->etscfg))) {
5161		/* If Priority Table has changed reconfig is needed */
5162		if (memcmp(&new_cfg->etscfg.prioritytable,
5163			   &old_cfg->etscfg.prioritytable,
5164			   sizeof(new_cfg->etscfg.prioritytable))) {
5165			need_reconfig = true;
5166			dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5167		}
5168
5169		if (memcmp(&new_cfg->etscfg.tcbwtable,
5170			   &old_cfg->etscfg.tcbwtable,
5171			   sizeof(new_cfg->etscfg.tcbwtable)))
5172			dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5173
5174		if (memcmp(&new_cfg->etscfg.tsatable,
5175			   &old_cfg->etscfg.tsatable,
5176			   sizeof(new_cfg->etscfg.tsatable)))
5177			dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5178	}
5179
5180	/* Check if PFC configuration has changed */
5181	if (memcmp(&new_cfg->pfc,
5182		   &old_cfg->pfc,
5183		   sizeof(new_cfg->pfc))) {
5184		need_reconfig = true;
5185		dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5186	}
5187
5188	/* Check if APP Table has changed */
5189	if (memcmp(&new_cfg->app,
5190		   &old_cfg->app,
5191		   sizeof(new_cfg->app))) {
5192		need_reconfig = true;
5193		dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5194	}
5195
5196	dev_dbg(&pf->pdev->dev, "%s: need_reconfig=%d\n", __func__,
5197		need_reconfig);
5198	return need_reconfig;
5199}
5200
5201/**
5202 * i40e_handle_lldp_event - Handle LLDP Change MIB event
5203 * @pf: board private structure
5204 * @e: event info posted on ARQ
5205 **/
5206static int i40e_handle_lldp_event(struct i40e_pf *pf,
5207				  struct i40e_arq_event_info *e)
5208{
5209	struct i40e_aqc_lldp_get_mib *mib =
5210		(struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5211	struct i40e_hw *hw = &pf->hw;
5212	struct i40e_dcbx_config tmp_dcbx_cfg;
5213	bool need_reconfig = false;
5214	int ret = 0;
5215	u8 type;
5216
5217	/* Not DCB capable or capability disabled */
5218	if (!(pf->flags & I40E_FLAG_DCB_CAPABLE))
5219		return ret;
5220
5221	/* Ignore if event is not for Nearest Bridge */
5222	type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5223		& I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5224	dev_dbg(&pf->pdev->dev,
5225		"%s: LLDP event mib bridge type 0x%x\n", __func__, type);
5226	if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5227		return ret;
5228
5229	/* Check MIB Type and return if event for Remote MIB update */
5230	type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5231	dev_dbg(&pf->pdev->dev,
5232		"%s: LLDP event mib type %s\n", __func__,
5233		type ? "remote" : "local");
5234	if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5235		/* Update the remote cached instance and return */
5236		ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5237				I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5238				&hw->remote_dcbx_config);
5239		goto exit;
5240	}
5241
5242	/* Store the old configuration */
5243	tmp_dcbx_cfg = hw->local_dcbx_config;
5244
5245	/* Reset the old DCBx configuration data */
5246	memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5247	/* Get updated DCBX data from firmware */
5248	ret = i40e_get_dcb_config(&pf->hw);
5249	if (ret) {
5250		dev_info(&pf->pdev->dev, "Failed querying DCB configuration data from firmware.\n");
5251		goto exit;
5252	}
5253
5254	/* No change detected in DCBX configs */
5255	if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5256		    sizeof(tmp_dcbx_cfg))) {
5257		dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5258		goto exit;
5259	}
5260
5261	need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5262					       &hw->local_dcbx_config);
5263
5264	i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5265
5266	if (!need_reconfig)
5267		goto exit;
5268
5269	/* Enable DCB tagging only when more than one TC */
5270	if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5271		pf->flags |= I40E_FLAG_DCB_ENABLED;
5272	else
5273		pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5274
5275	set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5276	/* Reconfiguration needed quiesce all VSIs */
5277	i40e_pf_quiesce_all_vsi(pf);
5278
5279	/* Changes in configuration update VEB/VSI */
5280	i40e_dcb_reconfigure(pf);
5281
5282	ret = i40e_resume_port_tx(pf);
5283
5284	clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5285	/* In case of error no point in resuming VSIs */
5286	if (ret)
5287		goto exit;
5288
5289	/* Wait for the PF's Tx queues to be disabled */
5290	ret = i40e_pf_wait_txq_disabled(pf);
5291	if (ret) {
5292		/* Schedule PF reset to recover */
5293		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5294		i40e_service_event_schedule(pf);
5295	} else {
5296		i40e_pf_unquiesce_all_vsi(pf);
5297	}
5298
5299exit:
5300	return ret;
5301}
5302#endif /* CONFIG_I40E_DCB */
5303
5304/**
5305 * i40e_do_reset_safe - Protected reset path for userland calls.
5306 * @pf: board private structure
5307 * @reset_flags: which reset is requested
5308 *
5309 **/
5310void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5311{
5312	rtnl_lock();
5313	i40e_do_reset(pf, reset_flags);
5314	rtnl_unlock();
5315}
5316
5317/**
5318 * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5319 * @pf: board private structure
5320 * @e: event info posted on ARQ
5321 *
5322 * Handler for LAN Queue Overflow Event generated by the firmware for PF
5323 * and VF queues
5324 **/
5325static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5326					   struct i40e_arq_event_info *e)
5327{
5328	struct i40e_aqc_lan_overflow *data =
5329		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5330	u32 queue = le32_to_cpu(data->prtdcb_rupto);
5331	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5332	struct i40e_hw *hw = &pf->hw;
5333	struct i40e_vf *vf;
5334	u16 vf_id;
5335
5336	dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5337		queue, qtx_ctl);
5338
5339	/* Queue belongs to VF, find the VF and issue VF reset */
5340	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5341	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5342		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5343			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5344		vf_id -= hw->func_caps.vf_base_id;
5345		vf = &pf->vf[vf_id];
5346		i40e_vc_notify_vf_reset(vf);
5347		/* Allow VF to process pending reset notification */
5348		msleep(20);
5349		i40e_reset_vf(vf, false);
5350	}
5351}
5352
5353/**
5354 * i40e_service_event_complete - Finish up the service event
5355 * @pf: board private structure
5356 **/
5357static void i40e_service_event_complete(struct i40e_pf *pf)
5358{
5359	BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5360
5361	/* flush memory to make sure state is correct before next watchog */
5362	smp_mb__before_atomic();
5363	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5364}
5365
5366/**
5367 * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5368 * @pf: board private structure
5369 **/
5370u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5371{
5372	u32 val, fcnt_prog;
5373
5374	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5375	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5376	return fcnt_prog;
5377}
5378
5379/**
5380 * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5381 * @pf: board private structure
5382 **/
5383u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5384{
5385	u32 val, fcnt_prog;
5386
5387	val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5388	fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5389		    ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5390		      I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5391	return fcnt_prog;
5392}
5393
5394/**
5395 * i40e_get_global_fd_count - Get total FD filters programmed on device
5396 * @pf: board private structure
5397 **/
5398u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5399{
5400	u32 val, fcnt_prog;
5401
5402	val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5403	fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5404		    ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5405		     I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5406	return fcnt_prog;
5407}
5408
5409/**
5410 * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5411 * @pf: board private structure
5412 **/
5413void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5414{
5415	u32 fcnt_prog, fcnt_avail;
5416
5417	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5418		return;
5419
5420	/* Check if, FD SB or ATR was auto disabled and if there is enough room
5421	 * to re-enable
5422	 */
5423	fcnt_prog = i40e_get_global_fd_count(pf);
5424	fcnt_avail = pf->fdir_pf_filter_count;
5425	if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5426	    (pf->fd_add_err == 0) ||
5427	    (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5428		if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5429		    (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5430			pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5431			dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5432		}
5433	}
5434	/* Wait for some more space to be available to turn on ATR */
5435	if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5436		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5437		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5438			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5439			dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5440		}
5441	}
5442}
5443
5444#define I40E_MIN_FD_FLUSH_INTERVAL 10
5445#define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5446/**
5447 * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5448 * @pf: board private structure
5449 **/
5450static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5451{
5452	unsigned long min_flush_time;
5453	int flush_wait_retry = 50;
5454	bool disable_atr = false;
5455	int fd_room;
5456	int reg;
5457
5458	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5459		return;
5460
5461	if (time_after(jiffies, pf->fd_flush_timestamp +
5462				(I40E_MIN_FD_FLUSH_INTERVAL * HZ))) {
5463		/* If the flush is happening too quick and we have mostly
5464		 * SB rules we should not re-enable ATR for some time.
5465		 */
5466		min_flush_time = pf->fd_flush_timestamp
5467				+ (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5468		fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5469
5470		if (!(time_after(jiffies, min_flush_time)) &&
5471		    (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5472			dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5473			disable_atr = true;
5474		}
5475
5476		pf->fd_flush_timestamp = jiffies;
5477		pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5478		/* flush all filters */
5479		wr32(&pf->hw, I40E_PFQF_CTL_1,
5480		     I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
5481		i40e_flush(&pf->hw);
5482		pf->fd_flush_cnt++;
5483		pf->fd_add_err = 0;
5484		do {
5485			/* Check FD flush status every 5-6msec */
5486			usleep_range(5000, 6000);
5487			reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
5488			if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
5489				break;
5490		} while (flush_wait_retry--);
5491		if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
5492			dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
5493		} else {
5494			/* replay sideband filters */
5495			i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
5496			if (!disable_atr)
5497				pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
5498			clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
5499			dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
5500		}
5501	}
5502}
5503
5504/**
5505 * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
5506 * @pf: board private structure
5507 **/
5508u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
5509{
5510	return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
5511}
5512
5513/* We can see up to 256 filter programming desc in transit if the filters are
5514 * being applied really fast; before we see the first
5515 * filter miss error on Rx queue 0. Accumulating enough error messages before
5516 * reacting will make sure we don't cause flush too often.
5517 */
5518#define I40E_MAX_FD_PROGRAM_ERROR 256
5519
5520/**
5521 * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
5522 * @pf: board private structure
5523 **/
5524static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
5525{
5526
5527	/* if interface is down do nothing */
5528	if (test_bit(__I40E_DOWN, &pf->state))
5529		return;
5530
5531	if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5532		return;
5533
5534	if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5535		i40e_fdir_flush_and_replay(pf);
5536
5537	i40e_fdir_check_and_reenable(pf);
5538
5539}
5540
5541/**
5542 * i40e_vsi_link_event - notify VSI of a link event
5543 * @vsi: vsi to be notified
5544 * @link_up: link up or down
5545 **/
5546static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
5547{
5548	if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
5549		return;
5550
5551	switch (vsi->type) {
5552	case I40E_VSI_MAIN:
5553#ifdef I40E_FCOE
5554	case I40E_VSI_FCOE:
5555#endif
5556		if (!vsi->netdev || !vsi->netdev_registered)
5557			break;
5558
5559		if (link_up) {
5560			netif_carrier_on(vsi->netdev);
5561			netif_tx_wake_all_queues(vsi->netdev);
5562		} else {
5563			netif_carrier_off(vsi->netdev);
5564			netif_tx_stop_all_queues(vsi->netdev);
5565		}
5566		break;
5567
5568	case I40E_VSI_SRIOV:
5569	case I40E_VSI_VMDQ2:
5570	case I40E_VSI_CTRL:
5571	case I40E_VSI_MIRROR:
5572	default:
5573		/* there is no notification for other VSIs */
5574		break;
5575	}
5576}
5577
5578/**
5579 * i40e_veb_link_event - notify elements on the veb of a link event
5580 * @veb: veb to be notified
5581 * @link_up: link up or down
5582 **/
5583static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
5584{
5585	struct i40e_pf *pf;
5586	int i;
5587
5588	if (!veb || !veb->pf)
5589		return;
5590	pf = veb->pf;
5591
5592	/* depth first... */
5593	for (i = 0; i < I40E_MAX_VEB; i++)
5594		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
5595			i40e_veb_link_event(pf->veb[i], link_up);
5596
5597	/* ... now the local VSIs */
5598	for (i = 0; i < pf->num_alloc_vsi; i++)
5599		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
5600			i40e_vsi_link_event(pf->vsi[i], link_up);
5601}
5602
5603/**
5604 * i40e_link_event - Update netif_carrier status
5605 * @pf: board private structure
5606 **/
5607static void i40e_link_event(struct i40e_pf *pf)
5608{
5609	bool new_link, old_link;
5610	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5611	u8 new_link_speed, old_link_speed;
5612
5613	/* set this to force the get_link_status call to refresh state */
5614	pf->hw.phy.get_link_info = true;
5615
5616	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
5617	new_link = i40e_get_link_status(&pf->hw);
5618	old_link_speed = pf->hw.phy.link_info_old.link_speed;
5619	new_link_speed = pf->hw.phy.link_info.link_speed;
5620
5621	if (new_link == old_link &&
5622	    new_link_speed == old_link_speed &&
5623	    (test_bit(__I40E_DOWN, &vsi->state) ||
5624	     new_link == netif_carrier_ok(vsi->netdev)))
5625		return;
5626
5627	if (!test_bit(__I40E_DOWN, &vsi->state))
5628		i40e_print_link_message(vsi, new_link);
5629
5630	/* Notify the base of the switch tree connected to
5631	 * the link.  Floating VEBs are not notified.
5632	 */
5633	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
5634		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
5635	else
5636		i40e_vsi_link_event(vsi, new_link);
5637
5638	if (pf->vf)
5639		i40e_vc_notify_link_state(pf);
5640
5641	if (pf->flags & I40E_FLAG_PTP)
5642		i40e_ptp_set_increment(pf);
5643}
5644
5645/**
5646 * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
5647 * @pf: board private structure
5648 *
5649 * Set the per-queue flags to request a check for stuck queues in the irq
5650 * clean functions, then force interrupts to be sure the irq clean is called.
5651 **/
5652static void i40e_check_hang_subtask(struct i40e_pf *pf)
5653{
5654	int i, v;
5655
5656	/* If we're down or resetting, just bail */
5657	if (test_bit(__I40E_DOWN, &pf->state) ||
5658	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
5659		return;
5660
5661	/* for each VSI/netdev
5662	 *     for each Tx queue
5663	 *         set the check flag
5664	 *     for each q_vector
5665	 *         force an interrupt
5666	 */
5667	for (v = 0; v < pf->num_alloc_vsi; v++) {
5668		struct i40e_vsi *vsi = pf->vsi[v];
5669		int armed = 0;
5670
5671		if (!pf->vsi[v] ||
5672		    test_bit(__I40E_DOWN, &vsi->state) ||
5673		    (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
5674			continue;
5675
5676		for (i = 0; i < vsi->num_queue_pairs; i++) {
5677			set_check_for_tx_hang(vsi->tx_rings[i]);
5678			if (test_bit(__I40E_HANG_CHECK_ARMED,
5679				     &vsi->tx_rings[i]->state))
5680				armed++;
5681		}
5682
5683		if (armed) {
5684			if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
5685				wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
5686				     (I40E_PFINT_DYN_CTL0_INTENA_MASK |
5687				      I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK |
5688				      I40E_PFINT_DYN_CTL0_ITR_INDX_MASK |
5689				      I40E_PFINT_DYN_CTL0_SW_ITR_INDX_ENA_MASK |
5690				      I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK));
5691			} else {
5692				u16 vec = vsi->base_vector - 1;
5693				u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
5694				      I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK |
5695				      I40E_PFINT_DYN_CTLN_ITR_INDX_MASK |
5696				      I40E_PFINT_DYN_CTLN_SW_ITR_INDX_ENA_MASK |
5697				      I40E_PFINT_DYN_CTLN_SW_ITR_INDX_MASK);
5698				for (i = 0; i < vsi->num_q_vectors; i++, vec++)
5699					wr32(&vsi->back->hw,
5700					     I40E_PFINT_DYN_CTLN(vec), val);
5701			}
5702			i40e_flush(&vsi->back->hw);
5703		}
5704	}
5705}
5706
5707/**
5708 * i40e_watchdog_subtask - periodic checks not using event driven response
5709 * @pf: board private structure
5710 **/
5711static void i40e_watchdog_subtask(struct i40e_pf *pf)
5712{
5713	int i;
5714
5715	/* if interface is down do nothing */
5716	if (test_bit(__I40E_DOWN, &pf->state) ||
5717	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
5718		return;
5719
5720	/* make sure we don't do these things too often */
5721	if (time_before(jiffies, (pf->service_timer_previous +
5722				  pf->service_timer_period)))
5723		return;
5724	pf->service_timer_previous = jiffies;
5725
5726	i40e_check_hang_subtask(pf);
5727	i40e_link_event(pf);
5728
5729	/* Update the stats for active netdevs so the network stack
5730	 * can look at updated numbers whenever it cares to
5731	 */
5732	for (i = 0; i < pf->num_alloc_vsi; i++)
5733		if (pf->vsi[i] && pf->vsi[i]->netdev)
5734			i40e_update_stats(pf->vsi[i]);
5735
5736	/* Update the stats for the active switching components */
5737	for (i = 0; i < I40E_MAX_VEB; i++)
5738		if (pf->veb[i])
5739			i40e_update_veb_stats(pf->veb[i]);
5740
5741	i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
5742}
5743
5744/**
5745 * i40e_reset_subtask - Set up for resetting the device and driver
5746 * @pf: board private structure
5747 **/
5748static void i40e_reset_subtask(struct i40e_pf *pf)
5749{
5750	u32 reset_flags = 0;
5751
5752	rtnl_lock();
5753	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
5754		reset_flags |= (1 << __I40E_REINIT_REQUESTED);
5755		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
5756	}
5757	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
5758		reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
5759		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5760	}
5761	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
5762		reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
5763		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
5764	}
5765	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
5766		reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
5767		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
5768	}
5769	if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
5770		reset_flags |= (1 << __I40E_DOWN_REQUESTED);
5771		clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
5772	}
5773
5774	/* If there's a recovery already waiting, it takes
5775	 * precedence before starting a new reset sequence.
5776	 */
5777	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
5778		i40e_handle_reset_warning(pf);
5779		goto unlock;
5780	}
5781
5782	/* If we're already down or resetting, just bail */
5783	if (reset_flags &&
5784	    !test_bit(__I40E_DOWN, &pf->state) &&
5785	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
5786		i40e_do_reset(pf, reset_flags);
5787
5788unlock:
5789	rtnl_unlock();
5790}
5791
5792/**
5793 * i40e_handle_link_event - Handle link event
5794 * @pf: board private structure
5795 * @e: event info posted on ARQ
5796 **/
5797static void i40e_handle_link_event(struct i40e_pf *pf,
5798				   struct i40e_arq_event_info *e)
5799{
5800	struct i40e_hw *hw = &pf->hw;
5801	struct i40e_aqc_get_link_status *status =
5802		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
5803
5804	/* save off old link status information */
5805	hw->phy.link_info_old = hw->phy.link_info;
5806
5807	/* Do a new status request to re-enable LSE reporting
5808	 * and load new status information into the hw struct
5809	 * This completely ignores any state information
5810	 * in the ARQ event info, instead choosing to always
5811	 * issue the AQ update link status command.
5812	 */
5813	i40e_link_event(pf);
5814
5815	/* check for unqualified module, if link is down */
5816	if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
5817	    (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
5818	    (!(status->link_info & I40E_AQ_LINK_UP)))
5819		dev_err(&pf->pdev->dev,
5820			"The driver failed to link because an unqualified module was detected.\n");
5821}
5822
5823/**
5824 * i40e_clean_adminq_subtask - Clean the AdminQ rings
5825 * @pf: board private structure
5826 **/
5827static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
5828{
5829	struct i40e_arq_event_info event;
5830	struct i40e_hw *hw = &pf->hw;
5831	u16 pending, i = 0;
5832	i40e_status ret;
5833	u16 opcode;
5834	u32 oldval;
5835	u32 val;
5836
5837	/* Do not run clean AQ when PF reset fails */
5838	if (test_bit(__I40E_RESET_FAILED, &pf->state))
5839		return;
5840
5841	/* check for error indications */
5842	val = rd32(&pf->hw, pf->hw.aq.arq.len);
5843	oldval = val;
5844	if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
5845		dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
5846		val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
5847	}
5848	if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
5849		dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
5850		val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
5851	}
5852	if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
5853		dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
5854		val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
5855	}
5856	if (oldval != val)
5857		wr32(&pf->hw, pf->hw.aq.arq.len, val);
5858
5859	val = rd32(&pf->hw, pf->hw.aq.asq.len);
5860	oldval = val;
5861	if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
5862		dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
5863		val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
5864	}
5865	if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
5866		dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
5867		val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
5868	}
5869	if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
5870		dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
5871		val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
5872	}
5873	if (oldval != val)
5874		wr32(&pf->hw, pf->hw.aq.asq.len, val);
5875
5876	event.buf_len = I40E_MAX_AQ_BUF_SIZE;
5877	event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
5878	if (!event.msg_buf)
5879		return;
5880
5881	do {
5882		ret = i40e_clean_arq_element(hw, &event, &pending);
5883		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
5884			break;
5885		else if (ret) {
5886			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
5887			break;
5888		}
5889
5890		opcode = le16_to_cpu(event.desc.opcode);
5891		switch (opcode) {
5892
5893		case i40e_aqc_opc_get_link_status:
5894			i40e_handle_link_event(pf, &event);
5895			break;
5896		case i40e_aqc_opc_send_msg_to_pf:
5897			ret = i40e_vc_process_vf_msg(pf,
5898					le16_to_cpu(event.desc.retval),
5899					le32_to_cpu(event.desc.cookie_high),
5900					le32_to_cpu(event.desc.cookie_low),
5901					event.msg_buf,
5902					event.msg_len);
5903			break;
5904		case i40e_aqc_opc_lldp_update_mib:
5905			dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
5906#ifdef CONFIG_I40E_DCB
5907			rtnl_lock();
5908			ret = i40e_handle_lldp_event(pf, &event);
5909			rtnl_unlock();
5910#endif /* CONFIG_I40E_DCB */
5911			break;
5912		case i40e_aqc_opc_event_lan_overflow:
5913			dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
5914			i40e_handle_lan_overflow_event(pf, &event);
5915			break;
5916		case i40e_aqc_opc_send_msg_to_peer:
5917			dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
5918			break;
5919		case i40e_aqc_opc_nvm_erase:
5920		case i40e_aqc_opc_nvm_update:
5921			i40e_debug(&pf->hw, I40E_DEBUG_NVM, "ARQ NVM operation completed\n");
5922			break;
5923		default:
5924			dev_info(&pf->pdev->dev,
5925				 "ARQ Error: Unknown event 0x%04x received\n",
5926				 opcode);
5927			break;
5928		}
5929	} while (pending && (i++ < pf->adminq_work_limit));
5930
5931	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
5932	/* re-enable Admin queue interrupt cause */
5933	val = rd32(hw, I40E_PFINT_ICR0_ENA);
5934	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
5935	wr32(hw, I40E_PFINT_ICR0_ENA, val);
5936	i40e_flush(hw);
5937
5938	kfree(event.msg_buf);
5939}
5940
5941/**
5942 * i40e_verify_eeprom - make sure eeprom is good to use
5943 * @pf: board private structure
5944 **/
5945static void i40e_verify_eeprom(struct i40e_pf *pf)
5946{
5947	int err;
5948
5949	err = i40e_diag_eeprom_test(&pf->hw);
5950	if (err) {
5951		/* retry in case of garbage read */
5952		err = i40e_diag_eeprom_test(&pf->hw);
5953		if (err) {
5954			dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
5955				 err);
5956			set_bit(__I40E_BAD_EEPROM, &pf->state);
5957		}
5958	}
5959
5960	if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
5961		dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
5962		clear_bit(__I40E_BAD_EEPROM, &pf->state);
5963	}
5964}
5965
5966/**
5967 * i40e_enable_pf_switch_lb
5968 * @pf: pointer to the PF structure
5969 *
5970 * enable switch loop back or die - no point in a return value
5971 **/
5972static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
5973{
5974	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
5975	struct i40e_vsi_context ctxt;
5976	int aq_ret;
5977
5978	ctxt.seid = pf->main_vsi_seid;
5979	ctxt.pf_num = pf->hw.pf_id;
5980	ctxt.vf_num = 0;
5981	aq_ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
5982	if (aq_ret) {
5983		dev_info(&pf->pdev->dev,
5984			 "%s couldn't get PF vsi config, err %d, aq_err %d\n",
5985			 __func__, aq_ret, pf->hw.aq.asq_last_status);
5986		return;
5987	}
5988	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
5989	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5990	ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5991
5992	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
5993	if (aq_ret) {
5994		dev_info(&pf->pdev->dev,
5995			 "%s: update vsi switch failed, aq_err=%d\n",
5996			 __func__, vsi->back->hw.aq.asq_last_status);
5997	}
5998}
5999
6000/**
6001 * i40e_disable_pf_switch_lb
6002 * @pf: pointer to the PF structure
6003 *
6004 * disable switch loop back or die - no point in a return value
6005 **/
6006static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6007{
6008	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6009	struct i40e_vsi_context ctxt;
6010	int aq_ret;
6011
6012	ctxt.seid = pf->main_vsi_seid;
6013	ctxt.pf_num = pf->hw.pf_id;
6014	ctxt.vf_num = 0;
6015	aq_ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6016	if (aq_ret) {
6017		dev_info(&pf->pdev->dev,
6018			 "%s couldn't get PF vsi config, err %d, aq_err %d\n",
6019			 __func__, aq_ret, pf->hw.aq.asq_last_status);
6020		return;
6021	}
6022	ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6023	ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6024	ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6025
6026	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6027	if (aq_ret) {
6028		dev_info(&pf->pdev->dev,
6029			 "%s: update vsi switch failed, aq_err=%d\n",
6030			 __func__, vsi->back->hw.aq.asq_last_status);
6031	}
6032}
6033
6034/**
6035 * i40e_config_bridge_mode - Configure the HW bridge mode
6036 * @veb: pointer to the bridge instance
6037 *
6038 * Configure the loop back mode for the LAN VSI that is downlink to the
6039 * specified HW bridge instance. It is expected this function is called
6040 * when a new HW bridge is instantiated.
6041 **/
6042static void i40e_config_bridge_mode(struct i40e_veb *veb)
6043{
6044	struct i40e_pf *pf = veb->pf;
6045
6046	dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6047		 veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6048	if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6049		i40e_disable_pf_switch_lb(pf);
6050	else
6051		i40e_enable_pf_switch_lb(pf);
6052}
6053
6054/**
6055 * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6056 * @veb: pointer to the VEB instance
6057 *
6058 * This is a recursive function that first builds the attached VSIs then
6059 * recurses in to build the next layer of VEB.  We track the connections
6060 * through our own index numbers because the seid's from the HW could
6061 * change across the reset.
6062 **/
6063static int i40e_reconstitute_veb(struct i40e_veb *veb)
6064{
6065	struct i40e_vsi *ctl_vsi = NULL;
6066	struct i40e_pf *pf = veb->pf;
6067	int v, veb_idx;
6068	int ret;
6069
6070	/* build VSI that owns this VEB, temporarily attached to base VEB */
6071	for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6072		if (pf->vsi[v] &&
6073		    pf->vsi[v]->veb_idx == veb->idx &&
6074		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6075			ctl_vsi = pf->vsi[v];
6076			break;
6077		}
6078	}
6079	if (!ctl_vsi) {
6080		dev_info(&pf->pdev->dev,
6081			 "missing owner VSI for veb_idx %d\n", veb->idx);
6082		ret = -ENOENT;
6083		goto end_reconstitute;
6084	}
6085	if (ctl_vsi != pf->vsi[pf->lan_vsi])
6086		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6087	ret = i40e_add_vsi(ctl_vsi);
6088	if (ret) {
6089		dev_info(&pf->pdev->dev,
6090			 "rebuild of owner VSI failed: %d\n", ret);
6091		goto end_reconstitute;
6092	}
6093	i40e_vsi_reset_stats(ctl_vsi);
6094
6095	/* create the VEB in the switch and move the VSI onto the VEB */
6096	ret = i40e_add_veb(veb, ctl_vsi);
6097	if (ret)
6098		goto end_reconstitute;
6099
6100	if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6101		veb->bridge_mode = BRIDGE_MODE_VEB;
6102	else
6103		veb->bridge_mode = BRIDGE_MODE_VEPA;
6104	i40e_config_bridge_mode(veb);
6105
6106	/* create the remaining VSIs attached to this VEB */
6107	for (v = 0; v < pf->num_alloc_vsi; v++) {
6108		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6109			continue;
6110
6111		if (pf->vsi[v]->veb_idx == veb->idx) {
6112			struct i40e_vsi *vsi = pf->vsi[v];
6113			vsi->uplink_seid = veb->seid;
6114			ret = i40e_add_vsi(vsi);
6115			if (ret) {
6116				dev_info(&pf->pdev->dev,
6117					 "rebuild of vsi_idx %d failed: %d\n",
6118					 v, ret);
6119				goto end_reconstitute;
6120			}
6121			i40e_vsi_reset_stats(vsi);
6122		}
6123	}
6124
6125	/* create any VEBs attached to this VEB - RECURSION */
6126	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6127		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6128			pf->veb[veb_idx]->uplink_seid = veb->seid;
6129			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6130			if (ret)
6131				break;
6132		}
6133	}
6134
6135end_reconstitute:
6136	return ret;
6137}
6138
6139/**
6140 * i40e_get_capabilities - get info about the HW
6141 * @pf: the PF struct
6142 **/
6143static int i40e_get_capabilities(struct i40e_pf *pf)
6144{
6145	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6146	u16 data_size;
6147	int buf_len;
6148	int err;
6149
6150	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6151	do {
6152		cap_buf = kzalloc(buf_len, GFP_KERNEL);
6153		if (!cap_buf)
6154			return -ENOMEM;
6155
6156		/* this loads the data into the hw struct for us */
6157		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6158					    &data_size,
6159					    i40e_aqc_opc_list_func_capabilities,
6160					    NULL);
6161		/* data loaded, buffer no longer needed */
6162		kfree(cap_buf);
6163
6164		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6165			/* retry with a larger buffer */
6166			buf_len = data_size;
6167		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6168			dev_info(&pf->pdev->dev,
6169				 "capability discovery failed: aq=%d\n",
6170				 pf->hw.aq.asq_last_status);
6171			return -ENODEV;
6172		}
6173	} while (err);
6174
6175	if (((pf->hw.aq.fw_maj_ver == 2) && (pf->hw.aq.fw_min_ver < 22)) ||
6176	    (pf->hw.aq.fw_maj_ver < 2)) {
6177		pf->hw.func_caps.num_msix_vectors++;
6178		pf->hw.func_caps.num_msix_vectors_vf++;
6179	}
6180
6181	if (pf->hw.debug_mask & I40E_DEBUG_USER)
6182		dev_info(&pf->pdev->dev,
6183			 "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",
6184			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6185			 pf->hw.func_caps.num_msix_vectors,
6186			 pf->hw.func_caps.num_msix_vectors_vf,
6187			 pf->hw.func_caps.fd_filters_guaranteed,
6188			 pf->hw.func_caps.fd_filters_best_effort,
6189			 pf->hw.func_caps.num_tx_qp,
6190			 pf->hw.func_caps.num_vsis);
6191
6192#define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6193		       + pf->hw.func_caps.num_vfs)
6194	if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6195		dev_info(&pf->pdev->dev,
6196			 "got num_vsis %d, setting num_vsis to %d\n",
6197			 pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6198		pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6199	}
6200
6201	return 0;
6202}
6203
6204static int i40e_vsi_clear(struct i40e_vsi *vsi);
6205
6206/**
6207 * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6208 * @pf: board private structure
6209 **/
6210static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6211{
6212	struct i40e_vsi *vsi;
6213	int i;
6214
6215	/* quick workaround for an NVM issue that leaves a critical register
6216	 * uninitialized
6217	 */
6218	if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6219		static const u32 hkey[] = {
6220			0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6221			0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6222			0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6223			0x95b3a76d};
6224
6225		for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6226			wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6227	}
6228
6229	if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6230		return;
6231
6232	/* find existing VSI and see if it needs configuring */
6233	vsi = NULL;
6234	for (i = 0; i < pf->num_alloc_vsi; i++) {
6235		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6236			vsi = pf->vsi[i];
6237			break;
6238		}
6239	}
6240
6241	/* create a new VSI if none exists */
6242	if (!vsi) {
6243		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6244				     pf->vsi[pf->lan_vsi]->seid, 0);
6245		if (!vsi) {
6246			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6247			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6248			return;
6249		}
6250	}
6251
6252	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6253}
6254
6255/**
6256 * i40e_fdir_teardown - release the Flow Director resources
6257 * @pf: board private structure
6258 **/
6259static void i40e_fdir_teardown(struct i40e_pf *pf)
6260{
6261	int i;
6262
6263	i40e_fdir_filter_exit(pf);
6264	for (i = 0; i < pf->num_alloc_vsi; i++) {
6265		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6266			i40e_vsi_release(pf->vsi[i]);
6267			break;
6268		}
6269	}
6270}
6271
6272/**
6273 * i40e_prep_for_reset - prep for the core to reset
6274 * @pf: board private structure
6275 *
6276 * Close up the VFs and other things in prep for PF Reset.
6277  **/
6278static void i40e_prep_for_reset(struct i40e_pf *pf)
6279{
6280	struct i40e_hw *hw = &pf->hw;
6281	i40e_status ret = 0;
6282	u32 v;
6283
6284	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6285	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6286		return;
6287
6288	dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6289
6290	/* quiesce the VSIs and their queues that are not already DOWN */
6291	i40e_pf_quiesce_all_vsi(pf);
6292
6293	for (v = 0; v < pf->num_alloc_vsi; v++) {
6294		if (pf->vsi[v])
6295			pf->vsi[v]->seid = 0;
6296	}
6297
6298	i40e_shutdown_adminq(&pf->hw);
6299
6300	/* call shutdown HMC */
6301	if (hw->hmc.hmc_obj) {
6302		ret = i40e_shutdown_lan_hmc(hw);
6303		if (ret)
6304			dev_warn(&pf->pdev->dev,
6305				 "shutdown_lan_hmc failed: %d\n", ret);
6306	}
6307}
6308
6309/**
6310 * i40e_send_version - update firmware with driver version
6311 * @pf: PF struct
6312 */
6313static void i40e_send_version(struct i40e_pf *pf)
6314{
6315	struct i40e_driver_version dv;
6316
6317	dv.major_version = DRV_VERSION_MAJOR;
6318	dv.minor_version = DRV_VERSION_MINOR;
6319	dv.build_version = DRV_VERSION_BUILD;
6320	dv.subbuild_version = 0;
6321	strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6322	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6323}
6324
6325/**
6326 * i40e_reset_and_rebuild - reset and rebuild using a saved config
6327 * @pf: board private structure
6328 * @reinit: if the Main VSI needs to re-initialized.
6329 **/
6330static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6331{
6332	struct i40e_hw *hw = &pf->hw;
6333	u8 set_fc_aq_fail = 0;
6334	i40e_status ret;
6335	u32 v;
6336
6337	/* Now we wait for GRST to settle out.
6338	 * We don't have to delete the VEBs or VSIs from the hw switch
6339	 * because the reset will make them disappear.
6340	 */
6341	ret = i40e_pf_reset(hw);
6342	if (ret) {
6343		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6344		set_bit(__I40E_RESET_FAILED, &pf->state);
6345		goto clear_recovery;
6346	}
6347	pf->pfr_count++;
6348
6349	if (test_bit(__I40E_DOWN, &pf->state))
6350		goto clear_recovery;
6351	dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6352
6353	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6354	ret = i40e_init_adminq(&pf->hw);
6355	if (ret) {
6356		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
6357		goto clear_recovery;
6358	}
6359
6360	/* re-verify the eeprom if we just had an EMP reset */
6361	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6362		i40e_verify_eeprom(pf);
6363
6364	i40e_clear_pxe_mode(hw);
6365	ret = i40e_get_capabilities(pf);
6366	if (ret) {
6367		dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
6368			 ret);
6369		goto end_core_reset;
6370	}
6371
6372	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6373				hw->func_caps.num_rx_qp,
6374				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6375	if (ret) {
6376		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6377		goto end_core_reset;
6378	}
6379	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6380	if (ret) {
6381		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6382		goto end_core_reset;
6383	}
6384
6385#ifdef CONFIG_I40E_DCB
6386	ret = i40e_init_pf_dcb(pf);
6387	if (ret) {
6388		dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6389		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6390		/* Continue without DCB enabled */
6391	}
6392#endif /* CONFIG_I40E_DCB */
6393#ifdef I40E_FCOE
6394	ret = i40e_init_pf_fcoe(pf);
6395	if (ret)
6396		dev_info(&pf->pdev->dev, "init_pf_fcoe failed: %d\n", ret);
6397
6398#endif
6399	/* do basic switch setup */
6400	ret = i40e_setup_pf_switch(pf, reinit);
6401	if (ret)
6402		goto end_core_reset;
6403
6404	/* driver is only interested in link up/down and module qualification
6405	 * reports from firmware
6406	 */
6407	ret = i40e_aq_set_phy_int_mask(&pf->hw,
6408				       I40E_AQ_EVENT_LINK_UPDOWN |
6409				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
6410	if (ret)
6411		dev_info(&pf->pdev->dev, "set phy mask fail, aq_err %d\n", ret);
6412
6413	/* make sure our flow control settings are restored */
6414	ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6415	if (ret)
6416		dev_info(&pf->pdev->dev, "set fc fail, aq_err %d\n", ret);
6417
6418	/* Rebuild the VSIs and VEBs that existed before reset.
6419	 * They are still in our local switch element arrays, so only
6420	 * need to rebuild the switch model in the HW.
6421	 *
6422	 * If there were VEBs but the reconstitution failed, we'll try
6423	 * try to recover minimal use by getting the basic PF VSI working.
6424	 */
6425	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6426		dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6427		/* find the one VEB connected to the MAC, and find orphans */
6428		for (v = 0; v < I40E_MAX_VEB; v++) {
6429			if (!pf->veb[v])
6430				continue;
6431
6432			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6433			    pf->veb[v]->uplink_seid == 0) {
6434				ret = i40e_reconstitute_veb(pf->veb[v]);
6435
6436				if (!ret)
6437					continue;
6438
6439				/* If Main VEB failed, we're in deep doodoo,
6440				 * so give up rebuilding the switch and set up
6441				 * for minimal rebuild of PF VSI.
6442				 * If orphan failed, we'll report the error
6443				 * but try to keep going.
6444				 */
6445				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6446					dev_info(&pf->pdev->dev,
6447						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6448						 ret);
6449					pf->vsi[pf->lan_vsi]->uplink_seid
6450								= pf->mac_seid;
6451					break;
6452				} else if (pf->veb[v]->uplink_seid == 0) {
6453					dev_info(&pf->pdev->dev,
6454						 "rebuild of orphan VEB failed: %d\n",
6455						 ret);
6456				}
6457			}
6458		}
6459	}
6460
6461	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6462		dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6463		/* no VEB, so rebuild only the Main VSI */
6464		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6465		if (ret) {
6466			dev_info(&pf->pdev->dev,
6467				 "rebuild of Main VSI failed: %d\n", ret);
6468			goto end_core_reset;
6469		}
6470	}
6471
6472	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
6473	    (pf->hw.aq.fw_maj_ver < 4)) {
6474		msleep(75);
6475		ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6476		if (ret)
6477			dev_info(&pf->pdev->dev, "link restart failed, aq_err=%d\n",
6478				 pf->hw.aq.asq_last_status);
6479	}
6480	/* reinit the misc interrupt */
6481	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6482		ret = i40e_setup_misc_vector(pf);
6483
6484	/* restart the VSIs that were rebuilt and running before the reset */
6485	i40e_pf_unquiesce_all_vsi(pf);
6486
6487	if (pf->num_alloc_vfs) {
6488		for (v = 0; v < pf->num_alloc_vfs; v++)
6489			i40e_reset_vf(&pf->vf[v], true);
6490	}
6491
6492	/* tell the firmware that we're starting */
6493	i40e_send_version(pf);
6494
6495end_core_reset:
6496	clear_bit(__I40E_RESET_FAILED, &pf->state);
6497clear_recovery:
6498	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
6499}
6500
6501/**
6502 * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
6503 * @pf: board private structure
6504 *
6505 * Close up the VFs and other things in prep for a Core Reset,
6506 * then get ready to rebuild the world.
6507 **/
6508static void i40e_handle_reset_warning(struct i40e_pf *pf)
6509{
6510	i40e_prep_for_reset(pf);
6511	i40e_reset_and_rebuild(pf, false);
6512}
6513
6514/**
6515 * i40e_handle_mdd_event
6516 * @pf: pointer to the PF structure
6517 *
6518 * Called from the MDD irq handler to identify possibly malicious vfs
6519 **/
6520static void i40e_handle_mdd_event(struct i40e_pf *pf)
6521{
6522	struct i40e_hw *hw = &pf->hw;
6523	bool mdd_detected = false;
6524	bool pf_mdd_detected = false;
6525	struct i40e_vf *vf;
6526	u32 reg;
6527	int i;
6528
6529	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
6530		return;
6531
6532	/* find what triggered the MDD event */
6533	reg = rd32(hw, I40E_GL_MDET_TX);
6534	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
6535		u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
6536				I40E_GL_MDET_TX_PF_NUM_SHIFT;
6537		u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
6538				I40E_GL_MDET_TX_VF_NUM_SHIFT;
6539		u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
6540				I40E_GL_MDET_TX_EVENT_SHIFT;
6541		u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
6542				I40E_GL_MDET_TX_QUEUE_SHIFT) -
6543				pf->hw.func_caps.base_queue;
6544		if (netif_msg_tx_err(pf))
6545			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
6546				 event, queue, pf_num, vf_num);
6547		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
6548		mdd_detected = true;
6549	}
6550	reg = rd32(hw, I40E_GL_MDET_RX);
6551	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
6552		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
6553				I40E_GL_MDET_RX_FUNCTION_SHIFT;
6554		u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
6555				I40E_GL_MDET_RX_EVENT_SHIFT;
6556		u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
6557				I40E_GL_MDET_RX_QUEUE_SHIFT) -
6558				pf->hw.func_caps.base_queue;
6559		if (netif_msg_rx_err(pf))
6560			dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
6561				 event, queue, func);
6562		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
6563		mdd_detected = true;
6564	}
6565
6566	if (mdd_detected) {
6567		reg = rd32(hw, I40E_PF_MDET_TX);
6568		if (reg & I40E_PF_MDET_TX_VALID_MASK) {
6569			wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
6570			dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
6571			pf_mdd_detected = true;
6572		}
6573		reg = rd32(hw, I40E_PF_MDET_RX);
6574		if (reg & I40E_PF_MDET_RX_VALID_MASK) {
6575			wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
6576			dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
6577			pf_mdd_detected = true;
6578		}
6579		/* Queue belongs to the PF, initiate a reset */
6580		if (pf_mdd_detected) {
6581			set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6582			i40e_service_event_schedule(pf);
6583		}
6584	}
6585
6586	/* see if one of the VFs needs its hand slapped */
6587	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
6588		vf = &(pf->vf[i]);
6589		reg = rd32(hw, I40E_VP_MDET_TX(i));
6590		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
6591			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
6592			vf->num_mdd_events++;
6593			dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
6594				 i);
6595		}
6596
6597		reg = rd32(hw, I40E_VP_MDET_RX(i));
6598		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
6599			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
6600			vf->num_mdd_events++;
6601			dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
6602				 i);
6603		}
6604
6605		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
6606			dev_info(&pf->pdev->dev,
6607				 "Too many MDD events on VF %d, disabled\n", i);
6608			dev_info(&pf->pdev->dev,
6609				 "Use PF Control I/F to re-enable the VF\n");
6610			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
6611		}
6612	}
6613
6614	/* re-enable mdd interrupt cause */
6615	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
6616	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
6617	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
6618	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
6619	i40e_flush(hw);
6620}
6621
6622#ifdef CONFIG_I40E_VXLAN
6623/**
6624 * i40e_sync_vxlan_filters_subtask - Sync the VSI filter list with HW
6625 * @pf: board private structure
6626 **/
6627static void i40e_sync_vxlan_filters_subtask(struct i40e_pf *pf)
6628{
6629	struct i40e_hw *hw = &pf->hw;
6630	i40e_status ret;
6631	__be16 port;
6632	int i;
6633
6634	if (!(pf->flags & I40E_FLAG_VXLAN_FILTER_SYNC))
6635		return;
6636
6637	pf->flags &= ~I40E_FLAG_VXLAN_FILTER_SYNC;
6638
6639	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
6640		if (pf->pending_vxlan_bitmap & (1 << i)) {
6641			pf->pending_vxlan_bitmap &= ~(1 << i);
6642			port = pf->vxlan_ports[i];
6643			if (port)
6644				ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
6645						     I40E_AQC_TUNNEL_TYPE_VXLAN,
6646						     NULL, NULL);
6647			else
6648				ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
6649
6650			if (ret) {
6651				dev_info(&pf->pdev->dev,
6652					 "%s vxlan port %d, index %d failed, err %d, aq_err %d\n",
6653					 port ? "add" : "delete",
6654					 ntohs(port), i, ret,
6655					 pf->hw.aq.asq_last_status);
6656				pf->vxlan_ports[i] = 0;
6657			}
6658		}
6659	}
6660}
6661
6662#endif
6663/**
6664 * i40e_service_task - Run the driver's async subtasks
6665 * @work: pointer to work_struct containing our data
6666 **/
6667static void i40e_service_task(struct work_struct *work)
6668{
6669	struct i40e_pf *pf = container_of(work,
6670					  struct i40e_pf,
6671					  service_task);
6672	unsigned long start_time = jiffies;
6673
6674	/* don't bother with service tasks if a reset is in progress */
6675	if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
6676		i40e_service_event_complete(pf);
6677		return;
6678	}
6679
6680	i40e_reset_subtask(pf);
6681	i40e_handle_mdd_event(pf);
6682	i40e_vc_process_vflr_event(pf);
6683	i40e_watchdog_subtask(pf);
6684	i40e_fdir_reinit_subtask(pf);
6685	i40e_sync_filters_subtask(pf);
6686#ifdef CONFIG_I40E_VXLAN
6687	i40e_sync_vxlan_filters_subtask(pf);
6688#endif
6689	i40e_clean_adminq_subtask(pf);
6690
6691	i40e_service_event_complete(pf);
6692
6693	/* If the tasks have taken longer than one timer cycle or there
6694	 * is more work to be done, reschedule the service task now
6695	 * rather than wait for the timer to tick again.
6696	 */
6697	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
6698	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
6699	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
6700	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
6701		i40e_service_event_schedule(pf);
6702}
6703
6704/**
6705 * i40e_service_timer - timer callback
6706 * @data: pointer to PF struct
6707 **/
6708static void i40e_service_timer(unsigned long data)
6709{
6710	struct i40e_pf *pf = (struct i40e_pf *)data;
6711
6712	mod_timer(&pf->service_timer,
6713		  round_jiffies(jiffies + pf->service_timer_period));
6714	i40e_service_event_schedule(pf);
6715}
6716
6717/**
6718 * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
6719 * @vsi: the VSI being configured
6720 **/
6721static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
6722{
6723	struct i40e_pf *pf = vsi->back;
6724
6725	switch (vsi->type) {
6726	case I40E_VSI_MAIN:
6727		vsi->alloc_queue_pairs = pf->num_lan_qps;
6728		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6729				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6730		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6731			vsi->num_q_vectors = pf->num_lan_msix;
6732		else
6733			vsi->num_q_vectors = 1;
6734
6735		break;
6736
6737	case I40E_VSI_FDIR:
6738		vsi->alloc_queue_pairs = 1;
6739		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
6740				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6741		vsi->num_q_vectors = 1;
6742		break;
6743
6744	case I40E_VSI_VMDQ2:
6745		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
6746		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6747				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6748		vsi->num_q_vectors = pf->num_vmdq_msix;
6749		break;
6750
6751	case I40E_VSI_SRIOV:
6752		vsi->alloc_queue_pairs = pf->num_vf_qps;
6753		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6754				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6755		break;
6756
6757#ifdef I40E_FCOE
6758	case I40E_VSI_FCOE:
6759		vsi->alloc_queue_pairs = pf->num_fcoe_qps;
6760		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
6761				      I40E_REQ_DESCRIPTOR_MULTIPLE);
6762		vsi->num_q_vectors = pf->num_fcoe_msix;
6763		break;
6764
6765#endif /* I40E_FCOE */
6766	default:
6767		WARN_ON(1);
6768		return -ENODATA;
6769	}
6770
6771	return 0;
6772}
6773
6774/**
6775 * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
6776 * @type: VSI pointer
6777 * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
6778 *
6779 * On error: returns error code (negative)
6780 * On success: returns 0
6781 **/
6782static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
6783{
6784	int size;
6785	int ret = 0;
6786
6787	/* allocate memory for both Tx and Rx ring pointers */
6788	size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
6789	vsi->tx_rings = kzalloc(size, GFP_KERNEL);
6790	if (!vsi->tx_rings)
6791		return -ENOMEM;
6792	vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
6793
6794	if (alloc_qvectors) {
6795		/* allocate memory for q_vector pointers */
6796		size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
6797		vsi->q_vectors = kzalloc(size, GFP_KERNEL);
6798		if (!vsi->q_vectors) {
6799			ret = -ENOMEM;
6800			goto err_vectors;
6801		}
6802	}
6803	return ret;
6804
6805err_vectors:
6806	kfree(vsi->tx_rings);
6807	return ret;
6808}
6809
6810/**
6811 * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
6812 * @pf: board private structure
6813 * @type: type of VSI
6814 *
6815 * On error: returns error code (negative)
6816 * On success: returns vsi index in PF (positive)
6817 **/
6818static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
6819{
6820	int ret = -ENODEV;
6821	struct i40e_vsi *vsi;
6822	int vsi_idx;
6823	int i;
6824
6825	/* Need to protect the allocation of the VSIs at the PF level */
6826	mutex_lock(&pf->switch_mutex);
6827
6828	/* VSI list may be fragmented if VSI creation/destruction has
6829	 * been happening.  We can afford to do a quick scan to look
6830	 * for any free VSIs in the list.
6831	 *
6832	 * find next empty vsi slot, looping back around if necessary
6833	 */
6834	i = pf->next_vsi;
6835	while (i < pf->num_alloc_vsi && pf->vsi[i])
6836		i++;
6837	if (i >= pf->num_alloc_vsi) {
6838		i = 0;
6839		while (i < pf->next_vsi && pf->vsi[i])
6840			i++;
6841	}
6842
6843	if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
6844		vsi_idx = i;             /* Found one! */
6845	} else {
6846		ret = -ENODEV;
6847		goto unlock_pf;  /* out of VSI slots! */
6848	}
6849	pf->next_vsi = ++i;
6850
6851	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
6852	if (!vsi) {
6853		ret = -ENOMEM;
6854		goto unlock_pf;
6855	}
6856	vsi->type = type;
6857	vsi->back = pf;
6858	set_bit(__I40E_DOWN, &vsi->state);
6859	vsi->flags = 0;
6860	vsi->idx = vsi_idx;
6861	vsi->rx_itr_setting = pf->rx_itr_default;
6862	vsi->tx_itr_setting = pf->tx_itr_default;
6863	vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
6864				pf->rss_table_size : 64;
6865	vsi->netdev_registered = false;
6866	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
6867	INIT_LIST_HEAD(&vsi->mac_filter_list);
6868	vsi->irqs_ready = false;
6869
6870	ret = i40e_set_num_rings_in_vsi(vsi);
6871	if (ret)
6872		goto err_rings;
6873
6874	ret = i40e_vsi_alloc_arrays(vsi, true);
6875	if (ret)
6876		goto err_rings;
6877
6878	/* Setup default MSIX irq handler for VSI */
6879	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
6880
6881	pf->vsi[vsi_idx] = vsi;
6882	ret = vsi_idx;
6883	goto unlock_pf;
6884
6885err_rings:
6886	pf->next_vsi = i - 1;
6887	kfree(vsi);
6888unlock_pf:
6889	mutex_unlock(&pf->switch_mutex);
6890	return ret;
6891}
6892
6893/**
6894 * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
6895 * @type: VSI pointer
6896 * @free_qvectors: a bool to specify if q_vectors need to be freed.
6897 *
6898 * On error: returns error code (negative)
6899 * On success: returns 0
6900 **/
6901static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
6902{
6903	/* free the ring and vector containers */
6904	if (free_qvectors) {
6905		kfree(vsi->q_vectors);
6906		vsi->q_vectors = NULL;
6907	}
6908	kfree(vsi->tx_rings);
6909	vsi->tx_rings = NULL;
6910	vsi->rx_rings = NULL;
6911}
6912
6913/**
6914 * i40e_vsi_clear - Deallocate the VSI provided
6915 * @vsi: the VSI being un-configured
6916 **/
6917static int i40e_vsi_clear(struct i40e_vsi *vsi)
6918{
6919	struct i40e_pf *pf;
6920
6921	if (!vsi)
6922		return 0;
6923
6924	if (!vsi->back)
6925		goto free_vsi;
6926	pf = vsi->back;
6927
6928	mutex_lock(&pf->switch_mutex);
6929	if (!pf->vsi[vsi->idx]) {
6930		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
6931			vsi->idx, vsi->idx, vsi, vsi->type);
6932		goto unlock_vsi;
6933	}
6934
6935	if (pf->vsi[vsi->idx] != vsi) {
6936		dev_err(&pf->pdev->dev,
6937			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
6938			pf->vsi[vsi->idx]->idx,
6939			pf->vsi[vsi->idx],
6940			pf->vsi[vsi->idx]->type,
6941			vsi->idx, vsi, vsi->type);
6942		goto unlock_vsi;
6943	}
6944
6945	/* updates the PF for this cleared vsi */
6946	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
6947	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
6948
6949	i40e_vsi_free_arrays(vsi, true);
6950
6951	pf->vsi[vsi->idx] = NULL;
6952	if (vsi->idx < pf->next_vsi)
6953		pf->next_vsi = vsi->idx;
6954
6955unlock_vsi:
6956	mutex_unlock(&pf->switch_mutex);
6957free_vsi:
6958	kfree(vsi);
6959
6960	return 0;
6961}
6962
6963/**
6964 * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
6965 * @vsi: the VSI being cleaned
6966 **/
6967static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
6968{
6969	int i;
6970
6971	if (vsi->tx_rings && vsi->tx_rings[0]) {
6972		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6973			kfree_rcu(vsi->tx_rings[i], rcu);
6974			vsi->tx_rings[i] = NULL;
6975			vsi->rx_rings[i] = NULL;
6976		}
6977	}
6978}
6979
6980/**
6981 * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
6982 * @vsi: the VSI being configured
6983 **/
6984static int i40e_alloc_rings(struct i40e_vsi *vsi)
6985{
6986	struct i40e_ring *tx_ring, *rx_ring;
6987	struct i40e_pf *pf = vsi->back;
6988	int i;
6989
6990	/* Set basic values in the rings to be used later during open() */
6991	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
6992		/* allocate space for both Tx and Rx in one shot */
6993		tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
6994		if (!tx_ring)
6995			goto err_out;
6996
6997		tx_ring->queue_index = i;
6998		tx_ring->reg_idx = vsi->base_queue + i;
6999		tx_ring->ring_active = false;
7000		tx_ring->vsi = vsi;
7001		tx_ring->netdev = vsi->netdev;
7002		tx_ring->dev = &pf->pdev->dev;
7003		tx_ring->count = vsi->num_desc;
7004		tx_ring->size = 0;
7005		tx_ring->dcb_tc = 0;
7006		vsi->tx_rings[i] = tx_ring;
7007
7008		rx_ring = &tx_ring[1];
7009		rx_ring->queue_index = i;
7010		rx_ring->reg_idx = vsi->base_queue + i;
7011		rx_ring->ring_active = false;
7012		rx_ring->vsi = vsi;
7013		rx_ring->netdev = vsi->netdev;
7014		rx_ring->dev = &pf->pdev->dev;
7015		rx_ring->count = vsi->num_desc;
7016		rx_ring->size = 0;
7017		rx_ring->dcb_tc = 0;
7018		if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
7019			set_ring_16byte_desc_enabled(rx_ring);
7020		else
7021			clear_ring_16byte_desc_enabled(rx_ring);
7022		vsi->rx_rings[i] = rx_ring;
7023	}
7024
7025	return 0;
7026
7027err_out:
7028	i40e_vsi_clear_rings(vsi);
7029	return -ENOMEM;
7030}
7031
7032/**
7033 * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7034 * @pf: board private structure
7035 * @vectors: the number of MSI-X vectors to request
7036 *
7037 * Returns the number of vectors reserved, or error
7038 **/
7039static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7040{
7041	vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7042					I40E_MIN_MSIX, vectors);
7043	if (vectors < 0) {
7044		dev_info(&pf->pdev->dev,
7045			 "MSI-X vector reservation failed: %d\n", vectors);
7046		vectors = 0;
7047	}
7048
7049	return vectors;
7050}
7051
7052/**
7053 * i40e_init_msix - Setup the MSIX capability
7054 * @pf: board private structure
7055 *
7056 * Work with the OS to set up the MSIX vectors needed.
7057 *
7058 * Returns the number of vectors reserved or negative on failure
7059 **/
7060static int i40e_init_msix(struct i40e_pf *pf)
7061{
7062	struct i40e_hw *hw = &pf->hw;
7063	int vectors_left;
7064	int v_budget, i;
7065	int v_actual;
7066
7067	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7068		return -ENODEV;
7069
7070	/* The number of vectors we'll request will be comprised of:
7071	 *   - Add 1 for "other" cause for Admin Queue events, etc.
7072	 *   - The number of LAN queue pairs
7073	 *	- Queues being used for RSS.
7074	 *		We don't need as many as max_rss_size vectors.
7075	 *		use rss_size instead in the calculation since that
7076	 *		is governed by number of cpus in the system.
7077	 *	- assumes symmetric Tx/Rx pairing
7078	 *   - The number of VMDq pairs
7079#ifdef I40E_FCOE
7080	 *   - The number of FCOE qps.
7081#endif
7082	 * Once we count this up, try the request.
7083	 *
7084	 * If we can't get what we want, we'll simplify to nearly nothing
7085	 * and try again.  If that still fails, we punt.
7086	 */
7087	vectors_left = hw->func_caps.num_msix_vectors;
7088	v_budget = 0;
7089
7090	/* reserve one vector for miscellaneous handler */
7091	if (vectors_left) {
7092		v_budget++;
7093		vectors_left--;
7094	}
7095
7096	/* reserve vectors for the main PF traffic queues */
7097	pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7098	vectors_left -= pf->num_lan_msix;
7099	v_budget += pf->num_lan_msix;
7100
7101	/* reserve one vector for sideband flow director */
7102	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7103		if (vectors_left) {
7104			v_budget++;
7105			vectors_left--;
7106		} else {
7107			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7108		}
7109	}
7110
7111#ifdef I40E_FCOE
7112	/* can we reserve enough for FCoE? */
7113	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7114		if (!vectors_left)
7115			pf->num_fcoe_msix = 0;
7116		else if (vectors_left >= pf->num_fcoe_qps)
7117			pf->num_fcoe_msix = pf->num_fcoe_qps;
7118		else
7119			pf->num_fcoe_msix = 1;
7120		v_budget += pf->num_fcoe_msix;
7121		vectors_left -= pf->num_fcoe_msix;
7122	}
7123
7124#endif
7125	/* any vectors left over go for VMDq support */
7126	if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7127		int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7128		int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7129
7130		/* if we're short on vectors for what's desired, we limit
7131		 * the queues per vmdq.  If this is still more than are
7132		 * available, the user will need to change the number of
7133		 * queues/vectors used by the PF later with the ethtool
7134		 * channels command
7135		 */
7136		if (vmdq_vecs < vmdq_vecs_wanted)
7137			pf->num_vmdq_qps = 1;
7138		pf->num_vmdq_msix = pf->num_vmdq_qps;
7139
7140		v_budget += vmdq_vecs;
7141		vectors_left -= vmdq_vecs;
7142	}
7143
7144	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7145				   GFP_KERNEL);
7146	if (!pf->msix_entries)
7147		return -ENOMEM;
7148
7149	for (i = 0; i < v_budget; i++)
7150		pf->msix_entries[i].entry = i;
7151	v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7152
7153	if (v_actual != v_budget) {
7154		/* If we have limited resources, we will start with no vectors
7155		 * for the special features and then allocate vectors to some
7156		 * of these features based on the policy and at the end disable
7157		 * the features that did not get any vectors.
7158		 */
7159#ifdef I40E_FCOE
7160		pf->num_fcoe_qps = 0;
7161		pf->num_fcoe_msix = 0;
7162#endif
7163		pf->num_vmdq_msix = 0;
7164	}
7165
7166	if (v_actual < I40E_MIN_MSIX) {
7167		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7168		kfree(pf->msix_entries);
7169		pf->msix_entries = NULL;
7170		return -ENODEV;
7171
7172	} else if (v_actual == I40E_MIN_MSIX) {
7173		/* Adjust for minimal MSIX use */
7174		pf->num_vmdq_vsis = 0;
7175		pf->num_vmdq_qps = 0;
7176		pf->num_lan_qps = 1;
7177		pf->num_lan_msix = 1;
7178
7179	} else if (v_actual != v_budget) {
7180		int vec;
7181
7182		/* reserve the misc vector */
7183		vec = v_actual - 1;
7184
7185		/* Scale vector usage down */
7186		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7187		pf->num_vmdq_vsis = 1;
7188		pf->num_vmdq_qps = 1;
7189		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7190
7191		/* partition out the remaining vectors */
7192		switch (vec) {
7193		case 2:
7194			pf->num_lan_msix = 1;
7195			break;
7196		case 3:
7197#ifdef I40E_FCOE
7198			/* give one vector to FCoE */
7199			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7200				pf->num_lan_msix = 1;
7201				pf->num_fcoe_msix = 1;
7202			}
7203#else
7204			pf->num_lan_msix = 2;
7205#endif
7206			break;
7207		default:
7208#ifdef I40E_FCOE
7209			/* give one vector to FCoE */
7210			if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7211				pf->num_fcoe_msix = 1;
7212				vec--;
7213			}
7214#endif
7215			/* give the rest to the PF */
7216			pf->num_lan_msix = min_t(int, vec, pf->num_lan_qps);
7217			break;
7218		}
7219	}
7220
7221	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7222	    (pf->num_vmdq_msix == 0)) {
7223		dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7224		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7225	}
7226#ifdef I40E_FCOE
7227
7228	if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7229		dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7230		pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7231	}
7232#endif
7233	return v_actual;
7234}
7235
7236/**
7237 * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7238 * @vsi: the VSI being configured
7239 * @v_idx: index of the vector in the vsi struct
7240 *
7241 * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7242 **/
7243static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
7244{
7245	struct i40e_q_vector *q_vector;
7246
7247	/* allocate q_vector */
7248	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7249	if (!q_vector)
7250		return -ENOMEM;
7251
7252	q_vector->vsi = vsi;
7253	q_vector->v_idx = v_idx;
7254	cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
7255	if (vsi->netdev)
7256		netif_napi_add(vsi->netdev, &q_vector->napi,
7257			       i40e_napi_poll, NAPI_POLL_WEIGHT);
7258
7259	q_vector->rx.latency_range = I40E_LOW_LATENCY;
7260	q_vector->tx.latency_range = I40E_LOW_LATENCY;
7261
7262	/* tie q_vector and vsi together */
7263	vsi->q_vectors[v_idx] = q_vector;
7264
7265	return 0;
7266}
7267
7268/**
7269 * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7270 * @vsi: the VSI being configured
7271 *
7272 * We allocate one q_vector per queue interrupt.  If allocation fails we
7273 * return -ENOMEM.
7274 **/
7275static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7276{
7277	struct i40e_pf *pf = vsi->back;
7278	int v_idx, num_q_vectors;
7279	int err;
7280
7281	/* if not MSIX, give the one vector only to the LAN VSI */
7282	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7283		num_q_vectors = vsi->num_q_vectors;
7284	else if (vsi == pf->vsi[pf->lan_vsi])
7285		num_q_vectors = 1;
7286	else
7287		return -EINVAL;
7288
7289	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7290		err = i40e_vsi_alloc_q_vector(vsi, v_idx);
7291		if (err)
7292			goto err_out;
7293	}
7294
7295	return 0;
7296
7297err_out:
7298	while (v_idx--)
7299		i40e_free_q_vector(vsi, v_idx);
7300
7301	return err;
7302}
7303
7304/**
7305 * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7306 * @pf: board private structure to initialize
7307 **/
7308static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7309{
7310	int vectors = 0;
7311	ssize_t size;
7312
7313	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7314		vectors = i40e_init_msix(pf);
7315		if (vectors < 0) {
7316			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	|
7317#ifdef I40E_FCOE
7318				       I40E_FLAG_FCOE_ENABLED	|
7319#endif
7320				       I40E_FLAG_RSS_ENABLED	|
7321				       I40E_FLAG_DCB_CAPABLE	|
7322				       I40E_FLAG_SRIOV_ENABLED	|
7323				       I40E_FLAG_FD_SB_ENABLED	|
7324				       I40E_FLAG_FD_ATR_ENABLED	|
7325				       I40E_FLAG_VMDQ_ENABLED);
7326
7327			/* rework the queue expectations without MSIX */
7328			i40e_determine_queue_usage(pf);
7329		}
7330	}
7331
7332	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7333	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7334		dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7335		vectors = pci_enable_msi(pf->pdev);
7336		if (vectors < 0) {
7337			dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7338				 vectors);
7339			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7340		}
7341		vectors = 1;  /* one MSI or Legacy vector */
7342	}
7343
7344	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7345		dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7346
7347	/* set up vector assignment tracking */
7348	size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7349	pf->irq_pile = kzalloc(size, GFP_KERNEL);
7350	if (!pf->irq_pile) {
7351		dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7352		return -ENOMEM;
7353	}
7354	pf->irq_pile->num_entries = vectors;
7355	pf->irq_pile->search_hint = 0;
7356
7357	/* track first vector for misc interrupts, ignore return */
7358	(void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7359
7360	return 0;
7361}
7362
7363/**
7364 * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7365 * @pf: board private structure
7366 *
7367 * This sets up the handler for MSIX 0, which is used to manage the
7368 * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7369 * when in MSI or Legacy interrupt mode.
7370 **/
7371static int i40e_setup_misc_vector(struct i40e_pf *pf)
7372{
7373	struct i40e_hw *hw = &pf->hw;
7374	int err = 0;
7375
7376	/* Only request the irq if this is the first time through, and
7377	 * not when we're rebuilding after a Reset
7378	 */
7379	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7380		err = request_irq(pf->msix_entries[0].vector,
7381				  i40e_intr, 0, pf->int_name, pf);
7382		if (err) {
7383			dev_info(&pf->pdev->dev,
7384				 "request_irq for %s failed: %d\n",
7385				 pf->int_name, err);
7386			return -EFAULT;
7387		}
7388	}
7389
7390	i40e_enable_misc_int_causes(pf);
7391
7392	/* associate no queues to the misc vector */
7393	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7394	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7395
7396	i40e_flush(hw);
7397
7398	i40e_irq_dynamic_enable_icr0(pf);
7399
7400	return err;
7401}
7402
7403/**
7404 * i40e_config_rss - Prepare for RSS if used
7405 * @pf: board private structure
7406 **/
7407static int i40e_config_rss(struct i40e_pf *pf)
7408{
7409	u32 rss_key[I40E_PFQF_HKEY_MAX_INDEX + 1];
7410	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7411	struct i40e_hw *hw = &pf->hw;
7412	u32 lut = 0;
7413	int i, j;
7414	u64 hena;
7415	u32 reg_val;
7416
7417	netdev_rss_key_fill(rss_key, sizeof(rss_key));
7418	for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
7419		wr32(hw, I40E_PFQF_HKEY(i), rss_key[i]);
7420
7421	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
7422	hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
7423		((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
7424	hena |= I40E_DEFAULT_RSS_HENA;
7425	wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
7426	wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
7427
7428	vsi->rss_size = min_t(int, pf->rss_size, vsi->num_queue_pairs);
7429
7430	/* Check capability and Set table size and register per hw expectation*/
7431	reg_val = rd32(hw, I40E_PFQF_CTL_0);
7432	if (pf->rss_table_size == 512)
7433		reg_val |= I40E_PFQF_CTL_0_HASHLUTSIZE_512;
7434	else
7435		reg_val &= ~I40E_PFQF_CTL_0_HASHLUTSIZE_512;
7436	wr32(hw, I40E_PFQF_CTL_0, reg_val);
7437
7438	/* Populate the LUT with max no. of queues in round robin fashion */
7439	for (i = 0, j = 0; i < pf->rss_table_size; i++, j++) {
7440
7441		/* The assumption is that lan qp count will be the highest
7442		 * qp count for any PF VSI that needs RSS.
7443		 * If multiple VSIs need RSS support, all the qp counts
7444		 * for those VSIs should be a power of 2 for RSS to work.
7445		 * If LAN VSI is the only consumer for RSS then this requirement
7446		 * is not necessary.
7447		 */
7448		if (j == vsi->rss_size)
7449			j = 0;
7450		/* lut = 4-byte sliding window of 4 lut entries */
7451		lut = (lut << 8) | (j &
7452			 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
7453		/* On i = 3, we have 4 entries in lut; write to the register */
7454		if ((i & 3) == 3)
7455			wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
7456	}
7457	i40e_flush(hw);
7458
7459	return 0;
7460}
7461
7462/**
7463 * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
7464 * @pf: board private structure
7465 * @queue_count: the requested queue count for rss.
7466 *
7467 * returns 0 if rss is not enabled, if enabled returns the final rss queue
7468 * count which may be different from the requested queue count.
7469 **/
7470int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
7471{
7472	struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
7473	int new_rss_size;
7474
7475	if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
7476		return 0;
7477
7478	new_rss_size = min_t(int, queue_count, pf->rss_size_max);
7479
7480	if (queue_count != vsi->num_queue_pairs) {
7481		vsi->req_queue_pairs = queue_count;
7482		i40e_prep_for_reset(pf);
7483
7484		pf->rss_size = new_rss_size;
7485
7486		i40e_reset_and_rebuild(pf, true);
7487		i40e_config_rss(pf);
7488	}
7489	dev_info(&pf->pdev->dev, "RSS count:  %d\n", pf->rss_size);
7490	return pf->rss_size;
7491}
7492
7493/**
7494 * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
7495 * @pf: board private structure
7496 **/
7497i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
7498{
7499	i40e_status status;
7500	bool min_valid, max_valid;
7501	u32 max_bw, min_bw;
7502
7503	status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
7504					   &min_valid, &max_valid);
7505
7506	if (!status) {
7507		if (min_valid)
7508			pf->npar_min_bw = min_bw;
7509		if (max_valid)
7510			pf->npar_max_bw = max_bw;
7511	}
7512
7513	return status;
7514}
7515
7516/**
7517 * i40e_set_npar_bw_setting - Set BW settings for this PF partition
7518 * @pf: board private structure
7519 **/
7520i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
7521{
7522	struct i40e_aqc_configure_partition_bw_data bw_data;
7523	i40e_status status;
7524
7525	/* Set the valid bit for this PF */
7526	bw_data.pf_valid_bits = cpu_to_le16(1 << pf->hw.pf_id);
7527	bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
7528	bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
7529
7530	/* Set the new bandwidths */
7531	status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
7532
7533	return status;
7534}
7535
7536/**
7537 * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
7538 * @pf: board private structure
7539 **/
7540i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
7541{
7542	/* Commit temporary BW setting to permanent NVM image */
7543	enum i40e_admin_queue_err last_aq_status;
7544	i40e_status ret;
7545	u16 nvm_word;
7546
7547	if (pf->hw.partition_id != 1) {
7548		dev_info(&pf->pdev->dev,
7549			 "Commit BW only works on partition 1! This is partition %d",
7550			 pf->hw.partition_id);
7551		ret = I40E_NOT_SUPPORTED;
7552		goto bw_commit_out;
7553	}
7554
7555	/* Acquire NVM for read access */
7556	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
7557	last_aq_status = pf->hw.aq.asq_last_status;
7558	if (ret) {
7559		dev_info(&pf->pdev->dev,
7560			 "Cannot acquire NVM for read access, err %d: aq_err %d\n",
7561			 ret, last_aq_status);
7562		goto bw_commit_out;
7563	}
7564
7565	/* Read word 0x10 of NVM - SW compatibility word 1 */
7566	ret = i40e_aq_read_nvm(&pf->hw,
7567			       I40E_SR_NVM_CONTROL_WORD,
7568			       0x10, sizeof(nvm_word), &nvm_word,
7569			       false, NULL);
7570	/* Save off last admin queue command status before releasing
7571	 * the NVM
7572	 */
7573	last_aq_status = pf->hw.aq.asq_last_status;
7574	i40e_release_nvm(&pf->hw);
7575	if (ret) {
7576		dev_info(&pf->pdev->dev, "NVM read error, err %d aq_err %d\n",
7577			 ret, last_aq_status);
7578		goto bw_commit_out;
7579	}
7580
7581	/* Wait a bit for NVM release to complete */
7582	msleep(50);
7583
7584	/* Acquire NVM for write access */
7585	ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
7586	last_aq_status = pf->hw.aq.asq_last_status;
7587	if (ret) {
7588		dev_info(&pf->pdev->dev,
7589			 "Cannot acquire NVM for write access, err %d: aq_err %d\n",
7590			 ret, last_aq_status);
7591		goto bw_commit_out;
7592	}
7593	/* Write it back out unchanged to initiate update NVM,
7594	 * which will force a write of the shadow (alt) RAM to
7595	 * the NVM - thus storing the bandwidth values permanently.
7596	 */
7597	ret = i40e_aq_update_nvm(&pf->hw,
7598				 I40E_SR_NVM_CONTROL_WORD,
7599				 0x10, sizeof(nvm_word),
7600				 &nvm_word, true, NULL);
7601	/* Save off last admin queue command status before releasing
7602	 * the NVM
7603	 */
7604	last_aq_status = pf->hw.aq.asq_last_status;
7605	i40e_release_nvm(&pf->hw);
7606	if (ret)
7607		dev_info(&pf->pdev->dev,
7608			 "BW settings NOT SAVED, err %d aq_err %d\n",
7609			 ret, last_aq_status);
7610bw_commit_out:
7611
7612	return ret;
7613}
7614
7615/**
7616 * i40e_sw_init - Initialize general software structures (struct i40e_pf)
7617 * @pf: board private structure to initialize
7618 *
7619 * i40e_sw_init initializes the Adapter private data structure.
7620 * Fields are initialized based on PCI device information and
7621 * OS network device settings (MTU size).
7622 **/
7623static int i40e_sw_init(struct i40e_pf *pf)
7624{
7625	int err = 0;
7626	int size;
7627
7628	pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
7629				(NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
7630	pf->hw.debug_mask = pf->msg_enable | I40E_DEBUG_DIAG;
7631	if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
7632		if (I40E_DEBUG_USER & debug)
7633			pf->hw.debug_mask = debug;
7634		pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
7635						I40E_DEFAULT_MSG_ENABLE);
7636	}
7637
7638	/* Set default capability flags */
7639	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
7640		    I40E_FLAG_MSI_ENABLED     |
7641		    I40E_FLAG_MSIX_ENABLED;
7642
7643	if (iommu_present(&pci_bus_type))
7644		pf->flags |= I40E_FLAG_RX_PS_ENABLED;
7645	else
7646		pf->flags |= I40E_FLAG_RX_1BUF_ENABLED;
7647
7648	/* Set default ITR */
7649	pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
7650	pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
7651
7652	/* Depending on PF configurations, it is possible that the RSS
7653	 * maximum might end up larger than the available queues
7654	 */
7655	pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
7656	pf->rss_size = 1;
7657	pf->rss_table_size = pf->hw.func_caps.rss_table_size;
7658	pf->rss_size_max = min_t(int, pf->rss_size_max,
7659				 pf->hw.func_caps.num_tx_qp);
7660	if (pf->hw.func_caps.rss) {
7661		pf->flags |= I40E_FLAG_RSS_ENABLED;
7662		pf->rss_size = min_t(int, pf->rss_size_max, num_online_cpus());
7663	}
7664
7665	/* MFP mode enabled */
7666	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
7667		pf->flags |= I40E_FLAG_MFP_ENABLED;
7668		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
7669		if (i40e_get_npar_bw_setting(pf))
7670			dev_warn(&pf->pdev->dev,
7671				 "Could not get NPAR bw settings\n");
7672		else
7673			dev_info(&pf->pdev->dev,
7674				 "Min BW = %8.8x, Max BW = %8.8x\n",
7675				 pf->npar_min_bw, pf->npar_max_bw);
7676	}
7677
7678	/* FW/NVM is not yet fixed in this regard */
7679	if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
7680	    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
7681		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
7682		pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
7683		/* Setup a counter for fd_atr per PF */
7684		pf->fd_atr_cnt_idx = I40E_FD_ATR_STAT_IDX(pf->hw.pf_id);
7685		if (!(pf->flags & I40E_FLAG_MFP_ENABLED)) {
7686			pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7687			/* Setup a counter for fd_sb per PF */
7688			pf->fd_sb_cnt_idx = I40E_FD_SB_STAT_IDX(pf->hw.pf_id);
7689		} else {
7690			dev_info(&pf->pdev->dev,
7691				 "Flow Director Sideband mode Disabled in MFP mode\n");
7692		}
7693		pf->fdir_pf_filter_count =
7694				 pf->hw.func_caps.fd_filters_guaranteed;
7695		pf->hw.fdir_shared_filter_count =
7696				 pf->hw.func_caps.fd_filters_best_effort;
7697	}
7698
7699	if (pf->hw.func_caps.vmdq) {
7700		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
7701		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
7702		pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
7703	}
7704
7705#ifdef I40E_FCOE
7706	err = i40e_init_pf_fcoe(pf);
7707	if (err)
7708		dev_info(&pf->pdev->dev, "init_pf_fcoe failed: %d\n", err);
7709
7710#endif /* I40E_FCOE */
7711#ifdef CONFIG_PCI_IOV
7712	if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
7713		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
7714		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
7715		pf->num_req_vfs = min_t(int,
7716					pf->hw.func_caps.num_vfs,
7717					I40E_MAX_VF_COUNT);
7718	}
7719#endif /* CONFIG_PCI_IOV */
7720	pf->eeprom_version = 0xDEAD;
7721	pf->lan_veb = I40E_NO_VEB;
7722	pf->lan_vsi = I40E_NO_VSI;
7723
7724	/* set up queue assignment tracking */
7725	size = sizeof(struct i40e_lump_tracking)
7726		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
7727	pf->qp_pile = kzalloc(size, GFP_KERNEL);
7728	if (!pf->qp_pile) {
7729		err = -ENOMEM;
7730		goto sw_init_done;
7731	}
7732	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
7733	pf->qp_pile->search_hint = 0;
7734
7735	pf->tx_timeout_recovery_level = 1;
7736
7737	mutex_init(&pf->switch_mutex);
7738
7739	/* If NPAR is enabled nudge the Tx scheduler */
7740	if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
7741		i40e_set_npar_bw_setting(pf);
7742
7743sw_init_done:
7744	return err;
7745}
7746
7747/**
7748 * i40e_set_ntuple - set the ntuple feature flag and take action
7749 * @pf: board private structure to initialize
7750 * @features: the feature set that the stack is suggesting
7751 *
7752 * returns a bool to indicate if reset needs to happen
7753 **/
7754bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
7755{
7756	bool need_reset = false;
7757
7758	/* Check if Flow Director n-tuple support was enabled or disabled.  If
7759	 * the state changed, we need to reset.
7760	 */
7761	if (features & NETIF_F_NTUPLE) {
7762		/* Enable filters and mark for reset */
7763		if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
7764			need_reset = true;
7765		pf->flags |= I40E_FLAG_FD_SB_ENABLED;
7766	} else {
7767		/* turn off filters, mark for reset and clear SW filter list */
7768		if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7769			need_reset = true;
7770			i40e_fdir_filter_exit(pf);
7771		}
7772		pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7773		pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
7774		/* reset fd counters */
7775		pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
7776		pf->fdir_pf_active_filters = 0;
7777		pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
7778		dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
7779		/* if ATR was auto disabled it can be re-enabled. */
7780		if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
7781		    (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
7782			pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
7783	}
7784	return need_reset;
7785}
7786
7787/**
7788 * i40e_set_features - set the netdev feature flags
7789 * @netdev: ptr to the netdev being adjusted
7790 * @features: the feature set that the stack is suggesting
7791 **/
7792static int i40e_set_features(struct net_device *netdev,
7793			     netdev_features_t features)
7794{
7795	struct i40e_netdev_priv *np = netdev_priv(netdev);
7796	struct i40e_vsi *vsi = np->vsi;
7797	struct i40e_pf *pf = vsi->back;
7798	bool need_reset;
7799
7800	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7801		i40e_vlan_stripping_enable(vsi);
7802	else
7803		i40e_vlan_stripping_disable(vsi);
7804
7805	need_reset = i40e_set_ntuple(pf, features);
7806
7807	if (need_reset)
7808		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
7809
7810	return 0;
7811}
7812
7813#ifdef CONFIG_I40E_VXLAN
7814/**
7815 * i40e_get_vxlan_port_idx - Lookup a possibly offloaded for Rx UDP port
7816 * @pf: board private structure
7817 * @port: The UDP port to look up
7818 *
7819 * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
7820 **/
7821static u8 i40e_get_vxlan_port_idx(struct i40e_pf *pf, __be16 port)
7822{
7823	u8 i;
7824
7825	for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7826		if (pf->vxlan_ports[i] == port)
7827			return i;
7828	}
7829
7830	return i;
7831}
7832
7833/**
7834 * i40e_add_vxlan_port - Get notifications about VXLAN ports that come up
7835 * @netdev: This physical port's netdev
7836 * @sa_family: Socket Family that VXLAN is notifying us about
7837 * @port: New UDP port number that VXLAN started listening to
7838 **/
7839static void i40e_add_vxlan_port(struct net_device *netdev,
7840				sa_family_t sa_family, __be16 port)
7841{
7842	struct i40e_netdev_priv *np = netdev_priv(netdev);
7843	struct i40e_vsi *vsi = np->vsi;
7844	struct i40e_pf *pf = vsi->back;
7845	u8 next_idx;
7846	u8 idx;
7847
7848	if (sa_family == AF_INET6)
7849		return;
7850
7851	idx = i40e_get_vxlan_port_idx(pf, port);
7852
7853	/* Check if port already exists */
7854	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7855		netdev_info(netdev, "vxlan port %d already offloaded\n",
7856			    ntohs(port));
7857		return;
7858	}
7859
7860	/* Now check if there is space to add the new port */
7861	next_idx = i40e_get_vxlan_port_idx(pf, 0);
7862
7863	if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7864		netdev_info(netdev, "maximum number of vxlan UDP ports reached, not adding port %d\n",
7865			    ntohs(port));
7866		return;
7867	}
7868
7869	/* New port: add it and mark its index in the bitmap */
7870	pf->vxlan_ports[next_idx] = port;
7871	pf->pending_vxlan_bitmap |= (1 << next_idx);
7872	pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
7873
7874	dev_info(&pf->pdev->dev, "adding vxlan port %d\n", ntohs(port));
7875}
7876
7877/**
7878 * i40e_del_vxlan_port - Get notifications about VXLAN ports that go away
7879 * @netdev: This physical port's netdev
7880 * @sa_family: Socket Family that VXLAN is notifying us about
7881 * @port: UDP port number that VXLAN stopped listening to
7882 **/
7883static void i40e_del_vxlan_port(struct net_device *netdev,
7884				sa_family_t sa_family, __be16 port)
7885{
7886	struct i40e_netdev_priv *np = netdev_priv(netdev);
7887	struct i40e_vsi *vsi = np->vsi;
7888	struct i40e_pf *pf = vsi->back;
7889	u8 idx;
7890
7891	if (sa_family == AF_INET6)
7892		return;
7893
7894	idx = i40e_get_vxlan_port_idx(pf, port);
7895
7896	/* Check if port already exists */
7897	if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
7898		/* if port exists, set it to 0 (mark for deletion)
7899		 * and make it pending
7900		 */
7901		pf->vxlan_ports[idx] = 0;
7902		pf->pending_vxlan_bitmap |= (1 << idx);
7903		pf->flags |= I40E_FLAG_VXLAN_FILTER_SYNC;
7904
7905		dev_info(&pf->pdev->dev, "deleting vxlan port %d\n",
7906			 ntohs(port));
7907	} else {
7908		netdev_warn(netdev, "vxlan port %d was not found, not deleting\n",
7909			    ntohs(port));
7910	}
7911}
7912
7913#endif
7914static int i40e_get_phys_port_id(struct net_device *netdev,
7915				 struct netdev_phys_item_id *ppid)
7916{
7917	struct i40e_netdev_priv *np = netdev_priv(netdev);
7918	struct i40e_pf *pf = np->vsi->back;
7919	struct i40e_hw *hw = &pf->hw;
7920
7921	if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
7922		return -EOPNOTSUPP;
7923
7924	ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
7925	memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
7926
7927	return 0;
7928}
7929
7930/**
7931 * i40e_ndo_fdb_add - add an entry to the hardware database
7932 * @ndm: the input from the stack
7933 * @tb: pointer to array of nladdr (unused)
7934 * @dev: the net device pointer
7935 * @addr: the MAC address entry being added
7936 * @flags: instructions from stack about fdb operation
7937 */
7938static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
7939			    struct net_device *dev,
7940			    const unsigned char *addr, u16 vid,
7941			    u16 flags)
7942{
7943	struct i40e_netdev_priv *np = netdev_priv(dev);
7944	struct i40e_pf *pf = np->vsi->back;
7945	int err = 0;
7946
7947	if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
7948		return -EOPNOTSUPP;
7949
7950	if (vid) {
7951		pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
7952		return -EINVAL;
7953	}
7954
7955	/* Hardware does not support aging addresses so if a
7956	 * ndm_state is given only allow permanent addresses
7957	 */
7958	if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
7959		netdev_info(dev, "FDB only supports static addresses\n");
7960		return -EINVAL;
7961	}
7962
7963	if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
7964		err = dev_uc_add_excl(dev, addr);
7965	else if (is_multicast_ether_addr(addr))
7966		err = dev_mc_add_excl(dev, addr);
7967	else
7968		err = -EINVAL;
7969
7970	/* Only return duplicate errors if NLM_F_EXCL is set */
7971	if (err == -EEXIST && !(flags & NLM_F_EXCL))
7972		err = 0;
7973
7974	return err;
7975}
7976
7977#ifdef HAVE_BRIDGE_ATTRIBS
7978/**
7979 * i40e_ndo_bridge_setlink - Set the hardware bridge mode
7980 * @dev: the netdev being configured
7981 * @nlh: RTNL message
7982 *
7983 * Inserts a new hardware bridge if not already created and
7984 * enables the bridging mode requested (VEB or VEPA). If the
7985 * hardware bridge has already been inserted and the request
7986 * is to change the mode then that requires a PF reset to
7987 * allow rebuild of the components with required hardware
7988 * bridge mode enabled.
7989 **/
7990static int i40e_ndo_bridge_setlink(struct net_device *dev,
7991				   struct nlmsghdr *nlh)
7992{
7993	struct i40e_netdev_priv *np = netdev_priv(dev);
7994	struct i40e_vsi *vsi = np->vsi;
7995	struct i40e_pf *pf = vsi->back;
7996	struct i40e_veb *veb = NULL;
7997	struct nlattr *attr, *br_spec;
7998	int i, rem;
7999
8000	/* Only for PF VSI for now */
8001	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8002		return -EOPNOTSUPP;
8003
8004	/* Find the HW bridge for PF VSI */
8005	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8006		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8007			veb = pf->veb[i];
8008	}
8009
8010	br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8011
8012	nla_for_each_nested(attr, br_spec, rem) {
8013		__u16 mode;
8014
8015		if (nla_type(attr) != IFLA_BRIDGE_MODE)
8016			continue;
8017
8018		mode = nla_get_u16(attr);
8019		if ((mode != BRIDGE_MODE_VEPA) &&
8020		    (mode != BRIDGE_MODE_VEB))
8021			return -EINVAL;
8022
8023		/* Insert a new HW bridge */
8024		if (!veb) {
8025			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8026					     vsi->tc_config.enabled_tc);
8027			if (veb) {
8028				veb->bridge_mode = mode;
8029				i40e_config_bridge_mode(veb);
8030			} else {
8031				/* No Bridge HW offload available */
8032				return -ENOENT;
8033			}
8034			break;
8035		} else if (mode != veb->bridge_mode) {
8036			/* Existing HW bridge but different mode needs reset */
8037			veb->bridge_mode = mode;
8038			/* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8039			if (mode == BRIDGE_MODE_VEB)
8040				pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8041			else
8042				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8043			i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8044			break;
8045		}
8046	}
8047
8048	return 0;
8049}
8050
8051/**
8052 * i40e_ndo_bridge_getlink - Get the hardware bridge mode
8053 * @skb: skb buff
8054 * @pid: process id
8055 * @seq: RTNL message seq #
8056 * @dev: the netdev being configured
8057 * @filter_mask: unused
8058 *
8059 * Return the mode in which the hardware bridge is operating in
8060 * i.e VEB or VEPA.
8061 **/
8062#ifdef HAVE_BRIDGE_FILTER
8063static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8064				   struct net_device *dev,
8065				   u32 __always_unused filter_mask, int nlflags)
8066#else
8067static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
8068				   struct net_device *dev, int nlflags)
8069#endif /* HAVE_BRIDGE_FILTER */
8070{
8071	struct i40e_netdev_priv *np = netdev_priv(dev);
8072	struct i40e_vsi *vsi = np->vsi;
8073	struct i40e_pf *pf = vsi->back;
8074	struct i40e_veb *veb = NULL;
8075	int i;
8076
8077	/* Only for PF VSI for now */
8078	if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8079		return -EOPNOTSUPP;
8080
8081	/* Find the HW bridge for the PF VSI */
8082	for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8083		if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8084			veb = pf->veb[i];
8085	}
8086
8087	if (!veb)
8088		return 0;
8089
8090	return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
8091				       nlflags);
8092}
8093#endif /* HAVE_BRIDGE_ATTRIBS */
8094
8095static const struct net_device_ops i40e_netdev_ops = {
8096	.ndo_open		= i40e_open,
8097	.ndo_stop		= i40e_close,
8098	.ndo_start_xmit		= i40e_lan_xmit_frame,
8099	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
8100	.ndo_set_rx_mode	= i40e_set_rx_mode,
8101	.ndo_validate_addr	= eth_validate_addr,
8102	.ndo_set_mac_address	= i40e_set_mac,
8103	.ndo_change_mtu		= i40e_change_mtu,
8104	.ndo_do_ioctl		= i40e_ioctl,
8105	.ndo_tx_timeout		= i40e_tx_timeout,
8106	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
8107	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
8108#ifdef CONFIG_NET_POLL_CONTROLLER
8109	.ndo_poll_controller	= i40e_netpoll,
8110#endif
8111	.ndo_setup_tc		= i40e_setup_tc,
8112#ifdef I40E_FCOE
8113	.ndo_fcoe_enable	= i40e_fcoe_enable,
8114	.ndo_fcoe_disable	= i40e_fcoe_disable,
8115#endif
8116	.ndo_set_features	= i40e_set_features,
8117	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
8118	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
8119	.ndo_set_vf_rate	= i40e_ndo_set_vf_bw,
8120	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
8121	.ndo_set_vf_link_state	= i40e_ndo_set_vf_link_state,
8122	.ndo_set_vf_spoofchk	= i40e_ndo_set_vf_spoofchk,
8123#ifdef CONFIG_I40E_VXLAN
8124	.ndo_add_vxlan_port	= i40e_add_vxlan_port,
8125	.ndo_del_vxlan_port	= i40e_del_vxlan_port,
8126#endif
8127	.ndo_get_phys_port_id	= i40e_get_phys_port_id,
8128	.ndo_fdb_add		= i40e_ndo_fdb_add,
8129#ifdef HAVE_BRIDGE_ATTRIBS
8130	.ndo_bridge_getlink	= i40e_ndo_bridge_getlink,
8131	.ndo_bridge_setlink	= i40e_ndo_bridge_setlink,
8132#endif /* HAVE_BRIDGE_ATTRIBS */
8133};
8134
8135/**
8136 * i40e_config_netdev - Setup the netdev flags
8137 * @vsi: the VSI being configured
8138 *
8139 * Returns 0 on success, negative value on failure
8140 **/
8141static int i40e_config_netdev(struct i40e_vsi *vsi)
8142{
8143	u8 brdcast[ETH_ALEN] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
8144	struct i40e_pf *pf = vsi->back;
8145	struct i40e_hw *hw = &pf->hw;
8146	struct i40e_netdev_priv *np;
8147	struct net_device *netdev;
8148	u8 mac_addr[ETH_ALEN];
8149	int etherdev_size;
8150
8151	etherdev_size = sizeof(struct i40e_netdev_priv);
8152	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
8153	if (!netdev)
8154		return -ENOMEM;
8155
8156	vsi->netdev = netdev;
8157	np = netdev_priv(netdev);
8158	np->vsi = vsi;
8159
8160	netdev->hw_enc_features |= NETIF_F_IP_CSUM	 |
8161				  NETIF_F_GSO_UDP_TUNNEL |
8162				  NETIF_F_TSO;
8163
8164	netdev->features = NETIF_F_SG		       |
8165			   NETIF_F_IP_CSUM	       |
8166			   NETIF_F_SCTP_CSUM	       |
8167			   NETIF_F_HIGHDMA	       |
8168			   NETIF_F_GSO_UDP_TUNNEL      |
8169			   NETIF_F_HW_VLAN_CTAG_TX     |
8170			   NETIF_F_HW_VLAN_CTAG_RX     |
8171			   NETIF_F_HW_VLAN_CTAG_FILTER |
8172			   NETIF_F_IPV6_CSUM	       |
8173			   NETIF_F_TSO		       |
8174			   NETIF_F_TSO_ECN	       |
8175			   NETIF_F_TSO6		       |
8176			   NETIF_F_RXCSUM	       |
8177			   NETIF_F_RXHASH	       |
8178			   0;
8179
8180	if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
8181		netdev->features |= NETIF_F_NTUPLE;
8182
8183	/* copy netdev features into list of user selectable features */
8184	netdev->hw_features |= netdev->features;
8185
8186	if (vsi->type == I40E_VSI_MAIN) {
8187		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
8188		ether_addr_copy(mac_addr, hw->mac.perm_addr);
8189		/* The following steps are necessary to prevent reception
8190		 * of tagged packets - some older NVM configurations load a
8191		 * default a MAC-VLAN filter that accepts any tagged packet
8192		 * which must be replaced by a normal filter.
8193		 */
8194		if (!i40e_rm_default_mac_filter(vsi, mac_addr))
8195			i40e_add_filter(vsi, mac_addr,
8196					I40E_VLAN_ANY, false, true);
8197	} else {
8198		/* relate the VSI_VMDQ name to the VSI_MAIN name */
8199		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
8200			 pf->vsi[pf->lan_vsi]->netdev->name);
8201		random_ether_addr(mac_addr);
8202		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
8203	}
8204	i40e_add_filter(vsi, brdcast, I40E_VLAN_ANY, false, false);
8205
8206	ether_addr_copy(netdev->dev_addr, mac_addr);
8207	ether_addr_copy(netdev->perm_addr, mac_addr);
8208	/* vlan gets same features (except vlan offload)
8209	 * after any tweaks for specific VSI types
8210	 */
8211	netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
8212						     NETIF_F_HW_VLAN_CTAG_RX |
8213						   NETIF_F_HW_VLAN_CTAG_FILTER);
8214	netdev->priv_flags |= IFF_UNICAST_FLT;
8215	netdev->priv_flags |= IFF_SUPP_NOFCS;
8216	/* Setup netdev TC information */
8217	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
8218
8219	netdev->netdev_ops = &i40e_netdev_ops;
8220	netdev->watchdog_timeo = 5 * HZ;
8221	i40e_set_ethtool_ops(netdev);
8222#ifdef I40E_FCOE
8223	i40e_fcoe_config_netdev(netdev, vsi);
8224#endif
8225
8226	return 0;
8227}
8228
8229/**
8230 * i40e_vsi_delete - Delete a VSI from the switch
8231 * @vsi: the VSI being removed
8232 *
8233 * Returns 0 on success, negative value on failure
8234 **/
8235static void i40e_vsi_delete(struct i40e_vsi *vsi)
8236{
8237	/* remove default VSI is not allowed */
8238	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
8239		return;
8240
8241	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
8242}
8243
8244/**
8245 * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
8246 * @vsi: the VSI being queried
8247 *
8248 * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
8249 **/
8250int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
8251{
8252	struct i40e_veb *veb;
8253	struct i40e_pf *pf = vsi->back;
8254
8255	/* Uplink is not a bridge so default to VEB */
8256	if (vsi->veb_idx == I40E_NO_VEB)
8257		return 1;
8258
8259	veb = pf->veb[vsi->veb_idx];
8260	/* Uplink is a bridge in VEPA mode */
8261	if (veb && (veb->bridge_mode & BRIDGE_MODE_VEPA))
8262		return 0;
8263
8264	/* Uplink is a bridge in VEB mode */
8265	return 1;
8266}
8267
8268/**
8269 * i40e_add_vsi - Add a VSI to the switch
8270 * @vsi: the VSI being configured
8271 *
8272 * This initializes a VSI context depending on the VSI type to be added and
8273 * passes it down to the add_vsi aq command.
8274 **/
8275static int i40e_add_vsi(struct i40e_vsi *vsi)
8276{
8277	int ret = -ENODEV;
8278	struct i40e_mac_filter *f, *ftmp;
8279	struct i40e_pf *pf = vsi->back;
8280	struct i40e_hw *hw = &pf->hw;
8281	struct i40e_vsi_context ctxt;
8282	u8 enabled_tc = 0x1; /* TC0 enabled */
8283	int f_count = 0;
8284
8285	memset(&ctxt, 0, sizeof(ctxt));
8286	switch (vsi->type) {
8287	case I40E_VSI_MAIN:
8288		/* The PF's main VSI is already setup as part of the
8289		 * device initialization, so we'll not bother with
8290		 * the add_vsi call, but we will retrieve the current
8291		 * VSI context.
8292		 */
8293		ctxt.seid = pf->main_vsi_seid;
8294		ctxt.pf_num = pf->hw.pf_id;
8295		ctxt.vf_num = 0;
8296		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
8297		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8298		if (ret) {
8299			dev_info(&pf->pdev->dev,
8300				 "couldn't get PF vsi config, err %d, aq_err %d\n",
8301				 ret, pf->hw.aq.asq_last_status);
8302			return -ENOENT;
8303		}
8304		vsi->info = ctxt.info;
8305		vsi->info.valid_sections = 0;
8306
8307		vsi->seid = ctxt.seid;
8308		vsi->id = ctxt.vsi_number;
8309
8310		enabled_tc = i40e_pf_get_tc_map(pf);
8311
8312		/* MFP mode setup queue map and update VSI */
8313		if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
8314		    !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
8315			memset(&ctxt, 0, sizeof(ctxt));
8316			ctxt.seid = pf->main_vsi_seid;
8317			ctxt.pf_num = pf->hw.pf_id;
8318			ctxt.vf_num = 0;
8319			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
8320			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
8321			if (ret) {
8322				dev_info(&pf->pdev->dev,
8323					 "update vsi failed, aq_err=%d\n",
8324					 pf->hw.aq.asq_last_status);
8325				ret = -ENOENT;
8326				goto err;
8327			}
8328			/* update the local VSI info queue map */
8329			i40e_vsi_update_queue_map(vsi, &ctxt);
8330			vsi->info.valid_sections = 0;
8331		} else {
8332			/* Default/Main VSI is only enabled for TC0
8333			 * reconfigure it to enable all TCs that are
8334			 * available on the port in SFP mode.
8335			 * For MFP case the iSCSI PF would use this
8336			 * flow to enable LAN+iSCSI TC.
8337			 */
8338			ret = i40e_vsi_config_tc(vsi, enabled_tc);
8339			if (ret) {
8340				dev_info(&pf->pdev->dev,
8341					 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
8342					 enabled_tc, ret,
8343					 pf->hw.aq.asq_last_status);
8344				ret = -ENOENT;
8345			}
8346		}
8347		break;
8348
8349	case I40E_VSI_FDIR:
8350		ctxt.pf_num = hw->pf_id;
8351		ctxt.vf_num = 0;
8352		ctxt.uplink_seid = vsi->uplink_seid;
8353		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8354		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
8355		if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
8356		    (i40e_is_vsi_uplink_mode_veb(vsi))) {
8357			ctxt.info.valid_sections |=
8358			     cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8359			ctxt.info.switch_id =
8360			   cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8361		}
8362		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8363		break;
8364
8365	case I40E_VSI_VMDQ2:
8366		ctxt.pf_num = hw->pf_id;
8367		ctxt.vf_num = 0;
8368		ctxt.uplink_seid = vsi->uplink_seid;
8369		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8370		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
8371
8372		/* This VSI is connected to VEB so the switch_id
8373		 * should be set to zero by default.
8374		 */
8375		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8376			ctxt.info.valid_sections |=
8377				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8378			ctxt.info.switch_id =
8379				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8380		}
8381
8382		/* Setup the VSI tx/rx queue map for TC0 only for now */
8383		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8384		break;
8385
8386	case I40E_VSI_SRIOV:
8387		ctxt.pf_num = hw->pf_id;
8388		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
8389		ctxt.uplink_seid = vsi->uplink_seid;
8390		ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
8391		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
8392
8393		/* This VSI is connected to VEB so the switch_id
8394		 * should be set to zero by default.
8395		 */
8396		if (i40e_is_vsi_uplink_mode_veb(vsi)) {
8397			ctxt.info.valid_sections |=
8398				cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
8399			ctxt.info.switch_id =
8400				cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
8401		}
8402
8403		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
8404		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
8405		if (pf->vf[vsi->vf_id].spoofchk) {
8406			ctxt.info.valid_sections |=
8407				cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
8408			ctxt.info.sec_flags |=
8409				(I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
8410				 I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
8411		}
8412		/* Setup the VSI tx/rx queue map for TC0 only for now */
8413		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
8414		break;
8415
8416#ifdef I40E_FCOE
8417	case I40E_VSI_FCOE:
8418		ret = i40e_fcoe_vsi_init(vsi, &ctxt);
8419		if (ret) {
8420			dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
8421			return ret;
8422		}
8423		break;
8424
8425#endif /* I40E_FCOE */
8426	default:
8427		return -ENODEV;
8428	}
8429
8430	if (vsi->type != I40E_VSI_MAIN) {
8431		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
8432		if (ret) {
8433			dev_info(&vsi->back->pdev->dev,
8434				 "add vsi failed, aq_err=%d\n",
8435				 vsi->back->hw.aq.asq_last_status);
8436			ret = -ENOENT;
8437			goto err;
8438		}
8439		vsi->info = ctxt.info;
8440		vsi->info.valid_sections = 0;
8441		vsi->seid = ctxt.seid;
8442		vsi->id = ctxt.vsi_number;
8443	}
8444
8445	/* If macvlan filters already exist, force them to get loaded */
8446	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
8447		f->changed = true;
8448		f_count++;
8449
8450		if (f->is_laa && vsi->type == I40E_VSI_MAIN) {
8451			struct i40e_aqc_remove_macvlan_element_data element;
8452
8453			memset(&element, 0, sizeof(element));
8454			ether_addr_copy(element.mac_addr, f->macaddr);
8455			element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
8456			ret = i40e_aq_remove_macvlan(hw, vsi->seid,
8457						     &element, 1, NULL);
8458			if (ret) {
8459				/* some older FW has a different default */
8460				element.flags |=
8461					       I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
8462				i40e_aq_remove_macvlan(hw, vsi->seid,
8463						       &element, 1, NULL);
8464			}
8465
8466			i40e_aq_mac_address_write(hw,
8467						  I40E_AQC_WRITE_TYPE_LAA_WOL,
8468						  f->macaddr, NULL);
8469		}
8470	}
8471	if (f_count) {
8472		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
8473		pf->flags |= I40E_FLAG_FILTER_SYNC;
8474	}
8475
8476	/* Update VSI BW information */
8477	ret = i40e_vsi_get_bw_info(vsi);
8478	if (ret) {
8479		dev_info(&pf->pdev->dev,
8480			 "couldn't get vsi bw info, err %d, aq_err %d\n",
8481			 ret, pf->hw.aq.asq_last_status);
8482		/* VSI is already added so not tearing that up */
8483		ret = 0;
8484	}
8485
8486err:
8487	return ret;
8488}
8489
8490/**
8491 * i40e_vsi_release - Delete a VSI and free its resources
8492 * @vsi: the VSI being removed
8493 *
8494 * Returns 0 on success or < 0 on error
8495 **/
8496int i40e_vsi_release(struct i40e_vsi *vsi)
8497{
8498	struct i40e_mac_filter *f, *ftmp;
8499	struct i40e_veb *veb = NULL;
8500	struct i40e_pf *pf;
8501	u16 uplink_seid;
8502	int i, n;
8503
8504	pf = vsi->back;
8505
8506	/* release of a VEB-owner or last VSI is not allowed */
8507	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
8508		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
8509			 vsi->seid, vsi->uplink_seid);
8510		return -ENODEV;
8511	}
8512	if (vsi == pf->vsi[pf->lan_vsi] &&
8513	    !test_bit(__I40E_DOWN, &pf->state)) {
8514		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
8515		return -ENODEV;
8516	}
8517
8518	uplink_seid = vsi->uplink_seid;
8519	if (vsi->type != I40E_VSI_SRIOV) {
8520		if (vsi->netdev_registered) {
8521			vsi->netdev_registered = false;
8522			if (vsi->netdev) {
8523				/* results in a call to i40e_close() */
8524				unregister_netdev(vsi->netdev);
8525			}
8526		} else {
8527			i40e_vsi_close(vsi);
8528		}
8529		i40e_vsi_disable_irq(vsi);
8530	}
8531
8532	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
8533		i40e_del_filter(vsi, f->macaddr, f->vlan,
8534				f->is_vf, f->is_netdev);
8535	i40e_sync_vsi_filters(vsi);
8536
8537	i40e_vsi_delete(vsi);
8538	i40e_vsi_free_q_vectors(vsi);
8539	if (vsi->netdev) {
8540		free_netdev(vsi->netdev);
8541		vsi->netdev = NULL;
8542	}
8543	i40e_vsi_clear_rings(vsi);
8544	i40e_vsi_clear(vsi);
8545
8546	/* If this was the last thing on the VEB, except for the
8547	 * controlling VSI, remove the VEB, which puts the controlling
8548	 * VSI onto the next level down in the switch.
8549	 *
8550	 * Well, okay, there's one more exception here: don't remove
8551	 * the orphan VEBs yet.  We'll wait for an explicit remove request
8552	 * from up the network stack.
8553	 */
8554	for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
8555		if (pf->vsi[i] &&
8556		    pf->vsi[i]->uplink_seid == uplink_seid &&
8557		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
8558			n++;      /* count the VSIs */
8559		}
8560	}
8561	for (i = 0; i < I40E_MAX_VEB; i++) {
8562		if (!pf->veb[i])
8563			continue;
8564		if (pf->veb[i]->uplink_seid == uplink_seid)
8565			n++;     /* count the VEBs */
8566		if (pf->veb[i]->seid == uplink_seid)
8567			veb = pf->veb[i];
8568	}
8569	if (n == 0 && veb && veb->uplink_seid != 0)
8570		i40e_veb_release(veb);
8571
8572	return 0;
8573}
8574
8575/**
8576 * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
8577 * @vsi: ptr to the VSI
8578 *
8579 * This should only be called after i40e_vsi_mem_alloc() which allocates the
8580 * corresponding SW VSI structure and initializes num_queue_pairs for the
8581 * newly allocated VSI.
8582 *
8583 * Returns 0 on success or negative on failure
8584 **/
8585static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
8586{
8587	int ret = -ENOENT;
8588	struct i40e_pf *pf = vsi->back;
8589
8590	if (vsi->q_vectors[0]) {
8591		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
8592			 vsi->seid);
8593		return -EEXIST;
8594	}
8595
8596	if (vsi->base_vector) {
8597		dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
8598			 vsi->seid, vsi->base_vector);
8599		return -EEXIST;
8600	}
8601
8602	ret = i40e_vsi_alloc_q_vectors(vsi);
8603	if (ret) {
8604		dev_info(&pf->pdev->dev,
8605			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
8606			 vsi->num_q_vectors, vsi->seid, ret);
8607		vsi->num_q_vectors = 0;
8608		goto vector_setup_out;
8609	}
8610
8611	if (vsi->num_q_vectors)
8612		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
8613						 vsi->num_q_vectors, vsi->idx);
8614	if (vsi->base_vector < 0) {
8615		dev_info(&pf->pdev->dev,
8616			 "failed to get tracking for %d vectors for VSI %d, err=%d\n",
8617			 vsi->num_q_vectors, vsi->seid, vsi->base_vector);
8618		i40e_vsi_free_q_vectors(vsi);
8619		ret = -ENOENT;
8620		goto vector_setup_out;
8621	}
8622
8623vector_setup_out:
8624	return ret;
8625}
8626
8627/**
8628 * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
8629 * @vsi: pointer to the vsi.
8630 *
8631 * This re-allocates a vsi's queue resources.
8632 *
8633 * Returns pointer to the successfully allocated and configured VSI sw struct
8634 * on success, otherwise returns NULL on failure.
8635 **/
8636static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
8637{
8638	struct i40e_pf *pf = vsi->back;
8639	u8 enabled_tc;
8640	int ret;
8641
8642	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
8643	i40e_vsi_clear_rings(vsi);
8644
8645	i40e_vsi_free_arrays(vsi, false);
8646	i40e_set_num_rings_in_vsi(vsi);
8647	ret = i40e_vsi_alloc_arrays(vsi, false);
8648	if (ret)
8649		goto err_vsi;
8650
8651	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
8652	if (ret < 0) {
8653		dev_info(&pf->pdev->dev,
8654			 "failed to get tracking for %d queues for VSI %d err=%d\n",
8655			 vsi->alloc_queue_pairs, vsi->seid, ret);
8656		goto err_vsi;
8657	}
8658	vsi->base_queue = ret;
8659
8660	/* Update the FW view of the VSI. Force a reset of TC and queue
8661	 * layout configurations.
8662	 */
8663	enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
8664	pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
8665	pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
8666	i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
8667
8668	/* assign it some queues */
8669	ret = i40e_alloc_rings(vsi);
8670	if (ret)
8671		goto err_rings;
8672
8673	/* map all of the rings to the q_vectors */
8674	i40e_vsi_map_rings_to_vectors(vsi);
8675	return vsi;
8676
8677err_rings:
8678	i40e_vsi_free_q_vectors(vsi);
8679	if (vsi->netdev_registered) {
8680		vsi->netdev_registered = false;
8681		unregister_netdev(vsi->netdev);
8682		free_netdev(vsi->netdev);
8683		vsi->netdev = NULL;
8684	}
8685	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
8686err_vsi:
8687	i40e_vsi_clear(vsi);
8688	return NULL;
8689}
8690
8691/**
8692 * i40e_vsi_setup - Set up a VSI by a given type
8693 * @pf: board private structure
8694 * @type: VSI type
8695 * @uplink_seid: the switch element to link to
8696 * @param1: usage depends upon VSI type. For VF types, indicates VF id
8697 *
8698 * This allocates the sw VSI structure and its queue resources, then add a VSI
8699 * to the identified VEB.
8700 *
8701 * Returns pointer to the successfully allocated and configure VSI sw struct on
8702 * success, otherwise returns NULL on failure.
8703 **/
8704struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
8705				u16 uplink_seid, u32 param1)
8706{
8707	struct i40e_vsi *vsi = NULL;
8708	struct i40e_veb *veb = NULL;
8709	int ret, i;
8710	int v_idx;
8711
8712	/* The requested uplink_seid must be either
8713	 *     - the PF's port seid
8714	 *              no VEB is needed because this is the PF
8715	 *              or this is a Flow Director special case VSI
8716	 *     - seid of an existing VEB
8717	 *     - seid of a VSI that owns an existing VEB
8718	 *     - seid of a VSI that doesn't own a VEB
8719	 *              a new VEB is created and the VSI becomes the owner
8720	 *     - seid of the PF VSI, which is what creates the first VEB
8721	 *              this is a special case of the previous
8722	 *
8723	 * Find which uplink_seid we were given and create a new VEB if needed
8724	 */
8725	for (i = 0; i < I40E_MAX_VEB; i++) {
8726		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
8727			veb = pf->veb[i];
8728			break;
8729		}
8730	}
8731
8732	if (!veb && uplink_seid != pf->mac_seid) {
8733
8734		for (i = 0; i < pf->num_alloc_vsi; i++) {
8735			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
8736				vsi = pf->vsi[i];
8737				break;
8738			}
8739		}
8740		if (!vsi) {
8741			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
8742				 uplink_seid);
8743			return NULL;
8744		}
8745
8746		if (vsi->uplink_seid == pf->mac_seid)
8747			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
8748					     vsi->tc_config.enabled_tc);
8749		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
8750			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8751					     vsi->tc_config.enabled_tc);
8752		if (veb) {
8753			if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
8754				dev_info(&vsi->back->pdev->dev,
8755					 "%s: New VSI creation error, uplink seid of LAN VSI expected.\n",
8756					 __func__);
8757				return NULL;
8758			}
8759			/* We come up by default in VEPA mode if SRIOV is not
8760			 * already enabled, in which case we can't force VEPA
8761			 * mode.
8762			 */
8763			if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
8764				veb->bridge_mode = BRIDGE_MODE_VEPA;
8765				pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8766			}
8767			i40e_config_bridge_mode(veb);
8768		}
8769		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8770			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8771				veb = pf->veb[i];
8772		}
8773		if (!veb) {
8774			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
8775			return NULL;
8776		}
8777
8778		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
8779		uplink_seid = veb->seid;
8780	}
8781
8782	/* get vsi sw struct */
8783	v_idx = i40e_vsi_mem_alloc(pf, type);
8784	if (v_idx < 0)
8785		goto err_alloc;
8786	vsi = pf->vsi[v_idx];
8787	if (!vsi)
8788		goto err_alloc;
8789	vsi->type = type;
8790	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
8791
8792	if (type == I40E_VSI_MAIN)
8793		pf->lan_vsi = v_idx;
8794	else if (type == I40E_VSI_SRIOV)
8795		vsi->vf_id = param1;
8796	/* assign it some queues */
8797	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
8798				vsi->idx);
8799	if (ret < 0) {
8800		dev_info(&pf->pdev->dev,
8801			 "failed to get tracking for %d queues for VSI %d err=%d\n",
8802			 vsi->alloc_queue_pairs, vsi->seid, ret);
8803		goto err_vsi;
8804	}
8805	vsi->base_queue = ret;
8806
8807	/* get a VSI from the hardware */
8808	vsi->uplink_seid = uplink_seid;
8809	ret = i40e_add_vsi(vsi);
8810	if (ret)
8811		goto err_vsi;
8812
8813	switch (vsi->type) {
8814	/* setup the netdev if needed */
8815	case I40E_VSI_MAIN:
8816	case I40E_VSI_VMDQ2:
8817	case I40E_VSI_FCOE:
8818		ret = i40e_config_netdev(vsi);
8819		if (ret)
8820			goto err_netdev;
8821		ret = register_netdev(vsi->netdev);
8822		if (ret)
8823			goto err_netdev;
8824		vsi->netdev_registered = true;
8825		netif_carrier_off(vsi->netdev);
8826#ifdef CONFIG_I40E_DCB
8827		/* Setup DCB netlink interface */
8828		i40e_dcbnl_setup(vsi);
8829#endif /* CONFIG_I40E_DCB */
8830		/* fall through */
8831
8832	case I40E_VSI_FDIR:
8833		/* set up vectors and rings if needed */
8834		ret = i40e_vsi_setup_vectors(vsi);
8835		if (ret)
8836			goto err_msix;
8837
8838		ret = i40e_alloc_rings(vsi);
8839		if (ret)
8840			goto err_rings;
8841
8842		/* map all of the rings to the q_vectors */
8843		i40e_vsi_map_rings_to_vectors(vsi);
8844
8845		i40e_vsi_reset_stats(vsi);
8846		break;
8847
8848	default:
8849		/* no netdev or rings for the other VSI types */
8850		break;
8851	}
8852
8853	return vsi;
8854
8855err_rings:
8856	i40e_vsi_free_q_vectors(vsi);
8857err_msix:
8858	if (vsi->netdev_registered) {
8859		vsi->netdev_registered = false;
8860		unregister_netdev(vsi->netdev);
8861		free_netdev(vsi->netdev);
8862		vsi->netdev = NULL;
8863	}
8864err_netdev:
8865	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
8866err_vsi:
8867	i40e_vsi_clear(vsi);
8868err_alloc:
8869	return NULL;
8870}
8871
8872/**
8873 * i40e_veb_get_bw_info - Query VEB BW information
8874 * @veb: the veb to query
8875 *
8876 * Query the Tx scheduler BW configuration data for given VEB
8877 **/
8878static int i40e_veb_get_bw_info(struct i40e_veb *veb)
8879{
8880	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
8881	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
8882	struct i40e_pf *pf = veb->pf;
8883	struct i40e_hw *hw = &pf->hw;
8884	u32 tc_bw_max;
8885	int ret = 0;
8886	int i;
8887
8888	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
8889						  &bw_data, NULL);
8890	if (ret) {
8891		dev_info(&pf->pdev->dev,
8892			 "query veb bw config failed, aq_err=%d\n",
8893			 hw->aq.asq_last_status);
8894		goto out;
8895	}
8896
8897	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
8898						   &ets_data, NULL);
8899	if (ret) {
8900		dev_info(&pf->pdev->dev,
8901			 "query veb bw ets config failed, aq_err=%d\n",
8902			 hw->aq.asq_last_status);
8903		goto out;
8904	}
8905
8906	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
8907	veb->bw_max_quanta = ets_data.tc_bw_max;
8908	veb->is_abs_credits = bw_data.absolute_credits_enable;
8909	veb->enabled_tc = ets_data.tc_valid_bits;
8910	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
8911		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
8912	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
8913		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
8914		veb->bw_tc_limit_credits[i] =
8915					le16_to_cpu(bw_data.tc_bw_limits[i]);
8916		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
8917	}
8918
8919out:
8920	return ret;
8921}
8922
8923/**
8924 * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
8925 * @pf: board private structure
8926 *
8927 * On error: returns error code (negative)
8928 * On success: returns vsi index in PF (positive)
8929 **/
8930static int i40e_veb_mem_alloc(struct i40e_pf *pf)
8931{
8932	int ret = -ENOENT;
8933	struct i40e_veb *veb;
8934	int i;
8935
8936	/* Need to protect the allocation of switch elements at the PF level */
8937	mutex_lock(&pf->switch_mutex);
8938
8939	/* VEB list may be fragmented if VEB creation/destruction has
8940	 * been happening.  We can afford to do a quick scan to look
8941	 * for any free slots in the list.
8942	 *
8943	 * find next empty veb slot, looping back around if necessary
8944	 */
8945	i = 0;
8946	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
8947		i++;
8948	if (i >= I40E_MAX_VEB) {
8949		ret = -ENOMEM;
8950		goto err_alloc_veb;  /* out of VEB slots! */
8951	}
8952
8953	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
8954	if (!veb) {
8955		ret = -ENOMEM;
8956		goto err_alloc_veb;
8957	}
8958	veb->pf = pf;
8959	veb->idx = i;
8960	veb->enabled_tc = 1;
8961
8962	pf->veb[i] = veb;
8963	ret = i;
8964err_alloc_veb:
8965	mutex_unlock(&pf->switch_mutex);
8966	return ret;
8967}
8968
8969/**
8970 * i40e_switch_branch_release - Delete a branch of the switch tree
8971 * @branch: where to start deleting
8972 *
8973 * This uses recursion to find the tips of the branch to be
8974 * removed, deleting until we get back to and can delete this VEB.
8975 **/
8976static void i40e_switch_branch_release(struct i40e_veb *branch)
8977{
8978	struct i40e_pf *pf = branch->pf;
8979	u16 branch_seid = branch->seid;
8980	u16 veb_idx = branch->idx;
8981	int i;
8982
8983	/* release any VEBs on this VEB - RECURSION */
8984	for (i = 0; i < I40E_MAX_VEB; i++) {
8985		if (!pf->veb[i])
8986			continue;
8987		if (pf->veb[i]->uplink_seid == branch->seid)
8988			i40e_switch_branch_release(pf->veb[i]);
8989	}
8990
8991	/* Release the VSIs on this VEB, but not the owner VSI.
8992	 *
8993	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
8994	 *       the VEB itself, so don't use (*branch) after this loop.
8995	 */
8996	for (i = 0; i < pf->num_alloc_vsi; i++) {
8997		if (!pf->vsi[i])
8998			continue;
8999		if (pf->vsi[i]->uplink_seid == branch_seid &&
9000		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9001			i40e_vsi_release(pf->vsi[i]);
9002		}
9003	}
9004
9005	/* There's one corner case where the VEB might not have been
9006	 * removed, so double check it here and remove it if needed.
9007	 * This case happens if the veb was created from the debugfs
9008	 * commands and no VSIs were added to it.
9009	 */
9010	if (pf->veb[veb_idx])
9011		i40e_veb_release(pf->veb[veb_idx]);
9012}
9013
9014/**
9015 * i40e_veb_clear - remove veb struct
9016 * @veb: the veb to remove
9017 **/
9018static void i40e_veb_clear(struct i40e_veb *veb)
9019{
9020	if (!veb)
9021		return;
9022
9023	if (veb->pf) {
9024		struct i40e_pf *pf = veb->pf;
9025
9026		mutex_lock(&pf->switch_mutex);
9027		if (pf->veb[veb->idx] == veb)
9028			pf->veb[veb->idx] = NULL;
9029		mutex_unlock(&pf->switch_mutex);
9030	}
9031
9032	kfree(veb);
9033}
9034
9035/**
9036 * i40e_veb_release - Delete a VEB and free its resources
9037 * @veb: the VEB being removed
9038 **/
9039void i40e_veb_release(struct i40e_veb *veb)
9040{
9041	struct i40e_vsi *vsi = NULL;
9042	struct i40e_pf *pf;
9043	int i, n = 0;
9044
9045	pf = veb->pf;
9046
9047	/* find the remaining VSI and check for extras */
9048	for (i = 0; i < pf->num_alloc_vsi; i++) {
9049		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
9050			n++;
9051			vsi = pf->vsi[i];
9052		}
9053	}
9054	if (n != 1) {
9055		dev_info(&pf->pdev->dev,
9056			 "can't remove VEB %d with %d VSIs left\n",
9057			 veb->seid, n);
9058		return;
9059	}
9060
9061	/* move the remaining VSI to uplink veb */
9062	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
9063	if (veb->uplink_seid) {
9064		vsi->uplink_seid = veb->uplink_seid;
9065		if (veb->uplink_seid == pf->mac_seid)
9066			vsi->veb_idx = I40E_NO_VEB;
9067		else
9068			vsi->veb_idx = veb->veb_idx;
9069	} else {
9070		/* floating VEB */
9071		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
9072		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
9073	}
9074
9075	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
9076	i40e_veb_clear(veb);
9077}
9078
9079/**
9080 * i40e_add_veb - create the VEB in the switch
9081 * @veb: the VEB to be instantiated
9082 * @vsi: the controlling VSI
9083 **/
9084static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
9085{
9086	bool is_default = false;
9087	bool is_cloud = false;
9088	int ret;
9089
9090	/* get a VEB from the hardware */
9091	ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
9092			      veb->enabled_tc, is_default,
9093			      is_cloud, &veb->seid, NULL);
9094	if (ret) {
9095		dev_info(&veb->pf->pdev->dev,
9096			 "couldn't add VEB, err %d, aq_err %d\n",
9097			 ret, veb->pf->hw.aq.asq_last_status);
9098		return -EPERM;
9099	}
9100
9101	/* get statistics counter */
9102	ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
9103					 &veb->stats_idx, NULL, NULL, NULL);
9104	if (ret) {
9105		dev_info(&veb->pf->pdev->dev,
9106			 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
9107			 ret, veb->pf->hw.aq.asq_last_status);
9108		return -EPERM;
9109	}
9110	ret = i40e_veb_get_bw_info(veb);
9111	if (ret) {
9112		dev_info(&veb->pf->pdev->dev,
9113			 "couldn't get VEB bw info, err %d, aq_err %d\n",
9114			 ret, veb->pf->hw.aq.asq_last_status);
9115		i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
9116		return -ENOENT;
9117	}
9118
9119	vsi->uplink_seid = veb->seid;
9120	vsi->veb_idx = veb->idx;
9121	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9122
9123	return 0;
9124}
9125
9126/**
9127 * i40e_veb_setup - Set up a VEB
9128 * @pf: board private structure
9129 * @flags: VEB setup flags
9130 * @uplink_seid: the switch element to link to
9131 * @vsi_seid: the initial VSI seid
9132 * @enabled_tc: Enabled TC bit-map
9133 *
9134 * This allocates the sw VEB structure and links it into the switch
9135 * It is possible and legal for this to be a duplicate of an already
9136 * existing VEB.  It is also possible for both uplink and vsi seids
9137 * to be zero, in order to create a floating VEB.
9138 *
9139 * Returns pointer to the successfully allocated VEB sw struct on
9140 * success, otherwise returns NULL on failure.
9141 **/
9142struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
9143				u16 uplink_seid, u16 vsi_seid,
9144				u8 enabled_tc)
9145{
9146	struct i40e_veb *veb, *uplink_veb = NULL;
9147	int vsi_idx, veb_idx;
9148	int ret;
9149
9150	/* if one seid is 0, the other must be 0 to create a floating relay */
9151	if ((uplink_seid == 0 || vsi_seid == 0) &&
9152	    (uplink_seid + vsi_seid != 0)) {
9153		dev_info(&pf->pdev->dev,
9154			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
9155			 uplink_seid, vsi_seid);
9156		return NULL;
9157	}
9158
9159	/* make sure there is such a vsi and uplink */
9160	for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
9161		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
9162			break;
9163	if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
9164		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
9165			 vsi_seid);
9166		return NULL;
9167	}
9168
9169	if (uplink_seid && uplink_seid != pf->mac_seid) {
9170		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
9171			if (pf->veb[veb_idx] &&
9172			    pf->veb[veb_idx]->seid == uplink_seid) {
9173				uplink_veb = pf->veb[veb_idx];
9174				break;
9175			}
9176		}
9177		if (!uplink_veb) {
9178			dev_info(&pf->pdev->dev,
9179				 "uplink seid %d not found\n", uplink_seid);
9180			return NULL;
9181		}
9182	}
9183
9184	/* get veb sw struct */
9185	veb_idx = i40e_veb_mem_alloc(pf);
9186	if (veb_idx < 0)
9187		goto err_alloc;
9188	veb = pf->veb[veb_idx];
9189	veb->flags = flags;
9190	veb->uplink_seid = uplink_seid;
9191	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
9192	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
9193
9194	/* create the VEB in the switch */
9195	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
9196	if (ret)
9197		goto err_veb;
9198	if (vsi_idx == pf->lan_vsi)
9199		pf->lan_veb = veb->idx;
9200
9201	return veb;
9202
9203err_veb:
9204	i40e_veb_clear(veb);
9205err_alloc:
9206	return NULL;
9207}
9208
9209/**
9210 * i40e_setup_pf_switch_element - set PF vars based on switch type
9211 * @pf: board private structure
9212 * @ele: element we are building info from
9213 * @num_reported: total number of elements
9214 * @printconfig: should we print the contents
9215 *
9216 * helper function to assist in extracting a few useful SEID values.
9217 **/
9218static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
9219				struct i40e_aqc_switch_config_element_resp *ele,
9220				u16 num_reported, bool printconfig)
9221{
9222	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
9223	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
9224	u8 element_type = ele->element_type;
9225	u16 seid = le16_to_cpu(ele->seid);
9226
9227	if (printconfig)
9228		dev_info(&pf->pdev->dev,
9229			 "type=%d seid=%d uplink=%d downlink=%d\n",
9230			 element_type, seid, uplink_seid, downlink_seid);
9231
9232	switch (element_type) {
9233	case I40E_SWITCH_ELEMENT_TYPE_MAC:
9234		pf->mac_seid = seid;
9235		break;
9236	case I40E_SWITCH_ELEMENT_TYPE_VEB:
9237		/* Main VEB? */
9238		if (uplink_seid != pf->mac_seid)
9239			break;
9240		if (pf->lan_veb == I40E_NO_VEB) {
9241			int v;
9242
9243			/* find existing or else empty VEB */
9244			for (v = 0; v < I40E_MAX_VEB; v++) {
9245				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
9246					pf->lan_veb = v;
9247					break;
9248				}
9249			}
9250			if (pf->lan_veb == I40E_NO_VEB) {
9251				v = i40e_veb_mem_alloc(pf);
9252				if (v < 0)
9253					break;
9254				pf->lan_veb = v;
9255			}
9256		}
9257
9258		pf->veb[pf->lan_veb]->seid = seid;
9259		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
9260		pf->veb[pf->lan_veb]->pf = pf;
9261		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
9262		break;
9263	case I40E_SWITCH_ELEMENT_TYPE_VSI:
9264		if (num_reported != 1)
9265			break;
9266		/* This is immediately after a reset so we can assume this is
9267		 * the PF's VSI
9268		 */
9269		pf->mac_seid = uplink_seid;
9270		pf->pf_seid = downlink_seid;
9271		pf->main_vsi_seid = seid;
9272		if (printconfig)
9273			dev_info(&pf->pdev->dev,
9274				 "pf_seid=%d main_vsi_seid=%d\n",
9275				 pf->pf_seid, pf->main_vsi_seid);
9276		break;
9277	case I40E_SWITCH_ELEMENT_TYPE_PF:
9278	case I40E_SWITCH_ELEMENT_TYPE_VF:
9279	case I40E_SWITCH_ELEMENT_TYPE_EMP:
9280	case I40E_SWITCH_ELEMENT_TYPE_BMC:
9281	case I40E_SWITCH_ELEMENT_TYPE_PE:
9282	case I40E_SWITCH_ELEMENT_TYPE_PA:
9283		/* ignore these for now */
9284		break;
9285	default:
9286		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
9287			 element_type, seid);
9288		break;
9289	}
9290}
9291
9292/**
9293 * i40e_fetch_switch_configuration - Get switch config from firmware
9294 * @pf: board private structure
9295 * @printconfig: should we print the contents
9296 *
9297 * Get the current switch configuration from the device and
9298 * extract a few useful SEID values.
9299 **/
9300int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
9301{
9302	struct i40e_aqc_get_switch_config_resp *sw_config;
9303	u16 next_seid = 0;
9304	int ret = 0;
9305	u8 *aq_buf;
9306	int i;
9307
9308	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
9309	if (!aq_buf)
9310		return -ENOMEM;
9311
9312	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
9313	do {
9314		u16 num_reported, num_total;
9315
9316		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
9317						I40E_AQ_LARGE_BUF,
9318						&next_seid, NULL);
9319		if (ret) {
9320			dev_info(&pf->pdev->dev,
9321				 "get switch config failed %d aq_err=%x\n",
9322				 ret, pf->hw.aq.asq_last_status);
9323			kfree(aq_buf);
9324			return -ENOENT;
9325		}
9326
9327		num_reported = le16_to_cpu(sw_config->header.num_reported);
9328		num_total = le16_to_cpu(sw_config->header.num_total);
9329
9330		if (printconfig)
9331			dev_info(&pf->pdev->dev,
9332				 "header: %d reported %d total\n",
9333				 num_reported, num_total);
9334
9335		for (i = 0; i < num_reported; i++) {
9336			struct i40e_aqc_switch_config_element_resp *ele =
9337				&sw_config->element[i];
9338
9339			i40e_setup_pf_switch_element(pf, ele, num_reported,
9340						     printconfig);
9341		}
9342	} while (next_seid != 0);
9343
9344	kfree(aq_buf);
9345	return ret;
9346}
9347
9348/**
9349 * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
9350 * @pf: board private structure
9351 * @reinit: if the Main VSI needs to re-initialized.
9352 *
9353 * Returns 0 on success, negative value on failure
9354 **/
9355static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
9356{
9357	int ret;
9358
9359	/* find out what's out there already */
9360	ret = i40e_fetch_switch_configuration(pf, false);
9361	if (ret) {
9362		dev_info(&pf->pdev->dev,
9363			 "couldn't fetch switch config, err %d, aq_err %d\n",
9364			 ret, pf->hw.aq.asq_last_status);
9365		return ret;
9366	}
9367	i40e_pf_reset_stats(pf);
9368
9369	/* first time setup */
9370	if (pf->lan_vsi == I40E_NO_VSI || reinit) {
9371		struct i40e_vsi *vsi = NULL;
9372		u16 uplink_seid;
9373
9374		/* Set up the PF VSI associated with the PF's main VSI
9375		 * that is already in the HW switch
9376		 */
9377		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
9378			uplink_seid = pf->veb[pf->lan_veb]->seid;
9379		else
9380			uplink_seid = pf->mac_seid;
9381		if (pf->lan_vsi == I40E_NO_VSI)
9382			vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
9383		else if (reinit)
9384			vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
9385		if (!vsi) {
9386			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
9387			i40e_fdir_teardown(pf);
9388			return -EAGAIN;
9389		}
9390	} else {
9391		/* force a reset of TC and queue layout configurations */
9392		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9393		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9394		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9395		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9396	}
9397	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
9398
9399	i40e_fdir_sb_setup(pf);
9400
9401	/* Setup static PF queue filter control settings */
9402	ret = i40e_setup_pf_filter_control(pf);
9403	if (ret) {
9404		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
9405			 ret);
9406		/* Failure here should not stop continuing other steps */
9407	}
9408
9409	/* enable RSS in the HW, even for only one queue, as the stack can use
9410	 * the hash
9411	 */
9412	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
9413		i40e_config_rss(pf);
9414
9415	/* fill in link information and enable LSE reporting */
9416	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
9417	i40e_link_event(pf);
9418
9419	/* Initialize user-specific link properties */
9420	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
9421				  I40E_AQ_AN_COMPLETED) ? true : false);
9422
9423	i40e_ptp_init(pf);
9424
9425	return ret;
9426}
9427
9428/**
9429 * i40e_determine_queue_usage - Work out queue distribution
9430 * @pf: board private structure
9431 **/
9432static void i40e_determine_queue_usage(struct i40e_pf *pf)
9433{
9434	int queues_left;
9435
9436	pf->num_lan_qps = 0;
9437#ifdef I40E_FCOE
9438	pf->num_fcoe_qps = 0;
9439#endif
9440
9441	/* Find the max queues to be put into basic use.  We'll always be
9442	 * using TC0, whether or not DCB is running, and TC0 will get the
9443	 * big RSS set.
9444	 */
9445	queues_left = pf->hw.func_caps.num_tx_qp;
9446
9447	if ((queues_left == 1) ||
9448	    !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
9449		/* one qp for PF, no queues for anything else */
9450		queues_left = 0;
9451		pf->rss_size = pf->num_lan_qps = 1;
9452
9453		/* make sure all the fancies are disabled */
9454		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
9455#ifdef I40E_FCOE
9456			       I40E_FLAG_FCOE_ENABLED	|
9457#endif
9458			       I40E_FLAG_FD_SB_ENABLED	|
9459			       I40E_FLAG_FD_ATR_ENABLED	|
9460			       I40E_FLAG_DCB_CAPABLE	|
9461			       I40E_FLAG_SRIOV_ENABLED	|
9462			       I40E_FLAG_VMDQ_ENABLED);
9463	} else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
9464				  I40E_FLAG_FD_SB_ENABLED |
9465				  I40E_FLAG_FD_ATR_ENABLED |
9466				  I40E_FLAG_DCB_CAPABLE))) {
9467		/* one qp for PF */
9468		pf->rss_size = pf->num_lan_qps = 1;
9469		queues_left -= pf->num_lan_qps;
9470
9471		pf->flags &= ~(I40E_FLAG_RSS_ENABLED	|
9472#ifdef I40E_FCOE
9473			       I40E_FLAG_FCOE_ENABLED	|
9474#endif
9475			       I40E_FLAG_FD_SB_ENABLED	|
9476			       I40E_FLAG_FD_ATR_ENABLED	|
9477			       I40E_FLAG_DCB_ENABLED	|
9478			       I40E_FLAG_VMDQ_ENABLED);
9479	} else {
9480		/* Not enough queues for all TCs */
9481		if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
9482		    (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
9483			pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
9484			dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
9485		}
9486		pf->num_lan_qps = max_t(int, pf->rss_size_max,
9487					num_online_cpus());
9488		pf->num_lan_qps = min_t(int, pf->num_lan_qps,
9489					pf->hw.func_caps.num_tx_qp);
9490
9491		queues_left -= pf->num_lan_qps;
9492	}
9493
9494#ifdef I40E_FCOE
9495	if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
9496		if (I40E_DEFAULT_FCOE <= queues_left) {
9497			pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
9498		} else if (I40E_MINIMUM_FCOE <= queues_left) {
9499			pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
9500		} else {
9501			pf->num_fcoe_qps = 0;
9502			pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
9503			dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
9504		}
9505
9506		queues_left -= pf->num_fcoe_qps;
9507	}
9508
9509#endif
9510	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9511		if (queues_left > 1) {
9512			queues_left -= 1; /* save 1 queue for FD */
9513		} else {
9514			pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
9515			dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
9516		}
9517	}
9518
9519	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
9520	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
9521		pf->num_req_vfs = min_t(int, pf->num_req_vfs,
9522					(queues_left / pf->num_vf_qps));
9523		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
9524	}
9525
9526	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
9527	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
9528		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
9529					  (queues_left / pf->num_vmdq_qps));
9530		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
9531	}
9532
9533	pf->queues_left = queues_left;
9534#ifdef I40E_FCOE
9535	dev_info(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
9536#endif
9537}
9538
9539/**
9540 * i40e_setup_pf_filter_control - Setup PF static filter control
9541 * @pf: PF to be setup
9542 *
9543 * i40e_setup_pf_filter_control sets up a PF's initial filter control
9544 * settings. If PE/FCoE are enabled then it will also set the per PF
9545 * based filter sizes required for them. It also enables Flow director,
9546 * ethertype and macvlan type filter settings for the pf.
9547 *
9548 * Returns 0 on success, negative on failure
9549 **/
9550static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
9551{
9552	struct i40e_filter_control_settings *settings = &pf->filter_settings;
9553
9554	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
9555
9556	/* Flow Director is enabled */
9557	if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
9558		settings->enable_fdir = true;
9559
9560	/* Ethtype and MACVLAN filters enabled for PF */
9561	settings->enable_ethtype = true;
9562	settings->enable_macvlan = true;
9563
9564	if (i40e_set_filter_control(&pf->hw, settings))
9565		return -ENOENT;
9566
9567	return 0;
9568}
9569
9570#define INFO_STRING_LEN 255
9571static void i40e_print_features(struct i40e_pf *pf)
9572{
9573	struct i40e_hw *hw = &pf->hw;
9574	char *buf, *string;
9575
9576	string = kzalloc(INFO_STRING_LEN, GFP_KERNEL);
9577	if (!string) {
9578		dev_err(&pf->pdev->dev, "Features string allocation failed\n");
9579		return;
9580	}
9581
9582	buf = string;
9583
9584	buf += sprintf(string, "Features: PF-id[%d] ", hw->pf_id);
9585#ifdef CONFIG_PCI_IOV
9586	buf += sprintf(buf, "VFs: %d ", pf->num_req_vfs);
9587#endif
9588	buf += sprintf(buf, "VSIs: %d QP: %d RX: %s ",
9589		       pf->hw.func_caps.num_vsis,
9590		       pf->vsi[pf->lan_vsi]->num_queue_pairs,
9591		       pf->flags & I40E_FLAG_RX_PS_ENABLED ? "PS" : "1BUF");
9592
9593	if (pf->flags & I40E_FLAG_RSS_ENABLED)
9594		buf += sprintf(buf, "RSS ");
9595	if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
9596		buf += sprintf(buf, "FD_ATR ");
9597	if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
9598		buf += sprintf(buf, "FD_SB ");
9599		buf += sprintf(buf, "NTUPLE ");
9600	}
9601	if (pf->flags & I40E_FLAG_DCB_CAPABLE)
9602		buf += sprintf(buf, "DCB ");
9603	if (pf->flags & I40E_FLAG_PTP)
9604		buf += sprintf(buf, "PTP ");
9605#ifdef I40E_FCOE
9606	if (pf->flags & I40E_FLAG_FCOE_ENABLED)
9607		buf += sprintf(buf, "FCOE ");
9608#endif
9609
9610	BUG_ON(buf > (string + INFO_STRING_LEN));
9611	dev_info(&pf->pdev->dev, "%s\n", string);
9612	kfree(string);
9613}
9614
9615/**
9616 * i40e_probe - Device initialization routine
9617 * @pdev: PCI device information struct
9618 * @ent: entry in i40e_pci_tbl
9619 *
9620 * i40e_probe initializes a PF identified by a pci_dev structure.
9621 * The OS initialization, configuring of the PF private structure,
9622 * and a hardware reset occur.
9623 *
9624 * Returns 0 on success, negative on failure
9625 **/
9626static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
9627{
9628	struct i40e_aq_get_phy_abilities_resp abilities;
9629	unsigned long ioremap_len;
9630	struct i40e_pf *pf;
9631	struct i40e_hw *hw;
9632	static u16 pfs_found;
9633	u16 link_status;
9634	int err = 0;
9635	u32 len;
9636	u32 i;
9637
9638	err = pci_enable_device_mem(pdev);
9639	if (err)
9640		return err;
9641
9642	/* set up for high or low dma */
9643	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
9644	if (err) {
9645		err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
9646		if (err) {
9647			dev_err(&pdev->dev,
9648				"DMA configuration failed: 0x%x\n", err);
9649			goto err_dma;
9650		}
9651	}
9652
9653	/* set up pci connections */
9654	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
9655					   IORESOURCE_MEM), i40e_driver_name);
9656	if (err) {
9657		dev_info(&pdev->dev,
9658			 "pci_request_selected_regions failed %d\n", err);
9659		goto err_pci_reg;
9660	}
9661
9662	pci_enable_pcie_error_reporting(pdev);
9663	pci_set_master(pdev);
9664
9665	/* Now that we have a PCI connection, we need to do the
9666	 * low level device setup.  This is primarily setting up
9667	 * the Admin Queue structures and then querying for the
9668	 * device's current profile information.
9669	 */
9670	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
9671	if (!pf) {
9672		err = -ENOMEM;
9673		goto err_pf_alloc;
9674	}
9675	pf->next_vsi = 0;
9676	pf->pdev = pdev;
9677	set_bit(__I40E_DOWN, &pf->state);
9678
9679	hw = &pf->hw;
9680	hw->back = pf;
9681
9682	ioremap_len = min_t(unsigned long, pci_resource_len(pdev, 0),
9683			    I40E_MAX_CSR_SPACE);
9684
9685	hw->hw_addr = ioremap(pci_resource_start(pdev, 0), ioremap_len);
9686	if (!hw->hw_addr) {
9687		err = -EIO;
9688		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
9689			 (unsigned int)pci_resource_start(pdev, 0),
9690			 (unsigned int)pci_resource_len(pdev, 0), err);
9691		goto err_ioremap;
9692	}
9693	hw->vendor_id = pdev->vendor;
9694	hw->device_id = pdev->device;
9695	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
9696	hw->subsystem_vendor_id = pdev->subsystem_vendor;
9697	hw->subsystem_device_id = pdev->subsystem_device;
9698	hw->bus.device = PCI_SLOT(pdev->devfn);
9699	hw->bus.func = PCI_FUNC(pdev->devfn);
9700	pf->instance = pfs_found;
9701
9702	if (debug != -1) {
9703		pf->msg_enable = pf->hw.debug_mask;
9704		pf->msg_enable = debug;
9705	}
9706
9707	/* do a special CORER for clearing PXE mode once at init */
9708	if (hw->revision_id == 0 &&
9709	    (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
9710		wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
9711		i40e_flush(hw);
9712		msleep(200);
9713		pf->corer_count++;
9714
9715		i40e_clear_pxe_mode(hw);
9716	}
9717
9718	/* Reset here to make sure all is clean and to define PF 'n' */
9719	i40e_clear_hw(hw);
9720	err = i40e_pf_reset(hw);
9721	if (err) {
9722		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
9723		goto err_pf_reset;
9724	}
9725	pf->pfr_count++;
9726
9727	hw->aq.num_arq_entries = I40E_AQ_LEN;
9728	hw->aq.num_asq_entries = I40E_AQ_LEN;
9729	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
9730	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
9731	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
9732
9733	snprintf(pf->int_name, sizeof(pf->int_name) - 1,
9734		 "%s-%s:misc",
9735		 dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
9736
9737	err = i40e_init_shared_code(hw);
9738	if (err) {
9739		dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
9740		goto err_pf_reset;
9741	}
9742
9743	/* set up a default setting for link flow control */
9744	pf->hw.fc.requested_mode = I40E_FC_NONE;
9745
9746	err = i40e_init_adminq(hw);
9747	dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
9748	if (err) {
9749		dev_info(&pdev->dev,
9750			 "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");
9751		goto err_pf_reset;
9752	}
9753
9754	if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
9755	    hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
9756		dev_info(&pdev->dev,
9757			 "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");
9758	else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
9759		 hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
9760		dev_info(&pdev->dev,
9761			 "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
9762
9763	i40e_verify_eeprom(pf);
9764
9765	/* Rev 0 hardware was never productized */
9766	if (hw->revision_id < 1)
9767		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");
9768
9769	i40e_clear_pxe_mode(hw);
9770	err = i40e_get_capabilities(pf);
9771	if (err)
9772		goto err_adminq_setup;
9773
9774	err = i40e_sw_init(pf);
9775	if (err) {
9776		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
9777		goto err_sw_init;
9778	}
9779
9780	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
9781				hw->func_caps.num_rx_qp,
9782				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
9783	if (err) {
9784		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
9785		goto err_init_lan_hmc;
9786	}
9787
9788	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
9789	if (err) {
9790		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
9791		err = -ENOENT;
9792		goto err_configure_lan_hmc;
9793	}
9794
9795	/* Disable LLDP for NICs that have firmware versions lower than v4.3.
9796	 * Ignore error return codes because if it was already disabled via
9797	 * hardware settings this will fail
9798	 */
9799	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
9800	    (pf->hw.aq.fw_maj_ver < 4)) {
9801		dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
9802		i40e_aq_stop_lldp(hw, true, NULL);
9803	}
9804
9805	i40e_get_mac_addr(hw, hw->mac.addr);
9806	if (!is_valid_ether_addr(hw->mac.addr)) {
9807		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
9808		err = -EIO;
9809		goto err_mac_addr;
9810	}
9811	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
9812	ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
9813	i40e_get_port_mac_addr(hw, hw->mac.port_addr);
9814	if (is_valid_ether_addr(hw->mac.port_addr))
9815		pf->flags |= I40E_FLAG_PORT_ID_VALID;
9816#ifdef I40E_FCOE
9817	err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
9818	if (err)
9819		dev_info(&pdev->dev,
9820			 "(non-fatal) SAN MAC retrieval failed: %d\n", err);
9821	if (!is_valid_ether_addr(hw->mac.san_addr)) {
9822		dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
9823			 hw->mac.san_addr);
9824		ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
9825	}
9826	dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
9827#endif /* I40E_FCOE */
9828
9829	pci_set_drvdata(pdev, pf);
9830	pci_save_state(pdev);
9831#ifdef CONFIG_I40E_DCB
9832	err = i40e_init_pf_dcb(pf);
9833	if (err) {
9834		dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
9835		pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
9836		/* Continue without DCB enabled */
9837	}
9838#endif /* CONFIG_I40E_DCB */
9839
9840	/* set up periodic task facility */
9841	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
9842	pf->service_timer_period = HZ;
9843
9844	INIT_WORK(&pf->service_task, i40e_service_task);
9845	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
9846	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
9847	pf->link_check_timeout = jiffies;
9848
9849	/* WoL defaults to disabled */
9850	pf->wol_en = false;
9851	device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
9852
9853	/* set up the main switch operations */
9854	i40e_determine_queue_usage(pf);
9855	err = i40e_init_interrupt_scheme(pf);
9856	if (err)
9857		goto err_switch_setup;
9858
9859	/* The number of VSIs reported by the FW is the minimum guaranteed
9860	 * to us; HW supports far more and we share the remaining pool with
9861	 * the other PFs. We allocate space for more than the guarantee with
9862	 * the understanding that we might not get them all later.
9863	 */
9864	if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
9865		pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
9866	else
9867		pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
9868
9869	/* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
9870	len = sizeof(struct i40e_vsi *) * pf->num_alloc_vsi;
9871	pf->vsi = kzalloc(len, GFP_KERNEL);
9872	if (!pf->vsi) {
9873		err = -ENOMEM;
9874		goto err_switch_setup;
9875	}
9876
9877#ifdef CONFIG_PCI_IOV
9878	/* prep for VF support */
9879	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
9880	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
9881	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
9882		if (pci_num_vf(pdev))
9883			pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
9884	}
9885#endif
9886	err = i40e_setup_pf_switch(pf, false);
9887	if (err) {
9888		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
9889		goto err_vsis;
9890	}
9891	/* if FDIR VSI was set up, start it now */
9892	for (i = 0; i < pf->num_alloc_vsi; i++) {
9893		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
9894			i40e_vsi_open(pf->vsi[i]);
9895			break;
9896		}
9897	}
9898
9899	/* driver is only interested in link up/down and module qualification
9900	 * reports from firmware
9901	 */
9902	err = i40e_aq_set_phy_int_mask(&pf->hw,
9903				       I40E_AQ_EVENT_LINK_UPDOWN |
9904				       I40E_AQ_EVENT_MODULE_QUAL_FAIL, NULL);
9905	if (err)
9906		dev_info(&pf->pdev->dev, "set phy mask fail, aq_err %d\n", err);
9907
9908	if (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
9909	    (pf->hw.aq.fw_maj_ver < 4)) {
9910		msleep(75);
9911		err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
9912		if (err)
9913			dev_info(&pf->pdev->dev, "link restart failed, aq_err=%d\n",
9914				 pf->hw.aq.asq_last_status);
9915	}
9916	/* The main driver is (mostly) up and happy. We need to set this state
9917	 * before setting up the misc vector or we get a race and the vector
9918	 * ends up disabled forever.
9919	 */
9920	clear_bit(__I40E_DOWN, &pf->state);
9921
9922	/* In case of MSIX we are going to setup the misc vector right here
9923	 * to handle admin queue events etc. In case of legacy and MSI
9924	 * the misc functionality and queue processing is combined in
9925	 * the same vector and that gets setup at open.
9926	 */
9927	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
9928		err = i40e_setup_misc_vector(pf);
9929		if (err) {
9930			dev_info(&pdev->dev,
9931				 "setup of misc vector failed: %d\n", err);
9932			goto err_vsis;
9933		}
9934	}
9935
9936#ifdef CONFIG_PCI_IOV
9937	/* prep for VF support */
9938	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
9939	    (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
9940	    !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
9941		u32 val;
9942
9943		/* disable link interrupts for VFs */
9944		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
9945		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
9946		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
9947		i40e_flush(hw);
9948
9949		if (pci_num_vf(pdev)) {
9950			dev_info(&pdev->dev,
9951				 "Active VFs found, allocating resources.\n");
9952			err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
9953			if (err)
9954				dev_info(&pdev->dev,
9955					 "Error %d allocating resources for existing VFs\n",
9956					 err);
9957		}
9958	}
9959#endif /* CONFIG_PCI_IOV */
9960
9961	pfs_found++;
9962
9963	i40e_dbg_pf_init(pf);
9964
9965	/* tell the firmware that we're starting */
9966	i40e_send_version(pf);
9967
9968	/* since everything's happy, start the service_task timer */
9969	mod_timer(&pf->service_timer,
9970		  round_jiffies(jiffies + pf->service_timer_period));
9971
9972#ifdef I40E_FCOE
9973	/* create FCoE interface */
9974	i40e_fcoe_vsi_setup(pf);
9975
9976#endif
9977	/* Get the negotiated link width and speed from PCI config space */
9978	pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA, &link_status);
9979
9980	i40e_set_pci_config_data(hw, link_status);
9981
9982	dev_info(&pdev->dev, "PCI-Express: %s %s\n",
9983		(hw->bus.speed == i40e_bus_speed_8000 ? "Speed 8.0GT/s" :
9984		 hw->bus.speed == i40e_bus_speed_5000 ? "Speed 5.0GT/s" :
9985		 hw->bus.speed == i40e_bus_speed_2500 ? "Speed 2.5GT/s" :
9986		 "Unknown"),
9987		(hw->bus.width == i40e_bus_width_pcie_x8 ? "Width x8" :
9988		 hw->bus.width == i40e_bus_width_pcie_x4 ? "Width x4" :
9989		 hw->bus.width == i40e_bus_width_pcie_x2 ? "Width x2" :
9990		 hw->bus.width == i40e_bus_width_pcie_x1 ? "Width x1" :
9991		 "Unknown"));
9992
9993	if (hw->bus.width < i40e_bus_width_pcie_x8 ||
9994	    hw->bus.speed < i40e_bus_speed_8000) {
9995		dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
9996		dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
9997	}
9998
9999	/* get the requested speeds from the fw */
10000	err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
10001	if (err)
10002		dev_info(&pf->pdev->dev, "get phy abilities failed, aq_err %d, advertised speed settings may not be correct\n",
10003			 err);
10004	pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
10005
10006	/* print a string summarizing features */
10007	i40e_print_features(pf);
10008
10009	return 0;
10010
10011	/* Unwind what we've done if something failed in the setup */
10012err_vsis:
10013	set_bit(__I40E_DOWN, &pf->state);
10014	i40e_clear_interrupt_scheme(pf);
10015	kfree(pf->vsi);
10016err_switch_setup:
10017	i40e_reset_interrupt_capability(pf);
10018	del_timer_sync(&pf->service_timer);
10019err_mac_addr:
10020err_configure_lan_hmc:
10021	(void)i40e_shutdown_lan_hmc(hw);
10022err_init_lan_hmc:
10023	kfree(pf->qp_pile);
10024err_sw_init:
10025err_adminq_setup:
10026	(void)i40e_shutdown_adminq(hw);
10027err_pf_reset:
10028	iounmap(hw->hw_addr);
10029err_ioremap:
10030	kfree(pf);
10031err_pf_alloc:
10032	pci_disable_pcie_error_reporting(pdev);
10033	pci_release_selected_regions(pdev,
10034				     pci_select_bars(pdev, IORESOURCE_MEM));
10035err_pci_reg:
10036err_dma:
10037	pci_disable_device(pdev);
10038	return err;
10039}
10040
10041/**
10042 * i40e_remove - Device removal routine
10043 * @pdev: PCI device information struct
10044 *
10045 * i40e_remove is called by the PCI subsystem to alert the driver
10046 * that is should release a PCI device.  This could be caused by a
10047 * Hot-Plug event, or because the driver is going to be removed from
10048 * memory.
10049 **/
10050static void i40e_remove(struct pci_dev *pdev)
10051{
10052	struct i40e_pf *pf = pci_get_drvdata(pdev);
10053	i40e_status ret_code;
10054	int i;
10055
10056	i40e_dbg_pf_exit(pf);
10057
10058	i40e_ptp_stop(pf);
10059
10060	/* no more scheduling of any task */
10061	set_bit(__I40E_DOWN, &pf->state);
10062	del_timer_sync(&pf->service_timer);
10063	cancel_work_sync(&pf->service_task);
10064	i40e_fdir_teardown(pf);
10065
10066	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
10067		i40e_free_vfs(pf);
10068		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
10069	}
10070
10071	i40e_fdir_teardown(pf);
10072
10073	/* If there is a switch structure or any orphans, remove them.
10074	 * This will leave only the PF's VSI remaining.
10075	 */
10076	for (i = 0; i < I40E_MAX_VEB; i++) {
10077		if (!pf->veb[i])
10078			continue;
10079
10080		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
10081		    pf->veb[i]->uplink_seid == 0)
10082			i40e_switch_branch_release(pf->veb[i]);
10083	}
10084
10085	/* Now we can shutdown the PF's VSI, just before we kill
10086	 * adminq and hmc.
10087	 */
10088	if (pf->vsi[pf->lan_vsi])
10089		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
10090
10091	/* shutdown and destroy the HMC */
10092	if (pf->hw.hmc.hmc_obj) {
10093		ret_code = i40e_shutdown_lan_hmc(&pf->hw);
10094		if (ret_code)
10095			dev_warn(&pdev->dev,
10096				 "Failed to destroy the HMC resources: %d\n",
10097				 ret_code);
10098	}
10099
10100	/* shutdown the adminq */
10101	ret_code = i40e_shutdown_adminq(&pf->hw);
10102	if (ret_code)
10103		dev_warn(&pdev->dev,
10104			 "Failed to destroy the Admin Queue resources: %d\n",
10105			 ret_code);
10106
10107	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
10108	i40e_clear_interrupt_scheme(pf);
10109	for (i = 0; i < pf->num_alloc_vsi; i++) {
10110		if (pf->vsi[i]) {
10111			i40e_vsi_clear_rings(pf->vsi[i]);
10112			i40e_vsi_clear(pf->vsi[i]);
10113			pf->vsi[i] = NULL;
10114		}
10115	}
10116
10117	for (i = 0; i < I40E_MAX_VEB; i++) {
10118		kfree(pf->veb[i]);
10119		pf->veb[i] = NULL;
10120	}
10121
10122	kfree(pf->qp_pile);
10123	kfree(pf->vsi);
10124
10125	iounmap(pf->hw.hw_addr);
10126	kfree(pf);
10127	pci_release_selected_regions(pdev,
10128				     pci_select_bars(pdev, IORESOURCE_MEM));
10129
10130	pci_disable_pcie_error_reporting(pdev);
10131	pci_disable_device(pdev);
10132}
10133
10134/**
10135 * i40e_pci_error_detected - warning that something funky happened in PCI land
10136 * @pdev: PCI device information struct
10137 *
10138 * Called to warn that something happened and the error handling steps
10139 * are in progress.  Allows the driver to quiesce things, be ready for
10140 * remediation.
10141 **/
10142static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
10143						enum pci_channel_state error)
10144{
10145	struct i40e_pf *pf = pci_get_drvdata(pdev);
10146
10147	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
10148
10149	/* shutdown all operations */
10150	if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
10151		rtnl_lock();
10152		i40e_prep_for_reset(pf);
10153		rtnl_unlock();
10154	}
10155
10156	/* Request a slot reset */
10157	return PCI_ERS_RESULT_NEED_RESET;
10158}
10159
10160/**
10161 * i40e_pci_error_slot_reset - a PCI slot reset just happened
10162 * @pdev: PCI device information struct
10163 *
10164 * Called to find if the driver can work with the device now that
10165 * the pci slot has been reset.  If a basic connection seems good
10166 * (registers are readable and have sane content) then return a
10167 * happy little PCI_ERS_RESULT_xxx.
10168 **/
10169static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
10170{
10171	struct i40e_pf *pf = pci_get_drvdata(pdev);
10172	pci_ers_result_t result;
10173	int err;
10174	u32 reg;
10175
10176	dev_info(&pdev->dev, "%s\n", __func__);
10177	if (pci_enable_device_mem(pdev)) {
10178		dev_info(&pdev->dev,
10179			 "Cannot re-enable PCI device after reset.\n");
10180		result = PCI_ERS_RESULT_DISCONNECT;
10181	} else {
10182		pci_set_master(pdev);
10183		pci_restore_state(pdev);
10184		pci_save_state(pdev);
10185		pci_wake_from_d3(pdev, false);
10186
10187		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
10188		if (reg == 0)
10189			result = PCI_ERS_RESULT_RECOVERED;
10190		else
10191			result = PCI_ERS_RESULT_DISCONNECT;
10192	}
10193
10194	err = pci_cleanup_aer_uncorrect_error_status(pdev);
10195	if (err) {
10196		dev_info(&pdev->dev,
10197			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
10198			 err);
10199		/* non-fatal, continue */
10200	}
10201
10202	return result;
10203}
10204
10205/**
10206 * i40e_pci_error_resume - restart operations after PCI error recovery
10207 * @pdev: PCI device information struct
10208 *
10209 * Called to allow the driver to bring things back up after PCI error
10210 * and/or reset recovery has finished.
10211 **/
10212static void i40e_pci_error_resume(struct pci_dev *pdev)
10213{
10214	struct i40e_pf *pf = pci_get_drvdata(pdev);
10215
10216	dev_info(&pdev->dev, "%s\n", __func__);
10217	if (test_bit(__I40E_SUSPENDED, &pf->state))
10218		return;
10219
10220	rtnl_lock();
10221	i40e_handle_reset_warning(pf);
10222	rtnl_lock();
10223}
10224
10225/**
10226 * i40e_shutdown - PCI callback for shutting down
10227 * @pdev: PCI device information struct
10228 **/
10229static void i40e_shutdown(struct pci_dev *pdev)
10230{
10231	struct i40e_pf *pf = pci_get_drvdata(pdev);
10232	struct i40e_hw *hw = &pf->hw;
10233
10234	set_bit(__I40E_SUSPENDED, &pf->state);
10235	set_bit(__I40E_DOWN, &pf->state);
10236	rtnl_lock();
10237	i40e_prep_for_reset(pf);
10238	rtnl_unlock();
10239
10240	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10241	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10242
10243	i40e_clear_interrupt_scheme(pf);
10244
10245	if (system_state == SYSTEM_POWER_OFF) {
10246		pci_wake_from_d3(pdev, pf->wol_en);
10247		pci_set_power_state(pdev, PCI_D3hot);
10248	}
10249}
10250
10251#ifdef CONFIG_PM
10252/**
10253 * i40e_suspend - PCI callback for moving to D3
10254 * @pdev: PCI device information struct
10255 **/
10256static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
10257{
10258	struct i40e_pf *pf = pci_get_drvdata(pdev);
10259	struct i40e_hw *hw = &pf->hw;
10260
10261	set_bit(__I40E_SUSPENDED, &pf->state);
10262	set_bit(__I40E_DOWN, &pf->state);
10263	del_timer_sync(&pf->service_timer);
10264	cancel_work_sync(&pf->service_task);
10265	i40e_fdir_teardown(pf);
10266
10267	rtnl_lock();
10268	i40e_prep_for_reset(pf);
10269	rtnl_unlock();
10270
10271	wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
10272	wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
10273
10274	pci_wake_from_d3(pdev, pf->wol_en);
10275	pci_set_power_state(pdev, PCI_D3hot);
10276
10277	return 0;
10278}
10279
10280/**
10281 * i40e_resume - PCI callback for waking up from D3
10282 * @pdev: PCI device information struct
10283 **/
10284static int i40e_resume(struct pci_dev *pdev)
10285{
10286	struct i40e_pf *pf = pci_get_drvdata(pdev);
10287	u32 err;
10288
10289	pci_set_power_state(pdev, PCI_D0);
10290	pci_restore_state(pdev);
10291	/* pci_restore_state() clears dev->state_saves, so
10292	 * call pci_save_state() again to restore it.
10293	 */
10294	pci_save_state(pdev);
10295
10296	err = pci_enable_device_mem(pdev);
10297	if (err) {
10298		dev_err(&pdev->dev,
10299			"%s: Cannot enable PCI device from suspend\n",
10300			__func__);
10301		return err;
10302	}
10303	pci_set_master(pdev);
10304
10305	/* no wakeup events while running */
10306	pci_wake_from_d3(pdev, false);
10307
10308	/* handling the reset will rebuild the device state */
10309	if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
10310		clear_bit(__I40E_DOWN, &pf->state);
10311		rtnl_lock();
10312		i40e_reset_and_rebuild(pf, false);
10313		rtnl_unlock();
10314	}
10315
10316	return 0;
10317}
10318
10319#endif
10320static const struct pci_error_handlers i40e_err_handler = {
10321	.error_detected = i40e_pci_error_detected,
10322	.slot_reset = i40e_pci_error_slot_reset,
10323	.resume = i40e_pci_error_resume,
10324};
10325
10326static struct pci_driver i40e_driver = {
10327	.name     = i40e_driver_name,
10328	.id_table = i40e_pci_tbl,
10329	.probe    = i40e_probe,
10330	.remove   = i40e_remove,
10331#ifdef CONFIG_PM
10332	.suspend  = i40e_suspend,
10333	.resume   = i40e_resume,
10334#endif
10335	.shutdown = i40e_shutdown,
10336	.err_handler = &i40e_err_handler,
10337	.sriov_configure = i40e_pci_sriov_configure,
10338};
10339
10340/**
10341 * i40e_init_module - Driver registration routine
10342 *
10343 * i40e_init_module is the first routine called when the driver is
10344 * loaded. All it does is register with the PCI subsystem.
10345 **/
10346static int __init i40e_init_module(void)
10347{
10348	pr_info("%s: %s - version %s\n", i40e_driver_name,
10349		i40e_driver_string, i40e_driver_version_str);
10350	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
10351
10352	i40e_dbg_init();
10353	return pci_register_driver(&i40e_driver);
10354}
10355module_init(i40e_init_module);
10356
10357/**
10358 * i40e_exit_module - Driver exit cleanup routine
10359 *
10360 * i40e_exit_module is called just before the driver is removed
10361 * from memory.
10362 **/
10363static void __exit i40e_exit_module(void)
10364{
10365	pci_unregister_driver(&i40e_driver);
10366	i40e_dbg_exit();
10367}
10368module_exit(i40e_exit_module);
10369