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 9
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 0;
106 
107 	return -ENOMEM;
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 0;
141 
142 	return -ENOMEM;
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 = 0;
178 	int j = 0;
179 
180 	if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
181 		dev_info(&pf->pdev->dev,
182 			 "param err: pile=%p needed=%d id=0x%04x\n",
183 			 pile, needed, id);
184 		return -EINVAL;
185 	}
186 
187 	/* start the linear search with an imperfect hint */
188 	i = pile->search_hint;
189 	while (i < pile->num_entries && ret < 0) {
190 		/* skip already allocated entries */
191 		if (pile->list[i] & I40E_PILE_VALID_BIT) {
192 			i++;
193 			continue;
194 		}
195 
196 		/* do we have enough in this lump? */
197 		for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
198 			if (pile->list[i+j] & I40E_PILE_VALID_BIT)
199 				break;
200 		}
201 
202 		if (j == needed) {
203 			/* there was enough, so assign it to the requestor */
204 			for (j = 0; j < needed; j++)
205 				pile->list[i+j] = id | I40E_PILE_VALID_BIT;
206 			ret = i;
207 			pile->search_hint = i + j;
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 *storage)
351 {
352 	struct i40e_netdev_priv *np = netdev_priv(netdev);
353 	struct i40e_vsi *vsi = np->vsi;
354 
355 	*storage = *i40e_get_vsi_stats_struct(vsi);
356 
357 	return storage;
358 }
359 
360 /**
361  * i40e_vsi_reset_stats - Resets all stats of the given vsi
362  * @vsi: the VSI to have its stats reset
363  **/
364 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
365 {
366 	struct rtnl_link_stats64 *ns;
367 	int i;
368 
369 	if (!vsi)
370 		return;
371 
372 	ns = i40e_get_vsi_stats_struct(vsi);
373 	memset(ns, 0, sizeof(*ns));
374 	memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
375 	memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
376 	memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
377 	if (vsi->rx_rings)
378 		for (i = 0; i < vsi->num_queue_pairs; i++) {
379 			memset(&vsi->rx_rings[i].rx_stats, 0 ,
380 			       sizeof(vsi->rx_rings[i].rx_stats));
381 			memset(&vsi->tx_rings[i].tx_stats, 0,
382 			       sizeof(vsi->tx_rings[i].tx_stats));
383 		}
384 	vsi->stat_offsets_loaded = false;
385 }
386 
387 /**
388  * i40e_pf_reset_stats - Reset all of the stats for the given pf
389  * @pf: the PF to be reset
390  **/
391 void i40e_pf_reset_stats(struct i40e_pf *pf)
392 {
393 	memset(&pf->stats, 0, sizeof(pf->stats));
394 	memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
395 	pf->stat_offsets_loaded = false;
396 }
397 
398 /**
399  * i40e_stat_update48 - read and update a 48 bit stat from the chip
400  * @hw: ptr to the hardware info
401  * @hireg: the high 32 bit reg to read
402  * @loreg: the low 32 bit reg to read
403  * @offset_loaded: has the initial offset been loaded yet
404  * @offset: ptr to current offset value
405  * @stat: ptr to the stat
406  *
407  * Since the device stats are not reset at PFReset, they likely will not
408  * be zeroed when the driver starts.  We'll save the first values read
409  * and use them as offsets to be subtracted from the raw values in order
410  * to report stats that count from zero.  In the process, we also manage
411  * the potential roll-over.
412  **/
413 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
414 			       bool offset_loaded, u64 *offset, u64 *stat)
415 {
416 	u64 new_data;
417 
418 	if (hw->device_id == I40E_QEMU_DEVICE_ID) {
419 		new_data = rd32(hw, loreg);
420 		new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
421 	} else {
422 		new_data = rd64(hw, loreg);
423 	}
424 	if (!offset_loaded)
425 		*offset = new_data;
426 	if (likely(new_data >= *offset))
427 		*stat = new_data - *offset;
428 	else
429 		*stat = (new_data + ((u64)1 << 48)) - *offset;
430 	*stat &= 0xFFFFFFFFFFFFULL;
431 }
432 
433 /**
434  * i40e_stat_update32 - read and update a 32 bit stat from the chip
435  * @hw: ptr to the hardware info
436  * @reg: the hw reg to read
437  * @offset_loaded: has the initial offset been loaded yet
438  * @offset: ptr to current offset value
439  * @stat: ptr to the stat
440  **/
441 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
442 			       bool offset_loaded, u64 *offset, u64 *stat)
443 {
444 	u32 new_data;
445 
446 	new_data = rd32(hw, reg);
447 	if (!offset_loaded)
448 		*offset = new_data;
449 	if (likely(new_data >= *offset))
450 		*stat = (u32)(new_data - *offset);
451 	else
452 		*stat = (u32)((new_data + ((u64)1 << 32)) - *offset);
453 }
454 
455 /**
456  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
457  * @vsi: the VSI to be updated
458  **/
459 void i40e_update_eth_stats(struct i40e_vsi *vsi)
460 {
461 	int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
462 	struct i40e_pf *pf = vsi->back;
463 	struct i40e_hw *hw = &pf->hw;
464 	struct i40e_eth_stats *oes;
465 	struct i40e_eth_stats *es;     /* device's eth stats */
466 
467 	es = &vsi->eth_stats;
468 	oes = &vsi->eth_stats_offsets;
469 
470 	/* Gather up the stats that the hw collects */
471 	i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
472 			   vsi->stat_offsets_loaded,
473 			   &oes->tx_errors, &es->tx_errors);
474 	i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
475 			   vsi->stat_offsets_loaded,
476 			   &oes->rx_discards, &es->rx_discards);
477 
478 	i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
479 			   I40E_GLV_GORCL(stat_idx),
480 			   vsi->stat_offsets_loaded,
481 			   &oes->rx_bytes, &es->rx_bytes);
482 	i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
483 			   I40E_GLV_UPRCL(stat_idx),
484 			   vsi->stat_offsets_loaded,
485 			   &oes->rx_unicast, &es->rx_unicast);
486 	i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
487 			   I40E_GLV_MPRCL(stat_idx),
488 			   vsi->stat_offsets_loaded,
489 			   &oes->rx_multicast, &es->rx_multicast);
490 	i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
491 			   I40E_GLV_BPRCL(stat_idx),
492 			   vsi->stat_offsets_loaded,
493 			   &oes->rx_broadcast, &es->rx_broadcast);
494 
495 	i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
496 			   I40E_GLV_GOTCL(stat_idx),
497 			   vsi->stat_offsets_loaded,
498 			   &oes->tx_bytes, &es->tx_bytes);
499 	i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
500 			   I40E_GLV_UPTCL(stat_idx),
501 			   vsi->stat_offsets_loaded,
502 			   &oes->tx_unicast, &es->tx_unicast);
503 	i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
504 			   I40E_GLV_MPTCL(stat_idx),
505 			   vsi->stat_offsets_loaded,
506 			   &oes->tx_multicast, &es->tx_multicast);
507 	i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
508 			   I40E_GLV_BPTCL(stat_idx),
509 			   vsi->stat_offsets_loaded,
510 			   &oes->tx_broadcast, &es->tx_broadcast);
511 	vsi->stat_offsets_loaded = true;
512 }
513 
514 /**
515  * i40e_update_veb_stats - Update Switch component statistics
516  * @veb: the VEB being updated
517  **/
518 static void i40e_update_veb_stats(struct i40e_veb *veb)
519 {
520 	struct i40e_pf *pf = veb->pf;
521 	struct i40e_hw *hw = &pf->hw;
522 	struct i40e_eth_stats *oes;
523 	struct i40e_eth_stats *es;     /* device's eth stats */
524 	int idx = 0;
525 
526 	idx = veb->stats_idx;
527 	es = &veb->stats;
528 	oes = &veb->stats_offsets;
529 
530 	/* Gather up the stats that the hw collects */
531 	i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
532 			   veb->stat_offsets_loaded,
533 			   &oes->tx_discards, &es->tx_discards);
534 	i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
535 			   veb->stat_offsets_loaded,
536 			   &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
537 
538 	i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
539 			   veb->stat_offsets_loaded,
540 			   &oes->rx_bytes, &es->rx_bytes);
541 	i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
542 			   veb->stat_offsets_loaded,
543 			   &oes->rx_unicast, &es->rx_unicast);
544 	i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
545 			   veb->stat_offsets_loaded,
546 			   &oes->rx_multicast, &es->rx_multicast);
547 	i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
548 			   veb->stat_offsets_loaded,
549 			   &oes->rx_broadcast, &es->rx_broadcast);
550 
551 	i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
552 			   veb->stat_offsets_loaded,
553 			   &oes->tx_bytes, &es->tx_bytes);
554 	i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
555 			   veb->stat_offsets_loaded,
556 			   &oes->tx_unicast, &es->tx_unicast);
557 	i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
558 			   veb->stat_offsets_loaded,
559 			   &oes->tx_multicast, &es->tx_multicast);
560 	i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
561 			   veb->stat_offsets_loaded,
562 			   &oes->tx_broadcast, &es->tx_broadcast);
563 	veb->stat_offsets_loaded = true;
564 }
565 
566 /**
567  * i40e_update_link_xoff_rx - Update XOFF received in link flow control mode
568  * @pf: the corresponding PF
569  *
570  * Update the Rx XOFF counter (PAUSE frames) in link flow control mode
571  **/
572 static void i40e_update_link_xoff_rx(struct i40e_pf *pf)
573 {
574 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
575 	struct i40e_hw_port_stats *nsd = &pf->stats;
576 	struct i40e_hw *hw = &pf->hw;
577 	u64 xoff = 0;
578 	u16 i, v;
579 
580 	if ((hw->fc.current_mode != I40E_FC_FULL) &&
581 	    (hw->fc.current_mode != I40E_FC_RX_PAUSE))
582 		return;
583 
584 	xoff = nsd->link_xoff_rx;
585 	i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
586 			   pf->stat_offsets_loaded,
587 			   &osd->link_xoff_rx, &nsd->link_xoff_rx);
588 
589 	/* No new LFC xoff rx */
590 	if (!(nsd->link_xoff_rx - xoff))
591 		return;
592 
593 	/* Clear the __I40E_HANG_CHECK_ARMED bit for all Tx rings */
594 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
595 		struct i40e_vsi *vsi = pf->vsi[v];
596 
597 		if (!vsi)
598 			continue;
599 
600 		for (i = 0; i < vsi->num_queue_pairs; i++) {
601 			struct i40e_ring *ring = &vsi->tx_rings[i];
602 			clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
603 		}
604 	}
605 }
606 
607 /**
608  * i40e_update_prio_xoff_rx - Update XOFF received in PFC mode
609  * @pf: the corresponding PF
610  *
611  * Update the Rx XOFF counter (PAUSE frames) in PFC mode
612  **/
613 static void i40e_update_prio_xoff_rx(struct i40e_pf *pf)
614 {
615 	struct i40e_hw_port_stats *osd = &pf->stats_offsets;
616 	struct i40e_hw_port_stats *nsd = &pf->stats;
617 	bool xoff[I40E_MAX_TRAFFIC_CLASS] = {false};
618 	struct i40e_dcbx_config *dcb_cfg;
619 	struct i40e_hw *hw = &pf->hw;
620 	u16 i, v;
621 	u8 tc;
622 
623 	dcb_cfg = &hw->local_dcbx_config;
624 
625 	/* See if DCB enabled with PFC TC */
626 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED) ||
627 	    !(dcb_cfg->pfc.pfcenable)) {
628 		i40e_update_link_xoff_rx(pf);
629 		return;
630 	}
631 
632 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
633 		u64 prio_xoff = nsd->priority_xoff_rx[i];
634 		i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
635 				   pf->stat_offsets_loaded,
636 				   &osd->priority_xoff_rx[i],
637 				   &nsd->priority_xoff_rx[i]);
638 
639 		/* No new PFC xoff rx */
640 		if (!(nsd->priority_xoff_rx[i] - prio_xoff))
641 			continue;
642 		/* Get the TC for given priority */
643 		tc = dcb_cfg->etscfg.prioritytable[i];
644 		xoff[tc] = true;
645 	}
646 
647 	/* Clear the __I40E_HANG_CHECK_ARMED bit for Tx rings */
648 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
649 		struct i40e_vsi *vsi = pf->vsi[v];
650 
651 		if (!vsi)
652 			continue;
653 
654 		for (i = 0; i < vsi->num_queue_pairs; i++) {
655 			struct i40e_ring *ring = &vsi->tx_rings[i];
656 
657 			tc = ring->dcb_tc;
658 			if (xoff[tc])
659 				clear_bit(__I40E_HANG_CHECK_ARMED,
660 					  &ring->state);
661 		}
662 	}
663 }
664 
665 /**
666  * i40e_update_stats - Update the board statistics counters.
667  * @vsi: the VSI to be updated
668  *
669  * There are a few instances where we store the same stat in a
670  * couple of different structs.  This is partly because we have
671  * the netdev stats that need to be filled out, which is slightly
672  * different from the "eth_stats" defined by the chip and used in
673  * VF communications.  We sort it all out here in a central place.
674  **/
675 void i40e_update_stats(struct i40e_vsi *vsi)
676 {
677 	struct i40e_pf *pf = vsi->back;
678 	struct i40e_hw *hw = &pf->hw;
679 	struct rtnl_link_stats64 *ons;
680 	struct rtnl_link_stats64 *ns;   /* netdev stats */
681 	struct i40e_eth_stats *oes;
682 	struct i40e_eth_stats *es;     /* device's eth stats */
683 	u32 tx_restart, tx_busy;
684 	u32 rx_page, rx_buf;
685 	u64 rx_p, rx_b;
686 	u64 tx_p, tx_b;
687 	int i;
688 	u16 q;
689 
690 	if (test_bit(__I40E_DOWN, &vsi->state) ||
691 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
692 		return;
693 
694 	ns = i40e_get_vsi_stats_struct(vsi);
695 	ons = &vsi->net_stats_offsets;
696 	es = &vsi->eth_stats;
697 	oes = &vsi->eth_stats_offsets;
698 
699 	/* Gather up the netdev and vsi stats that the driver collects
700 	 * on the fly during packet processing
701 	 */
702 	rx_b = rx_p = 0;
703 	tx_b = tx_p = 0;
704 	tx_restart = tx_busy = 0;
705 	rx_page = 0;
706 	rx_buf = 0;
707 	for (q = 0; q < vsi->num_queue_pairs; q++) {
708 		struct i40e_ring *p;
709 
710 		p = &vsi->rx_rings[q];
711 		rx_b += p->rx_stats.bytes;
712 		rx_p += p->rx_stats.packets;
713 		rx_buf += p->rx_stats.alloc_rx_buff_failed;
714 		rx_page += p->rx_stats.alloc_rx_page_failed;
715 
716 		p = &vsi->tx_rings[q];
717 		tx_b += p->tx_stats.bytes;
718 		tx_p += p->tx_stats.packets;
719 		tx_restart += p->tx_stats.restart_queue;
720 		tx_busy += p->tx_stats.tx_busy;
721 	}
722 	vsi->tx_restart = tx_restart;
723 	vsi->tx_busy = tx_busy;
724 	vsi->rx_page_failed = rx_page;
725 	vsi->rx_buf_failed = rx_buf;
726 
727 	ns->rx_packets = rx_p;
728 	ns->rx_bytes = rx_b;
729 	ns->tx_packets = tx_p;
730 	ns->tx_bytes = tx_b;
731 
732 	i40e_update_eth_stats(vsi);
733 	/* update netdev stats from eth stats */
734 	ons->rx_errors = oes->rx_errors;
735 	ns->rx_errors = es->rx_errors;
736 	ons->tx_errors = oes->tx_errors;
737 	ns->tx_errors = es->tx_errors;
738 	ons->multicast = oes->rx_multicast;
739 	ns->multicast = es->rx_multicast;
740 	ons->tx_dropped = oes->tx_discards;
741 	ns->tx_dropped = es->tx_discards;
742 
743 	/* Get the port data only if this is the main PF VSI */
744 	if (vsi == pf->vsi[pf->lan_vsi]) {
745 		struct i40e_hw_port_stats *nsd = &pf->stats;
746 		struct i40e_hw_port_stats *osd = &pf->stats_offsets;
747 
748 		i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
749 				   I40E_GLPRT_GORCL(hw->port),
750 				   pf->stat_offsets_loaded,
751 				   &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
752 		i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
753 				   I40E_GLPRT_GOTCL(hw->port),
754 				   pf->stat_offsets_loaded,
755 				   &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
756 		i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
757 				   pf->stat_offsets_loaded,
758 				   &osd->eth.rx_discards,
759 				   &nsd->eth.rx_discards);
760 		i40e_stat_update32(hw, I40E_GLPRT_TDPC(hw->port),
761 				   pf->stat_offsets_loaded,
762 				   &osd->eth.tx_discards,
763 				   &nsd->eth.tx_discards);
764 		i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
765 				   I40E_GLPRT_MPRCL(hw->port),
766 				   pf->stat_offsets_loaded,
767 				   &osd->eth.rx_multicast,
768 				   &nsd->eth.rx_multicast);
769 
770 		i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
771 				   pf->stat_offsets_loaded,
772 				   &osd->tx_dropped_link_down,
773 				   &nsd->tx_dropped_link_down);
774 
775 		i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
776 				   pf->stat_offsets_loaded,
777 				   &osd->crc_errors, &nsd->crc_errors);
778 		ns->rx_crc_errors = nsd->crc_errors;
779 
780 		i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
781 				   pf->stat_offsets_loaded,
782 				   &osd->illegal_bytes, &nsd->illegal_bytes);
783 		ns->rx_errors = nsd->crc_errors
784 				+ nsd->illegal_bytes;
785 
786 		i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
787 				   pf->stat_offsets_loaded,
788 				   &osd->mac_local_faults,
789 				   &nsd->mac_local_faults);
790 		i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
791 				   pf->stat_offsets_loaded,
792 				   &osd->mac_remote_faults,
793 				   &nsd->mac_remote_faults);
794 
795 		i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
796 				   pf->stat_offsets_loaded,
797 				   &osd->rx_length_errors,
798 				   &nsd->rx_length_errors);
799 		ns->rx_length_errors = nsd->rx_length_errors;
800 
801 		i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
802 				   pf->stat_offsets_loaded,
803 				   &osd->link_xon_rx, &nsd->link_xon_rx);
804 		i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
805 				   pf->stat_offsets_loaded,
806 				   &osd->link_xon_tx, &nsd->link_xon_tx);
807 		i40e_update_prio_xoff_rx(pf);  /* handles I40E_GLPRT_LXOFFRXC */
808 		i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
809 				   pf->stat_offsets_loaded,
810 				   &osd->link_xoff_tx, &nsd->link_xoff_tx);
811 
812 		for (i = 0; i < 8; i++) {
813 			i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
814 					   pf->stat_offsets_loaded,
815 					   &osd->priority_xon_rx[i],
816 					   &nsd->priority_xon_rx[i]);
817 			i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
818 					   pf->stat_offsets_loaded,
819 					   &osd->priority_xon_tx[i],
820 					   &nsd->priority_xon_tx[i]);
821 			i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
822 					   pf->stat_offsets_loaded,
823 					   &osd->priority_xoff_tx[i],
824 					   &nsd->priority_xoff_tx[i]);
825 			i40e_stat_update32(hw,
826 					   I40E_GLPRT_RXON2OFFCNT(hw->port, i),
827 					   pf->stat_offsets_loaded,
828 					   &osd->priority_xon_2_xoff[i],
829 					   &nsd->priority_xon_2_xoff[i]);
830 		}
831 
832 		i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
833 				   I40E_GLPRT_PRC64L(hw->port),
834 				   pf->stat_offsets_loaded,
835 				   &osd->rx_size_64, &nsd->rx_size_64);
836 		i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
837 				   I40E_GLPRT_PRC127L(hw->port),
838 				   pf->stat_offsets_loaded,
839 				   &osd->rx_size_127, &nsd->rx_size_127);
840 		i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
841 				   I40E_GLPRT_PRC255L(hw->port),
842 				   pf->stat_offsets_loaded,
843 				   &osd->rx_size_255, &nsd->rx_size_255);
844 		i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
845 				   I40E_GLPRT_PRC511L(hw->port),
846 				   pf->stat_offsets_loaded,
847 				   &osd->rx_size_511, &nsd->rx_size_511);
848 		i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
849 				   I40E_GLPRT_PRC1023L(hw->port),
850 				   pf->stat_offsets_loaded,
851 				   &osd->rx_size_1023, &nsd->rx_size_1023);
852 		i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
853 				   I40E_GLPRT_PRC1522L(hw->port),
854 				   pf->stat_offsets_loaded,
855 				   &osd->rx_size_1522, &nsd->rx_size_1522);
856 		i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
857 				   I40E_GLPRT_PRC9522L(hw->port),
858 				   pf->stat_offsets_loaded,
859 				   &osd->rx_size_big, &nsd->rx_size_big);
860 
861 		i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
862 				   I40E_GLPRT_PTC64L(hw->port),
863 				   pf->stat_offsets_loaded,
864 				   &osd->tx_size_64, &nsd->tx_size_64);
865 		i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
866 				   I40E_GLPRT_PTC127L(hw->port),
867 				   pf->stat_offsets_loaded,
868 				   &osd->tx_size_127, &nsd->tx_size_127);
869 		i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
870 				   I40E_GLPRT_PTC255L(hw->port),
871 				   pf->stat_offsets_loaded,
872 				   &osd->tx_size_255, &nsd->tx_size_255);
873 		i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
874 				   I40E_GLPRT_PTC511L(hw->port),
875 				   pf->stat_offsets_loaded,
876 				   &osd->tx_size_511, &nsd->tx_size_511);
877 		i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
878 				   I40E_GLPRT_PTC1023L(hw->port),
879 				   pf->stat_offsets_loaded,
880 				   &osd->tx_size_1023, &nsd->tx_size_1023);
881 		i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
882 				   I40E_GLPRT_PTC1522L(hw->port),
883 				   pf->stat_offsets_loaded,
884 				   &osd->tx_size_1522, &nsd->tx_size_1522);
885 		i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
886 				   I40E_GLPRT_PTC9522L(hw->port),
887 				   pf->stat_offsets_loaded,
888 				   &osd->tx_size_big, &nsd->tx_size_big);
889 
890 		i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
891 				   pf->stat_offsets_loaded,
892 				   &osd->rx_undersize, &nsd->rx_undersize);
893 		i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
894 				   pf->stat_offsets_loaded,
895 				   &osd->rx_fragments, &nsd->rx_fragments);
896 		i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
897 				   pf->stat_offsets_loaded,
898 				   &osd->rx_oversize, &nsd->rx_oversize);
899 		i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
900 				   pf->stat_offsets_loaded,
901 				   &osd->rx_jabber, &nsd->rx_jabber);
902 	}
903 
904 	pf->stat_offsets_loaded = true;
905 }
906 
907 /**
908  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
909  * @vsi: the VSI to be searched
910  * @macaddr: the MAC address
911  * @vlan: the vlan
912  * @is_vf: make sure its a vf filter, else doesn't matter
913  * @is_netdev: make sure its a netdev filter, else doesn't matter
914  *
915  * Returns ptr to the filter object or NULL
916  **/
917 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
918 						u8 *macaddr, s16 vlan,
919 						bool is_vf, bool is_netdev)
920 {
921 	struct i40e_mac_filter *f;
922 
923 	if (!vsi || !macaddr)
924 		return NULL;
925 
926 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
927 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
928 		    (vlan == f->vlan)    &&
929 		    (!is_vf || f->is_vf) &&
930 		    (!is_netdev || f->is_netdev))
931 			return f;
932 	}
933 	return NULL;
934 }
935 
936 /**
937  * i40e_find_mac - Find a mac addr in the macvlan filters list
938  * @vsi: the VSI to be searched
939  * @macaddr: the MAC address we are searching for
940  * @is_vf: make sure its a vf filter, else doesn't matter
941  * @is_netdev: make sure its a netdev filter, else doesn't matter
942  *
943  * Returns the first filter with the provided MAC address or NULL if
944  * MAC address was not found
945  **/
946 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
947 				      bool is_vf, bool is_netdev)
948 {
949 	struct i40e_mac_filter *f;
950 
951 	if (!vsi || !macaddr)
952 		return NULL;
953 
954 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
955 		if ((ether_addr_equal(macaddr, f->macaddr)) &&
956 		    (!is_vf || f->is_vf) &&
957 		    (!is_netdev || f->is_netdev))
958 			return f;
959 	}
960 	return NULL;
961 }
962 
963 /**
964  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
965  * @vsi: the VSI to be searched
966  *
967  * Returns true if VSI is in vlan mode or false otherwise
968  **/
969 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
970 {
971 	struct i40e_mac_filter *f;
972 
973 	/* Only -1 for all the filters denotes not in vlan mode
974 	 * so we have to go through all the list in order to make sure
975 	 */
976 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
977 		if (f->vlan >= 0)
978 			return true;
979 	}
980 
981 	return false;
982 }
983 
984 /**
985  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
986  * @vsi: the VSI to be searched
987  * @macaddr: the mac address to be filtered
988  * @is_vf: true if it is a vf
989  * @is_netdev: true if it is a netdev
990  *
991  * Goes through all the macvlan filters and adds a
992  * macvlan filter for each unique vlan that already exists
993  *
994  * Returns first filter found on success, else NULL
995  **/
996 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
997 					     bool is_vf, bool is_netdev)
998 {
999 	struct i40e_mac_filter *f;
1000 
1001 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1002 		if (!i40e_find_filter(vsi, macaddr, f->vlan,
1003 				      is_vf, is_netdev)) {
1004 			if (!i40e_add_filter(vsi, macaddr, f->vlan,
1005 						is_vf, is_netdev))
1006 				return NULL;
1007 		}
1008 	}
1009 
1010 	return list_first_entry_or_null(&vsi->mac_filter_list,
1011 					struct i40e_mac_filter, list);
1012 }
1013 
1014 /**
1015  * i40e_add_filter - Add a mac/vlan filter to the VSI
1016  * @vsi: the VSI to be searched
1017  * @macaddr: the MAC address
1018  * @vlan: the vlan
1019  * @is_vf: make sure its a vf filter, else doesn't matter
1020  * @is_netdev: make sure its a netdev filter, else doesn't matter
1021  *
1022  * Returns ptr to the filter object or NULL when no memory available.
1023  **/
1024 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1025 					u8 *macaddr, s16 vlan,
1026 					bool is_vf, bool is_netdev)
1027 {
1028 	struct i40e_mac_filter *f;
1029 
1030 	if (!vsi || !macaddr)
1031 		return NULL;
1032 
1033 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1034 	if (!f) {
1035 		f = kzalloc(sizeof(*f), GFP_ATOMIC);
1036 		if (!f)
1037 			goto add_filter_out;
1038 
1039 		memcpy(f->macaddr, macaddr, ETH_ALEN);
1040 		f->vlan = vlan;
1041 		f->changed = true;
1042 
1043 		INIT_LIST_HEAD(&f->list);
1044 		list_add(&f->list, &vsi->mac_filter_list);
1045 	}
1046 
1047 	/* increment counter and add a new flag if needed */
1048 	if (is_vf) {
1049 		if (!f->is_vf) {
1050 			f->is_vf = true;
1051 			f->counter++;
1052 		}
1053 	} else if (is_netdev) {
1054 		if (!f->is_netdev) {
1055 			f->is_netdev = true;
1056 			f->counter++;
1057 		}
1058 	} else {
1059 		f->counter++;
1060 	}
1061 
1062 	/* changed tells sync_filters_subtask to
1063 	 * push the filter down to the firmware
1064 	 */
1065 	if (f->changed) {
1066 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1067 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1068 	}
1069 
1070 add_filter_out:
1071 	return f;
1072 }
1073 
1074 /**
1075  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1076  * @vsi: the VSI to be searched
1077  * @macaddr: the MAC address
1078  * @vlan: the vlan
1079  * @is_vf: make sure it's a vf filter, else doesn't matter
1080  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1081  **/
1082 void i40e_del_filter(struct i40e_vsi *vsi,
1083 		     u8 *macaddr, s16 vlan,
1084 		     bool is_vf, bool is_netdev)
1085 {
1086 	struct i40e_mac_filter *f;
1087 
1088 	if (!vsi || !macaddr)
1089 		return;
1090 
1091 	f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1092 	if (!f || f->counter == 0)
1093 		return;
1094 
1095 	if (is_vf) {
1096 		if (f->is_vf) {
1097 			f->is_vf = false;
1098 			f->counter--;
1099 		}
1100 	} else if (is_netdev) {
1101 		if (f->is_netdev) {
1102 			f->is_netdev = false;
1103 			f->counter--;
1104 		}
1105 	} else {
1106 		/* make sure we don't remove a filter in use by vf or netdev */
1107 		int min_f = 0;
1108 		min_f += (f->is_vf ? 1 : 0);
1109 		min_f += (f->is_netdev ? 1 : 0);
1110 
1111 		if (f->counter > min_f)
1112 			f->counter--;
1113 	}
1114 
1115 	/* counter == 0 tells sync_filters_subtask to
1116 	 * remove the filter from the firmware's list
1117 	 */
1118 	if (f->counter == 0) {
1119 		f->changed = true;
1120 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1121 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1122 	}
1123 }
1124 
1125 /**
1126  * i40e_set_mac - NDO callback to set mac address
1127  * @netdev: network interface device structure
1128  * @p: pointer to an address structure
1129  *
1130  * Returns 0 on success, negative on failure
1131  **/
1132 static int i40e_set_mac(struct net_device *netdev, void *p)
1133 {
1134 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1135 	struct i40e_vsi *vsi = np->vsi;
1136 	struct sockaddr *addr = p;
1137 	struct i40e_mac_filter *f;
1138 
1139 	if (!is_valid_ether_addr(addr->sa_data))
1140 		return -EADDRNOTAVAIL;
1141 
1142 	netdev_info(netdev, "set mac address=%pM\n", addr->sa_data);
1143 
1144 	if (ether_addr_equal(netdev->dev_addr, addr->sa_data))
1145 		return 0;
1146 
1147 	if (vsi->type == I40E_VSI_MAIN) {
1148 		i40e_status ret;
1149 		ret = i40e_aq_mac_address_write(&vsi->back->hw,
1150 						I40E_AQC_WRITE_TYPE_LAA_ONLY,
1151 						addr->sa_data, NULL);
1152 		if (ret) {
1153 			netdev_info(netdev,
1154 				    "Addr change for Main VSI failed: %d\n",
1155 				    ret);
1156 			return -EADDRNOTAVAIL;
1157 		}
1158 
1159 		memcpy(vsi->back->hw.mac.addr, addr->sa_data, netdev->addr_len);
1160 	}
1161 
1162 	/* In order to be sure to not drop any packets, add the new address
1163 	 * then delete the old one.
1164 	 */
1165 	f = i40e_add_filter(vsi, addr->sa_data, I40E_VLAN_ANY, false, false);
1166 	if (!f)
1167 		return -ENOMEM;
1168 
1169 	i40e_sync_vsi_filters(vsi);
1170 	i40e_del_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY, false, false);
1171 	i40e_sync_vsi_filters(vsi);
1172 
1173 	memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
1174 
1175 	return 0;
1176 }
1177 
1178 /**
1179  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1180  * @vsi: the VSI being setup
1181  * @ctxt: VSI context structure
1182  * @enabled_tc: Enabled TCs bitmap
1183  * @is_add: True if called before Add VSI
1184  *
1185  * Setup VSI queue mapping for enabled traffic classes.
1186  **/
1187 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1188 				     struct i40e_vsi_context *ctxt,
1189 				     u8 enabled_tc,
1190 				     bool is_add)
1191 {
1192 	struct i40e_pf *pf = vsi->back;
1193 	u16 sections = 0;
1194 	u8 netdev_tc = 0;
1195 	u16 numtc = 0;
1196 	u16 qcount;
1197 	u8 offset;
1198 	u16 qmap;
1199 	int i;
1200 
1201 	sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1202 	offset = 0;
1203 
1204 	if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1205 		/* Find numtc from enabled TC bitmap */
1206 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1207 			if (enabled_tc & (1 << i)) /* TC is enabled */
1208 				numtc++;
1209 		}
1210 		if (!numtc) {
1211 			dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1212 			numtc = 1;
1213 		}
1214 	} else {
1215 		/* At least TC0 is enabled in case of non-DCB case */
1216 		numtc = 1;
1217 	}
1218 
1219 	vsi->tc_config.numtc = numtc;
1220 	vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1221 
1222 	/* Setup queue offset/count for all TCs for given VSI */
1223 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1224 		/* See if the given TC is enabled for the given VSI */
1225 		if (vsi->tc_config.enabled_tc & (1 << i)) { /* TC is enabled */
1226 			int pow, num_qps;
1227 
1228 			vsi->tc_config.tc_info[i].qoffset = offset;
1229 			switch (vsi->type) {
1230 			case I40E_VSI_MAIN:
1231 				if (i == 0)
1232 					qcount = pf->rss_size;
1233 				else
1234 					qcount = pf->num_tc_qps;
1235 				vsi->tc_config.tc_info[i].qcount = qcount;
1236 				break;
1237 			case I40E_VSI_FDIR:
1238 			case I40E_VSI_SRIOV:
1239 			case I40E_VSI_VMDQ2:
1240 			default:
1241 				qcount = vsi->alloc_queue_pairs;
1242 				vsi->tc_config.tc_info[i].qcount = qcount;
1243 				WARN_ON(i != 0);
1244 				break;
1245 			}
1246 
1247 			/* find the power-of-2 of the number of queue pairs */
1248 			num_qps = vsi->tc_config.tc_info[i].qcount;
1249 			pow = 0;
1250 			while (num_qps &&
1251 			      ((1 << pow) < vsi->tc_config.tc_info[i].qcount)) {
1252 				pow++;
1253 				num_qps >>= 1;
1254 			}
1255 
1256 			vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1257 			qmap =
1258 			    (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1259 			    (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1260 
1261 			offset += vsi->tc_config.tc_info[i].qcount;
1262 		} else {
1263 			/* TC is not enabled so set the offset to
1264 			 * default queue and allocate one queue
1265 			 * for the given TC.
1266 			 */
1267 			vsi->tc_config.tc_info[i].qoffset = 0;
1268 			vsi->tc_config.tc_info[i].qcount = 1;
1269 			vsi->tc_config.tc_info[i].netdev_tc = 0;
1270 
1271 			qmap = 0;
1272 		}
1273 		ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1274 	}
1275 
1276 	/* Set actual Tx/Rx queue pairs */
1277 	vsi->num_queue_pairs = offset;
1278 
1279 	/* Scheduler section valid can only be set for ADD VSI */
1280 	if (is_add) {
1281 		sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1282 
1283 		ctxt->info.up_enable_bits = enabled_tc;
1284 	}
1285 	if (vsi->type == I40E_VSI_SRIOV) {
1286 		ctxt->info.mapping_flags |=
1287 				     cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1288 		for (i = 0; i < vsi->num_queue_pairs; i++)
1289 			ctxt->info.queue_mapping[i] =
1290 					       cpu_to_le16(vsi->base_queue + i);
1291 	} else {
1292 		ctxt->info.mapping_flags |=
1293 					cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1294 		ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1295 	}
1296 	ctxt->info.valid_sections |= cpu_to_le16(sections);
1297 }
1298 
1299 /**
1300  * i40e_set_rx_mode - NDO callback to set the netdev filters
1301  * @netdev: network interface device structure
1302  **/
1303 static void i40e_set_rx_mode(struct net_device *netdev)
1304 {
1305 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1306 	struct i40e_mac_filter *f, *ftmp;
1307 	struct i40e_vsi *vsi = np->vsi;
1308 	struct netdev_hw_addr *uca;
1309 	struct netdev_hw_addr *mca;
1310 	struct netdev_hw_addr *ha;
1311 
1312 	/* add addr if not already in the filter list */
1313 	netdev_for_each_uc_addr(uca, netdev) {
1314 		if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1315 			if (i40e_is_vsi_in_vlan(vsi))
1316 				i40e_put_mac_in_vlan(vsi, uca->addr,
1317 						     false, true);
1318 			else
1319 				i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1320 						false, true);
1321 		}
1322 	}
1323 
1324 	netdev_for_each_mc_addr(mca, netdev) {
1325 		if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1326 			if (i40e_is_vsi_in_vlan(vsi))
1327 				i40e_put_mac_in_vlan(vsi, mca->addr,
1328 						     false, true);
1329 			else
1330 				i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1331 						false, true);
1332 		}
1333 	}
1334 
1335 	/* remove filter if not in netdev list */
1336 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1337 		bool found = false;
1338 
1339 		if (!f->is_netdev)
1340 			continue;
1341 
1342 		if (is_multicast_ether_addr(f->macaddr)) {
1343 			netdev_for_each_mc_addr(mca, netdev) {
1344 				if (ether_addr_equal(mca->addr, f->macaddr)) {
1345 					found = true;
1346 					break;
1347 				}
1348 			}
1349 		} else {
1350 			netdev_for_each_uc_addr(uca, netdev) {
1351 				if (ether_addr_equal(uca->addr, f->macaddr)) {
1352 					found = true;
1353 					break;
1354 				}
1355 			}
1356 
1357 			for_each_dev_addr(netdev, ha) {
1358 				if (ether_addr_equal(ha->addr, f->macaddr)) {
1359 					found = true;
1360 					break;
1361 				}
1362 			}
1363 		}
1364 		if (!found)
1365 			i40e_del_filter(
1366 			   vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1367 	}
1368 
1369 	/* check for other flag changes */
1370 	if (vsi->current_netdev_flags != vsi->netdev->flags) {
1371 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1372 		vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1373 	}
1374 }
1375 
1376 /**
1377  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1378  * @vsi: ptr to the VSI
1379  *
1380  * Push any outstanding VSI filter changes through the AdminQ.
1381  *
1382  * Returns 0 or error value
1383  **/
1384 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1385 {
1386 	struct i40e_mac_filter *f, *ftmp;
1387 	bool promisc_forced_on = false;
1388 	bool add_happened = false;
1389 	int filter_list_len = 0;
1390 	u32 changed_flags = 0;
1391 	i40e_status ret = 0;
1392 	struct i40e_pf *pf;
1393 	int num_add = 0;
1394 	int num_del = 0;
1395 	u16 cmd_flags;
1396 
1397 	/* empty array typed pointers, kcalloc later */
1398 	struct i40e_aqc_add_macvlan_element_data *add_list;
1399 	struct i40e_aqc_remove_macvlan_element_data *del_list;
1400 
1401 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1402 		usleep_range(1000, 2000);
1403 	pf = vsi->back;
1404 
1405 	if (vsi->netdev) {
1406 		changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1407 		vsi->current_netdev_flags = vsi->netdev->flags;
1408 	}
1409 
1410 	if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1411 		vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1412 
1413 		filter_list_len = pf->hw.aq.asq_buf_size /
1414 			    sizeof(struct i40e_aqc_remove_macvlan_element_data);
1415 		del_list = kcalloc(filter_list_len,
1416 			    sizeof(struct i40e_aqc_remove_macvlan_element_data),
1417 			    GFP_KERNEL);
1418 		if (!del_list)
1419 			return -ENOMEM;
1420 
1421 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1422 			if (!f->changed)
1423 				continue;
1424 
1425 			if (f->counter != 0)
1426 				continue;
1427 			f->changed = false;
1428 			cmd_flags = 0;
1429 
1430 			/* add to delete list */
1431 			memcpy(del_list[num_del].mac_addr,
1432 			       f->macaddr, ETH_ALEN);
1433 			del_list[num_del].vlan_tag =
1434 				cpu_to_le16((u16)(f->vlan ==
1435 					    I40E_VLAN_ANY ? 0 : f->vlan));
1436 
1437 			/* vlan0 as wild card to allow packets from all vlans */
1438 			if (f->vlan == I40E_VLAN_ANY ||
1439 			    (vsi->netdev && !(vsi->netdev->features &
1440 					      NETIF_F_HW_VLAN_CTAG_FILTER)))
1441 				cmd_flags |= I40E_AQC_MACVLAN_DEL_IGNORE_VLAN;
1442 			cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1443 			del_list[num_del].flags = cmd_flags;
1444 			num_del++;
1445 
1446 			/* unlink from filter list */
1447 			list_del(&f->list);
1448 			kfree(f);
1449 
1450 			/* flush a full buffer */
1451 			if (num_del == filter_list_len) {
1452 				ret = i40e_aq_remove_macvlan(&pf->hw,
1453 					    vsi->seid, del_list, num_del,
1454 					    NULL);
1455 				num_del = 0;
1456 				memset(del_list, 0, sizeof(*del_list));
1457 
1458 				if (ret)
1459 					dev_info(&pf->pdev->dev,
1460 						 "ignoring delete macvlan error, err %d, aq_err %d while flushing a full buffer\n",
1461 						 ret,
1462 						 pf->hw.aq.asq_last_status);
1463 			}
1464 		}
1465 		if (num_del) {
1466 			ret = i40e_aq_remove_macvlan(&pf->hw, vsi->seid,
1467 						     del_list, num_del, NULL);
1468 			num_del = 0;
1469 
1470 			if (ret)
1471 				dev_info(&pf->pdev->dev,
1472 					 "ignoring delete macvlan error, err %d, aq_err %d\n",
1473 					 ret, pf->hw.aq.asq_last_status);
1474 		}
1475 
1476 		kfree(del_list);
1477 		del_list = NULL;
1478 
1479 		/* do all the adds now */
1480 		filter_list_len = pf->hw.aq.asq_buf_size /
1481 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1482 		add_list = kcalloc(filter_list_len,
1483 			       sizeof(struct i40e_aqc_add_macvlan_element_data),
1484 			       GFP_KERNEL);
1485 		if (!add_list)
1486 			return -ENOMEM;
1487 
1488 		list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1489 			if (!f->changed)
1490 				continue;
1491 
1492 			if (f->counter == 0)
1493 				continue;
1494 			f->changed = false;
1495 			add_happened = true;
1496 			cmd_flags = 0;
1497 
1498 			/* add to add array */
1499 			memcpy(add_list[num_add].mac_addr,
1500 			       f->macaddr, ETH_ALEN);
1501 			add_list[num_add].vlan_tag =
1502 				cpu_to_le16(
1503 				 (u16)(f->vlan == I40E_VLAN_ANY ? 0 : f->vlan));
1504 			add_list[num_add].queue_number = 0;
1505 
1506 			cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
1507 
1508 			/* vlan0 as wild card to allow packets from all vlans */
1509 			if (f->vlan == I40E_VLAN_ANY || (vsi->netdev &&
1510 			    !(vsi->netdev->features &
1511 						 NETIF_F_HW_VLAN_CTAG_FILTER)))
1512 				cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1513 			add_list[num_add].flags = cpu_to_le16(cmd_flags);
1514 			num_add++;
1515 
1516 			/* flush a full buffer */
1517 			if (num_add == filter_list_len) {
1518 				ret = i40e_aq_add_macvlan(&pf->hw,
1519 							  vsi->seid,
1520 							  add_list,
1521 							  num_add,
1522 							  NULL);
1523 				num_add = 0;
1524 
1525 				if (ret)
1526 					break;
1527 				memset(add_list, 0, sizeof(*add_list));
1528 			}
1529 		}
1530 		if (num_add) {
1531 			ret = i40e_aq_add_macvlan(&pf->hw, vsi->seid,
1532 						  add_list, num_add, NULL);
1533 			num_add = 0;
1534 		}
1535 		kfree(add_list);
1536 		add_list = NULL;
1537 
1538 		if (add_happened && (!ret)) {
1539 			/* do nothing */;
1540 		} else if (add_happened && (ret)) {
1541 			dev_info(&pf->pdev->dev,
1542 				 "add filter failed, err %d, aq_err %d\n",
1543 				 ret, pf->hw.aq.asq_last_status);
1544 			if ((pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOSPC) &&
1545 			    !test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1546 				      &vsi->state)) {
1547 				promisc_forced_on = true;
1548 				set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1549 					&vsi->state);
1550 				dev_info(&pf->pdev->dev, "promiscuous mode forced on\n");
1551 			}
1552 		}
1553 	}
1554 
1555 	/* check for changes in promiscuous modes */
1556 	if (changed_flags & IFF_ALLMULTI) {
1557 		bool cur_multipromisc;
1558 		cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
1559 		ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
1560 							    vsi->seid,
1561 							    cur_multipromisc,
1562 							    NULL);
1563 		if (ret)
1564 			dev_info(&pf->pdev->dev,
1565 				 "set multi promisc failed, err %d, aq_err %d\n",
1566 				 ret, pf->hw.aq.asq_last_status);
1567 	}
1568 	if ((changed_flags & IFF_PROMISC) || promisc_forced_on) {
1569 		bool cur_promisc;
1570 		cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
1571 			       test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1572 					&vsi->state));
1573 		ret = i40e_aq_set_vsi_unicast_promiscuous(&vsi->back->hw,
1574 							  vsi->seid,
1575 							  cur_promisc,
1576 							  NULL);
1577 		if (ret)
1578 			dev_info(&pf->pdev->dev,
1579 				 "set uni promisc failed, err %d, aq_err %d\n",
1580 				 ret, pf->hw.aq.asq_last_status);
1581 	}
1582 
1583 	clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
1584 	return 0;
1585 }
1586 
1587 /**
1588  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
1589  * @pf: board private structure
1590  **/
1591 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
1592 {
1593 	int v;
1594 
1595 	if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
1596 		return;
1597 	pf->flags &= ~I40E_FLAG_FILTER_SYNC;
1598 
1599 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
1600 		if (pf->vsi[v] &&
1601 		    (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED))
1602 			i40e_sync_vsi_filters(pf->vsi[v]);
1603 	}
1604 }
1605 
1606 /**
1607  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
1608  * @netdev: network interface device structure
1609  * @new_mtu: new value for maximum frame size
1610  *
1611  * Returns 0 on success, negative on failure
1612  **/
1613 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
1614 {
1615 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1616 	int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN;
1617 	struct i40e_vsi *vsi = np->vsi;
1618 
1619 	/* MTU < 68 is an error and causes problems on some kernels */
1620 	if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
1621 		return -EINVAL;
1622 
1623 	netdev_info(netdev, "changing MTU from %d to %d\n",
1624 		    netdev->mtu, new_mtu);
1625 	netdev->mtu = new_mtu;
1626 	if (netif_running(netdev))
1627 		i40e_vsi_reinit_locked(vsi);
1628 
1629 	return 0;
1630 }
1631 
1632 /**
1633  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
1634  * @vsi: the vsi being adjusted
1635  **/
1636 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
1637 {
1638 	struct i40e_vsi_context ctxt;
1639 	i40e_status ret;
1640 
1641 	if ((vsi->info.valid_sections &
1642 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1643 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
1644 		return;  /* already enabled */
1645 
1646 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1647 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1648 				    I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
1649 
1650 	ctxt.seid = vsi->seid;
1651 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1652 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1653 	if (ret) {
1654 		dev_info(&vsi->back->pdev->dev,
1655 			 "%s: update vsi failed, aq_err=%d\n",
1656 			 __func__, vsi->back->hw.aq.asq_last_status);
1657 	}
1658 }
1659 
1660 /**
1661  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
1662  * @vsi: the vsi being adjusted
1663  **/
1664 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
1665 {
1666 	struct i40e_vsi_context ctxt;
1667 	i40e_status ret;
1668 
1669 	if ((vsi->info.valid_sections &
1670 	     cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
1671 	    ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
1672 	     I40E_AQ_VSI_PVLAN_EMOD_MASK))
1673 		return;  /* already disabled */
1674 
1675 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1676 	vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
1677 				    I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
1678 
1679 	ctxt.seid = vsi->seid;
1680 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1681 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1682 	if (ret) {
1683 		dev_info(&vsi->back->pdev->dev,
1684 			 "%s: update vsi failed, aq_err=%d\n",
1685 			 __func__, vsi->back->hw.aq.asq_last_status);
1686 	}
1687 }
1688 
1689 /**
1690  * i40e_vlan_rx_register - Setup or shutdown vlan offload
1691  * @netdev: network interface to be adjusted
1692  * @features: netdev features to test if VLAN offload is enabled or not
1693  **/
1694 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
1695 {
1696 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1697 	struct i40e_vsi *vsi = np->vsi;
1698 
1699 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
1700 		i40e_vlan_stripping_enable(vsi);
1701 	else
1702 		i40e_vlan_stripping_disable(vsi);
1703 }
1704 
1705 /**
1706  * i40e_vsi_add_vlan - Add vsi membership for given vlan
1707  * @vsi: the vsi being configured
1708  * @vid: vlan id to be added (0 = untagged only , -1 = any)
1709  **/
1710 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
1711 {
1712 	struct i40e_mac_filter *f, *add_f;
1713 	bool is_netdev, is_vf;
1714 	int ret;
1715 
1716 	is_vf = (vsi->type == I40E_VSI_SRIOV);
1717 	is_netdev = !!(vsi->netdev);
1718 
1719 	if (is_netdev) {
1720 		add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
1721 					is_vf, is_netdev);
1722 		if (!add_f) {
1723 			dev_info(&vsi->back->pdev->dev,
1724 				 "Could not add vlan filter %d for %pM\n",
1725 				 vid, vsi->netdev->dev_addr);
1726 			return -ENOMEM;
1727 		}
1728 	}
1729 
1730 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1731 		add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1732 		if (!add_f) {
1733 			dev_info(&vsi->back->pdev->dev,
1734 				 "Could not add vlan filter %d for %pM\n",
1735 				 vid, f->macaddr);
1736 			return -ENOMEM;
1737 		}
1738 	}
1739 
1740 	ret = i40e_sync_vsi_filters(vsi);
1741 	if (ret) {
1742 		dev_info(&vsi->back->pdev->dev,
1743 			 "Could not sync filters for vid %d\n", vid);
1744 		return ret;
1745 	}
1746 
1747 	/* Now if we add a vlan tag, make sure to check if it is the first
1748 	 * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
1749 	 * with 0, so we now accept untagged and specified tagged traffic
1750 	 * (and not any taged and untagged)
1751 	 */
1752 	if (vid > 0) {
1753 		if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
1754 						  I40E_VLAN_ANY,
1755 						  is_vf, is_netdev)) {
1756 			i40e_del_filter(vsi, vsi->netdev->dev_addr,
1757 					I40E_VLAN_ANY, is_vf, is_netdev);
1758 			add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
1759 						is_vf, is_netdev);
1760 			if (!add_f) {
1761 				dev_info(&vsi->back->pdev->dev,
1762 					 "Could not add filter 0 for %pM\n",
1763 					 vsi->netdev->dev_addr);
1764 				return -ENOMEM;
1765 			}
1766 		}
1767 
1768 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
1769 			if (i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1770 					     is_vf, is_netdev)) {
1771 				i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1772 						is_vf, is_netdev);
1773 				add_f = i40e_add_filter(vsi, f->macaddr,
1774 							0, is_vf, is_netdev);
1775 				if (!add_f) {
1776 					dev_info(&vsi->back->pdev->dev,
1777 						 "Could not add filter 0 for %pM\n",
1778 						 f->macaddr);
1779 					return -ENOMEM;
1780 				}
1781 			}
1782 		}
1783 		ret = i40e_sync_vsi_filters(vsi);
1784 	}
1785 
1786 	return ret;
1787 }
1788 
1789 /**
1790  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
1791  * @vsi: the vsi being configured
1792  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
1793  **/
1794 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
1795 {
1796 	struct net_device *netdev = vsi->netdev;
1797 	struct i40e_mac_filter *f, *add_f;
1798 	bool is_vf, is_netdev;
1799 	int filter_count = 0;
1800 	int ret;
1801 
1802 	is_vf = (vsi->type == I40E_VSI_SRIOV);
1803 	is_netdev = !!(netdev);
1804 
1805 	if (is_netdev)
1806 		i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
1807 
1808 	list_for_each_entry(f, &vsi->mac_filter_list, list)
1809 		i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
1810 
1811 	ret = i40e_sync_vsi_filters(vsi);
1812 	if (ret) {
1813 		dev_info(&vsi->back->pdev->dev, "Could not sync filters\n");
1814 		return ret;
1815 	}
1816 
1817 	/* go through all the filters for this VSI and if there is only
1818 	 * vid == 0 it means there are no other filters, so vid 0 must
1819 	 * be replaced with -1. This signifies that we should from now
1820 	 * on accept any traffic (with any tag present, or untagged)
1821 	 */
1822 	list_for_each_entry(f, &vsi->mac_filter_list, list) {
1823 		if (is_netdev) {
1824 			if (f->vlan &&
1825 			    ether_addr_equal(netdev->dev_addr, f->macaddr))
1826 				filter_count++;
1827 		}
1828 
1829 		if (f->vlan)
1830 			filter_count++;
1831 	}
1832 
1833 	if (!filter_count && is_netdev) {
1834 		i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
1835 		f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
1836 				    is_vf, is_netdev);
1837 		if (!f) {
1838 			dev_info(&vsi->back->pdev->dev,
1839 				 "Could not add filter %d for %pM\n",
1840 				 I40E_VLAN_ANY, netdev->dev_addr);
1841 			return -ENOMEM;
1842 		}
1843 	}
1844 
1845 	if (!filter_count) {
1846 		list_for_each_entry(f, &vsi->mac_filter_list, list) {
1847 			i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
1848 			add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
1849 					    is_vf, is_netdev);
1850 			if (!add_f) {
1851 				dev_info(&vsi->back->pdev->dev,
1852 					 "Could not add filter %d for %pM\n",
1853 					 I40E_VLAN_ANY, f->macaddr);
1854 				return -ENOMEM;
1855 			}
1856 		}
1857 	}
1858 
1859 	return i40e_sync_vsi_filters(vsi);
1860 }
1861 
1862 /**
1863  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
1864  * @netdev: network interface to be adjusted
1865  * @vid: vlan id to be added
1866  **/
1867 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
1868 				__always_unused __be16 proto, u16 vid)
1869 {
1870 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1871 	struct i40e_vsi *vsi = np->vsi;
1872 	int ret;
1873 
1874 	if (vid > 4095)
1875 		return 0;
1876 
1877 	netdev_info(vsi->netdev, "adding %pM vid=%d\n",
1878 		    netdev->dev_addr, vid);
1879 	/* If the network stack called us with vid = 0, we should
1880 	 * indicate to i40e_vsi_add_vlan() that we want to receive
1881 	 * any traffic (i.e. with any vlan tag, or untagged)
1882 	 */
1883 	ret = i40e_vsi_add_vlan(vsi, vid ? vid : I40E_VLAN_ANY);
1884 
1885 	if (!ret) {
1886 		if (vid < VLAN_N_VID)
1887 			set_bit(vid, vsi->active_vlans);
1888 	}
1889 
1890 	return 0;
1891 }
1892 
1893 /**
1894  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
1895  * @netdev: network interface to be adjusted
1896  * @vid: vlan id to be removed
1897  **/
1898 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
1899 				 __always_unused __be16 proto, u16 vid)
1900 {
1901 	struct i40e_netdev_priv *np = netdev_priv(netdev);
1902 	struct i40e_vsi *vsi = np->vsi;
1903 
1904 	netdev_info(vsi->netdev, "removing %pM vid=%d\n",
1905 		    netdev->dev_addr, vid);
1906 	/* return code is ignored as there is nothing a user
1907 	 * can do about failure to remove and a log message was
1908 	 * already printed from another function
1909 	 */
1910 	i40e_vsi_kill_vlan(vsi, vid);
1911 
1912 	clear_bit(vid, vsi->active_vlans);
1913 	return 0;
1914 }
1915 
1916 /**
1917  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
1918  * @vsi: the vsi being brought back up
1919  **/
1920 static void i40e_restore_vlan(struct i40e_vsi *vsi)
1921 {
1922 	u16 vid;
1923 
1924 	if (!vsi->netdev)
1925 		return;
1926 
1927 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
1928 
1929 	for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
1930 		i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
1931 				     vid);
1932 }
1933 
1934 /**
1935  * i40e_vsi_add_pvid - Add pvid for the VSI
1936  * @vsi: the vsi being adjusted
1937  * @vid: the vlan id to set as a PVID
1938  **/
1939 i40e_status i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
1940 {
1941 	struct i40e_vsi_context ctxt;
1942 	i40e_status ret;
1943 
1944 	vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
1945 	vsi->info.pvid = cpu_to_le16(vid);
1946 	vsi->info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_INSERT_PVID;
1947 	vsi->info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_UNTAGGED;
1948 
1949 	ctxt.seid = vsi->seid;
1950 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
1951 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
1952 	if (ret) {
1953 		dev_info(&vsi->back->pdev->dev,
1954 			 "%s: update vsi failed, aq_err=%d\n",
1955 			 __func__, vsi->back->hw.aq.asq_last_status);
1956 	}
1957 
1958 	return ret;
1959 }
1960 
1961 /**
1962  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
1963  * @vsi: the vsi being adjusted
1964  *
1965  * Just use the vlan_rx_register() service to put it back to normal
1966  **/
1967 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
1968 {
1969 	vsi->info.pvid = 0;
1970 	i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
1971 }
1972 
1973 /**
1974  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
1975  * @vsi: ptr to the VSI
1976  *
1977  * If this function returns with an error, then it's possible one or
1978  * more of the rings is populated (while the rest are not).  It is the
1979  * callers duty to clean those orphaned rings.
1980  *
1981  * Return 0 on success, negative on failure
1982  **/
1983 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
1984 {
1985 	int i, err = 0;
1986 
1987 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
1988 		err = i40e_setup_tx_descriptors(&vsi->tx_rings[i]);
1989 
1990 	return err;
1991 }
1992 
1993 /**
1994  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
1995  * @vsi: ptr to the VSI
1996  *
1997  * Free VSI's transmit software resources
1998  **/
1999 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2000 {
2001 	int i;
2002 
2003 	for (i = 0; i < vsi->num_queue_pairs; i++)
2004 		if (vsi->tx_rings[i].desc)
2005 			i40e_free_tx_resources(&vsi->tx_rings[i]);
2006 }
2007 
2008 /**
2009  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2010  * @vsi: ptr to the VSI
2011  *
2012  * If this function returns with an error, then it's possible one or
2013  * more of the rings is populated (while the rest are not).  It is the
2014  * callers duty to clean those orphaned rings.
2015  *
2016  * Return 0 on success, negative on failure
2017  **/
2018 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2019 {
2020 	int i, err = 0;
2021 
2022 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2023 		err = i40e_setup_rx_descriptors(&vsi->rx_rings[i]);
2024 	return err;
2025 }
2026 
2027 /**
2028  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2029  * @vsi: ptr to the VSI
2030  *
2031  * Free all receive software resources
2032  **/
2033 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2034 {
2035 	int i;
2036 
2037 	for (i = 0; i < vsi->num_queue_pairs; i++)
2038 		if (vsi->rx_rings[i].desc)
2039 			i40e_free_rx_resources(&vsi->rx_rings[i]);
2040 }
2041 
2042 /**
2043  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2044  * @ring: The Tx ring to configure
2045  *
2046  * Configure the Tx descriptor ring in the HMC context.
2047  **/
2048 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2049 {
2050 	struct i40e_vsi *vsi = ring->vsi;
2051 	u16 pf_q = vsi->base_queue + ring->queue_index;
2052 	struct i40e_hw *hw = &vsi->back->hw;
2053 	struct i40e_hmc_obj_txq tx_ctx;
2054 	i40e_status err = 0;
2055 	u32 qtx_ctl = 0;
2056 
2057 	/* some ATR related tx ring init */
2058 	if (vsi->back->flags & I40E_FLAG_FDIR_ATR_ENABLED) {
2059 		ring->atr_sample_rate = vsi->back->atr_sample_rate;
2060 		ring->atr_count = 0;
2061 	} else {
2062 		ring->atr_sample_rate = 0;
2063 	}
2064 
2065 	/* initialize XPS */
2066 	if (ring->q_vector && ring->netdev &&
2067 	    !test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2068 		netif_set_xps_queue(ring->netdev,
2069 				    &ring->q_vector->affinity_mask,
2070 				    ring->queue_index);
2071 
2072 	/* clear the context structure first */
2073 	memset(&tx_ctx, 0, sizeof(tx_ctx));
2074 
2075 	tx_ctx.new_context = 1;
2076 	tx_ctx.base = (ring->dma / 128);
2077 	tx_ctx.qlen = ring->count;
2078 	tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FDIR_ENABLED |
2079 			I40E_FLAG_FDIR_ATR_ENABLED));
2080 
2081 	/* As part of VSI creation/update, FW allocates certain
2082 	 * Tx arbitration queue sets for each TC enabled for
2083 	 * the VSI. The FW returns the handles to these queue
2084 	 * sets as part of the response buffer to Add VSI,
2085 	 * Update VSI, etc. AQ commands. It is expected that
2086 	 * these queue set handles be associated with the Tx
2087 	 * queues by the driver as part of the TX queue context
2088 	 * initialization. This has to be done regardless of
2089 	 * DCB as by default everything is mapped to TC0.
2090 	 */
2091 	tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2092 	tx_ctx.rdylist_act = 0;
2093 
2094 	/* clear the context in the HMC */
2095 	err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2096 	if (err) {
2097 		dev_info(&vsi->back->pdev->dev,
2098 			 "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2099 			 ring->queue_index, pf_q, err);
2100 		return -ENOMEM;
2101 	}
2102 
2103 	/* set the context in the HMC */
2104 	err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2105 	if (err) {
2106 		dev_info(&vsi->back->pdev->dev,
2107 			 "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2108 			 ring->queue_index, pf_q, err);
2109 		return -ENOMEM;
2110 	}
2111 
2112 	/* Now associate this queue with this PCI function */
2113 	qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2114 	qtx_ctl |= ((hw->hmc.hmc_fn_id << I40E_QTX_CTL_PF_INDX_SHIFT)
2115 						& I40E_QTX_CTL_PF_INDX_MASK);
2116 	wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2117 	i40e_flush(hw);
2118 
2119 	clear_bit(__I40E_HANG_CHECK_ARMED, &ring->state);
2120 
2121 	/* cache tail off for easier writes later */
2122 	ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2123 
2124 	return 0;
2125 }
2126 
2127 /**
2128  * i40e_configure_rx_ring - Configure a receive ring context
2129  * @ring: The Rx ring to configure
2130  *
2131  * Configure the Rx descriptor ring in the HMC context.
2132  **/
2133 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2134 {
2135 	struct i40e_vsi *vsi = ring->vsi;
2136 	u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2137 	u16 pf_q = vsi->base_queue + ring->queue_index;
2138 	struct i40e_hw *hw = &vsi->back->hw;
2139 	struct i40e_hmc_obj_rxq rx_ctx;
2140 	i40e_status err = 0;
2141 
2142 	ring->state = 0;
2143 
2144 	/* clear the context structure first */
2145 	memset(&rx_ctx, 0, sizeof(rx_ctx));
2146 
2147 	ring->rx_buf_len = vsi->rx_buf_len;
2148 	ring->rx_hdr_len = vsi->rx_hdr_len;
2149 
2150 	rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2151 	rx_ctx.hbuff = ring->rx_hdr_len >> I40E_RXQ_CTX_HBUFF_SHIFT;
2152 
2153 	rx_ctx.base = (ring->dma / 128);
2154 	rx_ctx.qlen = ring->count;
2155 
2156 	if (vsi->back->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED) {
2157 		set_ring_16byte_desc_enabled(ring);
2158 		rx_ctx.dsize = 0;
2159 	} else {
2160 		rx_ctx.dsize = 1;
2161 	}
2162 
2163 	rx_ctx.dtype = vsi->dtype;
2164 	if (vsi->dtype) {
2165 		set_ring_ps_enabled(ring);
2166 		rx_ctx.hsplit_0 = I40E_RX_SPLIT_L2      |
2167 				  I40E_RX_SPLIT_IP      |
2168 				  I40E_RX_SPLIT_TCP_UDP |
2169 				  I40E_RX_SPLIT_SCTP;
2170 	} else {
2171 		rx_ctx.hsplit_0 = 0;
2172 	}
2173 
2174 	rx_ctx.rxmax = min_t(u16, vsi->max_frame,
2175 				  (chain_len * ring->rx_buf_len));
2176 	rx_ctx.tphrdesc_ena = 1;
2177 	rx_ctx.tphwdesc_ena = 1;
2178 	rx_ctx.tphdata_ena = 1;
2179 	rx_ctx.tphhead_ena = 1;
2180 	rx_ctx.lrxqthresh = 2;
2181 	rx_ctx.crcstrip = 1;
2182 	rx_ctx.l2tsel = 1;
2183 	rx_ctx.showiv = 1;
2184 
2185 	/* clear the context in the HMC */
2186 	err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2187 	if (err) {
2188 		dev_info(&vsi->back->pdev->dev,
2189 			 "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2190 			 ring->queue_index, pf_q, err);
2191 		return -ENOMEM;
2192 	}
2193 
2194 	/* set the context in the HMC */
2195 	err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2196 	if (err) {
2197 		dev_info(&vsi->back->pdev->dev,
2198 			 "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2199 			 ring->queue_index, pf_q, err);
2200 		return -ENOMEM;
2201 	}
2202 
2203 	/* cache tail for quicker writes, and clear the reg before use */
2204 	ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2205 	writel(0, ring->tail);
2206 
2207 	i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2208 
2209 	return 0;
2210 }
2211 
2212 /**
2213  * i40e_vsi_configure_tx - Configure the VSI for Tx
2214  * @vsi: VSI structure describing this set of rings and resources
2215  *
2216  * Configure the Tx VSI for operation.
2217  **/
2218 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2219 {
2220 	int err = 0;
2221 	u16 i;
2222 
2223 	for (i = 0; (i < vsi->num_queue_pairs) && (!err); i++)
2224 		err = i40e_configure_tx_ring(&vsi->tx_rings[i]);
2225 
2226 	return err;
2227 }
2228 
2229 /**
2230  * i40e_vsi_configure_rx - Configure the VSI for Rx
2231  * @vsi: the VSI being configured
2232  *
2233  * Configure the Rx VSI for operation.
2234  **/
2235 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
2236 {
2237 	int err = 0;
2238 	u16 i;
2239 
2240 	if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
2241 		vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
2242 			       + ETH_FCS_LEN + VLAN_HLEN;
2243 	else
2244 		vsi->max_frame = I40E_RXBUFFER_2048;
2245 
2246 	/* figure out correct receive buffer length */
2247 	switch (vsi->back->flags & (I40E_FLAG_RX_1BUF_ENABLED |
2248 				    I40E_FLAG_RX_PS_ENABLED)) {
2249 	case I40E_FLAG_RX_1BUF_ENABLED:
2250 		vsi->rx_hdr_len = 0;
2251 		vsi->rx_buf_len = vsi->max_frame;
2252 		vsi->dtype = I40E_RX_DTYPE_NO_SPLIT;
2253 		break;
2254 	case I40E_FLAG_RX_PS_ENABLED:
2255 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2256 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2257 		vsi->dtype = I40E_RX_DTYPE_HEADER_SPLIT;
2258 		break;
2259 	default:
2260 		vsi->rx_hdr_len = I40E_RX_HDR_SIZE;
2261 		vsi->rx_buf_len = I40E_RXBUFFER_2048;
2262 		vsi->dtype = I40E_RX_DTYPE_SPLIT_ALWAYS;
2263 		break;
2264 	}
2265 
2266 	/* round up for the chip's needs */
2267 	vsi->rx_hdr_len = ALIGN(vsi->rx_hdr_len,
2268 				(1 << I40E_RXQ_CTX_HBUFF_SHIFT));
2269 	vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
2270 				(1 << I40E_RXQ_CTX_DBUFF_SHIFT));
2271 
2272 	/* set up individual rings */
2273 	for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2274 		err = i40e_configure_rx_ring(&vsi->rx_rings[i]);
2275 
2276 	return err;
2277 }
2278 
2279 /**
2280  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
2281  * @vsi: ptr to the VSI
2282  **/
2283 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
2284 {
2285 	u16 qoffset, qcount;
2286 	int i, n;
2287 
2288 	if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED))
2289 		return;
2290 
2291 	for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
2292 		if (!(vsi->tc_config.enabled_tc & (1 << n)))
2293 			continue;
2294 
2295 		qoffset = vsi->tc_config.tc_info[n].qoffset;
2296 		qcount = vsi->tc_config.tc_info[n].qcount;
2297 		for (i = qoffset; i < (qoffset + qcount); i++) {
2298 			struct i40e_ring *rx_ring = &vsi->rx_rings[i];
2299 			struct i40e_ring *tx_ring = &vsi->tx_rings[i];
2300 			rx_ring->dcb_tc = n;
2301 			tx_ring->dcb_tc = n;
2302 		}
2303 	}
2304 }
2305 
2306 /**
2307  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
2308  * @vsi: ptr to the VSI
2309  **/
2310 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
2311 {
2312 	if (vsi->netdev)
2313 		i40e_set_rx_mode(vsi->netdev);
2314 }
2315 
2316 /**
2317  * i40e_vsi_configure - Set up the VSI for action
2318  * @vsi: the VSI being configured
2319  **/
2320 static int i40e_vsi_configure(struct i40e_vsi *vsi)
2321 {
2322 	int err;
2323 
2324 	i40e_set_vsi_rx_mode(vsi);
2325 	i40e_restore_vlan(vsi);
2326 	i40e_vsi_config_dcb_rings(vsi);
2327 	err = i40e_vsi_configure_tx(vsi);
2328 	if (!err)
2329 		err = i40e_vsi_configure_rx(vsi);
2330 
2331 	return err;
2332 }
2333 
2334 /**
2335  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
2336  * @vsi: the VSI being configured
2337  **/
2338 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
2339 {
2340 	struct i40e_pf *pf = vsi->back;
2341 	struct i40e_q_vector *q_vector;
2342 	struct i40e_hw *hw = &pf->hw;
2343 	u16 vector;
2344 	int i, q;
2345 	u32 val;
2346 	u32 qp;
2347 
2348 	/* The interrupt indexing is offset by 1 in the PFINT_ITRn
2349 	 * and PFINT_LNKLSTn registers, e.g.:
2350 	 *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
2351 	 */
2352 	qp = vsi->base_queue;
2353 	vector = vsi->base_vector;
2354 	q_vector = vsi->q_vectors;
2355 	for (i = 0; i < vsi->num_q_vectors; i++, q_vector++, vector++) {
2356 		q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2357 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
2358 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
2359 		     q_vector->rx.itr);
2360 		q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2361 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
2362 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
2363 		     q_vector->tx.itr);
2364 
2365 		/* Linked list for the queuepairs assigned to this vector */
2366 		wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
2367 		for (q = 0; q < q_vector->num_ringpairs; q++) {
2368 			val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
2369 			      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
2370 			      (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
2371 			      (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
2372 			      (I40E_QUEUE_TYPE_TX
2373 				      << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
2374 
2375 			wr32(hw, I40E_QINT_RQCTL(qp), val);
2376 
2377 			val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
2378 			      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
2379 			      (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
2380 			      ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
2381 			      (I40E_QUEUE_TYPE_RX
2382 				      << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2383 
2384 			/* Terminate the linked list */
2385 			if (q == (q_vector->num_ringpairs - 1))
2386 				val |= (I40E_QUEUE_END_OF_LIST
2387 					   << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2388 
2389 			wr32(hw, I40E_QINT_TQCTL(qp), val);
2390 			qp++;
2391 		}
2392 	}
2393 
2394 	i40e_flush(hw);
2395 }
2396 
2397 /**
2398  * i40e_enable_misc_int_causes - enable the non-queue interrupts
2399  * @hw: ptr to the hardware info
2400  **/
2401 static void i40e_enable_misc_int_causes(struct i40e_hw *hw)
2402 {
2403 	u32 val;
2404 
2405 	/* clear things first */
2406 	wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
2407 	rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
2408 
2409 	val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
2410 	      I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
2411 	      I40E_PFINT_ICR0_ENA_GRST_MASK          |
2412 	      I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
2413 	      I40E_PFINT_ICR0_ENA_GPIO_MASK          |
2414 	      I40E_PFINT_ICR0_ENA_STORM_DETECT_MASK  |
2415 	      I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
2416 	      I40E_PFINT_ICR0_ENA_VFLR_MASK          |
2417 	      I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2418 
2419 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
2420 
2421 	/* SW_ITR_IDX = 0, but don't change INTENA */
2422 	wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTLN_SW_ITR_INDX_MASK |
2423 					I40E_PFINT_DYN_CTLN_INTENA_MSK_MASK);
2424 
2425 	/* OTHER_ITR_IDX = 0 */
2426 	wr32(hw, I40E_PFINT_STAT_CTL0, 0);
2427 }
2428 
2429 /**
2430  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
2431  * @vsi: the VSI being configured
2432  **/
2433 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
2434 {
2435 	struct i40e_q_vector *q_vector = vsi->q_vectors;
2436 	struct i40e_pf *pf = vsi->back;
2437 	struct i40e_hw *hw = &pf->hw;
2438 	u32 val;
2439 
2440 	/* set the ITR configuration */
2441 	q_vector->rx.itr = ITR_TO_REG(vsi->rx_itr_setting);
2442 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
2443 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
2444 	q_vector->tx.itr = ITR_TO_REG(vsi->tx_itr_setting);
2445 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
2446 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
2447 
2448 	i40e_enable_misc_int_causes(hw);
2449 
2450 	/* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
2451 	wr32(hw, I40E_PFINT_LNKLST0, 0);
2452 
2453 	/* Associate the queue pair to the vector and enable the q int */
2454 	val = I40E_QINT_RQCTL_CAUSE_ENA_MASK		      |
2455 	      (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
2456 	      (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
2457 
2458 	wr32(hw, I40E_QINT_RQCTL(0), val);
2459 
2460 	val = I40E_QINT_TQCTL_CAUSE_ENA_MASK		      |
2461 	      (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
2462 	      (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
2463 
2464 	wr32(hw, I40E_QINT_TQCTL(0), val);
2465 	i40e_flush(hw);
2466 }
2467 
2468 /**
2469  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
2470  * @pf: board private structure
2471  **/
2472 static void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf)
2473 {
2474 	struct i40e_hw *hw = &pf->hw;
2475 	u32 val;
2476 
2477 	val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
2478 	      I40E_PFINT_DYN_CTL0_CLEARPBA_MASK |
2479 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
2480 
2481 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2482 	i40e_flush(hw);
2483 }
2484 
2485 /**
2486  * i40e_irq_dynamic_enable - Enable default interrupt generation settings
2487  * @vsi: pointer to a vsi
2488  * @vector: enable a particular Hw Interrupt vector
2489  **/
2490 void i40e_irq_dynamic_enable(struct i40e_vsi *vsi, int vector)
2491 {
2492 	struct i40e_pf *pf = vsi->back;
2493 	struct i40e_hw *hw = &pf->hw;
2494 	u32 val;
2495 
2496 	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
2497 	      I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
2498 	      (I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
2499 	wr32(hw, I40E_PFINT_DYN_CTLN(vector - 1), val);
2500 	i40e_flush(hw);
2501 }
2502 
2503 /**
2504  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
2505  * @irq: interrupt number
2506  * @data: pointer to a q_vector
2507  **/
2508 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
2509 {
2510 	struct i40e_q_vector *q_vector = data;
2511 
2512 	if (!q_vector->tx.ring[0] && !q_vector->rx.ring[0])
2513 		return IRQ_HANDLED;
2514 
2515 	napi_schedule(&q_vector->napi);
2516 
2517 	return IRQ_HANDLED;
2518 }
2519 
2520 /**
2521  * i40e_fdir_clean_rings - Interrupt Handler for FDIR rings
2522  * @irq: interrupt number
2523  * @data: pointer to a q_vector
2524  **/
2525 static irqreturn_t i40e_fdir_clean_rings(int irq, void *data)
2526 {
2527 	struct i40e_q_vector *q_vector = data;
2528 
2529 	if (!q_vector->tx.ring[0] && !q_vector->rx.ring[0])
2530 		return IRQ_HANDLED;
2531 
2532 	pr_info("fdir ring cleaning needed\n");
2533 
2534 	return IRQ_HANDLED;
2535 }
2536 
2537 /**
2538  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
2539  * @vsi: the VSI being configured
2540  * @basename: name for the vector
2541  *
2542  * Allocates MSI-X vectors and requests interrupts from the kernel.
2543  **/
2544 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
2545 {
2546 	int q_vectors = vsi->num_q_vectors;
2547 	struct i40e_pf *pf = vsi->back;
2548 	int base = vsi->base_vector;
2549 	int rx_int_idx = 0;
2550 	int tx_int_idx = 0;
2551 	int vector, err;
2552 
2553 	for (vector = 0; vector < q_vectors; vector++) {
2554 		struct i40e_q_vector *q_vector = &(vsi->q_vectors[vector]);
2555 
2556 		if (q_vector->tx.ring[0] && q_vector->rx.ring[0]) {
2557 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2558 				 "%s-%s-%d", basename, "TxRx", rx_int_idx++);
2559 			tx_int_idx++;
2560 		} else if (q_vector->rx.ring[0]) {
2561 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2562 				 "%s-%s-%d", basename, "rx", rx_int_idx++);
2563 		} else if (q_vector->tx.ring[0]) {
2564 			snprintf(q_vector->name, sizeof(q_vector->name) - 1,
2565 				 "%s-%s-%d", basename, "tx", tx_int_idx++);
2566 		} else {
2567 			/* skip this unused q_vector */
2568 			continue;
2569 		}
2570 		err = request_irq(pf->msix_entries[base + vector].vector,
2571 				  vsi->irq_handler,
2572 				  0,
2573 				  q_vector->name,
2574 				  q_vector);
2575 		if (err) {
2576 			dev_info(&pf->pdev->dev,
2577 				 "%s: request_irq failed, error: %d\n",
2578 				 __func__, err);
2579 			goto free_queue_irqs;
2580 		}
2581 		/* assign the mask for this irq */
2582 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2583 				      &q_vector->affinity_mask);
2584 	}
2585 
2586 	return 0;
2587 
2588 free_queue_irqs:
2589 	while (vector) {
2590 		vector--;
2591 		irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
2592 				      NULL);
2593 		free_irq(pf->msix_entries[base + vector].vector,
2594 			 &(vsi->q_vectors[vector]));
2595 	}
2596 	return err;
2597 }
2598 
2599 /**
2600  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
2601  * @vsi: the VSI being un-configured
2602  **/
2603 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
2604 {
2605 	struct i40e_pf *pf = vsi->back;
2606 	struct i40e_hw *hw = &pf->hw;
2607 	int base = vsi->base_vector;
2608 	int i;
2609 
2610 	for (i = 0; i < vsi->num_queue_pairs; i++) {
2611 		wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i].reg_idx), 0);
2612 		wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i].reg_idx), 0);
2613 	}
2614 
2615 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2616 		for (i = vsi->base_vector;
2617 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
2618 			wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
2619 
2620 		i40e_flush(hw);
2621 		for (i = 0; i < vsi->num_q_vectors; i++)
2622 			synchronize_irq(pf->msix_entries[i + base].vector);
2623 	} else {
2624 		/* Legacy and MSI mode - this stops all interrupt handling */
2625 		wr32(hw, I40E_PFINT_ICR0_ENA, 0);
2626 		wr32(hw, I40E_PFINT_DYN_CTL0, 0);
2627 		i40e_flush(hw);
2628 		synchronize_irq(pf->pdev->irq);
2629 	}
2630 }
2631 
2632 /**
2633  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
2634  * @vsi: the VSI being configured
2635  **/
2636 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
2637 {
2638 	struct i40e_pf *pf = vsi->back;
2639 	int i;
2640 
2641 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2642 		for (i = vsi->base_vector;
2643 		     i < (vsi->num_q_vectors + vsi->base_vector); i++)
2644 			i40e_irq_dynamic_enable(vsi, i);
2645 	} else {
2646 		i40e_irq_dynamic_enable_icr0(pf);
2647 	}
2648 
2649 	return 0;
2650 }
2651 
2652 /**
2653  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
2654  * @pf: board private structure
2655  **/
2656 static void i40e_stop_misc_vector(struct i40e_pf *pf)
2657 {
2658 	/* Disable ICR 0 */
2659 	wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
2660 	i40e_flush(&pf->hw);
2661 }
2662 
2663 /**
2664  * i40e_intr - MSI/Legacy and non-queue interrupt handler
2665  * @irq: interrupt number
2666  * @data: pointer to a q_vector
2667  *
2668  * This is the handler used for all MSI/Legacy interrupts, and deals
2669  * with both queue and non-queue interrupts.  This is also used in
2670  * MSIX mode to handle the non-queue interrupts.
2671  **/
2672 static irqreturn_t i40e_intr(int irq, void *data)
2673 {
2674 	struct i40e_pf *pf = (struct i40e_pf *)data;
2675 	struct i40e_hw *hw = &pf->hw;
2676 	u32 icr0, icr0_remaining;
2677 	u32 val, ena_mask;
2678 
2679 	icr0 = rd32(hw, I40E_PFINT_ICR0);
2680 
2681 	/* if sharing a legacy IRQ, we might get called w/o an intr pending */
2682 	if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
2683 		return IRQ_NONE;
2684 
2685 	val = rd32(hw, I40E_PFINT_DYN_CTL0);
2686 	val = val | I40E_PFINT_DYN_CTL0_CLEARPBA_MASK;
2687 	wr32(hw, I40E_PFINT_DYN_CTL0, val);
2688 
2689 	ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
2690 
2691 	/* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
2692 	if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
2693 
2694 		/* temporarily disable queue cause for NAPI processing */
2695 		u32 qval = rd32(hw, I40E_QINT_RQCTL(0));
2696 		qval &= ~I40E_QINT_RQCTL_CAUSE_ENA_MASK;
2697 		wr32(hw, I40E_QINT_RQCTL(0), qval);
2698 
2699 		qval = rd32(hw, I40E_QINT_TQCTL(0));
2700 		qval &= ~I40E_QINT_TQCTL_CAUSE_ENA_MASK;
2701 		wr32(hw, I40E_QINT_TQCTL(0), qval);
2702 		i40e_flush(hw);
2703 
2704 		if (!test_bit(__I40E_DOWN, &pf->state))
2705 			napi_schedule(&pf->vsi[pf->lan_vsi]->q_vectors[0].napi);
2706 	}
2707 
2708 	if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
2709 		ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
2710 		set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
2711 	}
2712 
2713 	if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
2714 		ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
2715 		set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
2716 	}
2717 
2718 	if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
2719 		ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
2720 		set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
2721 	}
2722 
2723 	if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
2724 		if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
2725 			set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
2726 		ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
2727 		val = rd32(hw, I40E_GLGEN_RSTAT);
2728 		val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
2729 		       >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
2730 		if (val & I40E_RESET_CORER)
2731 			pf->corer_count++;
2732 		else if (val & I40E_RESET_GLOBR)
2733 			pf->globr_count++;
2734 		else if (val & I40E_RESET_EMPR)
2735 			pf->empr_count++;
2736 	}
2737 
2738 	/* If a critical error is pending we have no choice but to reset the
2739 	 * device.
2740 	 * Report and mask out any remaining unexpected interrupts.
2741 	 */
2742 	icr0_remaining = icr0 & ena_mask;
2743 	if (icr0_remaining) {
2744 		dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
2745 			 icr0_remaining);
2746 		if ((icr0_remaining & I40E_PFINT_ICR0_HMC_ERR_MASK) ||
2747 		    (icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
2748 		    (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
2749 		    (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK) ||
2750 		    (icr0_remaining & I40E_PFINT_ICR0_MAL_DETECT_MASK)) {
2751 			if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
2752 				dev_info(&pf->pdev->dev, "HMC error interrupt\n");
2753 			} else {
2754 				dev_info(&pf->pdev->dev, "device will be reset\n");
2755 				set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
2756 				i40e_service_event_schedule(pf);
2757 			}
2758 		}
2759 		ena_mask &= ~icr0_remaining;
2760 	}
2761 
2762 	/* re-enable interrupt causes */
2763 	wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
2764 	i40e_flush(hw);
2765 	if (!test_bit(__I40E_DOWN, &pf->state)) {
2766 		i40e_service_event_schedule(pf);
2767 		i40e_irq_dynamic_enable_icr0(pf);
2768 	}
2769 
2770 	return IRQ_HANDLED;
2771 }
2772 
2773 /**
2774  * i40e_map_vector_to_rxq - Assigns the Rx queue to the vector
2775  * @vsi: the VSI being configured
2776  * @v_idx: vector index
2777  * @r_idx: rx queue index
2778  **/
2779 static void map_vector_to_rxq(struct i40e_vsi *vsi, int v_idx, int r_idx)
2780 {
2781 	struct i40e_q_vector *q_vector = &(vsi->q_vectors[v_idx]);
2782 	struct i40e_ring *rx_ring = &(vsi->rx_rings[r_idx]);
2783 
2784 	rx_ring->q_vector = q_vector;
2785 	q_vector->rx.ring[q_vector->rx.count] = rx_ring;
2786 	q_vector->rx.count++;
2787 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
2788 	q_vector->vsi = vsi;
2789 }
2790 
2791 /**
2792  * i40e_map_vector_to_txq - Assigns the Tx queue to the vector
2793  * @vsi: the VSI being configured
2794  * @v_idx: vector index
2795  * @t_idx: tx queue index
2796  **/
2797 static void map_vector_to_txq(struct i40e_vsi *vsi, int v_idx, int t_idx)
2798 {
2799 	struct i40e_q_vector *q_vector = &(vsi->q_vectors[v_idx]);
2800 	struct i40e_ring *tx_ring = &(vsi->tx_rings[t_idx]);
2801 
2802 	tx_ring->q_vector = q_vector;
2803 	q_vector->tx.ring[q_vector->tx.count] = tx_ring;
2804 	q_vector->tx.count++;
2805 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
2806 	q_vector->num_ringpairs++;
2807 	q_vector->vsi = vsi;
2808 }
2809 
2810 /**
2811  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
2812  * @vsi: the VSI being configured
2813  *
2814  * This function maps descriptor rings to the queue-specific vectors
2815  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
2816  * one vector per queue pair, but on a constrained vector budget, we
2817  * group the queue pairs as "efficiently" as possible.
2818  **/
2819 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
2820 {
2821 	int qp_remaining = vsi->num_queue_pairs;
2822 	int q_vectors = vsi->num_q_vectors;
2823 	int qp_per_vector;
2824 	int v_start = 0;
2825 	int qp_idx = 0;
2826 
2827 	/* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
2828 	 * group them so there are multiple queues per vector.
2829 	 */
2830 	for (; v_start < q_vectors && qp_remaining; v_start++) {
2831 		qp_per_vector = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
2832 		for (; qp_per_vector;
2833 		     qp_per_vector--, qp_idx++, qp_remaining--)	{
2834 			map_vector_to_rxq(vsi, v_start, qp_idx);
2835 			map_vector_to_txq(vsi, v_start, qp_idx);
2836 		}
2837 	}
2838 }
2839 
2840 /**
2841  * i40e_vsi_request_irq - Request IRQ from the OS
2842  * @vsi: the VSI being configured
2843  * @basename: name for the vector
2844  **/
2845 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
2846 {
2847 	struct i40e_pf *pf = vsi->back;
2848 	int err;
2849 
2850 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
2851 		err = i40e_vsi_request_irq_msix(vsi, basename);
2852 	else if (pf->flags & I40E_FLAG_MSI_ENABLED)
2853 		err = request_irq(pf->pdev->irq, i40e_intr, 0,
2854 				  pf->misc_int_name, pf);
2855 	else
2856 		err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
2857 				  pf->misc_int_name, pf);
2858 
2859 	if (err)
2860 		dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
2861 
2862 	return err;
2863 }
2864 
2865 #ifdef CONFIG_NET_POLL_CONTROLLER
2866 /**
2867  * i40e_netpoll - A Polling 'interrupt'handler
2868  * @netdev: network interface device structure
2869  *
2870  * This is used by netconsole to send skbs without having to re-enable
2871  * interrupts.  It's not called while the normal interrupt routine is executing.
2872  **/
2873 static void i40e_netpoll(struct net_device *netdev)
2874 {
2875 	struct i40e_netdev_priv *np = netdev_priv(netdev);
2876 	struct i40e_vsi *vsi = np->vsi;
2877 	struct i40e_pf *pf = vsi->back;
2878 	int i;
2879 
2880 	/* if interface is down do nothing */
2881 	if (test_bit(__I40E_DOWN, &vsi->state))
2882 		return;
2883 
2884 	pf->flags |= I40E_FLAG_IN_NETPOLL;
2885 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
2886 		for (i = 0; i < vsi->num_q_vectors; i++)
2887 			i40e_msix_clean_rings(0, &vsi->q_vectors[i]);
2888 	} else {
2889 		i40e_intr(pf->pdev->irq, netdev);
2890 	}
2891 	pf->flags &= ~I40E_FLAG_IN_NETPOLL;
2892 }
2893 #endif
2894 
2895 /**
2896  * i40e_vsi_control_tx - Start or stop a VSI's rings
2897  * @vsi: the VSI being configured
2898  * @enable: start or stop the rings
2899  **/
2900 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
2901 {
2902 	struct i40e_pf *pf = vsi->back;
2903 	struct i40e_hw *hw = &pf->hw;
2904 	int i, j, pf_q;
2905 	u32 tx_reg;
2906 
2907 	pf_q = vsi->base_queue;
2908 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
2909 		j = 1000;
2910 		do {
2911 			usleep_range(1000, 2000);
2912 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
2913 		} while (j-- && ((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT)
2914 			       ^ (tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT)) & 1);
2915 
2916 		if (enable) {
2917 			/* is STAT set ? */
2918 			if ((tx_reg & I40E_QTX_ENA_QENA_STAT_MASK)) {
2919 				dev_info(&pf->pdev->dev,
2920 					 "Tx %d already enabled\n", i);
2921 				continue;
2922 			}
2923 		} else {
2924 			/* is !STAT set ? */
2925 			if (!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK)) {
2926 				dev_info(&pf->pdev->dev,
2927 					 "Tx %d already disabled\n", i);
2928 				continue;
2929 			}
2930 		}
2931 
2932 		/* turn on/off the queue */
2933 		if (enable)
2934 			tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK |
2935 				  I40E_QTX_ENA_QENA_STAT_MASK;
2936 		else
2937 			tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
2938 
2939 		wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
2940 
2941 		/* wait for the change to finish */
2942 		for (j = 0; j < 10; j++) {
2943 			tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
2944 			if (enable) {
2945 				if ((tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
2946 					break;
2947 			} else {
2948 				if (!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
2949 					break;
2950 			}
2951 
2952 			udelay(10);
2953 		}
2954 		if (j >= 10) {
2955 			dev_info(&pf->pdev->dev, "Tx ring %d %sable timeout\n",
2956 				 pf_q, (enable ? "en" : "dis"));
2957 			return -ETIMEDOUT;
2958 		}
2959 	}
2960 
2961 	return 0;
2962 }
2963 
2964 /**
2965  * i40e_vsi_control_rx - Start or stop a VSI's rings
2966  * @vsi: the VSI being configured
2967  * @enable: start or stop the rings
2968  **/
2969 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
2970 {
2971 	struct i40e_pf *pf = vsi->back;
2972 	struct i40e_hw *hw = &pf->hw;
2973 	int i, j, pf_q;
2974 	u32 rx_reg;
2975 
2976 	pf_q = vsi->base_queue;
2977 	for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
2978 		j = 1000;
2979 		do {
2980 			usleep_range(1000, 2000);
2981 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
2982 		} while (j-- && ((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT)
2983 			       ^ (rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT)) & 1);
2984 
2985 		if (enable) {
2986 			/* is STAT set ? */
2987 			if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
2988 				continue;
2989 		} else {
2990 			/* is !STAT set ? */
2991 			if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
2992 				continue;
2993 		}
2994 
2995 		/* turn on/off the queue */
2996 		if (enable)
2997 			rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK |
2998 				  I40E_QRX_ENA_QENA_STAT_MASK;
2999 		else
3000 			rx_reg &= ~(I40E_QRX_ENA_QENA_REQ_MASK |
3001 				  I40E_QRX_ENA_QENA_STAT_MASK);
3002 		wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3003 
3004 		/* wait for the change to finish */
3005 		for (j = 0; j < 10; j++) {
3006 			rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3007 
3008 			if (enable) {
3009 				if ((rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3010 					break;
3011 			} else {
3012 				if (!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3013 					break;
3014 			}
3015 
3016 			udelay(10);
3017 		}
3018 		if (j >= 10) {
3019 			dev_info(&pf->pdev->dev, "Rx ring %d %sable timeout\n",
3020 				 pf_q, (enable ? "en" : "dis"));
3021 			return -ETIMEDOUT;
3022 		}
3023 	}
3024 
3025 	return 0;
3026 }
3027 
3028 /**
3029  * i40e_vsi_control_rings - Start or stop a VSI's rings
3030  * @vsi: the VSI being configured
3031  * @enable: start or stop the rings
3032  **/
3033 static int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3034 {
3035 	int ret;
3036 
3037 	/* do rx first for enable and last for disable */
3038 	if (request) {
3039 		ret = i40e_vsi_control_rx(vsi, request);
3040 		if (ret)
3041 			return ret;
3042 		ret = i40e_vsi_control_tx(vsi, request);
3043 	} else {
3044 		ret = i40e_vsi_control_tx(vsi, request);
3045 		if (ret)
3046 			return ret;
3047 		ret = i40e_vsi_control_rx(vsi, request);
3048 	}
3049 
3050 	return ret;
3051 }
3052 
3053 /**
3054  * i40e_vsi_free_irq - Free the irq association with the OS
3055  * @vsi: the VSI being configured
3056  **/
3057 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
3058 {
3059 	struct i40e_pf *pf = vsi->back;
3060 	struct i40e_hw *hw = &pf->hw;
3061 	int base = vsi->base_vector;
3062 	u32 val, qp;
3063 	int i;
3064 
3065 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3066 		if (!vsi->q_vectors)
3067 			return;
3068 
3069 		for (i = 0; i < vsi->num_q_vectors; i++) {
3070 			u16 vector = i + base;
3071 
3072 			/* free only the irqs that were actually requested */
3073 			if (vsi->q_vectors[i].num_ringpairs == 0)
3074 				continue;
3075 
3076 			/* clear the affinity_mask in the IRQ descriptor */
3077 			irq_set_affinity_hint(pf->msix_entries[vector].vector,
3078 					      NULL);
3079 			free_irq(pf->msix_entries[vector].vector,
3080 				 &vsi->q_vectors[i]);
3081 
3082 			/* Tear down the interrupt queue link list
3083 			 *
3084 			 * We know that they come in pairs and always
3085 			 * the Rx first, then the Tx.  To clear the
3086 			 * link list, stick the EOL value into the
3087 			 * next_q field of the registers.
3088 			 */
3089 			val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
3090 			qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3091 				>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3092 			val |= I40E_QUEUE_END_OF_LIST
3093 				<< I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3094 			wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
3095 
3096 			while (qp != I40E_QUEUE_END_OF_LIST) {
3097 				u32 next;
3098 
3099 				val = rd32(hw, I40E_QINT_RQCTL(qp));
3100 
3101 				val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3102 					 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3103 					 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3104 					 I40E_QINT_RQCTL_INTEVENT_MASK);
3105 
3106 				val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3107 					 I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3108 
3109 				wr32(hw, I40E_QINT_RQCTL(qp), val);
3110 
3111 				val = rd32(hw, I40E_QINT_TQCTL(qp));
3112 
3113 				next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
3114 					>> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
3115 
3116 				val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3117 					 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3118 					 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3119 					 I40E_QINT_TQCTL_INTEVENT_MASK);
3120 
3121 				val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3122 					 I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3123 
3124 				wr32(hw, I40E_QINT_TQCTL(qp), val);
3125 				qp = next;
3126 			}
3127 		}
3128 	} else {
3129 		free_irq(pf->pdev->irq, pf);
3130 
3131 		val = rd32(hw, I40E_PFINT_LNKLST0);
3132 		qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
3133 			>> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
3134 		val |= I40E_QUEUE_END_OF_LIST
3135 			<< I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
3136 		wr32(hw, I40E_PFINT_LNKLST0, val);
3137 
3138 		val = rd32(hw, I40E_QINT_RQCTL(qp));
3139 		val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
3140 			 I40E_QINT_RQCTL_MSIX0_INDX_MASK |
3141 			 I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
3142 			 I40E_QINT_RQCTL_INTEVENT_MASK);
3143 
3144 		val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
3145 			I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
3146 
3147 		wr32(hw, I40E_QINT_RQCTL(qp), val);
3148 
3149 		val = rd32(hw, I40E_QINT_TQCTL(qp));
3150 
3151 		val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
3152 			 I40E_QINT_TQCTL_MSIX0_INDX_MASK |
3153 			 I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
3154 			 I40E_QINT_TQCTL_INTEVENT_MASK);
3155 
3156 		val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
3157 			I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
3158 
3159 		wr32(hw, I40E_QINT_TQCTL(qp), val);
3160 	}
3161 }
3162 
3163 /**
3164  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
3165  * @vsi: the VSI being un-configured
3166  *
3167  * This frees the memory allocated to the q_vectors and
3168  * deletes references to the NAPI struct.
3169  **/
3170 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
3171 {
3172 	int v_idx;
3173 
3174 	for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++) {
3175 		struct i40e_q_vector *q_vector = &vsi->q_vectors[v_idx];
3176 		int r_idx;
3177 
3178 		if (!q_vector)
3179 			continue;
3180 
3181 		/* disassociate q_vector from rings */
3182 		for (r_idx = 0; r_idx < q_vector->tx.count; r_idx++)
3183 			q_vector->tx.ring[r_idx]->q_vector = NULL;
3184 		for (r_idx = 0; r_idx < q_vector->rx.count; r_idx++)
3185 			q_vector->rx.ring[r_idx]->q_vector = NULL;
3186 
3187 		/* only VSI w/ an associated netdev is set up w/ NAPI */
3188 		if (vsi->netdev)
3189 			netif_napi_del(&q_vector->napi);
3190 	}
3191 	kfree(vsi->q_vectors);
3192 }
3193 
3194 /**
3195  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
3196  * @pf: board private structure
3197  **/
3198 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
3199 {
3200 	/* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
3201 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3202 		pci_disable_msix(pf->pdev);
3203 		kfree(pf->msix_entries);
3204 		pf->msix_entries = NULL;
3205 	} else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
3206 		pci_disable_msi(pf->pdev);
3207 	}
3208 	pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
3209 }
3210 
3211 /**
3212  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
3213  * @pf: board private structure
3214  *
3215  * We go through and clear interrupt specific resources and reset the structure
3216  * to pre-load conditions
3217  **/
3218 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
3219 {
3220 	int i;
3221 
3222 	i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
3223 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
3224 		if (pf->vsi[i])
3225 			i40e_vsi_free_q_vectors(pf->vsi[i]);
3226 	i40e_reset_interrupt_capability(pf);
3227 }
3228 
3229 /**
3230  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
3231  * @vsi: the VSI being configured
3232  **/
3233 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
3234 {
3235 	int q_idx;
3236 
3237 	if (!vsi->netdev)
3238 		return;
3239 
3240 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3241 		napi_enable(&vsi->q_vectors[q_idx].napi);
3242 }
3243 
3244 /**
3245  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
3246  * @vsi: the VSI being configured
3247  **/
3248 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
3249 {
3250 	int q_idx;
3251 
3252 	if (!vsi->netdev)
3253 		return;
3254 
3255 	for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
3256 		napi_disable(&vsi->q_vectors[q_idx].napi);
3257 }
3258 
3259 /**
3260  * i40e_quiesce_vsi - Pause a given VSI
3261  * @vsi: the VSI being paused
3262  **/
3263 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
3264 {
3265 	if (test_bit(__I40E_DOWN, &vsi->state))
3266 		return;
3267 
3268 	set_bit(__I40E_NEEDS_RESTART, &vsi->state);
3269 	if (vsi->netdev && netif_running(vsi->netdev)) {
3270 		vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
3271 	} else {
3272 		set_bit(__I40E_DOWN, &vsi->state);
3273 		i40e_down(vsi);
3274 	}
3275 }
3276 
3277 /**
3278  * i40e_unquiesce_vsi - Resume a given VSI
3279  * @vsi: the VSI being resumed
3280  **/
3281 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
3282 {
3283 	if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
3284 		return;
3285 
3286 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
3287 	if (vsi->netdev && netif_running(vsi->netdev))
3288 		vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
3289 	else
3290 		i40e_up(vsi);   /* this clears the DOWN bit */
3291 }
3292 
3293 /**
3294  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
3295  * @pf: the PF
3296  **/
3297 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
3298 {
3299 	int v;
3300 
3301 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3302 		if (pf->vsi[v])
3303 			i40e_quiesce_vsi(pf->vsi[v]);
3304 	}
3305 }
3306 
3307 /**
3308  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
3309  * @pf: the PF
3310  **/
3311 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
3312 {
3313 	int v;
3314 
3315 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3316 		if (pf->vsi[v])
3317 			i40e_unquiesce_vsi(pf->vsi[v]);
3318 	}
3319 }
3320 
3321 /**
3322  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
3323  * @dcbcfg: the corresponding DCBx configuration structure
3324  *
3325  * Return the number of TCs from given DCBx configuration
3326  **/
3327 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
3328 {
3329 	int num_tc = 0, i;
3330 
3331 	/* Scan the ETS Config Priority Table to find
3332 	 * traffic class enabled for a given priority
3333 	 * and use the traffic class index to get the
3334 	 * number of traffic classes enabled
3335 	 */
3336 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3337 		if (dcbcfg->etscfg.prioritytable[i] > num_tc)
3338 			num_tc = dcbcfg->etscfg.prioritytable[i];
3339 	}
3340 
3341 	/* Traffic class index starts from zero so
3342 	 * increment to return the actual count
3343 	 */
3344 	num_tc++;
3345 
3346 	return num_tc;
3347 }
3348 
3349 /**
3350  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
3351  * @dcbcfg: the corresponding DCBx configuration structure
3352  *
3353  * Query the current DCB configuration and return the number of
3354  * traffic classes enabled from the given DCBX config
3355  **/
3356 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
3357 {
3358 	u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
3359 	u8 enabled_tc = 1;
3360 	u8 i;
3361 
3362 	for (i = 0; i < num_tc; i++)
3363 		enabled_tc |= 1 << i;
3364 
3365 	return enabled_tc;
3366 }
3367 
3368 /**
3369  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
3370  * @pf: PF being queried
3371  *
3372  * Return number of traffic classes enabled for the given PF
3373  **/
3374 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
3375 {
3376 	struct i40e_hw *hw = &pf->hw;
3377 	u8 i, enabled_tc;
3378 	u8 num_tc = 0;
3379 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3380 
3381 	/* If DCB is not enabled then always in single TC */
3382 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3383 		return 1;
3384 
3385 	/* MFP mode return count of enabled TCs for this PF */
3386 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3387 		enabled_tc = pf->hw.func_caps.enabled_tcmap;
3388 		for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3389 			if (enabled_tc & (1 << i))
3390 				num_tc++;
3391 		}
3392 		return num_tc;
3393 	}
3394 
3395 	/* SFP mode will be enabled for all TCs on port */
3396 	return i40e_dcb_get_num_tc(dcbcfg);
3397 }
3398 
3399 /**
3400  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
3401  * @pf: PF being queried
3402  *
3403  * Return a bitmap for first enabled traffic class for this PF.
3404  **/
3405 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
3406 {
3407 	u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
3408 	u8 i = 0;
3409 
3410 	if (!enabled_tc)
3411 		return 0x1; /* TC0 */
3412 
3413 	/* Find the first enabled TC */
3414 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3415 		if (enabled_tc & (1 << i))
3416 			break;
3417 	}
3418 
3419 	return 1 << i;
3420 }
3421 
3422 /**
3423  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
3424  * @pf: PF being queried
3425  *
3426  * Return a bitmap for enabled traffic classes for this PF.
3427  **/
3428 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
3429 {
3430 	/* If DCB is not enabled for this PF then just return default TC */
3431 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
3432 		return i40e_pf_get_default_tc(pf);
3433 
3434 	/* MFP mode will have enabled TCs set by FW */
3435 	if (pf->flags & I40E_FLAG_MFP_ENABLED)
3436 		return pf->hw.func_caps.enabled_tcmap;
3437 
3438 	/* SFP mode we want PF to be enabled for all TCs */
3439 	return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
3440 }
3441 
3442 /**
3443  * i40e_vsi_get_bw_info - Query VSI BW Information
3444  * @vsi: the VSI being queried
3445  *
3446  * Returns 0 on success, negative value on failure
3447  **/
3448 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
3449 {
3450 	struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
3451 	struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
3452 	struct i40e_pf *pf = vsi->back;
3453 	struct i40e_hw *hw = &pf->hw;
3454 	u32 tc_bw_max;
3455 	int ret;
3456 	int i;
3457 
3458 	/* Get the VSI level BW configuration */
3459 	ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
3460 	if (ret) {
3461 		dev_info(&pf->pdev->dev,
3462 			 "couldn't get pf vsi bw config, err %d, aq_err %d\n",
3463 			 ret, pf->hw.aq.asq_last_status);
3464 		return ret;
3465 	}
3466 
3467 	/* Get the VSI level BW configuration per TC */
3468 	ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid,
3469 					       &bw_ets_config,
3470 					       NULL);
3471 	if (ret) {
3472 		dev_info(&pf->pdev->dev,
3473 			 "couldn't get pf vsi ets bw config, err %d, aq_err %d\n",
3474 			 ret, pf->hw.aq.asq_last_status);
3475 		return ret;
3476 	}
3477 
3478 	if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
3479 		dev_info(&pf->pdev->dev,
3480 			 "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
3481 			 bw_config.tc_valid_bits,
3482 			 bw_ets_config.tc_valid_bits);
3483 		/* Still continuing */
3484 	}
3485 
3486 	vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
3487 	vsi->bw_max_quanta = bw_config.max_bw;
3488 	tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
3489 		    (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
3490 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3491 		vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
3492 		vsi->bw_ets_limit_credits[i] =
3493 					le16_to_cpu(bw_ets_config.credits[i]);
3494 		/* 3 bits out of 4 for each TC */
3495 		vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
3496 	}
3497 	return ret;
3498 }
3499 
3500 /**
3501  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
3502  * @vsi: the VSI being configured
3503  * @enabled_tc: TC bitmap
3504  * @bw_credits: BW shared credits per TC
3505  *
3506  * Returns 0 on success, negative value on failure
3507  **/
3508 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi,
3509 				       u8 enabled_tc,
3510 				       u8 *bw_share)
3511 {
3512 	struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
3513 	int i, ret = 0;
3514 
3515 	bw_data.tc_valid_bits = enabled_tc;
3516 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3517 		bw_data.tc_bw_credits[i] = bw_share[i];
3518 
3519 	ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid,
3520 				       &bw_data, NULL);
3521 	if (ret) {
3522 		dev_info(&vsi->back->pdev->dev,
3523 			 "%s: AQ command Config VSI BW allocation per TC failed = %d\n",
3524 			 __func__, vsi->back->hw.aq.asq_last_status);
3525 		return ret;
3526 	}
3527 
3528 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
3529 		vsi->info.qs_handle[i] = bw_data.qs_handles[i];
3530 
3531 	return ret;
3532 }
3533 
3534 /**
3535  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
3536  * @vsi: the VSI being configured
3537  * @enabled_tc: TC map to be enabled
3538  *
3539  **/
3540 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3541 {
3542 	struct net_device *netdev = vsi->netdev;
3543 	struct i40e_pf *pf = vsi->back;
3544 	struct i40e_hw *hw = &pf->hw;
3545 	u8 netdev_tc = 0;
3546 	int i;
3547 	struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
3548 
3549 	if (!netdev)
3550 		return;
3551 
3552 	if (!enabled_tc) {
3553 		netdev_reset_tc(netdev);
3554 		return;
3555 	}
3556 
3557 	/* Set up actual enabled TCs on the VSI */
3558 	if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
3559 		return;
3560 
3561 	/* set per TC queues for the VSI */
3562 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3563 		/* Only set TC queues for enabled tcs
3564 		 *
3565 		 * e.g. For a VSI that has TC0 and TC3 enabled the
3566 		 * enabled_tc bitmap would be 0x00001001; the driver
3567 		 * will set the numtc for netdev as 2 that will be
3568 		 * referenced by the netdev layer as TC 0 and 1.
3569 		 */
3570 		if (vsi->tc_config.enabled_tc & (1 << i))
3571 			netdev_set_tc_queue(netdev,
3572 					vsi->tc_config.tc_info[i].netdev_tc,
3573 					vsi->tc_config.tc_info[i].qcount,
3574 					vsi->tc_config.tc_info[i].qoffset);
3575 	}
3576 
3577 	/* Assign UP2TC map for the VSI */
3578 	for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
3579 		/* Get the actual TC# for the UP */
3580 		u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
3581 		/* Get the mapped netdev TC# for the UP */
3582 		netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
3583 		netdev_set_prio_tc_map(netdev, i, netdev_tc);
3584 	}
3585 }
3586 
3587 /**
3588  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
3589  * @vsi: the VSI being configured
3590  * @ctxt: the ctxt buffer returned from AQ VSI update param command
3591  **/
3592 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
3593 				      struct i40e_vsi_context *ctxt)
3594 {
3595 	/* copy just the sections touched not the entire info
3596 	 * since not all sections are valid as returned by
3597 	 * update vsi params
3598 	 */
3599 	vsi->info.mapping_flags = ctxt->info.mapping_flags;
3600 	memcpy(&vsi->info.queue_mapping,
3601 	       &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
3602 	memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
3603 	       sizeof(vsi->info.tc_mapping));
3604 }
3605 
3606 /**
3607  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
3608  * @vsi: VSI to be configured
3609  * @enabled_tc: TC bitmap
3610  *
3611  * This configures a particular VSI for TCs that are mapped to the
3612  * given TC bitmap. It uses default bandwidth share for TCs across
3613  * VSIs to configure TC for a particular VSI.
3614  *
3615  * NOTE:
3616  * It is expected that the VSI queues have been quisced before calling
3617  * this function.
3618  **/
3619 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
3620 {
3621 	u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
3622 	struct i40e_vsi_context ctxt;
3623 	int ret = 0;
3624 	int i;
3625 
3626 	/* Check if enabled_tc is same as existing or new TCs */
3627 	if (vsi->tc_config.enabled_tc == enabled_tc)
3628 		return ret;
3629 
3630 	/* Enable ETS TCs with equal BW Share for now across all VSIs */
3631 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
3632 		if (enabled_tc & (1 << i))
3633 			bw_share[i] = 1;
3634 	}
3635 
3636 	ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
3637 	if (ret) {
3638 		dev_info(&vsi->back->pdev->dev,
3639 			 "Failed configuring TC map %d for VSI %d\n",
3640 			 enabled_tc, vsi->seid);
3641 		goto out;
3642 	}
3643 
3644 	/* Update Queue Pairs Mapping for currently enabled UPs */
3645 	ctxt.seid = vsi->seid;
3646 	ctxt.pf_num = vsi->back->hw.pf_id;
3647 	ctxt.vf_num = 0;
3648 	ctxt.uplink_seid = vsi->uplink_seid;
3649 	memcpy(&ctxt.info, &vsi->info, sizeof(vsi->info));
3650 	i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
3651 
3652 	/* Update the VSI after updating the VSI queue-mapping information */
3653 	ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
3654 	if (ret) {
3655 		dev_info(&vsi->back->pdev->dev,
3656 			 "update vsi failed, aq_err=%d\n",
3657 			 vsi->back->hw.aq.asq_last_status);
3658 		goto out;
3659 	}
3660 	/* update the local VSI info with updated queue map */
3661 	i40e_vsi_update_queue_map(vsi, &ctxt);
3662 	vsi->info.valid_sections = 0;
3663 
3664 	/* Update current VSI BW information */
3665 	ret = i40e_vsi_get_bw_info(vsi);
3666 	if (ret) {
3667 		dev_info(&vsi->back->pdev->dev,
3668 			 "Failed updating vsi bw info, aq_err=%d\n",
3669 			 vsi->back->hw.aq.asq_last_status);
3670 		goto out;
3671 	}
3672 
3673 	/* Update the netdev TC setup */
3674 	i40e_vsi_config_netdev_tc(vsi, enabled_tc);
3675 out:
3676 	return ret;
3677 }
3678 
3679 /**
3680  * i40e_up_complete - Finish the last steps of bringing up a connection
3681  * @vsi: the VSI being configured
3682  **/
3683 static int i40e_up_complete(struct i40e_vsi *vsi)
3684 {
3685 	struct i40e_pf *pf = vsi->back;
3686 	int err;
3687 
3688 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3689 		i40e_vsi_configure_msix(vsi);
3690 	else
3691 		i40e_configure_msi_and_legacy(vsi);
3692 
3693 	/* start rings */
3694 	err = i40e_vsi_control_rings(vsi, true);
3695 	if (err)
3696 		return err;
3697 
3698 	clear_bit(__I40E_DOWN, &vsi->state);
3699 	i40e_napi_enable_all(vsi);
3700 	i40e_vsi_enable_irq(vsi);
3701 
3702 	if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
3703 	    (vsi->netdev)) {
3704 		netif_tx_start_all_queues(vsi->netdev);
3705 		netif_carrier_on(vsi->netdev);
3706 	}
3707 	i40e_service_event_schedule(pf);
3708 
3709 	return 0;
3710 }
3711 
3712 /**
3713  * i40e_vsi_reinit_locked - Reset the VSI
3714  * @vsi: the VSI being configured
3715  *
3716  * Rebuild the ring structs after some configuration
3717  * has changed, e.g. MTU size.
3718  **/
3719 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
3720 {
3721 	struct i40e_pf *pf = vsi->back;
3722 
3723 	WARN_ON(in_interrupt());
3724 	while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
3725 		usleep_range(1000, 2000);
3726 	i40e_down(vsi);
3727 
3728 	/* Give a VF some time to respond to the reset.  The
3729 	 * two second wait is based upon the watchdog cycle in
3730 	 * the VF driver.
3731 	 */
3732 	if (vsi->type == I40E_VSI_SRIOV)
3733 		msleep(2000);
3734 	i40e_up(vsi);
3735 	clear_bit(__I40E_CONFIG_BUSY, &pf->state);
3736 }
3737 
3738 /**
3739  * i40e_up - Bring the connection back up after being down
3740  * @vsi: the VSI being configured
3741  **/
3742 int i40e_up(struct i40e_vsi *vsi)
3743 {
3744 	int err;
3745 
3746 	err = i40e_vsi_configure(vsi);
3747 	if (!err)
3748 		err = i40e_up_complete(vsi);
3749 
3750 	return err;
3751 }
3752 
3753 /**
3754  * i40e_down - Shutdown the connection processing
3755  * @vsi: the VSI being stopped
3756  **/
3757 void i40e_down(struct i40e_vsi *vsi)
3758 {
3759 	int i;
3760 
3761 	/* It is assumed that the caller of this function
3762 	 * sets the vsi->state __I40E_DOWN bit.
3763 	 */
3764 	if (vsi->netdev) {
3765 		netif_carrier_off(vsi->netdev);
3766 		netif_tx_disable(vsi->netdev);
3767 	}
3768 	i40e_vsi_disable_irq(vsi);
3769 	i40e_vsi_control_rings(vsi, false);
3770 	i40e_napi_disable_all(vsi);
3771 
3772 	for (i = 0; i < vsi->num_queue_pairs; i++) {
3773 		i40e_clean_tx_ring(&vsi->tx_rings[i]);
3774 		i40e_clean_rx_ring(&vsi->rx_rings[i]);
3775 	}
3776 }
3777 
3778 /**
3779  * i40e_setup_tc - configure multiple traffic classes
3780  * @netdev: net device to configure
3781  * @tc: number of traffic classes to enable
3782  **/
3783 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
3784 {
3785 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3786 	struct i40e_vsi *vsi = np->vsi;
3787 	struct i40e_pf *pf = vsi->back;
3788 	u8 enabled_tc = 0;
3789 	int ret = -EINVAL;
3790 	int i;
3791 
3792 	/* Check if DCB enabled to continue */
3793 	if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
3794 		netdev_info(netdev, "DCB is not enabled for adapter\n");
3795 		goto exit;
3796 	}
3797 
3798 	/* Check if MFP enabled */
3799 	if (pf->flags & I40E_FLAG_MFP_ENABLED) {
3800 		netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
3801 		goto exit;
3802 	}
3803 
3804 	/* Check whether tc count is within enabled limit */
3805 	if (tc > i40e_pf_get_num_tc(pf)) {
3806 		netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
3807 		goto exit;
3808 	}
3809 
3810 	/* Generate TC map for number of tc requested */
3811 	for (i = 0; i < tc; i++)
3812 		enabled_tc |= (1 << i);
3813 
3814 	/* Requesting same TC configuration as already enabled */
3815 	if (enabled_tc == vsi->tc_config.enabled_tc)
3816 		return 0;
3817 
3818 	/* Quiesce VSI queues */
3819 	i40e_quiesce_vsi(vsi);
3820 
3821 	/* Configure VSI for enabled TCs */
3822 	ret = i40e_vsi_config_tc(vsi, enabled_tc);
3823 	if (ret) {
3824 		netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
3825 			    vsi->seid);
3826 		goto exit;
3827 	}
3828 
3829 	/* Unquiesce VSI */
3830 	i40e_unquiesce_vsi(vsi);
3831 
3832 exit:
3833 	return ret;
3834 }
3835 
3836 /**
3837  * i40e_open - Called when a network interface is made active
3838  * @netdev: network interface device structure
3839  *
3840  * The open entry point is called when a network interface is made
3841  * active by the system (IFF_UP).  At this point all resources needed
3842  * for transmit and receive operations are allocated, the interrupt
3843  * handler is registered with the OS, the netdev watchdog subtask is
3844  * enabled, and the stack is notified that the interface is ready.
3845  *
3846  * Returns 0 on success, negative value on failure
3847  **/
3848 static int i40e_open(struct net_device *netdev)
3849 {
3850 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3851 	struct i40e_vsi *vsi = np->vsi;
3852 	struct i40e_pf *pf = vsi->back;
3853 	char int_name[IFNAMSIZ];
3854 	int err;
3855 
3856 	/* disallow open during test */
3857 	if (test_bit(__I40E_TESTING, &pf->state))
3858 		return -EBUSY;
3859 
3860 	netif_carrier_off(netdev);
3861 
3862 	/* allocate descriptors */
3863 	err = i40e_vsi_setup_tx_resources(vsi);
3864 	if (err)
3865 		goto err_setup_tx;
3866 	err = i40e_vsi_setup_rx_resources(vsi);
3867 	if (err)
3868 		goto err_setup_rx;
3869 
3870 	err = i40e_vsi_configure(vsi);
3871 	if (err)
3872 		goto err_setup_rx;
3873 
3874 	snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
3875 		 dev_driver_string(&pf->pdev->dev), netdev->name);
3876 	err = i40e_vsi_request_irq(vsi, int_name);
3877 	if (err)
3878 		goto err_setup_rx;
3879 
3880 	err = i40e_up_complete(vsi);
3881 	if (err)
3882 		goto err_up_complete;
3883 
3884 	if ((vsi->type == I40E_VSI_MAIN) || (vsi->type == I40E_VSI_VMDQ2)) {
3885 		err = i40e_aq_set_vsi_broadcast(&pf->hw, vsi->seid, true, NULL);
3886 		if (err)
3887 			netdev_info(netdev,
3888 				    "couldn't set broadcast err %d aq_err %d\n",
3889 				    err, pf->hw.aq.asq_last_status);
3890 	}
3891 
3892 	return 0;
3893 
3894 err_up_complete:
3895 	i40e_down(vsi);
3896 	i40e_vsi_free_irq(vsi);
3897 err_setup_rx:
3898 	i40e_vsi_free_rx_resources(vsi);
3899 err_setup_tx:
3900 	i40e_vsi_free_tx_resources(vsi);
3901 	if (vsi == pf->vsi[pf->lan_vsi])
3902 		i40e_do_reset(pf, (1 << __I40E_PF_RESET_REQUESTED));
3903 
3904 	return err;
3905 }
3906 
3907 /**
3908  * i40e_close - Disables a network interface
3909  * @netdev: network interface device structure
3910  *
3911  * The close entry point is called when an interface is de-activated
3912  * by the OS.  The hardware is still under the driver's control, but
3913  * this netdev interface is disabled.
3914  *
3915  * Returns 0, this is not allowed to fail
3916  **/
3917 static int i40e_close(struct net_device *netdev)
3918 {
3919 	struct i40e_netdev_priv *np = netdev_priv(netdev);
3920 	struct i40e_vsi *vsi = np->vsi;
3921 
3922 	if (test_and_set_bit(__I40E_DOWN, &vsi->state))
3923 		return 0;
3924 
3925 	i40e_down(vsi);
3926 	i40e_vsi_free_irq(vsi);
3927 
3928 	i40e_vsi_free_tx_resources(vsi);
3929 	i40e_vsi_free_rx_resources(vsi);
3930 
3931 	return 0;
3932 }
3933 
3934 /**
3935  * i40e_do_reset - Start a PF or Core Reset sequence
3936  * @pf: board private structure
3937  * @reset_flags: which reset is requested
3938  *
3939  * The essential difference in resets is that the PF Reset
3940  * doesn't clear the packet buffers, doesn't reset the PE
3941  * firmware, and doesn't bother the other PFs on the chip.
3942  **/
3943 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
3944 {
3945 	u32 val;
3946 
3947 	WARN_ON(in_interrupt());
3948 
3949 	/* do the biggest reset indicated */
3950 	if (reset_flags & (1 << __I40E_GLOBAL_RESET_REQUESTED)) {
3951 
3952 		/* Request a Global Reset
3953 		 *
3954 		 * This will start the chip's countdown to the actual full
3955 		 * chip reset event, and a warning interrupt to be sent
3956 		 * to all PFs, including the requestor.  Our handler
3957 		 * for the warning interrupt will deal with the shutdown
3958 		 * and recovery of the switch setup.
3959 		 */
3960 		dev_info(&pf->pdev->dev, "GlobalR requested\n");
3961 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
3962 		val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
3963 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
3964 
3965 	} else if (reset_flags & (1 << __I40E_CORE_RESET_REQUESTED)) {
3966 
3967 		/* Request a Core Reset
3968 		 *
3969 		 * Same as Global Reset, except does *not* include the MAC/PHY
3970 		 */
3971 		dev_info(&pf->pdev->dev, "CoreR requested\n");
3972 		val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
3973 		val |= I40E_GLGEN_RTRIG_CORER_MASK;
3974 		wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
3975 		i40e_flush(&pf->hw);
3976 
3977 	} else if (reset_flags & (1 << __I40E_PF_RESET_REQUESTED)) {
3978 
3979 		/* Request a PF Reset
3980 		 *
3981 		 * Resets only the PF-specific registers
3982 		 *
3983 		 * This goes directly to the tear-down and rebuild of
3984 		 * the switch, since we need to do all the recovery as
3985 		 * for the Core Reset.
3986 		 */
3987 		dev_info(&pf->pdev->dev, "PFR requested\n");
3988 		i40e_handle_reset_warning(pf);
3989 
3990 	} else if (reset_flags & (1 << __I40E_REINIT_REQUESTED)) {
3991 		int v;
3992 
3993 		/* Find the VSI(s) that requested a re-init */
3994 		dev_info(&pf->pdev->dev,
3995 			 "VSI reinit requested\n");
3996 		for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
3997 			struct i40e_vsi *vsi = pf->vsi[v];
3998 			if (vsi != NULL &&
3999 			    test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
4000 				i40e_vsi_reinit_locked(pf->vsi[v]);
4001 				clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
4002 			}
4003 		}
4004 
4005 		/* no further action needed, so return now */
4006 		return;
4007 	} else {
4008 		dev_info(&pf->pdev->dev,
4009 			 "bad reset request 0x%08x\n", reset_flags);
4010 		return;
4011 	}
4012 }
4013 
4014 /**
4015  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
4016  * @pf: board private structure
4017  * @e: event info posted on ARQ
4018  *
4019  * Handler for LAN Queue Overflow Event generated by the firmware for PF
4020  * and VF queues
4021  **/
4022 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
4023 					   struct i40e_arq_event_info *e)
4024 {
4025 	struct i40e_aqc_lan_overflow *data =
4026 		(struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
4027 	u32 queue = le32_to_cpu(data->prtdcb_rupto);
4028 	u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
4029 	struct i40e_hw *hw = &pf->hw;
4030 	struct i40e_vf *vf;
4031 	u16 vf_id;
4032 
4033 	dev_info(&pf->pdev->dev, "%s: Rx Queue Number = %d QTX_CTL=0x%08x\n",
4034 		 __func__, queue, qtx_ctl);
4035 
4036 	/* Queue belongs to VF, find the VF and issue VF reset */
4037 	if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
4038 	    >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
4039 		vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
4040 			 >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
4041 		vf_id -= hw->func_caps.vf_base_id;
4042 		vf = &pf->vf[vf_id];
4043 		i40e_vc_notify_vf_reset(vf);
4044 		/* Allow VF to process pending reset notification */
4045 		msleep(20);
4046 		i40e_reset_vf(vf, false);
4047 	}
4048 }
4049 
4050 /**
4051  * i40e_service_event_complete - Finish up the service event
4052  * @pf: board private structure
4053  **/
4054 static void i40e_service_event_complete(struct i40e_pf *pf)
4055 {
4056 	BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
4057 
4058 	/* flush memory to make sure state is correct before next watchog */
4059 	smp_mb__before_clear_bit();
4060 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
4061 }
4062 
4063 /**
4064  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
4065  * @pf: board private structure
4066  **/
4067 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
4068 {
4069 	if (!(pf->flags & I40E_FLAG_FDIR_REQUIRES_REINIT))
4070 		return;
4071 
4072 	pf->flags &= ~I40E_FLAG_FDIR_REQUIRES_REINIT;
4073 
4074 	/* if interface is down do nothing */
4075 	if (test_bit(__I40E_DOWN, &pf->state))
4076 		return;
4077 }
4078 
4079 /**
4080  * i40e_vsi_link_event - notify VSI of a link event
4081  * @vsi: vsi to be notified
4082  * @link_up: link up or down
4083  **/
4084 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
4085 {
4086 	if (!vsi)
4087 		return;
4088 
4089 	switch (vsi->type) {
4090 	case I40E_VSI_MAIN:
4091 		if (!vsi->netdev || !vsi->netdev_registered)
4092 			break;
4093 
4094 		if (link_up) {
4095 			netif_carrier_on(vsi->netdev);
4096 			netif_tx_wake_all_queues(vsi->netdev);
4097 		} else {
4098 			netif_carrier_off(vsi->netdev);
4099 			netif_tx_stop_all_queues(vsi->netdev);
4100 		}
4101 		break;
4102 
4103 	case I40E_VSI_SRIOV:
4104 		break;
4105 
4106 	case I40E_VSI_VMDQ2:
4107 	case I40E_VSI_CTRL:
4108 	case I40E_VSI_MIRROR:
4109 	default:
4110 		/* there is no notification for other VSIs */
4111 		break;
4112 	}
4113 }
4114 
4115 /**
4116  * i40e_veb_link_event - notify elements on the veb of a link event
4117  * @veb: veb to be notified
4118  * @link_up: link up or down
4119  **/
4120 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
4121 {
4122 	struct i40e_pf *pf;
4123 	int i;
4124 
4125 	if (!veb || !veb->pf)
4126 		return;
4127 	pf = veb->pf;
4128 
4129 	/* depth first... */
4130 	for (i = 0; i < I40E_MAX_VEB; i++)
4131 		if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
4132 			i40e_veb_link_event(pf->veb[i], link_up);
4133 
4134 	/* ... now the local VSIs */
4135 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4136 		if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
4137 			i40e_vsi_link_event(pf->vsi[i], link_up);
4138 }
4139 
4140 /**
4141  * i40e_link_event - Update netif_carrier status
4142  * @pf: board private structure
4143  **/
4144 static void i40e_link_event(struct i40e_pf *pf)
4145 {
4146 	bool new_link, old_link;
4147 
4148 	new_link = (pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP);
4149 	old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
4150 
4151 	if (new_link == old_link)
4152 		return;
4153 
4154 	netdev_info(pf->vsi[pf->lan_vsi]->netdev,
4155 		    "NIC Link is %s\n", (new_link ? "Up" : "Down"));
4156 
4157 	/* Notify the base of the switch tree connected to
4158 	 * the link.  Floating VEBs are not notified.
4159 	 */
4160 	if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
4161 		i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
4162 	else
4163 		i40e_vsi_link_event(pf->vsi[pf->lan_vsi], new_link);
4164 
4165 	if (pf->vf)
4166 		i40e_vc_notify_link_state(pf);
4167 }
4168 
4169 /**
4170  * i40e_check_hang_subtask - Check for hung queues and dropped interrupts
4171  * @pf: board private structure
4172  *
4173  * Set the per-queue flags to request a check for stuck queues in the irq
4174  * clean functions, then force interrupts to be sure the irq clean is called.
4175  **/
4176 static void i40e_check_hang_subtask(struct i40e_pf *pf)
4177 {
4178 	int i, v;
4179 
4180 	/* If we're down or resetting, just bail */
4181 	if (test_bit(__I40E_CONFIG_BUSY, &pf->state))
4182 		return;
4183 
4184 	/* for each VSI/netdev
4185 	 *     for each Tx queue
4186 	 *         set the check flag
4187 	 *     for each q_vector
4188 	 *         force an interrupt
4189 	 */
4190 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4191 		struct i40e_vsi *vsi = pf->vsi[v];
4192 		int armed = 0;
4193 
4194 		if (!pf->vsi[v] ||
4195 		    test_bit(__I40E_DOWN, &vsi->state) ||
4196 		    (vsi->netdev && !netif_carrier_ok(vsi->netdev)))
4197 			continue;
4198 
4199 		for (i = 0; i < vsi->num_queue_pairs; i++) {
4200 			set_check_for_tx_hang(&vsi->tx_rings[i]);
4201 			if (test_bit(__I40E_HANG_CHECK_ARMED,
4202 				     &vsi->tx_rings[i].state))
4203 				armed++;
4204 		}
4205 
4206 		if (armed) {
4207 			if (!(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
4208 				wr32(&vsi->back->hw, I40E_PFINT_DYN_CTL0,
4209 				     (I40E_PFINT_DYN_CTL0_INTENA_MASK |
4210 				      I40E_PFINT_DYN_CTL0_SWINT_TRIG_MASK));
4211 			} else {
4212 				u16 vec = vsi->base_vector - 1;
4213 				u32 val = (I40E_PFINT_DYN_CTLN_INTENA_MASK |
4214 					   I40E_PFINT_DYN_CTLN_SWINT_TRIG_MASK);
4215 				for (i = 0; i < vsi->num_q_vectors; i++, vec++)
4216 					wr32(&vsi->back->hw,
4217 					     I40E_PFINT_DYN_CTLN(vec), val);
4218 			}
4219 			i40e_flush(&vsi->back->hw);
4220 		}
4221 	}
4222 }
4223 
4224 /**
4225  * i40e_watchdog_subtask - Check and bring link up
4226  * @pf: board private structure
4227  **/
4228 static void i40e_watchdog_subtask(struct i40e_pf *pf)
4229 {
4230 	int i;
4231 
4232 	/* if interface is down do nothing */
4233 	if (test_bit(__I40E_DOWN, &pf->state) ||
4234 	    test_bit(__I40E_CONFIG_BUSY, &pf->state))
4235 		return;
4236 
4237 	/* Update the stats for active netdevs so the network stack
4238 	 * can look at updated numbers whenever it cares to
4239 	 */
4240 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4241 		if (pf->vsi[i] && pf->vsi[i]->netdev)
4242 			i40e_update_stats(pf->vsi[i]);
4243 
4244 	/* Update the stats for the active switching components */
4245 	for (i = 0; i < I40E_MAX_VEB; i++)
4246 		if (pf->veb[i])
4247 			i40e_update_veb_stats(pf->veb[i]);
4248 }
4249 
4250 /**
4251  * i40e_reset_subtask - Set up for resetting the device and driver
4252  * @pf: board private structure
4253  **/
4254 static void i40e_reset_subtask(struct i40e_pf *pf)
4255 {
4256 	u32 reset_flags = 0;
4257 
4258 	if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
4259 		reset_flags |= (1 << __I40E_REINIT_REQUESTED);
4260 		clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
4261 	}
4262 	if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
4263 		reset_flags |= (1 << __I40E_PF_RESET_REQUESTED);
4264 		clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
4265 	}
4266 	if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
4267 		reset_flags |= (1 << __I40E_CORE_RESET_REQUESTED);
4268 		clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
4269 	}
4270 	if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
4271 		reset_flags |= (1 << __I40E_GLOBAL_RESET_REQUESTED);
4272 		clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
4273 	}
4274 
4275 	/* If there's a recovery already waiting, it takes
4276 	 * precedence before starting a new reset sequence.
4277 	 */
4278 	if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
4279 		i40e_handle_reset_warning(pf);
4280 		return;
4281 	}
4282 
4283 	/* If we're already down or resetting, just bail */
4284 	if (reset_flags &&
4285 	    !test_bit(__I40E_DOWN, &pf->state) &&
4286 	    !test_bit(__I40E_CONFIG_BUSY, &pf->state))
4287 		i40e_do_reset(pf, reset_flags);
4288 }
4289 
4290 /**
4291  * i40e_handle_link_event - Handle link event
4292  * @pf: board private structure
4293  * @e: event info posted on ARQ
4294  **/
4295 static void i40e_handle_link_event(struct i40e_pf *pf,
4296 				   struct i40e_arq_event_info *e)
4297 {
4298 	struct i40e_hw *hw = &pf->hw;
4299 	struct i40e_aqc_get_link_status *status =
4300 		(struct i40e_aqc_get_link_status *)&e->desc.params.raw;
4301 	struct i40e_link_status *hw_link_info = &hw->phy.link_info;
4302 
4303 	/* save off old link status information */
4304 	memcpy(&pf->hw.phy.link_info_old, hw_link_info,
4305 	       sizeof(pf->hw.phy.link_info_old));
4306 
4307 	/* update link status */
4308 	hw_link_info->phy_type = (enum i40e_aq_phy_type)status->phy_type;
4309 	hw_link_info->link_speed = (enum i40e_aq_link_speed)status->link_speed;
4310 	hw_link_info->link_info = status->link_info;
4311 	hw_link_info->an_info = status->an_info;
4312 	hw_link_info->ext_info = status->ext_info;
4313 	hw_link_info->lse_enable =
4314 		le16_to_cpu(status->command_flags) &
4315 			    I40E_AQ_LSE_ENABLE;
4316 
4317 	/* process the event */
4318 	i40e_link_event(pf);
4319 
4320 	/* Do a new status request to re-enable LSE reporting
4321 	 * and load new status information into the hw struct,
4322 	 * then see if the status changed while processing the
4323 	 * initial event.
4324 	 */
4325 	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
4326 	i40e_link_event(pf);
4327 }
4328 
4329 /**
4330  * i40e_clean_adminq_subtask - Clean the AdminQ rings
4331  * @pf: board private structure
4332  **/
4333 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
4334 {
4335 	struct i40e_arq_event_info event;
4336 	struct i40e_hw *hw = &pf->hw;
4337 	u16 pending, i = 0;
4338 	i40e_status ret;
4339 	u16 opcode;
4340 	u32 val;
4341 
4342 	if (!test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state))
4343 		return;
4344 
4345 	event.msg_size = I40E_MAX_AQ_BUF_SIZE;
4346 	event.msg_buf = kzalloc(event.msg_size, GFP_KERNEL);
4347 	if (!event.msg_buf)
4348 		return;
4349 
4350 	do {
4351 		ret = i40e_clean_arq_element(hw, &event, &pending);
4352 		if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK) {
4353 			dev_info(&pf->pdev->dev, "No ARQ event found\n");
4354 			break;
4355 		} else if (ret) {
4356 			dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
4357 			break;
4358 		}
4359 
4360 		opcode = le16_to_cpu(event.desc.opcode);
4361 		switch (opcode) {
4362 
4363 		case i40e_aqc_opc_get_link_status:
4364 			i40e_handle_link_event(pf, &event);
4365 			break;
4366 		case i40e_aqc_opc_send_msg_to_pf:
4367 			ret = i40e_vc_process_vf_msg(pf,
4368 					le16_to_cpu(event.desc.retval),
4369 					le32_to_cpu(event.desc.cookie_high),
4370 					le32_to_cpu(event.desc.cookie_low),
4371 					event.msg_buf,
4372 					event.msg_size);
4373 			break;
4374 		case i40e_aqc_opc_lldp_update_mib:
4375 			dev_info(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
4376 			break;
4377 		case i40e_aqc_opc_event_lan_overflow:
4378 			dev_info(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
4379 			i40e_handle_lan_overflow_event(pf, &event);
4380 			break;
4381 		default:
4382 			dev_info(&pf->pdev->dev,
4383 				 "ARQ Error: Unknown event %d received\n",
4384 				 event.desc.opcode);
4385 			break;
4386 		}
4387 	} while (pending && (i++ < pf->adminq_work_limit));
4388 
4389 	clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
4390 	/* re-enable Admin queue interrupt cause */
4391 	val = rd32(hw, I40E_PFINT_ICR0_ENA);
4392 	val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
4393 	wr32(hw, I40E_PFINT_ICR0_ENA, val);
4394 	i40e_flush(hw);
4395 
4396 	kfree(event.msg_buf);
4397 }
4398 
4399 /**
4400  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
4401  * @veb: pointer to the VEB instance
4402  *
4403  * This is a recursive function that first builds the attached VSIs then
4404  * recurses in to build the next layer of VEB.  We track the connections
4405  * through our own index numbers because the seid's from the HW could
4406  * change across the reset.
4407  **/
4408 static int i40e_reconstitute_veb(struct i40e_veb *veb)
4409 {
4410 	struct i40e_vsi *ctl_vsi = NULL;
4411 	struct i40e_pf *pf = veb->pf;
4412 	int v, veb_idx;
4413 	int ret;
4414 
4415 	/* build VSI that owns this VEB, temporarily attached to base VEB */
4416 	for (v = 0; v < pf->hw.func_caps.num_vsis && !ctl_vsi; v++) {
4417 		if (pf->vsi[v] &&
4418 		    pf->vsi[v]->veb_idx == veb->idx &&
4419 		    pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
4420 			ctl_vsi = pf->vsi[v];
4421 			break;
4422 		}
4423 	}
4424 	if (!ctl_vsi) {
4425 		dev_info(&pf->pdev->dev,
4426 			 "missing owner VSI for veb_idx %d\n", veb->idx);
4427 		ret = -ENOENT;
4428 		goto end_reconstitute;
4429 	}
4430 	if (ctl_vsi != pf->vsi[pf->lan_vsi])
4431 		ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
4432 	ret = i40e_add_vsi(ctl_vsi);
4433 	if (ret) {
4434 		dev_info(&pf->pdev->dev,
4435 			 "rebuild of owner VSI failed: %d\n", ret);
4436 		goto end_reconstitute;
4437 	}
4438 	i40e_vsi_reset_stats(ctl_vsi);
4439 
4440 	/* create the VEB in the switch and move the VSI onto the VEB */
4441 	ret = i40e_add_veb(veb, ctl_vsi);
4442 	if (ret)
4443 		goto end_reconstitute;
4444 
4445 	/* create the remaining VSIs attached to this VEB */
4446 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4447 		if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
4448 			continue;
4449 
4450 		if (pf->vsi[v]->veb_idx == veb->idx) {
4451 			struct i40e_vsi *vsi = pf->vsi[v];
4452 			vsi->uplink_seid = veb->seid;
4453 			ret = i40e_add_vsi(vsi);
4454 			if (ret) {
4455 				dev_info(&pf->pdev->dev,
4456 					 "rebuild of vsi_idx %d failed: %d\n",
4457 					 v, ret);
4458 				goto end_reconstitute;
4459 			}
4460 			i40e_vsi_reset_stats(vsi);
4461 		}
4462 	}
4463 
4464 	/* create any VEBs attached to this VEB - RECURSION */
4465 	for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
4466 		if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
4467 			pf->veb[veb_idx]->uplink_seid = veb->seid;
4468 			ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
4469 			if (ret)
4470 				break;
4471 		}
4472 	}
4473 
4474 end_reconstitute:
4475 	return ret;
4476 }
4477 
4478 /**
4479  * i40e_get_capabilities - get info about the HW
4480  * @pf: the PF struct
4481  **/
4482 static int i40e_get_capabilities(struct i40e_pf *pf)
4483 {
4484 	struct i40e_aqc_list_capabilities_element_resp *cap_buf;
4485 	u16 data_size;
4486 	int buf_len;
4487 	int err;
4488 
4489 	buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
4490 	do {
4491 		cap_buf = kzalloc(buf_len, GFP_KERNEL);
4492 		if (!cap_buf)
4493 			return -ENOMEM;
4494 
4495 		/* this loads the data into the hw struct for us */
4496 		err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
4497 					    &data_size,
4498 					    i40e_aqc_opc_list_func_capabilities,
4499 					    NULL);
4500 		/* data loaded, buffer no longer needed */
4501 		kfree(cap_buf);
4502 
4503 		if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
4504 			/* retry with a larger buffer */
4505 			buf_len = data_size;
4506 		} else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
4507 			dev_info(&pf->pdev->dev,
4508 				 "capability discovery failed: aq=%d\n",
4509 				 pf->hw.aq.asq_last_status);
4510 			return -ENODEV;
4511 		}
4512 	} while (err);
4513 
4514 	if (pf->hw.debug_mask & I40E_DEBUG_USER)
4515 		dev_info(&pf->pdev->dev,
4516 			 "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",
4517 			 pf->hw.pf_id, pf->hw.func_caps.num_vfs,
4518 			 pf->hw.func_caps.num_msix_vectors,
4519 			 pf->hw.func_caps.num_msix_vectors_vf,
4520 			 pf->hw.func_caps.fd_filters_guaranteed,
4521 			 pf->hw.func_caps.fd_filters_best_effort,
4522 			 pf->hw.func_caps.num_tx_qp,
4523 			 pf->hw.func_caps.num_vsis);
4524 
4525 	return 0;
4526 }
4527 
4528 /**
4529  * i40e_fdir_setup - initialize the Flow Director resources
4530  * @pf: board private structure
4531  **/
4532 static void i40e_fdir_setup(struct i40e_pf *pf)
4533 {
4534 	struct i40e_vsi *vsi;
4535 	bool new_vsi = false;
4536 	int err, i;
4537 
4538 	if (!(pf->flags & (I40E_FLAG_FDIR_ENABLED|I40E_FLAG_FDIR_ATR_ENABLED)))
4539 		return;
4540 
4541 	pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
4542 
4543 	/* find existing or make new FDIR VSI */
4544 	vsi = NULL;
4545 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++)
4546 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR)
4547 			vsi = pf->vsi[i];
4548 	if (!vsi) {
4549 		vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR, pf->mac_seid, 0);
4550 		if (!vsi) {
4551 			dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
4552 			pf->flags &= ~I40E_FLAG_FDIR_ENABLED;
4553 			return;
4554 		}
4555 		new_vsi = true;
4556 	}
4557 	WARN_ON(vsi->base_queue != I40E_FDIR_RING);
4558 	i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_rings);
4559 
4560 	err = i40e_vsi_setup_tx_resources(vsi);
4561 	if (!err)
4562 		err = i40e_vsi_setup_rx_resources(vsi);
4563 	if (!err)
4564 		err = i40e_vsi_configure(vsi);
4565 	if (!err && new_vsi) {
4566 		char int_name[IFNAMSIZ + 9];
4567 		snprintf(int_name, sizeof(int_name) - 1, "%s-fdir",
4568 			 dev_driver_string(&pf->pdev->dev));
4569 		err = i40e_vsi_request_irq(vsi, int_name);
4570 	}
4571 	if (!err)
4572 		err = i40e_up_complete(vsi);
4573 
4574 	clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4575 }
4576 
4577 /**
4578  * i40e_fdir_teardown - release the Flow Director resources
4579  * @pf: board private structure
4580  **/
4581 static void i40e_fdir_teardown(struct i40e_pf *pf)
4582 {
4583 	int i;
4584 
4585 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
4586 		if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
4587 			i40e_vsi_release(pf->vsi[i]);
4588 			break;
4589 		}
4590 	}
4591 }
4592 
4593 /**
4594  * i40e_handle_reset_warning - prep for the core to reset
4595  * @pf: board private structure
4596  *
4597  * Close up the VFs and other things in prep for a Core Reset,
4598  * then get ready to rebuild the world.
4599  **/
4600 static void i40e_handle_reset_warning(struct i40e_pf *pf)
4601 {
4602 	struct i40e_driver_version dv;
4603 	struct i40e_hw *hw = &pf->hw;
4604 	i40e_status ret;
4605 	u32 v;
4606 
4607 	clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
4608 	if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
4609 		return;
4610 
4611 	dev_info(&pf->pdev->dev, "Tearing down internal switch for reset\n");
4612 
4613 	i40e_vc_notify_reset(pf);
4614 
4615 	/* quiesce the VSIs and their queues that are not already DOWN */
4616 	i40e_pf_quiesce_all_vsi(pf);
4617 
4618 	for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4619 		if (pf->vsi[v])
4620 			pf->vsi[v]->seid = 0;
4621 	}
4622 
4623 	i40e_shutdown_adminq(&pf->hw);
4624 
4625 	/* Now we wait for GRST to settle out.
4626 	 * We don't have to delete the VEBs or VSIs from the hw switch
4627 	 * because the reset will make them disappear.
4628 	 */
4629 	ret = i40e_pf_reset(hw);
4630 	if (ret)
4631 		dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
4632 	pf->pfr_count++;
4633 
4634 	if (test_bit(__I40E_DOWN, &pf->state))
4635 		goto end_core_reset;
4636 	dev_info(&pf->pdev->dev, "Rebuilding internal switch\n");
4637 
4638 	/* rebuild the basics for the AdminQ, HMC, and initial HW switch */
4639 	ret = i40e_init_adminq(&pf->hw);
4640 	if (ret) {
4641 		dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, %d\n", ret);
4642 		goto end_core_reset;
4643 	}
4644 
4645 	ret = i40e_get_capabilities(pf);
4646 	if (ret) {
4647 		dev_info(&pf->pdev->dev, "i40e_get_capabilities failed, %d\n",
4648 			 ret);
4649 		goto end_core_reset;
4650 	}
4651 
4652 	/* call shutdown HMC */
4653 	ret = i40e_shutdown_lan_hmc(hw);
4654 	if (ret) {
4655 		dev_info(&pf->pdev->dev, "shutdown_lan_hmc failed: %d\n", ret);
4656 		goto end_core_reset;
4657 	}
4658 
4659 	ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
4660 				hw->func_caps.num_rx_qp,
4661 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
4662 	if (ret) {
4663 		dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
4664 		goto end_core_reset;
4665 	}
4666 	ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
4667 	if (ret) {
4668 		dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
4669 		goto end_core_reset;
4670 	}
4671 
4672 	/* do basic switch setup */
4673 	ret = i40e_setup_pf_switch(pf);
4674 	if (ret)
4675 		goto end_core_reset;
4676 
4677 	/* Rebuild the VSIs and VEBs that existed before reset.
4678 	 * They are still in our local switch element arrays, so only
4679 	 * need to rebuild the switch model in the HW.
4680 	 *
4681 	 * If there were VEBs but the reconstitution failed, we'll try
4682 	 * try to recover minimal use by getting the basic PF VSI working.
4683 	 */
4684 	if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
4685 		dev_info(&pf->pdev->dev, "attempting to rebuild switch\n");
4686 		/* find the one VEB connected to the MAC, and find orphans */
4687 		for (v = 0; v < I40E_MAX_VEB; v++) {
4688 			if (!pf->veb[v])
4689 				continue;
4690 
4691 			if (pf->veb[v]->uplink_seid == pf->mac_seid ||
4692 			    pf->veb[v]->uplink_seid == 0) {
4693 				ret = i40e_reconstitute_veb(pf->veb[v]);
4694 
4695 				if (!ret)
4696 					continue;
4697 
4698 				/* If Main VEB failed, we're in deep doodoo,
4699 				 * so give up rebuilding the switch and set up
4700 				 * for minimal rebuild of PF VSI.
4701 				 * If orphan failed, we'll report the error
4702 				 * but try to keep going.
4703 				 */
4704 				if (pf->veb[v]->uplink_seid == pf->mac_seid) {
4705 					dev_info(&pf->pdev->dev,
4706 						 "rebuild of switch failed: %d, will try to set up simple PF connection\n",
4707 						 ret);
4708 					pf->vsi[pf->lan_vsi]->uplink_seid
4709 								= pf->mac_seid;
4710 					break;
4711 				} else if (pf->veb[v]->uplink_seid == 0) {
4712 					dev_info(&pf->pdev->dev,
4713 						 "rebuild of orphan VEB failed: %d\n",
4714 						 ret);
4715 				}
4716 			}
4717 		}
4718 	}
4719 
4720 	if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
4721 		dev_info(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
4722 		/* no VEB, so rebuild only the Main VSI */
4723 		ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
4724 		if (ret) {
4725 			dev_info(&pf->pdev->dev,
4726 				 "rebuild of Main VSI failed: %d\n", ret);
4727 			goto end_core_reset;
4728 		}
4729 	}
4730 
4731 	/* reinit the misc interrupt */
4732 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4733 		ret = i40e_setup_misc_vector(pf);
4734 
4735 	/* restart the VSIs that were rebuilt and running before the reset */
4736 	i40e_pf_unquiesce_all_vsi(pf);
4737 
4738 	/* tell the firmware that we're starting */
4739 	dv.major_version = DRV_VERSION_MAJOR;
4740 	dv.minor_version = DRV_VERSION_MINOR;
4741 	dv.build_version = DRV_VERSION_BUILD;
4742 	dv.subbuild_version = 0;
4743 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
4744 
4745 	dev_info(&pf->pdev->dev, "PF reset done\n");
4746 
4747 end_core_reset:
4748 	clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
4749 }
4750 
4751 /**
4752  * i40e_handle_mdd_event
4753  * @pf: pointer to the pf structure
4754  *
4755  * Called from the MDD irq handler to identify possibly malicious vfs
4756  **/
4757 static void i40e_handle_mdd_event(struct i40e_pf *pf)
4758 {
4759 	struct i40e_hw *hw = &pf->hw;
4760 	bool mdd_detected = false;
4761 	struct i40e_vf *vf;
4762 	u32 reg;
4763 	int i;
4764 
4765 	if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
4766 		return;
4767 
4768 	/* find what triggered the MDD event */
4769 	reg = rd32(hw, I40E_GL_MDET_TX);
4770 	if (reg & I40E_GL_MDET_TX_VALID_MASK) {
4771 		u8 func = (reg & I40E_GL_MDET_TX_FUNCTION_MASK)
4772 				>> I40E_GL_MDET_TX_FUNCTION_SHIFT;
4773 		u8 event = (reg & I40E_GL_MDET_TX_EVENT_SHIFT)
4774 				>> I40E_GL_MDET_TX_EVENT_SHIFT;
4775 		u8 queue = (reg & I40E_GL_MDET_TX_QUEUE_MASK)
4776 				>> I40E_GL_MDET_TX_QUEUE_SHIFT;
4777 		dev_info(&pf->pdev->dev,
4778 			 "Malicious Driver Detection TX event 0x%02x on q %d of function 0x%02x\n",
4779 			 event, queue, func);
4780 		wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
4781 		mdd_detected = true;
4782 	}
4783 	reg = rd32(hw, I40E_GL_MDET_RX);
4784 	if (reg & I40E_GL_MDET_RX_VALID_MASK) {
4785 		u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK)
4786 				>> I40E_GL_MDET_RX_FUNCTION_SHIFT;
4787 		u8 event = (reg & I40E_GL_MDET_RX_EVENT_SHIFT)
4788 				>> I40E_GL_MDET_RX_EVENT_SHIFT;
4789 		u8 queue = (reg & I40E_GL_MDET_RX_QUEUE_MASK)
4790 				>> I40E_GL_MDET_RX_QUEUE_SHIFT;
4791 		dev_info(&pf->pdev->dev,
4792 			 "Malicious Driver Detection RX event 0x%02x on q %d of function 0x%02x\n",
4793 			 event, queue, func);
4794 		wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
4795 		mdd_detected = true;
4796 	}
4797 
4798 	/* see if one of the VFs needs its hand slapped */
4799 	for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
4800 		vf = &(pf->vf[i]);
4801 		reg = rd32(hw, I40E_VP_MDET_TX(i));
4802 		if (reg & I40E_VP_MDET_TX_VALID_MASK) {
4803 			wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
4804 			vf->num_mdd_events++;
4805 			dev_info(&pf->pdev->dev, "MDD TX event on VF %d\n", i);
4806 		}
4807 
4808 		reg = rd32(hw, I40E_VP_MDET_RX(i));
4809 		if (reg & I40E_VP_MDET_RX_VALID_MASK) {
4810 			wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
4811 			vf->num_mdd_events++;
4812 			dev_info(&pf->pdev->dev, "MDD RX event on VF %d\n", i);
4813 		}
4814 
4815 		if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
4816 			dev_info(&pf->pdev->dev,
4817 				 "Too many MDD events on VF %d, disabled\n", i);
4818 			dev_info(&pf->pdev->dev,
4819 				 "Use PF Control I/F to re-enable the VF\n");
4820 			set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
4821 		}
4822 	}
4823 
4824 	/* re-enable mdd interrupt cause */
4825 	clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
4826 	reg = rd32(hw, I40E_PFINT_ICR0_ENA);
4827 	reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
4828 	wr32(hw, I40E_PFINT_ICR0_ENA, reg);
4829 	i40e_flush(hw);
4830 }
4831 
4832 /**
4833  * i40e_service_task - Run the driver's async subtasks
4834  * @work: pointer to work_struct containing our data
4835  **/
4836 static void i40e_service_task(struct work_struct *work)
4837 {
4838 	struct i40e_pf *pf = container_of(work,
4839 					  struct i40e_pf,
4840 					  service_task);
4841 	unsigned long start_time = jiffies;
4842 
4843 	i40e_reset_subtask(pf);
4844 	i40e_handle_mdd_event(pf);
4845 	i40e_vc_process_vflr_event(pf);
4846 	i40e_watchdog_subtask(pf);
4847 	i40e_fdir_reinit_subtask(pf);
4848 	i40e_check_hang_subtask(pf);
4849 	i40e_sync_filters_subtask(pf);
4850 	i40e_clean_adminq_subtask(pf);
4851 
4852 	i40e_service_event_complete(pf);
4853 
4854 	/* If the tasks have taken longer than one timer cycle or there
4855 	 * is more work to be done, reschedule the service task now
4856 	 * rather than wait for the timer to tick again.
4857 	 */
4858 	if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
4859 	    test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)		 ||
4860 	    test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)		 ||
4861 	    test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
4862 		i40e_service_event_schedule(pf);
4863 }
4864 
4865 /**
4866  * i40e_service_timer - timer callback
4867  * @data: pointer to PF struct
4868  **/
4869 static void i40e_service_timer(unsigned long data)
4870 {
4871 	struct i40e_pf *pf = (struct i40e_pf *)data;
4872 
4873 	mod_timer(&pf->service_timer,
4874 		  round_jiffies(jiffies + pf->service_timer_period));
4875 	i40e_service_event_schedule(pf);
4876 }
4877 
4878 /**
4879  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
4880  * @vsi: the VSI being configured
4881  **/
4882 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
4883 {
4884 	struct i40e_pf *pf = vsi->back;
4885 
4886 	switch (vsi->type) {
4887 	case I40E_VSI_MAIN:
4888 		vsi->alloc_queue_pairs = pf->num_lan_qps;
4889 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4890 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4891 		if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4892 			vsi->num_q_vectors = pf->num_lan_msix;
4893 		else
4894 			vsi->num_q_vectors = 1;
4895 
4896 		break;
4897 
4898 	case I40E_VSI_FDIR:
4899 		vsi->alloc_queue_pairs = 1;
4900 		vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
4901 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4902 		vsi->num_q_vectors = 1;
4903 		break;
4904 
4905 	case I40E_VSI_VMDQ2:
4906 		vsi->alloc_queue_pairs = pf->num_vmdq_qps;
4907 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4908 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4909 		vsi->num_q_vectors = pf->num_vmdq_msix;
4910 		break;
4911 
4912 	case I40E_VSI_SRIOV:
4913 		vsi->alloc_queue_pairs = pf->num_vf_qps;
4914 		vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
4915 				      I40E_REQ_DESCRIPTOR_MULTIPLE);
4916 		break;
4917 
4918 	default:
4919 		WARN_ON(1);
4920 		return -ENODATA;
4921 	}
4922 
4923 	return 0;
4924 }
4925 
4926 /**
4927  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
4928  * @pf: board private structure
4929  * @type: type of VSI
4930  *
4931  * On error: returns error code (negative)
4932  * On success: returns vsi index in PF (positive)
4933  **/
4934 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
4935 {
4936 	int ret = -ENODEV;
4937 	struct i40e_vsi *vsi;
4938 	int vsi_idx;
4939 	int i;
4940 
4941 	/* Need to protect the allocation of the VSIs at the PF level */
4942 	mutex_lock(&pf->switch_mutex);
4943 
4944 	/* VSI list may be fragmented if VSI creation/destruction has
4945 	 * been happening.  We can afford to do a quick scan to look
4946 	 * for any free VSIs in the list.
4947 	 *
4948 	 * find next empty vsi slot, looping back around if necessary
4949 	 */
4950 	i = pf->next_vsi;
4951 	while (i < pf->hw.func_caps.num_vsis && pf->vsi[i])
4952 		i++;
4953 	if (i >= pf->hw.func_caps.num_vsis) {
4954 		i = 0;
4955 		while (i < pf->next_vsi && pf->vsi[i])
4956 			i++;
4957 	}
4958 
4959 	if (i < pf->hw.func_caps.num_vsis && !pf->vsi[i]) {
4960 		vsi_idx = i;             /* Found one! */
4961 	} else {
4962 		ret = -ENODEV;
4963 		goto err_alloc_vsi;  /* out of VSI slots! */
4964 	}
4965 	pf->next_vsi = ++i;
4966 
4967 	vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
4968 	if (!vsi) {
4969 		ret = -ENOMEM;
4970 		goto err_alloc_vsi;
4971 	}
4972 	vsi->type = type;
4973 	vsi->back = pf;
4974 	set_bit(__I40E_DOWN, &vsi->state);
4975 	vsi->flags = 0;
4976 	vsi->idx = vsi_idx;
4977 	vsi->rx_itr_setting = pf->rx_itr_default;
4978 	vsi->tx_itr_setting = pf->tx_itr_default;
4979 	vsi->netdev_registered = false;
4980 	vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
4981 	INIT_LIST_HEAD(&vsi->mac_filter_list);
4982 
4983 	i40e_set_num_rings_in_vsi(vsi);
4984 
4985 	/* Setup default MSIX irq handler for VSI */
4986 	i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
4987 
4988 	pf->vsi[vsi_idx] = vsi;
4989 	ret = vsi_idx;
4990 err_alloc_vsi:
4991 	mutex_unlock(&pf->switch_mutex);
4992 	return ret;
4993 }
4994 
4995 /**
4996  * i40e_vsi_clear - Deallocate the VSI provided
4997  * @vsi: the VSI being un-configured
4998  **/
4999 static int i40e_vsi_clear(struct i40e_vsi *vsi)
5000 {
5001 	struct i40e_pf *pf;
5002 
5003 	if (!vsi)
5004 		return 0;
5005 
5006 	if (!vsi->back)
5007 		goto free_vsi;
5008 	pf = vsi->back;
5009 
5010 	mutex_lock(&pf->switch_mutex);
5011 	if (!pf->vsi[vsi->idx]) {
5012 		dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
5013 			vsi->idx, vsi->idx, vsi, vsi->type);
5014 		goto unlock_vsi;
5015 	}
5016 
5017 	if (pf->vsi[vsi->idx] != vsi) {
5018 		dev_err(&pf->pdev->dev,
5019 			"pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
5020 			pf->vsi[vsi->idx]->idx,
5021 			pf->vsi[vsi->idx],
5022 			pf->vsi[vsi->idx]->type,
5023 			vsi->idx, vsi, vsi->type);
5024 		goto unlock_vsi;
5025 	}
5026 
5027 	/* updates the pf for this cleared vsi */
5028 	i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
5029 	i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
5030 
5031 	pf->vsi[vsi->idx] = NULL;
5032 	if (vsi->idx < pf->next_vsi)
5033 		pf->next_vsi = vsi->idx;
5034 
5035 unlock_vsi:
5036 	mutex_unlock(&pf->switch_mutex);
5037 free_vsi:
5038 	kfree(vsi);
5039 
5040 	return 0;
5041 }
5042 
5043 /**
5044  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
5045  * @vsi: the VSI being configured
5046  **/
5047 static int i40e_alloc_rings(struct i40e_vsi *vsi)
5048 {
5049 	struct i40e_pf *pf = vsi->back;
5050 	int ret = 0;
5051 	int i;
5052 
5053 	vsi->rx_rings = kcalloc(vsi->alloc_queue_pairs,
5054 				sizeof(struct i40e_ring), GFP_KERNEL);
5055 	if (!vsi->rx_rings) {
5056 		ret = -ENOMEM;
5057 		goto err_alloc_rings;
5058 	}
5059 
5060 	vsi->tx_rings = kcalloc(vsi->alloc_queue_pairs,
5061 				sizeof(struct i40e_ring), GFP_KERNEL);
5062 	if (!vsi->tx_rings) {
5063 		ret = -ENOMEM;
5064 		kfree(vsi->rx_rings);
5065 		goto err_alloc_rings;
5066 	}
5067 
5068 	/* Set basic values in the rings to be used later during open() */
5069 	for (i = 0; i < vsi->alloc_queue_pairs; i++) {
5070 		struct i40e_ring *rx_ring = &vsi->rx_rings[i];
5071 		struct i40e_ring *tx_ring = &vsi->tx_rings[i];
5072 
5073 		tx_ring->queue_index = i;
5074 		tx_ring->reg_idx = vsi->base_queue + i;
5075 		tx_ring->ring_active = false;
5076 		tx_ring->vsi = vsi;
5077 		tx_ring->netdev = vsi->netdev;
5078 		tx_ring->dev = &pf->pdev->dev;
5079 		tx_ring->count = vsi->num_desc;
5080 		tx_ring->size = 0;
5081 		tx_ring->dcb_tc = 0;
5082 
5083 		rx_ring->queue_index = i;
5084 		rx_ring->reg_idx = vsi->base_queue + i;
5085 		rx_ring->ring_active = false;
5086 		rx_ring->vsi = vsi;
5087 		rx_ring->netdev = vsi->netdev;
5088 		rx_ring->dev = &pf->pdev->dev;
5089 		rx_ring->count = vsi->num_desc;
5090 		rx_ring->size = 0;
5091 		rx_ring->dcb_tc = 0;
5092 		if (pf->flags & I40E_FLAG_16BYTE_RX_DESC_ENABLED)
5093 			set_ring_16byte_desc_enabled(rx_ring);
5094 		else
5095 			clear_ring_16byte_desc_enabled(rx_ring);
5096 	}
5097 
5098 err_alloc_rings:
5099 	return ret;
5100 }
5101 
5102 /**
5103  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
5104  * @vsi: the VSI being cleaned
5105  **/
5106 static int i40e_vsi_clear_rings(struct i40e_vsi *vsi)
5107 {
5108 	if (vsi) {
5109 		kfree(vsi->rx_rings);
5110 		kfree(vsi->tx_rings);
5111 	}
5112 
5113 	return 0;
5114 }
5115 
5116 /**
5117  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
5118  * @pf: board private structure
5119  * @vectors: the number of MSI-X vectors to request
5120  *
5121  * Returns the number of vectors reserved, or error
5122  **/
5123 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
5124 {
5125 	int err = 0;
5126 
5127 	pf->num_msix_entries = 0;
5128 	while (vectors >= I40E_MIN_MSIX) {
5129 		err = pci_enable_msix(pf->pdev, pf->msix_entries, vectors);
5130 		if (err == 0) {
5131 			/* good to go */
5132 			pf->num_msix_entries = vectors;
5133 			break;
5134 		} else if (err < 0) {
5135 			/* total failure */
5136 			dev_info(&pf->pdev->dev,
5137 				 "MSI-X vector reservation failed: %d\n", err);
5138 			vectors = 0;
5139 			break;
5140 		} else {
5141 			/* err > 0 is the hint for retry */
5142 			dev_info(&pf->pdev->dev,
5143 				 "MSI-X vectors wanted %d, retrying with %d\n",
5144 				 vectors, err);
5145 			vectors = err;
5146 		}
5147 	}
5148 
5149 	if (vectors > 0 && vectors < I40E_MIN_MSIX) {
5150 		dev_info(&pf->pdev->dev,
5151 			 "Couldn't get enough vectors, only %d available\n",
5152 			 vectors);
5153 		vectors = 0;
5154 	}
5155 
5156 	return vectors;
5157 }
5158 
5159 /**
5160  * i40e_init_msix - Setup the MSIX capability
5161  * @pf: board private structure
5162  *
5163  * Work with the OS to set up the MSIX vectors needed.
5164  *
5165  * Returns 0 on success, negative on failure
5166  **/
5167 static int i40e_init_msix(struct i40e_pf *pf)
5168 {
5169 	i40e_status err = 0;
5170 	struct i40e_hw *hw = &pf->hw;
5171 	int v_budget, i;
5172 	int vec;
5173 
5174 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
5175 		return -ENODEV;
5176 
5177 	/* The number of vectors we'll request will be comprised of:
5178 	 *   - Add 1 for "other" cause for Admin Queue events, etc.
5179 	 *   - The number of LAN queue pairs
5180 	 *        already adjusted for the NUMA node
5181 	 *        assumes symmetric Tx/Rx pairing
5182 	 *   - The number of VMDq pairs
5183 	 * Once we count this up, try the request.
5184 	 *
5185 	 * If we can't get what we want, we'll simplify to nearly nothing
5186 	 * and try again.  If that still fails, we punt.
5187 	 */
5188 	pf->num_lan_msix = pf->num_lan_qps;
5189 	pf->num_vmdq_msix = pf->num_vmdq_qps;
5190 	v_budget = 1 + pf->num_lan_msix;
5191 	v_budget += (pf->num_vmdq_vsis * pf->num_vmdq_msix);
5192 	if (pf->flags & I40E_FLAG_FDIR_ENABLED)
5193 		v_budget++;
5194 
5195 	/* Scale down if necessary, and the rings will share vectors */
5196 	v_budget = min_t(int, v_budget, hw->func_caps.num_msix_vectors);
5197 
5198 	pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
5199 				   GFP_KERNEL);
5200 	if (!pf->msix_entries)
5201 		return -ENOMEM;
5202 
5203 	for (i = 0; i < v_budget; i++)
5204 		pf->msix_entries[i].entry = i;
5205 	vec = i40e_reserve_msix_vectors(pf, v_budget);
5206 	if (vec < I40E_MIN_MSIX) {
5207 		pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
5208 		kfree(pf->msix_entries);
5209 		pf->msix_entries = NULL;
5210 		return -ENODEV;
5211 
5212 	} else if (vec == I40E_MIN_MSIX) {
5213 		/* Adjust for minimal MSIX use */
5214 		dev_info(&pf->pdev->dev, "Features disabled, not enough MSIX vectors\n");
5215 		pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
5216 		pf->num_vmdq_vsis = 0;
5217 		pf->num_vmdq_qps = 0;
5218 		pf->num_vmdq_msix = 0;
5219 		pf->num_lan_qps = 1;
5220 		pf->num_lan_msix = 1;
5221 
5222 	} else if (vec != v_budget) {
5223 		/* Scale vector usage down */
5224 		pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
5225 		vec--;                    /* reserve the misc vector */
5226 
5227 		/* partition out the remaining vectors */
5228 		switch (vec) {
5229 		case 2:
5230 			pf->num_vmdq_vsis = 1;
5231 			pf->num_lan_msix = 1;
5232 			break;
5233 		case 3:
5234 			pf->num_vmdq_vsis = 1;
5235 			pf->num_lan_msix = 2;
5236 			break;
5237 		default:
5238 			pf->num_lan_msix = min_t(int, (vec / 2),
5239 						 pf->num_lan_qps);
5240 			pf->num_vmdq_vsis = min_t(int, (vec - pf->num_lan_msix),
5241 						  I40E_DEFAULT_NUM_VMDQ_VSI);
5242 			break;
5243 		}
5244 	}
5245 
5246 	return err;
5247 }
5248 
5249 /**
5250  * i40e_alloc_q_vectors - Allocate memory for interrupt vectors
5251  * @vsi: the VSI being configured
5252  *
5253  * We allocate one q_vector per queue interrupt.  If allocation fails we
5254  * return -ENOMEM.
5255  **/
5256 static int i40e_alloc_q_vectors(struct i40e_vsi *vsi)
5257 {
5258 	struct i40e_pf *pf = vsi->back;
5259 	int v_idx, num_q_vectors;
5260 
5261 	/* if not MSIX, give the one vector only to the LAN VSI */
5262 	if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5263 		num_q_vectors = vsi->num_q_vectors;
5264 	else if (vsi == pf->vsi[pf->lan_vsi])
5265 		num_q_vectors = 1;
5266 	else
5267 		return -EINVAL;
5268 
5269 	vsi->q_vectors = kcalloc(num_q_vectors,
5270 				 sizeof(struct i40e_q_vector),
5271 				 GFP_KERNEL);
5272 	if (!vsi->q_vectors)
5273 		return -ENOMEM;
5274 
5275 	for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
5276 		vsi->q_vectors[v_idx].vsi = vsi;
5277 		vsi->q_vectors[v_idx].v_idx = v_idx;
5278 		cpumask_set_cpu(v_idx, &vsi->q_vectors[v_idx].affinity_mask);
5279 		if (vsi->netdev)
5280 			netif_napi_add(vsi->netdev, &vsi->q_vectors[v_idx].napi,
5281 				       i40e_napi_poll, vsi->work_limit);
5282 	}
5283 
5284 	return 0;
5285 }
5286 
5287 /**
5288  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
5289  * @pf: board private structure to initialize
5290  **/
5291 static void i40e_init_interrupt_scheme(struct i40e_pf *pf)
5292 {
5293 	int err = 0;
5294 
5295 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
5296 		err = i40e_init_msix(pf);
5297 		if (err) {
5298 			pf->flags &= ~(I40E_FLAG_RSS_ENABLED	   |
5299 					I40E_FLAG_MQ_ENABLED	   |
5300 					I40E_FLAG_DCB_ENABLED	   |
5301 					I40E_FLAG_SRIOV_ENABLED	   |
5302 					I40E_FLAG_FDIR_ENABLED	   |
5303 					I40E_FLAG_FDIR_ATR_ENABLED |
5304 					I40E_FLAG_VMDQ_ENABLED);
5305 
5306 			/* rework the queue expectations without MSIX */
5307 			i40e_determine_queue_usage(pf);
5308 		}
5309 	}
5310 
5311 	if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
5312 	    (pf->flags & I40E_FLAG_MSI_ENABLED)) {
5313 		err = pci_enable_msi(pf->pdev);
5314 		if (err) {
5315 			dev_info(&pf->pdev->dev,
5316 				 "MSI init failed (%d), trying legacy.\n", err);
5317 			pf->flags &= ~I40E_FLAG_MSI_ENABLED;
5318 		}
5319 	}
5320 
5321 	/* track first vector for misc interrupts */
5322 	err = i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT-1);
5323 }
5324 
5325 /**
5326  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
5327  * @pf: board private structure
5328  *
5329  * This sets up the handler for MSIX 0, which is used to manage the
5330  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
5331  * when in MSI or Legacy interrupt mode.
5332  **/
5333 static int i40e_setup_misc_vector(struct i40e_pf *pf)
5334 {
5335 	struct i40e_hw *hw = &pf->hw;
5336 	int err = 0;
5337 
5338 	/* Only request the irq if this is the first time through, and
5339 	 * not when we're rebuilding after a Reset
5340 	 */
5341 	if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
5342 		err = request_irq(pf->msix_entries[0].vector,
5343 				  i40e_intr, 0, pf->misc_int_name, pf);
5344 		if (err) {
5345 			dev_info(&pf->pdev->dev,
5346 				 "request_irq for msix_misc failed: %d\n", err);
5347 			return -EFAULT;
5348 		}
5349 	}
5350 
5351 	i40e_enable_misc_int_causes(hw);
5352 
5353 	/* associate no queues to the misc vector */
5354 	wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
5355 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
5356 
5357 	i40e_flush(hw);
5358 
5359 	i40e_irq_dynamic_enable_icr0(pf);
5360 
5361 	return err;
5362 }
5363 
5364 /**
5365  * i40e_config_rss - Prepare for RSS if used
5366  * @pf: board private structure
5367  **/
5368 static int i40e_config_rss(struct i40e_pf *pf)
5369 {
5370 	struct i40e_hw *hw = &pf->hw;
5371 	u32 lut = 0;
5372 	int i, j;
5373 	u64 hena;
5374 	/* Set of random keys generated using kernel random number generator */
5375 	static const u32 seed[I40E_PFQF_HKEY_MAX_INDEX + 1] = {0x41b01687,
5376 				0x183cfd8c, 0xce880440, 0x580cbc3c, 0x35897377,
5377 				0x328b25e1, 0x4fa98922, 0xb7d90c14, 0xd5bad70d,
5378 				0xcd15a2c1, 0xe8580225, 0x4a1e9d11, 0xfe5731be};
5379 
5380 	/* Fill out hash function seed */
5381 	for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
5382 		wr32(hw, I40E_PFQF_HKEY(i), seed[i]);
5383 
5384 	/* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
5385 	hena = (u64)rd32(hw, I40E_PFQF_HENA(0)) |
5386 		((u64)rd32(hw, I40E_PFQF_HENA(1)) << 32);
5387 	hena |= ((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV4_UDP) |
5388 		((u64)1 << I40E_FILTER_PCTYPE_NONF_UNICAST_IPV4_UDP) |
5389 		((u64)1 << I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV4_UDP) |
5390 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV4_TCP) |
5391 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV6_TCP) |
5392 		((u64)1 << I40E_FILTER_PCTYPE_NONF_IPV6_UDP) |
5393 		((u64)1 << I40E_FILTER_PCTYPE_NONF_UNICAST_IPV6_UDP) |
5394 		((u64)1 << I40E_FILTER_PCTYPE_NONF_MULTICAST_IPV6_UDP) |
5395 		((u64)1 << I40E_FILTER_PCTYPE_FRAG_IPV4)|
5396 		((u64)1 << I40E_FILTER_PCTYPE_FRAG_IPV6);
5397 	wr32(hw, I40E_PFQF_HENA(0), (u32)hena);
5398 	wr32(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
5399 
5400 	/* Populate the LUT with max no. of queues in round robin fashion */
5401 	for (i = 0, j = 0; i < pf->hw.func_caps.rss_table_size; i++, j++) {
5402 
5403 		/* The assumption is that lan qp count will be the highest
5404 		 * qp count for any PF VSI that needs RSS.
5405 		 * If multiple VSIs need RSS support, all the qp counts
5406 		 * for those VSIs should be a power of 2 for RSS to work.
5407 		 * If LAN VSI is the only consumer for RSS then this requirement
5408 		 * is not necessary.
5409 		 */
5410 		if (j == pf->rss_size)
5411 			j = 0;
5412 		/* lut = 4-byte sliding window of 4 lut entries */
5413 		lut = (lut << 8) | (j &
5414 			 ((0x1 << pf->hw.func_caps.rss_table_entry_width) - 1));
5415 		/* On i = 3, we have 4 entries in lut; write to the register */
5416 		if ((i & 3) == 3)
5417 			wr32(hw, I40E_PFQF_HLUT(i >> 2), lut);
5418 	}
5419 	i40e_flush(hw);
5420 
5421 	return 0;
5422 }
5423 
5424 /**
5425  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
5426  * @pf: board private structure to initialize
5427  *
5428  * i40e_sw_init initializes the Adapter private data structure.
5429  * Fields are initialized based on PCI device information and
5430  * OS network device settings (MTU size).
5431  **/
5432 static int i40e_sw_init(struct i40e_pf *pf)
5433 {
5434 	int err = 0;
5435 	int size;
5436 
5437 	pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
5438 				(NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
5439 	if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
5440 		if (I40E_DEBUG_USER & debug)
5441 			pf->hw.debug_mask = debug;
5442 		pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
5443 						I40E_DEFAULT_MSG_ENABLE);
5444 	}
5445 
5446 	/* Set default capability flags */
5447 	pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
5448 		    I40E_FLAG_MSI_ENABLED     |
5449 		    I40E_FLAG_MSIX_ENABLED    |
5450 		    I40E_FLAG_RX_PS_ENABLED   |
5451 		    I40E_FLAG_MQ_ENABLED      |
5452 		    I40E_FLAG_RX_1BUF_ENABLED;
5453 
5454 	pf->rss_size_max = 0x1 << pf->hw.func_caps.rss_table_entry_width;
5455 	if (pf->hw.func_caps.rss) {
5456 		pf->flags |= I40E_FLAG_RSS_ENABLED;
5457 		pf->rss_size = min_t(int, pf->rss_size_max,
5458 				     nr_cpus_node(numa_node_id()));
5459 	} else {
5460 		pf->rss_size = 1;
5461 	}
5462 
5463 	if (pf->hw.func_caps.dcb)
5464 		pf->num_tc_qps = I40E_DEFAULT_QUEUES_PER_TC;
5465 	else
5466 		pf->num_tc_qps = 0;
5467 
5468 	if (pf->hw.func_caps.fd) {
5469 		/* FW/NVM is not yet fixed in this regard */
5470 		if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
5471 		    (pf->hw.func_caps.fd_filters_best_effort > 0)) {
5472 			pf->flags |= I40E_FLAG_FDIR_ATR_ENABLED;
5473 			dev_info(&pf->pdev->dev,
5474 				 "Flow Director ATR mode Enabled\n");
5475 			pf->flags |= I40E_FLAG_FDIR_ENABLED;
5476 			dev_info(&pf->pdev->dev,
5477 				 "Flow Director Side Band mode Enabled\n");
5478 			pf->fdir_pf_filter_count =
5479 					 pf->hw.func_caps.fd_filters_guaranteed;
5480 		}
5481 	} else {
5482 		pf->fdir_pf_filter_count = 0;
5483 	}
5484 
5485 	if (pf->hw.func_caps.vmdq) {
5486 		pf->flags |= I40E_FLAG_VMDQ_ENABLED;
5487 		pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
5488 		pf->num_vmdq_qps = I40E_DEFAULT_QUEUES_PER_VMDQ;
5489 	}
5490 
5491 	/* MFP mode enabled */
5492 	if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.mfp_mode_1) {
5493 		pf->flags |= I40E_FLAG_MFP_ENABLED;
5494 		dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
5495 	}
5496 
5497 #ifdef CONFIG_PCI_IOV
5498 	if (pf->hw.func_caps.num_vfs) {
5499 		pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
5500 		pf->flags |= I40E_FLAG_SRIOV_ENABLED;
5501 		pf->num_req_vfs = min_t(int,
5502 					pf->hw.func_caps.num_vfs,
5503 					I40E_MAX_VF_COUNT);
5504 	}
5505 #endif /* CONFIG_PCI_IOV */
5506 	pf->eeprom_version = 0xDEAD;
5507 	pf->lan_veb = I40E_NO_VEB;
5508 	pf->lan_vsi = I40E_NO_VSI;
5509 
5510 	/* set up queue assignment tracking */
5511 	size = sizeof(struct i40e_lump_tracking)
5512 		+ (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
5513 	pf->qp_pile = kzalloc(size, GFP_KERNEL);
5514 	if (!pf->qp_pile) {
5515 		err = -ENOMEM;
5516 		goto sw_init_done;
5517 	}
5518 	pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
5519 	pf->qp_pile->search_hint = 0;
5520 
5521 	/* set up vector assignment tracking */
5522 	size = sizeof(struct i40e_lump_tracking)
5523 		+ (sizeof(u16) * pf->hw.func_caps.num_msix_vectors);
5524 	pf->irq_pile = kzalloc(size, GFP_KERNEL);
5525 	if (!pf->irq_pile) {
5526 		kfree(pf->qp_pile);
5527 		err = -ENOMEM;
5528 		goto sw_init_done;
5529 	}
5530 	pf->irq_pile->num_entries = pf->hw.func_caps.num_msix_vectors;
5531 	pf->irq_pile->search_hint = 0;
5532 
5533 	mutex_init(&pf->switch_mutex);
5534 
5535 sw_init_done:
5536 	return err;
5537 }
5538 
5539 /**
5540  * i40e_set_features - set the netdev feature flags
5541  * @netdev: ptr to the netdev being adjusted
5542  * @features: the feature set that the stack is suggesting
5543  **/
5544 static int i40e_set_features(struct net_device *netdev,
5545 			     netdev_features_t features)
5546 {
5547 	struct i40e_netdev_priv *np = netdev_priv(netdev);
5548 	struct i40e_vsi *vsi = np->vsi;
5549 
5550 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
5551 		i40e_vlan_stripping_enable(vsi);
5552 	else
5553 		i40e_vlan_stripping_disable(vsi);
5554 
5555 	return 0;
5556 }
5557 
5558 static const struct net_device_ops i40e_netdev_ops = {
5559 	.ndo_open		= i40e_open,
5560 	.ndo_stop		= i40e_close,
5561 	.ndo_start_xmit		= i40e_lan_xmit_frame,
5562 	.ndo_get_stats64	= i40e_get_netdev_stats_struct,
5563 	.ndo_set_rx_mode	= i40e_set_rx_mode,
5564 	.ndo_validate_addr	= eth_validate_addr,
5565 	.ndo_set_mac_address	= i40e_set_mac,
5566 	.ndo_change_mtu		= i40e_change_mtu,
5567 	.ndo_tx_timeout		= i40e_tx_timeout,
5568 	.ndo_vlan_rx_add_vid	= i40e_vlan_rx_add_vid,
5569 	.ndo_vlan_rx_kill_vid	= i40e_vlan_rx_kill_vid,
5570 #ifdef CONFIG_NET_POLL_CONTROLLER
5571 	.ndo_poll_controller	= i40e_netpoll,
5572 #endif
5573 	.ndo_setup_tc		= i40e_setup_tc,
5574 	.ndo_set_features	= i40e_set_features,
5575 	.ndo_set_vf_mac		= i40e_ndo_set_vf_mac,
5576 	.ndo_set_vf_vlan	= i40e_ndo_set_vf_port_vlan,
5577 	.ndo_set_vf_tx_rate	= i40e_ndo_set_vf_bw,
5578 	.ndo_get_vf_config	= i40e_ndo_get_vf_config,
5579 };
5580 
5581 /**
5582  * i40e_config_netdev - Setup the netdev flags
5583  * @vsi: the VSI being configured
5584  *
5585  * Returns 0 on success, negative value on failure
5586  **/
5587 static int i40e_config_netdev(struct i40e_vsi *vsi)
5588 {
5589 	struct i40e_pf *pf = vsi->back;
5590 	struct i40e_hw *hw = &pf->hw;
5591 	struct i40e_netdev_priv *np;
5592 	struct net_device *netdev;
5593 	u8 mac_addr[ETH_ALEN];
5594 	int etherdev_size;
5595 
5596 	etherdev_size = sizeof(struct i40e_netdev_priv);
5597 	netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
5598 	if (!netdev)
5599 		return -ENOMEM;
5600 
5601 	vsi->netdev = netdev;
5602 	np = netdev_priv(netdev);
5603 	np->vsi = vsi;
5604 
5605 	netdev->hw_enc_features = NETIF_F_IP_CSUM	 |
5606 				  NETIF_F_GSO_UDP_TUNNEL |
5607 				  NETIF_F_TSO		 |
5608 				  NETIF_F_SG;
5609 
5610 	netdev->features = NETIF_F_SG		       |
5611 			   NETIF_F_IP_CSUM	       |
5612 			   NETIF_F_SCTP_CSUM	       |
5613 			   NETIF_F_HIGHDMA	       |
5614 			   NETIF_F_GSO_UDP_TUNNEL      |
5615 			   NETIF_F_HW_VLAN_CTAG_TX     |
5616 			   NETIF_F_HW_VLAN_CTAG_RX     |
5617 			   NETIF_F_HW_VLAN_CTAG_FILTER |
5618 			   NETIF_F_IPV6_CSUM	       |
5619 			   NETIF_F_TSO		       |
5620 			   NETIF_F_TSO6		       |
5621 			   NETIF_F_RXCSUM	       |
5622 			   NETIF_F_RXHASH	       |
5623 			   0;
5624 
5625 	/* copy netdev features into list of user selectable features */
5626 	netdev->hw_features |= netdev->features;
5627 
5628 	if (vsi->type == I40E_VSI_MAIN) {
5629 		SET_NETDEV_DEV(netdev, &pf->pdev->dev);
5630 		memcpy(mac_addr, hw->mac.perm_addr, ETH_ALEN);
5631 	} else {
5632 		/* relate the VSI_VMDQ name to the VSI_MAIN name */
5633 		snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
5634 			 pf->vsi[pf->lan_vsi]->netdev->name);
5635 		random_ether_addr(mac_addr);
5636 		i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
5637 	}
5638 
5639 	memcpy(netdev->dev_addr, mac_addr, ETH_ALEN);
5640 	memcpy(netdev->perm_addr, mac_addr, ETH_ALEN);
5641 	/* vlan gets same features (except vlan offload)
5642 	 * after any tweaks for specific VSI types
5643 	 */
5644 	netdev->vlan_features = netdev->features & ~(NETIF_F_HW_VLAN_CTAG_TX |
5645 						     NETIF_F_HW_VLAN_CTAG_RX |
5646 						   NETIF_F_HW_VLAN_CTAG_FILTER);
5647 	netdev->priv_flags |= IFF_UNICAST_FLT;
5648 	netdev->priv_flags |= IFF_SUPP_NOFCS;
5649 	/* Setup netdev TC information */
5650 	i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
5651 
5652 	netdev->netdev_ops = &i40e_netdev_ops;
5653 	netdev->watchdog_timeo = 5 * HZ;
5654 	i40e_set_ethtool_ops(netdev);
5655 
5656 	return 0;
5657 }
5658 
5659 /**
5660  * i40e_vsi_delete - Delete a VSI from the switch
5661  * @vsi: the VSI being removed
5662  *
5663  * Returns 0 on success, negative value on failure
5664  **/
5665 static void i40e_vsi_delete(struct i40e_vsi *vsi)
5666 {
5667 	/* remove default VSI is not allowed */
5668 	if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
5669 		return;
5670 
5671 	/* there is no HW VSI for FDIR */
5672 	if (vsi->type == I40E_VSI_FDIR)
5673 		return;
5674 
5675 	i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
5676 	return;
5677 }
5678 
5679 /**
5680  * i40e_add_vsi - Add a VSI to the switch
5681  * @vsi: the VSI being configured
5682  *
5683  * This initializes a VSI context depending on the VSI type to be added and
5684  * passes it down to the add_vsi aq command.
5685  **/
5686 static int i40e_add_vsi(struct i40e_vsi *vsi)
5687 {
5688 	int ret = -ENODEV;
5689 	struct i40e_mac_filter *f, *ftmp;
5690 	struct i40e_pf *pf = vsi->back;
5691 	struct i40e_hw *hw = &pf->hw;
5692 	struct i40e_vsi_context ctxt;
5693 	u8 enabled_tc = 0x1; /* TC0 enabled */
5694 	int f_count = 0;
5695 
5696 	memset(&ctxt, 0, sizeof(ctxt));
5697 	switch (vsi->type) {
5698 	case I40E_VSI_MAIN:
5699 		/* The PF's main VSI is already setup as part of the
5700 		 * device initialization, so we'll not bother with
5701 		 * the add_vsi call, but we will retrieve the current
5702 		 * VSI context.
5703 		 */
5704 		ctxt.seid = pf->main_vsi_seid;
5705 		ctxt.pf_num = pf->hw.pf_id;
5706 		ctxt.vf_num = 0;
5707 		ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
5708 		ctxt.flags = I40E_AQ_VSI_TYPE_PF;
5709 		if (ret) {
5710 			dev_info(&pf->pdev->dev,
5711 				 "couldn't get pf vsi config, err %d, aq_err %d\n",
5712 				 ret, pf->hw.aq.asq_last_status);
5713 			return -ENOENT;
5714 		}
5715 		memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
5716 		vsi->info.valid_sections = 0;
5717 
5718 		vsi->seid = ctxt.seid;
5719 		vsi->id = ctxt.vsi_number;
5720 
5721 		enabled_tc = i40e_pf_get_tc_map(pf);
5722 
5723 		/* MFP mode setup queue map and update VSI */
5724 		if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5725 			memset(&ctxt, 0, sizeof(ctxt));
5726 			ctxt.seid = pf->main_vsi_seid;
5727 			ctxt.pf_num = pf->hw.pf_id;
5728 			ctxt.vf_num = 0;
5729 			i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
5730 			ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
5731 			if (ret) {
5732 				dev_info(&pf->pdev->dev,
5733 					 "update vsi failed, aq_err=%d\n",
5734 					 pf->hw.aq.asq_last_status);
5735 				ret = -ENOENT;
5736 				goto err;
5737 			}
5738 			/* update the local VSI info queue map */
5739 			i40e_vsi_update_queue_map(vsi, &ctxt);
5740 			vsi->info.valid_sections = 0;
5741 		} else {
5742 			/* Default/Main VSI is only enabled for TC0
5743 			 * reconfigure it to enable all TCs that are
5744 			 * available on the port in SFP mode.
5745 			 */
5746 			ret = i40e_vsi_config_tc(vsi, enabled_tc);
5747 			if (ret) {
5748 				dev_info(&pf->pdev->dev,
5749 					 "failed to configure TCs for main VSI tc_map 0x%08x, err %d, aq_err %d\n",
5750 					 enabled_tc, ret,
5751 					 pf->hw.aq.asq_last_status);
5752 				ret = -ENOENT;
5753 			}
5754 		}
5755 		break;
5756 
5757 	case I40E_VSI_FDIR:
5758 		/* no queue mapping or actual HW VSI needed */
5759 		vsi->info.valid_sections = 0;
5760 		vsi->seid = 0;
5761 		vsi->id = 0;
5762 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5763 		return 0;
5764 		break;
5765 
5766 	case I40E_VSI_VMDQ2:
5767 		ctxt.pf_num = hw->pf_id;
5768 		ctxt.vf_num = 0;
5769 		ctxt.uplink_seid = vsi->uplink_seid;
5770 		ctxt.connection_type = 0x1;     /* regular data port */
5771 		ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
5772 
5773 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5774 
5775 		/* This VSI is connected to VEB so the switch_id
5776 		 * should be set to zero by default.
5777 		 */
5778 		ctxt.info.switch_id = 0;
5779 		ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_LOCAL_LB);
5780 		ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5781 
5782 		/* Setup the VSI tx/rx queue map for TC0 only for now */
5783 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5784 		break;
5785 
5786 	case I40E_VSI_SRIOV:
5787 		ctxt.pf_num = hw->pf_id;
5788 		ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
5789 		ctxt.uplink_seid = vsi->uplink_seid;
5790 		ctxt.connection_type = 0x1;     /* regular data port */
5791 		ctxt.flags = I40E_AQ_VSI_TYPE_VF;
5792 
5793 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
5794 
5795 		/* This VSI is connected to VEB so the switch_id
5796 		 * should be set to zero by default.
5797 		 */
5798 		ctxt.info.switch_id = cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
5799 
5800 		ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
5801 		ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
5802 		/* Setup the VSI tx/rx queue map for TC0 only for now */
5803 		i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
5804 		break;
5805 
5806 	default:
5807 		return -ENODEV;
5808 	}
5809 
5810 	if (vsi->type != I40E_VSI_MAIN) {
5811 		ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
5812 		if (ret) {
5813 			dev_info(&vsi->back->pdev->dev,
5814 				 "add vsi failed, aq_err=%d\n",
5815 				 vsi->back->hw.aq.asq_last_status);
5816 			ret = -ENOENT;
5817 			goto err;
5818 		}
5819 		memcpy(&vsi->info, &ctxt.info, sizeof(ctxt.info));
5820 		vsi->info.valid_sections = 0;
5821 		vsi->seid = ctxt.seid;
5822 		vsi->id = ctxt.vsi_number;
5823 	}
5824 
5825 	/* If macvlan filters already exist, force them to get loaded */
5826 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
5827 		f->changed = true;
5828 		f_count++;
5829 	}
5830 	if (f_count) {
5831 		vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
5832 		pf->flags |= I40E_FLAG_FILTER_SYNC;
5833 	}
5834 
5835 	/* Update VSI BW information */
5836 	ret = i40e_vsi_get_bw_info(vsi);
5837 	if (ret) {
5838 		dev_info(&pf->pdev->dev,
5839 			 "couldn't get vsi bw info, err %d, aq_err %d\n",
5840 			 ret, pf->hw.aq.asq_last_status);
5841 		/* VSI is already added so not tearing that up */
5842 		ret = 0;
5843 	}
5844 
5845 err:
5846 	return ret;
5847 }
5848 
5849 /**
5850  * i40e_vsi_release - Delete a VSI and free its resources
5851  * @vsi: the VSI being removed
5852  *
5853  * Returns 0 on success or < 0 on error
5854  **/
5855 int i40e_vsi_release(struct i40e_vsi *vsi)
5856 {
5857 	struct i40e_mac_filter *f, *ftmp;
5858 	struct i40e_veb *veb = NULL;
5859 	struct i40e_pf *pf;
5860 	u16 uplink_seid;
5861 	int i, n;
5862 
5863 	pf = vsi->back;
5864 
5865 	/* release of a VEB-owner or last VSI is not allowed */
5866 	if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
5867 		dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
5868 			 vsi->seid, vsi->uplink_seid);
5869 		return -ENODEV;
5870 	}
5871 	if (vsi == pf->vsi[pf->lan_vsi] &&
5872 	    !test_bit(__I40E_DOWN, &pf->state)) {
5873 		dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
5874 		return -ENODEV;
5875 	}
5876 
5877 	uplink_seid = vsi->uplink_seid;
5878 	if (vsi->type != I40E_VSI_SRIOV) {
5879 		if (vsi->netdev_registered) {
5880 			vsi->netdev_registered = false;
5881 			if (vsi->netdev) {
5882 				/* results in a call to i40e_close() */
5883 				unregister_netdev(vsi->netdev);
5884 				free_netdev(vsi->netdev);
5885 				vsi->netdev = NULL;
5886 			}
5887 		} else {
5888 			if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
5889 				i40e_down(vsi);
5890 			i40e_vsi_free_irq(vsi);
5891 			i40e_vsi_free_tx_resources(vsi);
5892 			i40e_vsi_free_rx_resources(vsi);
5893 		}
5894 		i40e_vsi_disable_irq(vsi);
5895 	}
5896 
5897 	list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
5898 		i40e_del_filter(vsi, f->macaddr, f->vlan,
5899 				f->is_vf, f->is_netdev);
5900 	i40e_sync_vsi_filters(vsi);
5901 
5902 	i40e_vsi_delete(vsi);
5903 	i40e_vsi_free_q_vectors(vsi);
5904 	i40e_vsi_clear_rings(vsi);
5905 	i40e_vsi_clear(vsi);
5906 
5907 	/* If this was the last thing on the VEB, except for the
5908 	 * controlling VSI, remove the VEB, which puts the controlling
5909 	 * VSI onto the next level down in the switch.
5910 	 *
5911 	 * Well, okay, there's one more exception here: don't remove
5912 	 * the orphan VEBs yet.  We'll wait for an explicit remove request
5913 	 * from up the network stack.
5914 	 */
5915 	for (n = 0, i = 0; i < pf->hw.func_caps.num_vsis; i++) {
5916 		if (pf->vsi[i] &&
5917 		    pf->vsi[i]->uplink_seid == uplink_seid &&
5918 		    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
5919 			n++;      /* count the VSIs */
5920 		}
5921 	}
5922 	for (i = 0; i < I40E_MAX_VEB; i++) {
5923 		if (!pf->veb[i])
5924 			continue;
5925 		if (pf->veb[i]->uplink_seid == uplink_seid)
5926 			n++;     /* count the VEBs */
5927 		if (pf->veb[i]->seid == uplink_seid)
5928 			veb = pf->veb[i];
5929 	}
5930 	if (n == 0 && veb && veb->uplink_seid != 0)
5931 		i40e_veb_release(veb);
5932 
5933 	return 0;
5934 }
5935 
5936 /**
5937  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
5938  * @vsi: ptr to the VSI
5939  *
5940  * This should only be called after i40e_vsi_mem_alloc() which allocates the
5941  * corresponding SW VSI structure and initializes num_queue_pairs for the
5942  * newly allocated VSI.
5943  *
5944  * Returns 0 on success or negative on failure
5945  **/
5946 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
5947 {
5948 	int ret = -ENOENT;
5949 	struct i40e_pf *pf = vsi->back;
5950 
5951 	if (vsi->q_vectors) {
5952 		dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
5953 			 vsi->seid);
5954 		return -EEXIST;
5955 	}
5956 
5957 	if (vsi->base_vector) {
5958 		dev_info(&pf->pdev->dev,
5959 			 "VSI %d has non-zero base vector %d\n",
5960 			 vsi->seid, vsi->base_vector);
5961 		return -EEXIST;
5962 	}
5963 
5964 	ret = i40e_alloc_q_vectors(vsi);
5965 	if (ret) {
5966 		dev_info(&pf->pdev->dev,
5967 			 "failed to allocate %d q_vector for VSI %d, ret=%d\n",
5968 			 vsi->num_q_vectors, vsi->seid, ret);
5969 		vsi->num_q_vectors = 0;
5970 		goto vector_setup_out;
5971 	}
5972 
5973 	vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
5974 					 vsi->num_q_vectors, vsi->idx);
5975 	if (vsi->base_vector < 0) {
5976 		dev_info(&pf->pdev->dev,
5977 			 "failed to get q tracking for VSI %d, err=%d\n",
5978 			 vsi->seid, vsi->base_vector);
5979 		i40e_vsi_free_q_vectors(vsi);
5980 		ret = -ENOENT;
5981 		goto vector_setup_out;
5982 	}
5983 
5984 vector_setup_out:
5985 	return ret;
5986 }
5987 
5988 /**
5989  * i40e_vsi_setup - Set up a VSI by a given type
5990  * @pf: board private structure
5991  * @type: VSI type
5992  * @uplink_seid: the switch element to link to
5993  * @param1: usage depends upon VSI type. For VF types, indicates VF id
5994  *
5995  * This allocates the sw VSI structure and its queue resources, then add a VSI
5996  * to the identified VEB.
5997  *
5998  * Returns pointer to the successfully allocated and configure VSI sw struct on
5999  * success, otherwise returns NULL on failure.
6000  **/
6001 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
6002 				u16 uplink_seid, u32 param1)
6003 {
6004 	struct i40e_vsi *vsi = NULL;
6005 	struct i40e_veb *veb = NULL;
6006 	int ret, i;
6007 	int v_idx;
6008 
6009 	/* The requested uplink_seid must be either
6010 	 *     - the PF's port seid
6011 	 *              no VEB is needed because this is the PF
6012 	 *              or this is a Flow Director special case VSI
6013 	 *     - seid of an existing VEB
6014 	 *     - seid of a VSI that owns an existing VEB
6015 	 *     - seid of a VSI that doesn't own a VEB
6016 	 *              a new VEB is created and the VSI becomes the owner
6017 	 *     - seid of the PF VSI, which is what creates the first VEB
6018 	 *              this is a special case of the previous
6019 	 *
6020 	 * Find which uplink_seid we were given and create a new VEB if needed
6021 	 */
6022 	for (i = 0; i < I40E_MAX_VEB; i++) {
6023 		if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
6024 			veb = pf->veb[i];
6025 			break;
6026 		}
6027 	}
6028 
6029 	if (!veb && uplink_seid != pf->mac_seid) {
6030 
6031 		for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6032 			if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
6033 				vsi = pf->vsi[i];
6034 				break;
6035 			}
6036 		}
6037 		if (!vsi) {
6038 			dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
6039 				 uplink_seid);
6040 			return NULL;
6041 		}
6042 
6043 		if (vsi->uplink_seid == pf->mac_seid)
6044 			veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
6045 					     vsi->tc_config.enabled_tc);
6046 		else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
6047 			veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
6048 					     vsi->tc_config.enabled_tc);
6049 
6050 		for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
6051 			if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
6052 				veb = pf->veb[i];
6053 		}
6054 		if (!veb) {
6055 			dev_info(&pf->pdev->dev, "couldn't add VEB\n");
6056 			return NULL;
6057 		}
6058 
6059 		vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6060 		uplink_seid = veb->seid;
6061 	}
6062 
6063 	/* get vsi sw struct */
6064 	v_idx = i40e_vsi_mem_alloc(pf, type);
6065 	if (v_idx < 0)
6066 		goto err_alloc;
6067 	vsi = pf->vsi[v_idx];
6068 	vsi->type = type;
6069 	vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
6070 
6071 	if (type == I40E_VSI_MAIN)
6072 		pf->lan_vsi = v_idx;
6073 	else if (type == I40E_VSI_SRIOV)
6074 		vsi->vf_id = param1;
6075 	/* assign it some queues */
6076 	ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
6077 	if (ret < 0) {
6078 		dev_info(&pf->pdev->dev, "VSI %d get_lump failed %d\n",
6079 			 vsi->seid, ret);
6080 		goto err_vsi;
6081 	}
6082 	vsi->base_queue = ret;
6083 
6084 	/* get a VSI from the hardware */
6085 	vsi->uplink_seid = uplink_seid;
6086 	ret = i40e_add_vsi(vsi);
6087 	if (ret)
6088 		goto err_vsi;
6089 
6090 	switch (vsi->type) {
6091 	/* setup the netdev if needed */
6092 	case I40E_VSI_MAIN:
6093 	case I40E_VSI_VMDQ2:
6094 		ret = i40e_config_netdev(vsi);
6095 		if (ret)
6096 			goto err_netdev;
6097 		ret = register_netdev(vsi->netdev);
6098 		if (ret)
6099 			goto err_netdev;
6100 		vsi->netdev_registered = true;
6101 		netif_carrier_off(vsi->netdev);
6102 		/* fall through */
6103 
6104 	case I40E_VSI_FDIR:
6105 		/* set up vectors and rings if needed */
6106 		ret = i40e_vsi_setup_vectors(vsi);
6107 		if (ret)
6108 			goto err_msix;
6109 
6110 		ret = i40e_alloc_rings(vsi);
6111 		if (ret)
6112 			goto err_rings;
6113 
6114 		/* map all of the rings to the q_vectors */
6115 		i40e_vsi_map_rings_to_vectors(vsi);
6116 
6117 		i40e_vsi_reset_stats(vsi);
6118 		break;
6119 
6120 	default:
6121 		/* no netdev or rings for the other VSI types */
6122 		break;
6123 	}
6124 
6125 	return vsi;
6126 
6127 err_rings:
6128 	i40e_vsi_free_q_vectors(vsi);
6129 err_msix:
6130 	if (vsi->netdev_registered) {
6131 		vsi->netdev_registered = false;
6132 		unregister_netdev(vsi->netdev);
6133 		free_netdev(vsi->netdev);
6134 		vsi->netdev = NULL;
6135 	}
6136 err_netdev:
6137 	i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
6138 err_vsi:
6139 	i40e_vsi_clear(vsi);
6140 err_alloc:
6141 	return NULL;
6142 }
6143 
6144 /**
6145  * i40e_veb_get_bw_info - Query VEB BW information
6146  * @veb: the veb to query
6147  *
6148  * Query the Tx scheduler BW configuration data for given VEB
6149  **/
6150 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
6151 {
6152 	struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
6153 	struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
6154 	struct i40e_pf *pf = veb->pf;
6155 	struct i40e_hw *hw = &pf->hw;
6156 	u32 tc_bw_max;
6157 	int ret = 0;
6158 	int i;
6159 
6160 	ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
6161 						  &bw_data, NULL);
6162 	if (ret) {
6163 		dev_info(&pf->pdev->dev,
6164 			 "query veb bw config failed, aq_err=%d\n",
6165 			 hw->aq.asq_last_status);
6166 		goto out;
6167 	}
6168 
6169 	ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
6170 						   &ets_data, NULL);
6171 	if (ret) {
6172 		dev_info(&pf->pdev->dev,
6173 			 "query veb bw ets config failed, aq_err=%d\n",
6174 			 hw->aq.asq_last_status);
6175 		goto out;
6176 	}
6177 
6178 	veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
6179 	veb->bw_max_quanta = ets_data.tc_bw_max;
6180 	veb->is_abs_credits = bw_data.absolute_credits_enable;
6181 	tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
6182 		    (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
6183 	for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
6184 		veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
6185 		veb->bw_tc_limit_credits[i] =
6186 					le16_to_cpu(bw_data.tc_bw_limits[i]);
6187 		veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
6188 	}
6189 
6190 out:
6191 	return ret;
6192 }
6193 
6194 /**
6195  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
6196  * @pf: board private structure
6197  *
6198  * On error: returns error code (negative)
6199  * On success: returns vsi index in PF (positive)
6200  **/
6201 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
6202 {
6203 	int ret = -ENOENT;
6204 	struct i40e_veb *veb;
6205 	int i;
6206 
6207 	/* Need to protect the allocation of switch elements at the PF level */
6208 	mutex_lock(&pf->switch_mutex);
6209 
6210 	/* VEB list may be fragmented if VEB creation/destruction has
6211 	 * been happening.  We can afford to do a quick scan to look
6212 	 * for any free slots in the list.
6213 	 *
6214 	 * find next empty veb slot, looping back around if necessary
6215 	 */
6216 	i = 0;
6217 	while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
6218 		i++;
6219 	if (i >= I40E_MAX_VEB) {
6220 		ret = -ENOMEM;
6221 		goto err_alloc_veb;  /* out of VEB slots! */
6222 	}
6223 
6224 	veb = kzalloc(sizeof(*veb), GFP_KERNEL);
6225 	if (!veb) {
6226 		ret = -ENOMEM;
6227 		goto err_alloc_veb;
6228 	}
6229 	veb->pf = pf;
6230 	veb->idx = i;
6231 	veb->enabled_tc = 1;
6232 
6233 	pf->veb[i] = veb;
6234 	ret = i;
6235 err_alloc_veb:
6236 	mutex_unlock(&pf->switch_mutex);
6237 	return ret;
6238 }
6239 
6240 /**
6241  * i40e_switch_branch_release - Delete a branch of the switch tree
6242  * @branch: where to start deleting
6243  *
6244  * This uses recursion to find the tips of the branch to be
6245  * removed, deleting until we get back to and can delete this VEB.
6246  **/
6247 static void i40e_switch_branch_release(struct i40e_veb *branch)
6248 {
6249 	struct i40e_pf *pf = branch->pf;
6250 	u16 branch_seid = branch->seid;
6251 	u16 veb_idx = branch->idx;
6252 	int i;
6253 
6254 	/* release any VEBs on this VEB - RECURSION */
6255 	for (i = 0; i < I40E_MAX_VEB; i++) {
6256 		if (!pf->veb[i])
6257 			continue;
6258 		if (pf->veb[i]->uplink_seid == branch->seid)
6259 			i40e_switch_branch_release(pf->veb[i]);
6260 	}
6261 
6262 	/* Release the VSIs on this VEB, but not the owner VSI.
6263 	 *
6264 	 * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
6265 	 *       the VEB itself, so don't use (*branch) after this loop.
6266 	 */
6267 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6268 		if (!pf->vsi[i])
6269 			continue;
6270 		if (pf->vsi[i]->uplink_seid == branch_seid &&
6271 		   (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
6272 			i40e_vsi_release(pf->vsi[i]);
6273 		}
6274 	}
6275 
6276 	/* There's one corner case where the VEB might not have been
6277 	 * removed, so double check it here and remove it if needed.
6278 	 * This case happens if the veb was created from the debugfs
6279 	 * commands and no VSIs were added to it.
6280 	 */
6281 	if (pf->veb[veb_idx])
6282 		i40e_veb_release(pf->veb[veb_idx]);
6283 }
6284 
6285 /**
6286  * i40e_veb_clear - remove veb struct
6287  * @veb: the veb to remove
6288  **/
6289 static void i40e_veb_clear(struct i40e_veb *veb)
6290 {
6291 	if (!veb)
6292 		return;
6293 
6294 	if (veb->pf) {
6295 		struct i40e_pf *pf = veb->pf;
6296 
6297 		mutex_lock(&pf->switch_mutex);
6298 		if (pf->veb[veb->idx] == veb)
6299 			pf->veb[veb->idx] = NULL;
6300 		mutex_unlock(&pf->switch_mutex);
6301 	}
6302 
6303 	kfree(veb);
6304 }
6305 
6306 /**
6307  * i40e_veb_release - Delete a VEB and free its resources
6308  * @veb: the VEB being removed
6309  **/
6310 void i40e_veb_release(struct i40e_veb *veb)
6311 {
6312 	struct i40e_vsi *vsi = NULL;
6313 	struct i40e_pf *pf;
6314 	int i, n = 0;
6315 
6316 	pf = veb->pf;
6317 
6318 	/* find the remaining VSI and check for extras */
6319 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
6320 		if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
6321 			n++;
6322 			vsi = pf->vsi[i];
6323 		}
6324 	}
6325 	if (n != 1) {
6326 		dev_info(&pf->pdev->dev,
6327 			 "can't remove VEB %d with %d VSIs left\n",
6328 			 veb->seid, n);
6329 		return;
6330 	}
6331 
6332 	/* move the remaining VSI to uplink veb */
6333 	vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
6334 	if (veb->uplink_seid) {
6335 		vsi->uplink_seid = veb->uplink_seid;
6336 		if (veb->uplink_seid == pf->mac_seid)
6337 			vsi->veb_idx = I40E_NO_VEB;
6338 		else
6339 			vsi->veb_idx = veb->veb_idx;
6340 	} else {
6341 		/* floating VEB */
6342 		vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6343 		vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
6344 	}
6345 
6346 	i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
6347 	i40e_veb_clear(veb);
6348 
6349 	return;
6350 }
6351 
6352 /**
6353  * i40e_add_veb - create the VEB in the switch
6354  * @veb: the VEB to be instantiated
6355  * @vsi: the controlling VSI
6356  **/
6357 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
6358 {
6359 	bool is_default = (vsi->idx == vsi->back->lan_vsi);
6360 	int ret;
6361 
6362 	/* get a VEB from the hardware */
6363 	ret = i40e_aq_add_veb(&veb->pf->hw, veb->uplink_seid, vsi->seid,
6364 			      veb->enabled_tc, is_default, &veb->seid, NULL);
6365 	if (ret) {
6366 		dev_info(&veb->pf->pdev->dev,
6367 			 "couldn't add VEB, err %d, aq_err %d\n",
6368 			 ret, veb->pf->hw.aq.asq_last_status);
6369 		return -EPERM;
6370 	}
6371 
6372 	/* get statistics counter */
6373 	ret = i40e_aq_get_veb_parameters(&veb->pf->hw, veb->seid, NULL, NULL,
6374 					 &veb->stats_idx, NULL, NULL, NULL);
6375 	if (ret) {
6376 		dev_info(&veb->pf->pdev->dev,
6377 			 "couldn't get VEB statistics idx, err %d, aq_err %d\n",
6378 			 ret, veb->pf->hw.aq.asq_last_status);
6379 		return -EPERM;
6380 	}
6381 	ret = i40e_veb_get_bw_info(veb);
6382 	if (ret) {
6383 		dev_info(&veb->pf->pdev->dev,
6384 			 "couldn't get VEB bw info, err %d, aq_err %d\n",
6385 			 ret, veb->pf->hw.aq.asq_last_status);
6386 		i40e_aq_delete_element(&veb->pf->hw, veb->seid, NULL);
6387 		return -ENOENT;
6388 	}
6389 
6390 	vsi->uplink_seid = veb->seid;
6391 	vsi->veb_idx = veb->idx;
6392 	vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
6393 
6394 	return 0;
6395 }
6396 
6397 /**
6398  * i40e_veb_setup - Set up a VEB
6399  * @pf: board private structure
6400  * @flags: VEB setup flags
6401  * @uplink_seid: the switch element to link to
6402  * @vsi_seid: the initial VSI seid
6403  * @enabled_tc: Enabled TC bit-map
6404  *
6405  * This allocates the sw VEB structure and links it into the switch
6406  * It is possible and legal for this to be a duplicate of an already
6407  * existing VEB.  It is also possible for both uplink and vsi seids
6408  * to be zero, in order to create a floating VEB.
6409  *
6410  * Returns pointer to the successfully allocated VEB sw struct on
6411  * success, otherwise returns NULL on failure.
6412  **/
6413 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
6414 				u16 uplink_seid, u16 vsi_seid,
6415 				u8 enabled_tc)
6416 {
6417 	struct i40e_veb *veb, *uplink_veb = NULL;
6418 	int vsi_idx, veb_idx;
6419 	int ret;
6420 
6421 	/* if one seid is 0, the other must be 0 to create a floating relay */
6422 	if ((uplink_seid == 0 || vsi_seid == 0) &&
6423 	    (uplink_seid + vsi_seid != 0)) {
6424 		dev_info(&pf->pdev->dev,
6425 			 "one, not both seid's are 0: uplink=%d vsi=%d\n",
6426 			 uplink_seid, vsi_seid);
6427 		return NULL;
6428 	}
6429 
6430 	/* make sure there is such a vsi and uplink */
6431 	for (vsi_idx = 0; vsi_idx < pf->hw.func_caps.num_vsis; vsi_idx++)
6432 		if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
6433 			break;
6434 	if (vsi_idx >= pf->hw.func_caps.num_vsis && vsi_seid != 0) {
6435 		dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
6436 			 vsi_seid);
6437 		return NULL;
6438 	}
6439 
6440 	if (uplink_seid && uplink_seid != pf->mac_seid) {
6441 		for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6442 			if (pf->veb[veb_idx] &&
6443 			    pf->veb[veb_idx]->seid == uplink_seid) {
6444 				uplink_veb = pf->veb[veb_idx];
6445 				break;
6446 			}
6447 		}
6448 		if (!uplink_veb) {
6449 			dev_info(&pf->pdev->dev,
6450 				 "uplink seid %d not found\n", uplink_seid);
6451 			return NULL;
6452 		}
6453 	}
6454 
6455 	/* get veb sw struct */
6456 	veb_idx = i40e_veb_mem_alloc(pf);
6457 	if (veb_idx < 0)
6458 		goto err_alloc;
6459 	veb = pf->veb[veb_idx];
6460 	veb->flags = flags;
6461 	veb->uplink_seid = uplink_seid;
6462 	veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
6463 	veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
6464 
6465 	/* create the VEB in the switch */
6466 	ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
6467 	if (ret)
6468 		goto err_veb;
6469 
6470 	return veb;
6471 
6472 err_veb:
6473 	i40e_veb_clear(veb);
6474 err_alloc:
6475 	return NULL;
6476 }
6477 
6478 /**
6479  * i40e_setup_pf_switch_element - set pf vars based on switch type
6480  * @pf: board private structure
6481  * @ele: element we are building info from
6482  * @num_reported: total number of elements
6483  * @printconfig: should we print the contents
6484  *
6485  * helper function to assist in extracting a few useful SEID values.
6486  **/
6487 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
6488 				struct i40e_aqc_switch_config_element_resp *ele,
6489 				u16 num_reported, bool printconfig)
6490 {
6491 	u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
6492 	u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
6493 	u8 element_type = ele->element_type;
6494 	u16 seid = le16_to_cpu(ele->seid);
6495 
6496 	if (printconfig)
6497 		dev_info(&pf->pdev->dev,
6498 			 "type=%d seid=%d uplink=%d downlink=%d\n",
6499 			 element_type, seid, uplink_seid, downlink_seid);
6500 
6501 	switch (element_type) {
6502 	case I40E_SWITCH_ELEMENT_TYPE_MAC:
6503 		pf->mac_seid = seid;
6504 		break;
6505 	case I40E_SWITCH_ELEMENT_TYPE_VEB:
6506 		/* Main VEB? */
6507 		if (uplink_seid != pf->mac_seid)
6508 			break;
6509 		if (pf->lan_veb == I40E_NO_VEB) {
6510 			int v;
6511 
6512 			/* find existing or else empty VEB */
6513 			for (v = 0; v < I40E_MAX_VEB; v++) {
6514 				if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
6515 					pf->lan_veb = v;
6516 					break;
6517 				}
6518 			}
6519 			if (pf->lan_veb == I40E_NO_VEB) {
6520 				v = i40e_veb_mem_alloc(pf);
6521 				if (v < 0)
6522 					break;
6523 				pf->lan_veb = v;
6524 			}
6525 		}
6526 
6527 		pf->veb[pf->lan_veb]->seid = seid;
6528 		pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
6529 		pf->veb[pf->lan_veb]->pf = pf;
6530 		pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
6531 		break;
6532 	case I40E_SWITCH_ELEMENT_TYPE_VSI:
6533 		if (num_reported != 1)
6534 			break;
6535 		/* This is immediately after a reset so we can assume this is
6536 		 * the PF's VSI
6537 		 */
6538 		pf->mac_seid = uplink_seid;
6539 		pf->pf_seid = downlink_seid;
6540 		pf->main_vsi_seid = seid;
6541 		if (printconfig)
6542 			dev_info(&pf->pdev->dev,
6543 				 "pf_seid=%d main_vsi_seid=%d\n",
6544 				 pf->pf_seid, pf->main_vsi_seid);
6545 		break;
6546 	case I40E_SWITCH_ELEMENT_TYPE_PF:
6547 	case I40E_SWITCH_ELEMENT_TYPE_VF:
6548 	case I40E_SWITCH_ELEMENT_TYPE_EMP:
6549 	case I40E_SWITCH_ELEMENT_TYPE_BMC:
6550 	case I40E_SWITCH_ELEMENT_TYPE_PE:
6551 	case I40E_SWITCH_ELEMENT_TYPE_PA:
6552 		/* ignore these for now */
6553 		break;
6554 	default:
6555 		dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
6556 			 element_type, seid);
6557 		break;
6558 	}
6559 }
6560 
6561 /**
6562  * i40e_fetch_switch_configuration - Get switch config from firmware
6563  * @pf: board private structure
6564  * @printconfig: should we print the contents
6565  *
6566  * Get the current switch configuration from the device and
6567  * extract a few useful SEID values.
6568  **/
6569 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
6570 {
6571 	struct i40e_aqc_get_switch_config_resp *sw_config;
6572 	u16 next_seid = 0;
6573 	int ret = 0;
6574 	u8 *aq_buf;
6575 	int i;
6576 
6577 	aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
6578 	if (!aq_buf)
6579 		return -ENOMEM;
6580 
6581 	sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
6582 	do {
6583 		u16 num_reported, num_total;
6584 
6585 		ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
6586 						I40E_AQ_LARGE_BUF,
6587 						&next_seid, NULL);
6588 		if (ret) {
6589 			dev_info(&pf->pdev->dev,
6590 				 "get switch config failed %d aq_err=%x\n",
6591 				 ret, pf->hw.aq.asq_last_status);
6592 			kfree(aq_buf);
6593 			return -ENOENT;
6594 		}
6595 
6596 		num_reported = le16_to_cpu(sw_config->header.num_reported);
6597 		num_total = le16_to_cpu(sw_config->header.num_total);
6598 
6599 		if (printconfig)
6600 			dev_info(&pf->pdev->dev,
6601 				 "header: %d reported %d total\n",
6602 				 num_reported, num_total);
6603 
6604 		if (num_reported) {
6605 			int sz = sizeof(*sw_config) * num_reported;
6606 
6607 			kfree(pf->sw_config);
6608 			pf->sw_config = kzalloc(sz, GFP_KERNEL);
6609 			if (pf->sw_config)
6610 				memcpy(pf->sw_config, sw_config, sz);
6611 		}
6612 
6613 		for (i = 0; i < num_reported; i++) {
6614 			struct i40e_aqc_switch_config_element_resp *ele =
6615 				&sw_config->element[i];
6616 
6617 			i40e_setup_pf_switch_element(pf, ele, num_reported,
6618 						     printconfig);
6619 		}
6620 	} while (next_seid != 0);
6621 
6622 	kfree(aq_buf);
6623 	return ret;
6624 }
6625 
6626 /**
6627  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
6628  * @pf: board private structure
6629  *
6630  * Returns 0 on success, negative value on failure
6631  **/
6632 static int i40e_setup_pf_switch(struct i40e_pf *pf)
6633 {
6634 	int ret;
6635 
6636 	/* find out what's out there already */
6637 	ret = i40e_fetch_switch_configuration(pf, false);
6638 	if (ret) {
6639 		dev_info(&pf->pdev->dev,
6640 			 "couldn't fetch switch config, err %d, aq_err %d\n",
6641 			 ret, pf->hw.aq.asq_last_status);
6642 		return ret;
6643 	}
6644 	i40e_pf_reset_stats(pf);
6645 
6646 	/* fdir VSI must happen first to be sure it gets queue 0, but only
6647 	 * if there is enough room for the fdir VSI
6648 	 */
6649 	if (pf->num_lan_qps > 1)
6650 		i40e_fdir_setup(pf);
6651 
6652 	/* first time setup */
6653 	if (pf->lan_vsi == I40E_NO_VSI) {
6654 		struct i40e_vsi *vsi = NULL;
6655 		u16 uplink_seid;
6656 
6657 		/* Set up the PF VSI associated with the PF's main VSI
6658 		 * that is already in the HW switch
6659 		 */
6660 		if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6661 			uplink_seid = pf->veb[pf->lan_veb]->seid;
6662 		else
6663 			uplink_seid = pf->mac_seid;
6664 
6665 		vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
6666 		if (!vsi) {
6667 			dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
6668 			i40e_fdir_teardown(pf);
6669 			return -EAGAIN;
6670 		}
6671 		/* accommodate kcompat by copying the main VSI queue count
6672 		 * into the pf, since this newer code pushes the pf queue
6673 		 * info down a level into a VSI
6674 		 */
6675 		pf->num_rx_queues = vsi->alloc_queue_pairs;
6676 		pf->num_tx_queues = vsi->alloc_queue_pairs;
6677 	} else {
6678 		/* force a reset of TC and queue layout configurations */
6679 		u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
6680 		pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
6681 		pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
6682 		i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
6683 	}
6684 	i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
6685 
6686 	/* Setup static PF queue filter control settings */
6687 	ret = i40e_setup_pf_filter_control(pf);
6688 	if (ret) {
6689 		dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
6690 			 ret);
6691 		/* Failure here should not stop continuing other steps */
6692 	}
6693 
6694 	/* enable RSS in the HW, even for only one queue, as the stack can use
6695 	 * the hash
6696 	 */
6697 	if ((pf->flags & I40E_FLAG_RSS_ENABLED))
6698 		i40e_config_rss(pf);
6699 
6700 	/* fill in link information and enable LSE reporting */
6701 	i40e_aq_get_link_info(&pf->hw, true, NULL, NULL);
6702 	i40e_link_event(pf);
6703 
6704 	/* Initialize user-specifics link properties */
6705 	pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
6706 				  I40E_AQ_AN_COMPLETED) ? true : false);
6707 	pf->hw.fc.requested_mode = I40E_FC_DEFAULT;
6708 	if (pf->hw.phy.link_info.an_info &
6709 	   (I40E_AQ_LINK_PAUSE_TX | I40E_AQ_LINK_PAUSE_RX))
6710 		pf->hw.fc.current_mode = I40E_FC_FULL;
6711 	else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_TX)
6712 		pf->hw.fc.current_mode = I40E_FC_TX_PAUSE;
6713 	else if (pf->hw.phy.link_info.an_info & I40E_AQ_LINK_PAUSE_RX)
6714 		pf->hw.fc.current_mode = I40E_FC_RX_PAUSE;
6715 	else
6716 		pf->hw.fc.current_mode = I40E_FC_DEFAULT;
6717 
6718 	return ret;
6719 }
6720 
6721 /**
6722  * i40e_set_rss_size - helper to set rss_size
6723  * @pf: board private structure
6724  * @queues_left: how many queues
6725  */
6726 static u16 i40e_set_rss_size(struct i40e_pf *pf, int queues_left)
6727 {
6728 	int num_tc0;
6729 
6730 	num_tc0 = min_t(int, queues_left, pf->rss_size_max);
6731 	num_tc0 = min_t(int, num_tc0, nr_cpus_node(numa_node_id()));
6732 	num_tc0 = rounddown_pow_of_two(num_tc0);
6733 
6734 	return num_tc0;
6735 }
6736 
6737 /**
6738  * i40e_determine_queue_usage - Work out queue distribution
6739  * @pf: board private structure
6740  **/
6741 static void i40e_determine_queue_usage(struct i40e_pf *pf)
6742 {
6743 	int accum_tc_size;
6744 	int queues_left;
6745 
6746 	pf->num_lan_qps = 0;
6747 	pf->num_tc_qps = rounddown_pow_of_two(pf->num_tc_qps);
6748 	accum_tc_size = (I40E_MAX_TRAFFIC_CLASS - 1) * pf->num_tc_qps;
6749 
6750 	/* Find the max queues to be put into basic use.  We'll always be
6751 	 * using TC0, whether or not DCB is running, and TC0 will get the
6752 	 * big RSS set.
6753 	 */
6754 	queues_left = pf->hw.func_caps.num_tx_qp;
6755 
6756 	if   (!((pf->flags & I40E_FLAG_MSIX_ENABLED)		 &&
6757 		(pf->flags & I40E_FLAG_MQ_ENABLED))		 ||
6758 		!(pf->flags & (I40E_FLAG_RSS_ENABLED |
6759 		I40E_FLAG_FDIR_ENABLED | I40E_FLAG_DCB_ENABLED)) ||
6760 		(queues_left == 1)) {
6761 
6762 		/* one qp for PF, no queues for anything else */
6763 		queues_left = 0;
6764 		pf->rss_size = pf->num_lan_qps = 1;
6765 
6766 		/* make sure all the fancies are disabled */
6767 		pf->flags &= ~(I40E_FLAG_RSS_ENABLED       |
6768 				I40E_FLAG_MQ_ENABLED	   |
6769 				I40E_FLAG_FDIR_ENABLED	   |
6770 				I40E_FLAG_FDIR_ATR_ENABLED |
6771 				I40E_FLAG_DCB_ENABLED	   |
6772 				I40E_FLAG_SRIOV_ENABLED	   |
6773 				I40E_FLAG_VMDQ_ENABLED);
6774 
6775 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED	  &&
6776 		   !(pf->flags & I40E_FLAG_FDIR_ENABLED)  &&
6777 		   !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6778 
6779 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6780 
6781 		queues_left -= pf->rss_size;
6782 		pf->num_lan_qps = pf->rss_size;
6783 
6784 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED	  &&
6785 		   !(pf->flags & I40E_FLAG_FDIR_ENABLED)  &&
6786 		   (pf->flags & I40E_FLAG_DCB_ENABLED)) {
6787 
6788 		/* save num_tc_qps queues for TCs 1 thru 7 and the rest
6789 		 * are set up for RSS in TC0
6790 		 */
6791 		queues_left -= accum_tc_size;
6792 
6793 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6794 
6795 		queues_left -= pf->rss_size;
6796 		if (queues_left < 0) {
6797 			dev_info(&pf->pdev->dev, "not enough queues for DCB\n");
6798 			return;
6799 		}
6800 
6801 		pf->num_lan_qps = pf->rss_size + accum_tc_size;
6802 
6803 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED   &&
6804 		  (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
6805 		  !(pf->flags & I40E_FLAG_DCB_ENABLED)) {
6806 
6807 		queues_left -= 1; /* save 1 queue for FD */
6808 
6809 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6810 
6811 		queues_left -= pf->rss_size;
6812 		if (queues_left < 0) {
6813 			dev_info(&pf->pdev->dev, "not enough queues for Flow Director\n");
6814 			return;
6815 		}
6816 
6817 		pf->num_lan_qps = pf->rss_size;
6818 
6819 	} else if (pf->flags & I40E_FLAG_RSS_ENABLED   &&
6820 		  (pf->flags & I40E_FLAG_FDIR_ENABLED) &&
6821 		  (pf->flags & I40E_FLAG_DCB_ENABLED)) {
6822 
6823 		/* save 1 queue for TCs 1 thru 7,
6824 		 * 1 queue for flow director,
6825 		 * and the rest are set up for RSS in TC0
6826 		 */
6827 		queues_left -= 1;
6828 		queues_left -= accum_tc_size;
6829 
6830 		pf->rss_size = i40e_set_rss_size(pf, queues_left);
6831 		queues_left -= pf->rss_size;
6832 		if (queues_left < 0) {
6833 			dev_info(&pf->pdev->dev, "not enough queues for DCB and Flow Director\n");
6834 			return;
6835 		}
6836 
6837 		pf->num_lan_qps = pf->rss_size + accum_tc_size;
6838 
6839 	} else {
6840 		dev_info(&pf->pdev->dev,
6841 			 "Invalid configuration, flags=0x%08llx\n", pf->flags);
6842 		return;
6843 	}
6844 
6845 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
6846 	    pf->num_vf_qps && pf->num_req_vfs && queues_left) {
6847 		pf->num_req_vfs = min_t(int, pf->num_req_vfs, (queues_left /
6848 							       pf->num_vf_qps));
6849 		queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
6850 	}
6851 
6852 	if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
6853 	    pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
6854 		pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
6855 					  (queues_left / pf->num_vmdq_qps));
6856 		queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
6857 	}
6858 
6859 	return;
6860 }
6861 
6862 /**
6863  * i40e_setup_pf_filter_control - Setup PF static filter control
6864  * @pf: PF to be setup
6865  *
6866  * i40e_setup_pf_filter_control sets up a pf's initial filter control
6867  * settings. If PE/FCoE are enabled then it will also set the per PF
6868  * based filter sizes required for them. It also enables Flow director,
6869  * ethertype and macvlan type filter settings for the pf.
6870  *
6871  * Returns 0 on success, negative on failure
6872  **/
6873 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
6874 {
6875 	struct i40e_filter_control_settings *settings = &pf->filter_settings;
6876 
6877 	settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
6878 
6879 	/* Flow Director is enabled */
6880 	if (pf->flags & (I40E_FLAG_FDIR_ENABLED | I40E_FLAG_FDIR_ATR_ENABLED))
6881 		settings->enable_fdir = true;
6882 
6883 	/* Ethtype and MACVLAN filters enabled for PF */
6884 	settings->enable_ethtype = true;
6885 	settings->enable_macvlan = true;
6886 
6887 	if (i40e_set_filter_control(&pf->hw, settings))
6888 		return -ENOENT;
6889 
6890 	return 0;
6891 }
6892 
6893 /**
6894  * i40e_probe - Device initialization routine
6895  * @pdev: PCI device information struct
6896  * @ent: entry in i40e_pci_tbl
6897  *
6898  * i40e_probe initializes a pf identified by a pci_dev structure.
6899  * The OS initialization, configuring of the pf private structure,
6900  * and a hardware reset occur.
6901  *
6902  * Returns 0 on success, negative on failure
6903  **/
6904 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
6905 {
6906 	struct i40e_driver_version dv;
6907 	struct i40e_pf *pf;
6908 	struct i40e_hw *hw;
6909 	int err = 0;
6910 	u32 len;
6911 
6912 	err = pci_enable_device_mem(pdev);
6913 	if (err)
6914 		return err;
6915 
6916 	/* set up for high or low dma */
6917 	if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(64))) {
6918 		/* coherent mask for the same size will always succeed if
6919 		 * dma_set_mask does
6920 		 */
6921 		dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(64));
6922 	} else if (!dma_set_mask(&pdev->dev, DMA_BIT_MASK(32))) {
6923 		dma_set_coherent_mask(&pdev->dev, DMA_BIT_MASK(32));
6924 	} else {
6925 		dev_err(&pdev->dev, "DMA configuration failed: %d\n", err);
6926 		err = -EIO;
6927 		goto err_dma;
6928 	}
6929 
6930 	/* set up pci connections */
6931 	err = pci_request_selected_regions(pdev, pci_select_bars(pdev,
6932 					   IORESOURCE_MEM), i40e_driver_name);
6933 	if (err) {
6934 		dev_info(&pdev->dev,
6935 			 "pci_request_selected_regions failed %d\n", err);
6936 		goto err_pci_reg;
6937 	}
6938 
6939 	pci_enable_pcie_error_reporting(pdev);
6940 	pci_set_master(pdev);
6941 
6942 	/* Now that we have a PCI connection, we need to do the
6943 	 * low level device setup.  This is primarily setting up
6944 	 * the Admin Queue structures and then querying for the
6945 	 * device's current profile information.
6946 	 */
6947 	pf = kzalloc(sizeof(*pf), GFP_KERNEL);
6948 	if (!pf) {
6949 		err = -ENOMEM;
6950 		goto err_pf_alloc;
6951 	}
6952 	pf->next_vsi = 0;
6953 	pf->pdev = pdev;
6954 	set_bit(__I40E_DOWN, &pf->state);
6955 
6956 	hw = &pf->hw;
6957 	hw->back = pf;
6958 	hw->hw_addr = ioremap(pci_resource_start(pdev, 0),
6959 			      pci_resource_len(pdev, 0));
6960 	if (!hw->hw_addr) {
6961 		err = -EIO;
6962 		dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
6963 			 (unsigned int)pci_resource_start(pdev, 0),
6964 			 (unsigned int)pci_resource_len(pdev, 0), err);
6965 		goto err_ioremap;
6966 	}
6967 	hw->vendor_id = pdev->vendor;
6968 	hw->device_id = pdev->device;
6969 	pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
6970 	hw->subsystem_vendor_id = pdev->subsystem_vendor;
6971 	hw->subsystem_device_id = pdev->subsystem_device;
6972 	hw->bus.device = PCI_SLOT(pdev->devfn);
6973 	hw->bus.func = PCI_FUNC(pdev->devfn);
6974 
6975 	/* Reset here to make sure all is clean and to define PF 'n' */
6976 	err = i40e_pf_reset(hw);
6977 	if (err) {
6978 		dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
6979 		goto err_pf_reset;
6980 	}
6981 	pf->pfr_count++;
6982 
6983 	hw->aq.num_arq_entries = I40E_AQ_LEN;
6984 	hw->aq.num_asq_entries = I40E_AQ_LEN;
6985 	hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
6986 	hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
6987 	pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
6988 	snprintf(pf->misc_int_name, sizeof(pf->misc_int_name) - 1,
6989 		 "%s-pf%d:misc",
6990 		 dev_driver_string(&pf->pdev->dev), pf->hw.pf_id);
6991 
6992 	err = i40e_init_shared_code(hw);
6993 	if (err) {
6994 		dev_info(&pdev->dev, "init_shared_code failed: %d\n", err);
6995 		goto err_pf_reset;
6996 	}
6997 
6998 	err = i40e_init_adminq(hw);
6999 	dev_info(&pdev->dev, "%s\n", i40e_fw_version_str(hw));
7000 	if (err) {
7001 		dev_info(&pdev->dev,
7002 			 "init_adminq failed: %d expecting API %02x.%02x\n",
7003 			 err,
7004 			 I40E_FW_API_VERSION_MAJOR, I40E_FW_API_VERSION_MINOR);
7005 		goto err_pf_reset;
7006 	}
7007 
7008 	err = i40e_get_capabilities(pf);
7009 	if (err)
7010 		goto err_adminq_setup;
7011 
7012 	err = i40e_sw_init(pf);
7013 	if (err) {
7014 		dev_info(&pdev->dev, "sw_init failed: %d\n", err);
7015 		goto err_sw_init;
7016 	}
7017 
7018 	err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
7019 				hw->func_caps.num_rx_qp,
7020 				pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
7021 	if (err) {
7022 		dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
7023 		goto err_init_lan_hmc;
7024 	}
7025 
7026 	err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
7027 	if (err) {
7028 		dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
7029 		err = -ENOENT;
7030 		goto err_configure_lan_hmc;
7031 	}
7032 
7033 	i40e_get_mac_addr(hw, hw->mac.addr);
7034 	if (i40e_validate_mac_addr(hw->mac.addr)) {
7035 		dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
7036 		err = -EIO;
7037 		goto err_mac_addr;
7038 	}
7039 	dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
7040 	memcpy(hw->mac.perm_addr, hw->mac.addr, ETH_ALEN);
7041 
7042 	pci_set_drvdata(pdev, pf);
7043 	pci_save_state(pdev);
7044 
7045 	/* set up periodic task facility */
7046 	setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
7047 	pf->service_timer_period = HZ;
7048 
7049 	INIT_WORK(&pf->service_task, i40e_service_task);
7050 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
7051 	pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
7052 	pf->link_check_timeout = jiffies;
7053 
7054 	/* set up the main switch operations */
7055 	i40e_determine_queue_usage(pf);
7056 	i40e_init_interrupt_scheme(pf);
7057 
7058 	/* Set up the *vsi struct based on the number of VSIs in the HW,
7059 	 * and set up our local tracking of the MAIN PF vsi.
7060 	 */
7061 	len = sizeof(struct i40e_vsi *) * pf->hw.func_caps.num_vsis;
7062 	pf->vsi = kzalloc(len, GFP_KERNEL);
7063 	if (!pf->vsi)
7064 		goto err_switch_setup;
7065 
7066 	err = i40e_setup_pf_switch(pf);
7067 	if (err) {
7068 		dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
7069 		goto err_vsis;
7070 	}
7071 
7072 	/* The main driver is (mostly) up and happy. We need to set this state
7073 	 * before setting up the misc vector or we get a race and the vector
7074 	 * ends up disabled forever.
7075 	 */
7076 	clear_bit(__I40E_DOWN, &pf->state);
7077 
7078 	/* In case of MSIX we are going to setup the misc vector right here
7079 	 * to handle admin queue events etc. In case of legacy and MSI
7080 	 * the misc functionality and queue processing is combined in
7081 	 * the same vector and that gets setup at open.
7082 	 */
7083 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7084 		err = i40e_setup_misc_vector(pf);
7085 		if (err) {
7086 			dev_info(&pdev->dev,
7087 				 "setup of misc vector failed: %d\n", err);
7088 			goto err_vsis;
7089 		}
7090 	}
7091 
7092 	/* prep for VF support */
7093 	if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
7094 	    (pf->flags & I40E_FLAG_MSIX_ENABLED)) {
7095 		u32 val;
7096 
7097 		/* disable link interrupts for VFs */
7098 		val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
7099 		val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
7100 		wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
7101 		i40e_flush(hw);
7102 	}
7103 
7104 	i40e_dbg_pf_init(pf);
7105 
7106 	/* tell the firmware that we're starting */
7107 	dv.major_version = DRV_VERSION_MAJOR;
7108 	dv.minor_version = DRV_VERSION_MINOR;
7109 	dv.build_version = DRV_VERSION_BUILD;
7110 	dv.subbuild_version = 0;
7111 	i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
7112 
7113 	/* since everything's happy, start the service_task timer */
7114 	mod_timer(&pf->service_timer,
7115 		  round_jiffies(jiffies + pf->service_timer_period));
7116 
7117 	return 0;
7118 
7119 	/* Unwind what we've done if something failed in the setup */
7120 err_vsis:
7121 	set_bit(__I40E_DOWN, &pf->state);
7122 err_switch_setup:
7123 	i40e_clear_interrupt_scheme(pf);
7124 	kfree(pf->vsi);
7125 	del_timer_sync(&pf->service_timer);
7126 err_mac_addr:
7127 err_configure_lan_hmc:
7128 	(void)i40e_shutdown_lan_hmc(hw);
7129 err_init_lan_hmc:
7130 	kfree(pf->qp_pile);
7131 	kfree(pf->irq_pile);
7132 err_sw_init:
7133 err_adminq_setup:
7134 	(void)i40e_shutdown_adminq(hw);
7135 err_pf_reset:
7136 	iounmap(hw->hw_addr);
7137 err_ioremap:
7138 	kfree(pf);
7139 err_pf_alloc:
7140 	pci_disable_pcie_error_reporting(pdev);
7141 	pci_release_selected_regions(pdev,
7142 				     pci_select_bars(pdev, IORESOURCE_MEM));
7143 err_pci_reg:
7144 err_dma:
7145 	pci_disable_device(pdev);
7146 	return err;
7147 }
7148 
7149 /**
7150  * i40e_remove - Device removal routine
7151  * @pdev: PCI device information struct
7152  *
7153  * i40e_remove is called by the PCI subsystem to alert the driver
7154  * that is should release a PCI device.  This could be caused by a
7155  * Hot-Plug event, or because the driver is going to be removed from
7156  * memory.
7157  **/
7158 static void i40e_remove(struct pci_dev *pdev)
7159 {
7160 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7161 	i40e_status ret_code;
7162 	u32 reg;
7163 	int i;
7164 
7165 	i40e_dbg_pf_exit(pf);
7166 
7167 	if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
7168 		i40e_free_vfs(pf);
7169 		pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
7170 	}
7171 
7172 	/* no more scheduling of any task */
7173 	set_bit(__I40E_DOWN, &pf->state);
7174 	del_timer_sync(&pf->service_timer);
7175 	cancel_work_sync(&pf->service_task);
7176 
7177 	i40e_fdir_teardown(pf);
7178 
7179 	/* If there is a switch structure or any orphans, remove them.
7180 	 * This will leave only the PF's VSI remaining.
7181 	 */
7182 	for (i = 0; i < I40E_MAX_VEB; i++) {
7183 		if (!pf->veb[i])
7184 			continue;
7185 
7186 		if (pf->veb[i]->uplink_seid == pf->mac_seid ||
7187 		    pf->veb[i]->uplink_seid == 0)
7188 			i40e_switch_branch_release(pf->veb[i]);
7189 	}
7190 
7191 	/* Now we can shutdown the PF's VSI, just before we kill
7192 	 * adminq and hmc.
7193 	 */
7194 	if (pf->vsi[pf->lan_vsi])
7195 		i40e_vsi_release(pf->vsi[pf->lan_vsi]);
7196 
7197 	i40e_stop_misc_vector(pf);
7198 	if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7199 		synchronize_irq(pf->msix_entries[0].vector);
7200 		free_irq(pf->msix_entries[0].vector, pf);
7201 	}
7202 
7203 	/* shutdown and destroy the HMC */
7204 	ret_code = i40e_shutdown_lan_hmc(&pf->hw);
7205 	if (ret_code)
7206 		dev_warn(&pdev->dev,
7207 			 "Failed to destroy the HMC resources: %d\n", ret_code);
7208 
7209 	/* shutdown the adminq */
7210 	i40e_aq_queue_shutdown(&pf->hw, true);
7211 	ret_code = i40e_shutdown_adminq(&pf->hw);
7212 	if (ret_code)
7213 		dev_warn(&pdev->dev,
7214 			 "Failed to destroy the Admin Queue resources: %d\n",
7215 			 ret_code);
7216 
7217 	/* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
7218 	i40e_clear_interrupt_scheme(pf);
7219 	for (i = 0; i < pf->hw.func_caps.num_vsis; i++) {
7220 		if (pf->vsi[i]) {
7221 			i40e_vsi_clear_rings(pf->vsi[i]);
7222 			i40e_vsi_clear(pf->vsi[i]);
7223 			pf->vsi[i] = NULL;
7224 		}
7225 	}
7226 
7227 	for (i = 0; i < I40E_MAX_VEB; i++) {
7228 		kfree(pf->veb[i]);
7229 		pf->veb[i] = NULL;
7230 	}
7231 
7232 	kfree(pf->qp_pile);
7233 	kfree(pf->irq_pile);
7234 	kfree(pf->sw_config);
7235 	kfree(pf->vsi);
7236 
7237 	/* force a PF reset to clean anything leftover */
7238 	reg = rd32(&pf->hw, I40E_PFGEN_CTRL);
7239 	wr32(&pf->hw, I40E_PFGEN_CTRL, (reg | I40E_PFGEN_CTRL_PFSWR_MASK));
7240 	i40e_flush(&pf->hw);
7241 
7242 	iounmap(pf->hw.hw_addr);
7243 	kfree(pf);
7244 	pci_release_selected_regions(pdev,
7245 				     pci_select_bars(pdev, IORESOURCE_MEM));
7246 
7247 	pci_disable_pcie_error_reporting(pdev);
7248 	pci_disable_device(pdev);
7249 }
7250 
7251 /**
7252  * i40e_pci_error_detected - warning that something funky happened in PCI land
7253  * @pdev: PCI device information struct
7254  *
7255  * Called to warn that something happened and the error handling steps
7256  * are in progress.  Allows the driver to quiesce things, be ready for
7257  * remediation.
7258  **/
7259 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
7260 						enum pci_channel_state error)
7261 {
7262 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7263 
7264 	dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
7265 
7266 	/* shutdown all operations */
7267 	i40e_pf_quiesce_all_vsi(pf);
7268 
7269 	/* Request a slot reset */
7270 	return PCI_ERS_RESULT_NEED_RESET;
7271 }
7272 
7273 /**
7274  * i40e_pci_error_slot_reset - a PCI slot reset just happened
7275  * @pdev: PCI device information struct
7276  *
7277  * Called to find if the driver can work with the device now that
7278  * the pci slot has been reset.  If a basic connection seems good
7279  * (registers are readable and have sane content) then return a
7280  * happy little PCI_ERS_RESULT_xxx.
7281  **/
7282 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
7283 {
7284 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7285 	pci_ers_result_t result;
7286 	int err;
7287 	u32 reg;
7288 
7289 	dev_info(&pdev->dev, "%s\n", __func__);
7290 	if (pci_enable_device_mem(pdev)) {
7291 		dev_info(&pdev->dev,
7292 			 "Cannot re-enable PCI device after reset.\n");
7293 		result = PCI_ERS_RESULT_DISCONNECT;
7294 	} else {
7295 		pci_set_master(pdev);
7296 		pci_restore_state(pdev);
7297 		pci_save_state(pdev);
7298 		pci_wake_from_d3(pdev, false);
7299 
7300 		reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
7301 		if (reg == 0)
7302 			result = PCI_ERS_RESULT_RECOVERED;
7303 		else
7304 			result = PCI_ERS_RESULT_DISCONNECT;
7305 	}
7306 
7307 	err = pci_cleanup_aer_uncorrect_error_status(pdev);
7308 	if (err) {
7309 		dev_info(&pdev->dev,
7310 			 "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
7311 			 err);
7312 		/* non-fatal, continue */
7313 	}
7314 
7315 	return result;
7316 }
7317 
7318 /**
7319  * i40e_pci_error_resume - restart operations after PCI error recovery
7320  * @pdev: PCI device information struct
7321  *
7322  * Called to allow the driver to bring things back up after PCI error
7323  * and/or reset recovery has finished.
7324  **/
7325 static void i40e_pci_error_resume(struct pci_dev *pdev)
7326 {
7327 	struct i40e_pf *pf = pci_get_drvdata(pdev);
7328 
7329 	dev_info(&pdev->dev, "%s\n", __func__);
7330 	i40e_handle_reset_warning(pf);
7331 }
7332 
7333 static const struct pci_error_handlers i40e_err_handler = {
7334 	.error_detected = i40e_pci_error_detected,
7335 	.slot_reset = i40e_pci_error_slot_reset,
7336 	.resume = i40e_pci_error_resume,
7337 };
7338 
7339 static struct pci_driver i40e_driver = {
7340 	.name     = i40e_driver_name,
7341 	.id_table = i40e_pci_tbl,
7342 	.probe    = i40e_probe,
7343 	.remove   = i40e_remove,
7344 	.err_handler = &i40e_err_handler,
7345 	.sriov_configure = i40e_pci_sriov_configure,
7346 };
7347 
7348 /**
7349  * i40e_init_module - Driver registration routine
7350  *
7351  * i40e_init_module is the first routine called when the driver is
7352  * loaded. All it does is register with the PCI subsystem.
7353  **/
7354 static int __init i40e_init_module(void)
7355 {
7356 	pr_info("%s: %s - version %s\n", i40e_driver_name,
7357 		i40e_driver_string, i40e_driver_version_str);
7358 	pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
7359 	i40e_dbg_init();
7360 	return pci_register_driver(&i40e_driver);
7361 }
7362 module_init(i40e_init_module);
7363 
7364 /**
7365  * i40e_exit_module - Driver exit cleanup routine
7366  *
7367  * i40e_exit_module is called just before the driver is removed
7368  * from memory.
7369  **/
7370 static void __exit i40e_exit_module(void)
7371 {
7372 	pci_unregister_driver(&i40e_driver);
7373 	i40e_dbg_exit();
7374 }
7375 module_exit(i40e_exit_module);
7376