1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 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 with
16  * this program; if not, write to the Free Software Foundation, Inc.,
17  * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
18  *
19  * The full GNU General Public License is included in this distribution in
20  * the file called "COPYING".
21  *
22  * Contact Information:
23  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
24  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
25  *
26  ******************************************************************************/
27 
28 /* Local includes */
29 #include "i40e.h"
30 
31 const char i40e_driver_name[] = "i40e";
32 static const char i40e_driver_string[] =
33 			"Intel(R) Ethernet Connection XL710 Network Driver";
34 
35 #define DRV_KERN "-k"
36 
37 #define DRV_VERSION_MAJOR 0
38 #define DRV_VERSION_MINOR 3
39 #define DRV_VERSION_BUILD 11
40 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
41 	     __stringify(DRV_VERSION_MINOR) "." \
42 	     __stringify(DRV_VERSION_BUILD)    DRV_KERN
43 const char i40e_driver_version_str[] = DRV_VERSION;
44 static const char i40e_copyright[] = "Copyright (c) 2013 Intel Corporation.";
45 
46 /* a bit of forward declarations */
47 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
48 static void i40e_handle_reset_warning(struct i40e_pf *pf);
49 static int i40e_add_vsi(struct i40e_vsi *vsi);
50 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
51 static int i40e_setup_pf_switch(struct i40e_pf *pf);
52 static int i40e_setup_misc_vector(struct i40e_pf *pf);
53 static void i40e_determine_queue_usage(struct i40e_pf *pf);
54 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
55 
56 /* i40e_pci_tbl - PCI Device ID Table
57  *
58  * Last entry must be all 0s
59  *
60  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
61  *   Class, Class Mask, private data (not used) }
62  */
63 static DEFINE_PCI_DEVICE_TABLE(i40e_pci_tbl) = {
64 	{PCI_VDEVICE(INTEL, I40E_SFP_XL710_DEVICE_ID), 0},
65 	{PCI_VDEVICE(INTEL, I40E_SFP_X710_DEVICE_ID), 0},
66 	{PCI_VDEVICE(INTEL, I40E_QEMU_DEVICE_ID), 0},
67 	{PCI_VDEVICE(INTEL, I40E_KX_A_DEVICE_ID), 0},
68 	{PCI_VDEVICE(INTEL, I40E_KX_B_DEVICE_ID), 0},
69 	{PCI_VDEVICE(INTEL, I40E_KX_C_DEVICE_ID), 0},
70 	{PCI_VDEVICE(INTEL, I40E_KX_D_DEVICE_ID), 0},
71 	{PCI_VDEVICE(INTEL, I40E_QSFP_A_DEVICE_ID), 0},
72 	{PCI_VDEVICE(INTEL, I40E_QSFP_B_DEVICE_ID), 0},
73 	{PCI_VDEVICE(INTEL, I40E_QSFP_C_DEVICE_ID), 0},
74 	/* required last entry */
75 	{0, }
76 };
77 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
78 
79 #define I40E_MAX_VF_COUNT 128
80 static int debug = -1;
81 module_param(debug, int, 0);
82 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
83 
84 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
85 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
86 MODULE_LICENSE("GPL");
87 MODULE_VERSION(DRV_VERSION);
88 
89 /**
90  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
91  * @hw:   pointer to the HW structure
92  * @mem:  ptr to mem struct to fill out
93  * @size: size of memory requested
94  * @alignment: what to align the allocation to
95  **/
96 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
97 			    u64 size, u32 alignment)
98 {
99 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
100 
101 	mem->size = ALIGN(size, alignment);
102 	mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
103 				      &mem->pa, GFP_KERNEL);
104 	if (!mem->va)
105 		return -ENOMEM;
106 
107 	return 0;
108 }
109 
110 /**
111  * i40e_free_dma_mem_d - OS specific memory free for shared code
112  * @hw:   pointer to the HW structure
113  * @mem:  ptr to mem struct to free
114  **/
115 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
116 {
117 	struct i40e_pf *pf = (struct i40e_pf *)hw->back;
118 
119 	dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
120 	mem->va = NULL;
121 	mem->pa = 0;
122 	mem->size = 0;
123 
124 	return 0;
125 }
126 
127 /**
128  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
129  * @hw:   pointer to the HW structure
130  * @mem:  ptr to mem struct to fill out
131  * @size: size of memory requested
132  **/
133 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
134 			     u32 size)
135 {
136 	mem->size = size;
137 	mem->va = kzalloc(size, GFP_KERNEL);
138 
139 	if (!mem->va)
140 		return -ENOMEM;
141 
142 	return 0;
143 }
144 
145 /**
146  * i40e_free_virt_mem_d - OS specific memory free for shared code
147  * @hw:   pointer to the HW structure
148  * @mem:  ptr to mem struct to free
149  **/
150 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
151 {
152 	/* it's ok to kfree a NULL pointer */
153 	kfree(mem->va);
154 	mem->va = NULL;
155 	mem->size = 0;
156 
157 	return 0;
158 }
159 
160 /**
161  * i40e_get_lump - find a lump of free generic resource
162  * @pf: board private structure
163  * @pile: the pile of resource to search
164  * @needed: the number of items needed
165  * @id: an owner id to stick on the items assigned
166  *
167  * Returns the base item index of the lump, or negative for error
168  *
169  * The search_hint trick and lack of advanced fit-finding only work
170  * because we're highly likely to have all the same size lump requests.
171  * Linear search time and any fragmentation should be minimal.
172  **/
173 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
174 			 u16 needed, u16 id)
175 {
176 	int ret = -ENOMEM;
177 	int i, j;
178 
179 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
180 		dev_info(&pf->pdev->dev,
181 			 "param err: pile=%p needed=%d id=0x%04x\n",
182 			 pile, needed, id);
183 		return -EINVAL;
184 	}
185 
186 	/* start the linear search with an imperfect hint */
187 	i = pile->search_hint;
188 	while (i < pile->num_entries) {
189 		/* skip already allocated entries */
190 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
191 			i++;
192 			continue;
193 		}
194 
195 		/* do we have enough in this lump? */
196 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
197 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
198 				break;
199 		}
200 
201 		if (j == needed) {
202 			/* there was enough, so assign it to the requestor */
203 			for (j = 0; j < needed; j++)
204 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
205 			ret = i;
206 			pile->search_hint = i + j;
207 			break;
208 		} else {
209 			/* not enough, so skip over it and continue looking */
210 			i += j;
211 		}
212 	}
213 
214 	return ret;
215 }
216 
217 /**
218  * i40e_put_lump - return a lump of generic resource
219  * @pile: the pile of resource to search
220  * @index: the base item index
221  * @id: the owner id of the items assigned
222  *
223  * Returns the count of items in the lump
224  **/
225 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
226 {
227 	int valid_id = (id | I40E_PILE_VALID_BIT);
228 	int count = 0;
229 	int i;
230 
231 	if (!pile || index >= pile->num_entries)
232 		return -EINVAL;
233 
234 	for (i = index;
235 	     i < pile->num_entries && pile->list[i] == valid_id;
236 	     i++) {
237 		pile->list[i] = 0;
238 		count++;
239 	}
240 
241 	if (count && index < pile->search_hint)
242 		pile->search_hint = index;
243 
244 	return count;
245 }
246 
247 /**
248  * i40e_service_event_schedule - Schedule the service task to wake up
249  * @pf: board private structure
250  *
251  * If not already scheduled, this puts the task into the work queue
252  **/
253 static void i40e_service_event_schedule(struct i40e_pf *pf)
254 {
255 	if (!test_bit(__I40E_DOWN, &pf->state) &&
256 	    !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
257 	    !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
258 		schedule_work(&pf->service_task);
259 }
260 
261 /**
262  * i40e_tx_timeout - Respond to a Tx Hang
263  * @netdev: network interface device structure
264  *
265  * If any port has noticed a Tx timeout, it is likely that the whole
266  * device is munged, not just the one netdev port, so go for the full
267  * reset.
268  **/
269 static void i40e_tx_timeout(struct net_device *netdev)
270 {
271 	struct i40e_netdev_priv *np = netdev_priv(netdev);
272 	struct i40e_vsi *vsi = np->vsi;
273 	struct i40e_pf *pf = vsi->back;
274 
275 	pf->tx_timeout_count++;
276 
277 	if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
278 		pf->tx_timeout_recovery_level = 0;
279 	pf->tx_timeout_last_recovery = jiffies;
280 	netdev_info(netdev, "tx_timeout recovery level %d\n",
281 		    pf->tx_timeout_recovery_level);
282 
283 	switch (pf->tx_timeout_recovery_level) {
284 	case 0:
285 		/* disable and re-enable queues for the VSI */
286 		if (in_interrupt()) {
287 			set_bit(__I40E_REINIT_REQUESTED, &pf->state);
288 			set_bit(__I40E_REINIT_REQUESTED, &vsi->state);
289 		} else {
290 			i40e_vsi_reinit_locked(vsi);
291 		}
292 		break;
293 	case 1:
294 		set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
295 		break;
296 	case 2:
297 		set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
298 		break;
299 	case 3:
300 		set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
301 		break;
302 	default:
303 		netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
304 		i40e_down(vsi);
305 		break;
306 	}
307 	i40e_service_event_schedule(pf);
308 	pf->tx_timeout_recovery_level++;
309 }
310 
311 /**
312  * i40e_release_rx_desc - Store the new tail and head values
313  * @rx_ring: ring to bump
314  * @val: new head index
315  **/
316 static inline void i40e_release_rx_desc(struct i40e_ring *rx_ring, u32 val)
317 {
318 	rx_ring->next_to_use = val;
319 
320 	/* Force memory writes to complete before letting h/w
321 	 * know there are new descriptors to fetch.  (Only
322 	 * applicable for weak-ordered memory model archs,
323 	 * such as IA-64).
324 	 */
325 	wmb();
326 	writel(val, rx_ring->tail);
327 }
328 
329 /**
330  * i40e_get_vsi_stats_struct - Get System Network Statistics
331  * @vsi: the VSI we care about
332  *
333  * Returns the address of the device statistics structure.
334  * The statistics are actually updated from the service task.
335  **/
336 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
337 {
338 	return &vsi->net_stats;
339 }
340 
341 /**
342  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
343  * @netdev: network interface device structure
344  *
345  * Returns the address of the device statistics structure.
346  * The statistics are actually updated from the service task.
347  **/
348 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
349 					     struct net_device *netdev,
350 					     struct rtnl_link_stats64 *stats)
351 {
352 	struct i40e_netdev_priv *np = netdev_priv(netdev);
353 	struct i40e_vsi *vsi = np->vsi;
354 	struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
355 	int i;
356 
357 	rcu_read_lock();
358 	for (i = 0; i < vsi->num_queue_pairs; i++) {
359 		struct i40e_ring *tx_ring, *rx_ring;
360 		u64 bytes, packets;
361 		unsigned int start;
362 
363 		tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
364 		if (!tx_ring)
365 			continue;
366 
367 		do {
368 			start = u64_stats_fetch_begin_bh(&tx_ring->syncp);
369 			packets = tx_ring->stats.packets;
370 			bytes   = tx_ring->stats.bytes;
371 		} while (u64_stats_fetch_retry_bh(&tx_ring->syncp, start));
372 
373 		stats->tx_packets += packets;
374 		stats->tx_bytes   += bytes;
375 		rx_ring = &tx_ring[1];
376 
377 		do {
378 			start = u64_stats_fetch_begin_bh(&rx_ring->syncp);
379 			packets = rx_ring->stats.packets;
380 			bytes   = rx_ring->stats.bytes;
381 		} while (u64_stats_fetch_retry_bh(&rx_ring->syncp, start));
382 
383 		stats->rx_packets += packets;
384 		stats->rx_bytes   += bytes;
385 	}
386 	rcu_read_unlock();
387 
388 	/* following stats updated by ixgbe_watchdog_task() */
389 	stats->multicast	= vsi_stats->multicast;
390 	stats->tx_errors	= vsi_stats->tx_errors;
391 	stats->tx_dropped	= vsi_stats->tx_dropped;
392 	stats->rx_errors	= vsi_stats->rx_errors;
393 	stats->rx_crc_errors	= vsi_stats->rx_crc_errors;
394 	stats->rx_length_errors	= vsi_stats->rx_length_errors;
395 
396 	return stats;
397 }
398 
399 /**
400  * i40e_vsi_reset_stats - Resets all stats of the given vsi
401  * @vsi: the VSI to have its stats reset
402  **/
403 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
404 {
405 	struct rtnl_link_stats64 *ns;
406 	int i;
407 
408 	if (!vsi)
409 		return;
410 
411 	ns = i40e_get_vsi_stats_struct(vsi);
412 	memset(ns, 0, sizeof(*ns));
413 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
414 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
415 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
416 	if (vsi->rx_rings)
417 		for (i = 0; i < vsi->num_queue_pairs; i++) {
418 			memset(&vsi->rx_rings[i]->stats, 0 ,
419 			       sizeof(vsi->rx_rings[i]->stats));
420 			memset(&vsi->rx_rings[i]->rx_stats, 0 ,
421 			       sizeof(vsi->rx_rings[i]->rx_stats));
422 			memset(&vsi->tx_rings[i]->stats, 0 ,
423 			       sizeof(vsi->tx_rings[i]->stats));
424 			memset(&vsi->tx_rings[i]->tx_stats, 0,
425 			       sizeof(vsi->tx_rings[i]->tx_stats));
426 		}
427 	vsi->stat_offsets_loaded = false;
428 }
429 
430 /**
431  * i40e_pf_reset_stats - Reset all of the stats for the given pf
432  * @pf: the PF to be reset
433  **/
434 void i40e_pf_reset_stats(struct i40e_pf *pf)
435 {
436 	memset(&pf->stats, 0, sizeof(pf->stats));
437 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
438 	pf->stat_offsets_loaded = false;
439 }
440 
441 /**
442  * i40e_stat_update48 - read and update a 48 bit stat from the chip
443  * @hw: ptr to the hardware info
444  * @hireg: the high 32 bit reg to read
445  * @loreg: the low 32 bit reg to read
446  * @offset_loaded: has the initial offset been loaded yet
447  * @offset: ptr to current offset value
448  * @stat: ptr to the stat
449  *
450  * Since the device stats are not reset at PFReset, they likely will not
451  * be zeroed when the driver starts.  We'll save the first values read
452  * and use them as offsets to be subtracted from the raw values in order
453  * to report stats that count from zero.  In the process, we also manage
454  * the potential roll-over.
455  **/
456 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
457 			       bool offset_loaded, u64 *offset, u64 *stat)
458 {
459 	u64 new_data;
460 
461 	if (hw->device_id == I40E_QEMU_DEVICE_ID) {
462 		new_data = rd32(hw, loreg);
463 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
464 	} else {
465 		new_data = rd64(hw, loreg);
466 	}
467 	if (!offset_loaded)
468 		*offset = new_data;
469 	if (likely(new_data >= *offset))
470 		*stat = new_data - *offset;
471 	else
472 		*stat = (new_data + ((u64)1 << 48)) - *offset;
473 	*stat &= 0xFFFFFFFFFFFFULL;
474 }
475 
476 /**
477  * i40e_stat_update32 - read and update a 32 bit stat from the chip
478  * @hw: ptr to the hardware info
479  * @reg: the hw reg to read
480  * @offset_loaded: has the initial offset been loaded yet
481  * @offset: ptr to current offset value
482  * @stat: ptr to the stat
483  **/
484 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
485 			       bool offset_loaded, u64 *offset, u64 *stat)
486 {
487 	u32 new_data;
488 
489 	new_data = rd32(hw, reg);
490 	if (!offset_loaded)
491 		*offset = new_data;
492 	if (likely(new_data >= *offset))
493 		*stat = (u32)(new_data - *offset);
494 	else
495 		*stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
496 }
497 
498 /**
499  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
500  * @vsi: the VSI to be updated
501  **/
502 void i40e_update_eth_stats(struct i40e_vsi *vsi)
503 {
504 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
505 	struct i40e_pf *pf = vsi->back;
506 	struct i40e_hw *hw = &pf->hw;
507 	struct i40e_eth_stats *oes;
508 	struct i40e_eth_stats *es;     /* device's eth stats */
509 
510 	es = &vsi->eth_stats;
511 	oes = &vsi->eth_stats_offsets;
512 
513 	/* Gather up the stats that the hw collects */
514 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
515 			   vsi->stat_offsets_loaded,
516 			   &oes->tx_errors, &es->tx_errors);
517 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
518 			   vsi->stat_offsets_loaded,
519 			   &oes->rx_discards, &es->rx_discards);
520 
521 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
522 			   I40E_GLV_GORCL(stat_idx),
523 			   vsi->stat_offsets_loaded,
524 			   &oes->rx_bytes, &es->rx_bytes);
525 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
526 			   I40E_GLV_UPRCL(stat_idx),
527 			   vsi->stat_offsets_loaded,
528 			   &oes->rx_unicast, &es->rx_unicast);
529 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
530 			   I40E_GLV_MPRCL(stat_idx),
531 			   vsi->stat_offsets_loaded,
532 			   &oes->rx_multicast, &es->rx_multicast);
533 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
534 			   I40E_GLV_BPRCL(stat_idx),
535 			   vsi->stat_offsets_loaded,
536 			   &oes->rx_broadcast, &es->rx_broadcast);
537 
538 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
539 			   I40E_GLV_GOTCL(stat_idx),
540 			   vsi->stat_offsets_loaded,
541 			   &oes->tx_bytes, &es->tx_bytes);
542 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
543 			   I40E_GLV_UPTCL(stat_idx),
544 			   vsi->stat_offsets_loaded,
545 			   &oes->tx_unicast, &es->tx_unicast);
546 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
547 			   I40E_GLV_MPTCL(stat_idx),
548 			   vsi->stat_offsets_loaded,
549 			   &oes->tx_multicast, &es->tx_multicast);
550 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
551 			   I40E_GLV_BPTCL(stat_idx),
552 			   vsi->stat_offsets_loaded,
553 			   &oes->tx_broadcast, &es->tx_broadcast);
554 	vsi->stat_offsets_loaded = true;
555 }
556 
557 /**
558  * i40e_update_veb_stats - Update Switch component statistics
559  * @veb: the VEB being updated
560  **/
561 static void i40e_update_veb_stats(struct i40e_veb *veb)
562 {
563 	struct i40e_pf *pf = veb->pf;
564 	struct i40e_hw *hw = &pf->hw;
565 	struct i40e_eth_stats *oes;
566 	struct i40e_eth_stats *es;     /* device's eth stats */
567 	int idx = 0;
568 
569 	idx = veb->stats_idx;
570 	es = &veb->stats;
571 	oes = &veb->stats_offsets;
572 
573 	/* Gather up the stats that the hw collects */
574 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
575 			   veb->stat_offsets_loaded,
576 			   &oes->tx_discards, &es->tx_discards);
577 	i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
578 			   veb->stat_offsets_loaded,
579 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
580 
581 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
582 			   veb->stat_offsets_loaded,
583 			   &oes->rx_bytes, &es->rx_bytes);
584 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
585 			   veb->stat_offsets_loaded,
586 			   &oes->rx_unicast, &es->rx_unicast);
587 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
588 			   veb->stat_offsets_loaded,
589 			   &oes->rx_multicast, &es->rx_multicast);
590 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
591 			   veb->stat_offsets_loaded,
592 			   &oes->rx_broadcast, &es->rx_broadcast);
593 
594 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
595 			   veb->stat_offsets_loaded,
596 			   &oes->tx_bytes, &es->tx_bytes);
597 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
598 			   veb->stat_offsets_loaded,
599 			   &oes->tx_unicast, &es->tx_unicast);
600 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
601 			   veb->stat_offsets_loaded,
602 			   &oes->tx_multicast, &es->tx_multicast);
603 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
604 			   veb->stat_offsets_loaded,
605 			   &oes->tx_broadcast, &es->tx_broadcast);
606 	veb->stat_offsets_loaded = true;
607 }
608 
609 /**
610  * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
611  * @pf: the corresponding PF
612  *
613  * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
614  **/
615 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
616 {
617 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
618 	struct i40e_hw_port_stats *nsd = &pf->stats;
619 	struct i40e_hw *hw = &pf->hw;
620 	u64 xoff = 0;
621 	u16 i, v;
622 
623 	if ((hw->fc.current_mode != I40E_FC_FULL) &&
624 	    (hw->fc.current_mode != I40E_FC_RX_PAUSE))
625 		return;
626 
627 	xoff = nsd->link_xoff_rx;
628 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
629 			   pf->stat_offsets_loaded,
630 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
631 
632 	/* No new LFC xoff rx */
633 	if (!(nsd->link_xoff_rx - xoff))
634 		return;
635 
636 	/* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
637 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
638 		struct i40e_vsi *vsi = pf->vsi[v];
639 
640 		if (!vsi)
641 			continue;
642 
643 		for (i = 0; i < vsi->num_queue_pairs; i++) {
644 			struct i40e_ring *ring = vsi->tx_rings[i];
645 			clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
646 		}
647 	}
648 }
649 
650 /**
651  * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
652  * @pf: the corresponding PF
653  *
654  * Update the Rx XOFF counter (PAUSE frames) in PFC mode
655  **/
656 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
657 {
658 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
659 	struct i40e_hw_port_stats *nsd = &pf->stats;
660 	bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
661 	struct i40e_dcbx_config *dcb_cfg;
662 	struct i40e_hw *hw = &pf->hw;
663 	u16 i, v;
664 	u8 tc;
665 
666 	dcb_cfg = &hw->local_dcbx_config;
667 
668 	/* See if DCB enabled with PFC TC */
669 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
670 	    !(dcb_cfg->pfc.pfcenable)) {
671 		i40e_update_link_xoff_rx(pf);
672 		return;
673 	}
674 
675 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
676 		u64 prio_xoff = nsd->priority_xoff_rx[i];
677 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
678 				   pf->stat_offsets_loaded,
679 				   &osd->priority_xoff_rx[i],
680 				   &nsd->priority_xoff_rx[i]);
681 
682 		/* No new PFC xoff rx */
683 		if (!(nsd->priority_xoff_rx[i] - prio_xoff))
684 			continue;
685 		/* Get the TC for given priority */
686 		tc = dcb_cfg->etscfg.prioritytable[i];
687 		xoff[tc] = true;
688 	}
689 
690 	/* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
691 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
692 		struct i40e_vsi *vsi = pf->vsi[v];
693 
694 		if (!vsi)
695 			continue;
696 
697 		for (i = 0; i < vsi->num_queue_pairs; i++) {
698 			struct i40e_ring *ring = vsi->tx_rings[i];
699 
700 			tc = ring->dcb_tc;
701 			if (xoff[tc])
702 				clear_bit(__I40E_HANG_CHECK_ARMED,
703 					  &ring->state);
704 		}
705 	}
706 }
707 
708 /**
709  * i40e_update_stats - Update the board statistics counters.
710  * @vsi: the VSI to be updated
711  *
712  * There are a few instances where we store the same stat in a
713  * couple of different structs.  This is partly because we have
714  * the netdev stats that need to be filled out, which is slightly
715  * different from the "eth_stats" defined by the chip and used in
716  * VF communications.  We sort it all out here in a central place.
717  **/
718 void i40e_update_stats(struct i40e_vsi *vsi)
719 {
720 	struct i40e_pf *pf = vsi->back;
721 	struct i40e_hw *hw = &pf->hw;
722 	struct rtnl_link_stats64 *ons;
723 	struct rtnl_link_stats64 *ns;   /* netdev stats */
724 	struct i40e_eth_stats *oes;
725 	struct i40e_eth_stats *es;     /* device's eth stats */
726 	u32 tx_restart, tx_busy;
727 	u32 rx_page, rx_buf;
728 	u64 rx_p, rx_b;
729 	u64 tx_p, tx_b;
730 	int i;
731 	u16 q;
732 
733 	if (test_bit(__I40E_DOWN, &vsi->state) ||
734 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
735 		return;
736 
737 	ns = i40e_get_vsi_stats_struct(vsi);
738 	ons = &vsi->net_stats_offsets;
739 	es = &vsi->eth_stats;
740 	oes = &vsi->eth_stats_offsets;
741 
742 	/* Gather up the netdev and vsi stats that the driver collects
743 	 * on the fly during packet processing
744 	 */
745 	rx_b = rx_p = 0;
746 	tx_b = tx_p = 0;
747 	tx_restart = tx_busy = 0;
748 	rx_page = 0;
749 	rx_buf = 0;
750 	rcu_read_lock();
751 	for (q = 0; q < vsi->num_queue_pairs; q++) {
752 		struct i40e_ring *p;
753 		u64 bytes, packets;
754 		unsigned int start;
755 
756 		/* locate Tx ring */
757 		p = ACCESS_ONCE(vsi->tx_rings[q]);
758 
759 		do {
760 			start = u64_stats_fetch_begin_bh(&p->syncp);
761 			packets = p->stats.packets;
762 			bytes = p->stats.bytes;
763 		} while (u64_stats_fetch_retry_bh(&p->syncp, start));
764 		tx_b += bytes;
765 		tx_p += packets;
766 		tx_restart += p->tx_stats.restart_queue;
767 		tx_busy += p->tx_stats.tx_busy;
768 
769 		/* Rx queue is part of the same block as Tx queue */
770 		p = &p[1];
771 		do {
772 			start = u64_stats_fetch_begin_bh(&p->syncp);
773 			packets = p->stats.packets;
774 			bytes = p->stats.bytes;
775 		} while (u64_stats_fetch_retry_bh(&p->syncp, start));
776 		rx_b += bytes;
777 		rx_p += packets;
778 		rx_buf += p->rx_stats.alloc_rx_buff_failed;
779 		rx_page += p->rx_stats.alloc_rx_page_failed;
780 	}
781 	rcu_read_unlock();
782 	vsi->tx_restart = tx_restart;
783 	vsi->tx_busy = tx_busy;
784 	vsi->rx_page_failed = rx_page;
785 	vsi->rx_buf_failed = rx_buf;
786 
787 	ns->rx_packets = rx_p;
788 	ns->rx_bytes = rx_b;
789 	ns->tx_packets = tx_p;
790 	ns->tx_bytes = tx_b;
791 
792 	i40e_update_eth_stats(vsi);
793 	/* update netdev stats from eth stats */
794 	ons->rx_errors = oes->rx_errors;
795 	ns->rx_errors = es->rx_errors;
796 	ons->tx_errors = oes->tx_errors;
797 	ns->tx_errors = es->tx_errors;
798 	ons->multicast = oes->rx_multicast;
799 	ns->multicast = es->rx_multicast;
800 	ons->tx_dropped = oes->tx_discards;
801 	ns->tx_dropped = es->tx_discards;
802 
803 	/* Get the port data only if this is the main PF VSI */
804 	if (vsi == pf->vsi[pf->lan_vsi]) {
805 		struct i40e_hw_port_stats *nsd = &pf->stats;
806 		struct i40e_hw_port_stats *osd = &pf->stats_offsets;
807 
808 		i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
809 				   I40E_GLPRT_GORCL(hw->port),
810 				   pf->stat_offsets_loaded,
811 				   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
812 		i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
813 				   I40E_GLPRT_GOTCL(hw->port),
814 				   pf->stat_offsets_loaded,
815 				   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
816 		i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
817 				   pf->stat_offsets_loaded,
818 				   &osd->eth.rx_discards,
819 				   &nsd->eth.rx_discards);
820 		i40e_stat_update32(hw, I40E_GLPRT_TDPC(hw->port),
821 				   pf->stat_offsets_loaded,
822 				   &osd->eth.tx_discards,
823 				   &nsd->eth.tx_discards);
824 		i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
825 				   I40E_GLPRT_MPRCL(hw->port),
826 				   pf->stat_offsets_loaded,
827 				   &osd->eth.rx_multicast,
828 				   &nsd->eth.rx_multicast);
829 
830 		i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
831 				   pf->stat_offsets_loaded,
832 				   &osd->tx_dropped_link_down,
833 				   &nsd->tx_dropped_link_down);
834 
835 		i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
836 				   pf->stat_offsets_loaded,
837 				   &osd->crc_errors, &nsd->crc_errors);
838 		ns->rx_crc_errors = nsd->crc_errors;
839 
840 		i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
841 				   pf->stat_offsets_loaded,
842 				   &osd->illegal_bytes, &nsd->illegal_bytes);
843 		ns->rx_errors = nsd->crc_errors
844 				+ nsd->illegal_bytes;
845 
846 		i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
847 				   pf->stat_offsets_loaded,
848 				   &osd->mac_local_faults,
849 				   &nsd->mac_local_faults);
850 		i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
851 				   pf->stat_offsets_loaded,
852 				   &osd->mac_remote_faults,
853 				   &nsd->mac_remote_faults);
854 
855 		i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
856 				   pf->stat_offsets_loaded,
857 				   &osd->rx_length_errors,
858 				   &nsd->rx_length_errors);
859 		ns->rx_length_errors = nsd->rx_length_errors;
860 
861 		i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
862 				   pf->stat_offsets_loaded,
863 				   &osd->link_xon_rx, &nsd->link_xon_rx);
864 		i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
865 				   pf->stat_offsets_loaded,
866 				   &osd->link_xon_tx, &nsd->link_xon_tx);
867 		i40e_update_prio_xoff_rx(pf);  /* handles I40E_GLPRT_LXOFFRXC */
868 		i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
869 				   pf->stat_offsets_loaded,
870 				   &osd->link_xoff_tx, &nsd->link_xoff_tx);
871 
872 		for (i = 0; i < 8; i++) {
873 			i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
874 					   pf->stat_offsets_loaded,
875 					   &osd->priority_xon_rx[i],
876 					   &nsd->priority_xon_rx[i]);
877 			i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
878 					   pf->stat_offsets_loaded,
879 					   &osd->priority_xon_tx[i],
880 					   &nsd->priority_xon_tx[i]);
881 			i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
882 					   pf->stat_offsets_loaded,
883 					   &osd->priority_xoff_tx[i],
884 					   &nsd->priority_xoff_tx[i]);
885 			i40e_stat_update32(hw,
886 					   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
887 					   pf->stat_offsets_loaded,
888 					   &osd->priority_xon_2_xoff[i],
889 					   &nsd->priority_xon_2_xoff[i]);
890 		}
891 
892 		i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
893 				   I40E_GLPRT_PRC64L(hw->port),
894 				   pf->stat_offsets_loaded,
895 				   &osd->rx_size_64, &nsd->rx_size_64);
896 		i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
897 				   I40E_GLPRT_PRC127L(hw->port),
898 				   pf->stat_offsets_loaded,
899 				   &osd->rx_size_127, &nsd->rx_size_127);
900 		i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
901 				   I40E_GLPRT_PRC255L(hw->port),
902 				   pf->stat_offsets_loaded,
903 				   &osd->rx_size_255, &nsd->rx_size_255);
904 		i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
905 				   I40E_GLPRT_PRC511L(hw->port),
906 				   pf->stat_offsets_loaded,
907 				   &osd->rx_size_511, &nsd->rx_size_511);
908 		i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
909 				   I40E_GLPRT_PRC1023L(hw->port),
910 				   pf->stat_offsets_loaded,
911 				   &osd->rx_size_1023, &nsd->rx_size_1023);
912 		i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
913 				   I40E_GLPRT_PRC1522L(hw->port),
914 				   pf->stat_offsets_loaded,
915 				   &osd->rx_size_1522, &nsd->rx_size_1522);
916 		i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
917 				   I40E_GLPRT_PRC9522L(hw->port),
918 				   pf->stat_offsets_loaded,
919 				   &osd->rx_size_big, &nsd->rx_size_big);
920 
921 		i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
922 				   I40E_GLPRT_PTC64L(hw->port),
923 				   pf->stat_offsets_loaded,
924 				   &osd->tx_size_64, &nsd->tx_size_64);
925 		i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
926 				   I40E_GLPRT_PTC127L(hw->port),
927 				   pf->stat_offsets_loaded,
928 				   &osd->tx_size_127, &nsd->tx_size_127);
929 		i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
930 				   I40E_GLPRT_PTC255L(hw->port),
931 				   pf->stat_offsets_loaded,
932 				   &osd->tx_size_255, &nsd->tx_size_255);
933 		i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
934 				   I40E_GLPRT_PTC511L(hw->port),
935 				   pf->stat_offsets_loaded,
936 				   &osd->tx_size_511, &nsd->tx_size_511);
937 		i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
938 				   I40E_GLPRT_PTC1023L(hw->port),
939 				   pf->stat_offsets_loaded,
940 				   &osd->tx_size_1023, &nsd->tx_size_1023);
941 		i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
942 				   I40E_GLPRT_PTC1522L(hw->port),
943 				   pf->stat_offsets_loaded,
944 				   &osd->tx_size_1522, &nsd->tx_size_1522);
945 		i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
946 				   I40E_GLPRT_PTC9522L(hw->port),
947 				   pf->stat_offsets_loaded,
948 				   &osd->tx_size_big, &nsd->tx_size_big);
949 
950 		i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
951 				   pf->stat_offsets_loaded,
952 				   &osd->rx_undersize, &nsd->rx_undersize);
953 		i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
954 				   pf->stat_offsets_loaded,
955 				   &osd->rx_fragments, &nsd->rx_fragments);
956 		i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
957 				   pf->stat_offsets_loaded,
958 				   &osd->rx_oversize, &nsd->rx_oversize);
959 		i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
960 				   pf->stat_offsets_loaded,
961 				   &osd->rx_jabber, &nsd->rx_jabber);
962 	}
963 
964 	pf->stat_offsets_loaded = true;
965 }
966 
967 /**
968  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
969  * @vsi: the VSI to be searched
970  * @macaddr: the MAC address
971  * @vlan: the vlan
972  * @is_vf: make sure its a vf filter, else doesn't matter
973  * @is_netdev: make sure its a netdev filter, else doesn't matter
974  *
975  * Returns ptr to the filter object or NULL
976  **/
977 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
978 						u8 *macaddr, s16 vlan,
979 						bool is_vf, bool is_netdev)
980 {
981 	struct i40e_mac_filter *f;
982 
983 	if (!vsi || !macaddr)
984 		return NULL;
985 
986 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
987 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
988 		    (vlan == f->vlan)    &&
989 		    (!is_vf || f->is_vf) &&
990 		    (!is_netdev || f->is_netdev))
991 			return f;
992 	}
993 	return NULL;
994 }
995 
996 /**
997  * i40e_find_mac - Find a mac addr in the macvlan filters list
998  * @vsi: the VSI to be searched
999  * @macaddr: the MAC address we are searching for
1000  * @is_vf: make sure its a vf filter, else doesn't matter
1001  * @is_netdev: make sure its a netdev filter, else doesn't matter
1002  *
1003  * Returns the first filter with the provided MAC address or NULL if
1004  * MAC address was not found
1005  **/
1006 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1007 				      bool is_vf, bool is_netdev)
1008 {
1009 	struct i40e_mac_filter *f;
1010 
1011 	if (!vsi || !macaddr)
1012 		return NULL;
1013 
1014 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1015 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
1016 		    (!is_vf || f->is_vf) &&
1017 		    (!is_netdev || f->is_netdev))
1018 			return f;
1019 	}
1020 	return NULL;
1021 }
1022 
1023 /**
1024  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1025  * @vsi: the VSI to be searched
1026  *
1027  * Returns true if VSI is in vlan mode or false otherwise
1028  **/
1029 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1030 {
1031 	struct i40e_mac_filter *f;
1032 
1033 	/* Only -1 for all the filters denotes not in vlan mode
1034 	 * so we have to go through all the list in order to make sure
1035 	 */
1036 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1037 		if (f->vlan >= 0)
1038 			return true;
1039 	}
1040 
1041 	return false;
1042 }
1043 
1044 /**
1045  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1046  * @vsi: the VSI to be searched
1047  * @macaddr: the mac address to be filtered
1048  * @is_vf: true if it is a vf
1049  * @is_netdev: true if it is a netdev
1050  *
1051  * Goes through all the macvlan filters and adds a
1052  * macvlan filter for each unique vlan that already exists
1053  *
1054  * Returns first filter found on success, else NULL
1055  **/
1056 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1057 					     bool is_vf, bool is_netdev)
1058 {
1059 	struct i40e_mac_filter *f;
1060 
1061 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1062 		if (!i40e_find_filter(vsi, macaddr, f->vlan,
1063 				      is_vf, is_netdev)) {
1064 			if (!i40e_add_filter(vsi, macaddr, f->vlan,
1065 						is_vf, is_netdev))
1066 				return NULL;
1067 		}
1068 	}
1069 
1070 	return list_first_entry_or_null(&vsi->mac_filter_list,
1071 					struct i40e_mac_filter, list);
1072 }
1073 
1074 /**
1075  * i40e_add_filter - Add a mac/vlan filter to the VSI
1076  * @vsi: the VSI to be searched
1077  * @macaddr: the MAC address
1078  * @vlan: the vlan
1079  * @is_vf: make sure its a vf filter, else doesn't matter
1080  * @is_netdev: make sure its a netdev filter, else doesn't matter
1081  *
1082  * Returns ptr to the filter object or NULL when no memory available.
1083  **/
1084 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1085 					u8 *macaddr, s16 vlan,
1086 					bool is_vf, bool is_netdev)
1087 {
1088 	struct i40e_mac_filter *f;
1089 
1090 	if (!vsi || !macaddr)
1091 		return NULL;
1092 
1093 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1094 	if (!f) {
1095 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1096 		if (!f)
1097 			goto add_filter_out;
1098 
1099 		memcpy(f->macaddr, macaddr, ETH_ALEN);
1100 		f->vlan = vlan;
1101 		f->changed = true;
1102 
1103 		INIT_LIST_HEAD(&f->list);
1104 		list_add(&f->list, &vsi->mac_filter_list);
1105 	}
1106 
1107 	/* increment counter and add a new flag if needed */
1108 	if (is_vf) {
1109 		if (!f->is_vf) {
1110 			f->is_vf = true;
1111 			f->counter++;
1112 		}
1113 	} else if (is_netdev) {
1114 		if (!f->is_netdev) {
1115 			f->is_netdev = true;
1116 			f->counter++;
1117 		}
1118 	} else {
1119 		f->counter++;
1120 	}
1121 
1122 	/* changed tells sync_filters_subtask to
1123 	 * push the filter down to the firmware
1124 	 */
1125 	if (f->changed) {
1126 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1127 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1128 	}
1129 
1130 add_filter_out:
1131 	return f;
1132 }
1133 
1134 /**
1135  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1136  * @vsi: the VSI to be searched
1137  * @macaddr: the MAC address
1138  * @vlan: the vlan
1139  * @is_vf: make sure it's a vf filter, else doesn't matter
1140  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1141  **/
1142 void i40e_del_filter(struct i40e_vsi *vsi,
1143 		     u8 *macaddr, s16 vlan,
1144 		     bool is_vf, bool is_netdev)
1145 {
1146 	struct i40e_mac_filter *f;
1147 
1148 	if (!vsi || !macaddr)
1149 		return;
1150 
1151 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1152 	if (!f || f->counter == 0)
1153 		return;
1154 
1155 	if (is_vf) {
1156 		if (f->is_vf) {
1157 			f->is_vf = false;
1158 			f->counter--;
1159 		}
1160 	} else if (is_netdev) {
1161 		if (f->is_netdev) {
1162 			f->is_netdev = false;
1163 			f->counter--;
1164 		}
1165 	} else {
1166 		/* make sure we don't remove a filter in use by vf or netdev */
1167 		int min_f = 0;
1168 		min_f += (f->is_vf ? 1 : 0);
1169 		min_f += (f->is_netdev ? 1 : 0);
1170 
1171 		if (f->counter > min_f)
1172 			f->counter--;
1173 	}
1174 
1175 	/* counter == 0 tells sync_filters_subtask to
1176 	 * remove the filter from the firmware's list
1177 	 */
1178 	if (f->counter == 0) {
1179 		f->changed = true;
1180 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1181 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1182 	}
1183 }
1184 
1185 /**
1186  * i40e_set_mac - NDO callback to set mac address
1187  * @netdev: network interface device structure
1188  * @p: pointer to an address structure
1189  *
1190  * Returns 0 on success, negative on failure
1191  **/
1192 static int i40e_set_mac(struct net_device *netdev, void *p)
1193 {
1194 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1195 	struct i40e_vsi *vsi = np->vsi;
1196 	struct sockaddr *addr = p;
1197 	struct i40e_mac_filter *f;
1198 
1199 	if (!is_valid_ether_addr(addr->sa_data))
1200 		return -EADDRNOTAVAIL;
1201 
1202 	netdev_info(netdev, "set mac address=%pM\n", addr->sa_data);
1203 
1204 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data))
1205 		return 0;
1206 
1207 	if (vsi->type == I40E_VSI_MAIN) {
1208 		i40e_status ret;
1209 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1210 						I40E_AQC_WRITE_TYPE_LAA_ONLY,
1211 						addr->sa_data, NULL);
1212 		if (ret) {
1213 			netdev_info(netdev,
1214 				    "Addr change for Main VSI failed: %d\n",
1215 				    ret);
1216 			return -EADDRNOTAVAIL;
1217 		}
1218 
1219 		memcpy(vsi->back->hw.mac.addr, addr->sa_data, netdev->addr_len);
1220 	}
1221 
1222 	/* In order to be sure to not drop any packets, add the new address
1223 	 * then delete the old one.
1224 	 */
1225 	f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY, false, false);
1226 	if (!f)
1227 		return -ENOMEM;
1228 
1229 	i40e_sync_vsi_filters(vsi);
1230 	i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY, false, false);
1231 	i40e_sync_vsi_filters(vsi);
1232 
1233 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1234 
1235 	return 0;
1236 }
1237 
1238 /**
1239  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1240  * @vsi: the VSI being setup
1241  * @ctxt: VSI context structure
1242  * @enabled_tc: Enabled TCs bitmap
1243  * @is_add: True if called before Add VSI
1244  *
1245  * Setup VSI queue mapping for enabled traffic classes.
1246  **/
1247 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1248 				     struct i40e_vsi_context *ctxt,
1249 				     u8 enabled_tc,
1250 				     bool is_add)
1251 {
1252 	struct i40e_pf *pf = vsi->back;
1253 	u16 sections = 0;
1254 	u8 netdev_tc = 0;
1255 	u16 numtc = 0;
1256 	u16 qcount;
1257 	u8 offset;
1258 	u16 qmap;
1259 	int i;
1260 
1261 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1262 	offset = 0;
1263 
1264 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1265 		/* Find numtc from enabled TC bitmap */
1266 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1267 			if (enabled_tc & (1 << i)) /* TC is enabled */
1268 				numtc++;
1269 		}
1270 		if (!numtc) {
1271 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1272 			numtc = 1;
1273 		}
1274 	} else {
1275 		/* At least TC0 is enabled in case of non-DCB case */
1276 		numtc = 1;
1277 	}
1278 
1279 	vsi->tc_config.numtc = numtc;
1280 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1281 
1282 	/* Setup queue offset/count for all TCs for given VSI */
1283 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1284 		/* See if the given TC is enabled for the given VSI */
1285 		if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1286 			int pow, num_qps;
1287 
1288 			vsi->tc_config.tc_info[i].qoffset = offset;
1289 			switch (vsi->type) {
1290 			case I40E_VSI_MAIN:
1291 				if (i == 0)
1292 					qcount = pf->rss_size;
1293 				else
1294 					qcount = pf->num_tc_qps;
1295 				vsi->tc_config.tc_info[i].qcount = qcount;
1296 				break;
1297 			case I40E_VSI_FDIR:
1298 			case I40E_VSI_SRIOV:
1299 			case I40E_VSI_VMDQ2:
1300 			default:
1301 				qcount = vsi->alloc_queue_pairs;
1302 				vsi->tc_config.tc_info[i].qcount = qcount;
1303 				WARN_ON(i != 0);
1304 				break;
1305 			}
1306 
1307 			/* find the power-of-2 of the number of queue pairs */
1308 			num_qps = vsi->tc_config.tc_info[i].qcount;
1309 			pow = 0;
1310 			while (num_qps &&
1311 			      ((1 << pow) < vsi->tc_config.tc_info[i].qcount)) {
1312 				pow++;
1313 				num_qps >>= 1;
1314 			}
1315 
1316 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1317 			qmap =
1318 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1319 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1320 
1321 			offset += vsi->tc_config.tc_info[i].qcount;
1322 		} else {
1323 			/* TC is not enabled so set the offset to
1324 			 * default queue and allocate one queue
1325 			 * for the given TC.
1326 			 */
1327 			vsi->tc_config.tc_info[i].qoffset = 0;
1328 			vsi->tc_config.tc_info[i].qcount = 1;
1329 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1330 
1331 			qmap = 0;
1332 		}
1333 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1334 	}
1335 
1336 	/* Set actual Tx/Rx queue pairs */
1337 	vsi->num_queue_pairs = offset;
1338 
1339 	/* Scheduler section valid can only be set for ADD VSI */
1340 	if (is_add) {
1341 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1342 
1343 		ctxt->info.up_enable_bits = enabled_tc;
1344 	}
1345 	if (vsi->type == I40E_VSI_SRIOV) {
1346 		ctxt->info.mapping_flags |=
1347 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1348 		for (i = 0; i < vsi->num_queue_pairs; i++)
1349 			ctxt->info.queue_mapping[i] =
1350 					       cpu_to_le16(vsi->base_queue + i);
1351 	} else {
1352 		ctxt->info.mapping_flags |=
1353 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1354 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1355 	}
1356 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1357 }
1358 
1359 /**
1360  * i40e_set_rx_mode - NDO callback to set the netdev filters
1361  * @netdev: network interface device structure
1362  **/
1363 static void i40e_set_rx_mode(struct net_device *netdev)
1364 {
1365 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1366 	struct i40e_mac_filter *f, *ftmp;
1367 	struct i40e_vsi *vsi = np->vsi;
1368 	struct netdev_hw_addr *uca;
1369 	struct netdev_hw_addr *mca;
1370 	struct netdev_hw_addr *ha;
1371 
1372 	/* add addr if not already in the filter list */
1373 	netdev_for_each_uc_addr(uca, netdev) {
1374 		if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1375 			if (i40e_is_vsi_in_vlan(vsi))
1376 				i40e_put_mac_in_vlan(vsi, uca->addr,
1377 						     false, true);
1378 			else
1379 				i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1380 						false, true);
1381 		}
1382 	}
1383 
1384 	netdev_for_each_mc_addr(mca, netdev) {
1385 		if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1386 			if (i40e_is_vsi_in_vlan(vsi))
1387 				i40e_put_mac_in_vlan(vsi, mca->addr,
1388 						     false, true);
1389 			else
1390 				i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1391 						false, true);
1392 		}
1393 	}
1394 
1395 	/* remove filter if not in netdev list */
1396 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1397 		bool found = false;
1398 
1399 		if (!f->is_netdev)
1400 			continue;
1401 
1402 		if (is_multicast_ether_addr(f->macaddr)) {
1403 			netdev_for_each_mc_addr(mca, netdev) {
1404 				if (ether_addr_equal(mca->addr, f->macaddr)) {
1405 					found = true;
1406 					break;
1407 				}
1408 			}
1409 		} else {
1410 			netdev_for_each_uc_addr(uca, netdev) {
1411 				if (ether_addr_equal(uca->addr, f->macaddr)) {
1412 					found = true;
1413 					break;
1414 				}
1415 			}
1416 
1417 			for_each_dev_addr(netdev, ha) {
1418 				if (ether_addr_equal(ha->addr, f->macaddr)) {
1419 					found = true;
1420 					break;
1421 				}
1422 			}
1423 		}
1424 		if (!found)
1425 			i40e_del_filter(
1426 			   vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1427 	}
1428 
1429 	/* check for other flag changes */
1430 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1431 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1432 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1433 	}
1434 }
1435 
1436 /**
1437  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1438  * @vsi: ptr to the VSI
1439  *
1440  * Push any outstanding VSI filter changes through the AdminQ.
1441  *
1442  * Returns 0 or error value
1443  **/
1444 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1445 {
1446 	struct i40e_mac_filter *f, *ftmp;
1447 	bool promisc_forced_on = false;
1448 	bool add_happened = false;
1449 	int filter_list_len = 0;
1450 	u32 changed_flags = 0;
1451 	i40e_status aq_ret = 0;
1452 	struct i40e_pf *pf;
1453 	int num_add = 0;
1454 	int num_del = 0;
1455 	u16 cmd_flags;
1456 
1457 	/* empty array typed pointers, kcalloc later */
1458 	struct i40e_aqc_add_macvlan_element_data *add_list;
1459 	struct i40e_aqc_remove_macvlan_element_data *del_list;
1460 
1461 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1462 		usleep_range(1000, 2000);
1463 	pf = vsi->back;
1464 
1465 	if (vsi->netdev) {
1466 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1467 		vsi->current_netdev_flags = vsi->netdev->flags;
1468 	}
1469 
1470 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1471 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1472 
1473 		filter_list_len = pf->hw.aq.asq_buf_size /
1474 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1475 		del_list = kcalloc(filter_list_len,
1476 			    sizeof(struct i40e_aqc_remove_macvlan_element_data),
1477 			    GFP_KERNEL);
1478 		if (!del_list)
1479 			return -ENOMEM;
1480 
1481 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1482 			if (!f->changed)
1483 				continue;
1484 
1485 			if (f->counter != 0)
1486 				continue;
1487 			f->changed = false;
1488 			cmd_flags = 0;
1489 
1490 			/* add to delete list */
1491 			memcpy(del_list[num_del].mac_addr,
1492 			       f->macaddr, ETH_ALEN);
1493 			del_list[num_del].vlan_tag =
1494 				cpu_to_le16((u16)(f->vlan ==
1495 					    I40E_VLAN_ANY ? 0 : f->vlan));
1496 
1497 			/* vlan0 as wild card to allow packets from all vlans */
1498 			if (f->vlan == I40E_VLAN_ANY ||
1499 			    (vsi->netdev && !(vsi->netdev->features &
1500 					      NETIF_F_HW_VLAN_CTAG_FILTER)))
1501 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1502 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1503 			del_list[num_del].flags = cmd_flags;
1504 			num_del++;
1505 
1506 			/* unlink from filter list */
1507 			list_del(&f->list);
1508 			kfree(f);
1509 
1510 			/* flush a full buffer */
1511 			if (num_del == filter_list_len) {
1512 				aq_ret = i40e_aq_remove_macvlan(&pf->hw,
1513 					    vsi->seid, del_list, num_del,
1514 					    NULL);
1515 				num_del = 0;
1516 				memset(del_list, 0, sizeof(*del_list));
1517 
1518 				if (aq_ret)
1519 					dev_info(&pf->pdev->dev,
1520 						 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1521 						 aq_ret,
1522 						 pf->hw.aq.asq_last_status);
1523 			}
1524 		}
1525 		if (num_del) {
1526 			aq_ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1527 						     del_list, num_del, NULL);
1528 			num_del = 0;
1529 
1530 			if (aq_ret)
1531 				dev_info(&pf->pdev->dev,
1532 					 "ignoring delete macvlan error, err %d, aq_err %d\n",
1533 					 aq_ret, pf->hw.aq.asq_last_status);
1534 		}
1535 
1536 		kfree(del_list);
1537 		del_list = NULL;
1538 
1539 		/* do all the adds now */
1540 		filter_list_len = pf->hw.aq.asq_buf_size /
1541 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1542 		add_list = kcalloc(filter_list_len,
1543 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1544 			       GFP_KERNEL);
1545 		if (!add_list)
1546 			return -ENOMEM;
1547 
1548 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1549 			if (!f->changed)
1550 				continue;
1551 
1552 			if (f->counter == 0)
1553 				continue;
1554 			f->changed = false;
1555 			add_happened = true;
1556 			cmd_flags = 0;
1557 
1558 			/* add to add array */
1559 			memcpy(add_list[num_add].mac_addr,
1560 			       f->macaddr, ETH_ALEN);
1561 			add_list[num_add].vlan_tag =
1562 				cpu_to_le16(
1563 				 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1564 			add_list[num_add].queue_number = 0;
1565 
1566 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1567 
1568 			/* vlan0 as wild card to allow packets from all vlans */
1569 			if (f->vlan == I40E_VLAN_ANY || (vsi->netdev &&
1570 			    !(vsi->netdev->features &
1571 						 NETIF_F_HW_VLAN_CTAG_FILTER)))
1572 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1573 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
1574 			num_add++;
1575 
1576 			/* flush a full buffer */
1577 			if (num_add == filter_list_len) {
1578 				aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1579 							     add_list, num_add,
1580 							     NULL);
1581 				num_add = 0;
1582 
1583 				if (aq_ret)
1584 					break;
1585 				memset(add_list, 0, sizeof(*add_list));
1586 			}
1587 		}
1588 		if (num_add) {
1589 			aq_ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1590 						     add_list, num_add, NULL);
1591 			num_add = 0;
1592 		}
1593 		kfree(add_list);
1594 		add_list = NULL;
1595 
1596 		if (add_happened && (!aq_ret)) {
1597 			/* do nothing */;
1598 		} else if (add_happened && (aq_ret)) {
1599 			dev_info(&pf->pdev->dev,
1600 				 "add filter failed, err %d, aq_err %d\n",
1601 				 aq_ret, pf->hw.aq.asq_last_status);
1602 			if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1603 			    !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1604 				      &vsi->state)) {
1605 				promisc_forced_on = true;
1606 				set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1607 					&vsi->state);
1608 				dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1609 			}
1610 		}
1611 	}
1612 
1613 	/* check for changes in promiscuous modes */
1614 	if (changed_flags & IFF_ALLMULTI) {
1615 		bool cur_multipromisc;
1616 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1617 		aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1618 							       vsi->seid,
1619 							       cur_multipromisc,
1620 							       NULL);
1621 		if (aq_ret)
1622 			dev_info(&pf->pdev->dev,
1623 				 "set multi promisc failed, err %d, aq_err %d\n",
1624 				 aq_ret, pf->hw.aq.asq_last_status);
1625 	}
1626 	if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1627 		bool cur_promisc;
1628 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1629 			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1630 					&vsi->state));
1631 		aq_ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1632 							     vsi->seid,
1633 							     cur_promisc, NULL);
1634 		if (aq_ret)
1635 			dev_info(&pf->pdev->dev,
1636 				 "set uni promisc failed, err %d, aq_err %d\n",
1637 				 aq_ret, pf->hw.aq.asq_last_status);
1638 	}
1639 
1640 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1641 	return 0;
1642 }
1643 
1644 /**
1645  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1646  * @pf: board private structure
1647  **/
1648 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1649 {
1650 	int v;
1651 
1652 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1653 		return;
1654 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1655 
1656 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
1657 		if (pf->vsi[v] &&
1658 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1659 			i40e_sync_vsi_filters(pf->vsi[v]);
1660 	}
1661 }
1662 
1663 /**
1664  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1665  * @netdev: network interface device structure
1666  * @new_mtu: new value for maximum frame size
1667  *
1668  * Returns 0 on success, negative on failure
1669  **/
1670 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1671 {
1672 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1673 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
1674 	struct i40e_vsi *vsi = np->vsi;
1675 
1676 	/* MTU < 68 is an error and causes problems on some kernels */
1677 	if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1678 		return -EINVAL;
1679 
1680 	netdev_info(netdev, "changing MTU from %d to %d\n",
1681 		    netdev->mtu, new_mtu);
1682 	netdev->mtu = new_mtu;
1683 	if (netif_running(netdev))
1684 		i40e_vsi_reinit_locked(vsi);
1685 
1686 	return 0;
1687 }
1688 
1689 /**
1690  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1691  * @vsi: the vsi being adjusted
1692  **/
1693 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1694 {
1695 	struct i40e_vsi_context ctxt;
1696 	i40e_status ret;
1697 
1698 	if ((vsi->info.valid_sections &
1699 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1700 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1701 		return;  /* already enabled */
1702 
1703 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1704 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1705 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1706 
1707 	ctxt.seid = vsi->seid;
1708 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1709 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1710 	if (ret) {
1711 		dev_info(&vsi->back->pdev->dev,
1712 			 "%s: update vsi failed, aq_err=%d\n",
1713 			 __func__, vsi->back->hw.aq.asq_last_status);
1714 	}
1715 }
1716 
1717 /**
1718  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1719  * @vsi: the vsi being adjusted
1720  **/
1721 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
1722 {
1723 	struct i40e_vsi_context ctxt;
1724 	i40e_status ret;
1725 
1726 	if ((vsi->info.valid_sections &
1727 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1728 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
1729 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
1730 		return;  /* already disabled */
1731 
1732 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1733 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1734 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
1735 
1736 	ctxt.seid = vsi->seid;
1737 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1738 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1739 	if (ret) {
1740 		dev_info(&vsi->back->pdev->dev,
1741 			 "%s: update vsi failed, aq_err=%d\n",
1742 			 __func__, vsi->back->hw.aq.asq_last_status);
1743 	}
1744 }
1745 
1746 /**
1747  * i40e_vlan_rx_register - Setup or shutdown vlan offload
1748  * @netdev: network interface to be adjusted
1749  * @features: netdev features to test if VLAN offload is enabled or not
1750  **/
1751 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
1752 {
1753 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1754 	struct i40e_vsi *vsi = np->vsi;
1755 
1756 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
1757 		i40e_vlan_stripping_enable(vsi);
1758 	else
1759 		i40e_vlan_stripping_disable(vsi);
1760 }
1761 
1762 /**
1763  * i40e_vsi_add_vlan - Add vsi membership for given vlan
1764  * @vsi: the vsi being configured
1765  * @vid: vlan id to be added (0 = untagged only , -1 = any)
1766  **/
1767 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
1768 {
1769 	struct i40e_mac_filter *f, *add_f;
1770 	bool is_netdev, is_vf;
1771 	int ret;
1772 
1773 	is_vf = (vsi->type == I40E_VSI_SRIOV);
1774 	is_netdev = !!(vsi->netdev);
1775 
1776 	if (is_netdev) {
1777 		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
1778 					is_vf, is_netdev);
1779 		if (!add_f) {
1780 			dev_info(&vsi->back->pdev->dev,
1781 				 "Could not add vlan filter %d for %pM\n",
1782 				 vid, vsi->netdev->dev_addr);
1783 			return -ENOMEM;
1784 		}
1785 	}
1786 
1787 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1788 		add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1789 		if (!add_f) {
1790 			dev_info(&vsi->back->pdev->dev,
1791 				 "Could not add vlan filter %d for %pM\n",
1792 				 vid, f->macaddr);
1793 			return -ENOMEM;
1794 		}
1795 	}
1796 
1797 	ret = i40e_sync_vsi_filters(vsi);
1798 	if (ret) {
1799 		dev_info(&vsi->back->pdev->dev,
1800 			 "Could not sync filters for vid %d\n", vid);
1801 		return ret;
1802 	}
1803 
1804 	/* Now if we add a vlan tag, make sure to check if it is the first
1805 	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
1806 	 * with 0, so we now accept untagged and specified tagged traffic
1807 	 * (and not any taged and untagged)
1808 	 */
1809 	if (vid > 0) {
1810 		if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
1811 						  I40E_VLAN_ANY,
1812 						  is_vf, is_netdev)) {
1813 			i40e_del_filter(vsi, vsi->netdev->dev_addr,
1814 					I40E_VLAN_ANY, is_vf, is_netdev);
1815 			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
1816 						is_vf, is_netdev);
1817 			if (!add_f) {
1818 				dev_info(&vsi->back->pdev->dev,
1819 					 "Could not add filter 0 for %pM\n",
1820 					 vsi->netdev->dev_addr);
1821 				return -ENOMEM;
1822 			}
1823 		}
1824 
1825 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
1826 			if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1827 					     is_vf, is_netdev)) {
1828 				i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1829 						is_vf, is_netdev);
1830 				add_f = i40e_add_filter(vsi, f->macaddr,
1831 							0, is_vf, is_netdev);
1832 				if (!add_f) {
1833 					dev_info(&vsi->back->pdev->dev,
1834 						 "Could not add filter 0 for %pM\n",
1835 						 f->macaddr);
1836 					return -ENOMEM;
1837 				}
1838 			}
1839 		}
1840 		ret = i40e_sync_vsi_filters(vsi);
1841 	}
1842 
1843 	return ret;
1844 }
1845 
1846 /**
1847  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
1848  * @vsi: the vsi being configured
1849  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
1850  *
1851  * Return: 0 on success or negative otherwise
1852  **/
1853 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
1854 {
1855 	struct net_device *netdev = vsi->netdev;
1856 	struct i40e_mac_filter *f, *add_f;
1857 	bool is_vf, is_netdev;
1858 	int filter_count = 0;
1859 	int ret;
1860 
1861 	is_vf = (vsi->type == I40E_VSI_SRIOV);
1862 	is_netdev = !!(netdev);
1863 
1864 	if (is_netdev)
1865 		i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
1866 
1867 	list_for_each_entry(f, &vsi->mac_filter_list, list)
1868 		i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1869 
1870 	ret = i40e_sync_vsi_filters(vsi);
1871 	if (ret) {
1872 		dev_info(&vsi->back->pdev->dev, "Could not sync filters\n");
1873 		return ret;
1874 	}
1875 
1876 	/* go through all the filters for this VSI and if there is only
1877 	 * vid == 0 it means there are no other filters, so vid 0 must
1878 	 * be replaced with -1. This signifies that we should from now
1879 	 * on accept any traffic (with any tag present, or untagged)
1880 	 */
1881 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1882 		if (is_netdev) {
1883 			if (f->vlan &&
1884 			    ether_addr_equal(netdev->dev_addr, f->macaddr))
1885 				filter_count++;
1886 		}
1887 
1888 		if (f->vlan)
1889 			filter_count++;
1890 	}
1891 
1892 	if (!filter_count && is_netdev) {
1893 		i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
1894 		f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1895 				    is_vf, is_netdev);
1896 		if (!f) {
1897 			dev_info(&vsi->back->pdev->dev,
1898 				 "Could not add filter %d for %pM\n",
1899 				 I40E_VLAN_ANY, netdev->dev_addr);
1900 			return -ENOMEM;
1901 		}
1902 	}
1903 
1904 	if (!filter_count) {
1905 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
1906 			i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
1907 			add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1908 					    is_vf, is_netdev);
1909 			if (!add_f) {
1910 				dev_info(&vsi->back->pdev->dev,
1911 					 "Could not add filter %d for %pM\n",
1912 					 I40E_VLAN_ANY, f->macaddr);
1913 				return -ENOMEM;
1914 			}
1915 		}
1916 	}
1917 
1918 	return i40e_sync_vsi_filters(vsi);
1919 }
1920 
1921 /**
1922  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
1923  * @netdev: network interface to be adjusted
1924  * @vid: vlan id to be added
1925  *
1926  * net_device_ops implementation for adding vlan ids
1927  **/
1928 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
1929 				__always_unused __be16 proto, u16 vid)
1930 {
1931 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1932 	struct i40e_vsi *vsi = np->vsi;
1933 	int ret = 0;
1934 
1935 	if (vid > 4095)
1936 		return -EINVAL;
1937 
1938 	netdev_info(netdev, "adding %pM vid=%d\n", netdev->dev_addr, vid);
1939 
1940 	/* If the network stack called us with vid = 0, we should
1941 	 * indicate to i40e_vsi_add_vlan() that we want to receive
1942 	 * any traffic (i.e. with any vlan tag, or untagged)
1943 	 */
1944 	ret = i40e_vsi_add_vlan(vsi, vid ? vid : I40E_VLAN_ANY);
1945 
1946 	if (!ret && (vid < VLAN_N_VID))
1947 		set_bit(vid, vsi->active_vlans);
1948 
1949 	return ret;
1950 }
1951 
1952 /**
1953  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
1954  * @netdev: network interface to be adjusted
1955  * @vid: vlan id to be removed
1956  *
1957  * net_device_ops implementation for adding vlan ids
1958  **/
1959 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
1960 				 __always_unused __be16 proto, u16 vid)
1961 {
1962 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1963 	struct i40e_vsi *vsi = np->vsi;
1964 
1965 	netdev_info(netdev, "removing %pM vid=%d\n", netdev->dev_addr, vid);
1966 
1967 	/* return code is ignored as there is nothing a user
1968 	 * can do about failure to remove and a log message was
1969 	 * already printed from the other function
1970 	 */
1971 	i40e_vsi_kill_vlan(vsi, vid);
1972 
1973 	clear_bit(vid, vsi->active_vlans);
1974 
1975 	return 0;
1976 }
1977 
1978 /**
1979  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
1980  * @vsi: the vsi being brought back up
1981  **/
1982 static void i40e_restore_vlan(struct i40e_vsi *vsi)
1983 {
1984 	u16 vid;
1985 
1986 	if (!vsi->netdev)
1987 		return;
1988 
1989 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
1990 
1991 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
1992 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
1993 				     vid);
1994 }
1995 
1996 /**
1997  * i40e_vsi_add_pvid - Add pvid for the VSI
1998  * @vsi: the vsi being adjusted
1999  * @vid: the vlan id to set as a PVID
2000  **/
2001 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2002 {
2003 	struct i40e_vsi_context ctxt;
2004 	i40e_status aq_ret;
2005 
2006 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2007 	vsi->info.pvid = cpu_to_le16(vid);
2008 	vsi->info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_INSERT_PVID;
2009 	vsi->info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_UNTAGGED;
2010 
2011 	ctxt.seid = vsi->seid;
2012 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
2013 	aq_ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2014 	if (aq_ret) {
2015 		dev_info(&vsi->back->pdev->dev,
2016 			 "%s: update vsi failed, aq_err=%d\n",
2017 			 __func__, vsi->back->hw.aq.asq_last_status);
2018 		return -ENOENT;
2019 	}
2020 
2021 	return 0;
2022 }
2023 
2024 /**
2025  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2026  * @vsi: the vsi being adjusted
2027  *
2028  * Just use the vlan_rx_register() service to put it back to normal
2029  **/
2030 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2031 {
2032 	vsi->info.pvid = 0;
2033 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2034 }
2035 
2036 /**
2037  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2038  * @vsi: ptr to the VSI
2039  *
2040  * If this function returns with an error, then it's possible one or
2041  * more of the rings is populated (while the rest are not).  It is the
2042  * callers duty to clean those orphaned rings.
2043  *
2044  * Return 0 on success, negative on failure
2045  **/
2046 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2047 {
2048 	int i, err = 0;
2049 
2050 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2051 		err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2052 
2053 	return err;
2054 }
2055 
2056 /**
2057  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2058  * @vsi: ptr to the VSI
2059  *
2060  * Free VSI's transmit software resources
2061  **/
2062 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2063 {
2064 	int i;
2065 
2066 	for (i = 0; i < vsi->num_queue_pairs; i++)
2067 		if (vsi->tx_rings[i]->desc)
2068 			i40e_free_tx_resources(vsi->tx_rings[i]);
2069 }
2070 
2071 /**
2072  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2073  * @vsi: ptr to the VSI
2074  *
2075  * If this function returns with an error, then it's possible one or
2076  * more of the rings is populated (while the rest are not).  It is the
2077  * callers duty to clean those orphaned rings.
2078  *
2079  * Return 0 on success, negative on failure
2080  **/
2081 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2082 {
2083 	int i, err = 0;
2084 
2085 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2086 		err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2087 	return err;
2088 }
2089 
2090 /**
2091  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2092  * @vsi: ptr to the VSI
2093  *
2094  * Free all receive software resources
2095  **/
2096 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2097 {
2098 	int i;
2099 
2100 	for (i = 0; i < vsi->num_queue_pairs; i++)
2101 		if (vsi->rx_rings[i]->desc)
2102 			i40e_free_rx_resources(vsi->rx_rings[i]);
2103 }
2104 
2105 /**
2106  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2107  * @ring: The Tx ring to configure
2108  *
2109  * Configure the Tx descriptor ring in the HMC context.
2110  **/
2111 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2112 {
2113 	struct i40e_vsi *vsi = ring->vsi;
2114 	u16 pf_q = vsi->base_queue + ring->queue_index;
2115 	struct i40e_hw *hw = &vsi->back->hw;
2116 	struct i40e_hmc_obj_txq tx_ctx;
2117 	i40e_status err = 0;
2118 	u32 qtx_ctl = 0;
2119 
2120 	/* some ATR related tx ring init */
2121 	if (vsi->back->flags & I40E_FLAG_FDIR_ATR_ENABLED) {
2122 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2123 		ring->atr_count = 0;
2124 	} else {
2125 		ring->atr_sample_rate = 0;
2126 	}
2127 
2128 	/* initialize XPS */
2129 	if (ring->q_vector && ring->netdev &&
2130 	    !test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2131 		netif_set_xps_queue(ring->netdev,
2132 				    &ring->q_vector->affinity_mask,
2133 				    ring->queue_index);
2134 
2135 	/* clear the context structure first */
2136 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2137 
2138 	tx_ctx.new_context = 1;
2139 	tx_ctx.base = (ring->dma / 128);
2140 	tx_ctx.qlen = ring->count;
2141 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FDIR_ENABLED |
2142 			I40E_FLAG_FDIR_ATR_ENABLED));
2143 
2144 	/* As part of VSI creation/update, FW allocates certain
2145 	 * Tx arbitration queue sets for each TC enabled for
2146 	 * the VSI. The FW returns the handles to these queue
2147 	 * sets as part of the response buffer to Add VSI,
2148 	 * Update VSI, etc. AQ commands. It is expected that
2149 	 * these queue set handles be associated with the Tx
2150 	 * queues by the driver as part of the TX queue context
2151 	 * initialization. This has to be done regardless of
2152 	 * DCB as by default everything is mapped to TC0.
2153 	 */
2154 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2155 	tx_ctx.rdylist_act = 0;
2156 
2157 	/* clear the context in the HMC */
2158 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2159 	if (err) {
2160 		dev_info(&vsi->back->pdev->dev,
2161 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2162 			 ring->queue_index, pf_q, err);
2163 		return -ENOMEM;
2164 	}
2165 
2166 	/* set the context in the HMC */
2167 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2168 	if (err) {
2169 		dev_info(&vsi->back->pdev->dev,
2170 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2171 			 ring->queue_index, pf_q, err);
2172 		return -ENOMEM;
2173 	}
2174 
2175 	/* Now associate this queue with this PCI function */
2176 	qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2177 	qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2178 		    I40E_QTX_CTL_PF_INDX_MASK);
2179 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2180 	i40e_flush(hw);
2181 
2182 	clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2183 
2184 	/* cache tail off for easier writes later */
2185 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2186 
2187 	return 0;
2188 }
2189 
2190 /**
2191  * i40e_configure_rx_ring - Configure a receive ring context
2192  * @ring: The Rx ring to configure
2193  *
2194  * Configure the Rx descriptor ring in the HMC context.
2195  **/
2196 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2197 {
2198 	struct i40e_vsi *vsi = ring->vsi;
2199 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2200 	u16 pf_q = vsi->base_queue + ring->queue_index;
2201 	struct i40e_hw *hw = &vsi->back->hw;
2202 	struct i40e_hmc_obj_rxq rx_ctx;
2203 	i40e_status err = 0;
2204 
2205 	ring->state = 0;
2206 
2207 	/* clear the context structure first */
2208 	memset(&rx_ctx, 0, sizeof(rx_ctx));
2209 
2210 	ring->rx_buf_len = vsi->rx_buf_len;
2211 	ring->rx_hdr_len = vsi->rx_hdr_len;
2212 
2213 	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2214 	rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2215 
2216 	rx_ctx.base = (ring->dma / 128);
2217 	rx_ctx.qlen = ring->count;
2218 
2219 	if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2220 		set_ring_16byte_desc_enabled(ring);
2221 		rx_ctx.dsize = 0;
2222 	} else {
2223 		rx_ctx.dsize = 1;
2224 	}
2225 
2226 	rx_ctx.dtype = vsi->dtype;
2227 	if (vsi->dtype) {
2228 		set_ring_ps_enabled(ring);
2229 		rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2230 				  I40E_RX_SPLIT_IP      |
2231 				  I40E_RX_SPLIT_TCP_UDP |
2232 				  I40E_RX_SPLIT_SCTP;
2233 	} else {
2234 		rx_ctx.hsplit_0 = 0;
2235 	}
2236 
2237 	rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2238 				  (chain_len * ring->rx_buf_len));
2239 	rx_ctx.tphrdesc_ena = 1;
2240 	rx_ctx.tphwdesc_ena = 1;
2241 	rx_ctx.tphdata_ena = 1;
2242 	rx_ctx.tphhead_ena = 1;
2243 	rx_ctx.lrxqthresh = 2;
2244 	rx_ctx.crcstrip = 1;
2245 	rx_ctx.l2tsel = 1;
2246 	rx_ctx.showiv = 1;
2247 
2248 	/* clear the context in the HMC */
2249 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2250 	if (err) {
2251 		dev_info(&vsi->back->pdev->dev,
2252 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2253 			 ring->queue_index, pf_q, err);
2254 		return -ENOMEM;
2255 	}
2256 
2257 	/* set the context in the HMC */
2258 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2259 	if (err) {
2260 		dev_info(&vsi->back->pdev->dev,
2261 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2262 			 ring->queue_index, pf_q, err);
2263 		return -ENOMEM;
2264 	}
2265 
2266 	/* cache tail for quicker writes, and clear the reg before use */
2267 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2268 	writel(0, ring->tail);
2269 
2270 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2271 
2272 	return 0;
2273 }
2274 
2275 /**
2276  * i40e_vsi_configure_tx - Configure the VSI for Tx
2277  * @vsi: VSI structure describing this set of rings and resources
2278  *
2279  * Configure the Tx VSI for operation.
2280  **/
2281 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2282 {
2283 	int err = 0;
2284 	u16 i;
2285 
2286 	for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2287 		err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2288 
2289 	return err;
2290 }
2291 
2292 /**
2293  * i40e_vsi_configure_rx - Configure the VSI for Rx
2294  * @vsi: the VSI being configured
2295  *
2296  * Configure the Rx VSI for operation.
2297  **/
2298 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2299 {
2300 	int err = 0;
2301 	u16 i;
2302 
2303 	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2304 		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2305 			       + ETH_FCS_LEN + VLAN_HLEN;
2306 	else
2307 		vsi->max_frame = I40E_RXBUFFER_2048;
2308 
2309 	/* figure out correct receive buffer length */
2310 	switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2311 				    I40E_FLAG_RX_PS_ENABLED)) {
2312 	case I40E_FLAG_RX_1BUF_ENABLED:
2313 		vsi->rx_hdr_len = 0;
2314 		vsi->rx_buf_len = vsi->max_frame;
2315 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2316 		break;
2317 	case I40E_FLAG_RX_PS_ENABLED:
2318 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2319 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2320 		vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2321 		break;
2322 	default:
2323 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2324 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2325 		vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2326 		break;
2327 	}
2328 
2329 	/* round up for the chip's needs */
2330 	vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2331 				(1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2332 	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2333 				(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2334 
2335 	/* set up individual rings */
2336 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2337 		err = i40e_configure_rx_ring(vsi->rx_rings[i]);
2338 
2339 	return err;
2340 }
2341 
2342 /**
2343  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2344  * @vsi: ptr to the VSI
2345  **/
2346 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2347 {
2348 	u16 qoffset, qcount;
2349 	int i, n;
2350 
2351 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED))
2352 		return;
2353 
2354 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2355 		if (!(vsi->tc_config.enabled_tc & (1 << n)))
2356 			continue;
2357 
2358 		qoffset = vsi->tc_config.tc_info[n].qoffset;
2359 		qcount = vsi->tc_config.tc_info[n].qcount;
2360 		for (i = qoffset; i < (qoffset + qcount); i++) {
2361 			struct i40e_ring *rx_ring = vsi->rx_rings[i];
2362 			struct i40e_ring *tx_ring = vsi->tx_rings[i];
2363 			rx_ring->dcb_tc = n;
2364 			tx_ring->dcb_tc = n;
2365 		}
2366 	}
2367 }
2368 
2369 /**
2370  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2371  * @vsi: ptr to the VSI
2372  **/
2373 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2374 {
2375 	if (vsi->netdev)
2376 		i40e_set_rx_mode(vsi->netdev);
2377 }
2378 
2379 /**
2380  * i40e_vsi_configure - Set up the VSI for action
2381  * @vsi: the VSI being configured
2382  **/
2383 static int i40e_vsi_configure(struct i40e_vsi *vsi)
2384 {
2385 	int err;
2386 
2387 	i40e_set_vsi_rx_mode(vsi);
2388 	i40e_restore_vlan(vsi);
2389 	i40e_vsi_config_dcb_rings(vsi);
2390 	err = i40e_vsi_configure_tx(vsi);
2391 	if (!err)
2392 		err = i40e_vsi_configure_rx(vsi);
2393 
2394 	return err;
2395 }
2396 
2397 /**
2398  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2399  * @vsi: the VSI being configured
2400  **/
2401 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2402 {
2403 	struct i40e_pf *pf = vsi->back;
2404 	struct i40e_q_vector *q_vector;
2405 	struct i40e_hw *hw = &pf->hw;
2406 	u16 vector;
2407 	int i, q;
2408 	u32 val;
2409 	u32 qp;
2410 
2411 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
2412 	 * and PFINT_LNKLSTn registers, e.g.:
2413 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
2414 	 */
2415 	qp = vsi->base_queue;
2416 	vector = vsi->base_vector;
2417 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
2418 		q_vector = vsi->q_vectors[i];
2419 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2420 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
2421 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2422 		     q_vector->rx.itr);
2423 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2424 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
2425 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2426 		     q_vector->tx.itr);
2427 
2428 		/* Linked list for the queuepairs assigned to this vector */
2429 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2430 		for (q = 0; q < q_vector->num_ringpairs; q++) {
2431 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2432 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
2433 			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2434 			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2435 			      (I40E_QUEUE_TYPE_TX
2436 				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2437 
2438 			wr32(hw, I40E_QINT_RQCTL(qp), val);
2439 
2440 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2441 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
2442 			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2443 			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2444 			      (I40E_QUEUE_TYPE_RX
2445 				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2446 
2447 			/* Terminate the linked list */
2448 			if (q == (q_vector->num_ringpairs - 1))
2449 				val |= (I40E_QUEUE_END_OF_LIST
2450 					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2451 
2452 			wr32(hw, I40E_QINT_TQCTL(qp), val);
2453 			qp++;
2454 		}
2455 	}
2456 
2457 	i40e_flush(hw);
2458 }
2459 
2460 /**
2461  * i40e_enable_misc_int_causes - enable the non-queue interrupts
2462  * @hw: ptr to the hardware info
2463  **/
2464 static void i40e_enable_misc_int_causes(struct i40e_hw *hw)
2465 {
2466 	u32 val;
2467 
2468 	/* clear things first */
2469 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
2470 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
2471 
2472 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
2473 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
2474 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
2475 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2476 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
2477 	      I40E_PFINT_ICR0_ENA_STORM_DETECT_MASK  |
2478 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
2479 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
2480 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2481 
2482 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
2483 
2484 	/* SW_ITR_IDX = 0, but don't change INTENA */
2485 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTLN_SW_ITR_INDX_MASK |
2486 					I40E_PFINT_DYN_CTLN_INTENA_MSK_MASK);
2487 
2488 	/* OTHER_ITR_IDX = 0 */
2489 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2490 }
2491 
2492 /**
2493  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2494  * @vsi: the VSI being configured
2495  **/
2496 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2497 {
2498 	struct i40e_q_vector *q_vector = vsi->q_vectors[0];
2499 	struct i40e_pf *pf = vsi->back;
2500 	struct i40e_hw *hw = &pf->hw;
2501 	u32 val;
2502 
2503 	/* set the ITR configuration */
2504 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2505 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
2506 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2507 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2508 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
2509 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2510 
2511 	i40e_enable_misc_int_causes(hw);
2512 
2513 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2514 	wr32(hw, I40E_PFINT_LNKLST0, 0);
2515 
2516 	/* Associate the queue pair to the vector and enable the q int */
2517 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
2518 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2519 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2520 
2521 	wr32(hw, I40E_QINT_RQCTL(0), val);
2522 
2523 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
2524 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2525 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2526 
2527 	wr32(hw, I40E_QINT_TQCTL(0), val);
2528 	i40e_flush(hw);
2529 }
2530 
2531 /**
2532  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2533  * @pf: board private structure
2534  **/
2535 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2536 {
2537 	struct i40e_hw *hw = &pf->hw;
2538 	u32 val;
2539 
2540 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
2541 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2542 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2543 
2544 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2545 	i40e_flush(hw);
2546 }
2547 
2548 /**
2549  * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2550  * @vsi: pointer to a vsi
2551  * @vector: enable a particular Hw Interrupt vector
2552  **/
2553 void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2554 {
2555 	struct i40e_pf *pf = vsi->back;
2556 	struct i40e_hw *hw = &pf->hw;
2557 	u32 val;
2558 
2559 	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2560 	      I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2561 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2562 	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2563 	/* skip the flush */
2564 }
2565 
2566 /**
2567  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2568  * @irq: interrupt number
2569  * @data: pointer to a q_vector
2570  **/
2571 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2572 {
2573 	struct i40e_q_vector *q_vector = data;
2574 
2575 	if (!q_vector->tx.ring && !q_vector->rx.ring)
2576 		return IRQ_HANDLED;
2577 
2578 	napi_schedule(&q_vector->napi);
2579 
2580 	return IRQ_HANDLED;
2581 }
2582 
2583 /**
2584  * i40e_fdir_clean_rings - Interrupt Handler for FDIR rings
2585  * @irq: interrupt number
2586  * @data: pointer to a q_vector
2587  **/
2588 static irqreturn_t i40e_fdir_clean_rings(int irq, void *data)
2589 {
2590 	struct i40e_q_vector *q_vector = data;
2591 
2592 	if (!q_vector->tx.ring && !q_vector->rx.ring)
2593 		return IRQ_HANDLED;
2594 
2595 	pr_info("fdir ring cleaning needed\n");
2596 
2597 	return IRQ_HANDLED;
2598 }
2599 
2600 /**
2601  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
2602  * @vsi: the VSI being configured
2603  * @basename: name for the vector
2604  *
2605  * Allocates MSI-X vectors and requests interrupts from the kernel.
2606  **/
2607 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
2608 {
2609 	int q_vectors = vsi->num_q_vectors;
2610 	struct i40e_pf *pf = vsi->back;
2611 	int base = vsi->base_vector;
2612 	int rx_int_idx = 0;
2613 	int tx_int_idx = 0;
2614 	int vector, err;
2615 
2616 	for (vector = 0; vector < q_vectors; vector++) {
2617 		struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
2618 
2619 		if (q_vector->tx.ring && q_vector->rx.ring) {
2620 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2621 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2622 			tx_int_idx++;
2623 		} else if (q_vector->rx.ring) {
2624 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2625 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
2626 		} else if (q_vector->tx.ring) {
2627 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2628 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
2629 		} else {
2630 			/* skip this unused q_vector */
2631 			continue;
2632 		}
2633 		err = request_irq(pf->msix_entries[base + vector].vector,
2634 				  vsi->irq_handler,
2635 				  0,
2636 				  q_vector->name,
2637 				  q_vector);
2638 		if (err) {
2639 			dev_info(&pf->pdev->dev,
2640 				 "%s: request_irq failed, error: %d\n",
2641 				 __func__, err);
2642 			goto free_queue_irqs;
2643 		}
2644 		/* assign the mask for this irq */
2645 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2646 				      &q_vector->affinity_mask);
2647 	}
2648 
2649 	return 0;
2650 
2651 free_queue_irqs:
2652 	while (vector) {
2653 		vector--;
2654 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2655 				      NULL);
2656 		free_irq(pf->msix_entries[base + vector].vector,
2657 			 &(vsi->q_vectors[vector]));
2658 	}
2659 	return err;
2660 }
2661 
2662 /**
2663  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
2664  * @vsi: the VSI being un-configured
2665  **/
2666 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
2667 {
2668 	struct i40e_pf *pf = vsi->back;
2669 	struct i40e_hw *hw = &pf->hw;
2670 	int base = vsi->base_vector;
2671 	int i;
2672 
2673 	for (i = 0; i < vsi->num_queue_pairs; i++) {
2674 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
2675 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
2676 	}
2677 
2678 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2679 		for (i = vsi->base_vector;
2680 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
2681 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
2682 
2683 		i40e_flush(hw);
2684 		for (i = 0; i < vsi->num_q_vectors; i++)
2685 			synchronize_irq(pf->msix_entries[i + base].vector);
2686 	} else {
2687 		/* Legacy and MSI mode - this stops all interrupt handling */
2688 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
2689 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
2690 		i40e_flush(hw);
2691 		synchronize_irq(pf->pdev->irq);
2692 	}
2693 }
2694 
2695 /**
2696  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
2697  * @vsi: the VSI being configured
2698  **/
2699 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
2700 {
2701 	struct i40e_pf *pf = vsi->back;
2702 	int i;
2703 
2704 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2705 		for (i = vsi->base_vector;
2706 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
2707 			i40e_irq_dynamic_enable(vsi, i);
2708 	} else {
2709 		i40e_irq_dynamic_enable_icr0(pf);
2710 	}
2711 
2712 	i40e_flush(&pf->hw);
2713 	return 0;
2714 }
2715 
2716 /**
2717  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
2718  * @pf: board private structure
2719  **/
2720 static void i40e_stop_misc_vector(struct i40e_pf *pf)
2721 {
2722 	/* Disable ICR 0 */
2723 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
2724 	i40e_flush(&pf->hw);
2725 }
2726 
2727 /**
2728  * i40e_intr - MSI/Legacy and non-queue interrupt handler
2729  * @irq: interrupt number
2730  * @data: pointer to a q_vector
2731  *
2732  * This is the handler used for all MSI/Legacy interrupts, and deals
2733  * with both queue and non-queue interrupts.  This is also used in
2734  * MSIX mode to handle the non-queue interrupts.
2735  **/
2736 static irqreturn_t i40e_intr(int irq, void *data)
2737 {
2738 	struct i40e_pf *pf = (struct i40e_pf *)data;
2739 	struct i40e_hw *hw = &pf->hw;
2740 	u32 icr0, icr0_remaining;
2741 	u32 val, ena_mask;
2742 
2743 	icr0 = rd32(hw, I40E_PFINT_ICR0);
2744 
2745 	val = rd32(hw, I40E_PFINT_DYN_CTL0);
2746 	val = val | I40E_PFINT_DYN_CTL0_CLEARPBA_MASK;
2747 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2748 
2749 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
2750 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
2751 		return IRQ_NONE;
2752 
2753 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
2754 
2755 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
2756 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
2757 
2758 		/* temporarily disable queue cause for NAPI processing */
2759 		u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
2760 		qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
2761 		wr32(hw, I40E_QINT_RQCTL(0), qval);
2762 
2763 		qval = rd32(hw, I40E_QINT_TQCTL(0));
2764 		qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
2765 		wr32(hw, I40E_QINT_TQCTL(0), qval);
2766 
2767 		if (!test_bit(__I40E_DOWN, &pf->state))
2768 			napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0]->napi);
2769 	}
2770 
2771 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
2772 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2773 		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
2774 	}
2775 
2776 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
2777 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
2778 		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
2779 	}
2780 
2781 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
2782 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
2783 		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
2784 	}
2785 
2786 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
2787 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
2788 			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
2789 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
2790 		val = rd32(hw, I40E_GLGEN_RSTAT);
2791 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
2792 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
2793 		if (val & I40E_RESET_CORER)
2794 			pf->corer_count++;
2795 		else if (val & I40E_RESET_GLOBR)
2796 			pf->globr_count++;
2797 		else if (val & I40E_RESET_EMPR)
2798 			pf->empr_count++;
2799 	}
2800 
2801 	/* If a critical error is pending we have no choice but to reset the
2802 	 * device.
2803 	 * Report and mask out any remaining unexpected interrupts.
2804 	 */
2805 	icr0_remaining = icr0 & ena_mask;
2806 	if (icr0_remaining) {
2807 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
2808 			 icr0_remaining);
2809 		if ((icr0_remaining & I40E_PFINT_ICR0_HMC_ERR_MASK) ||
2810 		    (icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
2811 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
2812 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK) ||
2813 		    (icr0_remaining & I40E_PFINT_ICR0_MAL_DETECT_MASK)) {
2814 			if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
2815 				dev_info(&pf->pdev->dev, "HMC error interrupt\n");
2816 			} else {
2817 				dev_info(&pf->pdev->dev, "device will be reset\n");
2818 				set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2819 				i40e_service_event_schedule(pf);
2820 			}
2821 		}
2822 		ena_mask &= ~icr0_remaining;
2823 	}
2824 
2825 	/* re-enable interrupt causes */
2826 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
2827 	if (!test_bit(__I40E_DOWN, &pf->state)) {
2828 		i40e_service_event_schedule(pf);
2829 		i40e_irq_dynamic_enable_icr0(pf);
2830 	}
2831 
2832 	return IRQ_HANDLED;
2833 }
2834 
2835 /**
2836  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
2837  * @vsi: the VSI being configured
2838  * @v_idx: vector index
2839  * @qp_idx: queue pair index
2840  **/
2841 static void map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
2842 {
2843 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
2844 	struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
2845 	struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
2846 
2847 	tx_ring->q_vector = q_vector;
2848 	tx_ring->next = q_vector->tx.ring;
2849 	q_vector->tx.ring = tx_ring;
2850 	q_vector->tx.count++;
2851 
2852 	rx_ring->q_vector = q_vector;
2853 	rx_ring->next = q_vector->rx.ring;
2854 	q_vector->rx.ring = rx_ring;
2855 	q_vector->rx.count++;
2856 }
2857 
2858 /**
2859  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
2860  * @vsi: the VSI being configured
2861  *
2862  * This function maps descriptor rings to the queue-specific vectors
2863  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
2864  * one vector per queue pair, but on a constrained vector budget, we
2865  * group the queue pairs as "efficiently" as possible.
2866  **/
2867 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
2868 {
2869 	int qp_remaining = vsi->num_queue_pairs;
2870 	int q_vectors = vsi->num_q_vectors;
2871 	int num_ringpairs;
2872 	int v_start = 0;
2873 	int qp_idx = 0;
2874 
2875 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
2876 	 * group them so there are multiple queues per vector.
2877 	 */
2878 	for (; v_start < q_vectors && qp_remaining; v_start++) {
2879 		struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
2880 
2881 		num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
2882 
2883 		q_vector->num_ringpairs = num_ringpairs;
2884 
2885 		q_vector->rx.count = 0;
2886 		q_vector->tx.count = 0;
2887 		q_vector->rx.ring = NULL;
2888 		q_vector->tx.ring = NULL;
2889 
2890 		while (num_ringpairs--) {
2891 			map_vector_to_qp(vsi, v_start, qp_idx);
2892 			qp_idx++;
2893 			qp_remaining--;
2894 		}
2895 	}
2896 }
2897 
2898 /**
2899  * i40e_vsi_request_irq - Request IRQ from the OS
2900  * @vsi: the VSI being configured
2901  * @basename: name for the vector
2902  **/
2903 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
2904 {
2905 	struct i40e_pf *pf = vsi->back;
2906 	int err;
2907 
2908 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
2909 		err = i40e_vsi_request_irq_msix(vsi, basename);
2910 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
2911 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
2912 				  pf->misc_int_name, pf);
2913 	else
2914 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
2915 				  pf->misc_int_name, pf);
2916 
2917 	if (err)
2918 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
2919 
2920 	return err;
2921 }
2922 
2923 #ifdef CONFIG_NET_POLL_CONTROLLER
2924 /**
2925  * i40e_netpoll - A Polling 'interrupt'handler
2926  * @netdev: network interface device structure
2927  *
2928  * This is used by netconsole to send skbs without having to re-enable
2929  * interrupts.  It's not called while the normal interrupt routine is executing.
2930  **/
2931 static void i40e_netpoll(struct net_device *netdev)
2932 {
2933 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2934 	struct i40e_vsi *vsi = np->vsi;
2935 	struct i40e_pf *pf = vsi->back;
2936 	int i;
2937 
2938 	/* if interface is down do nothing */
2939 	if (test_bit(__I40E_DOWN, &vsi->state))
2940 		return;
2941 
2942 	pf->flags |= I40E_FLAG_IN_NETPOLL;
2943 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2944 		for (i = 0; i < vsi->num_q_vectors; i++)
2945 			i40e_msix_clean_rings(0, vsi->q_vectors[i]);
2946 	} else {
2947 		i40e_intr(pf->pdev->irq, netdev);
2948 	}
2949 	pf->flags &= ~I40E_FLAG_IN_NETPOLL;
2950 }
2951 #endif
2952 
2953 /**
2954  * i40e_vsi_control_tx - Start or stop a VSI's rings
2955  * @vsi: the VSI being configured
2956  * @enable: start or stop the rings
2957  **/
2958 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
2959 {
2960 	struct i40e_pf *pf = vsi->back;
2961 	struct i40e_hw *hw = &pf->hw;
2962 	int i, j, pf_q;
2963 	u32 tx_reg;
2964 
2965 	pf_q = vsi->base_queue;
2966 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
2967 		j = 1000;
2968 		do {
2969 			usleep_range(1000, 2000);
2970 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
2971 		} while (j-- && ((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT)
2972 			       ^ (tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT)) & 1);
2973 
2974 		if (enable) {
2975 			/* is STAT set ? */
2976 			if ((tx_reg & I40E_QTX_ENA_QENA_STAT_MASK)) {
2977 				dev_info(&pf->pdev->dev,
2978 					 "Tx %d already enabled\n", i);
2979 				continue;
2980 			}
2981 		} else {
2982 			/* is !STAT set ? */
2983 			if (!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK)) {
2984 				dev_info(&pf->pdev->dev,
2985 					 "Tx %d already disabled\n", i);
2986 				continue;
2987 			}
2988 		}
2989 
2990 		/* turn on/off the queue */
2991 		if (enable)
2992 			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK |
2993 				  I40E_QTX_ENA_QENA_STAT_MASK;
2994 		else
2995 			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
2996 
2997 		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
2998 
2999 		/* wait for the change to finish */
3000 		for (j = 0; j < 10; j++) {
3001 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3002 			if (enable) {
3003 				if ((tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3004 					break;
3005 			} else {
3006 				if (!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3007 					break;
3008 			}
3009 
3010 			udelay(10);
3011 		}
3012 		if (j >= 10) {
3013 			dev_info(&pf->pdev->dev, "Tx ring %d %sable timeout\n",
3014 				 pf_q, (enable ? "en" : "dis"));
3015 			return -ETIMEDOUT;
3016 		}
3017 	}
3018 
3019 	return 0;
3020 }
3021 
3022 /**
3023  * i40e_vsi_control_rx - Start or stop a VSI's rings
3024  * @vsi: the VSI being configured
3025  * @enable: start or stop the rings
3026  **/
3027 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3028 {
3029 	struct i40e_pf *pf = vsi->back;
3030 	struct i40e_hw *hw = &pf->hw;
3031 	int i, j, pf_q;
3032 	u32 rx_reg;
3033 
3034 	pf_q = vsi->base_queue;
3035 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3036 		j = 1000;
3037 		do {
3038 			usleep_range(1000, 2000);
3039 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3040 		} while (j-- && ((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT)
3041 			       ^ (rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT)) & 1);
3042 
3043 		if (enable) {
3044 			/* is STAT set ? */
3045 			if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3046 				continue;
3047 		} else {
3048 			/* is !STAT set ? */
3049 			if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3050 				continue;
3051 		}
3052 
3053 		/* turn on/off the queue */
3054 		if (enable)
3055 			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK |
3056 				  I40E_QRX_ENA_QENA_STAT_MASK;
3057 		else
3058 			rx_reg &= ~(I40E_QRX_ENA_QENA_REQ_MASK |
3059 				  I40E_QRX_ENA_QENA_STAT_MASK);
3060 		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3061 
3062 		/* wait for the change to finish */
3063 		for (j = 0; j < 10; j++) {
3064 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3065 
3066 			if (enable) {
3067 				if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3068 					break;
3069 			} else {
3070 				if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3071 					break;
3072 			}
3073 
3074 			udelay(10);
3075 		}
3076 		if (j >= 10) {
3077 			dev_info(&pf->pdev->dev, "Rx ring %d %sable timeout\n",
3078 				 pf_q, (enable ? "en" : "dis"));
3079 			return -ETIMEDOUT;
3080 		}
3081 	}
3082 
3083 	return 0;
3084 }
3085 
3086 /**
3087  * i40e_vsi_control_rings - Start or stop a VSI's rings
3088  * @vsi: the VSI being configured
3089  * @enable: start or stop the rings
3090  **/
3091 static int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3092 {
3093 	int ret;
3094 
3095 	/* do rx first for enable and last for disable */
3096 	if (request) {
3097 		ret = i40e_vsi_control_rx(vsi, request);
3098 		if (ret)
3099 			return ret;
3100 		ret = i40e_vsi_control_tx(vsi, request);
3101 	} else {
3102 		ret = i40e_vsi_control_tx(vsi, request);
3103 		if (ret)
3104 			return ret;
3105 		ret = i40e_vsi_control_rx(vsi, request);
3106 	}
3107 
3108 	return ret;
3109 }
3110 
3111 /**
3112  * i40e_vsi_free_irq - Free the irq association with the OS
3113  * @vsi: the VSI being configured
3114  **/
3115 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3116 {
3117 	struct i40e_pf *pf = vsi->back;
3118 	struct i40e_hw *hw = &pf->hw;
3119 	int base = vsi->base_vector;
3120 	u32 val, qp;
3121 	int i;
3122 
3123 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3124 		if (!vsi->q_vectors)
3125 			return;
3126 
3127 		for (i = 0; i < vsi->num_q_vectors; i++) {
3128 			u16 vector = i + base;
3129 
3130 			/* free only the irqs that were actually requested */
3131 			if (vsi->q_vectors[i]->num_ringpairs == 0)
3132 				continue;
3133 
3134 			/* clear the affinity_mask in the IRQ descriptor */
3135 			irq_set_affinity_hint(pf->msix_entries[vector].vector,
3136 					      NULL);
3137 			free_irq(pf->msix_entries[vector].vector,
3138 				 vsi->q_vectors[i]);
3139 
3140 			/* Tear down the interrupt queue link list
3141 			 *
3142 			 * We know that they come in pairs and always
3143 			 * the Rx first, then the Tx.  To clear the
3144 			 * link list, stick the EOL value into the
3145 			 * next_q field of the registers.
3146 			 */
3147 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3148 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3149 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3150 			val |= I40E_QUEUE_END_OF_LIST
3151 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3152 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3153 
3154 			while (qp != I40E_QUEUE_END_OF_LIST) {
3155 				u32 next;
3156 
3157 				val = rd32(hw, I40E_QINT_RQCTL(qp));
3158 
3159 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3160 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3161 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3162 					 I40E_QINT_RQCTL_INTEVENT_MASK);
3163 
3164 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3165 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3166 
3167 				wr32(hw, I40E_QINT_RQCTL(qp), val);
3168 
3169 				val = rd32(hw, I40E_QINT_TQCTL(qp));
3170 
3171 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3172 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3173 
3174 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3175 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3176 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3177 					 I40E_QINT_TQCTL_INTEVENT_MASK);
3178 
3179 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3180 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3181 
3182 				wr32(hw, I40E_QINT_TQCTL(qp), val);
3183 				qp = next;
3184 			}
3185 		}
3186 	} else {
3187 		free_irq(pf->pdev->irq, pf);
3188 
3189 		val = rd32(hw, I40E_PFINT_LNKLST0);
3190 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3191 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3192 		val |= I40E_QUEUE_END_OF_LIST
3193 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3194 		wr32(hw, I40E_PFINT_LNKLST0, val);
3195 
3196 		val = rd32(hw, I40E_QINT_RQCTL(qp));
3197 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3198 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3199 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3200 			 I40E_QINT_RQCTL_INTEVENT_MASK);
3201 
3202 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3203 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3204 
3205 		wr32(hw, I40E_QINT_RQCTL(qp), val);
3206 
3207 		val = rd32(hw, I40E_QINT_TQCTL(qp));
3208 
3209 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3210 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3211 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3212 			 I40E_QINT_TQCTL_INTEVENT_MASK);
3213 
3214 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3215 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3216 
3217 		wr32(hw, I40E_QINT_TQCTL(qp), val);
3218 	}
3219 }
3220 
3221 /**
3222  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
3223  * @vsi: the VSI being configured
3224  * @v_idx: Index of vector to be freed
3225  *
3226  * This function frees the memory allocated to the q_vector.  In addition if
3227  * NAPI is enabled it will delete any references to the NAPI struct prior
3228  * to freeing the q_vector.
3229  **/
3230 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
3231 {
3232 	struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3233 	struct i40e_ring *ring;
3234 
3235 	if (!q_vector)
3236 		return;
3237 
3238 	/* disassociate q_vector from rings */
3239 	i40e_for_each_ring(ring, q_vector->tx)
3240 		ring->q_vector = NULL;
3241 
3242 	i40e_for_each_ring(ring, q_vector->rx)
3243 		ring->q_vector = NULL;
3244 
3245 	/* only VSI w/ an associated netdev is set up w/ NAPI */
3246 	if (vsi->netdev)
3247 		netif_napi_del(&q_vector->napi);
3248 
3249 	vsi->q_vectors[v_idx] = NULL;
3250 
3251 	kfree_rcu(q_vector, rcu);
3252 }
3253 
3254 /**
3255  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3256  * @vsi: the VSI being un-configured
3257  *
3258  * This frees the memory allocated to the q_vectors and
3259  * deletes references to the NAPI struct.
3260  **/
3261 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3262 {
3263 	int v_idx;
3264 
3265 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
3266 		i40e_free_q_vector(vsi, v_idx);
3267 }
3268 
3269 /**
3270  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3271  * @pf: board private structure
3272  **/
3273 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3274 {
3275 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3276 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3277 		pci_disable_msix(pf->pdev);
3278 		kfree(pf->msix_entries);
3279 		pf->msix_entries = NULL;
3280 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3281 		pci_disable_msi(pf->pdev);
3282 	}
3283 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3284 }
3285 
3286 /**
3287  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3288  * @pf: board private structure
3289  *
3290  * We go through and clear interrupt specific resources and reset the structure
3291  * to pre-load conditions
3292  **/
3293 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3294 {
3295 	int i;
3296 
3297 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3298 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
3299 		if (pf->vsi[i])
3300 			i40e_vsi_free_q_vectors(pf->vsi[i]);
3301 	i40e_reset_interrupt_capability(pf);
3302 }
3303 
3304 /**
3305  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3306  * @vsi: the VSI being configured
3307  **/
3308 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3309 {
3310 	int q_idx;
3311 
3312 	if (!vsi->netdev)
3313 		return;
3314 
3315 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3316 		napi_enable(&vsi->q_vectors[q_idx]->napi);
3317 }
3318 
3319 /**
3320  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3321  * @vsi: the VSI being configured
3322  **/
3323 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3324 {
3325 	int q_idx;
3326 
3327 	if (!vsi->netdev)
3328 		return;
3329 
3330 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3331 		napi_disable(&vsi->q_vectors[q_idx]->napi);
3332 }
3333 
3334 /**
3335  * i40e_quiesce_vsi - Pause a given VSI
3336  * @vsi: the VSI being paused
3337  **/
3338 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3339 {
3340 	if (test_bit(__I40E_DOWN, &vsi->state))
3341 		return;
3342 
3343 	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3344 	if (vsi->netdev && netif_running(vsi->netdev)) {
3345 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3346 	} else {
3347 		set_bit(__I40E_DOWN, &vsi->state);
3348 		i40e_down(vsi);
3349 	}
3350 }
3351 
3352 /**
3353  * i40e_unquiesce_vsi - Resume a given VSI
3354  * @vsi: the VSI being resumed
3355  **/
3356 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3357 {
3358 	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3359 		return;
3360 
3361 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3362 	if (vsi->netdev && netif_running(vsi->netdev))
3363 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3364 	else
3365 		i40e_up(vsi);   /* this clears the DOWN bit */
3366 }
3367 
3368 /**
3369  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3370  * @pf: the PF
3371  **/
3372 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3373 {
3374 	int v;
3375 
3376 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3377 		if (pf->vsi[v])
3378 			i40e_quiesce_vsi(pf->vsi[v]);
3379 	}
3380 }
3381 
3382 /**
3383  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3384  * @pf: the PF
3385  **/
3386 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3387 {
3388 	int v;
3389 
3390 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3391 		if (pf->vsi[v])
3392 			i40e_unquiesce_vsi(pf->vsi[v]);
3393 	}
3394 }
3395 
3396 /**
3397  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
3398  * @dcbcfg: the corresponding DCBx configuration structure
3399  *
3400  * Return the number of TCs from given DCBx configuration
3401  **/
3402 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
3403 {
3404 	u8 num_tc = 0;
3405 	int i;
3406 
3407 	/* Scan the ETS Config Priority Table to find
3408 	 * traffic class enabled for a given priority
3409 	 * and use the traffic class index to get the
3410 	 * number of traffic classes enabled
3411 	 */
3412 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3413 		if (dcbcfg->etscfg.prioritytable[i] > num_tc)
3414 			num_tc = dcbcfg->etscfg.prioritytable[i];
3415 	}
3416 
3417 	/* Traffic class index starts from zero so
3418 	 * increment to return the actual count
3419 	 */
3420 	return num_tc + 1;
3421 }
3422 
3423 /**
3424  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
3425  * @dcbcfg: the corresponding DCBx configuration structure
3426  *
3427  * Query the current DCB configuration and return the number of
3428  * traffic classes enabled from the given DCBX config
3429  **/
3430 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
3431 {
3432 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
3433 	u8 enabled_tc = 1;
3434 	u8 i;
3435 
3436 	for (i = 0; i < num_tc; i++)
3437 		enabled_tc |= 1 << i;
3438 
3439 	return enabled_tc;
3440 }
3441 
3442 /**
3443  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
3444  * @pf: PF being queried
3445  *
3446  * Return number of traffic classes enabled for the given PF
3447  **/
3448 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
3449 {
3450 	struct i40e_hw *hw = &pf->hw;
3451 	u8 i, enabled_tc;
3452 	u8 num_tc = 0;
3453 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3454 
3455 	/* If DCB is not enabled then always in single TC */
3456 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3457 		return 1;
3458 
3459 	/* MFP mode return count of enabled TCs for this PF */
3460 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3461 		enabled_tc = pf->hw.func_caps.enabled_tcmap;
3462 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3463 			if (enabled_tc & (1 << i))
3464 				num_tc++;
3465 		}
3466 		return num_tc;
3467 	}
3468 
3469 	/* SFP mode will be enabled for all TCs on port */
3470 	return i40e_dcb_get_num_tc(dcbcfg);
3471 }
3472 
3473 /**
3474  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
3475  * @pf: PF being queried
3476  *
3477  * Return a bitmap for first enabled traffic class for this PF.
3478  **/
3479 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
3480 {
3481 	u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3482 	u8 i = 0;
3483 
3484 	if (!enabled_tc)
3485 		return 0x1; /* TC0 */
3486 
3487 	/* Find the first enabled TC */
3488 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3489 		if (enabled_tc & (1 << i))
3490 			break;
3491 	}
3492 
3493 	return 1 << i;
3494 }
3495 
3496 /**
3497  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
3498  * @pf: PF being queried
3499  *
3500  * Return a bitmap for enabled traffic classes for this PF.
3501  **/
3502 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
3503 {
3504 	/* If DCB is not enabled for this PF then just return default TC */
3505 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3506 		return i40e_pf_get_default_tc(pf);
3507 
3508 	/* MFP mode will have enabled TCs set by FW */
3509 	if (pf->flags & I40E_FLAG_MFP_ENABLED)
3510 		return pf->hw.func_caps.enabled_tcmap;
3511 
3512 	/* SFP mode we want PF to be enabled for all TCs */
3513 	return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
3514 }
3515 
3516 /**
3517  * i40e_vsi_get_bw_info - Query VSI BW Information
3518  * @vsi: the VSI being queried
3519  *
3520  * Returns 0 on success, negative value on failure
3521  **/
3522 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
3523 {
3524 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
3525 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
3526 	struct i40e_pf *pf = vsi->back;
3527 	struct i40e_hw *hw = &pf->hw;
3528 	i40e_status aq_ret;
3529 	u32 tc_bw_max;
3530 	int i;
3531 
3532 	/* Get the VSI level BW configuration */
3533 	aq_ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
3534 	if (aq_ret) {
3535 		dev_info(&pf->pdev->dev,
3536 			 "couldn't get pf vsi bw config, err %d, aq_err %d\n",
3537 			 aq_ret, pf->hw.aq.asq_last_status);
3538 		return -EINVAL;
3539 	}
3540 
3541 	/* Get the VSI level BW configuration per TC */
3542 	aq_ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
3543 					          NULL);
3544 	if (aq_ret) {
3545 		dev_info(&pf->pdev->dev,
3546 			 "couldn't get pf vsi ets bw config, err %d, aq_err %d\n",
3547 			 aq_ret, pf->hw.aq.asq_last_status);
3548 		return -EINVAL;
3549 	}
3550 
3551 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
3552 		dev_info(&pf->pdev->dev,
3553 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
3554 			 bw_config.tc_valid_bits,
3555 			 bw_ets_config.tc_valid_bits);
3556 		/* Still continuing */
3557 	}
3558 
3559 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
3560 	vsi->bw_max_quanta = bw_config.max_bw;
3561 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
3562 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
3563 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3564 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
3565 		vsi->bw_ets_limit_credits[i] =
3566 					le16_to_cpu(bw_ets_config.credits[i]);
3567 		/* 3 bits out of 4 for each TC */
3568 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
3569 	}
3570 
3571 	return 0;
3572 }
3573 
3574 /**
3575  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
3576  * @vsi: the VSI being configured
3577  * @enabled_tc: TC bitmap
3578  * @bw_credits: BW shared credits per TC
3579  *
3580  * Returns 0 on success, negative value on failure
3581  **/
3582 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
3583 				       u8 *bw_share)
3584 {
3585 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
3586 	i40e_status aq_ret;
3587 	int i;
3588 
3589 	bw_data.tc_valid_bits = enabled_tc;
3590 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3591 		bw_data.tc_bw_credits[i] = bw_share[i];
3592 
3593 	aq_ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
3594 					  NULL);
3595 	if (aq_ret) {
3596 		dev_info(&vsi->back->pdev->dev,
3597 			 "%s: AQ command Config VSI BW allocation per TC failed = %d\n",
3598 			 __func__, vsi->back->hw.aq.asq_last_status);
3599 		return -EINVAL;
3600 	}
3601 
3602 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3603 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
3604 
3605 	return 0;
3606 }
3607 
3608 /**
3609  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
3610  * @vsi: the VSI being configured
3611  * @enabled_tc: TC map to be enabled
3612  *
3613  **/
3614 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3615 {
3616 	struct net_device *netdev = vsi->netdev;
3617 	struct i40e_pf *pf = vsi->back;
3618 	struct i40e_hw *hw = &pf->hw;
3619 	u8 netdev_tc = 0;
3620 	int i;
3621 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3622 
3623 	if (!netdev)
3624 		return;
3625 
3626 	if (!enabled_tc) {
3627 		netdev_reset_tc(netdev);
3628 		return;
3629 	}
3630 
3631 	/* Set up actual enabled TCs on the VSI */
3632 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
3633 		return;
3634 
3635 	/* set per TC queues for the VSI */
3636 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3637 		/* Only set TC queues for enabled tcs
3638 		 *
3639 		 * e.g. For a VSI that has TC0 and TC3 enabled the
3640 		 * enabled_tc bitmap would be 0x00001001; the driver
3641 		 * will set the numtc for netdev as 2 that will be
3642 		 * referenced by the netdev layer as TC 0 and 1.
3643 		 */
3644 		if (vsi->tc_config.enabled_tc & (1 << i))
3645 			netdev_set_tc_queue(netdev,
3646 					vsi->tc_config.tc_info[i].netdev_tc,
3647 					vsi->tc_config.tc_info[i].qcount,
3648 					vsi->tc_config.tc_info[i].qoffset);
3649 	}
3650 
3651 	/* Assign UP2TC map for the VSI */
3652 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3653 		/* Get the actual TC# for the UP */
3654 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
3655 		/* Get the mapped netdev TC# for the UP */
3656 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
3657 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
3658 	}
3659 }
3660 
3661 /**
3662  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
3663  * @vsi: the VSI being configured
3664  * @ctxt: the ctxt buffer returned from AQ VSI update param command
3665  **/
3666 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
3667 				      struct i40e_vsi_context *ctxt)
3668 {
3669 	/* copy just the sections touched not the entire info
3670 	 * since not all sections are valid as returned by
3671 	 * update vsi params
3672 	 */
3673 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
3674 	memcpy(&vsi->info.queue_mapping,
3675 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
3676 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
3677 	       sizeof(vsi->info.tc_mapping));
3678 }
3679 
3680 /**
3681  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
3682  * @vsi: VSI to be configured
3683  * @enabled_tc: TC bitmap
3684  *
3685  * This configures a particular VSI for TCs that are mapped to the
3686  * given TC bitmap. It uses default bandwidth share for TCs across
3687  * VSIs to configure TC for a particular VSI.
3688  *
3689  * NOTE:
3690  * It is expected that the VSI queues have been quisced before calling
3691  * this function.
3692  **/
3693 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3694 {
3695 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
3696 	struct i40e_vsi_context ctxt;
3697 	int ret = 0;
3698 	int i;
3699 
3700 	/* Check if enabled_tc is same as existing or new TCs */
3701 	if (vsi->tc_config.enabled_tc == enabled_tc)
3702 		return ret;
3703 
3704 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
3705 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3706 		if (enabled_tc & (1 << i))
3707 			bw_share[i] = 1;
3708 	}
3709 
3710 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
3711 	if (ret) {
3712 		dev_info(&vsi->back->pdev->dev,
3713 			 "Failed configuring TC map %d for VSI %d\n",
3714 			 enabled_tc, vsi->seid);
3715 		goto out;
3716 	}
3717 
3718 	/* Update Queue Pairs Mapping for currently enabled UPs */
3719 	ctxt.seid = vsi->seid;
3720 	ctxt.pf_num = vsi->back->hw.pf_id;
3721 	ctxt.vf_num = 0;
3722 	ctxt.uplink_seid = vsi->uplink_seid;
3723 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
3724 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
3725 
3726 	/* Update the VSI after updating the VSI queue-mapping information */
3727 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
3728 	if (ret) {
3729 		dev_info(&vsi->back->pdev->dev,
3730 			 "update vsi failed, aq_err=%d\n",
3731 			 vsi->back->hw.aq.asq_last_status);
3732 		goto out;
3733 	}
3734 	/* update the local VSI info with updated queue map */
3735 	i40e_vsi_update_queue_map(vsi, &ctxt);
3736 	vsi->info.valid_sections = 0;
3737 
3738 	/* Update current VSI BW information */
3739 	ret = i40e_vsi_get_bw_info(vsi);
3740 	if (ret) {
3741 		dev_info(&vsi->back->pdev->dev,
3742 			 "Failed updating vsi bw info, aq_err=%d\n",
3743 			 vsi->back->hw.aq.asq_last_status);
3744 		goto out;
3745 	}
3746 
3747 	/* Update the netdev TC setup */
3748 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
3749 out:
3750 	return ret;
3751 }
3752 
3753 /**
3754  * i40e_up_complete - Finish the last steps of bringing up a connection
3755  * @vsi: the VSI being configured
3756  **/
3757 static int i40e_up_complete(struct i40e_vsi *vsi)
3758 {
3759 	struct i40e_pf *pf = vsi->back;
3760 	int err;
3761 
3762 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3763 		i40e_vsi_configure_msix(vsi);
3764 	else
3765 		i40e_configure_msi_and_legacy(vsi);
3766 
3767 	/* start rings */
3768 	err = i40e_vsi_control_rings(vsi, true);
3769 	if (err)
3770 		return err;
3771 
3772 	clear_bit(__I40E_DOWN, &vsi->state);
3773 	i40e_napi_enable_all(vsi);
3774 	i40e_vsi_enable_irq(vsi);
3775 
3776 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
3777 	    (vsi->netdev)) {
3778 		netdev_info(vsi->netdev, "NIC Link is Up\n");
3779 		netif_tx_start_all_queues(vsi->netdev);
3780 		netif_carrier_on(vsi->netdev);
3781 	} else if (vsi->netdev) {
3782 		netdev_info(vsi->netdev, "NIC Link is Down\n");
3783 	}
3784 	i40e_service_event_schedule(pf);
3785 
3786 	return 0;
3787 }
3788 
3789 /**
3790  * i40e_vsi_reinit_locked - Reset the VSI
3791  * @vsi: the VSI being configured
3792  *
3793  * Rebuild the ring structs after some configuration
3794  * has changed, e.g. MTU size.
3795  **/
3796 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
3797 {
3798 	struct i40e_pf *pf = vsi->back;
3799 
3800 	WARN_ON(in_interrupt());
3801 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
3802 		usleep_range(1000, 2000);
3803 	i40e_down(vsi);
3804 
3805 	/* Give a VF some time to respond to the reset.  The
3806 	 * two second wait is based upon the watchdog cycle in
3807 	 * the VF driver.
3808 	 */
3809 	if (vsi->type == I40E_VSI_SRIOV)
3810 		msleep(2000);
3811 	i40e_up(vsi);
3812 	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
3813 }
3814 
3815 /**
3816  * i40e_up - Bring the connection back up after being down
3817  * @vsi: the VSI being configured
3818  **/
3819 int i40e_up(struct i40e_vsi *vsi)
3820 {
3821 	int err;
3822 
3823 	err = i40e_vsi_configure(vsi);
3824 	if (!err)
3825 		err = i40e_up_complete(vsi);
3826 
3827 	return err;
3828 }
3829 
3830 /**
3831  * i40e_down - Shutdown the connection processing
3832  * @vsi: the VSI being stopped
3833  **/
3834 void i40e_down(struct i40e_vsi *vsi)
3835 {
3836 	int i;
3837 
3838 	/* It is assumed that the caller of this function
3839 	 * sets the vsi->state __I40E_DOWN bit.
3840 	 */
3841 	if (vsi->netdev) {
3842 		netif_carrier_off(vsi->netdev);
3843 		netif_tx_disable(vsi->netdev);
3844 	}
3845 	i40e_vsi_disable_irq(vsi);
3846 	i40e_vsi_control_rings(vsi, false);
3847 	i40e_napi_disable_all(vsi);
3848 
3849 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3850 		i40e_clean_tx_ring(vsi->tx_rings[i]);
3851 		i40e_clean_rx_ring(vsi->rx_rings[i]);
3852 	}
3853 }
3854 
3855 /**
3856  * i40e_setup_tc - configure multiple traffic classes
3857  * @netdev: net device to configure
3858  * @tc: number of traffic classes to enable
3859  **/
3860 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
3861 {
3862 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3863 	struct i40e_vsi *vsi = np->vsi;
3864 	struct i40e_pf *pf = vsi->back;
3865 	u8 enabled_tc = 0;
3866 	int ret = -EINVAL;
3867 	int i;
3868 
3869 	/* Check if DCB enabled to continue */
3870 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
3871 		netdev_info(netdev, "DCB is not enabled for adapter\n");
3872 		goto exit;
3873 	}
3874 
3875 	/* Check if MFP enabled */
3876 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3877 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
3878 		goto exit;
3879 	}
3880 
3881 	/* Check whether tc count is within enabled limit */
3882 	if (tc > i40e_pf_get_num_tc(pf)) {
3883 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
3884 		goto exit;
3885 	}
3886 
3887 	/* Generate TC map for number of tc requested */
3888 	for (i = 0; i < tc; i++)
3889 		enabled_tc |= (1 << i);
3890 
3891 	/* Requesting same TC configuration as already enabled */
3892 	if (enabled_tc == vsi->tc_config.enabled_tc)
3893 		return 0;
3894 
3895 	/* Quiesce VSI queues */
3896 	i40e_quiesce_vsi(vsi);
3897 
3898 	/* Configure VSI for enabled TCs */
3899 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
3900 	if (ret) {
3901 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
3902 			    vsi->seid);
3903 		goto exit;
3904 	}
3905 
3906 	/* Unquiesce VSI */
3907 	i40e_unquiesce_vsi(vsi);
3908 
3909 exit:
3910 	return ret;
3911 }
3912 
3913 /**
3914  * i40e_open - Called when a network interface is made active
3915  * @netdev: network interface device structure
3916  *
3917  * The open entry point is called when a network interface is made
3918  * active by the system (IFF_UP).  At this point all resources needed
3919  * for transmit and receive operations are allocated, the interrupt
3920  * handler is registered with the OS, the netdev watchdog subtask is
3921  * enabled, and the stack is notified that the interface is ready.
3922  *
3923  * Returns 0 on success, negative value on failure
3924  **/
3925 static int i40e_open(struct net_device *netdev)
3926 {
3927 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3928 	struct i40e_vsi *vsi = np->vsi;
3929 	struct i40e_pf *pf = vsi->back;
3930 	char int_name[IFNAMSIZ];
3931 	int err;
3932 
3933 	/* disallow open during test */
3934 	if (test_bit(__I40E_TESTING, &pf->state))
3935 		return -EBUSY;
3936 
3937 	netif_carrier_off(netdev);
3938 
3939 	/* allocate descriptors */
3940 	err = i40e_vsi_setup_tx_resources(vsi);
3941 	if (err)
3942 		goto err_setup_tx;
3943 	err = i40e_vsi_setup_rx_resources(vsi);
3944 	if (err)
3945 		goto err_setup_rx;
3946 
3947 	err = i40e_vsi_configure(vsi);
3948 	if (err)
3949 		goto err_setup_rx;
3950 
3951 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3952 		 dev_driver_string(&pf->pdev->dev), netdev->name);
3953 	err = i40e_vsi_request_irq(vsi, int_name);
3954 	if (err)
3955 		goto err_setup_rx;
3956 
3957 	err = i40e_up_complete(vsi);
3958 	if (err)
3959 		goto err_up_complete;
3960 
3961 	if ((vsi->type == I40E_VSI_MAIN) || (vsi->type == I40E_VSI_VMDQ2)) {
3962 		err = i40e_aq_set_vsi_broadcast(&pf->hw, vsi->seid, true, NULL);
3963 		if (err)
3964 			netdev_info(netdev,
3965 				    "couldn't set broadcast err %d aq_err %d\n",
3966 				    err, pf->hw.aq.asq_last_status);
3967 	}
3968 
3969 	return 0;
3970 
3971 err_up_complete:
3972 	i40e_down(vsi);
3973 	i40e_vsi_free_irq(vsi);
3974 err_setup_rx:
3975 	i40e_vsi_free_rx_resources(vsi);
3976 err_setup_tx:
3977 	i40e_vsi_free_tx_resources(vsi);
3978 	if (vsi == pf->vsi[pf->lan_vsi])
3979 		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
3980 
3981 	return err;
3982 }
3983 
3984 /**
3985  * i40e_close - Disables a network interface
3986  * @netdev: network interface device structure
3987  *
3988  * The close entry point is called when an interface is de-activated
3989  * by the OS.  The hardware is still under the driver's control, but
3990  * this netdev interface is disabled.
3991  *
3992  * Returns 0, this is not allowed to fail
3993  **/
3994 static int i40e_close(struct net_device *netdev)
3995 {
3996 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3997 	struct i40e_vsi *vsi = np->vsi;
3998 
3999 	if (test_and_set_bit(__I40E_DOWN, &vsi->state))
4000 		return 0;
4001 
4002 	i40e_down(vsi);
4003 	i40e_vsi_free_irq(vsi);
4004 
4005 	i40e_vsi_free_tx_resources(vsi);
4006 	i40e_vsi_free_rx_resources(vsi);
4007 
4008 	return 0;
4009 }
4010 
4011 /**
4012  * i40e_do_reset - Start a PF or Core Reset sequence
4013  * @pf: board private structure
4014  * @reset_flags: which reset is requested
4015  *
4016  * The essential difference in resets is that the PF Reset
4017  * doesn't clear the packet buffers, doesn't reset the PE
4018  * firmware, and doesn't bother the other PFs on the chip.
4019  **/
4020 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
4021 {
4022 	u32 val;
4023 
4024 	WARN_ON(in_interrupt());
4025 
4026 	/* do the biggest reset indicated */
4027 	if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
4028 
4029 		/* Request a Global Reset
4030 		 *
4031 		 * This will start the chip's countdown to the actual full
4032 		 * chip reset event, and a warning interrupt to be sent
4033 		 * to all PFs, including the requestor.  Our handler
4034 		 * for the warning interrupt will deal with the shutdown
4035 		 * and recovery of the switch setup.
4036 		 */
4037 		dev_info(&pf->pdev->dev, "GlobalR requested\n");
4038 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4039 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
4040 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4041 
4042 	} else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
4043 
4044 		/* Request a Core Reset
4045 		 *
4046 		 * Same as Global Reset, except does *not* include the MAC/PHY
4047 		 */
4048 		dev_info(&pf->pdev->dev, "CoreR requested\n");
4049 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
4050 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
4051 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
4052 		i40e_flush(&pf->hw);
4053 
4054 	} else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
4055 
4056 		/* Request a PF Reset
4057 		 *
4058 		 * Resets only the PF-specific registers
4059 		 *
4060 		 * This goes directly to the tear-down and rebuild of
4061 		 * the switch, since we need to do all the recovery as
4062 		 * for the Core Reset.
4063 		 */
4064 		dev_info(&pf->pdev->dev, "PFR requested\n");
4065 		i40e_handle_reset_warning(pf);
4066 
4067 	} else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
4068 		int v;
4069 
4070 		/* Find the VSI(s) that requested a re-init */
4071 		dev_info(&pf->pdev->dev,
4072 			 "VSI reinit requested\n");
4073 		for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4074 			struct i40e_vsi *vsi = pf->vsi[v];
4075 			if (vsi != NULL &&
4076 			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
4077 				i40e_vsi_reinit_locked(pf->vsi[v]);
4078 				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
4079 			}
4080 		}
4081 
4082 		/* no further action needed, so return now */
4083 		return;
4084 	} else {
4085 		dev_info(&pf->pdev->dev,
4086 			 "bad reset request 0x%08x\n", reset_flags);
4087 		return;
4088 	}
4089 }
4090 
4091 /**
4092  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
4093  * @pf: board private structure
4094  * @e: event info posted on ARQ
4095  *
4096  * Handler for LAN Queue Overflow Event generated by the firmware for PF
4097  * and VF queues
4098  **/
4099 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
4100 					   struct i40e_arq_event_info *e)
4101 {
4102 	struct i40e_aqc_lan_overflow *data =
4103 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
4104 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
4105 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
4106 	struct i40e_hw *hw = &pf->hw;
4107 	struct i40e_vf *vf;
4108 	u16 vf_id;
4109 
4110 	dev_info(&pf->pdev->dev, "%s: Rx Queue Number = %d QTX_CTL=0x%08x\n",
4111 		 __func__, queue, qtx_ctl);
4112 
4113 	/* Queue belongs to VF, find the VF and issue VF reset */
4114 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
4115 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
4116 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
4117 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
4118 		vf_id -= hw->func_caps.vf_base_id;
4119 		vf = &pf->vf[vf_id];
4120 		i40e_vc_notify_vf_reset(vf);
4121 		/* Allow VF to process pending reset notification */
4122 		msleep(20);
4123 		i40e_reset_vf(vf, false);
4124 	}
4125 }
4126 
4127 /**
4128  * i40e_service_event_complete - Finish up the service event
4129  * @pf: board private structure
4130  **/
4131 static void i40e_service_event_complete(struct i40e_pf *pf)
4132 {
4133 	BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
4134 
4135 	/* flush memory to make sure state is correct before next watchog */
4136 	smp_mb__before_clear_bit();
4137 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
4138 }
4139 
4140 /**
4141  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
4142  * @pf: board private structure
4143  **/
4144 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
4145 {
4146 	if (!(pf->flags & I40E_FLAG_FDIR_REQUIRES_REINIT))
4147 		return;
4148 
4149 	pf->flags &= ~I40E_FLAG_FDIR_REQUIRES_REINIT;
4150 
4151 	/* if interface is down do nothing */
4152 	if (test_bit(__I40E_DOWN, &pf->state))
4153 		return;
4154 }
4155 
4156 /**
4157  * i40e_vsi_link_event - notify VSI of a link event
4158  * @vsi: vsi to be notified
4159  * @link_up: link up or down
4160  **/
4161 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
4162 {
4163 	if (!vsi)
4164 		return;
4165 
4166 	switch (vsi->type) {
4167 	case I40E_VSI_MAIN:
4168 		if (!vsi->netdev || !vsi->netdev_registered)
4169 			break;
4170 
4171 		if (link_up) {
4172 			netif_carrier_on(vsi->netdev);
4173 			netif_tx_wake_all_queues(vsi->netdev);
4174 		} else {
4175 			netif_carrier_off(vsi->netdev);
4176 			netif_tx_stop_all_queues(vsi->netdev);
4177 		}
4178 		break;
4179 
4180 	case I40E_VSI_SRIOV:
4181 		break;
4182 
4183 	case I40E_VSI_VMDQ2:
4184 	case I40E_VSI_CTRL:
4185 	case I40E_VSI_MIRROR:
4186 	default:
4187 		/* there is no notification for other VSIs */
4188 		break;
4189 	}
4190 }
4191 
4192 /**
4193  * i40e_veb_link_event - notify elements on the veb of a link event
4194  * @veb: veb to be notified
4195  * @link_up: link up or down
4196  **/
4197 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
4198 {
4199 	struct i40e_pf *pf;
4200 	int i;
4201 
4202 	if (!veb || !veb->pf)
4203 		return;
4204 	pf = veb->pf;
4205 
4206 	/* depth first... */
4207 	for (i = 0; i < I40E_MAX_VEB; i++)
4208 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
4209 			i40e_veb_link_event(pf->veb[i], link_up);
4210 
4211 	/* ... now the local VSIs */
4212 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4213 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
4214 			i40e_vsi_link_event(pf->vsi[i], link_up);
4215 }
4216 
4217 /**
4218  * i40e_link_event - Update netif_carrier status
4219  * @pf: board private structure
4220  **/
4221 static void i40e_link_event(struct i40e_pf *pf)
4222 {
4223 	bool new_link, old_link;
4224 
4225 	new_link = (pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP);
4226 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
4227 
4228 	if (new_link == old_link)
4229 		return;
4230 
4231 	if (!test_bit(__I40E_DOWN, &pf->vsi[pf->lan_vsi]->state))
4232 		netdev_info(pf->vsi[pf->lan_vsi]->netdev,
4233 			    "NIC Link is %s\n", (new_link ? "Up" : "Down"));
4234 
4235 	/* Notify the base of the switch tree connected to
4236 	 * the link.  Floating VEBs are not notified.
4237 	 */
4238 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
4239 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
4240 	else
4241 		i40e_vsi_link_event(pf->vsi[pf->lan_vsi], new_link);
4242 
4243 	if (pf->vf)
4244 		i40e_vc_notify_link_state(pf);
4245 }
4246 
4247 /**
4248  * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
4249  * @pf: board private structure
4250  *
4251  * Set the per-queue flags to request a check for stuck queues in the irq
4252  * clean functions, then force interrupts to be sure the irq clean is called.
4253  **/
4254 static void i40e_check_hang_subtask(struct i40e_pf *pf)
4255 {
4256 	int i, v;
4257 
4258 	/* If we're down or resetting, just bail */
4259 	if (test_bit(__I40E_CONFIG_BUSY, &pf->state))
4260 		return;
4261 
4262 	/* for each VSI/netdev
4263 	 *     for each Tx queue
4264 	 *         set the check flag
4265 	 *     for each q_vector
4266 	 *         force an interrupt
4267 	 */
4268 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4269 		struct i40e_vsi *vsi = pf->vsi[v];
4270 		int armed = 0;
4271 
4272 		if (!pf->vsi[v] ||
4273 		    test_bit(__I40E_DOWN, &vsi->state) ||
4274 		    (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
4275 			continue;
4276 
4277 		for (i = 0; i < vsi->num_queue_pairs; i++) {
4278 			set_check_for_tx_hang(vsi->tx_rings[i]);
4279 			if (test_bit(__I40E_HANG_CHECK_ARMED,
4280 				     &vsi->tx_rings[i]->state))
4281 				armed++;
4282 		}
4283 
4284 		if (armed) {
4285 			if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
4286 				wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
4287 				     (I40E_PFINT_DYN_CTL0_INTENA_MASK |
4288 				      I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK));
4289 			} else {
4290 				u16 vec = vsi->base_vector - 1;
4291 				u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
4292 					   I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK);
4293 				for (i = 0; i < vsi->num_q_vectors; i++, vec++)
4294 					wr32(&vsi->back->hw,
4295 					     I40E_PFINT_DYN_CTLN(vec), val);
4296 			}
4297 			i40e_flush(&vsi->back->hw);
4298 		}
4299 	}
4300 }
4301 
4302 /**
4303  * i40e_watchdog_subtask - Check and bring link up
4304  * @pf: board private structure
4305  **/
4306 static void i40e_watchdog_subtask(struct i40e_pf *pf)
4307 {
4308 	int i;
4309 
4310 	/* if interface is down do nothing */
4311 	if (test_bit(__I40E_DOWN, &pf->state) ||
4312 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
4313 		return;
4314 
4315 	/* Update the stats for active netdevs so the network stack
4316 	 * can look at updated numbers whenever it cares to
4317 	 */
4318 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4319 		if (pf->vsi[i] && pf->vsi[i]->netdev)
4320 			i40e_update_stats(pf->vsi[i]);
4321 
4322 	/* Update the stats for the active switching components */
4323 	for (i = 0; i < I40E_MAX_VEB; i++)
4324 		if (pf->veb[i])
4325 			i40e_update_veb_stats(pf->veb[i]);
4326 }
4327 
4328 /**
4329  * i40e_reset_subtask - Set up for resetting the device and driver
4330  * @pf: board private structure
4331  **/
4332 static void i40e_reset_subtask(struct i40e_pf *pf)
4333 {
4334 	u32 reset_flags = 0;
4335 
4336 	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
4337 		reset_flags |= (1 << __I40E_REINIT_REQUESTED);
4338 		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
4339 	}
4340 	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
4341 		reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
4342 		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4343 	}
4344 	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
4345 		reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
4346 		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
4347 	}
4348 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
4349 		reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
4350 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
4351 	}
4352 
4353 	/* If there's a recovery already waiting, it takes
4354 	 * precedence before starting a new reset sequence.
4355 	 */
4356 	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
4357 		i40e_handle_reset_warning(pf);
4358 		return;
4359 	}
4360 
4361 	/* If we're already down or resetting, just bail */
4362 	if (reset_flags &&
4363 	    !test_bit(__I40E_DOWN, &pf->state) &&
4364 	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
4365 		i40e_do_reset(pf, reset_flags);
4366 }
4367 
4368 /**
4369  * i40e_handle_link_event - Handle link event
4370  * @pf: board private structure
4371  * @e: event info posted on ARQ
4372  **/
4373 static void i40e_handle_link_event(struct i40e_pf *pf,
4374 				   struct i40e_arq_event_info *e)
4375 {
4376 	struct i40e_hw *hw = &pf->hw;
4377 	struct i40e_aqc_get_link_status *status =
4378 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
4379 	struct i40e_link_status *hw_link_info = &hw->phy.link_info;
4380 
4381 	/* save off old link status information */
4382 	memcpy(&pf->hw.phy.link_info_old, hw_link_info,
4383 	       sizeof(pf->hw.phy.link_info_old));
4384 
4385 	/* update link status */
4386 	hw_link_info->phy_type = (enum i40e_aq_phy_type)status->phy_type;
4387 	hw_link_info->link_speed = (enum i40e_aq_link_speed)status->link_speed;
4388 	hw_link_info->link_info = status->link_info;
4389 	hw_link_info->an_info = status->an_info;
4390 	hw_link_info->ext_info = status->ext_info;
4391 	hw_link_info->lse_enable =
4392 		le16_to_cpu(status->command_flags) &
4393 			    I40E_AQ_LSE_ENABLE;
4394 
4395 	/* process the event */
4396 	i40e_link_event(pf);
4397 
4398 	/* Do a new status request to re-enable LSE reporting
4399 	 * and load new status information into the hw struct,
4400 	 * then see if the status changed while processing the
4401 	 * initial event.
4402 	 */
4403 	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
4404 	i40e_link_event(pf);
4405 }
4406 
4407 /**
4408  * i40e_clean_adminq_subtask - Clean the AdminQ rings
4409  * @pf: board private structure
4410  **/
4411 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
4412 {
4413 	struct i40e_arq_event_info event;
4414 	struct i40e_hw *hw = &pf->hw;
4415 	u16 pending, i = 0;
4416 	i40e_status ret;
4417 	u16 opcode;
4418 	u32 val;
4419 
4420 	if (!test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state))
4421 		return;
4422 
4423 	event.msg_size = I40E_MAX_AQ_BUF_SIZE;
4424 	event.msg_buf = kzalloc(event.msg_size, GFP_KERNEL);
4425 	if (!event.msg_buf)
4426 		return;
4427 
4428 	do {
4429 		ret = i40e_clean_arq_element(hw, &event, &pending);
4430 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK) {
4431 			dev_info(&pf->pdev->dev, "No ARQ event found\n");
4432 			break;
4433 		} else if (ret) {
4434 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
4435 			break;
4436 		}
4437 
4438 		opcode = le16_to_cpu(event.desc.opcode);
4439 		switch (opcode) {
4440 
4441 		case i40e_aqc_opc_get_link_status:
4442 			i40e_handle_link_event(pf, &event);
4443 			break;
4444 		case i40e_aqc_opc_send_msg_to_pf:
4445 			ret = i40e_vc_process_vf_msg(pf,
4446 					le16_to_cpu(event.desc.retval),
4447 					le32_to_cpu(event.desc.cookie_high),
4448 					le32_to_cpu(event.desc.cookie_low),
4449 					event.msg_buf,
4450 					event.msg_size);
4451 			break;
4452 		case i40e_aqc_opc_lldp_update_mib:
4453 			dev_info(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
4454 			break;
4455 		case i40e_aqc_opc_event_lan_overflow:
4456 			dev_info(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
4457 			i40e_handle_lan_overflow_event(pf, &event);
4458 			break;
4459 		default:
4460 			dev_info(&pf->pdev->dev,
4461 				 "ARQ Error: Unknown event %d received\n",
4462 				 event.desc.opcode);
4463 			break;
4464 		}
4465 	} while (pending && (i++ < pf->adminq_work_limit));
4466 
4467 	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
4468 	/* re-enable Admin queue interrupt cause */
4469 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
4470 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
4471 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
4472 	i40e_flush(hw);
4473 
4474 	kfree(event.msg_buf);
4475 }
4476 
4477 /**
4478  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
4479  * @veb: pointer to the VEB instance
4480  *
4481  * This is a recursive function that first builds the attached VSIs then
4482  * recurses in to build the next layer of VEB.  We track the connections
4483  * through our own index numbers because the seid's from the HW could
4484  * change across the reset.
4485  **/
4486 static int i40e_reconstitute_veb(struct i40e_veb *veb)
4487 {
4488 	struct i40e_vsi *ctl_vsi = NULL;
4489 	struct i40e_pf *pf = veb->pf;
4490 	int v, veb_idx;
4491 	int ret;
4492 
4493 	/* build VSI that owns this VEB, temporarily attached to base VEB */
4494 	for (v = 0; v < pf->hw.func_caps.num_vsis && !ctl_vsi; v++) {
4495 		if (pf->vsi[v] &&
4496 		    pf->vsi[v]->veb_idx == veb->idx &&
4497 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
4498 			ctl_vsi = pf->vsi[v];
4499 			break;
4500 		}
4501 	}
4502 	if (!ctl_vsi) {
4503 		dev_info(&pf->pdev->dev,
4504 			 "missing owner VSI for veb_idx %d\n", veb->idx);
4505 		ret = -ENOENT;
4506 		goto end_reconstitute;
4507 	}
4508 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
4509 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
4510 	ret = i40e_add_vsi(ctl_vsi);
4511 	if (ret) {
4512 		dev_info(&pf->pdev->dev,
4513 			 "rebuild of owner VSI failed: %d\n", ret);
4514 		goto end_reconstitute;
4515 	}
4516 	i40e_vsi_reset_stats(ctl_vsi);
4517 
4518 	/* create the VEB in the switch and move the VSI onto the VEB */
4519 	ret = i40e_add_veb(veb, ctl_vsi);
4520 	if (ret)
4521 		goto end_reconstitute;
4522 
4523 	/* create the remaining VSIs attached to this VEB */
4524 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4525 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
4526 			continue;
4527 
4528 		if (pf->vsi[v]->veb_idx == veb->idx) {
4529 			struct i40e_vsi *vsi = pf->vsi[v];
4530 			vsi->uplink_seid = veb->seid;
4531 			ret = i40e_add_vsi(vsi);
4532 			if (ret) {
4533 				dev_info(&pf->pdev->dev,
4534 					 "rebuild of vsi_idx %d failed: %d\n",
4535 					 v, ret);
4536 				goto end_reconstitute;
4537 			}
4538 			i40e_vsi_reset_stats(vsi);
4539 		}
4540 	}
4541 
4542 	/* create any VEBs attached to this VEB - RECURSION */
4543 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
4544 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
4545 			pf->veb[veb_idx]->uplink_seid = veb->seid;
4546 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
4547 			if (ret)
4548 				break;
4549 		}
4550 	}
4551 
4552 end_reconstitute:
4553 	return ret;
4554 }
4555 
4556 /**
4557  * i40e_get_capabilities - get info about the HW
4558  * @pf: the PF struct
4559  **/
4560 static int i40e_get_capabilities(struct i40e_pf *pf)
4561 {
4562 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
4563 	u16 data_size;
4564 	int buf_len;
4565 	int err;
4566 
4567 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
4568 	do {
4569 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
4570 		if (!cap_buf)
4571 			return -ENOMEM;
4572 
4573 		/* this loads the data into the hw struct for us */
4574 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
4575 					    &data_size,
4576 					    i40e_aqc_opc_list_func_capabilities,
4577 					    NULL);
4578 		/* data loaded, buffer no longer needed */
4579 		kfree(cap_buf);
4580 
4581 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
4582 			/* retry with a larger buffer */
4583 			buf_len = data_size;
4584 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
4585 			dev_info(&pf->pdev->dev,
4586 				 "capability discovery failed: aq=%d\n",
4587 				 pf->hw.aq.asq_last_status);
4588 			return -ENODEV;
4589 		}
4590 	} while (err);
4591 
4592 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
4593 		dev_info(&pf->pdev->dev,
4594 			 "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",
4595 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
4596 			 pf->hw.func_caps.num_msix_vectors,
4597 			 pf->hw.func_caps.num_msix_vectors_vf,
4598 			 pf->hw.func_caps.fd_filters_guaranteed,
4599 			 pf->hw.func_caps.fd_filters_best_effort,
4600 			 pf->hw.func_caps.num_tx_qp,
4601 			 pf->hw.func_caps.num_vsis);
4602 
4603 	return 0;
4604 }
4605 
4606 /**
4607  * i40e_fdir_setup - initialize the Flow Director resources
4608  * @pf: board private structure
4609  **/
4610 static void i40e_fdir_setup(struct i40e_pf *pf)
4611 {
4612 	struct i40e_vsi *vsi;
4613 	bool new_vsi = false;
4614 	int err, i;
4615 
4616 	if (!(pf->flags & (I40E_FLAG_FDIR_ENABLED |
4617 			   I40E_FLAG_FDIR_ATR_ENABLED)))
4618 		return;
4619 
4620 	pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
4621 
4622 	/* find existing or make new FDIR VSI */
4623 	vsi = NULL;
4624 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4625 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR)
4626 			vsi = pf->vsi[i];
4627 	if (!vsi) {
4628 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR, pf->mac_seid, 0);
4629 		if (!vsi) {
4630 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
4631 			pf->flags &= ~I40E_FLAG_FDIR_ENABLED;
4632 			return;
4633 		}
4634 		new_vsi = true;
4635 	}
4636 	WARN_ON(vsi->base_queue != I40E_FDIR_RING);
4637 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_rings);
4638 
4639 	err = i40e_vsi_setup_tx_resources(vsi);
4640 	if (!err)
4641 		err = i40e_vsi_setup_rx_resources(vsi);
4642 	if (!err)
4643 		err = i40e_vsi_configure(vsi);
4644 	if (!err && new_vsi) {
4645 		char int_name[IFNAMSIZ + 9];
4646 		snprintf(int_name, sizeof(int_name) - 1, "%s-fdir",
4647 			 dev_driver_string(&pf->pdev->dev));
4648 		err = i40e_vsi_request_irq(vsi, int_name);
4649 	}
4650 	if (!err)
4651 		err = i40e_up_complete(vsi);
4652 
4653 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4654 }
4655 
4656 /**
4657  * i40e_fdir_teardown - release the Flow Director resources
4658  * @pf: board private structure
4659  **/
4660 static void i40e_fdir_teardown(struct i40e_pf *pf)
4661 {
4662 	int i;
4663 
4664 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
4665 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
4666 			i40e_vsi_release(pf->vsi[i]);
4667 			break;
4668 		}
4669 	}
4670 }
4671 
4672 /**
4673  * i40e_handle_reset_warning - prep for the core to reset
4674  * @pf: board private structure
4675  *
4676  * Close up the VFs and other things in prep for a Core Reset,
4677  * then get ready to rebuild the world.
4678  **/
4679 static void i40e_handle_reset_warning(struct i40e_pf *pf)
4680 {
4681 	struct i40e_driver_version dv;
4682 	struct i40e_hw *hw = &pf->hw;
4683 	i40e_status ret;
4684 	u32 v;
4685 
4686 	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
4687 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
4688 		return;
4689 
4690 	dev_info(&pf->pdev->dev, "Tearing down internal switch for reset\n");
4691 
4692 	i40e_vc_notify_reset(pf);
4693 
4694 	/* quiesce the VSIs and their queues that are not already DOWN */
4695 	i40e_pf_quiesce_all_vsi(pf);
4696 
4697 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4698 		if (pf->vsi[v])
4699 			pf->vsi[v]->seid = 0;
4700 	}
4701 
4702 	i40e_shutdown_adminq(&pf->hw);
4703 
4704 	/* Now we wait for GRST to settle out.
4705 	 * We don't have to delete the VEBs or VSIs from the hw switch
4706 	 * because the reset will make them disappear.
4707 	 */
4708 	ret = i40e_pf_reset(hw);
4709 	if (ret)
4710 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
4711 	pf->pfr_count++;
4712 
4713 	if (test_bit(__I40E_DOWN, &pf->state))
4714 		goto end_core_reset;
4715 	dev_info(&pf->pdev->dev, "Rebuilding internal switch\n");
4716 
4717 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
4718 	ret = i40e_init_adminq(&pf->hw);
4719 	if (ret) {
4720 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
4721 		goto end_core_reset;
4722 	}
4723 
4724 	ret = i40e_get_capabilities(pf);
4725 	if (ret) {
4726 		dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
4727 			 ret);
4728 		goto end_core_reset;
4729 	}
4730 
4731 	/* call shutdown HMC */
4732 	ret = i40e_shutdown_lan_hmc(hw);
4733 	if (ret) {
4734 		dev_info(&pf->pdev->dev, "shutdown_lan_hmc failed: %d\n", ret);
4735 		goto end_core_reset;
4736 	}
4737 
4738 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
4739 				hw->func_caps.num_rx_qp,
4740 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
4741 	if (ret) {
4742 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
4743 		goto end_core_reset;
4744 	}
4745 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
4746 	if (ret) {
4747 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
4748 		goto end_core_reset;
4749 	}
4750 
4751 	/* do basic switch setup */
4752 	ret = i40e_setup_pf_switch(pf);
4753 	if (ret)
4754 		goto end_core_reset;
4755 
4756 	/* Rebuild the VSIs and VEBs that existed before reset.
4757 	 * They are still in our local switch element arrays, so only
4758 	 * need to rebuild the switch model in the HW.
4759 	 *
4760 	 * If there were VEBs but the reconstitution failed, we'll try
4761 	 * try to recover minimal use by getting the basic PF VSI working.
4762 	 */
4763 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
4764 		dev_info(&pf->pdev->dev, "attempting to rebuild switch\n");
4765 		/* find the one VEB connected to the MAC, and find orphans */
4766 		for (v = 0; v < I40E_MAX_VEB; v++) {
4767 			if (!pf->veb[v])
4768 				continue;
4769 
4770 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
4771 			    pf->veb[v]->uplink_seid == 0) {
4772 				ret = i40e_reconstitute_veb(pf->veb[v]);
4773 
4774 				if (!ret)
4775 					continue;
4776 
4777 				/* If Main VEB failed, we're in deep doodoo,
4778 				 * so give up rebuilding the switch and set up
4779 				 * for minimal rebuild of PF VSI.
4780 				 * If orphan failed, we'll report the error
4781 				 * but try to keep going.
4782 				 */
4783 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
4784 					dev_info(&pf->pdev->dev,
4785 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
4786 						 ret);
4787 					pf->vsi[pf->lan_vsi]->uplink_seid
4788 								= pf->mac_seid;
4789 					break;
4790 				} else if (pf->veb[v]->uplink_seid == 0) {
4791 					dev_info(&pf->pdev->dev,
4792 						 "rebuild of orphan VEB failed: %d\n",
4793 						 ret);
4794 				}
4795 			}
4796 		}
4797 	}
4798 
4799 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
4800 		dev_info(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
4801 		/* no VEB, so rebuild only the Main VSI */
4802 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
4803 		if (ret) {
4804 			dev_info(&pf->pdev->dev,
4805 				 "rebuild of Main VSI failed: %d\n", ret);
4806 			goto end_core_reset;
4807 		}
4808 	}
4809 
4810 	/* reinit the misc interrupt */
4811 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4812 		ret = i40e_setup_misc_vector(pf);
4813 
4814 	/* restart the VSIs that were rebuilt and running before the reset */
4815 	i40e_pf_unquiesce_all_vsi(pf);
4816 
4817 	/* tell the firmware that we're starting */
4818 	dv.major_version = DRV_VERSION_MAJOR;
4819 	dv.minor_version = DRV_VERSION_MINOR;
4820 	dv.build_version = DRV_VERSION_BUILD;
4821 	dv.subbuild_version = 0;
4822 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
4823 
4824 	dev_info(&pf->pdev->dev, "PF reset done\n");
4825 
4826 end_core_reset:
4827 	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
4828 }
4829 
4830 /**
4831  * i40e_handle_mdd_event
4832  * @pf: pointer to the pf structure
4833  *
4834  * Called from the MDD irq handler to identify possibly malicious vfs
4835  **/
4836 static void i40e_handle_mdd_event(struct i40e_pf *pf)
4837 {
4838 	struct i40e_hw *hw = &pf->hw;
4839 	bool mdd_detected = false;
4840 	struct i40e_vf *vf;
4841 	u32 reg;
4842 	int i;
4843 
4844 	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
4845 		return;
4846 
4847 	/* find what triggered the MDD event */
4848 	reg = rd32(hw, I40E_GL_MDET_TX);
4849 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
4850 		u8 func = (reg & I40E_GL_MDET_TX_FUNCTION_MASK)
4851 				>> I40E_GL_MDET_TX_FUNCTION_SHIFT;
4852 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_SHIFT)
4853 				>> I40E_GL_MDET_TX_EVENT_SHIFT;
4854 		u8 queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK)
4855 				>> I40E_GL_MDET_TX_QUEUE_SHIFT;
4856 		dev_info(&pf->pdev->dev,
4857 			 "Malicious Driver Detection TX event 0x%02x on q %d of function 0x%02x\n",
4858 			 event, queue, func);
4859 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
4860 		mdd_detected = true;
4861 	}
4862 	reg = rd32(hw, I40E_GL_MDET_RX);
4863 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
4864 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK)
4865 				>> I40E_GL_MDET_RX_FUNCTION_SHIFT;
4866 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_SHIFT)
4867 				>> I40E_GL_MDET_RX_EVENT_SHIFT;
4868 		u8 queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK)
4869 				>> I40E_GL_MDET_RX_QUEUE_SHIFT;
4870 		dev_info(&pf->pdev->dev,
4871 			 "Malicious Driver Detection RX event 0x%02x on q %d of function 0x%02x\n",
4872 			 event, queue, func);
4873 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
4874 		mdd_detected = true;
4875 	}
4876 
4877 	/* see if one of the VFs needs its hand slapped */
4878 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
4879 		vf = &(pf->vf[i]);
4880 		reg = rd32(hw, I40E_VP_MDET_TX(i));
4881 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
4882 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
4883 			vf->num_mdd_events++;
4884 			dev_info(&pf->pdev->dev, "MDD TX event on VF %d\n", i);
4885 		}
4886 
4887 		reg = rd32(hw, I40E_VP_MDET_RX(i));
4888 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
4889 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
4890 			vf->num_mdd_events++;
4891 			dev_info(&pf->pdev->dev, "MDD RX event on VF %d\n", i);
4892 		}
4893 
4894 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
4895 			dev_info(&pf->pdev->dev,
4896 				 "Too many MDD events on VF %d, disabled\n", i);
4897 			dev_info(&pf->pdev->dev,
4898 				 "Use PF Control I/F to re-enable the VF\n");
4899 			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
4900 		}
4901 	}
4902 
4903 	/* re-enable mdd interrupt cause */
4904 	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
4905 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
4906 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
4907 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
4908 	i40e_flush(hw);
4909 }
4910 
4911 /**
4912  * i40e_service_task - Run the driver's async subtasks
4913  * @work: pointer to work_struct containing our data
4914  **/
4915 static void i40e_service_task(struct work_struct *work)
4916 {
4917 	struct i40e_pf *pf = container_of(work,
4918 					  struct i40e_pf,
4919 					  service_task);
4920 	unsigned long start_time = jiffies;
4921 
4922 	i40e_reset_subtask(pf);
4923 	i40e_handle_mdd_event(pf);
4924 	i40e_vc_process_vflr_event(pf);
4925 	i40e_watchdog_subtask(pf);
4926 	i40e_fdir_reinit_subtask(pf);
4927 	i40e_check_hang_subtask(pf);
4928 	i40e_sync_filters_subtask(pf);
4929 	i40e_clean_adminq_subtask(pf);
4930 
4931 	i40e_service_event_complete(pf);
4932 
4933 	/* If the tasks have taken longer than one timer cycle or there
4934 	 * is more work to be done, reschedule the service task now
4935 	 * rather than wait for the timer to tick again.
4936 	 */
4937 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
4938 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
4939 	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
4940 	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
4941 		i40e_service_event_schedule(pf);
4942 }
4943 
4944 /**
4945  * i40e_service_timer - timer callback
4946  * @data: pointer to PF struct
4947  **/
4948 static void i40e_service_timer(unsigned long data)
4949 {
4950 	struct i40e_pf *pf = (struct i40e_pf *)data;
4951 
4952 	mod_timer(&pf->service_timer,
4953 		  round_jiffies(jiffies + pf->service_timer_period));
4954 	i40e_service_event_schedule(pf);
4955 }
4956 
4957 /**
4958  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
4959  * @vsi: the VSI being configured
4960  **/
4961 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
4962 {
4963 	struct i40e_pf *pf = vsi->back;
4964 
4965 	switch (vsi->type) {
4966 	case I40E_VSI_MAIN:
4967 		vsi->alloc_queue_pairs = pf->num_lan_qps;
4968 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4969 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4970 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4971 			vsi->num_q_vectors = pf->num_lan_msix;
4972 		else
4973 			vsi->num_q_vectors = 1;
4974 
4975 		break;
4976 
4977 	case I40E_VSI_FDIR:
4978 		vsi->alloc_queue_pairs = 1;
4979 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
4980 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4981 		vsi->num_q_vectors = 1;
4982 		break;
4983 
4984 	case I40E_VSI_VMDQ2:
4985 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
4986 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4987 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4988 		vsi->num_q_vectors = pf->num_vmdq_msix;
4989 		break;
4990 
4991 	case I40E_VSI_SRIOV:
4992 		vsi->alloc_queue_pairs = pf->num_vf_qps;
4993 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4994 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4995 		break;
4996 
4997 	default:
4998 		WARN_ON(1);
4999 		return -ENODATA;
5000 	}
5001 
5002 	return 0;
5003 }
5004 
5005 /**
5006  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
5007  * @pf: board private structure
5008  * @type: type of VSI
5009  *
5010  * On error: returns error code (negative)
5011  * On success: returns vsi index in PF (positive)
5012  **/
5013 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
5014 {
5015 	int ret = -ENODEV;
5016 	struct i40e_vsi *vsi;
5017 	int sz_vectors;
5018 	int sz_rings;
5019 	int vsi_idx;
5020 	int i;
5021 
5022 	/* Need to protect the allocation of the VSIs at the PF level */
5023 	mutex_lock(&pf->switch_mutex);
5024 
5025 	/* VSI list may be fragmented if VSI creation/destruction has
5026 	 * been happening.  We can afford to do a quick scan to look
5027 	 * for any free VSIs in the list.
5028 	 *
5029 	 * find next empty vsi slot, looping back around if necessary
5030 	 */
5031 	i = pf->next_vsi;
5032 	while (i < pf->hw.func_caps.num_vsis && pf->vsi[i])
5033 		i++;
5034 	if (i >= pf->hw.func_caps.num_vsis) {
5035 		i = 0;
5036 		while (i < pf->next_vsi && pf->vsi[i])
5037 			i++;
5038 	}
5039 
5040 	if (i < pf->hw.func_caps.num_vsis && !pf->vsi[i]) {
5041 		vsi_idx = i;             /* Found one! */
5042 	} else {
5043 		ret = -ENODEV;
5044 		goto unlock_pf;  /* out of VSI slots! */
5045 	}
5046 	pf->next_vsi = ++i;
5047 
5048 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
5049 	if (!vsi) {
5050 		ret = -ENOMEM;
5051 		goto unlock_pf;
5052 	}
5053 	vsi->type = type;
5054 	vsi->back = pf;
5055 	set_bit(__I40E_DOWN, &vsi->state);
5056 	vsi->flags = 0;
5057 	vsi->idx = vsi_idx;
5058 	vsi->rx_itr_setting = pf->rx_itr_default;
5059 	vsi->tx_itr_setting = pf->tx_itr_default;
5060 	vsi->netdev_registered = false;
5061 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
5062 	INIT_LIST_HEAD(&vsi->mac_filter_list);
5063 
5064 	ret = i40e_set_num_rings_in_vsi(vsi);
5065 	if (ret)
5066 		goto err_rings;
5067 
5068 	/* allocate memory for ring pointers */
5069 	sz_rings = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
5070 	vsi->tx_rings = kzalloc(sz_rings, GFP_KERNEL);
5071 	if (!vsi->tx_rings) {
5072 		ret = -ENOMEM;
5073 		goto err_rings;
5074 	}
5075 	vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
5076 
5077 	/* allocate memory for q_vector pointers */
5078 	sz_vectors = sizeof(struct i40e_q_vectors *) * vsi->num_q_vectors;
5079 	vsi->q_vectors = kzalloc(sz_vectors, GFP_KERNEL);
5080 	if (!vsi->q_vectors) {
5081 		ret = -ENOMEM;
5082 		goto err_vectors;
5083 	}
5084 
5085 	/* Setup default MSIX irq handler for VSI */
5086 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
5087 
5088 	pf->vsi[vsi_idx] = vsi;
5089 	ret = vsi_idx;
5090 	goto unlock_pf;
5091 
5092 err_vectors:
5093  	kfree(vsi->tx_rings);
5094 err_rings:
5095 	pf->next_vsi = i - 1;
5096 	kfree(vsi);
5097 unlock_pf:
5098 	mutex_unlock(&pf->switch_mutex);
5099 	return ret;
5100 }
5101 
5102 /**
5103  * i40e_vsi_clear - Deallocate the VSI provided
5104  * @vsi: the VSI being un-configured
5105  **/
5106 static int i40e_vsi_clear(struct i40e_vsi *vsi)
5107 {
5108 	struct i40e_pf *pf;
5109 
5110 	if (!vsi)
5111 		return 0;
5112 
5113 	if (!vsi->back)
5114 		goto free_vsi;
5115 	pf = vsi->back;
5116 
5117 	mutex_lock(&pf->switch_mutex);
5118 	if (!pf->vsi[vsi->idx]) {
5119 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
5120 			vsi->idx, vsi->idx, vsi, vsi->type);
5121 		goto unlock_vsi;
5122 	}
5123 
5124 	if (pf->vsi[vsi->idx] != vsi) {
5125 		dev_err(&pf->pdev->dev,
5126 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
5127 			pf->vsi[vsi->idx]->idx,
5128 			pf->vsi[vsi->idx],
5129 			pf->vsi[vsi->idx]->type,
5130 			vsi->idx, vsi, vsi->type);
5131 		goto unlock_vsi;
5132 	}
5133 
5134 	/* updates the pf for this cleared vsi */
5135 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
5136 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
5137 
5138 	/* free the ring and vector containers */
5139 	kfree(vsi->q_vectors);
5140 	kfree(vsi->tx_rings);
5141 
5142 	pf->vsi[vsi->idx] = NULL;
5143 	if (vsi->idx < pf->next_vsi)
5144 		pf->next_vsi = vsi->idx;
5145 
5146 unlock_vsi:
5147 	mutex_unlock(&pf->switch_mutex);
5148 free_vsi:
5149 	kfree(vsi);
5150 
5151 	return 0;
5152 }
5153 
5154 /**
5155  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
5156  * @vsi: the VSI being cleaned
5157  **/
5158 static s32 i40e_vsi_clear_rings(struct i40e_vsi *vsi)
5159 {
5160 	int i;
5161 
5162 	if (vsi->tx_rings[0])
5163 		for (i = 0; i < vsi->alloc_queue_pairs; i++) {
5164 			kfree_rcu(vsi->tx_rings[i], rcu);
5165 			vsi->tx_rings[i] = NULL;
5166 			vsi->rx_rings[i] = NULL;
5167 		}
5168 
5169 	return 0;
5170 }
5171 
5172 /**
5173  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
5174  * @vsi: the VSI being configured
5175  **/
5176 static int i40e_alloc_rings(struct i40e_vsi *vsi)
5177 {
5178 	struct i40e_pf *pf = vsi->back;
5179 	int i;
5180 
5181 	/* Set basic values in the rings to be used later during open() */
5182 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
5183 		struct i40e_ring *tx_ring;
5184 		struct i40e_ring *rx_ring;
5185 
5186 		tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
5187 		if (!tx_ring)
5188 			goto err_out;
5189 
5190 		tx_ring->queue_index = i;
5191 		tx_ring->reg_idx = vsi->base_queue + i;
5192 		tx_ring->ring_active = false;
5193 		tx_ring->vsi = vsi;
5194 		tx_ring->netdev = vsi->netdev;
5195 		tx_ring->dev = &pf->pdev->dev;
5196 		tx_ring->count = vsi->num_desc;
5197 		tx_ring->size = 0;
5198 		tx_ring->dcb_tc = 0;
5199 		vsi->tx_rings[i] = tx_ring;
5200 
5201 		rx_ring = &tx_ring[1];
5202 		rx_ring->queue_index = i;
5203 		rx_ring->reg_idx = vsi->base_queue + i;
5204 		rx_ring->ring_active = false;
5205 		rx_ring->vsi = vsi;
5206 		rx_ring->netdev = vsi->netdev;
5207 		rx_ring->dev = &pf->pdev->dev;
5208 		rx_ring->count = vsi->num_desc;
5209 		rx_ring->size = 0;
5210 		rx_ring->dcb_tc = 0;
5211 		if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
5212 			set_ring_16byte_desc_enabled(rx_ring);
5213 		else
5214 			clear_ring_16byte_desc_enabled(rx_ring);
5215 		vsi->rx_rings[i] = rx_ring;
5216 	}
5217 
5218 	return 0;
5219 
5220 err_out:
5221 	i40e_vsi_clear_rings(vsi);
5222 	return -ENOMEM;
5223 }
5224 
5225 /**
5226  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
5227  * @pf: board private structure
5228  * @vectors: the number of MSI-X vectors to request
5229  *
5230  * Returns the number of vectors reserved, or error
5231  **/
5232 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
5233 {
5234 	int err = 0;
5235 
5236 	pf->num_msix_entries = 0;
5237 	while (vectors >= I40E_MIN_MSIX) {
5238 		err = pci_enable_msix(pf->pdev, pf->msix_entries, vectors);
5239 		if (err == 0) {
5240 			/* good to go */
5241 			pf->num_msix_entries = vectors;
5242 			break;
5243 		} else if (err < 0) {
5244 			/* total failure */
5245 			dev_info(&pf->pdev->dev,
5246 				 "MSI-X vector reservation failed: %d\n", err);
5247 			vectors = 0;
5248 			break;
5249 		} else {
5250 			/* err > 0 is the hint for retry */
5251 			dev_info(&pf->pdev->dev,
5252 				 "MSI-X vectors wanted %d, retrying with %d\n",
5253 				 vectors, err);
5254 			vectors = err;
5255 		}
5256 	}
5257 
5258 	if (vectors > 0 && vectors < I40E_MIN_MSIX) {
5259 		dev_info(&pf->pdev->dev,
5260 			 "Couldn't get enough vectors, only %d available\n",
5261 			 vectors);
5262 		vectors = 0;
5263 	}
5264 
5265 	return vectors;
5266 }
5267 
5268 /**
5269  * i40e_init_msix - Setup the MSIX capability
5270  * @pf: board private structure
5271  *
5272  * Work with the OS to set up the MSIX vectors needed.
5273  *
5274  * Returns 0 on success, negative on failure
5275  **/
5276 static int i40e_init_msix(struct i40e_pf *pf)
5277 {
5278 	i40e_status err = 0;
5279 	struct i40e_hw *hw = &pf->hw;
5280 	int v_budget, i;
5281 	int vec;
5282 
5283 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
5284 		return -ENODEV;
5285 
5286 	/* The number of vectors we'll request will be comprised of:
5287 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
5288 	 *   - The number of LAN queue pairs
5289 	 *        already adjusted for the NUMA node
5290 	 *        assumes symmetric Tx/Rx pairing
5291 	 *   - The number of VMDq pairs
5292 	 * Once we count this up, try the request.
5293 	 *
5294 	 * If we can't get what we want, we'll simplify to nearly nothing
5295 	 * and try again.  If that still fails, we punt.
5296 	 */
5297 	pf->num_lan_msix = pf->num_lan_qps;
5298 	pf->num_vmdq_msix = pf->num_vmdq_qps;
5299 	v_budget = 1 + pf->num_lan_msix;
5300 	v_budget += (pf->num_vmdq_vsis * pf->num_vmdq_msix);
5301 	if (pf->flags & I40E_FLAG_FDIR_ENABLED)
5302 		v_budget++;
5303 
5304 	/* Scale down if necessary, and the rings will share vectors */
5305 	v_budget = min_t(int, v_budget, hw->func_caps.num_msix_vectors);
5306 
5307 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
5308 				   GFP_KERNEL);
5309 	if (!pf->msix_entries)
5310 		return -ENOMEM;
5311 
5312 	for (i = 0; i < v_budget; i++)
5313 		pf->msix_entries[i].entry = i;
5314 	vec = i40e_reserve_msix_vectors(pf, v_budget);
5315 	if (vec < I40E_MIN_MSIX) {
5316 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
5317 		kfree(pf->msix_entries);
5318 		pf->msix_entries = NULL;
5319 		return -ENODEV;
5320 
5321 	} else if (vec == I40E_MIN_MSIX) {
5322 		/* Adjust for minimal MSIX use */
5323 		dev_info(&pf->pdev->dev, "Features disabled, not enough MSIX vectors\n");
5324 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
5325 		pf->num_vmdq_vsis = 0;
5326 		pf->num_vmdq_qps = 0;
5327 		pf->num_vmdq_msix = 0;
5328 		pf->num_lan_qps = 1;
5329 		pf->num_lan_msix = 1;
5330 
5331 	} else if (vec != v_budget) {
5332 		/* Scale vector usage down */
5333 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
5334 		vec--;                    /* reserve the misc vector */
5335 
5336 		/* partition out the remaining vectors */
5337 		switch (vec) {
5338 		case 2:
5339 			pf->num_vmdq_vsis = 1;
5340 			pf->num_lan_msix = 1;
5341 			break;
5342 		case 3:
5343 			pf->num_vmdq_vsis = 1;
5344 			pf->num_lan_msix = 2;
5345 			break;
5346 		default:
5347 			pf->num_lan_msix = min_t(int, (vec / 2),
5348 						 pf->num_lan_qps);
5349 			pf->num_vmdq_vsis = min_t(int, (vec - pf->num_lan_msix),
5350 						  I40E_DEFAULT_NUM_VMDQ_VSI);
5351 			break;
5352 		}
5353 	}
5354 
5355 	return err;
5356 }
5357 
5358 /**
5359  * i40e_alloc_q_vector - Allocate memory for a single interrupt vector
5360  * @vsi: the VSI being configured
5361  * @v_idx: index of the vector in the vsi struct
5362  *
5363  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
5364  **/
5365 static int i40e_alloc_q_vector(struct i40e_vsi *vsi, int v_idx)
5366 {
5367 	struct i40e_q_vector *q_vector;
5368 
5369 	/* allocate q_vector */
5370 	q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
5371 	if (!q_vector)
5372 		return -ENOMEM;
5373 
5374 	q_vector->vsi = vsi;
5375 	q_vector->v_idx = v_idx;
5376 	cpumask_set_cpu(v_idx, &q_vector->affinity_mask);
5377 	if (vsi->netdev)
5378 		netif_napi_add(vsi->netdev, &q_vector->napi,
5379 			       i40e_napi_poll, vsi->work_limit);
5380 
5381 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
5382 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
5383 
5384 	/* tie q_vector and vsi together */
5385 	vsi->q_vectors[v_idx] = q_vector;
5386 
5387 	return 0;
5388 }
5389 
5390 /**
5391  * i40e_alloc_q_vectors - Allocate memory for interrupt vectors
5392  * @vsi: the VSI being configured
5393  *
5394  * We allocate one q_vector per queue interrupt.  If allocation fails we
5395  * return -ENOMEM.
5396  **/
5397 static int i40e_alloc_q_vectors(struct i40e_vsi *vsi)
5398 {
5399 	struct i40e_pf *pf = vsi->back;
5400 	int v_idx, num_q_vectors;
5401 	int err;
5402 
5403 	/* if not MSIX, give the one vector only to the LAN VSI */
5404 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5405 		num_q_vectors = vsi->num_q_vectors;
5406 	else if (vsi == pf->vsi[pf->lan_vsi])
5407 		num_q_vectors = 1;
5408 	else
5409 		return -EINVAL;
5410 
5411 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
5412 		err = i40e_alloc_q_vector(vsi, v_idx);
5413 		if (err)
5414 			goto err_out;
5415 	}
5416 
5417 	return 0;
5418 
5419 err_out:
5420 	while (v_idx--)
5421 		i40e_free_q_vector(vsi, v_idx);
5422 
5423 	return err;
5424 }
5425 
5426 /**
5427  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
5428  * @pf: board private structure to initialize
5429  **/
5430 static void i40e_init_interrupt_scheme(struct i40e_pf *pf)
5431 {
5432 	int err = 0;
5433 
5434 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
5435 		err = i40e_init_msix(pf);
5436 		if (err) {
5437 			pf->flags &= ~(I40E_FLAG_MSIX_ENABLED	   |
5438 					I40E_FLAG_RSS_ENABLED	   |
5439 					I40E_FLAG_MQ_ENABLED	   |
5440 					I40E_FLAG_DCB_ENABLED	   |
5441 					I40E_FLAG_SRIOV_ENABLED	   |
5442 					I40E_FLAG_FDIR_ENABLED	   |
5443 					I40E_FLAG_FDIR_ATR_ENABLED |
5444 					I40E_FLAG_VMDQ_ENABLED);
5445 
5446 			/* rework the queue expectations without MSIX */
5447 			i40e_determine_queue_usage(pf);
5448 		}
5449 	}
5450 
5451 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
5452 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
5453 		dev_info(&pf->pdev->dev, "MSIX not available, trying MSI\n");
5454 		err = pci_enable_msi(pf->pdev);
5455 		if (err) {
5456 			dev_info(&pf->pdev->dev, "MSI init failed - %d\n", err);
5457 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
5458 		}
5459 	}
5460 
5461 	if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
5462 		dev_info(&pf->pdev->dev, "MSIX and MSI not available, falling back to Legacy IRQ\n");
5463 
5464 	/* track first vector for misc interrupts */
5465 	err = i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT-1);
5466 }
5467 
5468 /**
5469  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
5470  * @pf: board private structure
5471  *
5472  * This sets up the handler for MSIX 0, which is used to manage the
5473  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
5474  * when in MSI or Legacy interrupt mode.
5475  **/
5476 static int i40e_setup_misc_vector(struct i40e_pf *pf)
5477 {
5478 	struct i40e_hw *hw = &pf->hw;
5479 	int err = 0;
5480 
5481 	/* Only request the irq if this is the first time through, and
5482 	 * not when we're rebuilding after a Reset
5483 	 */
5484 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
5485 		err = request_irq(pf->msix_entries[0].vector,
5486 				  i40e_intr, 0, pf->misc_int_name, pf);
5487 		if (err) {
5488 			dev_info(&pf->pdev->dev,
5489 				 "request_irq for msix_misc failed: %d\n", err);
5490 			return -EFAULT;
5491 		}
5492 	}
5493 
5494 	i40e_enable_misc_int_causes(hw);
5495 
5496 	/* associate no queues to the misc vector */
5497 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
5498 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
5499 
5500 	i40e_flush(hw);
5501 
5502 	i40e_irq_dynamic_enable_icr0(pf);
5503 
5504 	return err;
5505 }
5506 
5507 /**
5508  * i40e_config_rss - Prepare for RSS if used
5509  * @pf: board private structure
5510  **/
5511 static int i40e_config_rss(struct i40e_pf *pf)
5512 {
5513 	struct i40e_hw *hw = &pf->hw;
5514 	u32 lut = 0;
5515 	int i, j;
5516 	u64 hena;
5517 	/* Set of random keys generated using kernel random number generator */
5518 	static const u32 seed[I40E_PFQF_HKEY_MAX_INDEX + 1] = {0x41b01687,
5519 				0x183cfd8c, 0xce880440, 0x580cbc3c, 0x35897377,
5520 				0x328b25e1, 0x4fa98922, 0xb7d90c14, 0xd5bad70d,
5521 				0xcd15a2c1, 0xe8580225, 0x4a1e9d11, 0xfe5731be};
5522 
5523 	/* Fill out hash function seed */
5524 	for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
5525 		wr32(hw, I40E_PFQF_HKEY(i), seed[i]);
5526 
5527 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
5528 	hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
5529 		((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
5530 	hena |= ((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV4_UDP) |
5531 		((u64)1 << I40E_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP) |
5532 		((u64)1 << I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP) |
5533 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV4_TCP) |
5534 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV6_TCP) |
5535 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV6_UDP) |
5536 		((u64)1 << I40E_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP) |
5537 		((u64)1 << I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP) |
5538 		((u64)1 << I40E_FILTER_PCTYPE_FRAG_IPV4)|
5539 		((u64)1 << I40E_FILTER_PCTYPE_FRAG_IPV6);
5540 	wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
5541 	wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
5542 
5543 	/* Populate the LUT with max no. of queues in round robin fashion */
5544 	for (i = 0, j = 0; i < pf->hw.func_caps.rss_table_size; i++, j++) {
5545 
5546 		/* The assumption is that lan qp count will be the highest
5547 		 * qp count for any PF VSI that needs RSS.
5548 		 * If multiple VSIs need RSS support, all the qp counts
5549 		 * for those VSIs should be a power of 2 for RSS to work.
5550 		 * If LAN VSI is the only consumer for RSS then this requirement
5551 		 * is not necessary.
5552 		 */
5553 		if (j == pf->rss_size)
5554 			j = 0;
5555 		/* lut = 4-byte sliding window of 4 lut entries */
5556 		lut = (lut << 8) | (j &
5557 			 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
5558 		/* On i = 3, we have 4 entries in lut; write to the register */
5559 		if ((i & 3) == 3)
5560 			wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
5561 	}
5562 	i40e_flush(hw);
5563 
5564 	return 0;
5565 }
5566 
5567 /**
5568  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
5569  * @pf: board private structure to initialize
5570  *
5571  * i40e_sw_init initializes the Adapter private data structure.
5572  * Fields are initialized based on PCI device information and
5573  * OS network device settings (MTU size).
5574  **/
5575 static int i40e_sw_init(struct i40e_pf *pf)
5576 {
5577 	int err = 0;
5578 	int size;
5579 
5580 	pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
5581 				(NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
5582 	if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
5583 		if (I40E_DEBUG_USER & debug)
5584 			pf->hw.debug_mask = debug;
5585 		pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
5586 						I40E_DEFAULT_MSG_ENABLE);
5587 	}
5588 
5589 	/* Set default capability flags */
5590 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
5591 		    I40E_FLAG_MSI_ENABLED     |
5592 		    I40E_FLAG_MSIX_ENABLED    |
5593 		    I40E_FLAG_RX_PS_ENABLED   |
5594 		    I40E_FLAG_MQ_ENABLED      |
5595 		    I40E_FLAG_RX_1BUF_ENABLED;
5596 
5597 	pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
5598 	if (pf->hw.func_caps.rss) {
5599 		pf->flags |= I40E_FLAG_RSS_ENABLED;
5600 		pf->rss_size = min_t(int, pf->rss_size_max,
5601 				     nr_cpus_node(numa_node_id()));
5602 	} else {
5603 		pf->rss_size = 1;
5604 	}
5605 
5606 	if (pf->hw.func_caps.dcb)
5607 		pf->num_tc_qps = I40E_DEFAULT_QUEUES_PER_TC;
5608 	else
5609 		pf->num_tc_qps = 0;
5610 
5611 	if (pf->hw.func_caps.fd) {
5612 		/* FW/NVM is not yet fixed in this regard */
5613 		if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
5614 		    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
5615 			pf->flags |= I40E_FLAG_FDIR_ATR_ENABLED;
5616 			dev_info(&pf->pdev->dev,
5617 				 "Flow Director ATR mode Enabled\n");
5618 			pf->flags |= I40E_FLAG_FDIR_ENABLED;
5619 			dev_info(&pf->pdev->dev,
5620 				 "Flow Director Side Band mode Enabled\n");
5621 			pf->fdir_pf_filter_count =
5622 					 pf->hw.func_caps.fd_filters_guaranteed;
5623 		}
5624 	} else {
5625 		pf->fdir_pf_filter_count = 0;
5626 	}
5627 
5628 	if (pf->hw.func_caps.vmdq) {
5629 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
5630 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
5631 		pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
5632 	}
5633 
5634 	/* MFP mode enabled */
5635 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
5636 		pf->flags |= I40E_FLAG_MFP_ENABLED;
5637 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
5638 	}
5639 
5640 #ifdef CONFIG_PCI_IOV
5641 	if (pf->hw.func_caps.num_vfs) {
5642 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
5643 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
5644 		pf->num_req_vfs = min_t(int,
5645 					pf->hw.func_caps.num_vfs,
5646 					I40E_MAX_VF_COUNT);
5647 	}
5648 #endif /* CONFIG_PCI_IOV */
5649 	pf->eeprom_version = 0xDEAD;
5650 	pf->lan_veb = I40E_NO_VEB;
5651 	pf->lan_vsi = I40E_NO_VSI;
5652 
5653 	/* set up queue assignment tracking */
5654 	size = sizeof(struct i40e_lump_tracking)
5655 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
5656 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
5657 	if (!pf->qp_pile) {
5658 		err = -ENOMEM;
5659 		goto sw_init_done;
5660 	}
5661 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
5662 	pf->qp_pile->search_hint = 0;
5663 
5664 	/* set up vector assignment tracking */
5665 	size = sizeof(struct i40e_lump_tracking)
5666 		+ (sizeof(u16) * pf->hw.func_caps.num_msix_vectors);
5667 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
5668 	if (!pf->irq_pile) {
5669 		kfree(pf->qp_pile);
5670 		err = -ENOMEM;
5671 		goto sw_init_done;
5672 	}
5673 	pf->irq_pile->num_entries = pf->hw.func_caps.num_msix_vectors;
5674 	pf->irq_pile->search_hint = 0;
5675 
5676 	mutex_init(&pf->switch_mutex);
5677 
5678 sw_init_done:
5679 	return err;
5680 }
5681 
5682 /**
5683  * i40e_set_features - set the netdev feature flags
5684  * @netdev: ptr to the netdev being adjusted
5685  * @features: the feature set that the stack is suggesting
5686  **/
5687 static int i40e_set_features(struct net_device *netdev,
5688 			     netdev_features_t features)
5689 {
5690 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5691 	struct i40e_vsi *vsi = np->vsi;
5692 
5693 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5694 		i40e_vlan_stripping_enable(vsi);
5695 	else
5696 		i40e_vlan_stripping_disable(vsi);
5697 
5698 	return 0;
5699 }
5700 
5701 static const struct net_device_ops i40e_netdev_ops = {
5702 	.ndo_open		= i40e_open,
5703 	.ndo_stop		= i40e_close,
5704 	.ndo_start_xmit		= i40e_lan_xmit_frame,
5705 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
5706 	.ndo_set_rx_mode	= i40e_set_rx_mode,
5707 	.ndo_validate_addr	= eth_validate_addr,
5708 	.ndo_set_mac_address	= i40e_set_mac,
5709 	.ndo_change_mtu		= i40e_change_mtu,
5710 	.ndo_tx_timeout		= i40e_tx_timeout,
5711 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
5712 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
5713 #ifdef CONFIG_NET_POLL_CONTROLLER
5714 	.ndo_poll_controller	= i40e_netpoll,
5715 #endif
5716 	.ndo_setup_tc		= i40e_setup_tc,
5717 	.ndo_set_features	= i40e_set_features,
5718 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
5719 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
5720 	.ndo_set_vf_tx_rate	= i40e_ndo_set_vf_bw,
5721 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
5722 };
5723 
5724 /**
5725  * i40e_config_netdev - Setup the netdev flags
5726  * @vsi: the VSI being configured
5727  *
5728  * Returns 0 on success, negative value on failure
5729  **/
5730 static int i40e_config_netdev(struct i40e_vsi *vsi)
5731 {
5732 	struct i40e_pf *pf = vsi->back;
5733 	struct i40e_hw *hw = &pf->hw;
5734 	struct i40e_netdev_priv *np;
5735 	struct net_device *netdev;
5736 	u8 mac_addr[ETH_ALEN];
5737 	int etherdev_size;
5738 
5739 	etherdev_size = sizeof(struct i40e_netdev_priv);
5740 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
5741 	if (!netdev)
5742 		return -ENOMEM;
5743 
5744 	vsi->netdev = netdev;
5745 	np = netdev_priv(netdev);
5746 	np->vsi = vsi;
5747 
5748 	netdev->hw_enc_features = NETIF_F_IP_CSUM	 |
5749 				  NETIF_F_GSO_UDP_TUNNEL |
5750 				  NETIF_F_TSO		 |
5751 				  NETIF_F_SG;
5752 
5753 	netdev->features = NETIF_F_SG		       |
5754 			   NETIF_F_IP_CSUM	       |
5755 			   NETIF_F_SCTP_CSUM	       |
5756 			   NETIF_F_HIGHDMA	       |
5757 			   NETIF_F_GSO_UDP_TUNNEL      |
5758 			   NETIF_F_HW_VLAN_CTAG_TX     |
5759 			   NETIF_F_HW_VLAN_CTAG_RX     |
5760 			   NETIF_F_HW_VLAN_CTAG_FILTER |
5761 			   NETIF_F_IPV6_CSUM	       |
5762 			   NETIF_F_TSO		       |
5763 			   NETIF_F_TSO6		       |
5764 			   NETIF_F_RXCSUM	       |
5765 			   NETIF_F_RXHASH	       |
5766 			   0;
5767 
5768 	/* copy netdev features into list of user selectable features */
5769 	netdev->hw_features |= netdev->features;
5770 
5771 	if (vsi->type == I40E_VSI_MAIN) {
5772 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
5773 		memcpy(mac_addr, hw->mac.perm_addr, ETH_ALEN);
5774 	} else {
5775 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
5776 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
5777 			 pf->vsi[pf->lan_vsi]->netdev->name);
5778 		random_ether_addr(mac_addr);
5779 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
5780 	}
5781 
5782 	memcpy(netdev->dev_addr, mac_addr, ETH_ALEN);
5783 	memcpy(netdev->perm_addr, mac_addr, ETH_ALEN);
5784 	/* vlan gets same features (except vlan offload)
5785 	 * after any tweaks for specific VSI types
5786 	 */
5787 	netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
5788 						     NETIF_F_HW_VLAN_CTAG_RX |
5789 						   NETIF_F_HW_VLAN_CTAG_FILTER);
5790 	netdev->priv_flags |= IFF_UNICAST_FLT;
5791 	netdev->priv_flags |= IFF_SUPP_NOFCS;
5792 	/* Setup netdev TC information */
5793 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
5794 
5795 	netdev->netdev_ops = &i40e_netdev_ops;
5796 	netdev->watchdog_timeo = 5 * HZ;
5797 	i40e_set_ethtool_ops(netdev);
5798 
5799 	return 0;
5800 }
5801 
5802 /**
5803  * i40e_vsi_delete - Delete a VSI from the switch
5804  * @vsi: the VSI being removed
5805  *
5806  * Returns 0 on success, negative value on failure
5807  **/
5808 static void i40e_vsi_delete(struct i40e_vsi *vsi)
5809 {
5810 	/* remove default VSI is not allowed */
5811 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
5812 		return;
5813 
5814 	/* there is no HW VSI for FDIR */
5815 	if (vsi->type == I40E_VSI_FDIR)
5816 		return;
5817 
5818 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
5819 	return;
5820 }
5821 
5822 /**
5823  * i40e_add_vsi - Add a VSI to the switch
5824  * @vsi: the VSI being configured
5825  *
5826  * This initializes a VSI context depending on the VSI type to be added and
5827  * passes it down to the add_vsi aq command.
5828  **/
5829 static int i40e_add_vsi(struct i40e_vsi *vsi)
5830 {
5831 	int ret = -ENODEV;
5832 	struct i40e_mac_filter *f, *ftmp;
5833 	struct i40e_pf *pf = vsi->back;
5834 	struct i40e_hw *hw = &pf->hw;
5835 	struct i40e_vsi_context ctxt;
5836 	u8 enabled_tc = 0x1; /* TC0 enabled */
5837 	int f_count = 0;
5838 
5839 	memset(&ctxt, 0, sizeof(ctxt));
5840 	switch (vsi->type) {
5841 	case I40E_VSI_MAIN:
5842 		/* The PF's main VSI is already setup as part of the
5843 		 * device initialization, so we'll not bother with
5844 		 * the add_vsi call, but we will retrieve the current
5845 		 * VSI context.
5846 		 */
5847 		ctxt.seid = pf->main_vsi_seid;
5848 		ctxt.pf_num = pf->hw.pf_id;
5849 		ctxt.vf_num = 0;
5850 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
5851 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
5852 		if (ret) {
5853 			dev_info(&pf->pdev->dev,
5854 				 "couldn't get pf vsi config, err %d, aq_err %d\n",
5855 				 ret, pf->hw.aq.asq_last_status);
5856 			return -ENOENT;
5857 		}
5858 		memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
5859 		vsi->info.valid_sections = 0;
5860 
5861 		vsi->seid = ctxt.seid;
5862 		vsi->id = ctxt.vsi_number;
5863 
5864 		enabled_tc = i40e_pf_get_tc_map(pf);
5865 
5866 		/* MFP mode setup queue map and update VSI */
5867 		if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5868 			memset(&ctxt, 0, sizeof(ctxt));
5869 			ctxt.seid = pf->main_vsi_seid;
5870 			ctxt.pf_num = pf->hw.pf_id;
5871 			ctxt.vf_num = 0;
5872 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5873 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
5874 			if (ret) {
5875 				dev_info(&pf->pdev->dev,
5876 					 "update vsi failed, aq_err=%d\n",
5877 					 pf->hw.aq.asq_last_status);
5878 				ret = -ENOENT;
5879 				goto err;
5880 			}
5881 			/* update the local VSI info queue map */
5882 			i40e_vsi_update_queue_map(vsi, &ctxt);
5883 			vsi->info.valid_sections = 0;
5884 		} else {
5885 			/* Default/Main VSI is only enabled for TC0
5886 			 * reconfigure it to enable all TCs that are
5887 			 * available on the port in SFP mode.
5888 			 */
5889 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
5890 			if (ret) {
5891 				dev_info(&pf->pdev->dev,
5892 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
5893 					 enabled_tc, ret,
5894 					 pf->hw.aq.asq_last_status);
5895 				ret = -ENOENT;
5896 			}
5897 		}
5898 		break;
5899 
5900 	case I40E_VSI_FDIR:
5901 		/* no queue mapping or actual HW VSI needed */
5902 		vsi->info.valid_sections = 0;
5903 		vsi->seid = 0;
5904 		vsi->id = 0;
5905 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5906 		return 0;
5907 		break;
5908 
5909 	case I40E_VSI_VMDQ2:
5910 		ctxt.pf_num = hw->pf_id;
5911 		ctxt.vf_num = 0;
5912 		ctxt.uplink_seid = vsi->uplink_seid;
5913 		ctxt.connection_type = 0x1;     /* regular data port */
5914 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5915 
5916 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5917 
5918 		/* This VSI is connected to VEB so the switch_id
5919 		 * should be set to zero by default.
5920 		 */
5921 		ctxt.info.switch_id = 0;
5922 		ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
5923 		ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5924 
5925 		/* Setup the VSI tx/rx queue map for TC0 only for now */
5926 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5927 		break;
5928 
5929 	case I40E_VSI_SRIOV:
5930 		ctxt.pf_num = hw->pf_id;
5931 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
5932 		ctxt.uplink_seid = vsi->uplink_seid;
5933 		ctxt.connection_type = 0x1;     /* regular data port */
5934 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
5935 
5936 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5937 
5938 		/* This VSI is connected to VEB so the switch_id
5939 		 * should be set to zero by default.
5940 		 */
5941 		ctxt.info.switch_id = cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5942 
5943 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
5944 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
5945 		/* Setup the VSI tx/rx queue map for TC0 only for now */
5946 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5947 		break;
5948 
5949 	default:
5950 		return -ENODEV;
5951 	}
5952 
5953 	if (vsi->type != I40E_VSI_MAIN) {
5954 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5955 		if (ret) {
5956 			dev_info(&vsi->back->pdev->dev,
5957 				 "add vsi failed, aq_err=%d\n",
5958 				 vsi->back->hw.aq.asq_last_status);
5959 			ret = -ENOENT;
5960 			goto err;
5961 		}
5962 		memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
5963 		vsi->info.valid_sections = 0;
5964 		vsi->seid = ctxt.seid;
5965 		vsi->id = ctxt.vsi_number;
5966 	}
5967 
5968 	/* If macvlan filters already exist, force them to get loaded */
5969 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
5970 		f->changed = true;
5971 		f_count++;
5972 	}
5973 	if (f_count) {
5974 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
5975 		pf->flags |= I40E_FLAG_FILTER_SYNC;
5976 	}
5977 
5978 	/* Update VSI BW information */
5979 	ret = i40e_vsi_get_bw_info(vsi);
5980 	if (ret) {
5981 		dev_info(&pf->pdev->dev,
5982 			 "couldn't get vsi bw info, err %d, aq_err %d\n",
5983 			 ret, pf->hw.aq.asq_last_status);
5984 		/* VSI is already added so not tearing that up */
5985 		ret = 0;
5986 	}
5987 
5988 err:
5989 	return ret;
5990 }
5991 
5992 /**
5993  * i40e_vsi_release - Delete a VSI and free its resources
5994  * @vsi: the VSI being removed
5995  *
5996  * Returns 0 on success or < 0 on error
5997  **/
5998 int i40e_vsi_release(struct i40e_vsi *vsi)
5999 {
6000 	struct i40e_mac_filter *f, *ftmp;
6001 	struct i40e_veb *veb = NULL;
6002 	struct i40e_pf *pf;
6003 	u16 uplink_seid;
6004 	int i, n;
6005 
6006 	pf = vsi->back;
6007 
6008 	/* release of a VEB-owner or last VSI is not allowed */
6009 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
6010 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
6011 			 vsi->seid, vsi->uplink_seid);
6012 		return -ENODEV;
6013 	}
6014 	if (vsi == pf->vsi[pf->lan_vsi] &&
6015 	    !test_bit(__I40E_DOWN, &pf->state)) {
6016 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
6017 		return -ENODEV;
6018 	}
6019 
6020 	uplink_seid = vsi->uplink_seid;
6021 	if (vsi->type != I40E_VSI_SRIOV) {
6022 		if (vsi->netdev_registered) {
6023 			vsi->netdev_registered = false;
6024 			if (vsi->netdev) {
6025 				/* results in a call to i40e_close() */
6026 				unregister_netdev(vsi->netdev);
6027 				free_netdev(vsi->netdev);
6028 				vsi->netdev = NULL;
6029 			}
6030 		} else {
6031 			if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
6032 				i40e_down(vsi);
6033 			i40e_vsi_free_irq(vsi);
6034 			i40e_vsi_free_tx_resources(vsi);
6035 			i40e_vsi_free_rx_resources(vsi);
6036 		}
6037 		i40e_vsi_disable_irq(vsi);
6038 	}
6039 
6040 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
6041 		i40e_del_filter(vsi, f->macaddr, f->vlan,
6042 				f->is_vf, f->is_netdev);
6043 	i40e_sync_vsi_filters(vsi);
6044 
6045 	i40e_vsi_delete(vsi);
6046 	i40e_vsi_free_q_vectors(vsi);
6047 	i40e_vsi_clear_rings(vsi);
6048 	i40e_vsi_clear(vsi);
6049 
6050 	/* If this was the last thing on the VEB, except for the
6051 	 * controlling VSI, remove the VEB, which puts the controlling
6052 	 * VSI onto the next level down in the switch.
6053 	 *
6054 	 * Well, okay, there's one more exception here: don't remove
6055 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
6056 	 * from up the network stack.
6057 	 */
6058 	for (n = 0, i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6059 		if (pf->vsi[i] &&
6060 		    pf->vsi[i]->uplink_seid == uplink_seid &&
6061 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
6062 			n++;      /* count the VSIs */
6063 		}
6064 	}
6065 	for (i = 0; i < I40E_MAX_VEB; i++) {
6066 		if (!pf->veb[i])
6067 			continue;
6068 		if (pf->veb[i]->uplink_seid == uplink_seid)
6069 			n++;     /* count the VEBs */
6070 		if (pf->veb[i]->seid == uplink_seid)
6071 			veb = pf->veb[i];
6072 	}
6073 	if (n == 0 && veb && veb->uplink_seid != 0)
6074 		i40e_veb_release(veb);
6075 
6076 	return 0;
6077 }
6078 
6079 /**
6080  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
6081  * @vsi: ptr to the VSI
6082  *
6083  * This should only be called after i40e_vsi_mem_alloc() which allocates the
6084  * corresponding SW VSI structure and initializes num_queue_pairs for the
6085  * newly allocated VSI.
6086  *
6087  * Returns 0 on success or negative on failure
6088  **/
6089 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
6090 {
6091 	int ret = -ENOENT;
6092 	struct i40e_pf *pf = vsi->back;
6093 
6094 	if (vsi->q_vectors[0]) {
6095 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
6096 			 vsi->seid);
6097 		return -EEXIST;
6098 	}
6099 
6100 	if (vsi->base_vector) {
6101 		dev_info(&pf->pdev->dev,
6102 			 "VSI %d has non-zero base vector %d\n",
6103 			 vsi->seid, vsi->base_vector);
6104 		return -EEXIST;
6105 	}
6106 
6107 	ret = i40e_alloc_q_vectors(vsi);
6108 	if (ret) {
6109 		dev_info(&pf->pdev->dev,
6110 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
6111 			 vsi->num_q_vectors, vsi->seid, ret);
6112 		vsi->num_q_vectors = 0;
6113 		goto vector_setup_out;
6114 	}
6115 
6116 	if (vsi->num_q_vectors)
6117 		vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
6118 						 vsi->num_q_vectors, vsi->idx);
6119 	if (vsi->base_vector < 0) {
6120 		dev_info(&pf->pdev->dev,
6121 			 "failed to get q tracking for VSI %d, err=%d\n",
6122 			 vsi->seid, vsi->base_vector);
6123 		i40e_vsi_free_q_vectors(vsi);
6124 		ret = -ENOENT;
6125 		goto vector_setup_out;
6126 	}
6127 
6128 vector_setup_out:
6129 	return ret;
6130 }
6131 
6132 /**
6133  * i40e_vsi_setup - Set up a VSI by a given type
6134  * @pf: board private structure
6135  * @type: VSI type
6136  * @uplink_seid: the switch element to link to
6137  * @param1: usage depends upon VSI type. For VF types, indicates VF id
6138  *
6139  * This allocates the sw VSI structure and its queue resources, then add a VSI
6140  * to the identified VEB.
6141  *
6142  * Returns pointer to the successfully allocated and configure VSI sw struct on
6143  * success, otherwise returns NULL on failure.
6144  **/
6145 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
6146 				u16 uplink_seid, u32 param1)
6147 {
6148 	struct i40e_vsi *vsi = NULL;
6149 	struct i40e_veb *veb = NULL;
6150 	int ret, i;
6151 	int v_idx;
6152 
6153 	/* The requested uplink_seid must be either
6154 	 *     - the PF's port seid
6155 	 *              no VEB is needed because this is the PF
6156 	 *              or this is a Flow Director special case VSI
6157 	 *     - seid of an existing VEB
6158 	 *     - seid of a VSI that owns an existing VEB
6159 	 *     - seid of a VSI that doesn't own a VEB
6160 	 *              a new VEB is created and the VSI becomes the owner
6161 	 *     - seid of the PF VSI, which is what creates the first VEB
6162 	 *              this is a special case of the previous
6163 	 *
6164 	 * Find which uplink_seid we were given and create a new VEB if needed
6165 	 */
6166 	for (i = 0; i < I40E_MAX_VEB; i++) {
6167 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
6168 			veb = pf->veb[i];
6169 			break;
6170 		}
6171 	}
6172 
6173 	if (!veb && uplink_seid != pf->mac_seid) {
6174 
6175 		for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6176 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
6177 				vsi = pf->vsi[i];
6178 				break;
6179 			}
6180 		}
6181 		if (!vsi) {
6182 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
6183 				 uplink_seid);
6184 			return NULL;
6185 		}
6186 
6187 		if (vsi->uplink_seid == pf->mac_seid)
6188 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
6189 					     vsi->tc_config.enabled_tc);
6190 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
6191 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
6192 					     vsi->tc_config.enabled_tc);
6193 
6194 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
6195 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
6196 				veb = pf->veb[i];
6197 		}
6198 		if (!veb) {
6199 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
6200 			return NULL;
6201 		}
6202 
6203 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6204 		uplink_seid = veb->seid;
6205 	}
6206 
6207 	/* get vsi sw struct */
6208 	v_idx = i40e_vsi_mem_alloc(pf, type);
6209 	if (v_idx < 0)
6210 		goto err_alloc;
6211 	vsi = pf->vsi[v_idx];
6212 	vsi->type = type;
6213 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
6214 
6215 	if (type == I40E_VSI_MAIN)
6216 		pf->lan_vsi = v_idx;
6217 	else if (type == I40E_VSI_SRIOV)
6218 		vsi->vf_id = param1;
6219 	/* assign it some queues */
6220 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
6221 	if (ret < 0) {
6222 		dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
6223 			 vsi->seid, ret);
6224 		goto err_vsi;
6225 	}
6226 	vsi->base_queue = ret;
6227 
6228 	/* get a VSI from the hardware */
6229 	vsi->uplink_seid = uplink_seid;
6230 	ret = i40e_add_vsi(vsi);
6231 	if (ret)
6232 		goto err_vsi;
6233 
6234 	switch (vsi->type) {
6235 	/* setup the netdev if needed */
6236 	case I40E_VSI_MAIN:
6237 	case I40E_VSI_VMDQ2:
6238 		ret = i40e_config_netdev(vsi);
6239 		if (ret)
6240 			goto err_netdev;
6241 		ret = register_netdev(vsi->netdev);
6242 		if (ret)
6243 			goto err_netdev;
6244 		vsi->netdev_registered = true;
6245 		netif_carrier_off(vsi->netdev);
6246 		/* fall through */
6247 
6248 	case I40E_VSI_FDIR:
6249 		/* set up vectors and rings if needed */
6250 		ret = i40e_vsi_setup_vectors(vsi);
6251 		if (ret)
6252 			goto err_msix;
6253 
6254 		ret = i40e_alloc_rings(vsi);
6255 		if (ret)
6256 			goto err_rings;
6257 
6258 		/* map all of the rings to the q_vectors */
6259 		i40e_vsi_map_rings_to_vectors(vsi);
6260 
6261 		i40e_vsi_reset_stats(vsi);
6262 		break;
6263 
6264 	default:
6265 		/* no netdev or rings for the other VSI types */
6266 		break;
6267 	}
6268 
6269 	return vsi;
6270 
6271 err_rings:
6272 	i40e_vsi_free_q_vectors(vsi);
6273 err_msix:
6274 	if (vsi->netdev_registered) {
6275 		vsi->netdev_registered = false;
6276 		unregister_netdev(vsi->netdev);
6277 		free_netdev(vsi->netdev);
6278 		vsi->netdev = NULL;
6279 	}
6280 err_netdev:
6281 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
6282 err_vsi:
6283 	i40e_vsi_clear(vsi);
6284 err_alloc:
6285 	return NULL;
6286 }
6287 
6288 /**
6289  * i40e_veb_get_bw_info - Query VEB BW information
6290  * @veb: the veb to query
6291  *
6292  * Query the Tx scheduler BW configuration data for given VEB
6293  **/
6294 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
6295 {
6296 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
6297 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
6298 	struct i40e_pf *pf = veb->pf;
6299 	struct i40e_hw *hw = &pf->hw;
6300 	u32 tc_bw_max;
6301 	int ret = 0;
6302 	int i;
6303 
6304 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
6305 						  &bw_data, NULL);
6306 	if (ret) {
6307 		dev_info(&pf->pdev->dev,
6308 			 "query veb bw config failed, aq_err=%d\n",
6309 			 hw->aq.asq_last_status);
6310 		goto out;
6311 	}
6312 
6313 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
6314 						   &ets_data, NULL);
6315 	if (ret) {
6316 		dev_info(&pf->pdev->dev,
6317 			 "query veb bw ets config failed, aq_err=%d\n",
6318 			 hw->aq.asq_last_status);
6319 		goto out;
6320 	}
6321 
6322 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
6323 	veb->bw_max_quanta = ets_data.tc_bw_max;
6324 	veb->is_abs_credits = bw_data.absolute_credits_enable;
6325 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
6326 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
6327 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6328 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
6329 		veb->bw_tc_limit_credits[i] =
6330 					le16_to_cpu(bw_data.tc_bw_limits[i]);
6331 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
6332 	}
6333 
6334 out:
6335 	return ret;
6336 }
6337 
6338 /**
6339  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
6340  * @pf: board private structure
6341  *
6342  * On error: returns error code (negative)
6343  * On success: returns vsi index in PF (positive)
6344  **/
6345 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
6346 {
6347 	int ret = -ENOENT;
6348 	struct i40e_veb *veb;
6349 	int i;
6350 
6351 	/* Need to protect the allocation of switch elements at the PF level */
6352 	mutex_lock(&pf->switch_mutex);
6353 
6354 	/* VEB list may be fragmented if VEB creation/destruction has
6355 	 * been happening.  We can afford to do a quick scan to look
6356 	 * for any free slots in the list.
6357 	 *
6358 	 * find next empty veb slot, looping back around if necessary
6359 	 */
6360 	i = 0;
6361 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
6362 		i++;
6363 	if (i >= I40E_MAX_VEB) {
6364 		ret = -ENOMEM;
6365 		goto err_alloc_veb;  /* out of VEB slots! */
6366 	}
6367 
6368 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
6369 	if (!veb) {
6370 		ret = -ENOMEM;
6371 		goto err_alloc_veb;
6372 	}
6373 	veb->pf = pf;
6374 	veb->idx = i;
6375 	veb->enabled_tc = 1;
6376 
6377 	pf->veb[i] = veb;
6378 	ret = i;
6379 err_alloc_veb:
6380 	mutex_unlock(&pf->switch_mutex);
6381 	return ret;
6382 }
6383 
6384 /**
6385  * i40e_switch_branch_release - Delete a branch of the switch tree
6386  * @branch: where to start deleting
6387  *
6388  * This uses recursion to find the tips of the branch to be
6389  * removed, deleting until we get back to and can delete this VEB.
6390  **/
6391 static void i40e_switch_branch_release(struct i40e_veb *branch)
6392 {
6393 	struct i40e_pf *pf = branch->pf;
6394 	u16 branch_seid = branch->seid;
6395 	u16 veb_idx = branch->idx;
6396 	int i;
6397 
6398 	/* release any VEBs on this VEB - RECURSION */
6399 	for (i = 0; i < I40E_MAX_VEB; i++) {
6400 		if (!pf->veb[i])
6401 			continue;
6402 		if (pf->veb[i]->uplink_seid == branch->seid)
6403 			i40e_switch_branch_release(pf->veb[i]);
6404 	}
6405 
6406 	/* Release the VSIs on this VEB, but not the owner VSI.
6407 	 *
6408 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
6409 	 *       the VEB itself, so don't use (*branch) after this loop.
6410 	 */
6411 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6412 		if (!pf->vsi[i])
6413 			continue;
6414 		if (pf->vsi[i]->uplink_seid == branch_seid &&
6415 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
6416 			i40e_vsi_release(pf->vsi[i]);
6417 		}
6418 	}
6419 
6420 	/* There's one corner case where the VEB might not have been
6421 	 * removed, so double check it here and remove it if needed.
6422 	 * This case happens if the veb was created from the debugfs
6423 	 * commands and no VSIs were added to it.
6424 	 */
6425 	if (pf->veb[veb_idx])
6426 		i40e_veb_release(pf->veb[veb_idx]);
6427 }
6428 
6429 /**
6430  * i40e_veb_clear - remove veb struct
6431  * @veb: the veb to remove
6432  **/
6433 static void i40e_veb_clear(struct i40e_veb *veb)
6434 {
6435 	if (!veb)
6436 		return;
6437 
6438 	if (veb->pf) {
6439 		struct i40e_pf *pf = veb->pf;
6440 
6441 		mutex_lock(&pf->switch_mutex);
6442 		if (pf->veb[veb->idx] == veb)
6443 			pf->veb[veb->idx] = NULL;
6444 		mutex_unlock(&pf->switch_mutex);
6445 	}
6446 
6447 	kfree(veb);
6448 }
6449 
6450 /**
6451  * i40e_veb_release - Delete a VEB and free its resources
6452  * @veb: the VEB being removed
6453  **/
6454 void i40e_veb_release(struct i40e_veb *veb)
6455 {
6456 	struct i40e_vsi *vsi = NULL;
6457 	struct i40e_pf *pf;
6458 	int i, n = 0;
6459 
6460 	pf = veb->pf;
6461 
6462 	/* find the remaining VSI and check for extras */
6463 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6464 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
6465 			n++;
6466 			vsi = pf->vsi[i];
6467 		}
6468 	}
6469 	if (n != 1) {
6470 		dev_info(&pf->pdev->dev,
6471 			 "can't remove VEB %d with %d VSIs left\n",
6472 			 veb->seid, n);
6473 		return;
6474 	}
6475 
6476 	/* move the remaining VSI to uplink veb */
6477 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
6478 	if (veb->uplink_seid) {
6479 		vsi->uplink_seid = veb->uplink_seid;
6480 		if (veb->uplink_seid == pf->mac_seid)
6481 			vsi->veb_idx = I40E_NO_VEB;
6482 		else
6483 			vsi->veb_idx = veb->veb_idx;
6484 	} else {
6485 		/* floating VEB */
6486 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6487 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
6488 	}
6489 
6490 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
6491 	i40e_veb_clear(veb);
6492 
6493 	return;
6494 }
6495 
6496 /**
6497  * i40e_add_veb - create the VEB in the switch
6498  * @veb: the VEB to be instantiated
6499  * @vsi: the controlling VSI
6500  **/
6501 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
6502 {
6503 	bool is_default = (vsi->idx == vsi->back->lan_vsi);
6504 	int ret;
6505 
6506 	/* get a VEB from the hardware */
6507 	ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
6508 			      veb->enabled_tc, is_default, &veb->seid, NULL);
6509 	if (ret) {
6510 		dev_info(&veb->pf->pdev->dev,
6511 			 "couldn't add VEB, err %d, aq_err %d\n",
6512 			 ret, veb->pf->hw.aq.asq_last_status);
6513 		return -EPERM;
6514 	}
6515 
6516 	/* get statistics counter */
6517 	ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
6518 					 &veb->stats_idx, NULL, NULL, NULL);
6519 	if (ret) {
6520 		dev_info(&veb->pf->pdev->dev,
6521 			 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
6522 			 ret, veb->pf->hw.aq.asq_last_status);
6523 		return -EPERM;
6524 	}
6525 	ret = i40e_veb_get_bw_info(veb);
6526 	if (ret) {
6527 		dev_info(&veb->pf->pdev->dev,
6528 			 "couldn't get VEB bw info, err %d, aq_err %d\n",
6529 			 ret, veb->pf->hw.aq.asq_last_status);
6530 		i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
6531 		return -ENOENT;
6532 	}
6533 
6534 	vsi->uplink_seid = veb->seid;
6535 	vsi->veb_idx = veb->idx;
6536 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6537 
6538 	return 0;
6539 }
6540 
6541 /**
6542  * i40e_veb_setup - Set up a VEB
6543  * @pf: board private structure
6544  * @flags: VEB setup flags
6545  * @uplink_seid: the switch element to link to
6546  * @vsi_seid: the initial VSI seid
6547  * @enabled_tc: Enabled TC bit-map
6548  *
6549  * This allocates the sw VEB structure and links it into the switch
6550  * It is possible and legal for this to be a duplicate of an already
6551  * existing VEB.  It is also possible for both uplink and vsi seids
6552  * to be zero, in order to create a floating VEB.
6553  *
6554  * Returns pointer to the successfully allocated VEB sw struct on
6555  * success, otherwise returns NULL on failure.
6556  **/
6557 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
6558 				u16 uplink_seid, u16 vsi_seid,
6559 				u8 enabled_tc)
6560 {
6561 	struct i40e_veb *veb, *uplink_veb = NULL;
6562 	int vsi_idx, veb_idx;
6563 	int ret;
6564 
6565 	/* if one seid is 0, the other must be 0 to create a floating relay */
6566 	if ((uplink_seid == 0 || vsi_seid == 0) &&
6567 	    (uplink_seid + vsi_seid != 0)) {
6568 		dev_info(&pf->pdev->dev,
6569 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
6570 			 uplink_seid, vsi_seid);
6571 		return NULL;
6572 	}
6573 
6574 	/* make sure there is such a vsi and uplink */
6575 	for (vsi_idx = 0; vsi_idx < pf->hw.func_caps.num_vsis; vsi_idx++)
6576 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
6577 			break;
6578 	if (vsi_idx >= pf->hw.func_caps.num_vsis && vsi_seid != 0) {
6579 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
6580 			 vsi_seid);
6581 		return NULL;
6582 	}
6583 
6584 	if (uplink_seid && uplink_seid != pf->mac_seid) {
6585 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6586 			if (pf->veb[veb_idx] &&
6587 			    pf->veb[veb_idx]->seid == uplink_seid) {
6588 				uplink_veb = pf->veb[veb_idx];
6589 				break;
6590 			}
6591 		}
6592 		if (!uplink_veb) {
6593 			dev_info(&pf->pdev->dev,
6594 				 "uplink seid %d not found\n", uplink_seid);
6595 			return NULL;
6596 		}
6597 	}
6598 
6599 	/* get veb sw struct */
6600 	veb_idx = i40e_veb_mem_alloc(pf);
6601 	if (veb_idx < 0)
6602 		goto err_alloc;
6603 	veb = pf->veb[veb_idx];
6604 	veb->flags = flags;
6605 	veb->uplink_seid = uplink_seid;
6606 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
6607 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
6608 
6609 	/* create the VEB in the switch */
6610 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
6611 	if (ret)
6612 		goto err_veb;
6613 
6614 	return veb;
6615 
6616 err_veb:
6617 	i40e_veb_clear(veb);
6618 err_alloc:
6619 	return NULL;
6620 }
6621 
6622 /**
6623  * i40e_setup_pf_switch_element - set pf vars based on switch type
6624  * @pf: board private structure
6625  * @ele: element we are building info from
6626  * @num_reported: total number of elements
6627  * @printconfig: should we print the contents
6628  *
6629  * helper function to assist in extracting a few useful SEID values.
6630  **/
6631 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
6632 				struct i40e_aqc_switch_config_element_resp *ele,
6633 				u16 num_reported, bool printconfig)
6634 {
6635 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
6636 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
6637 	u8 element_type = ele->element_type;
6638 	u16 seid = le16_to_cpu(ele->seid);
6639 
6640 	if (printconfig)
6641 		dev_info(&pf->pdev->dev,
6642 			 "type=%d seid=%d uplink=%d downlink=%d\n",
6643 			 element_type, seid, uplink_seid, downlink_seid);
6644 
6645 	switch (element_type) {
6646 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
6647 		pf->mac_seid = seid;
6648 		break;
6649 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
6650 		/* Main VEB? */
6651 		if (uplink_seid != pf->mac_seid)
6652 			break;
6653 		if (pf->lan_veb == I40E_NO_VEB) {
6654 			int v;
6655 
6656 			/* find existing or else empty VEB */
6657 			for (v = 0; v < I40E_MAX_VEB; v++) {
6658 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
6659 					pf->lan_veb = v;
6660 					break;
6661 				}
6662 			}
6663 			if (pf->lan_veb == I40E_NO_VEB) {
6664 				v = i40e_veb_mem_alloc(pf);
6665 				if (v < 0)
6666 					break;
6667 				pf->lan_veb = v;
6668 			}
6669 		}
6670 
6671 		pf->veb[pf->lan_veb]->seid = seid;
6672 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
6673 		pf->veb[pf->lan_veb]->pf = pf;
6674 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
6675 		break;
6676 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
6677 		if (num_reported != 1)
6678 			break;
6679 		/* This is immediately after a reset so we can assume this is
6680 		 * the PF's VSI
6681 		 */
6682 		pf->mac_seid = uplink_seid;
6683 		pf->pf_seid = downlink_seid;
6684 		pf->main_vsi_seid = seid;
6685 		if (printconfig)
6686 			dev_info(&pf->pdev->dev,
6687 				 "pf_seid=%d main_vsi_seid=%d\n",
6688 				 pf->pf_seid, pf->main_vsi_seid);
6689 		break;
6690 	case I40E_SWITCH_ELEMENT_TYPE_PF:
6691 	case I40E_SWITCH_ELEMENT_TYPE_VF:
6692 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
6693 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
6694 	case I40E_SWITCH_ELEMENT_TYPE_PE:
6695 	case I40E_SWITCH_ELEMENT_TYPE_PA:
6696 		/* ignore these for now */
6697 		break;
6698 	default:
6699 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
6700 			 element_type, seid);
6701 		break;
6702 	}
6703 }
6704 
6705 /**
6706  * i40e_fetch_switch_configuration - Get switch config from firmware
6707  * @pf: board private structure
6708  * @printconfig: should we print the contents
6709  *
6710  * Get the current switch configuration from the device and
6711  * extract a few useful SEID values.
6712  **/
6713 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
6714 {
6715 	struct i40e_aqc_get_switch_config_resp *sw_config;
6716 	u16 next_seid = 0;
6717 	int ret = 0;
6718 	u8 *aq_buf;
6719 	int i;
6720 
6721 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
6722 	if (!aq_buf)
6723 		return -ENOMEM;
6724 
6725 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
6726 	do {
6727 		u16 num_reported, num_total;
6728 
6729 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
6730 						I40E_AQ_LARGE_BUF,
6731 						&next_seid, NULL);
6732 		if (ret) {
6733 			dev_info(&pf->pdev->dev,
6734 				 "get switch config failed %d aq_err=%x\n",
6735 				 ret, pf->hw.aq.asq_last_status);
6736 			kfree(aq_buf);
6737 			return -ENOENT;
6738 		}
6739 
6740 		num_reported = le16_to_cpu(sw_config->header.num_reported);
6741 		num_total = le16_to_cpu(sw_config->header.num_total);
6742 
6743 		if (printconfig)
6744 			dev_info(&pf->pdev->dev,
6745 				 "header: %d reported %d total\n",
6746 				 num_reported, num_total);
6747 
6748 		if (num_reported) {
6749 			int sz = sizeof(*sw_config) * num_reported;
6750 
6751 			kfree(pf->sw_config);
6752 			pf->sw_config = kzalloc(sz, GFP_KERNEL);
6753 			if (pf->sw_config)
6754 				memcpy(pf->sw_config, sw_config, sz);
6755 		}
6756 
6757 		for (i = 0; i < num_reported; i++) {
6758 			struct i40e_aqc_switch_config_element_resp *ele =
6759 				&sw_config->element[i];
6760 
6761 			i40e_setup_pf_switch_element(pf, ele, num_reported,
6762 						     printconfig);
6763 		}
6764 	} while (next_seid != 0);
6765 
6766 	kfree(aq_buf);
6767 	return ret;
6768 }
6769 
6770 /**
6771  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
6772  * @pf: board private structure
6773  *
6774  * Returns 0 on success, negative value on failure
6775  **/
6776 static int i40e_setup_pf_switch(struct i40e_pf *pf)
6777 {
6778 	int ret;
6779 
6780 	/* find out what's out there already */
6781 	ret = i40e_fetch_switch_configuration(pf, false);
6782 	if (ret) {
6783 		dev_info(&pf->pdev->dev,
6784 			 "couldn't fetch switch config, err %d, aq_err %d\n",
6785 			 ret, pf->hw.aq.asq_last_status);
6786 		return ret;
6787 	}
6788 	i40e_pf_reset_stats(pf);
6789 
6790 	/* fdir VSI must happen first to be sure it gets queue 0, but only
6791 	 * if there is enough room for the fdir VSI
6792 	 */
6793 	if (pf->num_lan_qps > 1)
6794 		i40e_fdir_setup(pf);
6795 
6796 	/* first time setup */
6797 	if (pf->lan_vsi == I40E_NO_VSI) {
6798 		struct i40e_vsi *vsi = NULL;
6799 		u16 uplink_seid;
6800 
6801 		/* Set up the PF VSI associated with the PF's main VSI
6802 		 * that is already in the HW switch
6803 		 */
6804 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6805 			uplink_seid = pf->veb[pf->lan_veb]->seid;
6806 		else
6807 			uplink_seid = pf->mac_seid;
6808 
6809 		vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
6810 		if (!vsi) {
6811 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
6812 			i40e_fdir_teardown(pf);
6813 			return -EAGAIN;
6814 		}
6815 		/* accommodate kcompat by copying the main VSI queue count
6816 		 * into the pf, since this newer code pushes the pf queue
6817 		 * info down a level into a VSI
6818 		 */
6819 		pf->num_rx_queues = vsi->alloc_queue_pairs;
6820 		pf->num_tx_queues = vsi->alloc_queue_pairs;
6821 	} else {
6822 		/* force a reset of TC and queue layout configurations */
6823 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
6824 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
6825 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
6826 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
6827 	}
6828 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
6829 
6830 	/* Setup static PF queue filter control settings */
6831 	ret = i40e_setup_pf_filter_control(pf);
6832 	if (ret) {
6833 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
6834 			 ret);
6835 		/* Failure here should not stop continuing other steps */
6836 	}
6837 
6838 	/* enable RSS in the HW, even for only one queue, as the stack can use
6839 	 * the hash
6840 	 */
6841 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
6842 		i40e_config_rss(pf);
6843 
6844 	/* fill in link information and enable LSE reporting */
6845 	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
6846 	i40e_link_event(pf);
6847 
6848 	/* Initialize user-specifics link properties */
6849 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
6850 				  I40E_AQ_AN_COMPLETED) ? true : false);
6851 	pf->hw.fc.requested_mode = I40E_FC_DEFAULT;
6852 	if (pf->hw.phy.link_info.an_info &
6853 	   (I40E_AQ_LINK_PAUSE_TX | I40E_AQ_LINK_PAUSE_RX))
6854 		pf->hw.fc.current_mode = I40E_FC_FULL;
6855 	else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX)
6856 		pf->hw.fc.current_mode = I40E_FC_TX_PAUSE;
6857 	else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX)
6858 		pf->hw.fc.current_mode = I40E_FC_RX_PAUSE;
6859 	else
6860 		pf->hw.fc.current_mode = I40E_FC_DEFAULT;
6861 
6862 	return ret;
6863 }
6864 
6865 /**
6866  * i40e_set_rss_size - helper to set rss_size
6867  * @pf: board private structure
6868  * @queues_left: how many queues
6869  */
6870 static u16 i40e_set_rss_size(struct i40e_pf *pf, int queues_left)
6871 {
6872 	int num_tc0;
6873 
6874 	num_tc0 = min_t(int, queues_left, pf->rss_size_max);
6875 	num_tc0 = min_t(int, num_tc0, nr_cpus_node(numa_node_id()));
6876 	num_tc0 = rounddown_pow_of_two(num_tc0);
6877 
6878 	return num_tc0;
6879 }
6880 
6881 /**
6882  * i40e_determine_queue_usage - Work out queue distribution
6883  * @pf: board private structure
6884  **/
6885 static void i40e_determine_queue_usage(struct i40e_pf *pf)
6886 {
6887 	int accum_tc_size;
6888 	int queues_left;
6889 
6890 	pf->num_lan_qps = 0;
6891 	pf->num_tc_qps = rounddown_pow_of_two(pf->num_tc_qps);
6892 	accum_tc_size = (I40E_MAX_TRAFFIC_CLASS - 1) * pf->num_tc_qps;
6893 
6894 	/* Find the max queues to be put into basic use.  We'll always be
6895 	 * using TC0, whether or not DCB is running, and TC0 will get the
6896 	 * big RSS set.
6897 	 */
6898 	queues_left = pf->hw.func_caps.num_tx_qp;
6899 
6900 	if   (!((pf->flags & I40E_FLAG_MSIX_ENABLED)		 &&
6901 		(pf->flags & I40E_FLAG_MQ_ENABLED))		 ||
6902 		!(pf->flags & (I40E_FLAG_RSS_ENABLED |
6903 		I40E_FLAG_FDIR_ENABLED | I40E_FLAG_DCB_ENABLED)) ||
6904 		(queues_left == 1)) {
6905 
6906 		/* one qp for PF, no queues for anything else */
6907 		queues_left = 0;
6908 		pf->rss_size = pf->num_lan_qps = 1;
6909 
6910 		/* make sure all the fancies are disabled */
6911 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED       |
6912 				I40E_FLAG_MQ_ENABLED	   |
6913 				I40E_FLAG_FDIR_ENABLED	   |
6914 				I40E_FLAG_FDIR_ATR_ENABLED |
6915 				I40E_FLAG_DCB_ENABLED	   |
6916 				I40E_FLAG_SRIOV_ENABLED	   |
6917 				I40E_FLAG_VMDQ_ENABLED);
6918 
6919 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED	  &&
6920 		   !(pf->flags & I40E_FLAG_FDIR_ENABLED)  &&
6921 		   !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6922 
6923 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6924 
6925 		queues_left -= pf->rss_size;
6926 		pf->num_lan_qps = pf->rss_size;
6927 
6928 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED	  &&
6929 		   !(pf->flags & I40E_FLAG_FDIR_ENABLED)  &&
6930 		   (pf->flags & I40E_FLAG_DCB_ENABLED)) {
6931 
6932 		/* save num_tc_qps queues for TCs 1 thru 7 and the rest
6933 		 * are set up for RSS in TC0
6934 		 */
6935 		queues_left -= accum_tc_size;
6936 
6937 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6938 
6939 		queues_left -= pf->rss_size;
6940 		if (queues_left < 0) {
6941 			dev_info(&pf->pdev->dev, "not enough queues for DCB\n");
6942 			return;
6943 		}
6944 
6945 		pf->num_lan_qps = pf->rss_size + accum_tc_size;
6946 
6947 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED   &&
6948 		  (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
6949 		  !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6950 
6951 		queues_left -= 1; /* save 1 queue for FD */
6952 
6953 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6954 
6955 		queues_left -= pf->rss_size;
6956 		if (queues_left < 0) {
6957 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director\n");
6958 			return;
6959 		}
6960 
6961 		pf->num_lan_qps = pf->rss_size;
6962 
6963 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED   &&
6964 		  (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
6965 		  (pf->flags & I40E_FLAG_DCB_ENABLED)) {
6966 
6967 		/* save 1 queue for TCs 1 thru 7,
6968 		 * 1 queue for flow director,
6969 		 * and the rest are set up for RSS in TC0
6970 		 */
6971 		queues_left -= 1;
6972 		queues_left -= accum_tc_size;
6973 
6974 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6975 		queues_left -= pf->rss_size;
6976 		if (queues_left < 0) {
6977 			dev_info(&pf->pdev->dev, "not enough queues for DCB and Flow Director\n");
6978 			return;
6979 		}
6980 
6981 		pf->num_lan_qps = pf->rss_size + accum_tc_size;
6982 
6983 	} else {
6984 		dev_info(&pf->pdev->dev,
6985 			 "Invalid configuration, flags=0x%08llx\n", pf->flags);
6986 		return;
6987 	}
6988 
6989 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
6990 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
6991 		pf->num_req_vfs = min_t(int, pf->num_req_vfs, (queues_left /
6992 							       pf->num_vf_qps));
6993 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
6994 	}
6995 
6996 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
6997 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
6998 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
6999 					  (queues_left / pf->num_vmdq_qps));
7000 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
7001 	}
7002 
7003 	return;
7004 }
7005 
7006 /**
7007  * i40e_setup_pf_filter_control - Setup PF static filter control
7008  * @pf: PF to be setup
7009  *
7010  * i40e_setup_pf_filter_control sets up a pf's initial filter control
7011  * settings. If PE/FCoE are enabled then it will also set the per PF
7012  * based filter sizes required for them. It also enables Flow director,
7013  * ethertype and macvlan type filter settings for the pf.
7014  *
7015  * Returns 0 on success, negative on failure
7016  **/
7017 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
7018 {
7019 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
7020 
7021 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
7022 
7023 	/* Flow Director is enabled */
7024 	if (pf->flags & (I40E_FLAG_FDIR_ENABLED | I40E_FLAG_FDIR_ATR_ENABLED))
7025 		settings->enable_fdir = true;
7026 
7027 	/* Ethtype and MACVLAN filters enabled for PF */
7028 	settings->enable_ethtype = true;
7029 	settings->enable_macvlan = true;
7030 
7031 	if (i40e_set_filter_control(&pf->hw, settings))
7032 		return -ENOENT;
7033 
7034 	return 0;
7035 }
7036 
7037 /**
7038  * i40e_probe - Device initialization routine
7039  * @pdev: PCI device information struct
7040  * @ent: entry in i40e_pci_tbl
7041  *
7042  * i40e_probe initializes a pf identified by a pci_dev structure.
7043  * The OS initialization, configuring of the pf private structure,
7044  * and a hardware reset occur.
7045  *
7046  * Returns 0 on success, negative on failure
7047  **/
7048 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7049 {
7050 	struct i40e_driver_version dv;
7051 	struct i40e_pf *pf;
7052 	struct i40e_hw *hw;
7053 	int err = 0;
7054 	u32 len;
7055 
7056 	err = pci_enable_device_mem(pdev);
7057 	if (err)
7058 		return err;
7059 
7060 	/* set up for high or low dma */
7061 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64))) {
7062 		/* coherent mask for the same size will always succeed if
7063 		 * dma_set_mask does
7064 		 */
7065 		dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
7066 	} else if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(32))) {
7067 		dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(32));
7068 	} else {
7069 		dev_err(&pdev->dev, "DMA configuration failed: %d\n", err);
7070 		err = -EIO;
7071 		goto err_dma;
7072 	}
7073 
7074 	/* set up pci connections */
7075 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
7076 					   IORESOURCE_MEM), i40e_driver_name);
7077 	if (err) {
7078 		dev_info(&pdev->dev,
7079 			 "pci_request_selected_regions failed %d\n", err);
7080 		goto err_pci_reg;
7081 	}
7082 
7083 	pci_enable_pcie_error_reporting(pdev);
7084 	pci_set_master(pdev);
7085 
7086 	/* Now that we have a PCI connection, we need to do the
7087 	 * low level device setup.  This is primarily setting up
7088 	 * the Admin Queue structures and then querying for the
7089 	 * device's current profile information.
7090 	 */
7091 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
7092 	if (!pf) {
7093 		err = -ENOMEM;
7094 		goto err_pf_alloc;
7095 	}
7096 	pf->next_vsi = 0;
7097 	pf->pdev = pdev;
7098 	set_bit(__I40E_DOWN, &pf->state);
7099 
7100 	hw = &pf->hw;
7101 	hw->back = pf;
7102 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
7103 			      pci_resource_len(pdev, 0));
7104 	if (!hw->hw_addr) {
7105 		err = -EIO;
7106 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
7107 			 (unsigned int)pci_resource_start(pdev, 0),
7108 			 (unsigned int)pci_resource_len(pdev, 0), err);
7109 		goto err_ioremap;
7110 	}
7111 	hw->vendor_id = pdev->vendor;
7112 	hw->device_id = pdev->device;
7113 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
7114 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
7115 	hw->subsystem_device_id = pdev->subsystem_device;
7116 	hw->bus.device = PCI_SLOT(pdev->devfn);
7117 	hw->bus.func = PCI_FUNC(pdev->devfn);
7118 
7119 	/* Reset here to make sure all is clean and to define PF 'n' */
7120 	err = i40e_pf_reset(hw);
7121 	if (err) {
7122 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
7123 		goto err_pf_reset;
7124 	}
7125 	pf->pfr_count++;
7126 
7127 	hw->aq.num_arq_entries = I40E_AQ_LEN;
7128 	hw->aq.num_asq_entries = I40E_AQ_LEN;
7129 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
7130 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
7131 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
7132 	snprintf(pf->misc_int_name, sizeof(pf->misc_int_name) - 1,
7133 		 "%s-pf%d:misc",
7134 		 dev_driver_string(&pf->pdev->dev), pf->hw.pf_id);
7135 
7136 	err = i40e_init_shared_code(hw);
7137 	if (err) {
7138 		dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
7139 		goto err_pf_reset;
7140 	}
7141 
7142 	err = i40e_init_adminq(hw);
7143 	dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
7144 	if (err) {
7145 		dev_info(&pdev->dev,
7146 			 "init_adminq failed: %d expecting API %02x.%02x\n",
7147 			 err,
7148 			 I40E_FW_API_VERSION_MAJOR, I40E_FW_API_VERSION_MINOR);
7149 		goto err_pf_reset;
7150 	}
7151 
7152 	err = i40e_get_capabilities(pf);
7153 	if (err)
7154 		goto err_adminq_setup;
7155 
7156 	err = i40e_sw_init(pf);
7157 	if (err) {
7158 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
7159 		goto err_sw_init;
7160 	}
7161 
7162 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7163 				hw->func_caps.num_rx_qp,
7164 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
7165 	if (err) {
7166 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
7167 		goto err_init_lan_hmc;
7168 	}
7169 
7170 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7171 	if (err) {
7172 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
7173 		err = -ENOENT;
7174 		goto err_configure_lan_hmc;
7175 	}
7176 
7177 	i40e_get_mac_addr(hw, hw->mac.addr);
7178 	if (i40e_validate_mac_addr(hw->mac.addr)) {
7179 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
7180 		err = -EIO;
7181 		goto err_mac_addr;
7182 	}
7183 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
7184 	memcpy(hw->mac.perm_addr, hw->mac.addr, ETH_ALEN);
7185 
7186 	pci_set_drvdata(pdev, pf);
7187 	pci_save_state(pdev);
7188 
7189 	/* set up periodic task facility */
7190 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
7191 	pf->service_timer_period = HZ;
7192 
7193 	INIT_WORK(&pf->service_task, i40e_service_task);
7194 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
7195 	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
7196 	pf->link_check_timeout = jiffies;
7197 
7198 	/* set up the main switch operations */
7199 	i40e_determine_queue_usage(pf);
7200 	i40e_init_interrupt_scheme(pf);
7201 
7202 	/* Set up the *vsi struct based on the number of VSIs in the HW,
7203 	 * and set up our local tracking of the MAIN PF vsi.
7204 	 */
7205 	len = sizeof(struct i40e_vsi *) * pf->hw.func_caps.num_vsis;
7206 	pf->vsi = kzalloc(len, GFP_KERNEL);
7207 	if (!pf->vsi) {
7208 		err = -ENOMEM;
7209 		goto err_switch_setup;
7210 	}
7211 
7212 	err = i40e_setup_pf_switch(pf);
7213 	if (err) {
7214 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
7215 		goto err_vsis;
7216 	}
7217 
7218 	/* The main driver is (mostly) up and happy. We need to set this state
7219 	 * before setting up the misc vector or we get a race and the vector
7220 	 * ends up disabled forever.
7221 	 */
7222 	clear_bit(__I40E_DOWN, &pf->state);
7223 
7224 	/* In case of MSIX we are going to setup the misc vector right here
7225 	 * to handle admin queue events etc. In case of legacy and MSI
7226 	 * the misc functionality and queue processing is combined in
7227 	 * the same vector and that gets setup at open.
7228 	 */
7229 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7230 		err = i40e_setup_misc_vector(pf);
7231 		if (err) {
7232 			dev_info(&pdev->dev,
7233 				 "setup of misc vector failed: %d\n", err);
7234 			goto err_vsis;
7235 		}
7236 	}
7237 
7238 	/* prep for VF support */
7239 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
7240 	    (pf->flags & I40E_FLAG_MSIX_ENABLED)) {
7241 		u32 val;
7242 
7243 		/* disable link interrupts for VFs */
7244 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
7245 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
7246 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
7247 		i40e_flush(hw);
7248 	}
7249 
7250 	i40e_dbg_pf_init(pf);
7251 
7252 	/* tell the firmware that we're starting */
7253 	dv.major_version = DRV_VERSION_MAJOR;
7254 	dv.minor_version = DRV_VERSION_MINOR;
7255 	dv.build_version = DRV_VERSION_BUILD;
7256 	dv.subbuild_version = 0;
7257 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
7258 
7259 	/* since everything's happy, start the service_task timer */
7260 	mod_timer(&pf->service_timer,
7261 		  round_jiffies(jiffies + pf->service_timer_period));
7262 
7263 	return 0;
7264 
7265 	/* Unwind what we've done if something failed in the setup */
7266 err_vsis:
7267 	set_bit(__I40E_DOWN, &pf->state);
7268 err_switch_setup:
7269 	i40e_clear_interrupt_scheme(pf);
7270 	kfree(pf->vsi);
7271 	del_timer_sync(&pf->service_timer);
7272 err_mac_addr:
7273 err_configure_lan_hmc:
7274 	(void)i40e_shutdown_lan_hmc(hw);
7275 err_init_lan_hmc:
7276 	kfree(pf->qp_pile);
7277 	kfree(pf->irq_pile);
7278 err_sw_init:
7279 err_adminq_setup:
7280 	(void)i40e_shutdown_adminq(hw);
7281 err_pf_reset:
7282 	iounmap(hw->hw_addr);
7283 err_ioremap:
7284 	kfree(pf);
7285 err_pf_alloc:
7286 	pci_disable_pcie_error_reporting(pdev);
7287 	pci_release_selected_regions(pdev,
7288 				     pci_select_bars(pdev, IORESOURCE_MEM));
7289 err_pci_reg:
7290 err_dma:
7291 	pci_disable_device(pdev);
7292 	return err;
7293 }
7294 
7295 /**
7296  * i40e_remove - Device removal routine
7297  * @pdev: PCI device information struct
7298  *
7299  * i40e_remove is called by the PCI subsystem to alert the driver
7300  * that is should release a PCI device.  This could be caused by a
7301  * Hot-Plug event, or because the driver is going to be removed from
7302  * memory.
7303  **/
7304 static void i40e_remove(struct pci_dev *pdev)
7305 {
7306 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7307 	i40e_status ret_code;
7308 	u32 reg;
7309 	int i;
7310 
7311 	i40e_dbg_pf_exit(pf);
7312 
7313 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
7314 		i40e_free_vfs(pf);
7315 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
7316 	}
7317 
7318 	/* no more scheduling of any task */
7319 	set_bit(__I40E_DOWN, &pf->state);
7320 	del_timer_sync(&pf->service_timer);
7321 	cancel_work_sync(&pf->service_task);
7322 
7323 	i40e_fdir_teardown(pf);
7324 
7325 	/* If there is a switch structure or any orphans, remove them.
7326 	 * This will leave only the PF's VSI remaining.
7327 	 */
7328 	for (i = 0; i < I40E_MAX_VEB; i++) {
7329 		if (!pf->veb[i])
7330 			continue;
7331 
7332 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
7333 		    pf->veb[i]->uplink_seid == 0)
7334 			i40e_switch_branch_release(pf->veb[i]);
7335 	}
7336 
7337 	/* Now we can shutdown the PF's VSI, just before we kill
7338 	 * adminq and hmc.
7339 	 */
7340 	if (pf->vsi[pf->lan_vsi])
7341 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
7342 
7343 	i40e_stop_misc_vector(pf);
7344 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7345 		synchronize_irq(pf->msix_entries[0].vector);
7346 		free_irq(pf->msix_entries[0].vector, pf);
7347 	}
7348 
7349 	/* shutdown and destroy the HMC */
7350 	ret_code = i40e_shutdown_lan_hmc(&pf->hw);
7351 	if (ret_code)
7352 		dev_warn(&pdev->dev,
7353 			 "Failed to destroy the HMC resources: %d\n", ret_code);
7354 
7355 	/* shutdown the adminq */
7356 	i40e_aq_queue_shutdown(&pf->hw, true);
7357 	ret_code = i40e_shutdown_adminq(&pf->hw);
7358 	if (ret_code)
7359 		dev_warn(&pdev->dev,
7360 			 "Failed to destroy the Admin Queue resources: %d\n",
7361 			 ret_code);
7362 
7363 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
7364 	i40e_clear_interrupt_scheme(pf);
7365 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7366 		if (pf->vsi[i]) {
7367 			i40e_vsi_clear_rings(pf->vsi[i]);
7368 			i40e_vsi_clear(pf->vsi[i]);
7369 			pf->vsi[i] = NULL;
7370 		}
7371 	}
7372 
7373 	for (i = 0; i < I40E_MAX_VEB; i++) {
7374 		kfree(pf->veb[i]);
7375 		pf->veb[i] = NULL;
7376 	}
7377 
7378 	kfree(pf->qp_pile);
7379 	kfree(pf->irq_pile);
7380 	kfree(pf->sw_config);
7381 	kfree(pf->vsi);
7382 
7383 	/* force a PF reset to clean anything leftover */
7384 	reg = rd32(&pf->hw, I40E_PFGEN_CTRL);
7385 	wr32(&pf->hw, I40E_PFGEN_CTRL, (reg | I40E_PFGEN_CTRL_PFSWR_MASK));
7386 	i40e_flush(&pf->hw);
7387 
7388 	iounmap(pf->hw.hw_addr);
7389 	kfree(pf);
7390 	pci_release_selected_regions(pdev,
7391 				     pci_select_bars(pdev, IORESOURCE_MEM));
7392 
7393 	pci_disable_pcie_error_reporting(pdev);
7394 	pci_disable_device(pdev);
7395 }
7396 
7397 /**
7398  * i40e_pci_error_detected - warning that something funky happened in PCI land
7399  * @pdev: PCI device information struct
7400  *
7401  * Called to warn that something happened and the error handling steps
7402  * are in progress.  Allows the driver to quiesce things, be ready for
7403  * remediation.
7404  **/
7405 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
7406 						enum pci_channel_state error)
7407 {
7408 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7409 
7410 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
7411 
7412 	/* shutdown all operations */
7413 	i40e_pf_quiesce_all_vsi(pf);
7414 
7415 	/* Request a slot reset */
7416 	return PCI_ERS_RESULT_NEED_RESET;
7417 }
7418 
7419 /**
7420  * i40e_pci_error_slot_reset - a PCI slot reset just happened
7421  * @pdev: PCI device information struct
7422  *
7423  * Called to find if the driver can work with the device now that
7424  * the pci slot has been reset.  If a basic connection seems good
7425  * (registers are readable and have sane content) then return a
7426  * happy little PCI_ERS_RESULT_xxx.
7427  **/
7428 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
7429 {
7430 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7431 	pci_ers_result_t result;
7432 	int err;
7433 	u32 reg;
7434 
7435 	dev_info(&pdev->dev, "%s\n", __func__);
7436 	if (pci_enable_device_mem(pdev)) {
7437 		dev_info(&pdev->dev,
7438 			 "Cannot re-enable PCI device after reset.\n");
7439 		result = PCI_ERS_RESULT_DISCONNECT;
7440 	} else {
7441 		pci_set_master(pdev);
7442 		pci_restore_state(pdev);
7443 		pci_save_state(pdev);
7444 		pci_wake_from_d3(pdev, false);
7445 
7446 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
7447 		if (reg == 0)
7448 			result = PCI_ERS_RESULT_RECOVERED;
7449 		else
7450 			result = PCI_ERS_RESULT_DISCONNECT;
7451 	}
7452 
7453 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7454 	if (err) {
7455 		dev_info(&pdev->dev,
7456 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7457 			 err);
7458 		/* non-fatal, continue */
7459 	}
7460 
7461 	return result;
7462 }
7463 
7464 /**
7465  * i40e_pci_error_resume - restart operations after PCI error recovery
7466  * @pdev: PCI device information struct
7467  *
7468  * Called to allow the driver to bring things back up after PCI error
7469  * and/or reset recovery has finished.
7470  **/
7471 static void i40e_pci_error_resume(struct pci_dev *pdev)
7472 {
7473 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7474 
7475 	dev_info(&pdev->dev, "%s\n", __func__);
7476 	i40e_handle_reset_warning(pf);
7477 }
7478 
7479 static const struct pci_error_handlers i40e_err_handler = {
7480 	.error_detected = i40e_pci_error_detected,
7481 	.slot_reset = i40e_pci_error_slot_reset,
7482 	.resume = i40e_pci_error_resume,
7483 };
7484 
7485 static struct pci_driver i40e_driver = {
7486 	.name     = i40e_driver_name,
7487 	.id_table = i40e_pci_tbl,
7488 	.probe    = i40e_probe,
7489 	.remove   = i40e_remove,
7490 	.err_handler = &i40e_err_handler,
7491 	.sriov_configure = i40e_pci_sriov_configure,
7492 };
7493 
7494 /**
7495  * i40e_init_module - Driver registration routine
7496  *
7497  * i40e_init_module is the first routine called when the driver is
7498  * loaded. All it does is register with the PCI subsystem.
7499  **/
7500 static int __init i40e_init_module(void)
7501 {
7502 	pr_info("%s: %s - version %s\n", i40e_driver_name,
7503 		i40e_driver_string, i40e_driver_version_str);
7504 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
7505 	i40e_dbg_init();
7506 	return pci_register_driver(&i40e_driver);
7507 }
7508 module_init(i40e_init_module);
7509 
7510 /**
7511  * i40e_exit_module - Driver exit cleanup routine
7512  *
7513  * i40e_exit_module is called just before the driver is removed
7514  * from memory.
7515  **/
7516 static void __exit i40e_exit_module(void)
7517 {
7518 	pci_unregister_driver(&i40e_driver);
7519 	i40e_dbg_exit();
7520 }
7521 module_exit(i40e_exit_module);
7522