Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[cascardo/linux.git] / drivers / net / ethernet / intel / i40e / i40e_main.c
1 /*******************************************************************************
2  *
3  * Intel Ethernet Controller XL710 Family Linux Driver
4  * Copyright(c) 2013 - 2016 Intel Corporation.
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along
16  * with this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * The full GNU General Public License is included in this distribution in
19  * the file called "COPYING".
20  *
21  * Contact Information:
22  * e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
23  * Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
24  *
25  ******************************************************************************/
26
27 #include <linux/etherdevice.h>
28 #include <linux/of_net.h>
29 #include <linux/pci.h>
30
31 /* Local includes */
32 #include "i40e.h"
33 #include "i40e_diag.h"
34 #include <net/udp_tunnel.h>
35
36 const char i40e_driver_name[] = "i40e";
37 static const char i40e_driver_string[] =
38                         "Intel(R) Ethernet Connection XL710 Network Driver";
39
40 #define DRV_KERN "-k"
41
42 #define DRV_VERSION_MAJOR 1
43 #define DRV_VERSION_MINOR 6
44 #define DRV_VERSION_BUILD 12
45 #define DRV_VERSION __stringify(DRV_VERSION_MAJOR) "." \
46              __stringify(DRV_VERSION_MINOR) "." \
47              __stringify(DRV_VERSION_BUILD)    DRV_KERN
48 const char i40e_driver_version_str[] = DRV_VERSION;
49 static const char i40e_copyright[] = "Copyright (c) 2013 - 2014 Intel Corporation.";
50
51 /* a bit of forward declarations */
52 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi);
53 static void i40e_handle_reset_warning(struct i40e_pf *pf);
54 static int i40e_add_vsi(struct i40e_vsi *vsi);
55 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi);
56 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit);
57 static int i40e_setup_misc_vector(struct i40e_pf *pf);
58 static void i40e_determine_queue_usage(struct i40e_pf *pf);
59 static int i40e_setup_pf_filter_control(struct i40e_pf *pf);
60 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
61                               u16 rss_table_size, u16 rss_size);
62 static void i40e_fdir_sb_setup(struct i40e_pf *pf);
63 static int i40e_veb_get_bw_info(struct i40e_veb *veb);
64
65 /* i40e_pci_tbl - PCI Device ID Table
66  *
67  * Last entry must be all 0s
68  *
69  * { Vendor ID, Device ID, SubVendor ID, SubDevice ID,
70  *   Class, Class Mask, private data (not used) }
71  */
72 static const struct pci_device_id i40e_pci_tbl[] = {
73         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_XL710), 0},
74         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QEMU), 0},
75         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_B), 0},
76         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_C), 0},
77         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_A), 0},
78         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_B), 0},
79         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_C), 0},
80         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T), 0},
81         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T4), 0},
82         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
83         {PCI_VDEVICE(INTEL, I40E_DEV_ID_KX_X722), 0},
84         {PCI_VDEVICE(INTEL, I40E_DEV_ID_QSFP_X722), 0},
85         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_X722), 0},
86         {PCI_VDEVICE(INTEL, I40E_DEV_ID_1G_BASE_T_X722), 0},
87         {PCI_VDEVICE(INTEL, I40E_DEV_ID_10G_BASE_T_X722), 0},
88         {PCI_VDEVICE(INTEL, I40E_DEV_ID_SFP_I_X722), 0},
89         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2), 0},
90         {PCI_VDEVICE(INTEL, I40E_DEV_ID_20G_KR2_A), 0},
91         /* required last entry */
92         {0, }
93 };
94 MODULE_DEVICE_TABLE(pci, i40e_pci_tbl);
95
96 #define I40E_MAX_VF_COUNT 128
97 static int debug = -1;
98 module_param(debug, int, 0);
99 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
100
101 MODULE_AUTHOR("Intel Corporation, <e1000-devel@lists.sourceforge.net>");
102 MODULE_DESCRIPTION("Intel(R) Ethernet Connection XL710 Network Driver");
103 MODULE_LICENSE("GPL");
104 MODULE_VERSION(DRV_VERSION);
105
106 static struct workqueue_struct *i40e_wq;
107
108 /**
109  * i40e_allocate_dma_mem_d - OS specific memory alloc for shared code
110  * @hw:   pointer to the HW structure
111  * @mem:  ptr to mem struct to fill out
112  * @size: size of memory requested
113  * @alignment: what to align the allocation to
114  **/
115 int i40e_allocate_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem,
116                             u64 size, u32 alignment)
117 {
118         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
119
120         mem->size = ALIGN(size, alignment);
121         mem->va = dma_zalloc_coherent(&pf->pdev->dev, mem->size,
122                                       &mem->pa, GFP_KERNEL);
123         if (!mem->va)
124                 return -ENOMEM;
125
126         return 0;
127 }
128
129 /**
130  * i40e_free_dma_mem_d - OS specific memory free for shared code
131  * @hw:   pointer to the HW structure
132  * @mem:  ptr to mem struct to free
133  **/
134 int i40e_free_dma_mem_d(struct i40e_hw *hw, struct i40e_dma_mem *mem)
135 {
136         struct i40e_pf *pf = (struct i40e_pf *)hw->back;
137
138         dma_free_coherent(&pf->pdev->dev, mem->size, mem->va, mem->pa);
139         mem->va = NULL;
140         mem->pa = 0;
141         mem->size = 0;
142
143         return 0;
144 }
145
146 /**
147  * i40e_allocate_virt_mem_d - OS specific memory alloc for shared code
148  * @hw:   pointer to the HW structure
149  * @mem:  ptr to mem struct to fill out
150  * @size: size of memory requested
151  **/
152 int i40e_allocate_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem,
153                              u32 size)
154 {
155         mem->size = size;
156         mem->va = kzalloc(size, GFP_KERNEL);
157
158         if (!mem->va)
159                 return -ENOMEM;
160
161         return 0;
162 }
163
164 /**
165  * i40e_free_virt_mem_d - OS specific memory free for shared code
166  * @hw:   pointer to the HW structure
167  * @mem:  ptr to mem struct to free
168  **/
169 int i40e_free_virt_mem_d(struct i40e_hw *hw, struct i40e_virt_mem *mem)
170 {
171         /* it's ok to kfree a NULL pointer */
172         kfree(mem->va);
173         mem->va = NULL;
174         mem->size = 0;
175
176         return 0;
177 }
178
179 /**
180  * i40e_get_lump - find a lump of free generic resource
181  * @pf: board private structure
182  * @pile: the pile of resource to search
183  * @needed: the number of items needed
184  * @id: an owner id to stick on the items assigned
185  *
186  * Returns the base item index of the lump, or negative for error
187  *
188  * The search_hint trick and lack of advanced fit-finding only work
189  * because we're highly likely to have all the same size lump requests.
190  * Linear search time and any fragmentation should be minimal.
191  **/
192 static int i40e_get_lump(struct i40e_pf *pf, struct i40e_lump_tracking *pile,
193                          u16 needed, u16 id)
194 {
195         int ret = -ENOMEM;
196         int i, j;
197
198         if (!pile || needed == 0 || id >= I40E_PILE_VALID_BIT) {
199                 dev_info(&pf->pdev->dev,
200                          "param err: pile=%p needed=%d id=0x%04x\n",
201                          pile, needed, id);
202                 return -EINVAL;
203         }
204
205         /* start the linear search with an imperfect hint */
206         i = pile->search_hint;
207         while (i < pile->num_entries) {
208                 /* skip already allocated entries */
209                 if (pile->list[i] & I40E_PILE_VALID_BIT) {
210                         i++;
211                         continue;
212                 }
213
214                 /* do we have enough in this lump? */
215                 for (j = 0; (j < needed) && ((i+j) < pile->num_entries); j++) {
216                         if (pile->list[i+j] & I40E_PILE_VALID_BIT)
217                                 break;
218                 }
219
220                 if (j == needed) {
221                         /* there was enough, so assign it to the requestor */
222                         for (j = 0; j < needed; j++)
223                                 pile->list[i+j] = id | I40E_PILE_VALID_BIT;
224                         ret = i;
225                         pile->search_hint = i + j;
226                         break;
227                 }
228
229                 /* not enough, so skip over it and continue looking */
230                 i += j;
231         }
232
233         return ret;
234 }
235
236 /**
237  * i40e_put_lump - return a lump of generic resource
238  * @pile: the pile of resource to search
239  * @index: the base item index
240  * @id: the owner id of the items assigned
241  *
242  * Returns the count of items in the lump
243  **/
244 static int i40e_put_lump(struct i40e_lump_tracking *pile, u16 index, u16 id)
245 {
246         int valid_id = (id | I40E_PILE_VALID_BIT);
247         int count = 0;
248         int i;
249
250         if (!pile || index >= pile->num_entries)
251                 return -EINVAL;
252
253         for (i = index;
254              i < pile->num_entries && pile->list[i] == valid_id;
255              i++) {
256                 pile->list[i] = 0;
257                 count++;
258         }
259
260         if (count && index < pile->search_hint)
261                 pile->search_hint = index;
262
263         return count;
264 }
265
266 /**
267  * i40e_find_vsi_from_id - searches for the vsi with the given id
268  * @pf - the pf structure to search for the vsi
269  * @id - id of the vsi it is searching for
270  **/
271 struct i40e_vsi *i40e_find_vsi_from_id(struct i40e_pf *pf, u16 id)
272 {
273         int i;
274
275         for (i = 0; i < pf->num_alloc_vsi; i++)
276                 if (pf->vsi[i] && (pf->vsi[i]->id == id))
277                         return pf->vsi[i];
278
279         return NULL;
280 }
281
282 /**
283  * i40e_service_event_schedule - Schedule the service task to wake up
284  * @pf: board private structure
285  *
286  * If not already scheduled, this puts the task into the work queue
287  **/
288 void i40e_service_event_schedule(struct i40e_pf *pf)
289 {
290         if (!test_bit(__I40E_DOWN, &pf->state) &&
291             !test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state) &&
292             !test_and_set_bit(__I40E_SERVICE_SCHED, &pf->state))
293                 queue_work(i40e_wq, &pf->service_task);
294 }
295
296 /**
297  * i40e_tx_timeout - Respond to a Tx Hang
298  * @netdev: network interface device structure
299  *
300  * If any port has noticed a Tx timeout, it is likely that the whole
301  * device is munged, not just the one netdev port, so go for the full
302  * reset.
303  **/
304 #ifdef I40E_FCOE
305 void i40e_tx_timeout(struct net_device *netdev)
306 #else
307 static void i40e_tx_timeout(struct net_device *netdev)
308 #endif
309 {
310         struct i40e_netdev_priv *np = netdev_priv(netdev);
311         struct i40e_vsi *vsi = np->vsi;
312         struct i40e_pf *pf = vsi->back;
313         struct i40e_ring *tx_ring = NULL;
314         unsigned int i, hung_queue = 0;
315         u32 head, val;
316
317         pf->tx_timeout_count++;
318
319         /* find the stopped queue the same way the stack does */
320         for (i = 0; i < netdev->num_tx_queues; i++) {
321                 struct netdev_queue *q;
322                 unsigned long trans_start;
323
324                 q = netdev_get_tx_queue(netdev, i);
325                 trans_start = q->trans_start;
326                 if (netif_xmit_stopped(q) &&
327                     time_after(jiffies,
328                                (trans_start + netdev->watchdog_timeo))) {
329                         hung_queue = i;
330                         break;
331                 }
332         }
333
334         if (i == netdev->num_tx_queues) {
335                 netdev_info(netdev, "tx_timeout: no netdev hung queue found\n");
336         } else {
337                 /* now that we have an index, find the tx_ring struct */
338                 for (i = 0; i < vsi->num_queue_pairs; i++) {
339                         if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
340                                 if (hung_queue ==
341                                     vsi->tx_rings[i]->queue_index) {
342                                         tx_ring = vsi->tx_rings[i];
343                                         break;
344                                 }
345                         }
346                 }
347         }
348
349         if (time_after(jiffies, (pf->tx_timeout_last_recovery + HZ*20)))
350                 pf->tx_timeout_recovery_level = 1;  /* reset after some time */
351         else if (time_before(jiffies,
352                       (pf->tx_timeout_last_recovery + netdev->watchdog_timeo)))
353                 return;   /* don't do any new action before the next timeout */
354
355         if (tx_ring) {
356                 head = i40e_get_head(tx_ring);
357                 /* Read interrupt register */
358                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
359                         val = rd32(&pf->hw,
360                              I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
361                                                 tx_ring->vsi->base_vector - 1));
362                 else
363                         val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
364
365                 netdev_info(netdev, "tx_timeout: VSI_seid: %d, Q %d, NTC: 0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x, INT: 0x%x\n",
366                             vsi->seid, hung_queue, tx_ring->next_to_clean,
367                             head, tx_ring->next_to_use,
368                             readl(tx_ring->tail), val);
369         }
370
371         pf->tx_timeout_last_recovery = jiffies;
372         netdev_info(netdev, "tx_timeout recovery level %d, hung_queue %d\n",
373                     pf->tx_timeout_recovery_level, hung_queue);
374
375         switch (pf->tx_timeout_recovery_level) {
376         case 1:
377                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
378                 break;
379         case 2:
380                 set_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
381                 break;
382         case 3:
383                 set_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
384                 break;
385         default:
386                 netdev_err(netdev, "tx_timeout recovery unsuccessful\n");
387                 break;
388         }
389
390         i40e_service_event_schedule(pf);
391         pf->tx_timeout_recovery_level++;
392 }
393
394 /**
395  * i40e_get_vsi_stats_struct - Get System Network Statistics
396  * @vsi: the VSI we care about
397  *
398  * Returns the address of the device statistics structure.
399  * The statistics are actually updated from the service task.
400  **/
401 struct rtnl_link_stats64 *i40e_get_vsi_stats_struct(struct i40e_vsi *vsi)
402 {
403         return &vsi->net_stats;
404 }
405
406 /**
407  * i40e_get_netdev_stats_struct - Get statistics for netdev interface
408  * @netdev: network interface device structure
409  *
410  * Returns the address of the device statistics structure.
411  * The statistics are actually updated from the service task.
412  **/
413 #ifdef I40E_FCOE
414 struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
415                                              struct net_device *netdev,
416                                              struct rtnl_link_stats64 *stats)
417 #else
418 static struct rtnl_link_stats64 *i40e_get_netdev_stats_struct(
419                                              struct net_device *netdev,
420                                              struct rtnl_link_stats64 *stats)
421 #endif
422 {
423         struct i40e_netdev_priv *np = netdev_priv(netdev);
424         struct i40e_ring *tx_ring, *rx_ring;
425         struct i40e_vsi *vsi = np->vsi;
426         struct rtnl_link_stats64 *vsi_stats = i40e_get_vsi_stats_struct(vsi);
427         int i;
428
429         if (test_bit(__I40E_DOWN, &vsi->state))
430                 return stats;
431
432         if (!vsi->tx_rings)
433                 return stats;
434
435         rcu_read_lock();
436         for (i = 0; i < vsi->num_queue_pairs; i++) {
437                 u64 bytes, packets;
438                 unsigned int start;
439
440                 tx_ring = ACCESS_ONCE(vsi->tx_rings[i]);
441                 if (!tx_ring)
442                         continue;
443
444                 do {
445                         start = u64_stats_fetch_begin_irq(&tx_ring->syncp);
446                         packets = tx_ring->stats.packets;
447                         bytes   = tx_ring->stats.bytes;
448                 } while (u64_stats_fetch_retry_irq(&tx_ring->syncp, start));
449
450                 stats->tx_packets += packets;
451                 stats->tx_bytes   += bytes;
452                 rx_ring = &tx_ring[1];
453
454                 do {
455                         start = u64_stats_fetch_begin_irq(&rx_ring->syncp);
456                         packets = rx_ring->stats.packets;
457                         bytes   = rx_ring->stats.bytes;
458                 } while (u64_stats_fetch_retry_irq(&rx_ring->syncp, start));
459
460                 stats->rx_packets += packets;
461                 stats->rx_bytes   += bytes;
462         }
463         rcu_read_unlock();
464
465         /* following stats updated by i40e_watchdog_subtask() */
466         stats->multicast        = vsi_stats->multicast;
467         stats->tx_errors        = vsi_stats->tx_errors;
468         stats->tx_dropped       = vsi_stats->tx_dropped;
469         stats->rx_errors        = vsi_stats->rx_errors;
470         stats->rx_dropped       = vsi_stats->rx_dropped;
471         stats->rx_crc_errors    = vsi_stats->rx_crc_errors;
472         stats->rx_length_errors = vsi_stats->rx_length_errors;
473
474         return stats;
475 }
476
477 /**
478  * i40e_vsi_reset_stats - Resets all stats of the given vsi
479  * @vsi: the VSI to have its stats reset
480  **/
481 void i40e_vsi_reset_stats(struct i40e_vsi *vsi)
482 {
483         struct rtnl_link_stats64 *ns;
484         int i;
485
486         if (!vsi)
487                 return;
488
489         ns = i40e_get_vsi_stats_struct(vsi);
490         memset(ns, 0, sizeof(*ns));
491         memset(&vsi->net_stats_offsets, 0, sizeof(vsi->net_stats_offsets));
492         memset(&vsi->eth_stats, 0, sizeof(vsi->eth_stats));
493         memset(&vsi->eth_stats_offsets, 0, sizeof(vsi->eth_stats_offsets));
494         if (vsi->rx_rings && vsi->rx_rings[0]) {
495                 for (i = 0; i < vsi->num_queue_pairs; i++) {
496                         memset(&vsi->rx_rings[i]->stats, 0,
497                                sizeof(vsi->rx_rings[i]->stats));
498                         memset(&vsi->rx_rings[i]->rx_stats, 0,
499                                sizeof(vsi->rx_rings[i]->rx_stats));
500                         memset(&vsi->tx_rings[i]->stats, 0,
501                                sizeof(vsi->tx_rings[i]->stats));
502                         memset(&vsi->tx_rings[i]->tx_stats, 0,
503                                sizeof(vsi->tx_rings[i]->tx_stats));
504                 }
505         }
506         vsi->stat_offsets_loaded = false;
507 }
508
509 /**
510  * i40e_pf_reset_stats - Reset all of the stats for the given PF
511  * @pf: the PF to be reset
512  **/
513 void i40e_pf_reset_stats(struct i40e_pf *pf)
514 {
515         int i;
516
517         memset(&pf->stats, 0, sizeof(pf->stats));
518         memset(&pf->stats_offsets, 0, sizeof(pf->stats_offsets));
519         pf->stat_offsets_loaded = false;
520
521         for (i = 0; i < I40E_MAX_VEB; i++) {
522                 if (pf->veb[i]) {
523                         memset(&pf->veb[i]->stats, 0,
524                                sizeof(pf->veb[i]->stats));
525                         memset(&pf->veb[i]->stats_offsets, 0,
526                                sizeof(pf->veb[i]->stats_offsets));
527                         pf->veb[i]->stat_offsets_loaded = false;
528                 }
529         }
530         pf->hw_csum_rx_error = 0;
531 }
532
533 /**
534  * i40e_stat_update48 - read and update a 48 bit stat from the chip
535  * @hw: ptr to the hardware info
536  * @hireg: the high 32 bit reg to read
537  * @loreg: the low 32 bit reg to read
538  * @offset_loaded: has the initial offset been loaded yet
539  * @offset: ptr to current offset value
540  * @stat: ptr to the stat
541  *
542  * Since the device stats are not reset at PFReset, they likely will not
543  * be zeroed when the driver starts.  We'll save the first values read
544  * and use them as offsets to be subtracted from the raw values in order
545  * to report stats that count from zero.  In the process, we also manage
546  * the potential roll-over.
547  **/
548 static void i40e_stat_update48(struct i40e_hw *hw, u32 hireg, u32 loreg,
549                                bool offset_loaded, u64 *offset, u64 *stat)
550 {
551         u64 new_data;
552
553         if (hw->device_id == I40E_DEV_ID_QEMU) {
554                 new_data = rd32(hw, loreg);
555                 new_data |= ((u64)(rd32(hw, hireg) & 0xFFFF)) << 32;
556         } else {
557                 new_data = rd64(hw, loreg);
558         }
559         if (!offset_loaded)
560                 *offset = new_data;
561         if (likely(new_data >= *offset))
562                 *stat = new_data - *offset;
563         else
564                 *stat = (new_data + BIT_ULL(48)) - *offset;
565         *stat &= 0xFFFFFFFFFFFFULL;
566 }
567
568 /**
569  * i40e_stat_update32 - read and update a 32 bit stat from the chip
570  * @hw: ptr to the hardware info
571  * @reg: the hw reg to read
572  * @offset_loaded: has the initial offset been loaded yet
573  * @offset: ptr to current offset value
574  * @stat: ptr to the stat
575  **/
576 static void i40e_stat_update32(struct i40e_hw *hw, u32 reg,
577                                bool offset_loaded, u64 *offset, u64 *stat)
578 {
579         u32 new_data;
580
581         new_data = rd32(hw, reg);
582         if (!offset_loaded)
583                 *offset = new_data;
584         if (likely(new_data >= *offset))
585                 *stat = (u32)(new_data - *offset);
586         else
587                 *stat = (u32)((new_data + BIT_ULL(32)) - *offset);
588 }
589
590 /**
591  * i40e_update_eth_stats - Update VSI-specific ethernet statistics counters.
592  * @vsi: the VSI to be updated
593  **/
594 void i40e_update_eth_stats(struct i40e_vsi *vsi)
595 {
596         int stat_idx = le16_to_cpu(vsi->info.stat_counter_idx);
597         struct i40e_pf *pf = vsi->back;
598         struct i40e_hw *hw = &pf->hw;
599         struct i40e_eth_stats *oes;
600         struct i40e_eth_stats *es;     /* device's eth stats */
601
602         es = &vsi->eth_stats;
603         oes = &vsi->eth_stats_offsets;
604
605         /* Gather up the stats that the hw collects */
606         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
607                            vsi->stat_offsets_loaded,
608                            &oes->tx_errors, &es->tx_errors);
609         i40e_stat_update32(hw, I40E_GLV_RDPC(stat_idx),
610                            vsi->stat_offsets_loaded,
611                            &oes->rx_discards, &es->rx_discards);
612         i40e_stat_update32(hw, I40E_GLV_RUPP(stat_idx),
613                            vsi->stat_offsets_loaded,
614                            &oes->rx_unknown_protocol, &es->rx_unknown_protocol);
615         i40e_stat_update32(hw, I40E_GLV_TEPC(stat_idx),
616                            vsi->stat_offsets_loaded,
617                            &oes->tx_errors, &es->tx_errors);
618
619         i40e_stat_update48(hw, I40E_GLV_GORCH(stat_idx),
620                            I40E_GLV_GORCL(stat_idx),
621                            vsi->stat_offsets_loaded,
622                            &oes->rx_bytes, &es->rx_bytes);
623         i40e_stat_update48(hw, I40E_GLV_UPRCH(stat_idx),
624                            I40E_GLV_UPRCL(stat_idx),
625                            vsi->stat_offsets_loaded,
626                            &oes->rx_unicast, &es->rx_unicast);
627         i40e_stat_update48(hw, I40E_GLV_MPRCH(stat_idx),
628                            I40E_GLV_MPRCL(stat_idx),
629                            vsi->stat_offsets_loaded,
630                            &oes->rx_multicast, &es->rx_multicast);
631         i40e_stat_update48(hw, I40E_GLV_BPRCH(stat_idx),
632                            I40E_GLV_BPRCL(stat_idx),
633                            vsi->stat_offsets_loaded,
634                            &oes->rx_broadcast, &es->rx_broadcast);
635
636         i40e_stat_update48(hw, I40E_GLV_GOTCH(stat_idx),
637                            I40E_GLV_GOTCL(stat_idx),
638                            vsi->stat_offsets_loaded,
639                            &oes->tx_bytes, &es->tx_bytes);
640         i40e_stat_update48(hw, I40E_GLV_UPTCH(stat_idx),
641                            I40E_GLV_UPTCL(stat_idx),
642                            vsi->stat_offsets_loaded,
643                            &oes->tx_unicast, &es->tx_unicast);
644         i40e_stat_update48(hw, I40E_GLV_MPTCH(stat_idx),
645                            I40E_GLV_MPTCL(stat_idx),
646                            vsi->stat_offsets_loaded,
647                            &oes->tx_multicast, &es->tx_multicast);
648         i40e_stat_update48(hw, I40E_GLV_BPTCH(stat_idx),
649                            I40E_GLV_BPTCL(stat_idx),
650                            vsi->stat_offsets_loaded,
651                            &oes->tx_broadcast, &es->tx_broadcast);
652         vsi->stat_offsets_loaded = true;
653 }
654
655 /**
656  * i40e_update_veb_stats - Update Switch component statistics
657  * @veb: the VEB being updated
658  **/
659 static void i40e_update_veb_stats(struct i40e_veb *veb)
660 {
661         struct i40e_pf *pf = veb->pf;
662         struct i40e_hw *hw = &pf->hw;
663         struct i40e_eth_stats *oes;
664         struct i40e_eth_stats *es;     /* device's eth stats */
665         struct i40e_veb_tc_stats *veb_oes;
666         struct i40e_veb_tc_stats *veb_es;
667         int i, idx = 0;
668
669         idx = veb->stats_idx;
670         es = &veb->stats;
671         oes = &veb->stats_offsets;
672         veb_es = &veb->tc_stats;
673         veb_oes = &veb->tc_stats_offsets;
674
675         /* Gather up the stats that the hw collects */
676         i40e_stat_update32(hw, I40E_GLSW_TDPC(idx),
677                            veb->stat_offsets_loaded,
678                            &oes->tx_discards, &es->tx_discards);
679         if (hw->revision_id > 0)
680                 i40e_stat_update32(hw, I40E_GLSW_RUPP(idx),
681                                    veb->stat_offsets_loaded,
682                                    &oes->rx_unknown_protocol,
683                                    &es->rx_unknown_protocol);
684         i40e_stat_update48(hw, I40E_GLSW_GORCH(idx), I40E_GLSW_GORCL(idx),
685                            veb->stat_offsets_loaded,
686                            &oes->rx_bytes, &es->rx_bytes);
687         i40e_stat_update48(hw, I40E_GLSW_UPRCH(idx), I40E_GLSW_UPRCL(idx),
688                            veb->stat_offsets_loaded,
689                            &oes->rx_unicast, &es->rx_unicast);
690         i40e_stat_update48(hw, I40E_GLSW_MPRCH(idx), I40E_GLSW_MPRCL(idx),
691                            veb->stat_offsets_loaded,
692                            &oes->rx_multicast, &es->rx_multicast);
693         i40e_stat_update48(hw, I40E_GLSW_BPRCH(idx), I40E_GLSW_BPRCL(idx),
694                            veb->stat_offsets_loaded,
695                            &oes->rx_broadcast, &es->rx_broadcast);
696
697         i40e_stat_update48(hw, I40E_GLSW_GOTCH(idx), I40E_GLSW_GOTCL(idx),
698                            veb->stat_offsets_loaded,
699                            &oes->tx_bytes, &es->tx_bytes);
700         i40e_stat_update48(hw, I40E_GLSW_UPTCH(idx), I40E_GLSW_UPTCL(idx),
701                            veb->stat_offsets_loaded,
702                            &oes->tx_unicast, &es->tx_unicast);
703         i40e_stat_update48(hw, I40E_GLSW_MPTCH(idx), I40E_GLSW_MPTCL(idx),
704                            veb->stat_offsets_loaded,
705                            &oes->tx_multicast, &es->tx_multicast);
706         i40e_stat_update48(hw, I40E_GLSW_BPTCH(idx), I40E_GLSW_BPTCL(idx),
707                            veb->stat_offsets_loaded,
708                            &oes->tx_broadcast, &es->tx_broadcast);
709         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
710                 i40e_stat_update48(hw, I40E_GLVEBTC_RPCH(i, idx),
711                                    I40E_GLVEBTC_RPCL(i, idx),
712                                    veb->stat_offsets_loaded,
713                                    &veb_oes->tc_rx_packets[i],
714                                    &veb_es->tc_rx_packets[i]);
715                 i40e_stat_update48(hw, I40E_GLVEBTC_RBCH(i, idx),
716                                    I40E_GLVEBTC_RBCL(i, idx),
717                                    veb->stat_offsets_loaded,
718                                    &veb_oes->tc_rx_bytes[i],
719                                    &veb_es->tc_rx_bytes[i]);
720                 i40e_stat_update48(hw, I40E_GLVEBTC_TPCH(i, idx),
721                                    I40E_GLVEBTC_TPCL(i, idx),
722                                    veb->stat_offsets_loaded,
723                                    &veb_oes->tc_tx_packets[i],
724                                    &veb_es->tc_tx_packets[i]);
725                 i40e_stat_update48(hw, I40E_GLVEBTC_TBCH(i, idx),
726                                    I40E_GLVEBTC_TBCL(i, idx),
727                                    veb->stat_offsets_loaded,
728                                    &veb_oes->tc_tx_bytes[i],
729                                    &veb_es->tc_tx_bytes[i]);
730         }
731         veb->stat_offsets_loaded = true;
732 }
733
734 #ifdef I40E_FCOE
735 /**
736  * i40e_update_fcoe_stats - Update FCoE-specific ethernet statistics counters.
737  * @vsi: the VSI that is capable of doing FCoE
738  **/
739 static void i40e_update_fcoe_stats(struct i40e_vsi *vsi)
740 {
741         struct i40e_pf *pf = vsi->back;
742         struct i40e_hw *hw = &pf->hw;
743         struct i40e_fcoe_stats *ofs;
744         struct i40e_fcoe_stats *fs;     /* device's eth stats */
745         int idx;
746
747         if (vsi->type != I40E_VSI_FCOE)
748                 return;
749
750         idx = hw->pf_id + I40E_FCOE_PF_STAT_OFFSET;
751         fs = &vsi->fcoe_stats;
752         ofs = &vsi->fcoe_stats_offsets;
753
754         i40e_stat_update32(hw, I40E_GL_FCOEPRC(idx),
755                            vsi->fcoe_stat_offsets_loaded,
756                            &ofs->rx_fcoe_packets, &fs->rx_fcoe_packets);
757         i40e_stat_update48(hw, I40E_GL_FCOEDWRCH(idx), I40E_GL_FCOEDWRCL(idx),
758                            vsi->fcoe_stat_offsets_loaded,
759                            &ofs->rx_fcoe_dwords, &fs->rx_fcoe_dwords);
760         i40e_stat_update32(hw, I40E_GL_FCOERPDC(idx),
761                            vsi->fcoe_stat_offsets_loaded,
762                            &ofs->rx_fcoe_dropped, &fs->rx_fcoe_dropped);
763         i40e_stat_update32(hw, I40E_GL_FCOEPTC(idx),
764                            vsi->fcoe_stat_offsets_loaded,
765                            &ofs->tx_fcoe_packets, &fs->tx_fcoe_packets);
766         i40e_stat_update48(hw, I40E_GL_FCOEDWTCH(idx), I40E_GL_FCOEDWTCL(idx),
767                            vsi->fcoe_stat_offsets_loaded,
768                            &ofs->tx_fcoe_dwords, &fs->tx_fcoe_dwords);
769         i40e_stat_update32(hw, I40E_GL_FCOECRC(idx),
770                            vsi->fcoe_stat_offsets_loaded,
771                            &ofs->fcoe_bad_fccrc, &fs->fcoe_bad_fccrc);
772         i40e_stat_update32(hw, I40E_GL_FCOELAST(idx),
773                            vsi->fcoe_stat_offsets_loaded,
774                            &ofs->fcoe_last_error, &fs->fcoe_last_error);
775         i40e_stat_update32(hw, I40E_GL_FCOEDDPC(idx),
776                            vsi->fcoe_stat_offsets_loaded,
777                            &ofs->fcoe_ddp_count, &fs->fcoe_ddp_count);
778
779         vsi->fcoe_stat_offsets_loaded = true;
780 }
781
782 #endif
783 /**
784  * i40e_update_vsi_stats - Update the vsi statistics counters.
785  * @vsi: the VSI to be updated
786  *
787  * There are a few instances where we store the same stat in a
788  * couple of different structs.  This is partly because we have
789  * the netdev stats that need to be filled out, which is slightly
790  * different from the "eth_stats" defined by the chip and used in
791  * VF communications.  We sort it out here.
792  **/
793 static void i40e_update_vsi_stats(struct i40e_vsi *vsi)
794 {
795         struct i40e_pf *pf = vsi->back;
796         struct rtnl_link_stats64 *ons;
797         struct rtnl_link_stats64 *ns;   /* netdev stats */
798         struct i40e_eth_stats *oes;
799         struct i40e_eth_stats *es;     /* device's eth stats */
800         u32 tx_restart, tx_busy;
801         u64 tx_lost_interrupt;
802         struct i40e_ring *p;
803         u32 rx_page, rx_buf;
804         u64 bytes, packets;
805         unsigned int start;
806         u64 tx_linearize;
807         u64 tx_force_wb;
808         u64 rx_p, rx_b;
809         u64 tx_p, tx_b;
810         u16 q;
811
812         if (test_bit(__I40E_DOWN, &vsi->state) ||
813             test_bit(__I40E_CONFIG_BUSY, &pf->state))
814                 return;
815
816         ns = i40e_get_vsi_stats_struct(vsi);
817         ons = &vsi->net_stats_offsets;
818         es = &vsi->eth_stats;
819         oes = &vsi->eth_stats_offsets;
820
821         /* Gather up the netdev and vsi stats that the driver collects
822          * on the fly during packet processing
823          */
824         rx_b = rx_p = 0;
825         tx_b = tx_p = 0;
826         tx_restart = tx_busy = tx_linearize = tx_force_wb = 0;
827         tx_lost_interrupt = 0;
828         rx_page = 0;
829         rx_buf = 0;
830         rcu_read_lock();
831         for (q = 0; q < vsi->num_queue_pairs; q++) {
832                 /* locate Tx ring */
833                 p = ACCESS_ONCE(vsi->tx_rings[q]);
834
835                 do {
836                         start = u64_stats_fetch_begin_irq(&p->syncp);
837                         packets = p->stats.packets;
838                         bytes = p->stats.bytes;
839                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
840                 tx_b += bytes;
841                 tx_p += packets;
842                 tx_restart += p->tx_stats.restart_queue;
843                 tx_busy += p->tx_stats.tx_busy;
844                 tx_linearize += p->tx_stats.tx_linearize;
845                 tx_force_wb += p->tx_stats.tx_force_wb;
846                 tx_lost_interrupt += p->tx_stats.tx_lost_interrupt;
847
848                 /* Rx queue is part of the same block as Tx queue */
849                 p = &p[1];
850                 do {
851                         start = u64_stats_fetch_begin_irq(&p->syncp);
852                         packets = p->stats.packets;
853                         bytes = p->stats.bytes;
854                 } while (u64_stats_fetch_retry_irq(&p->syncp, start));
855                 rx_b += bytes;
856                 rx_p += packets;
857                 rx_buf += p->rx_stats.alloc_buff_failed;
858                 rx_page += p->rx_stats.alloc_page_failed;
859         }
860         rcu_read_unlock();
861         vsi->tx_restart = tx_restart;
862         vsi->tx_busy = tx_busy;
863         vsi->tx_linearize = tx_linearize;
864         vsi->tx_force_wb = tx_force_wb;
865         vsi->tx_lost_interrupt = tx_lost_interrupt;
866         vsi->rx_page_failed = rx_page;
867         vsi->rx_buf_failed = rx_buf;
868
869         ns->rx_packets = rx_p;
870         ns->rx_bytes = rx_b;
871         ns->tx_packets = tx_p;
872         ns->tx_bytes = tx_b;
873
874         /* update netdev stats from eth stats */
875         i40e_update_eth_stats(vsi);
876         ons->tx_errors = oes->tx_errors;
877         ns->tx_errors = es->tx_errors;
878         ons->multicast = oes->rx_multicast;
879         ns->multicast = es->rx_multicast;
880         ons->rx_dropped = oes->rx_discards;
881         ns->rx_dropped = es->rx_discards;
882         ons->tx_dropped = oes->tx_discards;
883         ns->tx_dropped = es->tx_discards;
884
885         /* pull in a couple PF stats if this is the main vsi */
886         if (vsi == pf->vsi[pf->lan_vsi]) {
887                 ns->rx_crc_errors = pf->stats.crc_errors;
888                 ns->rx_errors = pf->stats.crc_errors + pf->stats.illegal_bytes;
889                 ns->rx_length_errors = pf->stats.rx_length_errors;
890         }
891 }
892
893 /**
894  * i40e_update_pf_stats - Update the PF statistics counters.
895  * @pf: the PF to be updated
896  **/
897 static void i40e_update_pf_stats(struct i40e_pf *pf)
898 {
899         struct i40e_hw_port_stats *osd = &pf->stats_offsets;
900         struct i40e_hw_port_stats *nsd = &pf->stats;
901         struct i40e_hw *hw = &pf->hw;
902         u32 val;
903         int i;
904
905         i40e_stat_update48(hw, I40E_GLPRT_GORCH(hw->port),
906                            I40E_GLPRT_GORCL(hw->port),
907                            pf->stat_offsets_loaded,
908                            &osd->eth.rx_bytes, &nsd->eth.rx_bytes);
909         i40e_stat_update48(hw, I40E_GLPRT_GOTCH(hw->port),
910                            I40E_GLPRT_GOTCL(hw->port),
911                            pf->stat_offsets_loaded,
912                            &osd->eth.tx_bytes, &nsd->eth.tx_bytes);
913         i40e_stat_update32(hw, I40E_GLPRT_RDPC(hw->port),
914                            pf->stat_offsets_loaded,
915                            &osd->eth.rx_discards,
916                            &nsd->eth.rx_discards);
917         i40e_stat_update48(hw, I40E_GLPRT_UPRCH(hw->port),
918                            I40E_GLPRT_UPRCL(hw->port),
919                            pf->stat_offsets_loaded,
920                            &osd->eth.rx_unicast,
921                            &nsd->eth.rx_unicast);
922         i40e_stat_update48(hw, I40E_GLPRT_MPRCH(hw->port),
923                            I40E_GLPRT_MPRCL(hw->port),
924                            pf->stat_offsets_loaded,
925                            &osd->eth.rx_multicast,
926                            &nsd->eth.rx_multicast);
927         i40e_stat_update48(hw, I40E_GLPRT_BPRCH(hw->port),
928                            I40E_GLPRT_BPRCL(hw->port),
929                            pf->stat_offsets_loaded,
930                            &osd->eth.rx_broadcast,
931                            &nsd->eth.rx_broadcast);
932         i40e_stat_update48(hw, I40E_GLPRT_UPTCH(hw->port),
933                            I40E_GLPRT_UPTCL(hw->port),
934                            pf->stat_offsets_loaded,
935                            &osd->eth.tx_unicast,
936                            &nsd->eth.tx_unicast);
937         i40e_stat_update48(hw, I40E_GLPRT_MPTCH(hw->port),
938                            I40E_GLPRT_MPTCL(hw->port),
939                            pf->stat_offsets_loaded,
940                            &osd->eth.tx_multicast,
941                            &nsd->eth.tx_multicast);
942         i40e_stat_update48(hw, I40E_GLPRT_BPTCH(hw->port),
943                            I40E_GLPRT_BPTCL(hw->port),
944                            pf->stat_offsets_loaded,
945                            &osd->eth.tx_broadcast,
946                            &nsd->eth.tx_broadcast);
947
948         i40e_stat_update32(hw, I40E_GLPRT_TDOLD(hw->port),
949                            pf->stat_offsets_loaded,
950                            &osd->tx_dropped_link_down,
951                            &nsd->tx_dropped_link_down);
952
953         i40e_stat_update32(hw, I40E_GLPRT_CRCERRS(hw->port),
954                            pf->stat_offsets_loaded,
955                            &osd->crc_errors, &nsd->crc_errors);
956
957         i40e_stat_update32(hw, I40E_GLPRT_ILLERRC(hw->port),
958                            pf->stat_offsets_loaded,
959                            &osd->illegal_bytes, &nsd->illegal_bytes);
960
961         i40e_stat_update32(hw, I40E_GLPRT_MLFC(hw->port),
962                            pf->stat_offsets_loaded,
963                            &osd->mac_local_faults,
964                            &nsd->mac_local_faults);
965         i40e_stat_update32(hw, I40E_GLPRT_MRFC(hw->port),
966                            pf->stat_offsets_loaded,
967                            &osd->mac_remote_faults,
968                            &nsd->mac_remote_faults);
969
970         i40e_stat_update32(hw, I40E_GLPRT_RLEC(hw->port),
971                            pf->stat_offsets_loaded,
972                            &osd->rx_length_errors,
973                            &nsd->rx_length_errors);
974
975         i40e_stat_update32(hw, I40E_GLPRT_LXONRXC(hw->port),
976                            pf->stat_offsets_loaded,
977                            &osd->link_xon_rx, &nsd->link_xon_rx);
978         i40e_stat_update32(hw, I40E_GLPRT_LXONTXC(hw->port),
979                            pf->stat_offsets_loaded,
980                            &osd->link_xon_tx, &nsd->link_xon_tx);
981         i40e_stat_update32(hw, I40E_GLPRT_LXOFFRXC(hw->port),
982                            pf->stat_offsets_loaded,
983                            &osd->link_xoff_rx, &nsd->link_xoff_rx);
984         i40e_stat_update32(hw, I40E_GLPRT_LXOFFTXC(hw->port),
985                            pf->stat_offsets_loaded,
986                            &osd->link_xoff_tx, &nsd->link_xoff_tx);
987
988         for (i = 0; i < 8; i++) {
989                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFRXC(hw->port, i),
990                                    pf->stat_offsets_loaded,
991                                    &osd->priority_xoff_rx[i],
992                                    &nsd->priority_xoff_rx[i]);
993                 i40e_stat_update32(hw, I40E_GLPRT_PXONRXC(hw->port, i),
994                                    pf->stat_offsets_loaded,
995                                    &osd->priority_xon_rx[i],
996                                    &nsd->priority_xon_rx[i]);
997                 i40e_stat_update32(hw, I40E_GLPRT_PXONTXC(hw->port, i),
998                                    pf->stat_offsets_loaded,
999                                    &osd->priority_xon_tx[i],
1000                                    &nsd->priority_xon_tx[i]);
1001                 i40e_stat_update32(hw, I40E_GLPRT_PXOFFTXC(hw->port, i),
1002                                    pf->stat_offsets_loaded,
1003                                    &osd->priority_xoff_tx[i],
1004                                    &nsd->priority_xoff_tx[i]);
1005                 i40e_stat_update32(hw,
1006                                    I40E_GLPRT_RXON2OFFCNT(hw->port, i),
1007                                    pf->stat_offsets_loaded,
1008                                    &osd->priority_xon_2_xoff[i],
1009                                    &nsd->priority_xon_2_xoff[i]);
1010         }
1011
1012         i40e_stat_update48(hw, I40E_GLPRT_PRC64H(hw->port),
1013                            I40E_GLPRT_PRC64L(hw->port),
1014                            pf->stat_offsets_loaded,
1015                            &osd->rx_size_64, &nsd->rx_size_64);
1016         i40e_stat_update48(hw, I40E_GLPRT_PRC127H(hw->port),
1017                            I40E_GLPRT_PRC127L(hw->port),
1018                            pf->stat_offsets_loaded,
1019                            &osd->rx_size_127, &nsd->rx_size_127);
1020         i40e_stat_update48(hw, I40E_GLPRT_PRC255H(hw->port),
1021                            I40E_GLPRT_PRC255L(hw->port),
1022                            pf->stat_offsets_loaded,
1023                            &osd->rx_size_255, &nsd->rx_size_255);
1024         i40e_stat_update48(hw, I40E_GLPRT_PRC511H(hw->port),
1025                            I40E_GLPRT_PRC511L(hw->port),
1026                            pf->stat_offsets_loaded,
1027                            &osd->rx_size_511, &nsd->rx_size_511);
1028         i40e_stat_update48(hw, I40E_GLPRT_PRC1023H(hw->port),
1029                            I40E_GLPRT_PRC1023L(hw->port),
1030                            pf->stat_offsets_loaded,
1031                            &osd->rx_size_1023, &nsd->rx_size_1023);
1032         i40e_stat_update48(hw, I40E_GLPRT_PRC1522H(hw->port),
1033                            I40E_GLPRT_PRC1522L(hw->port),
1034                            pf->stat_offsets_loaded,
1035                            &osd->rx_size_1522, &nsd->rx_size_1522);
1036         i40e_stat_update48(hw, I40E_GLPRT_PRC9522H(hw->port),
1037                            I40E_GLPRT_PRC9522L(hw->port),
1038                            pf->stat_offsets_loaded,
1039                            &osd->rx_size_big, &nsd->rx_size_big);
1040
1041         i40e_stat_update48(hw, I40E_GLPRT_PTC64H(hw->port),
1042                            I40E_GLPRT_PTC64L(hw->port),
1043                            pf->stat_offsets_loaded,
1044                            &osd->tx_size_64, &nsd->tx_size_64);
1045         i40e_stat_update48(hw, I40E_GLPRT_PTC127H(hw->port),
1046                            I40E_GLPRT_PTC127L(hw->port),
1047                            pf->stat_offsets_loaded,
1048                            &osd->tx_size_127, &nsd->tx_size_127);
1049         i40e_stat_update48(hw, I40E_GLPRT_PTC255H(hw->port),
1050                            I40E_GLPRT_PTC255L(hw->port),
1051                            pf->stat_offsets_loaded,
1052                            &osd->tx_size_255, &nsd->tx_size_255);
1053         i40e_stat_update48(hw, I40E_GLPRT_PTC511H(hw->port),
1054                            I40E_GLPRT_PTC511L(hw->port),
1055                            pf->stat_offsets_loaded,
1056                            &osd->tx_size_511, &nsd->tx_size_511);
1057         i40e_stat_update48(hw, I40E_GLPRT_PTC1023H(hw->port),
1058                            I40E_GLPRT_PTC1023L(hw->port),
1059                            pf->stat_offsets_loaded,
1060                            &osd->tx_size_1023, &nsd->tx_size_1023);
1061         i40e_stat_update48(hw, I40E_GLPRT_PTC1522H(hw->port),
1062                            I40E_GLPRT_PTC1522L(hw->port),
1063                            pf->stat_offsets_loaded,
1064                            &osd->tx_size_1522, &nsd->tx_size_1522);
1065         i40e_stat_update48(hw, I40E_GLPRT_PTC9522H(hw->port),
1066                            I40E_GLPRT_PTC9522L(hw->port),
1067                            pf->stat_offsets_loaded,
1068                            &osd->tx_size_big, &nsd->tx_size_big);
1069
1070         i40e_stat_update32(hw, I40E_GLPRT_RUC(hw->port),
1071                            pf->stat_offsets_loaded,
1072                            &osd->rx_undersize, &nsd->rx_undersize);
1073         i40e_stat_update32(hw, I40E_GLPRT_RFC(hw->port),
1074                            pf->stat_offsets_loaded,
1075                            &osd->rx_fragments, &nsd->rx_fragments);
1076         i40e_stat_update32(hw, I40E_GLPRT_ROC(hw->port),
1077                            pf->stat_offsets_loaded,
1078                            &osd->rx_oversize, &nsd->rx_oversize);
1079         i40e_stat_update32(hw, I40E_GLPRT_RJC(hw->port),
1080                            pf->stat_offsets_loaded,
1081                            &osd->rx_jabber, &nsd->rx_jabber);
1082
1083         /* FDIR stats */
1084         i40e_stat_update32(hw,
1085                            I40E_GLQF_PCNT(I40E_FD_ATR_STAT_IDX(pf->hw.pf_id)),
1086                            pf->stat_offsets_loaded,
1087                            &osd->fd_atr_match, &nsd->fd_atr_match);
1088         i40e_stat_update32(hw,
1089                            I40E_GLQF_PCNT(I40E_FD_SB_STAT_IDX(pf->hw.pf_id)),
1090                            pf->stat_offsets_loaded,
1091                            &osd->fd_sb_match, &nsd->fd_sb_match);
1092         i40e_stat_update32(hw,
1093                       I40E_GLQF_PCNT(I40E_FD_ATR_TUNNEL_STAT_IDX(pf->hw.pf_id)),
1094                       pf->stat_offsets_loaded,
1095                       &osd->fd_atr_tunnel_match, &nsd->fd_atr_tunnel_match);
1096
1097         val = rd32(hw, I40E_PRTPM_EEE_STAT);
1098         nsd->tx_lpi_status =
1099                        (val & I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_MASK) >>
1100                         I40E_PRTPM_EEE_STAT_TX_LPI_STATUS_SHIFT;
1101         nsd->rx_lpi_status =
1102                        (val & I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_MASK) >>
1103                         I40E_PRTPM_EEE_STAT_RX_LPI_STATUS_SHIFT;
1104         i40e_stat_update32(hw, I40E_PRTPM_TLPIC,
1105                            pf->stat_offsets_loaded,
1106                            &osd->tx_lpi_count, &nsd->tx_lpi_count);
1107         i40e_stat_update32(hw, I40E_PRTPM_RLPIC,
1108                            pf->stat_offsets_loaded,
1109                            &osd->rx_lpi_count, &nsd->rx_lpi_count);
1110
1111         if (pf->flags & I40E_FLAG_FD_SB_ENABLED &&
1112             !(pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED))
1113                 nsd->fd_sb_status = true;
1114         else
1115                 nsd->fd_sb_status = false;
1116
1117         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED &&
1118             !(pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
1119                 nsd->fd_atr_status = true;
1120         else
1121                 nsd->fd_atr_status = false;
1122
1123         pf->stat_offsets_loaded = true;
1124 }
1125
1126 /**
1127  * i40e_update_stats - Update the various statistics counters.
1128  * @vsi: the VSI to be updated
1129  *
1130  * Update the various stats for this VSI and its related entities.
1131  **/
1132 void i40e_update_stats(struct i40e_vsi *vsi)
1133 {
1134         struct i40e_pf *pf = vsi->back;
1135
1136         if (vsi == pf->vsi[pf->lan_vsi])
1137                 i40e_update_pf_stats(pf);
1138
1139         i40e_update_vsi_stats(vsi);
1140 #ifdef I40E_FCOE
1141         i40e_update_fcoe_stats(vsi);
1142 #endif
1143 }
1144
1145 /**
1146  * i40e_find_filter - Search VSI filter list for specific mac/vlan filter
1147  * @vsi: the VSI to be searched
1148  * @macaddr: the MAC address
1149  * @vlan: the vlan
1150  * @is_vf: make sure its a VF filter, else doesn't matter
1151  * @is_netdev: make sure its a netdev filter, else doesn't matter
1152  *
1153  * Returns ptr to the filter object or NULL
1154  **/
1155 static struct i40e_mac_filter *i40e_find_filter(struct i40e_vsi *vsi,
1156                                                 u8 *macaddr, s16 vlan,
1157                                                 bool is_vf, bool is_netdev)
1158 {
1159         struct i40e_mac_filter *f;
1160
1161         if (!vsi || !macaddr)
1162                 return NULL;
1163
1164         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1165                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1166                     (vlan == f->vlan)    &&
1167                     (!is_vf || f->is_vf) &&
1168                     (!is_netdev || f->is_netdev))
1169                         return f;
1170         }
1171         return NULL;
1172 }
1173
1174 /**
1175  * i40e_find_mac - Find a mac addr in the macvlan filters list
1176  * @vsi: the VSI to be searched
1177  * @macaddr: the MAC address we are searching for
1178  * @is_vf: make sure its a VF filter, else doesn't matter
1179  * @is_netdev: make sure its a netdev filter, else doesn't matter
1180  *
1181  * Returns the first filter with the provided MAC address or NULL if
1182  * MAC address was not found
1183  **/
1184 struct i40e_mac_filter *i40e_find_mac(struct i40e_vsi *vsi, u8 *macaddr,
1185                                       bool is_vf, bool is_netdev)
1186 {
1187         struct i40e_mac_filter *f;
1188
1189         if (!vsi || !macaddr)
1190                 return NULL;
1191
1192         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1193                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1194                     (!is_vf || f->is_vf) &&
1195                     (!is_netdev || f->is_netdev))
1196                         return f;
1197         }
1198         return NULL;
1199 }
1200
1201 /**
1202  * i40e_is_vsi_in_vlan - Check if VSI is in vlan mode
1203  * @vsi: the VSI to be searched
1204  *
1205  * Returns true if VSI is in vlan mode or false otherwise
1206  **/
1207 bool i40e_is_vsi_in_vlan(struct i40e_vsi *vsi)
1208 {
1209         struct i40e_mac_filter *f;
1210
1211         /* Only -1 for all the filters denotes not in vlan mode
1212          * so we have to go through all the list in order to make sure
1213          */
1214         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1215                 if (f->vlan >= 0 || vsi->info.pvid)
1216                         return true;
1217         }
1218
1219         return false;
1220 }
1221
1222 /**
1223  * i40e_put_mac_in_vlan - Make macvlan filters from macaddrs and vlans
1224  * @vsi: the VSI to be searched
1225  * @macaddr: the mac address to be filtered
1226  * @is_vf: true if it is a VF
1227  * @is_netdev: true if it is a netdev
1228  *
1229  * Goes through all the macvlan filters and adds a
1230  * macvlan filter for each unique vlan that already exists
1231  *
1232  * Returns first filter found on success, else NULL
1233  **/
1234 struct i40e_mac_filter *i40e_put_mac_in_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1235                                              bool is_vf, bool is_netdev)
1236 {
1237         struct i40e_mac_filter *f;
1238
1239         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1240                 if (vsi->info.pvid)
1241                         f->vlan = le16_to_cpu(vsi->info.pvid);
1242                 if (!i40e_find_filter(vsi, macaddr, f->vlan,
1243                                       is_vf, is_netdev)) {
1244                         if (!i40e_add_filter(vsi, macaddr, f->vlan,
1245                                              is_vf, is_netdev))
1246                                 return NULL;
1247                 }
1248         }
1249
1250         return list_first_entry_or_null(&vsi->mac_filter_list,
1251                                         struct i40e_mac_filter, list);
1252 }
1253
1254 /**
1255  * i40e_del_mac_all_vlan - Remove a MAC filter from all VLANS
1256  * @vsi: the VSI to be searched
1257  * @macaddr: the mac address to be removed
1258  * @is_vf: true if it is a VF
1259  * @is_netdev: true if it is a netdev
1260  *
1261  * Removes a given MAC address from a VSI, regardless of VLAN
1262  *
1263  * Returns 0 for success, or error
1264  **/
1265 int i40e_del_mac_all_vlan(struct i40e_vsi *vsi, u8 *macaddr,
1266                           bool is_vf, bool is_netdev)
1267 {
1268         struct i40e_mac_filter *f = NULL;
1269         int changed = 0;
1270
1271         WARN(!spin_is_locked(&vsi->mac_filter_list_lock),
1272              "Missing mac_filter_list_lock\n");
1273         list_for_each_entry(f, &vsi->mac_filter_list, list) {
1274                 if ((ether_addr_equal(macaddr, f->macaddr)) &&
1275                     (is_vf == f->is_vf) &&
1276                     (is_netdev == f->is_netdev)) {
1277                         f->counter--;
1278                         changed = 1;
1279                         if (f->counter == 0)
1280                                 f->state = I40E_FILTER_REMOVE;
1281                 }
1282         }
1283         if (changed) {
1284                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1285                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1286                 return 0;
1287         }
1288         return -ENOENT;
1289 }
1290
1291 /**
1292  * i40e_rm_default_mac_filter - Remove the default MAC filter set by NVM
1293  * @vsi: the PF Main VSI - inappropriate for any other VSI
1294  * @macaddr: the MAC address
1295  *
1296  * Remove whatever filter the firmware set up so the driver can manage
1297  * its own filtering intelligently.
1298  **/
1299 static void i40e_rm_default_mac_filter(struct i40e_vsi *vsi, u8 *macaddr)
1300 {
1301         struct i40e_aqc_remove_macvlan_element_data element;
1302         struct i40e_pf *pf = vsi->back;
1303
1304         /* Only appropriate for the PF main VSI */
1305         if (vsi->type != I40E_VSI_MAIN)
1306                 return;
1307
1308         memset(&element, 0, sizeof(element));
1309         ether_addr_copy(element.mac_addr, macaddr);
1310         element.vlan_tag = 0;
1311         /* Ignore error returns, some firmware does it this way... */
1312         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1313         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1314
1315         memset(&element, 0, sizeof(element));
1316         ether_addr_copy(element.mac_addr, macaddr);
1317         element.vlan_tag = 0;
1318         /* ...and some firmware does it this way. */
1319         element.flags = I40E_AQC_MACVLAN_DEL_PERFECT_MATCH |
1320                         I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1321         i40e_aq_remove_macvlan(&pf->hw, vsi->seid, &element, 1, NULL);
1322 }
1323
1324 /**
1325  * i40e_add_filter - Add a mac/vlan filter to the VSI
1326  * @vsi: the VSI to be searched
1327  * @macaddr: the MAC address
1328  * @vlan: the vlan
1329  * @is_vf: make sure its a VF filter, else doesn't matter
1330  * @is_netdev: make sure its a netdev filter, else doesn't matter
1331  *
1332  * Returns ptr to the filter object or NULL when no memory available.
1333  *
1334  * NOTE: This function is expected to be called with mac_filter_list_lock
1335  * being held.
1336  **/
1337 struct i40e_mac_filter *i40e_add_filter(struct i40e_vsi *vsi,
1338                                         u8 *macaddr, s16 vlan,
1339                                         bool is_vf, bool is_netdev)
1340 {
1341         struct i40e_mac_filter *f;
1342         int changed = false;
1343
1344         if (!vsi || !macaddr)
1345                 return NULL;
1346
1347         /* Do not allow broadcast filter to be added since broadcast filter
1348          * is added as part of add VSI for any newly created VSI except
1349          * FDIR VSI
1350          */
1351         if (is_broadcast_ether_addr(macaddr))
1352                 return NULL;
1353
1354         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1355         if (!f) {
1356                 f = kzalloc(sizeof(*f), GFP_ATOMIC);
1357                 if (!f)
1358                         goto add_filter_out;
1359
1360                 ether_addr_copy(f->macaddr, macaddr);
1361                 f->vlan = vlan;
1362                 /* If we're in overflow promisc mode, set the state directly
1363                  * to failed, so we don't bother to try sending the filter
1364                  * to the hardware.
1365                  */
1366                 if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))
1367                         f->state = I40E_FILTER_FAILED;
1368                 else
1369                         f->state = I40E_FILTER_NEW;
1370                 changed = true;
1371                 INIT_LIST_HEAD(&f->list);
1372                 list_add_tail(&f->list, &vsi->mac_filter_list);
1373         }
1374
1375         /* increment counter and add a new flag if needed */
1376         if (is_vf) {
1377                 if (!f->is_vf) {
1378                         f->is_vf = true;
1379                         f->counter++;
1380                 }
1381         } else if (is_netdev) {
1382                 if (!f->is_netdev) {
1383                         f->is_netdev = true;
1384                         f->counter++;
1385                 }
1386         } else {
1387                 f->counter++;
1388         }
1389
1390         if (changed) {
1391                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1392                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1393         }
1394
1395 add_filter_out:
1396         return f;
1397 }
1398
1399 /**
1400  * i40e_del_filter - Remove a mac/vlan filter from the VSI
1401  * @vsi: the VSI to be searched
1402  * @macaddr: the MAC address
1403  * @vlan: the vlan
1404  * @is_vf: make sure it's a VF filter, else doesn't matter
1405  * @is_netdev: make sure it's a netdev filter, else doesn't matter
1406  *
1407  * NOTE: This function is expected to be called with mac_filter_list_lock
1408  * being held.
1409  * ANOTHER NOTE: This function MUST be called from within the context of
1410  * the "safe" variants of any list iterators, e.g. list_for_each_entry_safe()
1411  * instead of list_for_each_entry().
1412  **/
1413 void i40e_del_filter(struct i40e_vsi *vsi,
1414                      u8 *macaddr, s16 vlan,
1415                      bool is_vf, bool is_netdev)
1416 {
1417         struct i40e_mac_filter *f;
1418
1419         if (!vsi || !macaddr)
1420                 return;
1421
1422         f = i40e_find_filter(vsi, macaddr, vlan, is_vf, is_netdev);
1423         if (!f || f->counter == 0)
1424                 return;
1425
1426         if (is_vf) {
1427                 if (f->is_vf) {
1428                         f->is_vf = false;
1429                         f->counter--;
1430                 }
1431         } else if (is_netdev) {
1432                 if (f->is_netdev) {
1433                         f->is_netdev = false;
1434                         f->counter--;
1435                 }
1436         } else {
1437                 /* make sure we don't remove a filter in use by VF or netdev */
1438                 int min_f = 0;
1439
1440                 min_f += (f->is_vf ? 1 : 0);
1441                 min_f += (f->is_netdev ? 1 : 0);
1442
1443                 if (f->counter > min_f)
1444                         f->counter--;
1445         }
1446
1447         /* counter == 0 tells sync_filters_subtask to
1448          * remove the filter from the firmware's list
1449          */
1450         if (f->counter == 0) {
1451                 if ((f->state == I40E_FILTER_FAILED) ||
1452                     (f->state == I40E_FILTER_NEW)) {
1453                         /* this one never got added by the FW. Just remove it,
1454                          * no need to sync anything.
1455                          */
1456                         list_del(&f->list);
1457                         kfree(f);
1458                 } else {
1459                         f->state = I40E_FILTER_REMOVE;
1460                         vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1461                         vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1462                 }
1463         }
1464 }
1465
1466 /**
1467  * i40e_set_mac - NDO callback to set mac address
1468  * @netdev: network interface device structure
1469  * @p: pointer to an address structure
1470  *
1471  * Returns 0 on success, negative on failure
1472  **/
1473 #ifdef I40E_FCOE
1474 int i40e_set_mac(struct net_device *netdev, void *p)
1475 #else
1476 static int i40e_set_mac(struct net_device *netdev, void *p)
1477 #endif
1478 {
1479         struct i40e_netdev_priv *np = netdev_priv(netdev);
1480         struct i40e_vsi *vsi = np->vsi;
1481         struct i40e_pf *pf = vsi->back;
1482         struct i40e_hw *hw = &pf->hw;
1483         struct sockaddr *addr = p;
1484
1485         if (!is_valid_ether_addr(addr->sa_data))
1486                 return -EADDRNOTAVAIL;
1487
1488         if (ether_addr_equal(netdev->dev_addr, addr->sa_data)) {
1489                 netdev_info(netdev, "already using mac address %pM\n",
1490                             addr->sa_data);
1491                 return 0;
1492         }
1493
1494         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
1495             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
1496                 return -EADDRNOTAVAIL;
1497
1498         if (ether_addr_equal(hw->mac.addr, addr->sa_data))
1499                 netdev_info(netdev, "returning to hw mac address %pM\n",
1500                             hw->mac.addr);
1501         else
1502                 netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
1503
1504         spin_lock_bh(&vsi->mac_filter_list_lock);
1505         i40e_del_mac_all_vlan(vsi, netdev->dev_addr, false, true);
1506         i40e_put_mac_in_vlan(vsi, addr->sa_data, false, true);
1507         spin_unlock_bh(&vsi->mac_filter_list_lock);
1508         ether_addr_copy(netdev->dev_addr, addr->sa_data);
1509         if (vsi->type == I40E_VSI_MAIN) {
1510                 i40e_status ret;
1511
1512                 ret = i40e_aq_mac_address_write(&vsi->back->hw,
1513                                                 I40E_AQC_WRITE_TYPE_LAA_WOL,
1514                                                 addr->sa_data, NULL);
1515                 if (ret)
1516                         netdev_info(netdev, "Ignoring error from firmware on LAA update, status %s, AQ ret %s\n",
1517                                     i40e_stat_str(hw, ret),
1518                                     i40e_aq_str(hw, hw->aq.asq_last_status));
1519         }
1520
1521         /* schedule our worker thread which will take care of
1522          * applying the new filter changes
1523          */
1524         i40e_service_event_schedule(vsi->back);
1525         return 0;
1526 }
1527
1528 /**
1529  * i40e_vsi_setup_queue_map - Setup a VSI queue map based on enabled_tc
1530  * @vsi: the VSI being setup
1531  * @ctxt: VSI context structure
1532  * @enabled_tc: Enabled TCs bitmap
1533  * @is_add: True if called before Add VSI
1534  *
1535  * Setup VSI queue mapping for enabled traffic classes.
1536  **/
1537 #ifdef I40E_FCOE
1538 void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1539                               struct i40e_vsi_context *ctxt,
1540                               u8 enabled_tc,
1541                               bool is_add)
1542 #else
1543 static void i40e_vsi_setup_queue_map(struct i40e_vsi *vsi,
1544                                      struct i40e_vsi_context *ctxt,
1545                                      u8 enabled_tc,
1546                                      bool is_add)
1547 #endif
1548 {
1549         struct i40e_pf *pf = vsi->back;
1550         u16 sections = 0;
1551         u8 netdev_tc = 0;
1552         u16 numtc = 0;
1553         u16 qcount;
1554         u8 offset;
1555         u16 qmap;
1556         int i;
1557         u16 num_tc_qps = 0;
1558
1559         sections = I40E_AQ_VSI_PROP_QUEUE_MAP_VALID;
1560         offset = 0;
1561
1562         if (enabled_tc && (vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
1563                 /* Find numtc from enabled TC bitmap */
1564                 for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1565                         if (enabled_tc & BIT(i)) /* TC is enabled */
1566                                 numtc++;
1567                 }
1568                 if (!numtc) {
1569                         dev_warn(&pf->pdev->dev, "DCB is enabled but no TC enabled, forcing TC0\n");
1570                         numtc = 1;
1571                 }
1572         } else {
1573                 /* At least TC0 is enabled in case of non-DCB case */
1574                 numtc = 1;
1575         }
1576
1577         vsi->tc_config.numtc = numtc;
1578         vsi->tc_config.enabled_tc = enabled_tc ? enabled_tc : 1;
1579         /* Number of queues per enabled TC */
1580         qcount = vsi->alloc_queue_pairs;
1581
1582         num_tc_qps = qcount / numtc;
1583         num_tc_qps = min_t(int, num_tc_qps, i40e_pf_get_max_q_per_tc(pf));
1584
1585         /* Setup queue offset/count for all TCs for given VSI */
1586         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
1587                 /* See if the given TC is enabled for the given VSI */
1588                 if (vsi->tc_config.enabled_tc & BIT(i)) {
1589                         /* TC is enabled */
1590                         int pow, num_qps;
1591
1592                         switch (vsi->type) {
1593                         case I40E_VSI_MAIN:
1594                                 qcount = min_t(int, pf->alloc_rss_size,
1595                                                num_tc_qps);
1596                                 break;
1597 #ifdef I40E_FCOE
1598                         case I40E_VSI_FCOE:
1599                                 qcount = num_tc_qps;
1600                                 break;
1601 #endif
1602                         case I40E_VSI_FDIR:
1603                         case I40E_VSI_SRIOV:
1604                         case I40E_VSI_VMDQ2:
1605                         default:
1606                                 qcount = num_tc_qps;
1607                                 WARN_ON(i != 0);
1608                                 break;
1609                         }
1610                         vsi->tc_config.tc_info[i].qoffset = offset;
1611                         vsi->tc_config.tc_info[i].qcount = qcount;
1612
1613                         /* find the next higher power-of-2 of num queue pairs */
1614                         num_qps = qcount;
1615                         pow = 0;
1616                         while (num_qps && (BIT_ULL(pow) < qcount)) {
1617                                 pow++;
1618                                 num_qps >>= 1;
1619                         }
1620
1621                         vsi->tc_config.tc_info[i].netdev_tc = netdev_tc++;
1622                         qmap =
1623                             (offset << I40E_AQ_VSI_TC_QUE_OFFSET_SHIFT) |
1624                             (pow << I40E_AQ_VSI_TC_QUE_NUMBER_SHIFT);
1625
1626                         offset += qcount;
1627                 } else {
1628                         /* TC is not enabled so set the offset to
1629                          * default queue and allocate one queue
1630                          * for the given TC.
1631                          */
1632                         vsi->tc_config.tc_info[i].qoffset = 0;
1633                         vsi->tc_config.tc_info[i].qcount = 1;
1634                         vsi->tc_config.tc_info[i].netdev_tc = 0;
1635
1636                         qmap = 0;
1637                 }
1638                 ctxt->info.tc_mapping[i] = cpu_to_le16(qmap);
1639         }
1640
1641         /* Set actual Tx/Rx queue pairs */
1642         vsi->num_queue_pairs = offset;
1643         if ((vsi->type == I40E_VSI_MAIN) && (numtc == 1)) {
1644                 if (vsi->req_queue_pairs > 0)
1645                         vsi->num_queue_pairs = vsi->req_queue_pairs;
1646                 else if (pf->flags & I40E_FLAG_MSIX_ENABLED)
1647                         vsi->num_queue_pairs = pf->num_lan_msix;
1648         }
1649
1650         /* Scheduler section valid can only be set for ADD VSI */
1651         if (is_add) {
1652                 sections |= I40E_AQ_VSI_PROP_SCHED_VALID;
1653
1654                 ctxt->info.up_enable_bits = enabled_tc;
1655         }
1656         if (vsi->type == I40E_VSI_SRIOV) {
1657                 ctxt->info.mapping_flags |=
1658                                      cpu_to_le16(I40E_AQ_VSI_QUE_MAP_NONCONTIG);
1659                 for (i = 0; i < vsi->num_queue_pairs; i++)
1660                         ctxt->info.queue_mapping[i] =
1661                                                cpu_to_le16(vsi->base_queue + i);
1662         } else {
1663                 ctxt->info.mapping_flags |=
1664                                         cpu_to_le16(I40E_AQ_VSI_QUE_MAP_CONTIG);
1665                 ctxt->info.queue_mapping[0] = cpu_to_le16(vsi->base_queue);
1666         }
1667         ctxt->info.valid_sections |= cpu_to_le16(sections);
1668 }
1669
1670 /**
1671  * i40e_set_rx_mode - NDO callback to set the netdev filters
1672  * @netdev: network interface device structure
1673  **/
1674 #ifdef I40E_FCOE
1675 void i40e_set_rx_mode(struct net_device *netdev)
1676 #else
1677 static void i40e_set_rx_mode(struct net_device *netdev)
1678 #endif
1679 {
1680         struct i40e_netdev_priv *np = netdev_priv(netdev);
1681         struct i40e_mac_filter *f, *ftmp;
1682         struct i40e_vsi *vsi = np->vsi;
1683         struct netdev_hw_addr *uca;
1684         struct netdev_hw_addr *mca;
1685         struct netdev_hw_addr *ha;
1686
1687         spin_lock_bh(&vsi->mac_filter_list_lock);
1688
1689         /* add addr if not already in the filter list */
1690         netdev_for_each_uc_addr(uca, netdev) {
1691                 if (!i40e_find_mac(vsi, uca->addr, false, true)) {
1692                         if (i40e_is_vsi_in_vlan(vsi))
1693                                 i40e_put_mac_in_vlan(vsi, uca->addr,
1694                                                      false, true);
1695                         else
1696                                 i40e_add_filter(vsi, uca->addr, I40E_VLAN_ANY,
1697                                                 false, true);
1698                 }
1699         }
1700
1701         netdev_for_each_mc_addr(mca, netdev) {
1702                 if (!i40e_find_mac(vsi, mca->addr, false, true)) {
1703                         if (i40e_is_vsi_in_vlan(vsi))
1704                                 i40e_put_mac_in_vlan(vsi, mca->addr,
1705                                                      false, true);
1706                         else
1707                                 i40e_add_filter(vsi, mca->addr, I40E_VLAN_ANY,
1708                                                 false, true);
1709                 }
1710         }
1711
1712         /* remove filter if not in netdev list */
1713         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1714
1715                 if (!f->is_netdev)
1716                         continue;
1717
1718                 netdev_for_each_mc_addr(mca, netdev)
1719                         if (ether_addr_equal(mca->addr, f->macaddr))
1720                                 goto bottom_of_search_loop;
1721
1722                 netdev_for_each_uc_addr(uca, netdev)
1723                         if (ether_addr_equal(uca->addr, f->macaddr))
1724                                 goto bottom_of_search_loop;
1725
1726                 for_each_dev_addr(netdev, ha)
1727                         if (ether_addr_equal(ha->addr, f->macaddr))
1728                                 goto bottom_of_search_loop;
1729
1730                 /* f->macaddr wasn't found in uc, mc, or ha list so delete it */
1731                 i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY, false, true);
1732
1733 bottom_of_search_loop:
1734                 continue;
1735         }
1736         spin_unlock_bh(&vsi->mac_filter_list_lock);
1737
1738         /* check for other flag changes */
1739         if (vsi->current_netdev_flags != vsi->netdev->flags) {
1740                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
1741                 vsi->back->flags |= I40E_FLAG_FILTER_SYNC;
1742         }
1743
1744         /* schedule our worker thread which will take care of
1745          * applying the new filter changes
1746          */
1747         i40e_service_event_schedule(vsi->back);
1748 }
1749
1750 /**
1751  * i40e_undo_del_filter_entries - Undo the changes made to MAC filter entries
1752  * @vsi: pointer to vsi struct
1753  * @from: Pointer to list which contains MAC filter entries - changes to
1754  *        those entries needs to be undone.
1755  *
1756  * MAC filter entries from list were slated to be removed from device.
1757  **/
1758 static void i40e_undo_del_filter_entries(struct i40e_vsi *vsi,
1759                                          struct list_head *from)
1760 {
1761         struct i40e_mac_filter *f, *ftmp;
1762
1763         list_for_each_entry_safe(f, ftmp, from, list) {
1764                 /* Move the element back into MAC filter list*/
1765                 list_move_tail(&f->list, &vsi->mac_filter_list);
1766         }
1767 }
1768
1769 /**
1770  * i40e_update_filter_state - Update filter state based on return data
1771  * from firmware
1772  * @count: Number of filters added
1773  * @add_list: return data from fw
1774  * @head: pointer to first filter in current batch
1775  * @aq_err: status from fw
1776  *
1777  * MAC filter entries from list were slated to be added to device. Returns
1778  * number of successful filters. Note that 0 does NOT mean success!
1779  **/
1780 static int
1781 i40e_update_filter_state(int count,
1782                          struct i40e_aqc_add_macvlan_element_data *add_list,
1783                          struct i40e_mac_filter *add_head, int aq_err)
1784 {
1785         int retval = 0;
1786         int i;
1787
1788
1789         if (!aq_err) {
1790                 retval = count;
1791                 /* Everything's good, mark all filters active. */
1792                 for (i = 0; i < count ; i++) {
1793                         add_head->state = I40E_FILTER_ACTIVE;
1794                         add_head = list_next_entry(add_head, list);
1795                 }
1796         } else if (aq_err == I40E_AQ_RC_ENOSPC) {
1797                 /* Device ran out of filter space. Check the return value
1798                  * for each filter to see which ones are active.
1799                  */
1800                 for (i = 0; i < count ; i++) {
1801                         if (add_list[i].match_method ==
1802                             I40E_AQC_MM_ERR_NO_RES) {
1803                                 add_head->state = I40E_FILTER_FAILED;
1804                         } else {
1805                                 add_head->state = I40E_FILTER_ACTIVE;
1806                                 retval++;
1807                         }
1808                         add_head = list_next_entry(add_head, list);
1809                 }
1810         } else {
1811                 /* Some other horrible thing happened, fail all filters */
1812                 retval = 0;
1813                 for (i = 0; i < count ; i++) {
1814                         add_head->state = I40E_FILTER_FAILED;
1815                         add_head = list_next_entry(add_head, list);
1816                 }
1817         }
1818         return retval;
1819 }
1820
1821 /**
1822  * i40e_sync_vsi_filters - Update the VSI filter list to the HW
1823  * @vsi: ptr to the VSI
1824  *
1825  * Push any outstanding VSI filter changes through the AdminQ.
1826  *
1827  * Returns 0 or error value
1828  **/
1829 int i40e_sync_vsi_filters(struct i40e_vsi *vsi)
1830 {
1831         struct i40e_mac_filter *f, *ftmp, *add_head = NULL;
1832         struct list_head tmp_add_list, tmp_del_list;
1833         struct i40e_hw *hw = &vsi->back->hw;
1834         bool promisc_changed = false;
1835         char vsi_name[16] = "PF";
1836         int filter_list_len = 0;
1837         u32 changed_flags = 0;
1838         i40e_status aq_ret = 0;
1839         int retval = 0;
1840         struct i40e_pf *pf;
1841         int num_add = 0;
1842         int num_del = 0;
1843         int aq_err = 0;
1844         u16 cmd_flags;
1845         int list_size;
1846         int fcnt;
1847
1848         /* empty array typed pointers, kcalloc later */
1849         struct i40e_aqc_add_macvlan_element_data *add_list;
1850         struct i40e_aqc_remove_macvlan_element_data *del_list;
1851
1852         while (test_and_set_bit(__I40E_CONFIG_BUSY, &vsi->state))
1853                 usleep_range(1000, 2000);
1854         pf = vsi->back;
1855
1856         if (vsi->netdev) {
1857                 changed_flags = vsi->current_netdev_flags ^ vsi->netdev->flags;
1858                 vsi->current_netdev_flags = vsi->netdev->flags;
1859         }
1860
1861         INIT_LIST_HEAD(&tmp_add_list);
1862         INIT_LIST_HEAD(&tmp_del_list);
1863
1864         if (vsi->type == I40E_VSI_SRIOV)
1865                 snprintf(vsi_name, sizeof(vsi_name) - 1, "VF %d", vsi->vf_id);
1866         else if (vsi->type != I40E_VSI_MAIN)
1867                 snprintf(vsi_name, sizeof(vsi_name) - 1, "vsi %d", vsi->seid);
1868
1869         if (vsi->flags & I40E_VSI_FLAG_FILTER_CHANGED) {
1870                 vsi->flags &= ~I40E_VSI_FLAG_FILTER_CHANGED;
1871
1872                 spin_lock_bh(&vsi->mac_filter_list_lock);
1873                 /* Create a list of filters to delete. */
1874                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
1875                         if (f->state == I40E_FILTER_REMOVE) {
1876                                 WARN_ON(f->counter != 0);
1877                                 /* Move the element into temporary del_list */
1878                                 list_move_tail(&f->list, &tmp_del_list);
1879                                 vsi->active_filters--;
1880                         }
1881                         if (f->state == I40E_FILTER_NEW) {
1882                                 WARN_ON(f->counter == 0);
1883                                 /* Move the element into temporary add_list */
1884                                 list_move_tail(&f->list, &tmp_add_list);
1885                         }
1886                 }
1887                 spin_unlock_bh(&vsi->mac_filter_list_lock);
1888         }
1889
1890         /* Now process 'del_list' outside the lock */
1891         if (!list_empty(&tmp_del_list)) {
1892                 filter_list_len = hw->aq.asq_buf_size /
1893                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1894                 list_size = filter_list_len *
1895                             sizeof(struct i40e_aqc_remove_macvlan_element_data);
1896                 del_list = kzalloc(list_size, GFP_ATOMIC);
1897                 if (!del_list) {
1898                         /* Undo VSI's MAC filter entry element updates */
1899                         spin_lock_bh(&vsi->mac_filter_list_lock);
1900                         i40e_undo_del_filter_entries(vsi, &tmp_del_list);
1901                         spin_unlock_bh(&vsi->mac_filter_list_lock);
1902                         retval = -ENOMEM;
1903                         goto out;
1904                 }
1905
1906                 list_for_each_entry_safe(f, ftmp, &tmp_del_list, list) {
1907                         cmd_flags = 0;
1908
1909                         /* add to delete list */
1910                         ether_addr_copy(del_list[num_del].mac_addr, f->macaddr);
1911                         if (f->vlan == I40E_VLAN_ANY) {
1912                                 del_list[num_del].vlan_tag = 0;
1913                                 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
1914                         } else {
1915                                 del_list[num_del].vlan_tag =
1916                                         cpu_to_le16((u16)(f->vlan));
1917                         }
1918
1919                         cmd_flags |= I40E_AQC_MACVLAN_DEL_PERFECT_MATCH;
1920                         del_list[num_del].flags = cmd_flags;
1921                         num_del++;
1922
1923                         /* flush a full buffer */
1924                         if (num_del == filter_list_len) {
1925                                 aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid,
1926                                                                 del_list,
1927                                                                 num_del, NULL);
1928                                 aq_err = hw->aq.asq_last_status;
1929                                 num_del = 0;
1930                                 memset(del_list, 0, list_size);
1931
1932                                 /* Explicitly ignore and do not report when
1933                                  * firmware returns ENOENT.
1934                                  */
1935                                 if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1936                                         retval = -EIO;
1937                                         dev_info(&pf->pdev->dev,
1938                                                  "ignoring delete macvlan error on %s, err %s, aq_err %s\n",
1939                                                  vsi_name,
1940                                                  i40e_stat_str(hw, aq_ret),
1941                                                  i40e_aq_str(hw, aq_err));
1942                                 }
1943                         }
1944                         /* Release memory for MAC filter entries which were
1945                          * synced up with HW.
1946                          */
1947                         list_del(&f->list);
1948                         kfree(f);
1949                 }
1950
1951                 if (num_del) {
1952                         aq_ret = i40e_aq_remove_macvlan(hw, vsi->seid, del_list,
1953                                                         num_del, NULL);
1954                         aq_err = hw->aq.asq_last_status;
1955                         num_del = 0;
1956
1957                         /* Explicitly ignore and do not report when firmware
1958                          * returns ENOENT.
1959                          */
1960                         if (aq_ret && !(aq_err == I40E_AQ_RC_ENOENT)) {
1961                                 retval = -EIO;
1962                                 dev_info(&pf->pdev->dev,
1963                                          "ignoring delete macvlan error on %s, err %s aq_err %s\n",
1964                                          vsi_name,
1965                                          i40e_stat_str(hw, aq_ret),
1966                                          i40e_aq_str(hw, aq_err));
1967                         }
1968                 }
1969
1970                 kfree(del_list);
1971                 del_list = NULL;
1972         }
1973
1974         if (!list_empty(&tmp_add_list)) {
1975                 /* Do all the adds now. */
1976                 filter_list_len = hw->aq.asq_buf_size /
1977                                sizeof(struct i40e_aqc_add_macvlan_element_data);
1978                 list_size = filter_list_len *
1979                                sizeof(struct i40e_aqc_add_macvlan_element_data);
1980                 add_list = kzalloc(list_size, GFP_ATOMIC);
1981                 if (!add_list) {
1982                         retval = -ENOMEM;
1983                         goto out;
1984                 }
1985                 num_add = 0;
1986                 list_for_each_entry(f, &tmp_add_list, list) {
1987                         if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
1988                                      &vsi->state)) {
1989                                 f->state = I40E_FILTER_FAILED;
1990                                 continue;
1991                         }
1992                         /* add to add array */
1993                         if (num_add == 0)
1994                                 add_head = f;
1995                         cmd_flags = 0;
1996                         ether_addr_copy(add_list[num_add].mac_addr, f->macaddr);
1997                         if (f->vlan == I40E_VLAN_ANY) {
1998                                 add_list[num_add].vlan_tag = 0;
1999                                 cmd_flags |= I40E_AQC_MACVLAN_ADD_IGNORE_VLAN;
2000                         } else {
2001                                 add_list[num_add].vlan_tag =
2002                                         cpu_to_le16((u16)(f->vlan));
2003                         }
2004                         add_list[num_add].queue_number = 0;
2005                         cmd_flags |= I40E_AQC_MACVLAN_ADD_PERFECT_MATCH;
2006                         add_list[num_add].flags = cpu_to_le16(cmd_flags);
2007                         num_add++;
2008
2009                         /* flush a full buffer */
2010                         if (num_add == filter_list_len) {
2011                                 aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2012                                                              add_list, num_add,
2013                                                              NULL);
2014                                 aq_err = hw->aq.asq_last_status;
2015                                 fcnt = i40e_update_filter_state(num_add,
2016                                                                 add_list,
2017                                                                 add_head,
2018                                                                 aq_ret);
2019                                 vsi->active_filters += fcnt;
2020
2021                                 if (fcnt != num_add) {
2022                                         promisc_changed = true;
2023                                         set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2024                                                 &vsi->state);
2025                                         vsi->promisc_threshold =
2026                                                 (vsi->active_filters * 3) / 4;
2027                                         dev_warn(&pf->pdev->dev,
2028                                                  "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2029                                                  i40e_aq_str(hw, aq_err),
2030                                                  vsi_name);
2031                                 }
2032                                 memset(add_list, 0, list_size);
2033                                 num_add = 0;
2034                         }
2035                 }
2036                 if (num_add) {
2037                         aq_ret = i40e_aq_add_macvlan(hw, vsi->seid,
2038                                                      add_list, num_add, NULL);
2039                         aq_err = hw->aq.asq_last_status;
2040                         fcnt = i40e_update_filter_state(num_add, add_list,
2041                                                         add_head, aq_ret);
2042                         vsi->active_filters += fcnt;
2043                         if (fcnt != num_add) {
2044                                 promisc_changed = true;
2045                                 set_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2046                                         &vsi->state);
2047                                 vsi->promisc_threshold =
2048                                                 (vsi->active_filters * 3) / 4;
2049                                 dev_warn(&pf->pdev->dev,
2050                                          "Error %s adding RX filters on %s, promiscuous mode forced on\n",
2051                                          i40e_aq_str(hw, aq_err), vsi_name);
2052                         }
2053                 }
2054                 /* Now move all of the filters from the temp add list back to
2055                  * the VSI's list.
2056                  */
2057                 spin_lock_bh(&vsi->mac_filter_list_lock);
2058                 list_for_each_entry_safe(f, ftmp, &tmp_add_list, list) {
2059                         list_move_tail(&f->list, &vsi->mac_filter_list);
2060                 }
2061                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2062                 kfree(add_list);
2063                 add_list = NULL;
2064         }
2065
2066         /* Check to see if we can drop out of overflow promiscuous mode. */
2067         if (test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state) &&
2068             (vsi->active_filters < vsi->promisc_threshold)) {
2069                 int failed_count = 0;
2070                 /* See if we have any failed filters. We can't drop out of
2071                  * promiscuous until these have all been deleted.
2072                  */
2073                 spin_lock_bh(&vsi->mac_filter_list_lock);
2074                 list_for_each_entry(f, &vsi->mac_filter_list, list) {
2075                         if (f->state == I40E_FILTER_FAILED)
2076                                 failed_count++;
2077                 }
2078                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2079                 if (!failed_count) {
2080                         dev_info(&pf->pdev->dev,
2081                                  "filter logjam cleared on %s, leaving overflow promiscuous mode\n",
2082                                  vsi_name);
2083                         clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2084                         promisc_changed = true;
2085                         vsi->promisc_threshold = 0;
2086                 }
2087         }
2088
2089         /* if the VF is not trusted do not do promisc */
2090         if ((vsi->type == I40E_VSI_SRIOV) && !pf->vf[vsi->vf_id].trusted) {
2091                 clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
2092                 goto out;
2093         }
2094
2095         /* check for changes in promiscuous modes */
2096         if (changed_flags & IFF_ALLMULTI) {
2097                 bool cur_multipromisc;
2098
2099                 cur_multipromisc = !!(vsi->current_netdev_flags & IFF_ALLMULTI);
2100                 aq_ret = i40e_aq_set_vsi_multicast_promiscuous(&vsi->back->hw,
2101                                                                vsi->seid,
2102                                                                cur_multipromisc,
2103                                                                NULL);
2104                 if (aq_ret) {
2105                         retval = i40e_aq_rc_to_posix(aq_ret,
2106                                                      hw->aq.asq_last_status);
2107                         dev_info(&pf->pdev->dev,
2108                                  "set multi promisc failed on %s, err %s aq_err %s\n",
2109                                  vsi_name,
2110                                  i40e_stat_str(hw, aq_ret),
2111                                  i40e_aq_str(hw, hw->aq.asq_last_status));
2112                 }
2113         }
2114         if ((changed_flags & IFF_PROMISC) ||
2115             (promisc_changed &&
2116              test_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state))) {
2117                 bool cur_promisc;
2118
2119                 cur_promisc = (!!(vsi->current_netdev_flags & IFF_PROMISC) ||
2120                                test_bit(__I40E_FILTER_OVERFLOW_PROMISC,
2121                                         &vsi->state));
2122                 if ((vsi->type == I40E_VSI_MAIN) &&
2123                     (pf->lan_veb != I40E_NO_VEB) &&
2124                     !(pf->flags & I40E_FLAG_MFP_ENABLED)) {
2125                         /* set defport ON for Main VSI instead of true promisc
2126                          * this way we will get all unicast/multicast and VLAN
2127                          * promisc behavior but will not get VF or VMDq traffic
2128                          * replicated on the Main VSI.
2129                          */
2130                         if (pf->cur_promisc != cur_promisc) {
2131                                 pf->cur_promisc = cur_promisc;
2132                                 if (cur_promisc)
2133                                         aq_ret =
2134                                               i40e_aq_set_default_vsi(hw,
2135                                                                       vsi->seid,
2136                                                                       NULL);
2137                                 else
2138                                         aq_ret =
2139                                             i40e_aq_clear_default_vsi(hw,
2140                                                                       vsi->seid,
2141                                                                       NULL);
2142                                 if (aq_ret) {
2143                                         retval = i40e_aq_rc_to_posix(aq_ret,
2144                                                         hw->aq.asq_last_status);
2145                                         dev_info(&pf->pdev->dev,
2146                                                  "Set default VSI failed on %s, err %s, aq_err %s\n",
2147                                                  vsi_name,
2148                                                  i40e_stat_str(hw, aq_ret),
2149                                                  i40e_aq_str(hw,
2150                                                      hw->aq.asq_last_status));
2151                                 }
2152                         }
2153                 } else {
2154                         aq_ret = i40e_aq_set_vsi_unicast_promiscuous(
2155                                                           hw,
2156                                                           vsi->seid,
2157                                                           cur_promisc, NULL,
2158                                                           true);
2159                         if (aq_ret) {
2160                                 retval =
2161                                 i40e_aq_rc_to_posix(aq_ret,
2162                                                     hw->aq.asq_last_status);
2163                                 dev_info(&pf->pdev->dev,
2164                                          "set unicast promisc failed on %s, err %s, aq_err %s\n",
2165                                          vsi_name,
2166                                          i40e_stat_str(hw, aq_ret),
2167                                          i40e_aq_str(hw,
2168                                                      hw->aq.asq_last_status));
2169                         }
2170                         aq_ret = i40e_aq_set_vsi_multicast_promiscuous(
2171                                                           hw,
2172                                                           vsi->seid,
2173                                                           cur_promisc, NULL);
2174                         if (aq_ret) {
2175                                 retval =
2176                                 i40e_aq_rc_to_posix(aq_ret,
2177                                                     hw->aq.asq_last_status);
2178                                 dev_info(&pf->pdev->dev,
2179                                          "set multicast promisc failed on %s, err %s, aq_err %s\n",
2180                                          vsi_name,
2181                                          i40e_stat_str(hw, aq_ret),
2182                                          i40e_aq_str(hw,
2183                                                      hw->aq.asq_last_status));
2184                         }
2185                 }
2186                 aq_ret = i40e_aq_set_vsi_broadcast(&vsi->back->hw,
2187                                                    vsi->seid,
2188                                                    cur_promisc, NULL);
2189                 if (aq_ret) {
2190                         retval = i40e_aq_rc_to_posix(aq_ret,
2191                                                      pf->hw.aq.asq_last_status);
2192                         dev_info(&pf->pdev->dev,
2193                                  "set brdcast promisc failed, err %s, aq_err %s\n",
2194                                          i40e_stat_str(hw, aq_ret),
2195                                          i40e_aq_str(hw,
2196                                                      hw->aq.asq_last_status));
2197                 }
2198         }
2199 out:
2200         /* if something went wrong then set the changed flag so we try again */
2201         if (retval)
2202                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
2203
2204         clear_bit(__I40E_CONFIG_BUSY, &vsi->state);
2205         return retval;
2206 }
2207
2208 /**
2209  * i40e_sync_filters_subtask - Sync the VSI filter list with HW
2210  * @pf: board private structure
2211  **/
2212 static void i40e_sync_filters_subtask(struct i40e_pf *pf)
2213 {
2214         int v;
2215
2216         if (!pf || !(pf->flags & I40E_FLAG_FILTER_SYNC))
2217                 return;
2218         pf->flags &= ~I40E_FLAG_FILTER_SYNC;
2219
2220         for (v = 0; v < pf->num_alloc_vsi; v++) {
2221                 if (pf->vsi[v] &&
2222                     (pf->vsi[v]->flags & I40E_VSI_FLAG_FILTER_CHANGED)) {
2223                         int ret = i40e_sync_vsi_filters(pf->vsi[v]);
2224
2225                         if (ret) {
2226                                 /* come back and try again later */
2227                                 pf->flags |= I40E_FLAG_FILTER_SYNC;
2228                                 break;
2229                         }
2230                 }
2231         }
2232 }
2233
2234 /**
2235  * i40e_change_mtu - NDO callback to change the Maximum Transfer Unit
2236  * @netdev: network interface device structure
2237  * @new_mtu: new value for maximum frame size
2238  *
2239  * Returns 0 on success, negative on failure
2240  **/
2241 static int i40e_change_mtu(struct net_device *netdev, int new_mtu)
2242 {
2243         struct i40e_netdev_priv *np = netdev_priv(netdev);
2244         int max_frame = new_mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN;
2245         struct i40e_vsi *vsi = np->vsi;
2246
2247         /* MTU < 68 is an error and causes problems on some kernels */
2248         if ((new_mtu < 68) || (max_frame > I40E_MAX_RXBUFFER))
2249                 return -EINVAL;
2250
2251         netdev_info(netdev, "changing MTU from %d to %d\n",
2252                     netdev->mtu, new_mtu);
2253         netdev->mtu = new_mtu;
2254         if (netif_running(netdev))
2255                 i40e_vsi_reinit_locked(vsi);
2256         i40e_notify_client_of_l2_param_changes(vsi);
2257         return 0;
2258 }
2259
2260 /**
2261  * i40e_ioctl - Access the hwtstamp interface
2262  * @netdev: network interface device structure
2263  * @ifr: interface request data
2264  * @cmd: ioctl command
2265  **/
2266 int i40e_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
2267 {
2268         struct i40e_netdev_priv *np = netdev_priv(netdev);
2269         struct i40e_pf *pf = np->vsi->back;
2270
2271         switch (cmd) {
2272         case SIOCGHWTSTAMP:
2273                 return i40e_ptp_get_ts_config(pf, ifr);
2274         case SIOCSHWTSTAMP:
2275                 return i40e_ptp_set_ts_config(pf, ifr);
2276         default:
2277                 return -EOPNOTSUPP;
2278         }
2279 }
2280
2281 /**
2282  * i40e_vlan_stripping_enable - Turn on vlan stripping for the VSI
2283  * @vsi: the vsi being adjusted
2284  **/
2285 void i40e_vlan_stripping_enable(struct i40e_vsi *vsi)
2286 {
2287         struct i40e_vsi_context ctxt;
2288         i40e_status ret;
2289
2290         if ((vsi->info.valid_sections &
2291              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2292             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_MODE_MASK) == 0))
2293                 return;  /* already enabled */
2294
2295         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2296         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2297                                     I40E_AQ_VSI_PVLAN_EMOD_STR_BOTH;
2298
2299         ctxt.seid = vsi->seid;
2300         ctxt.info = vsi->info;
2301         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2302         if (ret) {
2303                 dev_info(&vsi->back->pdev->dev,
2304                          "update vlan stripping failed, err %s aq_err %s\n",
2305                          i40e_stat_str(&vsi->back->hw, ret),
2306                          i40e_aq_str(&vsi->back->hw,
2307                                      vsi->back->hw.aq.asq_last_status));
2308         }
2309 }
2310
2311 /**
2312  * i40e_vlan_stripping_disable - Turn off vlan stripping for the VSI
2313  * @vsi: the vsi being adjusted
2314  **/
2315 void i40e_vlan_stripping_disable(struct i40e_vsi *vsi)
2316 {
2317         struct i40e_vsi_context ctxt;
2318         i40e_status ret;
2319
2320         if ((vsi->info.valid_sections &
2321              cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID)) &&
2322             ((vsi->info.port_vlan_flags & I40E_AQ_VSI_PVLAN_EMOD_MASK) ==
2323              I40E_AQ_VSI_PVLAN_EMOD_MASK))
2324                 return;  /* already disabled */
2325
2326         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2327         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_ALL |
2328                                     I40E_AQ_VSI_PVLAN_EMOD_NOTHING;
2329
2330         ctxt.seid = vsi->seid;
2331         ctxt.info = vsi->info;
2332         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2333         if (ret) {
2334                 dev_info(&vsi->back->pdev->dev,
2335                          "update vlan stripping failed, err %s aq_err %s\n",
2336                          i40e_stat_str(&vsi->back->hw, ret),
2337                          i40e_aq_str(&vsi->back->hw,
2338                                      vsi->back->hw.aq.asq_last_status));
2339         }
2340 }
2341
2342 /**
2343  * i40e_vlan_rx_register - Setup or shutdown vlan offload
2344  * @netdev: network interface to be adjusted
2345  * @features: netdev features to test if VLAN offload is enabled or not
2346  **/
2347 static void i40e_vlan_rx_register(struct net_device *netdev, u32 features)
2348 {
2349         struct i40e_netdev_priv *np = netdev_priv(netdev);
2350         struct i40e_vsi *vsi = np->vsi;
2351
2352         if (features & NETIF_F_HW_VLAN_CTAG_RX)
2353                 i40e_vlan_stripping_enable(vsi);
2354         else
2355                 i40e_vlan_stripping_disable(vsi);
2356 }
2357
2358 /**
2359  * i40e_vsi_add_vlan - Add vsi membership for given vlan
2360  * @vsi: the vsi being configured
2361  * @vid: vlan id to be added (0 = untagged only , -1 = any)
2362  **/
2363 int i40e_vsi_add_vlan(struct i40e_vsi *vsi, s16 vid)
2364 {
2365         struct i40e_mac_filter *f, *ftmp, *add_f;
2366         bool is_netdev, is_vf;
2367
2368         is_vf = (vsi->type == I40E_VSI_SRIOV);
2369         is_netdev = !!(vsi->netdev);
2370
2371         /* Locked once because all functions invoked below iterates list*/
2372         spin_lock_bh(&vsi->mac_filter_list_lock);
2373
2374         if (is_netdev) {
2375                 add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, vid,
2376                                         is_vf, is_netdev);
2377                 if (!add_f) {
2378                         dev_info(&vsi->back->pdev->dev,
2379                                  "Could not add vlan filter %d for %pM\n",
2380                                  vid, vsi->netdev->dev_addr);
2381                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2382                         return -ENOMEM;
2383                 }
2384         }
2385
2386         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2387                 add_f = i40e_add_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2388                 if (!add_f) {
2389                         dev_info(&vsi->back->pdev->dev,
2390                                  "Could not add vlan filter %d for %pM\n",
2391                                  vid, f->macaddr);
2392                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2393                         return -ENOMEM;
2394                 }
2395         }
2396
2397         /* Now if we add a vlan tag, make sure to check if it is the first
2398          * tag (i.e. a "tag" -1 does exist) and if so replace the -1 "tag"
2399          * with 0, so we now accept untagged and specified tagged traffic
2400          * (and not all tags along with untagged)
2401          */
2402         if (vid > 0) {
2403                 if (is_netdev && i40e_find_filter(vsi, vsi->netdev->dev_addr,
2404                                                   I40E_VLAN_ANY,
2405                                                   is_vf, is_netdev)) {
2406                         i40e_del_filter(vsi, vsi->netdev->dev_addr,
2407                                         I40E_VLAN_ANY, is_vf, is_netdev);
2408                         add_f = i40e_add_filter(vsi, vsi->netdev->dev_addr, 0,
2409                                                 is_vf, is_netdev);
2410                         if (!add_f) {
2411                                 dev_info(&vsi->back->pdev->dev,
2412                                          "Could not add filter 0 for %pM\n",
2413                                          vsi->netdev->dev_addr);
2414                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2415                                 return -ENOMEM;
2416                         }
2417                 }
2418         }
2419
2420         /* Do not assume that I40E_VLAN_ANY should be reset to VLAN 0 */
2421         if (vid > 0 && !vsi->info.pvid) {
2422                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2423                         if (!i40e_find_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2424                                               is_vf, is_netdev))
2425                                 continue;
2426                         i40e_del_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2427                                         is_vf, is_netdev);
2428                         add_f = i40e_add_filter(vsi, f->macaddr,
2429                                                 0, is_vf, is_netdev);
2430                         if (!add_f) {
2431                                 dev_info(&vsi->back->pdev->dev,
2432                                          "Could not add filter 0 for %pM\n",
2433                                         f->macaddr);
2434                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2435                                 return -ENOMEM;
2436                         }
2437                 }
2438         }
2439
2440         spin_unlock_bh(&vsi->mac_filter_list_lock);
2441
2442         /* schedule our worker thread which will take care of
2443          * applying the new filter changes
2444          */
2445         i40e_service_event_schedule(vsi->back);
2446         return 0;
2447 }
2448
2449 /**
2450  * i40e_vsi_kill_vlan - Remove vsi membership for given vlan
2451  * @vsi: the vsi being configured
2452  * @vid: vlan id to be removed (0 = untagged only , -1 = any)
2453  *
2454  * Return: 0 on success or negative otherwise
2455  **/
2456 int i40e_vsi_kill_vlan(struct i40e_vsi *vsi, s16 vid)
2457 {
2458         struct net_device *netdev = vsi->netdev;
2459         struct i40e_mac_filter *f, *ftmp, *add_f;
2460         bool is_vf, is_netdev;
2461         int filter_count = 0;
2462
2463         is_vf = (vsi->type == I40E_VSI_SRIOV);
2464         is_netdev = !!(netdev);
2465
2466         /* Locked once because all functions invoked below iterates list */
2467         spin_lock_bh(&vsi->mac_filter_list_lock);
2468
2469         if (is_netdev)
2470                 i40e_del_filter(vsi, netdev->dev_addr, vid, is_vf, is_netdev);
2471
2472         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
2473                 i40e_del_filter(vsi, f->macaddr, vid, is_vf, is_netdev);
2474
2475         /* go through all the filters for this VSI and if there is only
2476          * vid == 0 it means there are no other filters, so vid 0 must
2477          * be replaced with -1. This signifies that we should from now
2478          * on accept any traffic (with any tag present, or untagged)
2479          */
2480         list_for_each_entry(f, &vsi->mac_filter_list, list) {
2481                 if (is_netdev) {
2482                         if (f->vlan &&
2483                             ether_addr_equal(netdev->dev_addr, f->macaddr))
2484                                 filter_count++;
2485                 }
2486
2487                 if (f->vlan)
2488                         filter_count++;
2489         }
2490
2491         if (!filter_count && is_netdev) {
2492                 i40e_del_filter(vsi, netdev->dev_addr, 0, is_vf, is_netdev);
2493                 f = i40e_add_filter(vsi, netdev->dev_addr, I40E_VLAN_ANY,
2494                                     is_vf, is_netdev);
2495                 if (!f) {
2496                         dev_info(&vsi->back->pdev->dev,
2497                                  "Could not add filter %d for %pM\n",
2498                                  I40E_VLAN_ANY, netdev->dev_addr);
2499                         spin_unlock_bh(&vsi->mac_filter_list_lock);
2500                         return -ENOMEM;
2501                 }
2502         }
2503
2504         if (!filter_count) {
2505                 list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
2506                         i40e_del_filter(vsi, f->macaddr, 0, is_vf, is_netdev);
2507                         add_f = i40e_add_filter(vsi, f->macaddr, I40E_VLAN_ANY,
2508                                                 is_vf, is_netdev);
2509                         if (!add_f) {
2510                                 dev_info(&vsi->back->pdev->dev,
2511                                          "Could not add filter %d for %pM\n",
2512                                          I40E_VLAN_ANY, f->macaddr);
2513                                 spin_unlock_bh(&vsi->mac_filter_list_lock);
2514                                 return -ENOMEM;
2515                         }
2516                 }
2517         }
2518
2519         spin_unlock_bh(&vsi->mac_filter_list_lock);
2520
2521         /* schedule our worker thread which will take care of
2522          * applying the new filter changes
2523          */
2524         i40e_service_event_schedule(vsi->back);
2525         return 0;
2526 }
2527
2528 /**
2529  * i40e_vlan_rx_add_vid - Add a vlan id filter to HW offload
2530  * @netdev: network interface to be adjusted
2531  * @vid: vlan id to be added
2532  *
2533  * net_device_ops implementation for adding vlan ids
2534  **/
2535 #ifdef I40E_FCOE
2536 int i40e_vlan_rx_add_vid(struct net_device *netdev,
2537                          __always_unused __be16 proto, u16 vid)
2538 #else
2539 static int i40e_vlan_rx_add_vid(struct net_device *netdev,
2540                                 __always_unused __be16 proto, u16 vid)
2541 #endif
2542 {
2543         struct i40e_netdev_priv *np = netdev_priv(netdev);
2544         struct i40e_vsi *vsi = np->vsi;
2545         int ret = 0;
2546
2547         if (vid > 4095)
2548                 return -EINVAL;
2549
2550         /* If the network stack called us with vid = 0 then
2551          * it is asking to receive priority tagged packets with
2552          * vlan id 0.  Our HW receives them by default when configured
2553          * to receive untagged packets so there is no need to add an
2554          * extra filter for vlan 0 tagged packets.
2555          */
2556         if (vid)
2557                 ret = i40e_vsi_add_vlan(vsi, vid);
2558
2559         if (!ret && (vid < VLAN_N_VID))
2560                 set_bit(vid, vsi->active_vlans);
2561
2562         return ret;
2563 }
2564
2565 /**
2566  * i40e_vlan_rx_kill_vid - Remove a vlan id filter from HW offload
2567  * @netdev: network interface to be adjusted
2568  * @vid: vlan id to be removed
2569  *
2570  * net_device_ops implementation for removing vlan ids
2571  **/
2572 #ifdef I40E_FCOE
2573 int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2574                           __always_unused __be16 proto, u16 vid)
2575 #else
2576 static int i40e_vlan_rx_kill_vid(struct net_device *netdev,
2577                                  __always_unused __be16 proto, u16 vid)
2578 #endif
2579 {
2580         struct i40e_netdev_priv *np = netdev_priv(netdev);
2581         struct i40e_vsi *vsi = np->vsi;
2582
2583         /* return code is ignored as there is nothing a user
2584          * can do about failure to remove and a log message was
2585          * already printed from the other function
2586          */
2587         i40e_vsi_kill_vlan(vsi, vid);
2588
2589         clear_bit(vid, vsi->active_vlans);
2590
2591         return 0;
2592 }
2593
2594 /**
2595  * i40e_macaddr_init - explicitly write the mac address filters
2596  *
2597  * @vsi: pointer to the vsi
2598  * @macaddr: the MAC address
2599  *
2600  * This is needed when the macaddr has been obtained by other
2601  * means than the default, e.g., from Open Firmware or IDPROM.
2602  * Returns 0 on success, negative on failure
2603  **/
2604 static int i40e_macaddr_init(struct i40e_vsi *vsi, u8 *macaddr)
2605 {
2606         int ret;
2607         struct i40e_aqc_add_macvlan_element_data element;
2608
2609         ret = i40e_aq_mac_address_write(&vsi->back->hw,
2610                                         I40E_AQC_WRITE_TYPE_LAA_WOL,
2611                                         macaddr, NULL);
2612         if (ret) {
2613                 dev_info(&vsi->back->pdev->dev,
2614                          "Addr change for VSI failed: %d\n", ret);
2615                 return -EADDRNOTAVAIL;
2616         }
2617
2618         memset(&element, 0, sizeof(element));
2619         ether_addr_copy(element.mac_addr, macaddr);
2620         element.flags = cpu_to_le16(I40E_AQC_MACVLAN_ADD_PERFECT_MATCH);
2621         ret = i40e_aq_add_macvlan(&vsi->back->hw, vsi->seid, &element, 1, NULL);
2622         if (ret) {
2623                 dev_info(&vsi->back->pdev->dev,
2624                          "add filter failed err %s aq_err %s\n",
2625                          i40e_stat_str(&vsi->back->hw, ret),
2626                          i40e_aq_str(&vsi->back->hw,
2627                                      vsi->back->hw.aq.asq_last_status));
2628         }
2629         return ret;
2630 }
2631
2632 /**
2633  * i40e_restore_vlan - Reinstate vlans when vsi/netdev comes back up
2634  * @vsi: the vsi being brought back up
2635  **/
2636 static void i40e_restore_vlan(struct i40e_vsi *vsi)
2637 {
2638         u16 vid;
2639
2640         if (!vsi->netdev)
2641                 return;
2642
2643         i40e_vlan_rx_register(vsi->netdev, vsi->netdev->features);
2644
2645         for_each_set_bit(vid, vsi->active_vlans, VLAN_N_VID)
2646                 i40e_vlan_rx_add_vid(vsi->netdev, htons(ETH_P_8021Q),
2647                                      vid);
2648 }
2649
2650 /**
2651  * i40e_vsi_add_pvid - Add pvid for the VSI
2652  * @vsi: the vsi being adjusted
2653  * @vid: the vlan id to set as a PVID
2654  **/
2655 int i40e_vsi_add_pvid(struct i40e_vsi *vsi, u16 vid)
2656 {
2657         struct i40e_vsi_context ctxt;
2658         i40e_status ret;
2659
2660         vsi->info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
2661         vsi->info.pvid = cpu_to_le16(vid);
2662         vsi->info.port_vlan_flags = I40E_AQ_VSI_PVLAN_MODE_TAGGED |
2663                                     I40E_AQ_VSI_PVLAN_INSERT_PVID |
2664                                     I40E_AQ_VSI_PVLAN_EMOD_STR;
2665
2666         ctxt.seid = vsi->seid;
2667         ctxt.info = vsi->info;
2668         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
2669         if (ret) {
2670                 dev_info(&vsi->back->pdev->dev,
2671                          "add pvid failed, err %s aq_err %s\n",
2672                          i40e_stat_str(&vsi->back->hw, ret),
2673                          i40e_aq_str(&vsi->back->hw,
2674                                      vsi->back->hw.aq.asq_last_status));
2675                 return -ENOENT;
2676         }
2677
2678         return 0;
2679 }
2680
2681 /**
2682  * i40e_vsi_remove_pvid - Remove the pvid from the VSI
2683  * @vsi: the vsi being adjusted
2684  *
2685  * Just use the vlan_rx_register() service to put it back to normal
2686  **/
2687 void i40e_vsi_remove_pvid(struct i40e_vsi *vsi)
2688 {
2689         i40e_vlan_stripping_disable(vsi);
2690
2691         vsi->info.pvid = 0;
2692 }
2693
2694 /**
2695  * i40e_vsi_setup_tx_resources - Allocate VSI Tx queue resources
2696  * @vsi: ptr to the VSI
2697  *
2698  * If this function returns with an error, then it's possible one or
2699  * more of the rings is populated (while the rest are not).  It is the
2700  * callers duty to clean those orphaned rings.
2701  *
2702  * Return 0 on success, negative on failure
2703  **/
2704 static int i40e_vsi_setup_tx_resources(struct i40e_vsi *vsi)
2705 {
2706         int i, err = 0;
2707
2708         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2709                 err = i40e_setup_tx_descriptors(vsi->tx_rings[i]);
2710
2711         return err;
2712 }
2713
2714 /**
2715  * i40e_vsi_free_tx_resources - Free Tx resources for VSI queues
2716  * @vsi: ptr to the VSI
2717  *
2718  * Free VSI's transmit software resources
2719  **/
2720 static void i40e_vsi_free_tx_resources(struct i40e_vsi *vsi)
2721 {
2722         int i;
2723
2724         if (!vsi->tx_rings)
2725                 return;
2726
2727         for (i = 0; i < vsi->num_queue_pairs; i++)
2728                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc)
2729                         i40e_free_tx_resources(vsi->tx_rings[i]);
2730 }
2731
2732 /**
2733  * i40e_vsi_setup_rx_resources - Allocate VSI queues Rx resources
2734  * @vsi: ptr to the VSI
2735  *
2736  * If this function returns with an error, then it's possible one or
2737  * more of the rings is populated (while the rest are not).  It is the
2738  * callers duty to clean those orphaned rings.
2739  *
2740  * Return 0 on success, negative on failure
2741  **/
2742 static int i40e_vsi_setup_rx_resources(struct i40e_vsi *vsi)
2743 {
2744         int i, err = 0;
2745
2746         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
2747                 err = i40e_setup_rx_descriptors(vsi->rx_rings[i]);
2748 #ifdef I40E_FCOE
2749         i40e_fcoe_setup_ddp_resources(vsi);
2750 #endif
2751         return err;
2752 }
2753
2754 /**
2755  * i40e_vsi_free_rx_resources - Free Rx Resources for VSI queues
2756  * @vsi: ptr to the VSI
2757  *
2758  * Free all receive software resources
2759  **/
2760 static void i40e_vsi_free_rx_resources(struct i40e_vsi *vsi)
2761 {
2762         int i;
2763
2764         if (!vsi->rx_rings)
2765                 return;
2766
2767         for (i = 0; i < vsi->num_queue_pairs; i++)
2768                 if (vsi->rx_rings[i] && vsi->rx_rings[i]->desc)
2769                         i40e_free_rx_resources(vsi->rx_rings[i]);
2770 #ifdef I40E_FCOE
2771         i40e_fcoe_free_ddp_resources(vsi);
2772 #endif
2773 }
2774
2775 /**
2776  * i40e_config_xps_tx_ring - Configure XPS for a Tx ring
2777  * @ring: The Tx ring to configure
2778  *
2779  * This enables/disables XPS for a given Tx descriptor ring
2780  * based on the TCs enabled for the VSI that ring belongs to.
2781  **/
2782 static void i40e_config_xps_tx_ring(struct i40e_ring *ring)
2783 {
2784         struct i40e_vsi *vsi = ring->vsi;
2785         cpumask_var_t mask;
2786
2787         if (!ring->q_vector || !ring->netdev)
2788                 return;
2789
2790         /* Single TC mode enable XPS */
2791         if (vsi->tc_config.numtc <= 1) {
2792                 if (!test_and_set_bit(__I40E_TX_XPS_INIT_DONE, &ring->state))
2793                         netif_set_xps_queue(ring->netdev,
2794                                             &ring->q_vector->affinity_mask,
2795                                             ring->queue_index);
2796         } else if (alloc_cpumask_var(&mask, GFP_KERNEL)) {
2797                 /* Disable XPS to allow selection based on TC */
2798                 bitmap_zero(cpumask_bits(mask), nr_cpumask_bits);
2799                 netif_set_xps_queue(ring->netdev, mask, ring->queue_index);
2800                 free_cpumask_var(mask);
2801         }
2802
2803         /* schedule our worker thread which will take care of
2804          * applying the new filter changes
2805          */
2806         i40e_service_event_schedule(vsi->back);
2807 }
2808
2809 /**
2810  * i40e_configure_tx_ring - Configure a transmit ring context and rest
2811  * @ring: The Tx ring to configure
2812  *
2813  * Configure the Tx descriptor ring in the HMC context.
2814  **/
2815 static int i40e_configure_tx_ring(struct i40e_ring *ring)
2816 {
2817         struct i40e_vsi *vsi = ring->vsi;
2818         u16 pf_q = vsi->base_queue + ring->queue_index;
2819         struct i40e_hw *hw = &vsi->back->hw;
2820         struct i40e_hmc_obj_txq tx_ctx;
2821         i40e_status err = 0;
2822         u32 qtx_ctl = 0;
2823
2824         /* some ATR related tx ring init */
2825         if (vsi->back->flags & I40E_FLAG_FD_ATR_ENABLED) {
2826                 ring->atr_sample_rate = vsi->back->atr_sample_rate;
2827                 ring->atr_count = 0;
2828         } else {
2829                 ring->atr_sample_rate = 0;
2830         }
2831
2832         /* configure XPS */
2833         i40e_config_xps_tx_ring(ring);
2834
2835         /* clear the context structure first */
2836         memset(&tx_ctx, 0, sizeof(tx_ctx));
2837
2838         tx_ctx.new_context = 1;
2839         tx_ctx.base = (ring->dma / 128);
2840         tx_ctx.qlen = ring->count;
2841         tx_ctx.fd_ena = !!(vsi->back->flags & (I40E_FLAG_FD_SB_ENABLED |
2842                                                I40E_FLAG_FD_ATR_ENABLED));
2843 #ifdef I40E_FCOE
2844         tx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2845 #endif
2846         tx_ctx.timesync_ena = !!(vsi->back->flags & I40E_FLAG_PTP);
2847         /* FDIR VSI tx ring can still use RS bit and writebacks */
2848         if (vsi->type != I40E_VSI_FDIR)
2849                 tx_ctx.head_wb_ena = 1;
2850         tx_ctx.head_wb_addr = ring->dma +
2851                               (ring->count * sizeof(struct i40e_tx_desc));
2852
2853         /* As part of VSI creation/update, FW allocates certain
2854          * Tx arbitration queue sets for each TC enabled for
2855          * the VSI. The FW returns the handles to these queue
2856          * sets as part of the response buffer to Add VSI,
2857          * Update VSI, etc. AQ commands. It is expected that
2858          * these queue set handles be associated with the Tx
2859          * queues by the driver as part of the TX queue context
2860          * initialization. This has to be done regardless of
2861          * DCB as by default everything is mapped to TC0.
2862          */
2863         tx_ctx.rdylist = le16_to_cpu(vsi->info.qs_handle[ring->dcb_tc]);
2864         tx_ctx.rdylist_act = 0;
2865
2866         /* clear the context in the HMC */
2867         err = i40e_clear_lan_tx_queue_context(hw, pf_q);
2868         if (err) {
2869                 dev_info(&vsi->back->pdev->dev,
2870                          "Failed to clear LAN Tx queue context on Tx ring %d (pf_q %d), error: %d\n",
2871                          ring->queue_index, pf_q, err);
2872                 return -ENOMEM;
2873         }
2874
2875         /* set the context in the HMC */
2876         err = i40e_set_lan_tx_queue_context(hw, pf_q, &tx_ctx);
2877         if (err) {
2878                 dev_info(&vsi->back->pdev->dev,
2879                          "Failed to set LAN Tx queue context on Tx ring %d (pf_q %d, error: %d\n",
2880                          ring->queue_index, pf_q, err);
2881                 return -ENOMEM;
2882         }
2883
2884         /* Now associate this queue with this PCI function */
2885         if (vsi->type == I40E_VSI_VMDQ2) {
2886                 qtx_ctl = I40E_QTX_CTL_VM_QUEUE;
2887                 qtx_ctl |= ((vsi->id) << I40E_QTX_CTL_VFVM_INDX_SHIFT) &
2888                            I40E_QTX_CTL_VFVM_INDX_MASK;
2889         } else {
2890                 qtx_ctl = I40E_QTX_CTL_PF_QUEUE;
2891         }
2892
2893         qtx_ctl |= ((hw->pf_id << I40E_QTX_CTL_PF_INDX_SHIFT) &
2894                     I40E_QTX_CTL_PF_INDX_MASK);
2895         wr32(hw, I40E_QTX_CTL(pf_q), qtx_ctl);
2896         i40e_flush(hw);
2897
2898         /* cache tail off for easier writes later */
2899         ring->tail = hw->hw_addr + I40E_QTX_TAIL(pf_q);
2900
2901         return 0;
2902 }
2903
2904 /**
2905  * i40e_configure_rx_ring - Configure a receive ring context
2906  * @ring: The Rx ring to configure
2907  *
2908  * Configure the Rx descriptor ring in the HMC context.
2909  **/
2910 static int i40e_configure_rx_ring(struct i40e_ring *ring)
2911 {
2912         struct i40e_vsi *vsi = ring->vsi;
2913         u32 chain_len = vsi->back->hw.func_caps.rx_buf_chain_len;
2914         u16 pf_q = vsi->base_queue + ring->queue_index;
2915         struct i40e_hw *hw = &vsi->back->hw;
2916         struct i40e_hmc_obj_rxq rx_ctx;
2917         i40e_status err = 0;
2918
2919         ring->state = 0;
2920
2921         /* clear the context structure first */
2922         memset(&rx_ctx, 0, sizeof(rx_ctx));
2923
2924         ring->rx_buf_len = vsi->rx_buf_len;
2925
2926         rx_ctx.dbuff = ring->rx_buf_len >> I40E_RXQ_CTX_DBUFF_SHIFT;
2927
2928         rx_ctx.base = (ring->dma / 128);
2929         rx_ctx.qlen = ring->count;
2930
2931         /* use 32 byte descriptors */
2932         rx_ctx.dsize = 1;
2933
2934         /* descriptor type is always zero
2935          * rx_ctx.dtype = 0;
2936          */
2937         rx_ctx.hsplit_0 = 0;
2938
2939         rx_ctx.rxmax = min_t(u16, vsi->max_frame, chain_len * ring->rx_buf_len);
2940         if (hw->revision_id == 0)
2941                 rx_ctx.lrxqthresh = 0;
2942         else
2943                 rx_ctx.lrxqthresh = 2;
2944         rx_ctx.crcstrip = 1;
2945         rx_ctx.l2tsel = 1;
2946         /* this controls whether VLAN is stripped from inner headers */
2947         rx_ctx.showiv = 0;
2948 #ifdef I40E_FCOE
2949         rx_ctx.fc_ena = (vsi->type == I40E_VSI_FCOE);
2950 #endif
2951         /* set the prefena field to 1 because the manual says to */
2952         rx_ctx.prefena = 1;
2953
2954         /* clear the context in the HMC */
2955         err = i40e_clear_lan_rx_queue_context(hw, pf_q);
2956         if (err) {
2957                 dev_info(&vsi->back->pdev->dev,
2958                          "Failed to clear LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2959                          ring->queue_index, pf_q, err);
2960                 return -ENOMEM;
2961         }
2962
2963         /* set the context in the HMC */
2964         err = i40e_set_lan_rx_queue_context(hw, pf_q, &rx_ctx);
2965         if (err) {
2966                 dev_info(&vsi->back->pdev->dev,
2967                          "Failed to set LAN Rx queue context on Rx ring %d (pf_q %d), error: %d\n",
2968                          ring->queue_index, pf_q, err);
2969                 return -ENOMEM;
2970         }
2971
2972         /* cache tail for quicker writes, and clear the reg before use */
2973         ring->tail = hw->hw_addr + I40E_QRX_TAIL(pf_q);
2974         writel(0, ring->tail);
2975
2976         i40e_alloc_rx_buffers(ring, I40E_DESC_UNUSED(ring));
2977
2978         return 0;
2979 }
2980
2981 /**
2982  * i40e_vsi_configure_tx - Configure the VSI for Tx
2983  * @vsi: VSI structure describing this set of rings and resources
2984  *
2985  * Configure the Tx VSI for operation.
2986  **/
2987 static int i40e_vsi_configure_tx(struct i40e_vsi *vsi)
2988 {
2989         int err = 0;
2990         u16 i;
2991
2992         for (i = 0; (i < vsi->num_queue_pairs) && !err; i++)
2993                 err = i40e_configure_tx_ring(vsi->tx_rings[i]);
2994
2995         return err;
2996 }
2997
2998 /**
2999  * i40e_vsi_configure_rx - Configure the VSI for Rx
3000  * @vsi: the VSI being configured
3001  *
3002  * Configure the Rx VSI for operation.
3003  **/
3004 static int i40e_vsi_configure_rx(struct i40e_vsi *vsi)
3005 {
3006         int err = 0;
3007         u16 i;
3008
3009         if (vsi->netdev && (vsi->netdev->mtu > ETH_DATA_LEN))
3010                 vsi->max_frame = vsi->netdev->mtu + ETH_HLEN
3011                                + ETH_FCS_LEN + VLAN_HLEN;
3012         else
3013                 vsi->max_frame = I40E_RXBUFFER_2048;
3014
3015         vsi->rx_buf_len = I40E_RXBUFFER_2048;
3016
3017 #ifdef I40E_FCOE
3018         /* setup rx buffer for FCoE */
3019         if ((vsi->type == I40E_VSI_FCOE) &&
3020             (vsi->back->flags & I40E_FLAG_FCOE_ENABLED)) {
3021                 vsi->rx_buf_len = I40E_RXBUFFER_3072;
3022                 vsi->max_frame = I40E_RXBUFFER_3072;
3023         }
3024
3025 #endif /* I40E_FCOE */
3026         /* round up for the chip's needs */
3027         vsi->rx_buf_len = ALIGN(vsi->rx_buf_len,
3028                                 BIT_ULL(I40E_RXQ_CTX_DBUFF_SHIFT));
3029
3030         /* set up individual rings */
3031         for (i = 0; i < vsi->num_queue_pairs && !err; i++)
3032                 err = i40e_configure_rx_ring(vsi->rx_rings[i]);
3033
3034         return err;
3035 }
3036
3037 /**
3038  * i40e_vsi_config_dcb_rings - Update rings to reflect DCB TC
3039  * @vsi: ptr to the VSI
3040  **/
3041 static void i40e_vsi_config_dcb_rings(struct i40e_vsi *vsi)
3042 {
3043         struct i40e_ring *tx_ring, *rx_ring;
3044         u16 qoffset, qcount;
3045         int i, n;
3046
3047         if (!(vsi->back->flags & I40E_FLAG_DCB_ENABLED)) {
3048                 /* Reset the TC information */
3049                 for (i = 0; i < vsi->num_queue_pairs; i++) {
3050                         rx_ring = vsi->rx_rings[i];
3051                         tx_ring = vsi->tx_rings[i];
3052                         rx_ring->dcb_tc = 0;
3053                         tx_ring->dcb_tc = 0;
3054                 }
3055         }
3056
3057         for (n = 0; n < I40E_MAX_TRAFFIC_CLASS; n++) {
3058                 if (!(vsi->tc_config.enabled_tc & BIT_ULL(n)))
3059                         continue;
3060
3061                 qoffset = vsi->tc_config.tc_info[n].qoffset;
3062                 qcount = vsi->tc_config.tc_info[n].qcount;
3063                 for (i = qoffset; i < (qoffset + qcount); i++) {
3064                         rx_ring = vsi->rx_rings[i];
3065                         tx_ring = vsi->tx_rings[i];
3066                         rx_ring->dcb_tc = n;
3067                         tx_ring->dcb_tc = n;
3068                 }
3069         }
3070 }
3071
3072 /**
3073  * i40e_set_vsi_rx_mode - Call set_rx_mode on a VSI
3074  * @vsi: ptr to the VSI
3075  **/
3076 static void i40e_set_vsi_rx_mode(struct i40e_vsi *vsi)
3077 {
3078         struct i40e_pf *pf = vsi->back;
3079         int err;
3080
3081         if (vsi->netdev)
3082                 i40e_set_rx_mode(vsi->netdev);
3083
3084         if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
3085                 err = i40e_macaddr_init(vsi, pf->hw.mac.addr);
3086                 if (err) {
3087                         dev_warn(&pf->pdev->dev,
3088                                  "could not set up macaddr; err %d\n", err);
3089                 }
3090         }
3091 }
3092
3093 /**
3094  * i40e_fdir_filter_restore - Restore the Sideband Flow Director filters
3095  * @vsi: Pointer to the targeted VSI
3096  *
3097  * This function replays the hlist on the hw where all the SB Flow Director
3098  * filters were saved.
3099  **/
3100 static void i40e_fdir_filter_restore(struct i40e_vsi *vsi)
3101 {
3102         struct i40e_fdir_filter *filter;
3103         struct i40e_pf *pf = vsi->back;
3104         struct hlist_node *node;
3105
3106         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
3107                 return;
3108
3109         hlist_for_each_entry_safe(filter, node,
3110                                   &pf->fdir_filter_list, fdir_node) {
3111                 i40e_add_del_fdir(vsi, filter, true);
3112         }
3113 }
3114
3115 /**
3116  * i40e_vsi_configure - Set up the VSI for action
3117  * @vsi: the VSI being configured
3118  **/
3119 static int i40e_vsi_configure(struct i40e_vsi *vsi)
3120 {
3121         int err;
3122
3123         i40e_set_vsi_rx_mode(vsi);
3124         i40e_restore_vlan(vsi);
3125         i40e_vsi_config_dcb_rings(vsi);
3126         err = i40e_vsi_configure_tx(vsi);
3127         if (!err)
3128                 err = i40e_vsi_configure_rx(vsi);
3129
3130         return err;
3131 }
3132
3133 /**
3134  * i40e_vsi_configure_msix - MSIX mode Interrupt Config in the HW
3135  * @vsi: the VSI being configured
3136  **/
3137 static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
3138 {
3139         struct i40e_pf *pf = vsi->back;
3140         struct i40e_hw *hw = &pf->hw;
3141         u16 vector;
3142         int i, q;
3143         u32 qp;
3144
3145         /* The interrupt indexing is offset by 1 in the PFINT_ITRn
3146          * and PFINT_LNKLSTn registers, e.g.:
3147          *   PFINT_ITRn[0..n-1] gets msix-1..msix-n  (qpair interrupts)
3148          */
3149         qp = vsi->base_queue;
3150         vector = vsi->base_vector;
3151         for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
3152                 struct i40e_q_vector *q_vector = vsi->q_vectors[i];
3153
3154                 q_vector->itr_countdown = ITR_COUNTDOWN_START;
3155                 q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
3156                 q_vector->rx.latency_range = I40E_LOW_LATENCY;
3157                 wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
3158                      q_vector->rx.itr);
3159                 q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
3160                 q_vector->tx.latency_range = I40E_LOW_LATENCY;
3161                 wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
3162                      q_vector->tx.itr);
3163                 wr32(hw, I40E_PFINT_RATEN(vector - 1),
3164                      INTRL_USEC_TO_REG(vsi->int_rate_limit));
3165
3166                 /* Linked list for the queuepairs assigned to this vector */
3167                 wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), qp);
3168                 for (q = 0; q < q_vector->num_ringpairs; q++) {
3169                         u32 val;
3170
3171                         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK |
3172                               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT)  |
3173                               (vector      << I40E_QINT_RQCTL_MSIX_INDX_SHIFT) |
3174                               (qp          << I40E_QINT_RQCTL_NEXTQ_INDX_SHIFT)|
3175                               (I40E_QUEUE_TYPE_TX
3176                                       << I40E_QINT_RQCTL_NEXTQ_TYPE_SHIFT);
3177
3178                         wr32(hw, I40E_QINT_RQCTL(qp), val);
3179
3180                         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK |
3181                               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT)  |
3182                               (vector      << I40E_QINT_TQCTL_MSIX_INDX_SHIFT) |
3183                               ((qp+1)      << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT)|
3184                               (I40E_QUEUE_TYPE_RX
3185                                       << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3186
3187                         /* Terminate the linked list */
3188                         if (q == (q_vector->num_ringpairs - 1))
3189                                 val |= (I40E_QUEUE_END_OF_LIST
3190                                            << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3191
3192                         wr32(hw, I40E_QINT_TQCTL(qp), val);
3193                         qp++;
3194                 }
3195         }
3196
3197         i40e_flush(hw);
3198 }
3199
3200 /**
3201  * i40e_enable_misc_int_causes - enable the non-queue interrupts
3202  * @hw: ptr to the hardware info
3203  **/
3204 static void i40e_enable_misc_int_causes(struct i40e_pf *pf)
3205 {
3206         struct i40e_hw *hw = &pf->hw;
3207         u32 val;
3208
3209         /* clear things first */
3210         wr32(hw, I40E_PFINT_ICR0_ENA, 0);  /* disable all */
3211         rd32(hw, I40E_PFINT_ICR0);         /* read to clear */
3212
3213         val = I40E_PFINT_ICR0_ENA_ECC_ERR_MASK       |
3214               I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK    |
3215               I40E_PFINT_ICR0_ENA_GRST_MASK          |
3216               I40E_PFINT_ICR0_ENA_PCI_EXCEPTION_MASK |
3217               I40E_PFINT_ICR0_ENA_GPIO_MASK          |
3218               I40E_PFINT_ICR0_ENA_HMC_ERR_MASK       |
3219               I40E_PFINT_ICR0_ENA_VFLR_MASK          |
3220               I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3221
3222         if (pf->flags & I40E_FLAG_IWARP_ENABLED)
3223                 val |= I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3224
3225         if (pf->flags & I40E_FLAG_PTP)
3226                 val |= I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3227
3228         wr32(hw, I40E_PFINT_ICR0_ENA, val);
3229
3230         /* SW_ITR_IDX = 0, but don't change INTENA */
3231         wr32(hw, I40E_PFINT_DYN_CTL0, I40E_PFINT_DYN_CTL0_SW_ITR_INDX_MASK |
3232                                         I40E_PFINT_DYN_CTL0_INTENA_MSK_MASK);
3233
3234         /* OTHER_ITR_IDX = 0 */
3235         wr32(hw, I40E_PFINT_STAT_CTL0, 0);
3236 }
3237
3238 /**
3239  * i40e_configure_msi_and_legacy - Legacy mode interrupt config in the HW
3240  * @vsi: the VSI being configured
3241  **/
3242 static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
3243 {
3244         struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3245         struct i40e_pf *pf = vsi->back;
3246         struct i40e_hw *hw = &pf->hw;
3247         u32 val;
3248
3249         /* set the ITR configuration */
3250         q_vector->itr_countdown = ITR_COUNTDOWN_START;
3251         q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
3252         q_vector->rx.latency_range = I40E_LOW_LATENCY;
3253         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
3254         q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
3255         q_vector->tx.latency_range = I40E_LOW_LATENCY;
3256         wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
3257
3258         i40e_enable_misc_int_causes(pf);
3259
3260         /* FIRSTQ_INDX = 0, FIRSTQ_TYPE = 0 (rx) */
3261         wr32(hw, I40E_PFINT_LNKLST0, 0);
3262
3263         /* Associate the queue pair to the vector and enable the queue int */
3264         val = I40E_QINT_RQCTL_CAUSE_ENA_MASK                  |
3265               (I40E_RX_ITR << I40E_QINT_RQCTL_ITR_INDX_SHIFT) |
3266               (I40E_QUEUE_TYPE_TX << I40E_QINT_TQCTL_NEXTQ_TYPE_SHIFT);
3267
3268         wr32(hw, I40E_QINT_RQCTL(0), val);
3269
3270         val = I40E_QINT_TQCTL_CAUSE_ENA_MASK                  |
3271               (I40E_TX_ITR << I40E_QINT_TQCTL_ITR_INDX_SHIFT) |
3272               (I40E_QUEUE_END_OF_LIST << I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT);
3273
3274         wr32(hw, I40E_QINT_TQCTL(0), val);
3275         i40e_flush(hw);
3276 }
3277
3278 /**
3279  * i40e_irq_dynamic_disable_icr0 - Disable default interrupt generation for icr0
3280  * @pf: board private structure
3281  **/
3282 void i40e_irq_dynamic_disable_icr0(struct i40e_pf *pf)
3283 {
3284         struct i40e_hw *hw = &pf->hw;
3285
3286         wr32(hw, I40E_PFINT_DYN_CTL0,
3287              I40E_ITR_NONE << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT);
3288         i40e_flush(hw);
3289 }
3290
3291 /**
3292  * i40e_irq_dynamic_enable_icr0 - Enable default interrupt generation for icr0
3293  * @pf: board private structure
3294  * @clearpba: true when all pending interrupt events should be cleared
3295  **/
3296 void i40e_irq_dynamic_enable_icr0(struct i40e_pf *pf, bool clearpba)
3297 {
3298         struct i40e_hw *hw = &pf->hw;
3299         u32 val;
3300
3301         val = I40E_PFINT_DYN_CTL0_INTENA_MASK   |
3302               (clearpba ? I40E_PFINT_DYN_CTL0_CLEARPBA_MASK : 0) |
3303               (I40E_ITR_NONE << I40E_PFINT_DYN_CTL0_ITR_INDX_SHIFT);
3304
3305         wr32(hw, I40E_PFINT_DYN_CTL0, val);
3306         i40e_flush(hw);
3307 }
3308
3309 /**
3310  * i40e_msix_clean_rings - MSIX mode Interrupt Handler
3311  * @irq: interrupt number
3312  * @data: pointer to a q_vector
3313  **/
3314 static irqreturn_t i40e_msix_clean_rings(int irq, void *data)
3315 {
3316         struct i40e_q_vector *q_vector = data;
3317
3318         if (!q_vector->tx.ring && !q_vector->rx.ring)
3319                 return IRQ_HANDLED;
3320
3321         napi_schedule_irqoff(&q_vector->napi);
3322
3323         return IRQ_HANDLED;
3324 }
3325
3326 /**
3327  * i40e_vsi_request_irq_msix - Initialize MSI-X interrupts
3328  * @vsi: the VSI being configured
3329  * @basename: name for the vector
3330  *
3331  * Allocates MSI-X vectors and requests interrupts from the kernel.
3332  **/
3333 static int i40e_vsi_request_irq_msix(struct i40e_vsi *vsi, char *basename)
3334 {
3335         int q_vectors = vsi->num_q_vectors;
3336         struct i40e_pf *pf = vsi->back;
3337         int base = vsi->base_vector;
3338         int rx_int_idx = 0;
3339         int tx_int_idx = 0;
3340         int vector, err;
3341
3342         for (vector = 0; vector < q_vectors; vector++) {
3343                 struct i40e_q_vector *q_vector = vsi->q_vectors[vector];
3344
3345                 if (q_vector->tx.ring && q_vector->rx.ring) {
3346                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3347                                  "%s-%s-%d", basename, "TxRx", rx_int_idx++);
3348                         tx_int_idx++;
3349                 } else if (q_vector->rx.ring) {
3350                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3351                                  "%s-%s-%d", basename, "rx", rx_int_idx++);
3352                 } else if (q_vector->tx.ring) {
3353                         snprintf(q_vector->name, sizeof(q_vector->name) - 1,
3354                                  "%s-%s-%d", basename, "tx", tx_int_idx++);
3355                 } else {
3356                         /* skip this unused q_vector */
3357                         continue;
3358                 }
3359                 err = request_irq(pf->msix_entries[base + vector].vector,
3360                                   vsi->irq_handler,
3361                                   0,
3362                                   q_vector->name,
3363                                   q_vector);
3364                 if (err) {
3365                         dev_info(&pf->pdev->dev,
3366                                  "MSIX request_irq failed, error: %d\n", err);
3367                         goto free_queue_irqs;
3368                 }
3369                 /* assign the mask for this irq */
3370                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3371                                       &q_vector->affinity_mask);
3372         }
3373
3374         vsi->irqs_ready = true;
3375         return 0;
3376
3377 free_queue_irqs:
3378         while (vector) {
3379                 vector--;
3380                 irq_set_affinity_hint(pf->msix_entries[base + vector].vector,
3381                                       NULL);
3382                 free_irq(pf->msix_entries[base + vector].vector,
3383                          &(vsi->q_vectors[vector]));
3384         }
3385         return err;
3386 }
3387
3388 /**
3389  * i40e_vsi_disable_irq - Mask off queue interrupt generation on the VSI
3390  * @vsi: the VSI being un-configured
3391  **/
3392 static void i40e_vsi_disable_irq(struct i40e_vsi *vsi)
3393 {
3394         struct i40e_pf *pf = vsi->back;
3395         struct i40e_hw *hw = &pf->hw;
3396         int base = vsi->base_vector;
3397         int i;
3398
3399         for (i = 0; i < vsi->num_queue_pairs; i++) {
3400                 wr32(hw, I40E_QINT_TQCTL(vsi->tx_rings[i]->reg_idx), 0);
3401                 wr32(hw, I40E_QINT_RQCTL(vsi->rx_rings[i]->reg_idx), 0);
3402         }
3403
3404         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3405                 for (i = vsi->base_vector;
3406                      i < (vsi->num_q_vectors + vsi->base_vector); i++)
3407                         wr32(hw, I40E_PFINT_DYN_CTLN(i - 1), 0);
3408
3409                 i40e_flush(hw);
3410                 for (i = 0; i < vsi->num_q_vectors; i++)
3411                         synchronize_irq(pf->msix_entries[i + base].vector);
3412         } else {
3413                 /* Legacy and MSI mode - this stops all interrupt handling */
3414                 wr32(hw, I40E_PFINT_ICR0_ENA, 0);
3415                 wr32(hw, I40E_PFINT_DYN_CTL0, 0);
3416                 i40e_flush(hw);
3417                 synchronize_irq(pf->pdev->irq);
3418         }
3419 }
3420
3421 /**
3422  * i40e_vsi_enable_irq - Enable IRQ for the given VSI
3423  * @vsi: the VSI being configured
3424  **/
3425 static int i40e_vsi_enable_irq(struct i40e_vsi *vsi)
3426 {
3427         struct i40e_pf *pf = vsi->back;
3428         int i;
3429
3430         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3431                 for (i = 0; i < vsi->num_q_vectors; i++)
3432                         i40e_irq_dynamic_enable(vsi, i);
3433         } else {
3434                 i40e_irq_dynamic_enable_icr0(pf, true);
3435         }
3436
3437         i40e_flush(&pf->hw);
3438         return 0;
3439 }
3440
3441 /**
3442  * i40e_stop_misc_vector - Stop the vector that handles non-queue events
3443  * @pf: board private structure
3444  **/
3445 static void i40e_stop_misc_vector(struct i40e_pf *pf)
3446 {
3447         /* Disable ICR 0 */
3448         wr32(&pf->hw, I40E_PFINT_ICR0_ENA, 0);
3449         i40e_flush(&pf->hw);
3450 }
3451
3452 /**
3453  * i40e_intr - MSI/Legacy and non-queue interrupt handler
3454  * @irq: interrupt number
3455  * @data: pointer to a q_vector
3456  *
3457  * This is the handler used for all MSI/Legacy interrupts, and deals
3458  * with both queue and non-queue interrupts.  This is also used in
3459  * MSIX mode to handle the non-queue interrupts.
3460  **/
3461 static irqreturn_t i40e_intr(int irq, void *data)
3462 {
3463         struct i40e_pf *pf = (struct i40e_pf *)data;
3464         struct i40e_hw *hw = &pf->hw;
3465         irqreturn_t ret = IRQ_NONE;
3466         u32 icr0, icr0_remaining;
3467         u32 val, ena_mask;
3468
3469         icr0 = rd32(hw, I40E_PFINT_ICR0);
3470         ena_mask = rd32(hw, I40E_PFINT_ICR0_ENA);
3471
3472         /* if sharing a legacy IRQ, we might get called w/o an intr pending */
3473         if ((icr0 & I40E_PFINT_ICR0_INTEVENT_MASK) == 0)
3474                 goto enable_intr;
3475
3476         /* if interrupt but no bits showing, must be SWINT */
3477         if (((icr0 & ~I40E_PFINT_ICR0_INTEVENT_MASK) == 0) ||
3478             (icr0 & I40E_PFINT_ICR0_SWINT_MASK))
3479                 pf->sw_int_count++;
3480
3481         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
3482             (ena_mask & I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK)) {
3483                 ena_mask &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3484                 icr0 &= ~I40E_PFINT_ICR0_ENA_PE_CRITERR_MASK;
3485                 dev_info(&pf->pdev->dev, "cleared PE_CRITERR\n");
3486         }
3487
3488         /* only q0 is used in MSI/Legacy mode, and none are used in MSIX */
3489         if (icr0 & I40E_PFINT_ICR0_QUEUE_0_MASK) {
3490                 struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
3491                 struct i40e_q_vector *q_vector = vsi->q_vectors[0];
3492
3493                 /* We do not have a way to disarm Queue causes while leaving
3494                  * interrupt enabled for all other causes, ideally
3495                  * interrupt should be disabled while we are in NAPI but
3496                  * this is not a performance path and napi_schedule()
3497                  * can deal with rescheduling.
3498                  */
3499                 if (!test_bit(__I40E_DOWN, &pf->state))
3500                         napi_schedule_irqoff(&q_vector->napi);
3501         }
3502
3503         if (icr0 & I40E_PFINT_ICR0_ADMINQ_MASK) {
3504                 ena_mask &= ~I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
3505                 set_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
3506                 i40e_debug(&pf->hw, I40E_DEBUG_NVM, "AdminQ event\n");
3507         }
3508
3509         if (icr0 & I40E_PFINT_ICR0_MAL_DETECT_MASK) {
3510                 ena_mask &= ~I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
3511                 set_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
3512         }
3513
3514         if (icr0 & I40E_PFINT_ICR0_VFLR_MASK) {
3515                 ena_mask &= ~I40E_PFINT_ICR0_ENA_VFLR_MASK;
3516                 set_bit(__I40E_VFLR_EVENT_PENDING, &pf->state);
3517         }
3518
3519         if (icr0 & I40E_PFINT_ICR0_GRST_MASK) {
3520                 if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
3521                         set_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
3522                 ena_mask &= ~I40E_PFINT_ICR0_ENA_GRST_MASK;
3523                 val = rd32(hw, I40E_GLGEN_RSTAT);
3524                 val = (val & I40E_GLGEN_RSTAT_RESET_TYPE_MASK)
3525                        >> I40E_GLGEN_RSTAT_RESET_TYPE_SHIFT;
3526                 if (val == I40E_RESET_CORER) {
3527                         pf->corer_count++;
3528                 } else if (val == I40E_RESET_GLOBR) {
3529                         pf->globr_count++;
3530                 } else if (val == I40E_RESET_EMPR) {
3531                         pf->empr_count++;
3532                         set_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state);
3533                 }
3534         }
3535
3536         if (icr0 & I40E_PFINT_ICR0_HMC_ERR_MASK) {
3537                 icr0 &= ~I40E_PFINT_ICR0_HMC_ERR_MASK;
3538                 dev_info(&pf->pdev->dev, "HMC error interrupt\n");
3539                 dev_info(&pf->pdev->dev, "HMC error info 0x%x, HMC error data 0x%x\n",
3540                          rd32(hw, I40E_PFHMC_ERRORINFO),
3541                          rd32(hw, I40E_PFHMC_ERRORDATA));
3542         }
3543
3544         if (icr0 & I40E_PFINT_ICR0_TIMESYNC_MASK) {
3545                 u32 prttsyn_stat = rd32(hw, I40E_PRTTSYN_STAT_0);
3546
3547                 if (prttsyn_stat & I40E_PRTTSYN_STAT_0_TXTIME_MASK) {
3548                         icr0 &= ~I40E_PFINT_ICR0_ENA_TIMESYNC_MASK;
3549                         i40e_ptp_tx_hwtstamp(pf);
3550                 }
3551         }
3552
3553         /* If a critical error is pending we have no choice but to reset the
3554          * device.
3555          * Report and mask out any remaining unexpected interrupts.
3556          */
3557         icr0_remaining = icr0 & ena_mask;
3558         if (icr0_remaining) {
3559                 dev_info(&pf->pdev->dev, "unhandled interrupt icr0=0x%08x\n",
3560                          icr0_remaining);
3561                 if ((icr0_remaining & I40E_PFINT_ICR0_PE_CRITERR_MASK) ||
3562                     (icr0_remaining & I40E_PFINT_ICR0_PCI_EXCEPTION_MASK) ||
3563                     (icr0_remaining & I40E_PFINT_ICR0_ECC_ERR_MASK)) {
3564                         dev_info(&pf->pdev->dev, "device will be reset\n");
3565                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
3566                         i40e_service_event_schedule(pf);
3567                 }
3568                 ena_mask &= ~icr0_remaining;
3569         }
3570         ret = IRQ_HANDLED;
3571
3572 enable_intr:
3573         /* re-enable interrupt causes */
3574         wr32(hw, I40E_PFINT_ICR0_ENA, ena_mask);
3575         if (!test_bit(__I40E_DOWN, &pf->state)) {
3576                 i40e_service_event_schedule(pf);
3577                 i40e_irq_dynamic_enable_icr0(pf, false);
3578         }
3579
3580         return ret;
3581 }
3582
3583 /**
3584  * i40e_clean_fdir_tx_irq - Reclaim resources after transmit completes
3585  * @tx_ring:  tx ring to clean
3586  * @budget:   how many cleans we're allowed
3587  *
3588  * Returns true if there's any budget left (e.g. the clean is finished)
3589  **/
3590 static bool i40e_clean_fdir_tx_irq(struct i40e_ring *tx_ring, int budget)
3591 {
3592         struct i40e_vsi *vsi = tx_ring->vsi;
3593         u16 i = tx_ring->next_to_clean;
3594         struct i40e_tx_buffer *tx_buf;
3595         struct i40e_tx_desc *tx_desc;
3596
3597         tx_buf = &tx_ring->tx_bi[i];
3598         tx_desc = I40E_TX_DESC(tx_ring, i);
3599         i -= tx_ring->count;
3600
3601         do {
3602                 struct i40e_tx_desc *eop_desc = tx_buf->next_to_watch;
3603
3604                 /* if next_to_watch is not set then there is no work pending */
3605                 if (!eop_desc)
3606                         break;
3607
3608                 /* prevent any other reads prior to eop_desc */
3609                 read_barrier_depends();
3610
3611                 /* if the descriptor isn't done, no work yet to do */
3612                 if (!(eop_desc->cmd_type_offset_bsz &
3613                       cpu_to_le64(I40E_TX_DESC_DTYPE_DESC_DONE)))
3614                         break;
3615
3616                 /* clear next_to_watch to prevent false hangs */
3617                 tx_buf->next_to_watch = NULL;
3618
3619                 tx_desc->buffer_addr = 0;
3620                 tx_desc->cmd_type_offset_bsz = 0;
3621                 /* move past filter desc */
3622                 tx_buf++;
3623                 tx_desc++;
3624                 i++;
3625                 if (unlikely(!i)) {
3626                         i -= tx_ring->count;
3627                         tx_buf = tx_ring->tx_bi;
3628                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3629                 }
3630                 /* unmap skb header data */
3631                 dma_unmap_single(tx_ring->dev,
3632                                  dma_unmap_addr(tx_buf, dma),
3633                                  dma_unmap_len(tx_buf, len),
3634                                  DMA_TO_DEVICE);
3635                 if (tx_buf->tx_flags & I40E_TX_FLAGS_FD_SB)
3636                         kfree(tx_buf->raw_buf);
3637
3638                 tx_buf->raw_buf = NULL;
3639                 tx_buf->tx_flags = 0;
3640                 tx_buf->next_to_watch = NULL;
3641                 dma_unmap_len_set(tx_buf, len, 0);
3642                 tx_desc->buffer_addr = 0;
3643                 tx_desc->cmd_type_offset_bsz = 0;
3644
3645                 /* move us past the eop_desc for start of next FD desc */
3646                 tx_buf++;
3647                 tx_desc++;
3648                 i++;
3649                 if (unlikely(!i)) {
3650                         i -= tx_ring->count;
3651                         tx_buf = tx_ring->tx_bi;
3652                         tx_desc = I40E_TX_DESC(tx_ring, 0);
3653                 }
3654
3655                 /* update budget accounting */
3656                 budget--;
3657         } while (likely(budget));
3658
3659         i += tx_ring->count;
3660         tx_ring->next_to_clean = i;
3661
3662         if (vsi->back->flags & I40E_FLAG_MSIX_ENABLED)
3663                 i40e_irq_dynamic_enable(vsi, tx_ring->q_vector->v_idx);
3664
3665         return budget > 0;
3666 }
3667
3668 /**
3669  * i40e_fdir_clean_ring - Interrupt Handler for FDIR SB ring
3670  * @irq: interrupt number
3671  * @data: pointer to a q_vector
3672  **/
3673 static irqreturn_t i40e_fdir_clean_ring(int irq, void *data)
3674 {
3675         struct i40e_q_vector *q_vector = data;
3676         struct i40e_vsi *vsi;
3677
3678         if (!q_vector->tx.ring)
3679                 return IRQ_HANDLED;
3680
3681         vsi = q_vector->tx.ring->vsi;
3682         i40e_clean_fdir_tx_irq(q_vector->tx.ring, vsi->work_limit);
3683
3684         return IRQ_HANDLED;
3685 }
3686
3687 /**
3688  * i40e_map_vector_to_qp - Assigns the queue pair to the vector
3689  * @vsi: the VSI being configured
3690  * @v_idx: vector index
3691  * @qp_idx: queue pair index
3692  **/
3693 static void i40e_map_vector_to_qp(struct i40e_vsi *vsi, int v_idx, int qp_idx)
3694 {
3695         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
3696         struct i40e_ring *tx_ring = vsi->tx_rings[qp_idx];
3697         struct i40e_ring *rx_ring = vsi->rx_rings[qp_idx];
3698
3699         tx_ring->q_vector = q_vector;
3700         tx_ring->next = q_vector->tx.ring;
3701         q_vector->tx.ring = tx_ring;
3702         q_vector->tx.count++;
3703
3704         rx_ring->q_vector = q_vector;
3705         rx_ring->next = q_vector->rx.ring;
3706         q_vector->rx.ring = rx_ring;
3707         q_vector->rx.count++;
3708 }
3709
3710 /**
3711  * i40e_vsi_map_rings_to_vectors - Maps descriptor rings to vectors
3712  * @vsi: the VSI being configured
3713  *
3714  * This function maps descriptor rings to the queue-specific vectors
3715  * we were allotted through the MSI-X enabling code.  Ideally, we'd have
3716  * one vector per queue pair, but on a constrained vector budget, we
3717  * group the queue pairs as "efficiently" as possible.
3718  **/
3719 static void i40e_vsi_map_rings_to_vectors(struct i40e_vsi *vsi)
3720 {
3721         int qp_remaining = vsi->num_queue_pairs;
3722         int q_vectors = vsi->num_q_vectors;
3723         int num_ringpairs;
3724         int v_start = 0;
3725         int qp_idx = 0;
3726
3727         /* If we don't have enough vectors for a 1-to-1 mapping, we'll have to
3728          * group them so there are multiple queues per vector.
3729          * It is also important to go through all the vectors available to be
3730          * sure that if we don't use all the vectors, that the remaining vectors
3731          * are cleared. This is especially important when decreasing the
3732          * number of queues in use.
3733          */
3734         for (; v_start < q_vectors; v_start++) {
3735                 struct i40e_q_vector *q_vector = vsi->q_vectors[v_start];
3736
3737                 num_ringpairs = DIV_ROUND_UP(qp_remaining, q_vectors - v_start);
3738
3739                 q_vector->num_ringpairs = num_ringpairs;
3740
3741                 q_vector->rx.count = 0;
3742                 q_vector->tx.count = 0;
3743                 q_vector->rx.ring = NULL;
3744                 q_vector->tx.ring = NULL;
3745
3746                 while (num_ringpairs--) {
3747                         i40e_map_vector_to_qp(vsi, v_start, qp_idx);
3748                         qp_idx++;
3749                         qp_remaining--;
3750                 }
3751         }
3752 }
3753
3754 /**
3755  * i40e_vsi_request_irq - Request IRQ from the OS
3756  * @vsi: the VSI being configured
3757  * @basename: name for the vector
3758  **/
3759 static int i40e_vsi_request_irq(struct i40e_vsi *vsi, char *basename)
3760 {
3761         struct i40e_pf *pf = vsi->back;
3762         int err;
3763
3764         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
3765                 err = i40e_vsi_request_irq_msix(vsi, basename);
3766         else if (pf->flags & I40E_FLAG_MSI_ENABLED)
3767                 err = request_irq(pf->pdev->irq, i40e_intr, 0,
3768                                   pf->int_name, pf);
3769         else
3770                 err = request_irq(pf->pdev->irq, i40e_intr, IRQF_SHARED,
3771                                   pf->int_name, pf);
3772
3773         if (err)
3774                 dev_info(&pf->pdev->dev, "request_irq failed, Error %d\n", err);
3775
3776         return err;
3777 }
3778
3779 #ifdef CONFIG_NET_POLL_CONTROLLER
3780 /**
3781  * i40e_netpoll - A Polling 'interrupt' handler
3782  * @netdev: network interface device structure
3783  *
3784  * This is used by netconsole to send skbs without having to re-enable
3785  * interrupts.  It's not called while the normal interrupt routine is executing.
3786  **/
3787 #ifdef I40E_FCOE
3788 void i40e_netpoll(struct net_device *netdev)
3789 #else
3790 static void i40e_netpoll(struct net_device *netdev)
3791 #endif
3792 {
3793         struct i40e_netdev_priv *np = netdev_priv(netdev);
3794         struct i40e_vsi *vsi = np->vsi;
3795         struct i40e_pf *pf = vsi->back;
3796         int i;
3797
3798         /* if interface is down do nothing */
3799         if (test_bit(__I40E_DOWN, &vsi->state))
3800                 return;
3801
3802         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
3803                 for (i = 0; i < vsi->num_q_vectors; i++)
3804                         i40e_msix_clean_rings(0, vsi->q_vectors[i]);
3805         } else {
3806                 i40e_intr(pf->pdev->irq, netdev);
3807         }
3808 }
3809 #endif
3810
3811 /**
3812  * i40e_pf_txq_wait - Wait for a PF's Tx queue to be enabled or disabled
3813  * @pf: the PF being configured
3814  * @pf_q: the PF queue
3815  * @enable: enable or disable state of the queue
3816  *
3817  * This routine will wait for the given Tx queue of the PF to reach the
3818  * enabled or disabled state.
3819  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3820  * multiple retries; else will return 0 in case of success.
3821  **/
3822 static int i40e_pf_txq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3823 {
3824         int i;
3825         u32 tx_reg;
3826
3827         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3828                 tx_reg = rd32(&pf->hw, I40E_QTX_ENA(pf_q));
3829                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3830                         break;
3831
3832                 usleep_range(10, 20);
3833         }
3834         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3835                 return -ETIMEDOUT;
3836
3837         return 0;
3838 }
3839
3840 /**
3841  * i40e_vsi_control_tx - Start or stop a VSI's rings
3842  * @vsi: the VSI being configured
3843  * @enable: start or stop the rings
3844  **/
3845 static int i40e_vsi_control_tx(struct i40e_vsi *vsi, bool enable)
3846 {
3847         struct i40e_pf *pf = vsi->back;
3848         struct i40e_hw *hw = &pf->hw;
3849         int i, j, pf_q, ret = 0;
3850         u32 tx_reg;
3851
3852         pf_q = vsi->base_queue;
3853         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3854
3855                 /* warn the TX unit of coming changes */
3856                 i40e_pre_tx_queue_cfg(&pf->hw, pf_q, enable);
3857                 if (!enable)
3858                         usleep_range(10, 20);
3859
3860                 for (j = 0; j < 50; j++) {
3861                         tx_reg = rd32(hw, I40E_QTX_ENA(pf_q));
3862                         if (((tx_reg >> I40E_QTX_ENA_QENA_REQ_SHIFT) & 1) ==
3863                             ((tx_reg >> I40E_QTX_ENA_QENA_STAT_SHIFT) & 1))
3864                                 break;
3865                         usleep_range(1000, 2000);
3866                 }
3867                 /* Skip if the queue is already in the requested state */
3868                 if (enable == !!(tx_reg & I40E_QTX_ENA_QENA_STAT_MASK))
3869                         continue;
3870
3871                 /* turn on/off the queue */
3872                 if (enable) {
3873                         wr32(hw, I40E_QTX_HEAD(pf_q), 0);
3874                         tx_reg |= I40E_QTX_ENA_QENA_REQ_MASK;
3875                 } else {
3876                         tx_reg &= ~I40E_QTX_ENA_QENA_REQ_MASK;
3877                 }
3878
3879                 wr32(hw, I40E_QTX_ENA(pf_q), tx_reg);
3880                 /* No waiting for the Tx queue to disable */
3881                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3882                         continue;
3883
3884                 /* wait for the change to finish */
3885                 ret = i40e_pf_txq_wait(pf, pf_q, enable);
3886                 if (ret) {
3887                         dev_info(&pf->pdev->dev,
3888                                  "VSI seid %d Tx ring %d %sable timeout\n",
3889                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3890                         break;
3891                 }
3892         }
3893
3894         if (hw->revision_id == 0)
3895                 mdelay(50);
3896         return ret;
3897 }
3898
3899 /**
3900  * i40e_pf_rxq_wait - Wait for a PF's Rx queue to be enabled or disabled
3901  * @pf: the PF being configured
3902  * @pf_q: the PF queue
3903  * @enable: enable or disable state of the queue
3904  *
3905  * This routine will wait for the given Rx queue of the PF to reach the
3906  * enabled or disabled state.
3907  * Returns -ETIMEDOUT in case of failing to reach the requested state after
3908  * multiple retries; else will return 0 in case of success.
3909  **/
3910 static int i40e_pf_rxq_wait(struct i40e_pf *pf, int pf_q, bool enable)
3911 {
3912         int i;
3913         u32 rx_reg;
3914
3915         for (i = 0; i < I40E_QUEUE_WAIT_RETRY_LIMIT; i++) {
3916                 rx_reg = rd32(&pf->hw, I40E_QRX_ENA(pf_q));
3917                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3918                         break;
3919
3920                 usleep_range(10, 20);
3921         }
3922         if (i >= I40E_QUEUE_WAIT_RETRY_LIMIT)
3923                 return -ETIMEDOUT;
3924
3925         return 0;
3926 }
3927
3928 /**
3929  * i40e_vsi_control_rx - Start or stop a VSI's rings
3930  * @vsi: the VSI being configured
3931  * @enable: start or stop the rings
3932  **/
3933 static int i40e_vsi_control_rx(struct i40e_vsi *vsi, bool enable)
3934 {
3935         struct i40e_pf *pf = vsi->back;
3936         struct i40e_hw *hw = &pf->hw;
3937         int i, j, pf_q, ret = 0;
3938         u32 rx_reg;
3939
3940         pf_q = vsi->base_queue;
3941         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
3942                 for (j = 0; j < 50; j++) {
3943                         rx_reg = rd32(hw, I40E_QRX_ENA(pf_q));
3944                         if (((rx_reg >> I40E_QRX_ENA_QENA_REQ_SHIFT) & 1) ==
3945                             ((rx_reg >> I40E_QRX_ENA_QENA_STAT_SHIFT) & 1))
3946                                 break;
3947                         usleep_range(1000, 2000);
3948                 }
3949
3950                 /* Skip if the queue is already in the requested state */
3951                 if (enable == !!(rx_reg & I40E_QRX_ENA_QENA_STAT_MASK))
3952                         continue;
3953
3954                 /* turn on/off the queue */
3955                 if (enable)
3956                         rx_reg |= I40E_QRX_ENA_QENA_REQ_MASK;
3957                 else
3958                         rx_reg &= ~I40E_QRX_ENA_QENA_REQ_MASK;
3959                 wr32(hw, I40E_QRX_ENA(pf_q), rx_reg);
3960                 /* No waiting for the Tx queue to disable */
3961                 if (!enable && test_bit(__I40E_PORT_TX_SUSPENDED, &pf->state))
3962                         continue;
3963
3964                 /* wait for the change to finish */
3965                 ret = i40e_pf_rxq_wait(pf, pf_q, enable);
3966                 if (ret) {
3967                         dev_info(&pf->pdev->dev,
3968                                  "VSI seid %d Rx ring %d %sable timeout\n",
3969                                  vsi->seid, pf_q, (enable ? "en" : "dis"));
3970                         break;
3971                 }
3972         }
3973
3974         return ret;
3975 }
3976
3977 /**
3978  * i40e_vsi_control_rings - Start or stop a VSI's rings
3979  * @vsi: the VSI being configured
3980  * @enable: start or stop the rings
3981  **/
3982 int i40e_vsi_control_rings(struct i40e_vsi *vsi, bool request)
3983 {
3984         int ret = 0;
3985
3986         /* do rx first for enable and last for disable */
3987         if (request) {
3988                 ret = i40e_vsi_control_rx(vsi, request);
3989                 if (ret)
3990                         return ret;
3991                 ret = i40e_vsi_control_tx(vsi, request);
3992         } else {
3993                 /* Ignore return value, we need to shutdown whatever we can */
3994                 i40e_vsi_control_tx(vsi, request);
3995                 i40e_vsi_control_rx(vsi, request);
3996         }
3997
3998         return ret;
3999 }
4000
4001 /**
4002  * i40e_vsi_free_irq - Free the irq association with the OS
4003  * @vsi: the VSI being configured
4004  **/
4005 static void i40e_vsi_free_irq(struct i40e_vsi *vsi)
4006 {
4007         struct i40e_pf *pf = vsi->back;
4008         struct i40e_hw *hw = &pf->hw;
4009         int base = vsi->base_vector;
4010         u32 val, qp;
4011         int i;
4012
4013         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4014                 if (!vsi->q_vectors)
4015                         return;
4016
4017                 if (!vsi->irqs_ready)
4018                         return;
4019
4020                 vsi->irqs_ready = false;
4021                 for (i = 0; i < vsi->num_q_vectors; i++) {
4022                         u16 vector = i + base;
4023
4024                         /* free only the irqs that were actually requested */
4025                         if (!vsi->q_vectors[i] ||
4026                             !vsi->q_vectors[i]->num_ringpairs)
4027                                 continue;
4028
4029                         /* clear the affinity_mask in the IRQ descriptor */
4030                         irq_set_affinity_hint(pf->msix_entries[vector].vector,
4031                                               NULL);
4032                         synchronize_irq(pf->msix_entries[vector].vector);
4033                         free_irq(pf->msix_entries[vector].vector,
4034                                  vsi->q_vectors[i]);
4035
4036                         /* Tear down the interrupt queue link list
4037                          *
4038                          * We know that they come in pairs and always
4039                          * the Rx first, then the Tx.  To clear the
4040                          * link list, stick the EOL value into the
4041                          * next_q field of the registers.
4042                          */
4043                         val = rd32(hw, I40E_PFINT_LNKLSTN(vector - 1));
4044                         qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4045                                 >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4046                         val |= I40E_QUEUE_END_OF_LIST
4047                                 << I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4048                         wr32(hw, I40E_PFINT_LNKLSTN(vector - 1), val);
4049
4050                         while (qp != I40E_QUEUE_END_OF_LIST) {
4051                                 u32 next;
4052
4053                                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4054
4055                                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4056                                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4057                                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4058                                          I40E_QINT_RQCTL_INTEVENT_MASK);
4059
4060                                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4061                                          I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4062
4063                                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4064
4065                                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4066
4067                                 next = (val & I40E_QINT_TQCTL_NEXTQ_INDX_MASK)
4068                                         >> I40E_QINT_TQCTL_NEXTQ_INDX_SHIFT;
4069
4070                                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4071                                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4072                                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4073                                          I40E_QINT_TQCTL_INTEVENT_MASK);
4074
4075                                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4076                                          I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4077
4078                                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4079                                 qp = next;
4080                         }
4081                 }
4082         } else {
4083                 free_irq(pf->pdev->irq, pf);
4084
4085                 val = rd32(hw, I40E_PFINT_LNKLST0);
4086                 qp = (val & I40E_PFINT_LNKLSTN_FIRSTQ_INDX_MASK)
4087                         >> I40E_PFINT_LNKLSTN_FIRSTQ_INDX_SHIFT;
4088                 val |= I40E_QUEUE_END_OF_LIST
4089                         << I40E_PFINT_LNKLST0_FIRSTQ_INDX_SHIFT;
4090                 wr32(hw, I40E_PFINT_LNKLST0, val);
4091
4092                 val = rd32(hw, I40E_QINT_RQCTL(qp));
4093                 val &= ~(I40E_QINT_RQCTL_MSIX_INDX_MASK  |
4094                          I40E_QINT_RQCTL_MSIX0_INDX_MASK |
4095                          I40E_QINT_RQCTL_CAUSE_ENA_MASK  |
4096                          I40E_QINT_RQCTL_INTEVENT_MASK);
4097
4098                 val |= (I40E_QINT_RQCTL_ITR_INDX_MASK |
4099                         I40E_QINT_RQCTL_NEXTQ_INDX_MASK);
4100
4101                 wr32(hw, I40E_QINT_RQCTL(qp), val);
4102
4103                 val = rd32(hw, I40E_QINT_TQCTL(qp));
4104
4105                 val &= ~(I40E_QINT_TQCTL_MSIX_INDX_MASK  |
4106                          I40E_QINT_TQCTL_MSIX0_INDX_MASK |
4107                          I40E_QINT_TQCTL_CAUSE_ENA_MASK  |
4108                          I40E_QINT_TQCTL_INTEVENT_MASK);
4109
4110                 val |= (I40E_QINT_TQCTL_ITR_INDX_MASK |
4111                         I40E_QINT_TQCTL_NEXTQ_INDX_MASK);
4112
4113                 wr32(hw, I40E_QINT_TQCTL(qp), val);
4114         }
4115 }
4116
4117 /**
4118  * i40e_free_q_vector - Free memory allocated for specific interrupt vector
4119  * @vsi: the VSI being configured
4120  * @v_idx: Index of vector to be freed
4121  *
4122  * This function frees the memory allocated to the q_vector.  In addition if
4123  * NAPI is enabled it will delete any references to the NAPI struct prior
4124  * to freeing the q_vector.
4125  **/
4126 static void i40e_free_q_vector(struct i40e_vsi *vsi, int v_idx)
4127 {
4128         struct i40e_q_vector *q_vector = vsi->q_vectors[v_idx];
4129         struct i40e_ring *ring;
4130
4131         if (!q_vector)
4132                 return;
4133
4134         /* disassociate q_vector from rings */
4135         i40e_for_each_ring(ring, q_vector->tx)
4136                 ring->q_vector = NULL;
4137
4138         i40e_for_each_ring(ring, q_vector->rx)
4139                 ring->q_vector = NULL;
4140
4141         /* only VSI w/ an associated netdev is set up w/ NAPI */
4142         if (vsi->netdev)
4143                 netif_napi_del(&q_vector->napi);
4144
4145         vsi->q_vectors[v_idx] = NULL;
4146
4147         kfree_rcu(q_vector, rcu);
4148 }
4149
4150 /**
4151  * i40e_vsi_free_q_vectors - Free memory allocated for interrupt vectors
4152  * @vsi: the VSI being un-configured
4153  *
4154  * This frees the memory allocated to the q_vectors and
4155  * deletes references to the NAPI struct.
4156  **/
4157 static void i40e_vsi_free_q_vectors(struct i40e_vsi *vsi)
4158 {
4159         int v_idx;
4160
4161         for (v_idx = 0; v_idx < vsi->num_q_vectors; v_idx++)
4162                 i40e_free_q_vector(vsi, v_idx);
4163 }
4164
4165 /**
4166  * i40e_reset_interrupt_capability - Disable interrupt setup in OS
4167  * @pf: board private structure
4168  **/
4169 static void i40e_reset_interrupt_capability(struct i40e_pf *pf)
4170 {
4171         /* If we're in Legacy mode, the interrupt was cleaned in vsi_close */
4172         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
4173                 pci_disable_msix(pf->pdev);
4174                 kfree(pf->msix_entries);
4175                 pf->msix_entries = NULL;
4176                 kfree(pf->irq_pile);
4177                 pf->irq_pile = NULL;
4178         } else if (pf->flags & I40E_FLAG_MSI_ENABLED) {
4179                 pci_disable_msi(pf->pdev);
4180         }
4181         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED);
4182 }
4183
4184 /**
4185  * i40e_clear_interrupt_scheme - Clear the current interrupt scheme settings
4186  * @pf: board private structure
4187  *
4188  * We go through and clear interrupt specific resources and reset the structure
4189  * to pre-load conditions
4190  **/
4191 static void i40e_clear_interrupt_scheme(struct i40e_pf *pf)
4192 {
4193         int i;
4194
4195         i40e_stop_misc_vector(pf);
4196         if (pf->flags & I40E_FLAG_MSIX_ENABLED && pf->msix_entries) {
4197                 synchronize_irq(pf->msix_entries[0].vector);
4198                 free_irq(pf->msix_entries[0].vector, pf);
4199         }
4200
4201         i40e_put_lump(pf->irq_pile, pf->iwarp_base_vector,
4202                       I40E_IWARP_IRQ_PILE_ID);
4203
4204         i40e_put_lump(pf->irq_pile, 0, I40E_PILE_VALID_BIT-1);
4205         for (i = 0; i < pf->num_alloc_vsi; i++)
4206                 if (pf->vsi[i])
4207                         i40e_vsi_free_q_vectors(pf->vsi[i]);
4208         i40e_reset_interrupt_capability(pf);
4209 }
4210
4211 /**
4212  * i40e_napi_enable_all - Enable NAPI for all q_vectors in the VSI
4213  * @vsi: the VSI being configured
4214  **/
4215 static void i40e_napi_enable_all(struct i40e_vsi *vsi)
4216 {
4217         int q_idx;
4218
4219         if (!vsi->netdev)
4220                 return;
4221
4222         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4223                 napi_enable(&vsi->q_vectors[q_idx]->napi);
4224 }
4225
4226 /**
4227  * i40e_napi_disable_all - Disable NAPI for all q_vectors in the VSI
4228  * @vsi: the VSI being configured
4229  **/
4230 static void i40e_napi_disable_all(struct i40e_vsi *vsi)
4231 {
4232         int q_idx;
4233
4234         if (!vsi->netdev)
4235                 return;
4236
4237         for (q_idx = 0; q_idx < vsi->num_q_vectors; q_idx++)
4238                 napi_disable(&vsi->q_vectors[q_idx]->napi);
4239 }
4240
4241 /**
4242  * i40e_vsi_close - Shut down a VSI
4243  * @vsi: the vsi to be quelled
4244  **/
4245 static void i40e_vsi_close(struct i40e_vsi *vsi)
4246 {
4247         bool reset = false;
4248
4249         if (!test_and_set_bit(__I40E_DOWN, &vsi->state))
4250                 i40e_down(vsi);
4251         i40e_vsi_free_irq(vsi);
4252         i40e_vsi_free_tx_resources(vsi);
4253         i40e_vsi_free_rx_resources(vsi);
4254         vsi->current_netdev_flags = 0;
4255         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4256                 reset = true;
4257         i40e_notify_client_of_netdev_close(vsi, reset);
4258 }
4259
4260 /**
4261  * i40e_quiesce_vsi - Pause a given VSI
4262  * @vsi: the VSI being paused
4263  **/
4264 static void i40e_quiesce_vsi(struct i40e_vsi *vsi)
4265 {
4266         if (test_bit(__I40E_DOWN, &vsi->state))
4267                 return;
4268
4269         /* No need to disable FCoE VSI when Tx suspended */
4270         if ((test_bit(__I40E_PORT_TX_SUSPENDED, &vsi->back->state)) &&
4271             vsi->type == I40E_VSI_FCOE) {
4272                 dev_dbg(&vsi->back->pdev->dev,
4273                          "VSI seid %d skipping FCoE VSI disable\n", vsi->seid);
4274                 return;
4275         }
4276
4277         set_bit(__I40E_NEEDS_RESTART, &vsi->state);
4278         if (vsi->netdev && netif_running(vsi->netdev))
4279                 vsi->netdev->netdev_ops->ndo_stop(vsi->netdev);
4280         else
4281                 i40e_vsi_close(vsi);
4282 }
4283
4284 /**
4285  * i40e_unquiesce_vsi - Resume a given VSI
4286  * @vsi: the VSI being resumed
4287  **/
4288 static void i40e_unquiesce_vsi(struct i40e_vsi *vsi)
4289 {
4290         if (!test_bit(__I40E_NEEDS_RESTART, &vsi->state))
4291                 return;
4292
4293         clear_bit(__I40E_NEEDS_RESTART, &vsi->state);
4294         if (vsi->netdev && netif_running(vsi->netdev))
4295                 vsi->netdev->netdev_ops->ndo_open(vsi->netdev);
4296         else
4297                 i40e_vsi_open(vsi);   /* this clears the DOWN bit */
4298 }
4299
4300 /**
4301  * i40e_pf_quiesce_all_vsi - Pause all VSIs on a PF
4302  * @pf: the PF
4303  **/
4304 static void i40e_pf_quiesce_all_vsi(struct i40e_pf *pf)
4305 {
4306         int v;
4307
4308         for (v = 0; v < pf->num_alloc_vsi; v++) {
4309                 if (pf->vsi[v])
4310                         i40e_quiesce_vsi(pf->vsi[v]);
4311         }
4312 }
4313
4314 /**
4315  * i40e_pf_unquiesce_all_vsi - Resume all VSIs on a PF
4316  * @pf: the PF
4317  **/
4318 static void i40e_pf_unquiesce_all_vsi(struct i40e_pf *pf)
4319 {
4320         int v;
4321
4322         for (v = 0; v < pf->num_alloc_vsi; v++) {
4323                 if (pf->vsi[v])
4324                         i40e_unquiesce_vsi(pf->vsi[v]);
4325         }
4326 }
4327
4328 #ifdef CONFIG_I40E_DCB
4329 /**
4330  * i40e_vsi_wait_queues_disabled - Wait for VSI's queues to be disabled
4331  * @vsi: the VSI being configured
4332  *
4333  * This function waits for the given VSI's queues to be disabled.
4334  **/
4335 static int i40e_vsi_wait_queues_disabled(struct i40e_vsi *vsi)
4336 {
4337         struct i40e_pf *pf = vsi->back;
4338         int i, pf_q, ret;
4339
4340         pf_q = vsi->base_queue;
4341         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4342                 /* Check and wait for the disable status of the queue */
4343                 ret = i40e_pf_txq_wait(pf, pf_q, false);
4344                 if (ret) {
4345                         dev_info(&pf->pdev->dev,
4346                                  "VSI seid %d Tx ring %d disable timeout\n",
4347                                  vsi->seid, pf_q);
4348                         return ret;
4349                 }
4350         }
4351
4352         pf_q = vsi->base_queue;
4353         for (i = 0; i < vsi->num_queue_pairs; i++, pf_q++) {
4354                 /* Check and wait for the disable status of the queue */
4355                 ret = i40e_pf_rxq_wait(pf, pf_q, false);
4356                 if (ret) {
4357                         dev_info(&pf->pdev->dev,
4358                                  "VSI seid %d Rx ring %d disable timeout\n",
4359                                  vsi->seid, pf_q);
4360                         return ret;
4361                 }
4362         }
4363
4364         return 0;
4365 }
4366
4367 /**
4368  * i40e_pf_wait_queues_disabled - Wait for all queues of PF VSIs to be disabled
4369  * @pf: the PF
4370  *
4371  * This function waits for the queues to be in disabled state for all the
4372  * VSIs that are managed by this PF.
4373  **/
4374 static int i40e_pf_wait_queues_disabled(struct i40e_pf *pf)
4375 {
4376         int v, ret = 0;
4377
4378         for (v = 0; v < pf->hw.func_caps.num_vsis; v++) {
4379                 /* No need to wait for FCoE VSI queues */
4380                 if (pf->vsi[v] && pf->vsi[v]->type != I40E_VSI_FCOE) {
4381                         ret = i40e_vsi_wait_queues_disabled(pf->vsi[v]);
4382                         if (ret)
4383                                 break;
4384                 }
4385         }
4386
4387         return ret;
4388 }
4389
4390 #endif
4391
4392 /**
4393  * i40e_detect_recover_hung_queue - Function to detect and recover hung_queue
4394  * @q_idx: TX queue number
4395  * @vsi: Pointer to VSI struct
4396  *
4397  * This function checks specified queue for given VSI. Detects hung condition.
4398  * Sets hung bit since it is two step process. Before next run of service task
4399  * if napi_poll runs, it reset 'hung' bit for respective q_vector. If not,
4400  * hung condition remain unchanged and during subsequent run, this function
4401  * issues SW interrupt to recover from hung condition.
4402  **/
4403 static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi)
4404 {
4405         struct i40e_ring *tx_ring = NULL;
4406         struct i40e_pf  *pf;
4407         u32 head, val, tx_pending_hw;
4408         int i;
4409
4410         pf = vsi->back;
4411
4412         /* now that we have an index, find the tx_ring struct */
4413         for (i = 0; i < vsi->num_queue_pairs; i++) {
4414                 if (vsi->tx_rings[i] && vsi->tx_rings[i]->desc) {
4415                         if (q_idx == vsi->tx_rings[i]->queue_index) {
4416                                 tx_ring = vsi->tx_rings[i];
4417                                 break;
4418                         }
4419                 }
4420         }
4421
4422         if (!tx_ring)
4423                 return;
4424
4425         /* Read interrupt register */
4426         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
4427                 val = rd32(&pf->hw,
4428                            I40E_PFINT_DYN_CTLN(tx_ring->q_vector->v_idx +
4429                                                tx_ring->vsi->base_vector - 1));
4430         else
4431                 val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0);
4432
4433         head = i40e_get_head(tx_ring);
4434
4435         tx_pending_hw = i40e_get_tx_pending(tx_ring, false);
4436
4437         /* HW is done executing descriptors, updated HEAD write back,
4438          * but SW hasn't processed those descriptors. If interrupt is
4439          * not generated from this point ON, it could result into
4440          * dev_watchdog detecting timeout on those netdev_queue,
4441          * hence proactively trigger SW interrupt.
4442          */
4443         if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4444                 /* NAPI Poll didn't run and clear since it was set */
4445                 if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT,
4446                                        &tx_ring->q_vector->hung_detected)) {
4447                         netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n",
4448                                     vsi->seid, q_idx, tx_pending_hw,
4449                                     tx_ring->next_to_clean, head,
4450                                     tx_ring->next_to_use,
4451                                     readl(tx_ring->tail));
4452                         netdev_info(vsi->netdev, "VSI_seid %d, Issuing force_wb for TX queue %d, Interrupt Reg: 0x%x\n",
4453                                     vsi->seid, q_idx, val);
4454                         i40e_force_wb(vsi, tx_ring->q_vector);
4455                 } else {
4456                         /* First Chance - detected possible hung */
4457                         set_bit(I40E_Q_VECTOR_HUNG_DETECT,
4458                                 &tx_ring->q_vector->hung_detected);
4459                 }
4460         }
4461
4462         /* This is the case where we have interrupts missing,
4463          * so the tx_pending in HW will most likely be 0, but we
4464          * will have tx_pending in SW since the WB happened but the
4465          * interrupt got lost.
4466          */
4467         if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) &&
4468             (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) {
4469                 if (napi_reschedule(&tx_ring->q_vector->napi))
4470                         tx_ring->tx_stats.tx_lost_interrupt++;
4471         }
4472 }
4473
4474 /**
4475  * i40e_detect_recover_hung - Function to detect and recover hung_queues
4476  * @pf:  pointer to PF struct
4477  *
4478  * LAN VSI has netdev and netdev has TX queues. This function is to check
4479  * each of those TX queues if they are hung, trigger recovery by issuing
4480  * SW interrupt.
4481  **/
4482 static void i40e_detect_recover_hung(struct i40e_pf *pf)
4483 {
4484         struct net_device *netdev;
4485         struct i40e_vsi *vsi;
4486         int i;
4487
4488         /* Only for LAN VSI */
4489         vsi = pf->vsi[pf->lan_vsi];
4490
4491         if (!vsi)
4492                 return;
4493
4494         /* Make sure, VSI state is not DOWN/RECOVERY_PENDING */
4495         if (test_bit(__I40E_DOWN, &vsi->back->state) ||
4496             test_bit(__I40E_RESET_RECOVERY_PENDING, &vsi->back->state))
4497                 return;
4498
4499         /* Make sure type is MAIN VSI */
4500         if (vsi->type != I40E_VSI_MAIN)
4501                 return;
4502
4503         netdev = vsi->netdev;
4504         if (!netdev)
4505                 return;
4506
4507         /* Bail out if netif_carrier is not OK */
4508         if (!netif_carrier_ok(netdev))
4509                 return;
4510
4511         /* Go thru' TX queues for netdev */
4512         for (i = 0; i < netdev->num_tx_queues; i++) {
4513                 struct netdev_queue *q;
4514
4515                 q = netdev_get_tx_queue(netdev, i);
4516                 if (q)
4517                         i40e_detect_recover_hung_queue(i, vsi);
4518         }
4519 }
4520
4521 /**
4522  * i40e_get_iscsi_tc_map - Return TC map for iSCSI APP
4523  * @pf: pointer to PF
4524  *
4525  * Get TC map for ISCSI PF type that will include iSCSI TC
4526  * and LAN TC.
4527  **/
4528 static u8 i40e_get_iscsi_tc_map(struct i40e_pf *pf)
4529 {
4530         struct i40e_dcb_app_priority_table app;
4531         struct i40e_hw *hw = &pf->hw;
4532         u8 enabled_tc = 1; /* TC0 is always enabled */
4533         u8 tc, i;
4534         /* Get the iSCSI APP TLV */
4535         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4536
4537         for (i = 0; i < dcbcfg->numapps; i++) {
4538                 app = dcbcfg->app[i];
4539                 if (app.selector == I40E_APP_SEL_TCPIP &&
4540                     app.protocolid == I40E_APP_PROTOID_ISCSI) {
4541                         tc = dcbcfg->etscfg.prioritytable[app.priority];
4542                         enabled_tc |= BIT(tc);
4543                         break;
4544                 }
4545         }
4546
4547         return enabled_tc;
4548 }
4549
4550 /**
4551  * i40e_dcb_get_num_tc -  Get the number of TCs from DCBx config
4552  * @dcbcfg: the corresponding DCBx configuration structure
4553  *
4554  * Return the number of TCs from given DCBx configuration
4555  **/
4556 static u8 i40e_dcb_get_num_tc(struct i40e_dcbx_config *dcbcfg)
4557 {
4558         int i, tc_unused = 0;
4559         u8 num_tc = 0;
4560         u8 ret = 0;
4561
4562         /* Scan the ETS Config Priority Table to find
4563          * traffic class enabled for a given priority
4564          * and create a bitmask of enabled TCs
4565          */
4566         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++)
4567                 num_tc |= BIT(dcbcfg->etscfg.prioritytable[i]);
4568
4569         /* Now scan the bitmask to check for
4570          * contiguous TCs starting with TC0
4571          */
4572         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4573                 if (num_tc & BIT(i)) {
4574                         if (!tc_unused) {
4575                                 ret++;
4576                         } else {
4577                                 pr_err("Non-contiguous TC - Disabling DCB\n");
4578                                 return 1;
4579                         }
4580                 } else {
4581                         tc_unused = 1;
4582                 }
4583         }
4584
4585         /* There is always at least TC0 */
4586         if (!ret)
4587                 ret = 1;
4588
4589         return ret;
4590 }
4591
4592 /**
4593  * i40e_dcb_get_enabled_tc - Get enabled traffic classes
4594  * @dcbcfg: the corresponding DCBx configuration structure
4595  *
4596  * Query the current DCB configuration and return the number of
4597  * traffic classes enabled from the given DCBX config
4598  **/
4599 static u8 i40e_dcb_get_enabled_tc(struct i40e_dcbx_config *dcbcfg)
4600 {
4601         u8 num_tc = i40e_dcb_get_num_tc(dcbcfg);
4602         u8 enabled_tc = 1;
4603         u8 i;
4604
4605         for (i = 0; i < num_tc; i++)
4606                 enabled_tc |= BIT(i);
4607
4608         return enabled_tc;
4609 }
4610
4611 /**
4612  * i40e_pf_get_num_tc - Get enabled traffic classes for PF
4613  * @pf: PF being queried
4614  *
4615  * Return number of traffic classes enabled for the given PF
4616  **/
4617 static u8 i40e_pf_get_num_tc(struct i40e_pf *pf)
4618 {
4619         struct i40e_hw *hw = &pf->hw;
4620         u8 i, enabled_tc = 1;
4621         u8 num_tc = 0;
4622         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4623
4624         /* If DCB is not enabled then always in single TC */
4625         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4626                 return 1;
4627
4628         /* SFP mode will be enabled for all TCs on port */
4629         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4630                 return i40e_dcb_get_num_tc(dcbcfg);
4631
4632         /* MFP mode return count of enabled TCs for this PF */
4633         if (pf->hw.func_caps.iscsi)
4634                 enabled_tc =  i40e_get_iscsi_tc_map(pf);
4635         else
4636                 return 1; /* Only TC0 */
4637
4638         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4639                 if (enabled_tc & BIT(i))
4640                         num_tc++;
4641         }
4642         return num_tc;
4643 }
4644
4645 /**
4646  * i40e_pf_get_default_tc - Get bitmap for first enabled TC
4647  * @pf: PF being queried
4648  *
4649  * Return a bitmap for first enabled traffic class for this PF.
4650  **/
4651 static u8 i40e_pf_get_default_tc(struct i40e_pf *pf)
4652 {
4653         u8 enabled_tc = pf->hw.func_caps.enabled_tcmap;
4654         u8 i = 0;
4655
4656         if (!enabled_tc)
4657                 return 0x1; /* TC0 */
4658
4659         /* Find the first enabled TC */
4660         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4661                 if (enabled_tc & BIT(i))
4662                         break;
4663         }
4664
4665         return BIT(i);
4666 }
4667
4668 /**
4669  * i40e_pf_get_pf_tc_map - Get bitmap for enabled traffic classes
4670  * @pf: PF being queried
4671  *
4672  * Return a bitmap for enabled traffic classes for this PF.
4673  **/
4674 static u8 i40e_pf_get_tc_map(struct i40e_pf *pf)
4675 {
4676         /* If DCB is not enabled for this PF then just return default TC */
4677         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
4678                 return i40e_pf_get_default_tc(pf);
4679
4680         /* SFP mode we want PF to be enabled for all TCs */
4681         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
4682                 return i40e_dcb_get_enabled_tc(&pf->hw.local_dcbx_config);
4683
4684         /* MFP enabled and iSCSI PF type */
4685         if (pf->hw.func_caps.iscsi)
4686                 return i40e_get_iscsi_tc_map(pf);
4687         else
4688                 return i40e_pf_get_default_tc(pf);
4689 }
4690
4691 /**
4692  * i40e_vsi_get_bw_info - Query VSI BW Information
4693  * @vsi: the VSI being queried
4694  *
4695  * Returns 0 on success, negative value on failure
4696  **/
4697 static int i40e_vsi_get_bw_info(struct i40e_vsi *vsi)
4698 {
4699         struct i40e_aqc_query_vsi_ets_sla_config_resp bw_ets_config = {0};
4700         struct i40e_aqc_query_vsi_bw_config_resp bw_config = {0};
4701         struct i40e_pf *pf = vsi->back;
4702         struct i40e_hw *hw = &pf->hw;
4703         i40e_status ret;
4704         u32 tc_bw_max;
4705         int i;
4706
4707         /* Get the VSI level BW configuration */
4708         ret = i40e_aq_query_vsi_bw_config(hw, vsi->seid, &bw_config, NULL);
4709         if (ret) {
4710                 dev_info(&pf->pdev->dev,
4711                          "couldn't get PF vsi bw config, err %s aq_err %s\n",
4712                          i40e_stat_str(&pf->hw, ret),
4713                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4714                 return -EINVAL;
4715         }
4716
4717         /* Get the VSI level BW configuration per TC */
4718         ret = i40e_aq_query_vsi_ets_sla_config(hw, vsi->seid, &bw_ets_config,
4719                                                NULL);
4720         if (ret) {
4721                 dev_info(&pf->pdev->dev,
4722                          "couldn't get PF vsi ets bw config, err %s aq_err %s\n",
4723                          i40e_stat_str(&pf->hw, ret),
4724                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4725                 return -EINVAL;
4726         }
4727
4728         if (bw_config.tc_valid_bits != bw_ets_config.tc_valid_bits) {
4729                 dev_info(&pf->pdev->dev,
4730                          "Enabled TCs mismatch from querying VSI BW info 0x%08x 0x%08x\n",
4731                          bw_config.tc_valid_bits,
4732                          bw_ets_config.tc_valid_bits);
4733                 /* Still continuing */
4734         }
4735
4736         vsi->bw_limit = le16_to_cpu(bw_config.port_bw_limit);
4737         vsi->bw_max_quanta = bw_config.max_bw;
4738         tc_bw_max = le16_to_cpu(bw_ets_config.tc_bw_max[0]) |
4739                     (le16_to_cpu(bw_ets_config.tc_bw_max[1]) << 16);
4740         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4741                 vsi->bw_ets_share_credits[i] = bw_ets_config.share_credits[i];
4742                 vsi->bw_ets_limit_credits[i] =
4743                                         le16_to_cpu(bw_ets_config.credits[i]);
4744                 /* 3 bits out of 4 for each TC */
4745                 vsi->bw_ets_max_quanta[i] = (u8)((tc_bw_max >> (i*4)) & 0x7);
4746         }
4747
4748         return 0;
4749 }
4750
4751 /**
4752  * i40e_vsi_configure_bw_alloc - Configure VSI BW allocation per TC
4753  * @vsi: the VSI being configured
4754  * @enabled_tc: TC bitmap
4755  * @bw_credits: BW shared credits per TC
4756  *
4757  * Returns 0 on success, negative value on failure
4758  **/
4759 static int i40e_vsi_configure_bw_alloc(struct i40e_vsi *vsi, u8 enabled_tc,
4760                                        u8 *bw_share)
4761 {
4762         struct i40e_aqc_configure_vsi_tc_bw_data bw_data;
4763         i40e_status ret;
4764         int i;
4765
4766         bw_data.tc_valid_bits = enabled_tc;
4767         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4768                 bw_data.tc_bw_credits[i] = bw_share[i];
4769
4770         ret = i40e_aq_config_vsi_tc_bw(&vsi->back->hw, vsi->seid, &bw_data,
4771                                        NULL);
4772         if (ret) {
4773                 dev_info(&vsi->back->pdev->dev,
4774                          "AQ command Config VSI BW allocation per TC failed = %d\n",
4775                          vsi->back->hw.aq.asq_last_status);
4776                 return -EINVAL;
4777         }
4778
4779         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++)
4780                 vsi->info.qs_handle[i] = bw_data.qs_handles[i];
4781
4782         return 0;
4783 }
4784
4785 /**
4786  * i40e_vsi_config_netdev_tc - Setup the netdev TC configuration
4787  * @vsi: the VSI being configured
4788  * @enabled_tc: TC map to be enabled
4789  *
4790  **/
4791 static void i40e_vsi_config_netdev_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4792 {
4793         struct net_device *netdev = vsi->netdev;
4794         struct i40e_pf *pf = vsi->back;
4795         struct i40e_hw *hw = &pf->hw;
4796         u8 netdev_tc = 0;
4797         int i;
4798         struct i40e_dcbx_config *dcbcfg = &hw->local_dcbx_config;
4799
4800         if (!netdev)
4801                 return;
4802
4803         if (!enabled_tc) {
4804                 netdev_reset_tc(netdev);
4805                 return;
4806         }
4807
4808         /* Set up actual enabled TCs on the VSI */
4809         if (netdev_set_num_tc(netdev, vsi->tc_config.numtc))
4810                 return;
4811
4812         /* set per TC queues for the VSI */
4813         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4814                 /* Only set TC queues for enabled tcs
4815                  *
4816                  * e.g. For a VSI that has TC0 and TC3 enabled the
4817                  * enabled_tc bitmap would be 0x00001001; the driver
4818                  * will set the numtc for netdev as 2 that will be
4819                  * referenced by the netdev layer as TC 0 and 1.
4820                  */
4821                 if (vsi->tc_config.enabled_tc & BIT(i))
4822                         netdev_set_tc_queue(netdev,
4823                                         vsi->tc_config.tc_info[i].netdev_tc,
4824                                         vsi->tc_config.tc_info[i].qcount,
4825                                         vsi->tc_config.tc_info[i].qoffset);
4826         }
4827
4828         /* Assign UP2TC map for the VSI */
4829         for (i = 0; i < I40E_MAX_USER_PRIORITY; i++) {
4830                 /* Get the actual TC# for the UP */
4831                 u8 ets_tc = dcbcfg->etscfg.prioritytable[i];
4832                 /* Get the mapped netdev TC# for the UP */
4833                 netdev_tc =  vsi->tc_config.tc_info[ets_tc].netdev_tc;
4834                 netdev_set_prio_tc_map(netdev, i, netdev_tc);
4835         }
4836 }
4837
4838 /**
4839  * i40e_vsi_update_queue_map - Update our copy of VSi info with new queue map
4840  * @vsi: the VSI being configured
4841  * @ctxt: the ctxt buffer returned from AQ VSI update param command
4842  **/
4843 static void i40e_vsi_update_queue_map(struct i40e_vsi *vsi,
4844                                       struct i40e_vsi_context *ctxt)
4845 {
4846         /* copy just the sections touched not the entire info
4847          * since not all sections are valid as returned by
4848          * update vsi params
4849          */
4850         vsi->info.mapping_flags = ctxt->info.mapping_flags;
4851         memcpy(&vsi->info.queue_mapping,
4852                &ctxt->info.queue_mapping, sizeof(vsi->info.queue_mapping));
4853         memcpy(&vsi->info.tc_mapping, ctxt->info.tc_mapping,
4854                sizeof(vsi->info.tc_mapping));
4855 }
4856
4857 /**
4858  * i40e_vsi_config_tc - Configure VSI Tx Scheduler for given TC map
4859  * @vsi: VSI to be configured
4860  * @enabled_tc: TC bitmap
4861  *
4862  * This configures a particular VSI for TCs that are mapped to the
4863  * given TC bitmap. It uses default bandwidth share for TCs across
4864  * VSIs to configure TC for a particular VSI.
4865  *
4866  * NOTE:
4867  * It is expected that the VSI queues have been quisced before calling
4868  * this function.
4869  **/
4870 static int i40e_vsi_config_tc(struct i40e_vsi *vsi, u8 enabled_tc)
4871 {
4872         u8 bw_share[I40E_MAX_TRAFFIC_CLASS] = {0};
4873         struct i40e_vsi_context ctxt;
4874         int ret = 0;
4875         int i;
4876
4877         /* Check if enabled_tc is same as existing or new TCs */
4878         if (vsi->tc_config.enabled_tc == enabled_tc)
4879                 return ret;
4880
4881         /* Enable ETS TCs with equal BW Share for now across all VSIs */
4882         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4883                 if (enabled_tc & BIT(i))
4884                         bw_share[i] = 1;
4885         }
4886
4887         ret = i40e_vsi_configure_bw_alloc(vsi, enabled_tc, bw_share);
4888         if (ret) {
4889                 dev_info(&vsi->back->pdev->dev,
4890                          "Failed configuring TC map %d for VSI %d\n",
4891                          enabled_tc, vsi->seid);
4892                 goto out;
4893         }
4894
4895         /* Update Queue Pairs Mapping for currently enabled UPs */
4896         ctxt.seid = vsi->seid;
4897         ctxt.pf_num = vsi->back->hw.pf_id;
4898         ctxt.vf_num = 0;
4899         ctxt.uplink_seid = vsi->uplink_seid;
4900         ctxt.info = vsi->info;
4901         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
4902
4903         if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
4904                 ctxt.info.valid_sections |=
4905                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
4906                 ctxt.info.queueing_opt_flags |= I40E_AQ_VSI_QUE_OPT_TCP_ENA;
4907         }
4908
4909         /* Update the VSI after updating the VSI queue-mapping information */
4910         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
4911         if (ret) {
4912                 dev_info(&vsi->back->pdev->dev,
4913                          "Update vsi tc config failed, err %s aq_err %s\n",
4914                          i40e_stat_str(&vsi->back->hw, ret),
4915                          i40e_aq_str(&vsi->back->hw,
4916                                      vsi->back->hw.aq.asq_last_status));
4917                 goto out;
4918         }
4919         /* update the local VSI info with updated queue map */
4920         i40e_vsi_update_queue_map(vsi, &ctxt);
4921         vsi->info.valid_sections = 0;
4922
4923         /* Update current VSI BW information */
4924         ret = i40e_vsi_get_bw_info(vsi);
4925         if (ret) {
4926                 dev_info(&vsi->back->pdev->dev,
4927                          "Failed updating vsi bw info, err %s aq_err %s\n",
4928                          i40e_stat_str(&vsi->back->hw, ret),
4929                          i40e_aq_str(&vsi->back->hw,
4930                                      vsi->back->hw.aq.asq_last_status));
4931                 goto out;
4932         }
4933
4934         /* Update the netdev TC setup */
4935         i40e_vsi_config_netdev_tc(vsi, enabled_tc);
4936 out:
4937         return ret;
4938 }
4939
4940 /**
4941  * i40e_veb_config_tc - Configure TCs for given VEB
4942  * @veb: given VEB
4943  * @enabled_tc: TC bitmap
4944  *
4945  * Configures given TC bitmap for VEB (switching) element
4946  **/
4947 int i40e_veb_config_tc(struct i40e_veb *veb, u8 enabled_tc)
4948 {
4949         struct i40e_aqc_configure_switching_comp_bw_config_data bw_data = {0};
4950         struct i40e_pf *pf = veb->pf;
4951         int ret = 0;
4952         int i;
4953
4954         /* No TCs or already enabled TCs just return */
4955         if (!enabled_tc || veb->enabled_tc == enabled_tc)
4956                 return ret;
4957
4958         bw_data.tc_valid_bits = enabled_tc;
4959         /* bw_data.absolute_credits is not set (relative) */
4960
4961         /* Enable ETS TCs with equal BW Share for now */
4962         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
4963                 if (enabled_tc & BIT(i))
4964                         bw_data.tc_bw_share_credits[i] = 1;
4965         }
4966
4967         ret = i40e_aq_config_switch_comp_bw_config(&pf->hw, veb->seid,
4968                                                    &bw_data, NULL);
4969         if (ret) {
4970                 dev_info(&pf->pdev->dev,
4971                          "VEB bw config failed, err %s aq_err %s\n",
4972                          i40e_stat_str(&pf->hw, ret),
4973                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4974                 goto out;
4975         }
4976
4977         /* Update the BW information */
4978         ret = i40e_veb_get_bw_info(veb);
4979         if (ret) {
4980                 dev_info(&pf->pdev->dev,
4981                          "Failed getting veb bw config, err %s aq_err %s\n",
4982                          i40e_stat_str(&pf->hw, ret),
4983                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
4984         }
4985
4986 out:
4987         return ret;
4988 }
4989
4990 #ifdef CONFIG_I40E_DCB
4991 /**
4992  * i40e_dcb_reconfigure - Reconfigure all VEBs and VSIs
4993  * @pf: PF struct
4994  *
4995  * Reconfigure VEB/VSIs on a given PF; it is assumed that
4996  * the caller would've quiesce all the VSIs before calling
4997  * this function
4998  **/
4999 static void i40e_dcb_reconfigure(struct i40e_pf *pf)
5000 {
5001         u8 tc_map = 0;
5002         int ret;
5003         u8 v;
5004
5005         /* Enable the TCs available on PF to all VEBs */
5006         tc_map = i40e_pf_get_tc_map(pf);
5007         for (v = 0; v < I40E_MAX_VEB; v++) {
5008                 if (!pf->veb[v])
5009                         continue;
5010                 ret = i40e_veb_config_tc(pf->veb[v], tc_map);
5011                 if (ret) {
5012                         dev_info(&pf->pdev->dev,
5013                                  "Failed configuring TC for VEB seid=%d\n",
5014                                  pf->veb[v]->seid);
5015                         /* Will try to configure as many components */
5016                 }
5017         }
5018
5019         /* Update each VSI */
5020         for (v = 0; v < pf->num_alloc_vsi; v++) {
5021                 if (!pf->vsi[v])
5022                         continue;
5023
5024                 /* - Enable all TCs for the LAN VSI
5025 #ifdef I40E_FCOE
5026                  * - For FCoE VSI only enable the TC configured
5027                  *   as per the APP TLV
5028 #endif
5029                  * - For all others keep them at TC0 for now
5030                  */
5031                 if (v == pf->lan_vsi)
5032                         tc_map = i40e_pf_get_tc_map(pf);
5033                 else
5034                         tc_map = i40e_pf_get_default_tc(pf);
5035 #ifdef I40E_FCOE
5036                 if (pf->vsi[v]->type == I40E_VSI_FCOE)
5037                         tc_map = i40e_get_fcoe_tc_map(pf);
5038 #endif /* #ifdef I40E_FCOE */
5039
5040                 ret = i40e_vsi_config_tc(pf->vsi[v], tc_map);
5041                 if (ret) {
5042                         dev_info(&pf->pdev->dev,
5043                                  "Failed configuring TC for VSI seid=%d\n",
5044                                  pf->vsi[v]->seid);
5045                         /* Will try to configure as many components */
5046                 } else {
5047                         /* Re-configure VSI vectors based on updated TC map */
5048                         i40e_vsi_map_rings_to_vectors(pf->vsi[v]);
5049                         if (pf->vsi[v]->netdev)
5050                                 i40e_dcbnl_set_all(pf->vsi[v]);
5051                 }
5052         }
5053 }
5054
5055 /**
5056  * i40e_resume_port_tx - Resume port Tx
5057  * @pf: PF struct
5058  *
5059  * Resume a port's Tx and issue a PF reset in case of failure to
5060  * resume.
5061  **/
5062 static int i40e_resume_port_tx(struct i40e_pf *pf)
5063 {
5064         struct i40e_hw *hw = &pf->hw;
5065         int ret;
5066
5067         ret = i40e_aq_resume_port_tx(hw, NULL);
5068         if (ret) {
5069                 dev_info(&pf->pdev->dev,
5070                          "Resume Port Tx failed, err %s aq_err %s\n",
5071                           i40e_stat_str(&pf->hw, ret),
5072                           i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5073                 /* Schedule PF reset to recover */
5074                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5075                 i40e_service_event_schedule(pf);
5076         }
5077
5078         return ret;
5079 }
5080
5081 /**
5082  * i40e_init_pf_dcb - Initialize DCB configuration
5083  * @pf: PF being configured
5084  *
5085  * Query the current DCB configuration and cache it
5086  * in the hardware structure
5087  **/
5088 static int i40e_init_pf_dcb(struct i40e_pf *pf)
5089 {
5090         struct i40e_hw *hw = &pf->hw;
5091         int err = 0;
5092
5093         /* Do not enable DCB for SW1 and SW2 images even if the FW is capable */
5094         if (pf->flags & I40E_FLAG_NO_DCB_SUPPORT)
5095                 goto out;
5096
5097         /* Get the initial DCB configuration */
5098         err = i40e_init_dcb(hw);
5099         if (!err) {
5100                 /* Device/Function is not DCBX capable */
5101                 if ((!hw->func_caps.dcb) ||
5102                     (hw->dcbx_status == I40E_DCBX_STATUS_DISABLED)) {
5103                         dev_info(&pf->pdev->dev,
5104                                  "DCBX offload is not supported or is disabled for this PF.\n");
5105
5106                         if (pf->flags & I40E_FLAG_MFP_ENABLED)
5107                                 goto out;
5108
5109                 } else {
5110                         /* When status is not DISABLED then DCBX in FW */
5111                         pf->dcbx_cap = DCB_CAP_DCBX_LLD_MANAGED |
5112                                        DCB_CAP_DCBX_VER_IEEE;
5113
5114                         pf->flags |= I40E_FLAG_DCB_CAPABLE;
5115                         /* Enable DCB tagging only when more than one TC
5116                          * or explicitly disable if only one TC
5117                          */
5118                         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5119                                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5120                         else
5121                                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5122                         dev_dbg(&pf->pdev->dev,
5123                                 "DCBX offload is supported for this PF.\n");
5124                 }
5125         } else {
5126                 dev_info(&pf->pdev->dev,
5127                          "Query for DCB configuration failed, err %s aq_err %s\n",
5128                          i40e_stat_str(&pf->hw, err),
5129                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5130         }
5131
5132 out:
5133         return err;
5134 }
5135 #endif /* CONFIG_I40E_DCB */
5136 #define SPEED_SIZE 14
5137 #define FC_SIZE 8
5138 /**
5139  * i40e_print_link_message - print link up or down
5140  * @vsi: the VSI for which link needs a message
5141  */
5142 void i40e_print_link_message(struct i40e_vsi *vsi, bool isup)
5143 {
5144         char *speed = "Unknown";
5145         char *fc = "Unknown";
5146
5147         if (vsi->current_isup == isup)
5148                 return;
5149         vsi->current_isup = isup;
5150         if (!isup) {
5151                 netdev_info(vsi->netdev, "NIC Link is Down\n");
5152                 return;
5153         }
5154
5155         /* Warn user if link speed on NPAR enabled partition is not at
5156          * least 10GB
5157          */
5158         if (vsi->back->hw.func_caps.npar_enable &&
5159             (vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_1GB ||
5160              vsi->back->hw.phy.link_info.link_speed == I40E_LINK_SPEED_100MB))
5161                 netdev_warn(vsi->netdev,
5162                             "The partition detected link speed that is less than 10Gbps\n");
5163
5164         switch (vsi->back->hw.phy.link_info.link_speed) {
5165         case I40E_LINK_SPEED_40GB:
5166                 speed = "40 G";
5167                 break;
5168         case I40E_LINK_SPEED_20GB:
5169                 speed = "20 G";
5170                 break;
5171         case I40E_LINK_SPEED_10GB:
5172                 speed = "10 G";
5173                 break;
5174         case I40E_LINK_SPEED_1GB:
5175                 speed = "1000 M";
5176                 break;
5177         case I40E_LINK_SPEED_100MB:
5178                 speed = "100 M";
5179                 break;
5180         default:
5181                 break;
5182         }
5183
5184         switch (vsi->back->hw.fc.current_mode) {
5185         case I40E_FC_FULL:
5186                 fc = "RX/TX";
5187                 break;
5188         case I40E_FC_TX_PAUSE:
5189                 fc = "TX";
5190                 break;
5191         case I40E_FC_RX_PAUSE:
5192                 fc = "RX";
5193                 break;
5194         default:
5195                 fc = "None";
5196                 break;
5197         }
5198
5199         netdev_info(vsi->netdev, "NIC Link is Up %sbps Full Duplex, Flow Control: %s\n",
5200                     speed, fc);
5201 }
5202
5203 /**
5204  * i40e_up_complete - Finish the last steps of bringing up a connection
5205  * @vsi: the VSI being configured
5206  **/
5207 static int i40e_up_complete(struct i40e_vsi *vsi)
5208 {
5209         struct i40e_pf *pf = vsi->back;
5210         int err;
5211
5212         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
5213                 i40e_vsi_configure_msix(vsi);
5214         else
5215                 i40e_configure_msi_and_legacy(vsi);
5216
5217         /* start rings */
5218         err = i40e_vsi_control_rings(vsi, true);
5219         if (err)
5220                 return err;
5221
5222         clear_bit(__I40E_DOWN, &vsi->state);
5223         i40e_napi_enable_all(vsi);
5224         i40e_vsi_enable_irq(vsi);
5225
5226         if ((pf->hw.phy.link_info.link_info & I40E_AQ_LINK_UP) &&
5227             (vsi->netdev)) {
5228                 i40e_print_link_message(vsi, true);
5229                 netif_tx_start_all_queues(vsi->netdev);
5230                 netif_carrier_on(vsi->netdev);
5231         } else if (vsi->netdev) {
5232                 i40e_print_link_message(vsi, false);
5233                 /* need to check for qualified module here*/
5234                 if ((pf->hw.phy.link_info.link_info &
5235                         I40E_AQ_MEDIA_AVAILABLE) &&
5236                     (!(pf->hw.phy.link_info.an_info &
5237                         I40E_AQ_QUALIFIED_MODULE)))
5238                         netdev_err(vsi->netdev,
5239                                    "the driver failed to link because an unqualified module was detected.");
5240         }
5241
5242         /* replay FDIR SB filters */
5243         if (vsi->type == I40E_VSI_FDIR) {
5244                 /* reset fd counters */
5245                 pf->fd_add_err = pf->fd_atr_cnt = 0;
5246                 if (pf->fd_tcp_rule > 0) {
5247                         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5248                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5249                                 dev_info(&pf->pdev->dev, "Forcing ATR off, sideband rules for TCP/IPv4 exist\n");
5250                         pf->fd_tcp_rule = 0;
5251                 }
5252                 i40e_fdir_filter_restore(vsi);
5253         }
5254
5255         /* On the next run of the service_task, notify any clients of the new
5256          * opened netdev
5257          */
5258         pf->flags |= I40E_FLAG_SERVICE_CLIENT_REQUESTED;
5259         i40e_service_event_schedule(pf);
5260
5261         return 0;
5262 }
5263
5264 /**
5265  * i40e_vsi_reinit_locked - Reset the VSI
5266  * @vsi: the VSI being configured
5267  *
5268  * Rebuild the ring structs after some configuration
5269  * has changed, e.g. MTU size.
5270  **/
5271 static void i40e_vsi_reinit_locked(struct i40e_vsi *vsi)
5272 {
5273         struct i40e_pf *pf = vsi->back;
5274
5275         WARN_ON(in_interrupt());
5276         while (test_and_set_bit(__I40E_CONFIG_BUSY, &pf->state))
5277                 usleep_range(1000, 2000);
5278         i40e_down(vsi);
5279
5280         i40e_up(vsi);
5281         clear_bit(__I40E_CONFIG_BUSY, &pf->state);
5282 }
5283
5284 /**
5285  * i40e_up - Bring the connection back up after being down
5286  * @vsi: the VSI being configured
5287  **/
5288 int i40e_up(struct i40e_vsi *vsi)
5289 {
5290         int err;
5291
5292         err = i40e_vsi_configure(vsi);
5293         if (!err)
5294                 err = i40e_up_complete(vsi);
5295
5296         return err;
5297 }
5298
5299 /**
5300  * i40e_down - Shutdown the connection processing
5301  * @vsi: the VSI being stopped
5302  **/
5303 void i40e_down(struct i40e_vsi *vsi)
5304 {
5305         int i;
5306
5307         /* It is assumed that the caller of this function
5308          * sets the vsi->state __I40E_DOWN bit.
5309          */
5310         if (vsi->netdev) {
5311                 netif_carrier_off(vsi->netdev);
5312                 netif_tx_disable(vsi->netdev);
5313         }
5314         i40e_vsi_disable_irq(vsi);
5315         i40e_vsi_control_rings(vsi, false);
5316         i40e_napi_disable_all(vsi);
5317
5318         for (i = 0; i < vsi->num_queue_pairs; i++) {
5319                 i40e_clean_tx_ring(vsi->tx_rings[i]);
5320                 i40e_clean_rx_ring(vsi->rx_rings[i]);
5321         }
5322
5323         i40e_notify_client_of_netdev_close(vsi, false);
5324
5325 }
5326
5327 /**
5328  * i40e_setup_tc - configure multiple traffic classes
5329  * @netdev: net device to configure
5330  * @tc: number of traffic classes to enable
5331  **/
5332 static int i40e_setup_tc(struct net_device *netdev, u8 tc)
5333 {
5334         struct i40e_netdev_priv *np = netdev_priv(netdev);
5335         struct i40e_vsi *vsi = np->vsi;
5336         struct i40e_pf *pf = vsi->back;
5337         u8 enabled_tc = 0;
5338         int ret = -EINVAL;
5339         int i;
5340
5341         /* Check if DCB enabled to continue */
5342         if (!(pf->flags & I40E_FLAG_DCB_ENABLED)) {
5343                 netdev_info(netdev, "DCB is not enabled for adapter\n");
5344                 goto exit;
5345         }
5346
5347         /* Check if MFP enabled */
5348         if (pf->flags & I40E_FLAG_MFP_ENABLED) {
5349                 netdev_info(netdev, "Configuring TC not supported in MFP mode\n");
5350                 goto exit;
5351         }
5352
5353         /* Check whether tc count is within enabled limit */
5354         if (tc > i40e_pf_get_num_tc(pf)) {
5355                 netdev_info(netdev, "TC count greater than enabled on link for adapter\n");
5356                 goto exit;
5357         }
5358
5359         /* Generate TC map for number of tc requested */
5360         for (i = 0; i < tc; i++)
5361                 enabled_tc |= BIT(i);
5362
5363         /* Requesting same TC configuration as already enabled */
5364         if (enabled_tc == vsi->tc_config.enabled_tc)
5365                 return 0;
5366
5367         /* Quiesce VSI queues */
5368         i40e_quiesce_vsi(vsi);
5369
5370         /* Configure VSI for enabled TCs */
5371         ret = i40e_vsi_config_tc(vsi, enabled_tc);
5372         if (ret) {
5373                 netdev_info(netdev, "Failed configuring TC for VSI seid=%d\n",
5374                             vsi->seid);
5375                 goto exit;
5376         }
5377
5378         /* Unquiesce VSI */
5379         i40e_unquiesce_vsi(vsi);
5380
5381 exit:
5382         return ret;
5383 }
5384
5385 #ifdef I40E_FCOE
5386 int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5387                     struct tc_to_netdev *tc)
5388 #else
5389 static int __i40e_setup_tc(struct net_device *netdev, u32 handle, __be16 proto,
5390                            struct tc_to_netdev *tc)
5391 #endif
5392 {
5393         if (handle != TC_H_ROOT || tc->type != TC_SETUP_MQPRIO)
5394                 return -EINVAL;
5395         return i40e_setup_tc(netdev, tc->tc);
5396 }
5397
5398 /**
5399  * i40e_open - Called when a network interface is made active
5400  * @netdev: network interface device structure
5401  *
5402  * The open entry point is called when a network interface is made
5403  * active by the system (IFF_UP).  At this point all resources needed
5404  * for transmit and receive operations are allocated, the interrupt
5405  * handler is registered with the OS, the netdev watchdog subtask is
5406  * enabled, and the stack is notified that the interface is ready.
5407  *
5408  * Returns 0 on success, negative value on failure
5409  **/
5410 int i40e_open(struct net_device *netdev)
5411 {
5412         struct i40e_netdev_priv *np = netdev_priv(netdev);
5413         struct i40e_vsi *vsi = np->vsi;
5414         struct i40e_pf *pf = vsi->back;
5415         int err;
5416
5417         /* disallow open during test or if eeprom is broken */
5418         if (test_bit(__I40E_TESTING, &pf->state) ||
5419             test_bit(__I40E_BAD_EEPROM, &pf->state))
5420                 return -EBUSY;
5421
5422         netif_carrier_off(netdev);
5423
5424         err = i40e_vsi_open(vsi);
5425         if (err)
5426                 return err;
5427
5428         /* configure global TSO hardware offload settings */
5429         wr32(&pf->hw, I40E_GLLAN_TSOMSK_F, be32_to_cpu(TCP_FLAG_PSH |
5430                                                        TCP_FLAG_FIN) >> 16);
5431         wr32(&pf->hw, I40E_GLLAN_TSOMSK_M, be32_to_cpu(TCP_FLAG_PSH |
5432                                                        TCP_FLAG_FIN |
5433                                                        TCP_FLAG_CWR) >> 16);
5434         wr32(&pf->hw, I40E_GLLAN_TSOMSK_L, be32_to_cpu(TCP_FLAG_CWR) >> 16);
5435
5436         udp_tunnel_get_rx_info(netdev);
5437
5438         return 0;
5439 }
5440
5441 /**
5442  * i40e_vsi_open -
5443  * @vsi: the VSI to open
5444  *
5445  * Finish initialization of the VSI.
5446  *
5447  * Returns 0 on success, negative value on failure
5448  **/
5449 int i40e_vsi_open(struct i40e_vsi *vsi)
5450 {
5451         struct i40e_pf *pf = vsi->back;
5452         char int_name[I40E_INT_NAME_STR_LEN];
5453         int err;
5454
5455         /* allocate descriptors */
5456         err = i40e_vsi_setup_tx_resources(vsi);
5457         if (err)
5458                 goto err_setup_tx;
5459         err = i40e_vsi_setup_rx_resources(vsi);
5460         if (err)
5461                 goto err_setup_rx;
5462
5463         err = i40e_vsi_configure(vsi);
5464         if (err)
5465                 goto err_setup_rx;
5466
5467         if (vsi->netdev) {
5468                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s",
5469                          dev_driver_string(&pf->pdev->dev), vsi->netdev->name);
5470                 err = i40e_vsi_request_irq(vsi, int_name);
5471                 if (err)
5472                         goto err_setup_rx;
5473
5474                 /* Notify the stack of the actual queue counts. */
5475                 err = netif_set_real_num_tx_queues(vsi->netdev,
5476                                                    vsi->num_queue_pairs);
5477                 if (err)
5478                         goto err_set_queues;
5479
5480                 err = netif_set_real_num_rx_queues(vsi->netdev,
5481                                                    vsi->num_queue_pairs);
5482                 if (err)
5483                         goto err_set_queues;
5484
5485         } else if (vsi->type == I40E_VSI_FDIR) {
5486                 snprintf(int_name, sizeof(int_name) - 1, "%s-%s:fdir",
5487                          dev_driver_string(&pf->pdev->dev),
5488                          dev_name(&pf->pdev->dev));
5489                 err = i40e_vsi_request_irq(vsi, int_name);
5490
5491         } else {
5492                 err = -EINVAL;
5493                 goto err_setup_rx;
5494         }
5495
5496         err = i40e_up_complete(vsi);
5497         if (err)
5498                 goto err_up_complete;
5499
5500         return 0;
5501
5502 err_up_complete:
5503         i40e_down(vsi);
5504 err_set_queues:
5505         i40e_vsi_free_irq(vsi);
5506 err_setup_rx:
5507         i40e_vsi_free_rx_resources(vsi);
5508 err_setup_tx:
5509         i40e_vsi_free_tx_resources(vsi);
5510         if (vsi == pf->vsi[pf->lan_vsi])
5511                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
5512
5513         return err;
5514 }
5515
5516 /**
5517  * i40e_fdir_filter_exit - Cleans up the Flow Director accounting
5518  * @pf: Pointer to PF
5519  *
5520  * This function destroys the hlist where all the Flow Director
5521  * filters were saved.
5522  **/
5523 static void i40e_fdir_filter_exit(struct i40e_pf *pf)
5524 {
5525         struct i40e_fdir_filter *filter;
5526         struct hlist_node *node2;
5527
5528         hlist_for_each_entry_safe(filter, node2,
5529                                   &pf->fdir_filter_list, fdir_node) {
5530                 hlist_del(&filter->fdir_node);
5531                 kfree(filter);
5532         }
5533         pf->fdir_pf_active_filters = 0;
5534 }
5535
5536 /**
5537  * i40e_close - Disables a network interface
5538  * @netdev: network interface device structure
5539  *
5540  * The close entry point is called when an interface is de-activated
5541  * by the OS.  The hardware is still under the driver's control, but
5542  * this netdev interface is disabled.
5543  *
5544  * Returns 0, this is not allowed to fail
5545  **/
5546 int i40e_close(struct net_device *netdev)
5547 {
5548         struct i40e_netdev_priv *np = netdev_priv(netdev);
5549         struct i40e_vsi *vsi = np->vsi;
5550
5551         i40e_vsi_close(vsi);
5552
5553         return 0;
5554 }
5555
5556 /**
5557  * i40e_do_reset - Start a PF or Core Reset sequence
5558  * @pf: board private structure
5559  * @reset_flags: which reset is requested
5560  *
5561  * The essential difference in resets is that the PF Reset
5562  * doesn't clear the packet buffers, doesn't reset the PE
5563  * firmware, and doesn't bother the other PFs on the chip.
5564  **/
5565 void i40e_do_reset(struct i40e_pf *pf, u32 reset_flags)
5566 {
5567         u32 val;
5568
5569         WARN_ON(in_interrupt());
5570
5571
5572         /* do the biggest reset indicated */
5573         if (reset_flags & BIT_ULL(__I40E_GLOBAL_RESET_REQUESTED)) {
5574
5575                 /* Request a Global Reset
5576                  *
5577                  * This will start the chip's countdown to the actual full
5578                  * chip reset event, and a warning interrupt to be sent
5579                  * to all PFs, including the requestor.  Our handler
5580                  * for the warning interrupt will deal with the shutdown
5581                  * and recovery of the switch setup.
5582                  */
5583                 dev_dbg(&pf->pdev->dev, "GlobalR requested\n");
5584                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5585                 val |= I40E_GLGEN_RTRIG_GLOBR_MASK;
5586                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5587
5588         } else if (reset_flags & BIT_ULL(__I40E_CORE_RESET_REQUESTED)) {
5589
5590                 /* Request a Core Reset
5591                  *
5592                  * Same as Global Reset, except does *not* include the MAC/PHY
5593                  */
5594                 dev_dbg(&pf->pdev->dev, "CoreR requested\n");
5595                 val = rd32(&pf->hw, I40E_GLGEN_RTRIG);
5596                 val |= I40E_GLGEN_RTRIG_CORER_MASK;
5597                 wr32(&pf->hw, I40E_GLGEN_RTRIG, val);
5598                 i40e_flush(&pf->hw);
5599
5600         } else if (reset_flags & BIT_ULL(__I40E_PF_RESET_REQUESTED)) {
5601
5602                 /* Request a PF Reset
5603                  *
5604                  * Resets only the PF-specific registers
5605                  *
5606                  * This goes directly to the tear-down and rebuild of
5607                  * the switch, since we need to do all the recovery as
5608                  * for the Core Reset.
5609                  */
5610                 dev_dbg(&pf->pdev->dev, "PFR requested\n");
5611                 i40e_handle_reset_warning(pf);
5612
5613         } else if (reset_flags & BIT_ULL(__I40E_REINIT_REQUESTED)) {
5614                 int v;
5615
5616                 /* Find the VSI(s) that requested a re-init */
5617                 dev_info(&pf->pdev->dev,
5618                          "VSI reinit requested\n");
5619                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5620                         struct i40e_vsi *vsi = pf->vsi[v];
5621
5622                         if (vsi != NULL &&
5623                             test_bit(__I40E_REINIT_REQUESTED, &vsi->state)) {
5624                                 i40e_vsi_reinit_locked(pf->vsi[v]);
5625                                 clear_bit(__I40E_REINIT_REQUESTED, &vsi->state);
5626                         }
5627                 }
5628         } else if (reset_flags & BIT_ULL(__I40E_DOWN_REQUESTED)) {
5629                 int v;
5630
5631                 /* Find the VSI(s) that needs to be brought down */
5632                 dev_info(&pf->pdev->dev, "VSI down requested\n");
5633                 for (v = 0; v < pf->num_alloc_vsi; v++) {
5634                         struct i40e_vsi *vsi = pf->vsi[v];
5635
5636                         if (vsi != NULL &&
5637                             test_bit(__I40E_DOWN_REQUESTED, &vsi->state)) {
5638                                 set_bit(__I40E_DOWN, &vsi->state);
5639                                 i40e_down(vsi);
5640                                 clear_bit(__I40E_DOWN_REQUESTED, &vsi->state);
5641                         }
5642                 }
5643         } else {
5644                 dev_info(&pf->pdev->dev,
5645                          "bad reset request 0x%08x\n", reset_flags);
5646         }
5647 }
5648
5649 #ifdef CONFIG_I40E_DCB
5650 /**
5651  * i40e_dcb_need_reconfig - Check if DCB needs reconfig
5652  * @pf: board private structure
5653  * @old_cfg: current DCB config
5654  * @new_cfg: new DCB config
5655  **/
5656 bool i40e_dcb_need_reconfig(struct i40e_pf *pf,
5657                             struct i40e_dcbx_config *old_cfg,
5658                             struct i40e_dcbx_config *new_cfg)
5659 {
5660         bool need_reconfig = false;
5661
5662         /* Check if ETS configuration has changed */
5663         if (memcmp(&new_cfg->etscfg,
5664                    &old_cfg->etscfg,
5665                    sizeof(new_cfg->etscfg))) {
5666                 /* If Priority Table has changed reconfig is needed */
5667                 if (memcmp(&new_cfg->etscfg.prioritytable,
5668                            &old_cfg->etscfg.prioritytable,
5669                            sizeof(new_cfg->etscfg.prioritytable))) {
5670                         need_reconfig = true;
5671                         dev_dbg(&pf->pdev->dev, "ETS UP2TC changed.\n");
5672                 }
5673
5674                 if (memcmp(&new_cfg->etscfg.tcbwtable,
5675                            &old_cfg->etscfg.tcbwtable,
5676                            sizeof(new_cfg->etscfg.tcbwtable)))
5677                         dev_dbg(&pf->pdev->dev, "ETS TC BW Table changed.\n");
5678
5679                 if (memcmp(&new_cfg->etscfg.tsatable,
5680                            &old_cfg->etscfg.tsatable,
5681                            sizeof(new_cfg->etscfg.tsatable)))
5682                         dev_dbg(&pf->pdev->dev, "ETS TSA Table changed.\n");
5683         }
5684
5685         /* Check if PFC configuration has changed */
5686         if (memcmp(&new_cfg->pfc,
5687                    &old_cfg->pfc,
5688                    sizeof(new_cfg->pfc))) {
5689                 need_reconfig = true;
5690                 dev_dbg(&pf->pdev->dev, "PFC config change detected.\n");
5691         }
5692
5693         /* Check if APP Table has changed */
5694         if (memcmp(&new_cfg->app,
5695                    &old_cfg->app,
5696                    sizeof(new_cfg->app))) {
5697                 need_reconfig = true;
5698                 dev_dbg(&pf->pdev->dev, "APP Table change detected.\n");
5699         }
5700
5701         dev_dbg(&pf->pdev->dev, "dcb need_reconfig=%d\n", need_reconfig);
5702         return need_reconfig;
5703 }
5704
5705 /**
5706  * i40e_handle_lldp_event - Handle LLDP Change MIB event
5707  * @pf: board private structure
5708  * @e: event info posted on ARQ
5709  **/
5710 static int i40e_handle_lldp_event(struct i40e_pf *pf,
5711                                   struct i40e_arq_event_info *e)
5712 {
5713         struct i40e_aqc_lldp_get_mib *mib =
5714                 (struct i40e_aqc_lldp_get_mib *)&e->desc.params.raw;
5715         struct i40e_hw *hw = &pf->hw;
5716         struct i40e_dcbx_config tmp_dcbx_cfg;
5717         bool need_reconfig = false;
5718         int ret = 0;
5719         u8 type;
5720
5721         /* Not DCB capable or capability disabled */
5722         if (!(pf->flags & I40E_FLAG_DCB_ENABLED))
5723                 return ret;
5724
5725         /* Ignore if event is not for Nearest Bridge */
5726         type = ((mib->type >> I40E_AQ_LLDP_BRIDGE_TYPE_SHIFT)
5727                 & I40E_AQ_LLDP_BRIDGE_TYPE_MASK);
5728         dev_dbg(&pf->pdev->dev, "LLDP event mib bridge type 0x%x\n", type);
5729         if (type != I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE)
5730                 return ret;
5731
5732         /* Check MIB Type and return if event for Remote MIB update */
5733         type = mib->type & I40E_AQ_LLDP_MIB_TYPE_MASK;
5734         dev_dbg(&pf->pdev->dev,
5735                 "LLDP event mib type %s\n", type ? "remote" : "local");
5736         if (type == I40E_AQ_LLDP_MIB_REMOTE) {
5737                 /* Update the remote cached instance and return */
5738                 ret = i40e_aq_get_dcb_config(hw, I40E_AQ_LLDP_MIB_REMOTE,
5739                                 I40E_AQ_LLDP_BRIDGE_TYPE_NEAREST_BRIDGE,
5740                                 &hw->remote_dcbx_config);
5741                 goto exit;
5742         }
5743
5744         /* Store the old configuration */
5745         tmp_dcbx_cfg = hw->local_dcbx_config;
5746
5747         /* Reset the old DCBx configuration data */
5748         memset(&hw->local_dcbx_config, 0, sizeof(hw->local_dcbx_config));
5749         /* Get updated DCBX data from firmware */
5750         ret = i40e_get_dcb_config(&pf->hw);
5751         if (ret) {
5752                 dev_info(&pf->pdev->dev,
5753                          "Failed querying DCB configuration data from firmware, err %s aq_err %s\n",
5754                          i40e_stat_str(&pf->hw, ret),
5755                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
5756                 goto exit;
5757         }
5758
5759         /* No change detected in DCBX configs */
5760         if (!memcmp(&tmp_dcbx_cfg, &hw->local_dcbx_config,
5761                     sizeof(tmp_dcbx_cfg))) {
5762                 dev_dbg(&pf->pdev->dev, "No change detected in DCBX configuration.\n");
5763                 goto exit;
5764         }
5765
5766         need_reconfig = i40e_dcb_need_reconfig(pf, &tmp_dcbx_cfg,
5767                                                &hw->local_dcbx_config);
5768
5769         i40e_dcbnl_flush_apps(pf, &tmp_dcbx_cfg, &hw->local_dcbx_config);
5770
5771         if (!need_reconfig)
5772                 goto exit;
5773
5774         /* Enable DCB tagging only when more than one TC */
5775         if (i40e_dcb_get_num_tc(&hw->local_dcbx_config) > 1)
5776                 pf->flags |= I40E_FLAG_DCB_ENABLED;
5777         else
5778                 pf->flags &= ~I40E_FLAG_DCB_ENABLED;
5779
5780         set_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5781         /* Reconfiguration needed quiesce all VSIs */
5782         i40e_pf_quiesce_all_vsi(pf);
5783
5784         /* Changes in configuration update VEB/VSI */
5785         i40e_dcb_reconfigure(pf);
5786
5787         ret = i40e_resume_port_tx(pf);
5788
5789         clear_bit(__I40E_PORT_TX_SUSPENDED, &pf->state);
5790         /* In case of error no point in resuming VSIs */
5791         if (ret)
5792                 goto exit;
5793
5794         /* Wait for the PF's queues to be disabled */
5795         ret = i40e_pf_wait_queues_disabled(pf);
5796         if (ret) {
5797                 /* Schedule PF reset to recover */
5798                 set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
5799                 i40e_service_event_schedule(pf);
5800         } else {
5801                 i40e_pf_unquiesce_all_vsi(pf);
5802                 /* Notify the client for the DCB changes */
5803                 i40e_notify_client_of_l2_param_changes(pf->vsi[pf->lan_vsi]);
5804         }
5805
5806 exit:
5807         return ret;
5808 }
5809 #endif /* CONFIG_I40E_DCB */
5810
5811 /**
5812  * i40e_do_reset_safe - Protected reset path for userland calls.
5813  * @pf: board private structure
5814  * @reset_flags: which reset is requested
5815  *
5816  **/
5817 void i40e_do_reset_safe(struct i40e_pf *pf, u32 reset_flags)
5818 {
5819         rtnl_lock();
5820         i40e_do_reset(pf, reset_flags);
5821         rtnl_unlock();
5822 }
5823
5824 /**
5825  * i40e_handle_lan_overflow_event - Handler for LAN queue overflow event
5826  * @pf: board private structure
5827  * @e: event info posted on ARQ
5828  *
5829  * Handler for LAN Queue Overflow Event generated by the firmware for PF
5830  * and VF queues
5831  **/
5832 static void i40e_handle_lan_overflow_event(struct i40e_pf *pf,
5833                                            struct i40e_arq_event_info *e)
5834 {
5835         struct i40e_aqc_lan_overflow *data =
5836                 (struct i40e_aqc_lan_overflow *)&e->desc.params.raw;
5837         u32 queue = le32_to_cpu(data->prtdcb_rupto);
5838         u32 qtx_ctl = le32_to_cpu(data->otx_ctl);
5839         struct i40e_hw *hw = &pf->hw;
5840         struct i40e_vf *vf;
5841         u16 vf_id;
5842
5843         dev_dbg(&pf->pdev->dev, "overflow Rx Queue Number = %d QTX_CTL=0x%08x\n",
5844                 queue, qtx_ctl);
5845
5846         /* Queue belongs to VF, find the VF and issue VF reset */
5847         if (((qtx_ctl & I40E_QTX_CTL_PFVF_Q_MASK)
5848             >> I40E_QTX_CTL_PFVF_Q_SHIFT) == I40E_QTX_CTL_VF_QUEUE) {
5849                 vf_id = (u16)((qtx_ctl & I40E_QTX_CTL_VFVM_INDX_MASK)
5850                          >> I40E_QTX_CTL_VFVM_INDX_SHIFT);
5851                 vf_id -= hw->func_caps.vf_base_id;
5852                 vf = &pf->vf[vf_id];
5853                 i40e_vc_notify_vf_reset(vf);
5854                 /* Allow VF to process pending reset notification */
5855                 msleep(20);
5856                 i40e_reset_vf(vf, false);
5857         }
5858 }
5859
5860 /**
5861  * i40e_service_event_complete - Finish up the service event
5862  * @pf: board private structure
5863  **/
5864 static void i40e_service_event_complete(struct i40e_pf *pf)
5865 {
5866         WARN_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
5867
5868         /* flush memory to make sure state is correct before next watchog */
5869         smp_mb__before_atomic();
5870         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
5871 }
5872
5873 /**
5874  * i40e_get_cur_guaranteed_fd_count - Get the consumed guaranteed FD filters
5875  * @pf: board private structure
5876  **/
5877 u32 i40e_get_cur_guaranteed_fd_count(struct i40e_pf *pf)
5878 {
5879         u32 val, fcnt_prog;
5880
5881         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5882         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK);
5883         return fcnt_prog;
5884 }
5885
5886 /**
5887  * i40e_get_current_fd_count - Get total FD filters programmed for this PF
5888  * @pf: board private structure
5889  **/
5890 u32 i40e_get_current_fd_count(struct i40e_pf *pf)
5891 {
5892         u32 val, fcnt_prog;
5893
5894         val = rd32(&pf->hw, I40E_PFQF_FDSTAT);
5895         fcnt_prog = (val & I40E_PFQF_FDSTAT_GUARANT_CNT_MASK) +
5896                     ((val & I40E_PFQF_FDSTAT_BEST_CNT_MASK) >>
5897                       I40E_PFQF_FDSTAT_BEST_CNT_SHIFT);
5898         return fcnt_prog;
5899 }
5900
5901 /**
5902  * i40e_get_global_fd_count - Get total FD filters programmed on device
5903  * @pf: board private structure
5904  **/
5905 u32 i40e_get_global_fd_count(struct i40e_pf *pf)
5906 {
5907         u32 val, fcnt_prog;
5908
5909         val = rd32(&pf->hw, I40E_GLQF_FDCNT_0);
5910         fcnt_prog = (val & I40E_GLQF_FDCNT_0_GUARANT_CNT_MASK) +
5911                     ((val & I40E_GLQF_FDCNT_0_BESTCNT_MASK) >>
5912                      I40E_GLQF_FDCNT_0_BESTCNT_SHIFT);
5913         return fcnt_prog;
5914 }
5915
5916 /**
5917  * i40e_fdir_check_and_reenable - Function to reenabe FD ATR or SB if disabled
5918  * @pf: board private structure
5919  **/
5920 void i40e_fdir_check_and_reenable(struct i40e_pf *pf)
5921 {
5922         struct i40e_fdir_filter *filter;
5923         u32 fcnt_prog, fcnt_avail;
5924         struct hlist_node *node;
5925
5926         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
5927                 return;
5928
5929         /* Check if, FD SB or ATR was auto disabled and if there is enough room
5930          * to re-enable
5931          */
5932         fcnt_prog = i40e_get_global_fd_count(pf);
5933         fcnt_avail = pf->fdir_pf_filter_count;
5934         if ((fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM)) ||
5935             (pf->fd_add_err == 0) ||
5936             (i40e_get_current_atr_cnt(pf) < pf->fd_atr_cnt)) {
5937                 if ((pf->flags & I40E_FLAG_FD_SB_ENABLED) &&
5938                     (pf->auto_disable_flags & I40E_FLAG_FD_SB_ENABLED)) {
5939                         pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
5940                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5941                                 dev_info(&pf->pdev->dev, "FD Sideband/ntuple is being enabled since we have space in the table now\n");
5942                 }
5943         }
5944         /* Wait for some more space to be available to turn on ATR */
5945         if (fcnt_prog < (fcnt_avail - I40E_FDIR_BUFFER_HEAD_ROOM * 2)) {
5946                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
5947                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED)) {
5948                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
5949                         if (I40E_DEBUG_FD & pf->hw.debug_mask)
5950                                 dev_info(&pf->pdev->dev, "ATR is being enabled since we have space in the table now\n");
5951                 }
5952         }
5953
5954         /* if hw had a problem adding a filter, delete it */
5955         if (pf->fd_inv > 0) {
5956                 hlist_for_each_entry_safe(filter, node,
5957                                           &pf->fdir_filter_list, fdir_node) {
5958                         if (filter->fd_id == pf->fd_inv) {
5959                                 hlist_del(&filter->fdir_node);
5960                                 kfree(filter);
5961                                 pf->fdir_pf_active_filters--;
5962                         }
5963                 }
5964         }
5965 }
5966
5967 #define I40E_MIN_FD_FLUSH_INTERVAL 10
5968 #define I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE 30
5969 /**
5970  * i40e_fdir_flush_and_replay - Function to flush all FD filters and replay SB
5971  * @pf: board private structure
5972  **/
5973 static void i40e_fdir_flush_and_replay(struct i40e_pf *pf)
5974 {
5975         unsigned long min_flush_time;
5976         int flush_wait_retry = 50;
5977         bool disable_atr = false;
5978         int fd_room;
5979         int reg;
5980
5981         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
5982                 return;
5983
5984         if (!time_after(jiffies, pf->fd_flush_timestamp +
5985                                  (I40E_MIN_FD_FLUSH_INTERVAL * HZ)))
5986                 return;
5987
5988         /* If the flush is happening too quick and we have mostly SB rules we
5989          * should not re-enable ATR for some time.
5990          */
5991         min_flush_time = pf->fd_flush_timestamp +
5992                          (I40E_MIN_FD_FLUSH_SB_ATR_UNSTABLE * HZ);
5993         fd_room = pf->fdir_pf_filter_count - pf->fdir_pf_active_filters;
5994
5995         if (!(time_after(jiffies, min_flush_time)) &&
5996             (fd_room < I40E_FDIR_BUFFER_HEAD_ROOM_FOR_ATR)) {
5997                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
5998                         dev_info(&pf->pdev->dev, "ATR disabled, not enough FD filter space.\n");
5999                 disable_atr = true;
6000         }
6001
6002         pf->fd_flush_timestamp = jiffies;
6003         pf->flags &= ~I40E_FLAG_FD_ATR_ENABLED;
6004         /* flush all filters */
6005         wr32(&pf->hw, I40E_PFQF_CTL_1,
6006              I40E_PFQF_CTL_1_CLEARFDTABLE_MASK);
6007         i40e_flush(&pf->hw);
6008         pf->fd_flush_cnt++;
6009         pf->fd_add_err = 0;
6010         do {
6011                 /* Check FD flush status every 5-6msec */
6012                 usleep_range(5000, 6000);
6013                 reg = rd32(&pf->hw, I40E_PFQF_CTL_1);
6014                 if (!(reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK))
6015                         break;
6016         } while (flush_wait_retry--);
6017         if (reg & I40E_PFQF_CTL_1_CLEARFDTABLE_MASK) {
6018                 dev_warn(&pf->pdev->dev, "FD table did not flush, needs more time\n");
6019         } else {
6020                 /* replay sideband filters */
6021                 i40e_fdir_filter_restore(pf->vsi[pf->lan_vsi]);
6022                 if (!disable_atr)
6023                         pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
6024                 clear_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state);
6025                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
6026                         dev_info(&pf->pdev->dev, "FD Filter table flushed and FD-SB replayed.\n");
6027         }
6028 }
6029
6030 /**
6031  * i40e_get_current_atr_count - Get the count of total FD ATR filters programmed
6032  * @pf: board private structure
6033  **/
6034 u32 i40e_get_current_atr_cnt(struct i40e_pf *pf)
6035 {
6036         return i40e_get_current_fd_count(pf) - pf->fdir_pf_active_filters;
6037 }
6038
6039 /* We can see up to 256 filter programming desc in transit if the filters are
6040  * being applied really fast; before we see the first
6041  * filter miss error on Rx queue 0. Accumulating enough error messages before
6042  * reacting will make sure we don't cause flush too often.
6043  */
6044 #define I40E_MAX_FD_PROGRAM_ERROR 256
6045
6046 /**
6047  * i40e_fdir_reinit_subtask - Worker thread to reinit FDIR filter table
6048  * @pf: board private structure
6049  **/
6050 static void i40e_fdir_reinit_subtask(struct i40e_pf *pf)
6051 {
6052
6053         /* if interface is down do nothing */
6054         if (test_bit(__I40E_DOWN, &pf->state))
6055                 return;
6056
6057         if (!(pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED)))
6058                 return;
6059
6060         if (test_bit(__I40E_FD_FLUSH_REQUESTED, &pf->state))
6061                 i40e_fdir_flush_and_replay(pf);
6062
6063         i40e_fdir_check_and_reenable(pf);
6064
6065 }
6066
6067 /**
6068  * i40e_vsi_link_event - notify VSI of a link event
6069  * @vsi: vsi to be notified
6070  * @link_up: link up or down
6071  **/
6072 static void i40e_vsi_link_event(struct i40e_vsi *vsi, bool link_up)
6073 {
6074         if (!vsi || test_bit(__I40E_DOWN, &vsi->state))
6075                 return;
6076
6077         switch (vsi->type) {
6078         case I40E_VSI_MAIN:
6079 #ifdef I40E_FCOE
6080         case I40E_VSI_FCOE:
6081 #endif
6082                 if (!vsi->netdev || !vsi->netdev_registered)
6083                         break;
6084
6085                 if (link_up) {
6086                         netif_carrier_on(vsi->netdev);
6087                         netif_tx_wake_all_queues(vsi->netdev);
6088                 } else {
6089                         netif_carrier_off(vsi->netdev);
6090                         netif_tx_stop_all_queues(vsi->netdev);
6091                 }
6092                 break;
6093
6094         case I40E_VSI_SRIOV:
6095         case I40E_VSI_VMDQ2:
6096         case I40E_VSI_CTRL:
6097         case I40E_VSI_IWARP:
6098         case I40E_VSI_MIRROR:
6099         default:
6100                 /* there is no notification for other VSIs */
6101                 break;
6102         }
6103 }
6104
6105 /**
6106  * i40e_veb_link_event - notify elements on the veb of a link event
6107  * @veb: veb to be notified
6108  * @link_up: link up or down
6109  **/
6110 static void i40e_veb_link_event(struct i40e_veb *veb, bool link_up)
6111 {
6112         struct i40e_pf *pf;
6113         int i;
6114
6115         if (!veb || !veb->pf)
6116                 return;
6117         pf = veb->pf;
6118
6119         /* depth first... */
6120         for (i = 0; i < I40E_MAX_VEB; i++)
6121                 if (pf->veb[i] && (pf->veb[i]->uplink_seid == veb->seid))
6122                         i40e_veb_link_event(pf->veb[i], link_up);
6123
6124         /* ... now the local VSIs */
6125         for (i = 0; i < pf->num_alloc_vsi; i++)
6126                 if (pf->vsi[i] && (pf->vsi[i]->uplink_seid == veb->seid))
6127                         i40e_vsi_link_event(pf->vsi[i], link_up);
6128 }
6129
6130 /**
6131  * i40e_link_event - Update netif_carrier status
6132  * @pf: board private structure
6133  **/
6134 static void i40e_link_event(struct i40e_pf *pf)
6135 {
6136         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6137         u8 new_link_speed, old_link_speed;
6138         i40e_status status;
6139         bool new_link, old_link;
6140
6141         /* save off old link status information */
6142         pf->hw.phy.link_info_old = pf->hw.phy.link_info;
6143
6144         /* set this to force the get_link_status call to refresh state */
6145         pf->hw.phy.get_link_info = true;
6146
6147         old_link = (pf->hw.phy.link_info_old.link_info & I40E_AQ_LINK_UP);
6148
6149         status = i40e_get_link_status(&pf->hw, &new_link);
6150         if (status) {
6151                 dev_dbg(&pf->pdev->dev, "couldn't get link state, status: %d\n",
6152                         status);
6153                 return;
6154         }
6155
6156         old_link_speed = pf->hw.phy.link_info_old.link_speed;
6157         new_link_speed = pf->hw.phy.link_info.link_speed;
6158
6159         if (new_link == old_link &&
6160             new_link_speed == old_link_speed &&
6161             (test_bit(__I40E_DOWN, &vsi->state) ||
6162              new_link == netif_carrier_ok(vsi->netdev)))
6163                 return;
6164
6165         if (!test_bit(__I40E_DOWN, &vsi->state))
6166                 i40e_print_link_message(vsi, new_link);
6167
6168         /* Notify the base of the switch tree connected to
6169          * the link.  Floating VEBs are not notified.
6170          */
6171         if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
6172                 i40e_veb_link_event(pf->veb[pf->lan_veb], new_link);
6173         else
6174                 i40e_vsi_link_event(vsi, new_link);
6175
6176         if (pf->vf)
6177                 i40e_vc_notify_link_state(pf);
6178
6179         if (pf->flags & I40E_FLAG_PTP)
6180                 i40e_ptp_set_increment(pf);
6181 }
6182
6183 /**
6184  * i40e_watchdog_subtask - periodic checks not using event driven response
6185  * @pf: board private structure
6186  **/
6187 static void i40e_watchdog_subtask(struct i40e_pf *pf)
6188 {
6189         int i;
6190
6191         /* if interface is down do nothing */
6192         if (test_bit(__I40E_DOWN, &pf->state) ||
6193             test_bit(__I40E_CONFIG_BUSY, &pf->state))
6194                 return;
6195
6196         /* make sure we don't do these things too often */
6197         if (time_before(jiffies, (pf->service_timer_previous +
6198                                   pf->service_timer_period)))
6199                 return;
6200         pf->service_timer_previous = jiffies;
6201
6202         if (pf->flags & I40E_FLAG_LINK_POLLING_ENABLED)
6203                 i40e_link_event(pf);
6204
6205         /* Update the stats for active netdevs so the network stack
6206          * can look at updated numbers whenever it cares to
6207          */
6208         for (i = 0; i < pf->num_alloc_vsi; i++)
6209                 if (pf->vsi[i] && pf->vsi[i]->netdev)
6210                         i40e_update_stats(pf->vsi[i]);
6211
6212         if (pf->flags & I40E_FLAG_VEB_STATS_ENABLED) {
6213                 /* Update the stats for the active switching components */
6214                 for (i = 0; i < I40E_MAX_VEB; i++)
6215                         if (pf->veb[i])
6216                                 i40e_update_veb_stats(pf->veb[i]);
6217         }
6218
6219         i40e_ptp_rx_hang(pf->vsi[pf->lan_vsi]);
6220 }
6221
6222 /**
6223  * i40e_reset_subtask - Set up for resetting the device and driver
6224  * @pf: board private structure
6225  **/
6226 static void i40e_reset_subtask(struct i40e_pf *pf)
6227 {
6228         u32 reset_flags = 0;
6229
6230         rtnl_lock();
6231         if (test_bit(__I40E_REINIT_REQUESTED, &pf->state)) {
6232                 reset_flags |= BIT(__I40E_REINIT_REQUESTED);
6233                 clear_bit(__I40E_REINIT_REQUESTED, &pf->state);
6234         }
6235         if (test_bit(__I40E_PF_RESET_REQUESTED, &pf->state)) {
6236                 reset_flags |= BIT(__I40E_PF_RESET_REQUESTED);
6237                 clear_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
6238         }
6239         if (test_bit(__I40E_CORE_RESET_REQUESTED, &pf->state)) {
6240                 reset_flags |= BIT(__I40E_CORE_RESET_REQUESTED);
6241                 clear_bit(__I40E_CORE_RESET_REQUESTED, &pf->state);
6242         }
6243         if (test_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state)) {
6244                 reset_flags |= BIT(__I40E_GLOBAL_RESET_REQUESTED);
6245                 clear_bit(__I40E_GLOBAL_RESET_REQUESTED, &pf->state);
6246         }
6247         if (test_bit(__I40E_DOWN_REQUESTED, &pf->state)) {
6248                 reset_flags |= BIT(__I40E_DOWN_REQUESTED);
6249                 clear_bit(__I40E_DOWN_REQUESTED, &pf->state);
6250         }
6251
6252         /* If there's a recovery already waiting, it takes
6253          * precedence before starting a new reset sequence.
6254          */
6255         if (test_bit(__I40E_RESET_INTR_RECEIVED, &pf->state)) {
6256                 i40e_handle_reset_warning(pf);
6257                 goto unlock;
6258         }
6259
6260         /* If we're already down or resetting, just bail */
6261         if (reset_flags &&
6262             !test_bit(__I40E_DOWN, &pf->state) &&
6263             !test_bit(__I40E_CONFIG_BUSY, &pf->state))
6264                 i40e_do_reset(pf, reset_flags);
6265
6266 unlock:
6267         rtnl_unlock();
6268 }
6269
6270 /**
6271  * i40e_handle_link_event - Handle link event
6272  * @pf: board private structure
6273  * @e: event info posted on ARQ
6274  **/
6275 static void i40e_handle_link_event(struct i40e_pf *pf,
6276                                    struct i40e_arq_event_info *e)
6277 {
6278         struct i40e_aqc_get_link_status *status =
6279                 (struct i40e_aqc_get_link_status *)&e->desc.params.raw;
6280
6281         /* Do a new status request to re-enable LSE reporting
6282          * and load new status information into the hw struct
6283          * This completely ignores any state information
6284          * in the ARQ event info, instead choosing to always
6285          * issue the AQ update link status command.
6286          */
6287         i40e_link_event(pf);
6288
6289         /* check for unqualified module, if link is down */
6290         if ((status->link_info & I40E_AQ_MEDIA_AVAILABLE) &&
6291             (!(status->an_info & I40E_AQ_QUALIFIED_MODULE)) &&
6292             (!(status->link_info & I40E_AQ_LINK_UP)))
6293                 dev_err(&pf->pdev->dev,
6294                         "The driver failed to link because an unqualified module was detected.\n");
6295 }
6296
6297 /**
6298  * i40e_clean_adminq_subtask - Clean the AdminQ rings
6299  * @pf: board private structure
6300  **/
6301 static void i40e_clean_adminq_subtask(struct i40e_pf *pf)
6302 {
6303         struct i40e_arq_event_info event;
6304         struct i40e_hw *hw = &pf->hw;
6305         u16 pending, i = 0;
6306         i40e_status ret;
6307         u16 opcode;
6308         u32 oldval;
6309         u32 val;
6310
6311         /* Do not run clean AQ when PF reset fails */
6312         if (test_bit(__I40E_RESET_FAILED, &pf->state))
6313                 return;
6314
6315         /* check for error indications */
6316         val = rd32(&pf->hw, pf->hw.aq.arq.len);
6317         oldval = val;
6318         if (val & I40E_PF_ARQLEN_ARQVFE_MASK) {
6319                 if (hw->debug_mask & I40E_DEBUG_AQ)
6320                         dev_info(&pf->pdev->dev, "ARQ VF Error detected\n");
6321                 val &= ~I40E_PF_ARQLEN_ARQVFE_MASK;
6322         }
6323         if (val & I40E_PF_ARQLEN_ARQOVFL_MASK) {
6324                 if (hw->debug_mask & I40E_DEBUG_AQ)
6325                         dev_info(&pf->pdev->dev, "ARQ Overflow Error detected\n");
6326                 val &= ~I40E_PF_ARQLEN_ARQOVFL_MASK;
6327                 pf->arq_overflows++;
6328         }
6329         if (val & I40E_PF_ARQLEN_ARQCRIT_MASK) {
6330                 if (hw->debug_mask & I40E_DEBUG_AQ)
6331                         dev_info(&pf->pdev->dev, "ARQ Critical Error detected\n");
6332                 val &= ~I40E_PF_ARQLEN_ARQCRIT_MASK;
6333         }
6334         if (oldval != val)
6335                 wr32(&pf->hw, pf->hw.aq.arq.len, val);
6336
6337         val = rd32(&pf->hw, pf->hw.aq.asq.len);
6338         oldval = val;
6339         if (val & I40E_PF_ATQLEN_ATQVFE_MASK) {
6340                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6341                         dev_info(&pf->pdev->dev, "ASQ VF Error detected\n");
6342                 val &= ~I40E_PF_ATQLEN_ATQVFE_MASK;
6343         }
6344         if (val & I40E_PF_ATQLEN_ATQOVFL_MASK) {
6345                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6346                         dev_info(&pf->pdev->dev, "ASQ Overflow Error detected\n");
6347                 val &= ~I40E_PF_ATQLEN_ATQOVFL_MASK;
6348         }
6349         if (val & I40E_PF_ATQLEN_ATQCRIT_MASK) {
6350                 if (pf->hw.debug_mask & I40E_DEBUG_AQ)
6351                         dev_info(&pf->pdev->dev, "ASQ Critical Error detected\n");
6352                 val &= ~I40E_PF_ATQLEN_ATQCRIT_MASK;
6353         }
6354         if (oldval != val)
6355                 wr32(&pf->hw, pf->hw.aq.asq.len, val);
6356
6357         event.buf_len = I40E_MAX_AQ_BUF_SIZE;
6358         event.msg_buf = kzalloc(event.buf_len, GFP_KERNEL);
6359         if (!event.msg_buf)
6360                 return;
6361
6362         do {
6363                 ret = i40e_clean_arq_element(hw, &event, &pending);
6364                 if (ret == I40E_ERR_ADMIN_QUEUE_NO_WORK)
6365                         break;
6366                 else if (ret) {
6367                         dev_info(&pf->pdev->dev, "ARQ event error %d\n", ret);
6368                         break;
6369                 }
6370
6371                 opcode = le16_to_cpu(event.desc.opcode);
6372                 switch (opcode) {
6373
6374                 case i40e_aqc_opc_get_link_status:
6375                         i40e_handle_link_event(pf, &event);
6376                         break;
6377                 case i40e_aqc_opc_send_msg_to_pf:
6378                         ret = i40e_vc_process_vf_msg(pf,
6379                                         le16_to_cpu(event.desc.retval),
6380                                         le32_to_cpu(event.desc.cookie_high),
6381                                         le32_to_cpu(event.desc.cookie_low),
6382                                         event.msg_buf,
6383                                         event.msg_len);
6384                         break;
6385                 case i40e_aqc_opc_lldp_update_mib:
6386                         dev_dbg(&pf->pdev->dev, "ARQ: Update LLDP MIB event received\n");
6387 #ifdef CONFIG_I40E_DCB
6388                         rtnl_lock();
6389                         ret = i40e_handle_lldp_event(pf, &event);
6390                         rtnl_unlock();
6391 #endif /* CONFIG_I40E_DCB */
6392                         break;
6393                 case i40e_aqc_opc_event_lan_overflow:
6394                         dev_dbg(&pf->pdev->dev, "ARQ LAN queue overflow event received\n");
6395                         i40e_handle_lan_overflow_event(pf, &event);
6396                         break;
6397                 case i40e_aqc_opc_send_msg_to_peer:
6398                         dev_info(&pf->pdev->dev, "ARQ: Msg from other pf\n");
6399                         break;
6400                 case i40e_aqc_opc_nvm_erase:
6401                 case i40e_aqc_opc_nvm_update:
6402                 case i40e_aqc_opc_oem_post_update:
6403                         i40e_debug(&pf->hw, I40E_DEBUG_NVM,
6404                                    "ARQ NVM operation 0x%04x completed\n",
6405                                    opcode);
6406                         break;
6407                 default:
6408                         dev_info(&pf->pdev->dev,
6409                                  "ARQ: Unknown event 0x%04x ignored\n",
6410                                  opcode);
6411                         break;
6412                 }
6413         } while (pending && (i++ < pf->adminq_work_limit));
6414
6415         clear_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state);
6416         /* re-enable Admin queue interrupt cause */
6417         val = rd32(hw, I40E_PFINT_ICR0_ENA);
6418         val |=  I40E_PFINT_ICR0_ENA_ADMINQ_MASK;
6419         wr32(hw, I40E_PFINT_ICR0_ENA, val);
6420         i40e_flush(hw);
6421
6422         kfree(event.msg_buf);
6423 }
6424
6425 /**
6426  * i40e_verify_eeprom - make sure eeprom is good to use
6427  * @pf: board private structure
6428  **/
6429 static void i40e_verify_eeprom(struct i40e_pf *pf)
6430 {
6431         int err;
6432
6433         err = i40e_diag_eeprom_test(&pf->hw);
6434         if (err) {
6435                 /* retry in case of garbage read */
6436                 err = i40e_diag_eeprom_test(&pf->hw);
6437                 if (err) {
6438                         dev_info(&pf->pdev->dev, "eeprom check failed (%d), Tx/Rx traffic disabled\n",
6439                                  err);
6440                         set_bit(__I40E_BAD_EEPROM, &pf->state);
6441                 }
6442         }
6443
6444         if (!err && test_bit(__I40E_BAD_EEPROM, &pf->state)) {
6445                 dev_info(&pf->pdev->dev, "eeprom check passed, Tx/Rx traffic enabled\n");
6446                 clear_bit(__I40E_BAD_EEPROM, &pf->state);
6447         }
6448 }
6449
6450 /**
6451  * i40e_enable_pf_switch_lb
6452  * @pf: pointer to the PF structure
6453  *
6454  * enable switch loop back or die - no point in a return value
6455  **/
6456 static void i40e_enable_pf_switch_lb(struct i40e_pf *pf)
6457 {
6458         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6459         struct i40e_vsi_context ctxt;
6460         int ret;
6461
6462         ctxt.seid = pf->main_vsi_seid;
6463         ctxt.pf_num = pf->hw.pf_id;
6464         ctxt.vf_num = 0;
6465         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6466         if (ret) {
6467                 dev_info(&pf->pdev->dev,
6468                          "couldn't get PF vsi config, err %s aq_err %s\n",
6469                          i40e_stat_str(&pf->hw, ret),
6470                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6471                 return;
6472         }
6473         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6474         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6475         ctxt.info.switch_id |= cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6476
6477         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6478         if (ret) {
6479                 dev_info(&pf->pdev->dev,
6480                          "update vsi switch failed, err %s aq_err %s\n",
6481                          i40e_stat_str(&pf->hw, ret),
6482                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6483         }
6484 }
6485
6486 /**
6487  * i40e_disable_pf_switch_lb
6488  * @pf: pointer to the PF structure
6489  *
6490  * disable switch loop back or die - no point in a return value
6491  **/
6492 static void i40e_disable_pf_switch_lb(struct i40e_pf *pf)
6493 {
6494         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
6495         struct i40e_vsi_context ctxt;
6496         int ret;
6497
6498         ctxt.seid = pf->main_vsi_seid;
6499         ctxt.pf_num = pf->hw.pf_id;
6500         ctxt.vf_num = 0;
6501         ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
6502         if (ret) {
6503                 dev_info(&pf->pdev->dev,
6504                          "couldn't get PF vsi config, err %s aq_err %s\n",
6505                          i40e_stat_str(&pf->hw, ret),
6506                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6507                 return;
6508         }
6509         ctxt.flags = I40E_AQ_VSI_TYPE_PF;
6510         ctxt.info.valid_sections = cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
6511         ctxt.info.switch_id &= ~cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
6512
6513         ret = i40e_aq_update_vsi_params(&vsi->back->hw, &ctxt, NULL);
6514         if (ret) {
6515                 dev_info(&pf->pdev->dev,
6516                          "update vsi switch failed, err %s aq_err %s\n",
6517                          i40e_stat_str(&pf->hw, ret),
6518                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6519         }
6520 }
6521
6522 /**
6523  * i40e_config_bridge_mode - Configure the HW bridge mode
6524  * @veb: pointer to the bridge instance
6525  *
6526  * Configure the loop back mode for the LAN VSI that is downlink to the
6527  * specified HW bridge instance. It is expected this function is called
6528  * when a new HW bridge is instantiated.
6529  **/
6530 static void i40e_config_bridge_mode(struct i40e_veb *veb)
6531 {
6532         struct i40e_pf *pf = veb->pf;
6533
6534         if (pf->hw.debug_mask & I40E_DEBUG_LAN)
6535                 dev_info(&pf->pdev->dev, "enabling bridge mode: %s\n",
6536                          veb->bridge_mode == BRIDGE_MODE_VEPA ? "VEPA" : "VEB");
6537         if (veb->bridge_mode & BRIDGE_MODE_VEPA)
6538                 i40e_disable_pf_switch_lb(pf);
6539         else
6540                 i40e_enable_pf_switch_lb(pf);
6541 }
6542
6543 /**
6544  * i40e_reconstitute_veb - rebuild the VEB and anything connected to it
6545  * @veb: pointer to the VEB instance
6546  *
6547  * This is a recursive function that first builds the attached VSIs then
6548  * recurses in to build the next layer of VEB.  We track the connections
6549  * through our own index numbers because the seid's from the HW could
6550  * change across the reset.
6551  **/
6552 static int i40e_reconstitute_veb(struct i40e_veb *veb)
6553 {
6554         struct i40e_vsi *ctl_vsi = NULL;
6555         struct i40e_pf *pf = veb->pf;
6556         int v, veb_idx;
6557         int ret;
6558
6559         /* build VSI that owns this VEB, temporarily attached to base VEB */
6560         for (v = 0; v < pf->num_alloc_vsi && !ctl_vsi; v++) {
6561                 if (pf->vsi[v] &&
6562                     pf->vsi[v]->veb_idx == veb->idx &&
6563                     pf->vsi[v]->flags & I40E_VSI_FLAG_VEB_OWNER) {
6564                         ctl_vsi = pf->vsi[v];
6565                         break;
6566                 }
6567         }
6568         if (!ctl_vsi) {
6569                 dev_info(&pf->pdev->dev,
6570                          "missing owner VSI for veb_idx %d\n", veb->idx);
6571                 ret = -ENOENT;
6572                 goto end_reconstitute;
6573         }
6574         if (ctl_vsi != pf->vsi[pf->lan_vsi])
6575                 ctl_vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
6576         ret = i40e_add_vsi(ctl_vsi);
6577         if (ret) {
6578                 dev_info(&pf->pdev->dev,
6579                          "rebuild of veb_idx %d owner VSI failed: %d\n",
6580                          veb->idx, ret);
6581                 goto end_reconstitute;
6582         }
6583         i40e_vsi_reset_stats(ctl_vsi);
6584
6585         /* create the VEB in the switch and move the VSI onto the VEB */
6586         ret = i40e_add_veb(veb, ctl_vsi);
6587         if (ret)
6588                 goto end_reconstitute;
6589
6590         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
6591                 veb->bridge_mode = BRIDGE_MODE_VEB;
6592         else
6593                 veb->bridge_mode = BRIDGE_MODE_VEPA;
6594         i40e_config_bridge_mode(veb);
6595
6596         /* create the remaining VSIs attached to this VEB */
6597         for (v = 0; v < pf->num_alloc_vsi; v++) {
6598                 if (!pf->vsi[v] || pf->vsi[v] == ctl_vsi)
6599                         continue;
6600
6601                 if (pf->vsi[v]->veb_idx == veb->idx) {
6602                         struct i40e_vsi *vsi = pf->vsi[v];
6603
6604                         vsi->uplink_seid = veb->seid;
6605                         ret = i40e_add_vsi(vsi);
6606                         if (ret) {
6607                                 dev_info(&pf->pdev->dev,
6608                                          "rebuild of vsi_idx %d failed: %d\n",
6609                                          v, ret);
6610                                 goto end_reconstitute;
6611                         }
6612                         i40e_vsi_reset_stats(vsi);
6613                 }
6614         }
6615
6616         /* create any VEBs attached to this VEB - RECURSION */
6617         for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
6618                 if (pf->veb[veb_idx] && pf->veb[veb_idx]->veb_idx == veb->idx) {
6619                         pf->veb[veb_idx]->uplink_seid = veb->seid;
6620                         ret = i40e_reconstitute_veb(pf->veb[veb_idx]);
6621                         if (ret)
6622                                 break;
6623                 }
6624         }
6625
6626 end_reconstitute:
6627         return ret;
6628 }
6629
6630 /**
6631  * i40e_get_capabilities - get info about the HW
6632  * @pf: the PF struct
6633  **/
6634 static int i40e_get_capabilities(struct i40e_pf *pf)
6635 {
6636         struct i40e_aqc_list_capabilities_element_resp *cap_buf;
6637         u16 data_size;
6638         int buf_len;
6639         int err;
6640
6641         buf_len = 40 * sizeof(struct i40e_aqc_list_capabilities_element_resp);
6642         do {
6643                 cap_buf = kzalloc(buf_len, GFP_KERNEL);
6644                 if (!cap_buf)
6645                         return -ENOMEM;
6646
6647                 /* this loads the data into the hw struct for us */
6648                 err = i40e_aq_discover_capabilities(&pf->hw, cap_buf, buf_len,
6649                                             &data_size,
6650                                             i40e_aqc_opc_list_func_capabilities,
6651                                             NULL);
6652                 /* data loaded, buffer no longer needed */
6653                 kfree(cap_buf);
6654
6655                 if (pf->hw.aq.asq_last_status == I40E_AQ_RC_ENOMEM) {
6656                         /* retry with a larger buffer */
6657                         buf_len = data_size;
6658                 } else if (pf->hw.aq.asq_last_status != I40E_AQ_RC_OK) {
6659                         dev_info(&pf->pdev->dev,
6660                                  "capability discovery failed, err %s aq_err %s\n",
6661                                  i40e_stat_str(&pf->hw, err),
6662                                  i40e_aq_str(&pf->hw,
6663                                              pf->hw.aq.asq_last_status));
6664                         return -ENODEV;
6665                 }
6666         } while (err);
6667
6668         if (pf->hw.debug_mask & I40E_DEBUG_USER)
6669                 dev_info(&pf->pdev->dev,
6670                          "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",
6671                          pf->hw.pf_id, pf->hw.func_caps.num_vfs,
6672                          pf->hw.func_caps.num_msix_vectors,
6673                          pf->hw.func_caps.num_msix_vectors_vf,
6674                          pf->hw.func_caps.fd_filters_guaranteed,
6675                          pf->hw.func_caps.fd_filters_best_effort,
6676                          pf->hw.func_caps.num_tx_qp,
6677                          pf->hw.func_caps.num_vsis);
6678
6679 #define DEF_NUM_VSI (1 + (pf->hw.func_caps.fcoe ? 1 : 0) \
6680                        + pf->hw.func_caps.num_vfs)
6681         if (pf->hw.revision_id == 0 && (DEF_NUM_VSI > pf->hw.func_caps.num_vsis)) {
6682                 dev_info(&pf->pdev->dev,
6683                          "got num_vsis %d, setting num_vsis to %d\n",
6684                          pf->hw.func_caps.num_vsis, DEF_NUM_VSI);
6685                 pf->hw.func_caps.num_vsis = DEF_NUM_VSI;
6686         }
6687
6688         return 0;
6689 }
6690
6691 static int i40e_vsi_clear(struct i40e_vsi *vsi);
6692
6693 /**
6694  * i40e_fdir_sb_setup - initialize the Flow Director resources for Sideband
6695  * @pf: board private structure
6696  **/
6697 static void i40e_fdir_sb_setup(struct i40e_pf *pf)
6698 {
6699         struct i40e_vsi *vsi;
6700         int i;
6701
6702         /* quick workaround for an NVM issue that leaves a critical register
6703          * uninitialized
6704          */
6705         if (!rd32(&pf->hw, I40E_GLQF_HKEY(0))) {
6706                 static const u32 hkey[] = {
6707                         0xe640d33f, 0xcdfe98ab, 0x73fa7161, 0x0d7a7d36,
6708                         0xeacb7d61, 0xaa4f05b6, 0x9c5c89ed, 0xfc425ddb,
6709                         0xa4654832, 0xfc7461d4, 0x8f827619, 0xf5c63c21,
6710                         0x95b3a76d};
6711
6712                 for (i = 0; i <= I40E_GLQF_HKEY_MAX_INDEX; i++)
6713                         wr32(&pf->hw, I40E_GLQF_HKEY(i), hkey[i]);
6714         }
6715
6716         if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
6717                 return;
6718
6719         /* find existing VSI and see if it needs configuring */
6720         vsi = NULL;
6721         for (i = 0; i < pf->num_alloc_vsi; i++) {
6722                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6723                         vsi = pf->vsi[i];
6724                         break;
6725                 }
6726         }
6727
6728         /* create a new VSI if none exists */
6729         if (!vsi) {
6730                 vsi = i40e_vsi_setup(pf, I40E_VSI_FDIR,
6731                                      pf->vsi[pf->lan_vsi]->seid, 0);
6732                 if (!vsi) {
6733                         dev_info(&pf->pdev->dev, "Couldn't create FDir VSI\n");
6734                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
6735                         return;
6736                 }
6737         }
6738
6739         i40e_vsi_setup_irqhandler(vsi, i40e_fdir_clean_ring);
6740 }
6741
6742 /**
6743  * i40e_fdir_teardown - release the Flow Director resources
6744  * @pf: board private structure
6745  **/
6746 static void i40e_fdir_teardown(struct i40e_pf *pf)
6747 {
6748         int i;
6749
6750         i40e_fdir_filter_exit(pf);
6751         for (i = 0; i < pf->num_alloc_vsi; i++) {
6752                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
6753                         i40e_vsi_release(pf->vsi[i]);
6754                         break;
6755                 }
6756         }
6757 }
6758
6759 /**
6760  * i40e_prep_for_reset - prep for the core to reset
6761  * @pf: board private structure
6762  *
6763  * Close up the VFs and other things in prep for PF Reset.
6764   **/
6765 static void i40e_prep_for_reset(struct i40e_pf *pf)
6766 {
6767         struct i40e_hw *hw = &pf->hw;
6768         i40e_status ret = 0;
6769         u32 v;
6770
6771         clear_bit(__I40E_RESET_INTR_RECEIVED, &pf->state);
6772         if (test_and_set_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state))
6773                 return;
6774         if (i40e_check_asq_alive(&pf->hw))
6775                 i40e_vc_notify_reset(pf);
6776
6777         dev_dbg(&pf->pdev->dev, "Tearing down internal switch for reset\n");
6778
6779         /* quiesce the VSIs and their queues that are not already DOWN */
6780         i40e_pf_quiesce_all_vsi(pf);
6781
6782         for (v = 0; v < pf->num_alloc_vsi; v++) {
6783                 if (pf->vsi[v])
6784                         pf->vsi[v]->seid = 0;
6785         }
6786
6787         i40e_shutdown_adminq(&pf->hw);
6788
6789         /* call shutdown HMC */
6790         if (hw->hmc.hmc_obj) {
6791                 ret = i40e_shutdown_lan_hmc(hw);
6792                 if (ret)
6793                         dev_warn(&pf->pdev->dev,
6794                                  "shutdown_lan_hmc failed: %d\n", ret);
6795         }
6796 }
6797
6798 /**
6799  * i40e_send_version - update firmware with driver version
6800  * @pf: PF struct
6801  */
6802 static void i40e_send_version(struct i40e_pf *pf)
6803 {
6804         struct i40e_driver_version dv;
6805
6806         dv.major_version = DRV_VERSION_MAJOR;
6807         dv.minor_version = DRV_VERSION_MINOR;
6808         dv.build_version = DRV_VERSION_BUILD;
6809         dv.subbuild_version = 0;
6810         strlcpy(dv.driver_string, DRV_VERSION, sizeof(dv.driver_string));
6811         i40e_aq_send_driver_version(&pf->hw, &dv, NULL);
6812 }
6813
6814 /**
6815  * i40e_reset_and_rebuild - reset and rebuild using a saved config
6816  * @pf: board private structure
6817  * @reinit: if the Main VSI needs to re-initialized.
6818  **/
6819 static void i40e_reset_and_rebuild(struct i40e_pf *pf, bool reinit)
6820 {
6821         struct i40e_hw *hw = &pf->hw;
6822         u8 set_fc_aq_fail = 0;
6823         i40e_status ret;
6824         u32 val;
6825         u32 v;
6826
6827         /* Now we wait for GRST to settle out.
6828          * We don't have to delete the VEBs or VSIs from the hw switch
6829          * because the reset will make them disappear.
6830          */
6831         ret = i40e_pf_reset(hw);
6832         if (ret) {
6833                 dev_info(&pf->pdev->dev, "PF reset failed, %d\n", ret);
6834                 set_bit(__I40E_RESET_FAILED, &pf->state);
6835                 goto clear_recovery;
6836         }
6837         pf->pfr_count++;
6838
6839         if (test_bit(__I40E_DOWN, &pf->state))
6840                 goto clear_recovery;
6841         dev_dbg(&pf->pdev->dev, "Rebuilding internal switch\n");
6842
6843         /* rebuild the basics for the AdminQ, HMC, and initial HW switch */
6844         ret = i40e_init_adminq(&pf->hw);
6845         if (ret) {
6846                 dev_info(&pf->pdev->dev, "Rebuild AdminQ failed, err %s aq_err %s\n",
6847                          i40e_stat_str(&pf->hw, ret),
6848                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6849                 goto clear_recovery;
6850         }
6851
6852         /* re-verify the eeprom if we just had an EMP reset */
6853         if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, &pf->state))
6854                 i40e_verify_eeprom(pf);
6855
6856         i40e_clear_pxe_mode(hw);
6857         ret = i40e_get_capabilities(pf);
6858         if (ret)
6859                 goto end_core_reset;
6860
6861         ret = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
6862                                 hw->func_caps.num_rx_qp,
6863                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
6864         if (ret) {
6865                 dev_info(&pf->pdev->dev, "init_lan_hmc failed: %d\n", ret);
6866                 goto end_core_reset;
6867         }
6868         ret = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
6869         if (ret) {
6870                 dev_info(&pf->pdev->dev, "configure_lan_hmc failed: %d\n", ret);
6871                 goto end_core_reset;
6872         }
6873
6874 #ifdef CONFIG_I40E_DCB
6875         ret = i40e_init_pf_dcb(pf);
6876         if (ret) {
6877                 dev_info(&pf->pdev->dev, "DCB init failed %d, disabled\n", ret);
6878                 pf->flags &= ~I40E_FLAG_DCB_CAPABLE;
6879                 /* Continue without DCB enabled */
6880         }
6881 #endif /* CONFIG_I40E_DCB */
6882 #ifdef I40E_FCOE
6883         i40e_init_pf_fcoe(pf);
6884
6885 #endif
6886         /* do basic switch setup */
6887         ret = i40e_setup_pf_switch(pf, reinit);
6888         if (ret)
6889                 goto end_core_reset;
6890
6891         /* The driver only wants link up/down and module qualification
6892          * reports from firmware.  Note the negative logic.
6893          */
6894         ret = i40e_aq_set_phy_int_mask(&pf->hw,
6895                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
6896                                          I40E_AQ_EVENT_MEDIA_NA |
6897                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
6898         if (ret)
6899                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
6900                          i40e_stat_str(&pf->hw, ret),
6901                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6902
6903         /* make sure our flow control settings are restored */
6904         ret = i40e_set_fc(&pf->hw, &set_fc_aq_fail, true);
6905         if (ret)
6906                 dev_dbg(&pf->pdev->dev, "setting flow control: ret = %s last_status = %s\n",
6907                         i40e_stat_str(&pf->hw, ret),
6908                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
6909
6910         /* Rebuild the VSIs and VEBs that existed before reset.
6911          * They are still in our local switch element arrays, so only
6912          * need to rebuild the switch model in the HW.
6913          *
6914          * If there were VEBs but the reconstitution failed, we'll try
6915          * try to recover minimal use by getting the basic PF VSI working.
6916          */
6917         if (pf->vsi[pf->lan_vsi]->uplink_seid != pf->mac_seid) {
6918                 dev_dbg(&pf->pdev->dev, "attempting to rebuild switch\n");
6919                 /* find the one VEB connected to the MAC, and find orphans */
6920                 for (v = 0; v < I40E_MAX_VEB; v++) {
6921                         if (!pf->veb[v])
6922                                 continue;
6923
6924                         if (pf->veb[v]->uplink_seid == pf->mac_seid ||
6925                             pf->veb[v]->uplink_seid == 0) {
6926                                 ret = i40e_reconstitute_veb(pf->veb[v]);
6927
6928                                 if (!ret)
6929                                         continue;
6930
6931                                 /* If Main VEB failed, we're in deep doodoo,
6932                                  * so give up rebuilding the switch and set up
6933                                  * for minimal rebuild of PF VSI.
6934                                  * If orphan failed, we'll report the error
6935                                  * but try to keep going.
6936                                  */
6937                                 if (pf->veb[v]->uplink_seid == pf->mac_seid) {
6938                                         dev_info(&pf->pdev->dev,
6939                                                  "rebuild of switch failed: %d, will try to set up simple PF connection\n",
6940                                                  ret);
6941                                         pf->vsi[pf->lan_vsi]->uplink_seid
6942                                                                 = pf->mac_seid;
6943                                         break;
6944                                 } else if (pf->veb[v]->uplink_seid == 0) {
6945                                         dev_info(&pf->pdev->dev,
6946                                                  "rebuild of orphan VEB failed: %d\n",
6947                                                  ret);
6948                                 }
6949                         }
6950                 }
6951         }
6952
6953         if (pf->vsi[pf->lan_vsi]->uplink_seid == pf->mac_seid) {
6954                 dev_dbg(&pf->pdev->dev, "attempting to rebuild PF VSI\n");
6955                 /* no VEB, so rebuild only the Main VSI */
6956                 ret = i40e_add_vsi(pf->vsi[pf->lan_vsi]);
6957                 if (ret) {
6958                         dev_info(&pf->pdev->dev,
6959                                  "rebuild of Main VSI failed: %d\n", ret);
6960                         goto end_core_reset;
6961                 }
6962         }
6963
6964         /* Reconfigure hardware for allowing smaller MSS in the case
6965          * of TSO, so that we avoid the MDD being fired and causing
6966          * a reset in the case of small MSS+TSO.
6967          */
6968 #define I40E_REG_MSS          0x000E64DC
6969 #define I40E_REG_MSS_MIN_MASK 0x3FF0000
6970 #define I40E_64BYTE_MSS       0x400000
6971         val = rd32(hw, I40E_REG_MSS);
6972         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
6973                 val &= ~I40E_REG_MSS_MIN_MASK;
6974                 val |= I40E_64BYTE_MSS;
6975                 wr32(hw, I40E_REG_MSS, val);
6976         }
6977
6978         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
6979                 msleep(75);
6980                 ret = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
6981                 if (ret)
6982                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
6983                                  i40e_stat_str(&pf->hw, ret),
6984                                  i40e_aq_str(&pf->hw,
6985                                              pf->hw.aq.asq_last_status));
6986         }
6987         /* reinit the misc interrupt */
6988         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
6989                 ret = i40e_setup_misc_vector(pf);
6990
6991         /* Add a filter to drop all Flow control frames from any VSI from being
6992          * transmitted. By doing so we stop a malicious VF from sending out
6993          * PAUSE or PFC frames and potentially controlling traffic for other
6994          * PF/VF VSIs.
6995          * The FW can still send Flow control frames if enabled.
6996          */
6997         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
6998                                                        pf->main_vsi_seid);
6999
7000         /* restart the VSIs that were rebuilt and running before the reset */
7001         i40e_pf_unquiesce_all_vsi(pf);
7002
7003         if (pf->num_alloc_vfs) {
7004                 for (v = 0; v < pf->num_alloc_vfs; v++)
7005                         i40e_reset_vf(&pf->vf[v], true);
7006         }
7007
7008         /* tell the firmware that we're starting */
7009         i40e_send_version(pf);
7010
7011 end_core_reset:
7012         clear_bit(__I40E_RESET_FAILED, &pf->state);
7013 clear_recovery:
7014         clear_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state);
7015 }
7016
7017 /**
7018  * i40e_handle_reset_warning - prep for the PF to reset, reset and rebuild
7019  * @pf: board private structure
7020  *
7021  * Close up the VFs and other things in prep for a Core Reset,
7022  * then get ready to rebuild the world.
7023  **/
7024 static void i40e_handle_reset_warning(struct i40e_pf *pf)
7025 {
7026         i40e_prep_for_reset(pf);
7027         i40e_reset_and_rebuild(pf, false);
7028 }
7029
7030 /**
7031  * i40e_handle_mdd_event
7032  * @pf: pointer to the PF structure
7033  *
7034  * Called from the MDD irq handler to identify possibly malicious vfs
7035  **/
7036 static void i40e_handle_mdd_event(struct i40e_pf *pf)
7037 {
7038         struct i40e_hw *hw = &pf->hw;
7039         bool mdd_detected = false;
7040         bool pf_mdd_detected = false;
7041         struct i40e_vf *vf;
7042         u32 reg;
7043         int i;
7044
7045         if (!test_bit(__I40E_MDD_EVENT_PENDING, &pf->state))
7046                 return;
7047
7048         /* find what triggered the MDD event */
7049         reg = rd32(hw, I40E_GL_MDET_TX);
7050         if (reg & I40E_GL_MDET_TX_VALID_MASK) {
7051                 u8 pf_num = (reg & I40E_GL_MDET_TX_PF_NUM_MASK) >>
7052                                 I40E_GL_MDET_TX_PF_NUM_SHIFT;
7053                 u16 vf_num = (reg & I40E_GL_MDET_TX_VF_NUM_MASK) >>
7054                                 I40E_GL_MDET_TX_VF_NUM_SHIFT;
7055                 u8 event = (reg & I40E_GL_MDET_TX_EVENT_MASK) >>
7056                                 I40E_GL_MDET_TX_EVENT_SHIFT;
7057                 u16 queue = ((reg & I40E_GL_MDET_TX_QUEUE_MASK) >>
7058                                 I40E_GL_MDET_TX_QUEUE_SHIFT) -
7059                                 pf->hw.func_caps.base_queue;
7060                 if (netif_msg_tx_err(pf))
7061                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on TX queue %d PF number 0x%02x VF number 0x%02x\n",
7062                                  event, queue, pf_num, vf_num);
7063                 wr32(hw, I40E_GL_MDET_TX, 0xffffffff);
7064                 mdd_detected = true;
7065         }
7066         reg = rd32(hw, I40E_GL_MDET_RX);
7067         if (reg & I40E_GL_MDET_RX_VALID_MASK) {
7068                 u8 func = (reg & I40E_GL_MDET_RX_FUNCTION_MASK) >>
7069                                 I40E_GL_MDET_RX_FUNCTION_SHIFT;
7070                 u8 event = (reg & I40E_GL_MDET_RX_EVENT_MASK) >>
7071                                 I40E_GL_MDET_RX_EVENT_SHIFT;
7072                 u16 queue = ((reg & I40E_GL_MDET_RX_QUEUE_MASK) >>
7073                                 I40E_GL_MDET_RX_QUEUE_SHIFT) -
7074                                 pf->hw.func_caps.base_queue;
7075                 if (netif_msg_rx_err(pf))
7076                         dev_info(&pf->pdev->dev, "Malicious Driver Detection event 0x%02x on RX queue %d of function 0x%02x\n",
7077                                  event, queue, func);
7078                 wr32(hw, I40E_GL_MDET_RX, 0xffffffff);
7079                 mdd_detected = true;
7080         }
7081
7082         if (mdd_detected) {
7083                 reg = rd32(hw, I40E_PF_MDET_TX);
7084                 if (reg & I40E_PF_MDET_TX_VALID_MASK) {
7085                         wr32(hw, I40E_PF_MDET_TX, 0xFFFF);
7086                         dev_info(&pf->pdev->dev, "TX driver issue detected, PF reset issued\n");
7087                         pf_mdd_detected = true;
7088                 }
7089                 reg = rd32(hw, I40E_PF_MDET_RX);
7090                 if (reg & I40E_PF_MDET_RX_VALID_MASK) {
7091                         wr32(hw, I40E_PF_MDET_RX, 0xFFFF);
7092                         dev_info(&pf->pdev->dev, "RX driver issue detected, PF reset issued\n");
7093                         pf_mdd_detected = true;
7094                 }
7095                 /* Queue belongs to the PF, initiate a reset */
7096                 if (pf_mdd_detected) {
7097                         set_bit(__I40E_PF_RESET_REQUESTED, &pf->state);
7098                         i40e_service_event_schedule(pf);
7099                 }
7100         }
7101
7102         /* see if one of the VFs needs its hand slapped */
7103         for (i = 0; i < pf->num_alloc_vfs && mdd_detected; i++) {
7104                 vf = &(pf->vf[i]);
7105                 reg = rd32(hw, I40E_VP_MDET_TX(i));
7106                 if (reg & I40E_VP_MDET_TX_VALID_MASK) {
7107                         wr32(hw, I40E_VP_MDET_TX(i), 0xFFFF);
7108                         vf->num_mdd_events++;
7109                         dev_info(&pf->pdev->dev, "TX driver issue detected on VF %d\n",
7110                                  i);
7111                 }
7112
7113                 reg = rd32(hw, I40E_VP_MDET_RX(i));
7114                 if (reg & I40E_VP_MDET_RX_VALID_MASK) {
7115                         wr32(hw, I40E_VP_MDET_RX(i), 0xFFFF);
7116                         vf->num_mdd_events++;
7117                         dev_info(&pf->pdev->dev, "RX driver issue detected on VF %d\n",
7118                                  i);
7119                 }
7120
7121                 if (vf->num_mdd_events > I40E_DEFAULT_NUM_MDD_EVENTS_ALLOWED) {
7122                         dev_info(&pf->pdev->dev,
7123                                  "Too many MDD events on VF %d, disabled\n", i);
7124                         dev_info(&pf->pdev->dev,
7125                                  "Use PF Control I/F to re-enable the VF\n");
7126                         set_bit(I40E_VF_STAT_DISABLED, &vf->vf_states);
7127                 }
7128         }
7129
7130         /* re-enable mdd interrupt cause */
7131         clear_bit(__I40E_MDD_EVENT_PENDING, &pf->state);
7132         reg = rd32(hw, I40E_PFINT_ICR0_ENA);
7133         reg |=  I40E_PFINT_ICR0_ENA_MAL_DETECT_MASK;
7134         wr32(hw, I40E_PFINT_ICR0_ENA, reg);
7135         i40e_flush(hw);
7136 }
7137
7138 /**
7139  * i40e_sync_udp_filters_subtask - Sync the VSI filter list with HW
7140  * @pf: board private structure
7141  **/
7142 static void i40e_sync_udp_filters_subtask(struct i40e_pf *pf)
7143 {
7144         struct i40e_hw *hw = &pf->hw;
7145         i40e_status ret;
7146         __be16 port;
7147         int i;
7148
7149         if (!(pf->flags & I40E_FLAG_UDP_FILTER_SYNC))
7150                 return;
7151
7152         pf->flags &= ~I40E_FLAG_UDP_FILTER_SYNC;
7153
7154         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
7155                 if (pf->pending_udp_bitmap & BIT_ULL(i)) {
7156                         pf->pending_udp_bitmap &= ~BIT_ULL(i);
7157                         port = pf->udp_ports[i].index;
7158                         if (port)
7159                                 ret = i40e_aq_add_udp_tunnel(hw, ntohs(port),
7160                                                      pf->udp_ports[i].type,
7161                                                      NULL, NULL);
7162                         else
7163                                 ret = i40e_aq_del_udp_tunnel(hw, i, NULL);
7164
7165                         if (ret) {
7166                                 dev_dbg(&pf->pdev->dev,
7167                                         "%s %s port %d, index %d failed, err %s aq_err %s\n",
7168                                         pf->udp_ports[i].type ? "vxlan" : "geneve",
7169                                         port ? "add" : "delete",
7170                                         ntohs(port), i,
7171                                         i40e_stat_str(&pf->hw, ret),
7172                                         i40e_aq_str(&pf->hw,
7173                                                     pf->hw.aq.asq_last_status));
7174                                 pf->udp_ports[i].index = 0;
7175                         }
7176                 }
7177         }
7178 }
7179
7180 /**
7181  * i40e_service_task - Run the driver's async subtasks
7182  * @work: pointer to work_struct containing our data
7183  **/
7184 static void i40e_service_task(struct work_struct *work)
7185 {
7186         struct i40e_pf *pf = container_of(work,
7187                                           struct i40e_pf,
7188                                           service_task);
7189         unsigned long start_time = jiffies;
7190
7191         /* don't bother with service tasks if a reset is in progress */
7192         if (test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7193                 i40e_service_event_complete(pf);
7194                 return;
7195         }
7196
7197         i40e_detect_recover_hung(pf);
7198         i40e_sync_filters_subtask(pf);
7199         i40e_reset_subtask(pf);
7200         i40e_handle_mdd_event(pf);
7201         i40e_vc_process_vflr_event(pf);
7202         i40e_watchdog_subtask(pf);
7203         i40e_fdir_reinit_subtask(pf);
7204         i40e_client_subtask(pf);
7205         i40e_sync_filters_subtask(pf);
7206         i40e_sync_udp_filters_subtask(pf);
7207         i40e_clean_adminq_subtask(pf);
7208
7209         i40e_service_event_complete(pf);
7210
7211         /* If the tasks have taken longer than one timer cycle or there
7212          * is more work to be done, reschedule the service task now
7213          * rather than wait for the timer to tick again.
7214          */
7215         if (time_after(jiffies, (start_time + pf->service_timer_period)) ||
7216             test_bit(__I40E_ADMINQ_EVENT_PENDING, &pf->state)            ||
7217             test_bit(__I40E_MDD_EVENT_PENDING, &pf->state)               ||
7218             test_bit(__I40E_VFLR_EVENT_PENDING, &pf->state))
7219                 i40e_service_event_schedule(pf);
7220 }
7221
7222 /**
7223  * i40e_service_timer - timer callback
7224  * @data: pointer to PF struct
7225  **/
7226 static void i40e_service_timer(unsigned long data)
7227 {
7228         struct i40e_pf *pf = (struct i40e_pf *)data;
7229
7230         mod_timer(&pf->service_timer,
7231                   round_jiffies(jiffies + pf->service_timer_period));
7232         i40e_service_event_schedule(pf);
7233 }
7234
7235 /**
7236  * i40e_set_num_rings_in_vsi - Determine number of rings in the VSI
7237  * @vsi: the VSI being configured
7238  **/
7239 static int i40e_set_num_rings_in_vsi(struct i40e_vsi *vsi)
7240 {
7241         struct i40e_pf *pf = vsi->back;
7242
7243         switch (vsi->type) {
7244         case I40E_VSI_MAIN:
7245                 vsi->alloc_queue_pairs = pf->num_lan_qps;
7246                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7247                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7248                 if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7249                         vsi->num_q_vectors = pf->num_lan_msix;
7250                 else
7251                         vsi->num_q_vectors = 1;
7252
7253                 break;
7254
7255         case I40E_VSI_FDIR:
7256                 vsi->alloc_queue_pairs = 1;
7257                 vsi->num_desc = ALIGN(I40E_FDIR_RING_COUNT,
7258                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7259                 vsi->num_q_vectors = pf->num_fdsb_msix;
7260                 break;
7261
7262         case I40E_VSI_VMDQ2:
7263                 vsi->alloc_queue_pairs = pf->num_vmdq_qps;
7264                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7265                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7266                 vsi->num_q_vectors = pf->num_vmdq_msix;
7267                 break;
7268
7269         case I40E_VSI_SRIOV:
7270                 vsi->alloc_queue_pairs = pf->num_vf_qps;
7271                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7272                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7273                 break;
7274
7275 #ifdef I40E_FCOE
7276         case I40E_VSI_FCOE:
7277                 vsi->alloc_queue_pairs = pf->num_fcoe_qps;
7278                 vsi->num_desc = ALIGN(I40E_DEFAULT_NUM_DESCRIPTORS,
7279                                       I40E_REQ_DESCRIPTOR_MULTIPLE);
7280                 vsi->num_q_vectors = pf->num_fcoe_msix;
7281                 break;
7282
7283 #endif /* I40E_FCOE */
7284         default:
7285                 WARN_ON(1);
7286                 return -ENODATA;
7287         }
7288
7289         return 0;
7290 }
7291
7292 /**
7293  * i40e_vsi_alloc_arrays - Allocate queue and vector pointer arrays for the vsi
7294  * @type: VSI pointer
7295  * @alloc_qvectors: a bool to specify if q_vectors need to be allocated.
7296  *
7297  * On error: returns error code (negative)
7298  * On success: returns 0
7299  **/
7300 static int i40e_vsi_alloc_arrays(struct i40e_vsi *vsi, bool alloc_qvectors)
7301 {
7302         int size;
7303         int ret = 0;
7304
7305         /* allocate memory for both Tx and Rx ring pointers */
7306         size = sizeof(struct i40e_ring *) * vsi->alloc_queue_pairs * 2;
7307         vsi->tx_rings = kzalloc(size, GFP_KERNEL);
7308         if (!vsi->tx_rings)
7309                 return -ENOMEM;
7310         vsi->rx_rings = &vsi->tx_rings[vsi->alloc_queue_pairs];
7311
7312         if (alloc_qvectors) {
7313                 /* allocate memory for q_vector pointers */
7314                 size = sizeof(struct i40e_q_vector *) * vsi->num_q_vectors;
7315                 vsi->q_vectors = kzalloc(size, GFP_KERNEL);
7316                 if (!vsi->q_vectors) {
7317                         ret = -ENOMEM;
7318                         goto err_vectors;
7319                 }
7320         }
7321         return ret;
7322
7323 err_vectors:
7324         kfree(vsi->tx_rings);
7325         return ret;
7326 }
7327
7328 /**
7329  * i40e_vsi_mem_alloc - Allocates the next available struct vsi in the PF
7330  * @pf: board private structure
7331  * @type: type of VSI
7332  *
7333  * On error: returns error code (negative)
7334  * On success: returns vsi index in PF (positive)
7335  **/
7336 static int i40e_vsi_mem_alloc(struct i40e_pf *pf, enum i40e_vsi_type type)
7337 {
7338         int ret = -ENODEV;
7339         struct i40e_vsi *vsi;
7340         int vsi_idx;
7341         int i;
7342
7343         /* Need to protect the allocation of the VSIs at the PF level */
7344         mutex_lock(&pf->switch_mutex);
7345
7346         /* VSI list may be fragmented if VSI creation/destruction has
7347          * been happening.  We can afford to do a quick scan to look
7348          * for any free VSIs in the list.
7349          *
7350          * find next empty vsi slot, looping back around if necessary
7351          */
7352         i = pf->next_vsi;
7353         while (i < pf->num_alloc_vsi && pf->vsi[i])
7354                 i++;
7355         if (i >= pf->num_alloc_vsi) {
7356                 i = 0;
7357                 while (i < pf->next_vsi && pf->vsi[i])
7358                         i++;
7359         }
7360
7361         if (i < pf->num_alloc_vsi && !pf->vsi[i]) {
7362                 vsi_idx = i;             /* Found one! */
7363         } else {
7364                 ret = -ENODEV;
7365                 goto unlock_pf;  /* out of VSI slots! */
7366         }
7367         pf->next_vsi = ++i;
7368
7369         vsi = kzalloc(sizeof(*vsi), GFP_KERNEL);
7370         if (!vsi) {
7371                 ret = -ENOMEM;
7372                 goto unlock_pf;
7373         }
7374         vsi->type = type;
7375         vsi->back = pf;
7376         set_bit(__I40E_DOWN, &vsi->state);
7377         vsi->flags = 0;
7378         vsi->idx = vsi_idx;
7379         vsi->int_rate_limit = 0;
7380         vsi->rss_table_size = (vsi->type == I40E_VSI_MAIN) ?
7381                                 pf->rss_table_size : 64;
7382         vsi->netdev_registered = false;
7383         vsi->work_limit = I40E_DEFAULT_IRQ_WORK;
7384         INIT_LIST_HEAD(&vsi->mac_filter_list);
7385         vsi->irqs_ready = false;
7386
7387         ret = i40e_set_num_rings_in_vsi(vsi);
7388         if (ret)
7389                 goto err_rings;
7390
7391         ret = i40e_vsi_alloc_arrays(vsi, true);
7392         if (ret)
7393                 goto err_rings;
7394
7395         /* Setup default MSIX irq handler for VSI */
7396         i40e_vsi_setup_irqhandler(vsi, i40e_msix_clean_rings);
7397
7398         /* Initialize VSI lock */
7399         spin_lock_init(&vsi->mac_filter_list_lock);
7400         pf->vsi[vsi_idx] = vsi;
7401         ret = vsi_idx;
7402         goto unlock_pf;
7403
7404 err_rings:
7405         pf->next_vsi = i - 1;
7406         kfree(vsi);
7407 unlock_pf:
7408         mutex_unlock(&pf->switch_mutex);
7409         return ret;
7410 }
7411
7412 /**
7413  * i40e_vsi_free_arrays - Free queue and vector pointer arrays for the VSI
7414  * @type: VSI pointer
7415  * @free_qvectors: a bool to specify if q_vectors need to be freed.
7416  *
7417  * On error: returns error code (negative)
7418  * On success: returns 0
7419  **/
7420 static void i40e_vsi_free_arrays(struct i40e_vsi *vsi, bool free_qvectors)
7421 {
7422         /* free the ring and vector containers */
7423         if (free_qvectors) {
7424                 kfree(vsi->q_vectors);
7425                 vsi->q_vectors = NULL;
7426         }
7427         kfree(vsi->tx_rings);
7428         vsi->tx_rings = NULL;
7429         vsi->rx_rings = NULL;
7430 }
7431
7432 /**
7433  * i40e_clear_rss_config_user - clear the user configured RSS hash keys
7434  * and lookup table
7435  * @vsi: Pointer to VSI structure
7436  */
7437 static void i40e_clear_rss_config_user(struct i40e_vsi *vsi)
7438 {
7439         if (!vsi)
7440                 return;
7441
7442         kfree(vsi->rss_hkey_user);
7443         vsi->rss_hkey_user = NULL;
7444
7445         kfree(vsi->rss_lut_user);
7446         vsi->rss_lut_user = NULL;
7447 }
7448
7449 /**
7450  * i40e_vsi_clear - Deallocate the VSI provided
7451  * @vsi: the VSI being un-configured
7452  **/
7453 static int i40e_vsi_clear(struct i40e_vsi *vsi)
7454 {
7455         struct i40e_pf *pf;
7456
7457         if (!vsi)
7458                 return 0;
7459
7460         if (!vsi->back)
7461                 goto free_vsi;
7462         pf = vsi->back;
7463
7464         mutex_lock(&pf->switch_mutex);
7465         if (!pf->vsi[vsi->idx]) {
7466                 dev_err(&pf->pdev->dev, "pf->vsi[%d] is NULL, just free vsi[%d](%p,type %d)\n",
7467                         vsi->idx, vsi->idx, vsi, vsi->type);
7468                 goto unlock_vsi;
7469         }
7470
7471         if (pf->vsi[vsi->idx] != vsi) {
7472                 dev_err(&pf->pdev->dev,
7473                         "pf->vsi[%d](%p, type %d) != vsi[%d](%p,type %d): no free!\n",
7474                         pf->vsi[vsi->idx]->idx,
7475                         pf->vsi[vsi->idx],
7476                         pf->vsi[vsi->idx]->type,
7477                         vsi->idx, vsi, vsi->type);
7478                 goto unlock_vsi;
7479         }
7480
7481         /* updates the PF for this cleared vsi */
7482         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
7483         i40e_put_lump(pf->irq_pile, vsi->base_vector, vsi->idx);
7484
7485         i40e_vsi_free_arrays(vsi, true);
7486         i40e_clear_rss_config_user(vsi);
7487
7488         pf->vsi[vsi->idx] = NULL;
7489         if (vsi->idx < pf->next_vsi)
7490                 pf->next_vsi = vsi->idx;
7491
7492 unlock_vsi:
7493         mutex_unlock(&pf->switch_mutex);
7494 free_vsi:
7495         kfree(vsi);
7496
7497         return 0;
7498 }
7499
7500 /**
7501  * i40e_vsi_clear_rings - Deallocates the Rx and Tx rings for the provided VSI
7502  * @vsi: the VSI being cleaned
7503  **/
7504 static void i40e_vsi_clear_rings(struct i40e_vsi *vsi)
7505 {
7506         int i;
7507
7508         if (vsi->tx_rings && vsi->tx_rings[0]) {
7509                 for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7510                         kfree_rcu(vsi->tx_rings[i], rcu);
7511                         vsi->tx_rings[i] = NULL;
7512                         vsi->rx_rings[i] = NULL;
7513                 }
7514         }
7515 }
7516
7517 /**
7518  * i40e_alloc_rings - Allocates the Rx and Tx rings for the provided VSI
7519  * @vsi: the VSI being configured
7520  **/
7521 static int i40e_alloc_rings(struct i40e_vsi *vsi)
7522 {
7523         struct i40e_ring *tx_ring, *rx_ring;
7524         struct i40e_pf *pf = vsi->back;
7525         int i;
7526
7527         /* Set basic values in the rings to be used later during open() */
7528         for (i = 0; i < vsi->alloc_queue_pairs; i++) {
7529                 /* allocate space for both Tx and Rx in one shot */
7530                 tx_ring = kzalloc(sizeof(struct i40e_ring) * 2, GFP_KERNEL);
7531                 if (!tx_ring)
7532                         goto err_out;
7533
7534                 tx_ring->queue_index = i;
7535                 tx_ring->reg_idx = vsi->base_queue + i;
7536                 tx_ring->ring_active = false;
7537                 tx_ring->vsi = vsi;
7538                 tx_ring->netdev = vsi->netdev;
7539                 tx_ring->dev = &pf->pdev->dev;
7540                 tx_ring->count = vsi->num_desc;
7541                 tx_ring->size = 0;
7542                 tx_ring->dcb_tc = 0;
7543                 if (vsi->back->flags & I40E_FLAG_WB_ON_ITR_CAPABLE)
7544                         tx_ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
7545                 tx_ring->tx_itr_setting = pf->tx_itr_default;
7546                 vsi->tx_rings[i] = tx_ring;
7547
7548                 rx_ring = &tx_ring[1];
7549                 rx_ring->queue_index = i;
7550                 rx_ring->reg_idx = vsi->base_queue + i;
7551                 rx_ring->ring_active = false;
7552                 rx_ring->vsi = vsi;
7553                 rx_ring->netdev = vsi->netdev;
7554                 rx_ring->dev = &pf->pdev->dev;
7555                 rx_ring->count = vsi->num_desc;
7556                 rx_ring->size = 0;
7557                 rx_ring->dcb_tc = 0;
7558                 rx_ring->rx_itr_setting = pf->rx_itr_default;
7559                 vsi->rx_rings[i] = rx_ring;
7560         }
7561
7562         return 0;
7563
7564 err_out:
7565         i40e_vsi_clear_rings(vsi);
7566         return -ENOMEM;
7567 }
7568
7569 /**
7570  * i40e_reserve_msix_vectors - Reserve MSI-X vectors in the kernel
7571  * @pf: board private structure
7572  * @vectors: the number of MSI-X vectors to request
7573  *
7574  * Returns the number of vectors reserved, or error
7575  **/
7576 static int i40e_reserve_msix_vectors(struct i40e_pf *pf, int vectors)
7577 {
7578         vectors = pci_enable_msix_range(pf->pdev, pf->msix_entries,
7579                                         I40E_MIN_MSIX, vectors);
7580         if (vectors < 0) {
7581                 dev_info(&pf->pdev->dev,
7582                          "MSI-X vector reservation failed: %d\n", vectors);
7583                 vectors = 0;
7584         }
7585
7586         return vectors;
7587 }
7588
7589 /**
7590  * i40e_init_msix - Setup the MSIX capability
7591  * @pf: board private structure
7592  *
7593  * Work with the OS to set up the MSIX vectors needed.
7594  *
7595  * Returns the number of vectors reserved or negative on failure
7596  **/
7597 static int i40e_init_msix(struct i40e_pf *pf)
7598 {
7599         struct i40e_hw *hw = &pf->hw;
7600         int vectors_left;
7601         int v_budget, i;
7602         int v_actual;
7603         int iwarp_requested = 0;
7604
7605         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
7606                 return -ENODEV;
7607
7608         /* The number of vectors we'll request will be comprised of:
7609          *   - Add 1 for "other" cause for Admin Queue events, etc.
7610          *   - The number of LAN queue pairs
7611          *      - Queues being used for RSS.
7612          *              We don't need as many as max_rss_size vectors.
7613          *              use rss_size instead in the calculation since that
7614          *              is governed by number of cpus in the system.
7615          *      - assumes symmetric Tx/Rx pairing
7616          *   - The number of VMDq pairs
7617          *   - The CPU count within the NUMA node if iWARP is enabled
7618 #ifdef I40E_FCOE
7619          *   - The number of FCOE qps.
7620 #endif
7621          * Once we count this up, try the request.
7622          *
7623          * If we can't get what we want, we'll simplify to nearly nothing
7624          * and try again.  If that still fails, we punt.
7625          */
7626         vectors_left = hw->func_caps.num_msix_vectors;
7627         v_budget = 0;
7628
7629         /* reserve one vector for miscellaneous handler */
7630         if (vectors_left) {
7631                 v_budget++;
7632                 vectors_left--;
7633         }
7634
7635         /* reserve vectors for the main PF traffic queues */
7636         pf->num_lan_msix = min_t(int, num_online_cpus(), vectors_left);
7637         vectors_left -= pf->num_lan_msix;
7638         v_budget += pf->num_lan_msix;
7639
7640         /* reserve one vector for sideband flow director */
7641         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
7642                 if (vectors_left) {
7643                         pf->num_fdsb_msix = 1;
7644                         v_budget++;
7645                         vectors_left--;
7646                 } else {
7647                         pf->num_fdsb_msix = 0;
7648                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7649                 }
7650         }
7651
7652 #ifdef I40E_FCOE
7653         /* can we reserve enough for FCoE? */
7654         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7655                 if (!vectors_left)
7656                         pf->num_fcoe_msix = 0;
7657                 else if (vectors_left >= pf->num_fcoe_qps)
7658                         pf->num_fcoe_msix = pf->num_fcoe_qps;
7659                 else
7660                         pf->num_fcoe_msix = 1;
7661                 v_budget += pf->num_fcoe_msix;
7662                 vectors_left -= pf->num_fcoe_msix;
7663         }
7664
7665 #endif
7666         /* can we reserve enough for iWARP? */
7667         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7668                 if (!vectors_left)
7669                         pf->num_iwarp_msix = 0;
7670                 else if (vectors_left < pf->num_iwarp_msix)
7671                         pf->num_iwarp_msix = 1;
7672                 v_budget += pf->num_iwarp_msix;
7673                 vectors_left -= pf->num_iwarp_msix;
7674         }
7675
7676         /* any vectors left over go for VMDq support */
7677         if (pf->flags & I40E_FLAG_VMDQ_ENABLED) {
7678                 int vmdq_vecs_wanted = pf->num_vmdq_vsis * pf->num_vmdq_qps;
7679                 int vmdq_vecs = min_t(int, vectors_left, vmdq_vecs_wanted);
7680
7681                 /* if we're short on vectors for what's desired, we limit
7682                  * the queues per vmdq.  If this is still more than are
7683                  * available, the user will need to change the number of
7684                  * queues/vectors used by the PF later with the ethtool
7685                  * channels command
7686                  */
7687                 if (vmdq_vecs < vmdq_vecs_wanted)
7688                         pf->num_vmdq_qps = 1;
7689                 pf->num_vmdq_msix = pf->num_vmdq_qps;
7690
7691                 v_budget += vmdq_vecs;
7692                 vectors_left -= vmdq_vecs;
7693         }
7694
7695         pf->msix_entries = kcalloc(v_budget, sizeof(struct msix_entry),
7696                                    GFP_KERNEL);
7697         if (!pf->msix_entries)
7698                 return -ENOMEM;
7699
7700         for (i = 0; i < v_budget; i++)
7701                 pf->msix_entries[i].entry = i;
7702         v_actual = i40e_reserve_msix_vectors(pf, v_budget);
7703
7704         if (v_actual != v_budget) {
7705                 /* If we have limited resources, we will start with no vectors
7706                  * for the special features and then allocate vectors to some
7707                  * of these features based on the policy and at the end disable
7708                  * the features that did not get any vectors.
7709                  */
7710                 iwarp_requested = pf->num_iwarp_msix;
7711                 pf->num_iwarp_msix = 0;
7712 #ifdef I40E_FCOE
7713                 pf->num_fcoe_qps = 0;
7714                 pf->num_fcoe_msix = 0;
7715 #endif
7716                 pf->num_vmdq_msix = 0;
7717         }
7718
7719         if (v_actual < I40E_MIN_MSIX) {
7720                 pf->flags &= ~I40E_FLAG_MSIX_ENABLED;
7721                 kfree(pf->msix_entries);
7722                 pf->msix_entries = NULL;
7723                 return -ENODEV;
7724
7725         } else if (v_actual == I40E_MIN_MSIX) {
7726                 /* Adjust for minimal MSIX use */
7727                 pf->num_vmdq_vsis = 0;
7728                 pf->num_vmdq_qps = 0;
7729                 pf->num_lan_qps = 1;
7730                 pf->num_lan_msix = 1;
7731
7732         } else if (v_actual != v_budget) {
7733                 int vec;
7734
7735                 /* reserve the misc vector */
7736                 vec = v_actual - 1;
7737
7738                 /* Scale vector usage down */
7739                 pf->num_vmdq_msix = 1;    /* force VMDqs to only one vector */
7740                 pf->num_vmdq_vsis = 1;
7741                 pf->num_vmdq_qps = 1;
7742                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
7743
7744                 /* partition out the remaining vectors */
7745                 switch (vec) {
7746                 case 2:
7747                         pf->num_lan_msix = 1;
7748                         break;
7749                 case 3:
7750                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7751                                 pf->num_lan_msix = 1;
7752                                 pf->num_iwarp_msix = 1;
7753                         } else {
7754                                 pf->num_lan_msix = 2;
7755                         }
7756 #ifdef I40E_FCOE
7757                         /* give one vector to FCoE */
7758                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7759                                 pf->num_lan_msix = 1;
7760                                 pf->num_fcoe_msix = 1;
7761                         }
7762 #endif
7763                         break;
7764                 default:
7765                         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
7766                                 pf->num_iwarp_msix = min_t(int, (vec / 3),
7767                                                  iwarp_requested);
7768                                 pf->num_vmdq_vsis = min_t(int, (vec / 3),
7769                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7770                         } else {
7771                                 pf->num_vmdq_vsis = min_t(int, (vec / 2),
7772                                                   I40E_DEFAULT_NUM_VMDQ_VSI);
7773                         }
7774                         pf->num_lan_msix = min_t(int,
7775                                (vec - (pf->num_iwarp_msix + pf->num_vmdq_vsis)),
7776                                                               pf->num_lan_msix);
7777 #ifdef I40E_FCOE
7778                         /* give one vector to FCoE */
7779                         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
7780                                 pf->num_fcoe_msix = 1;
7781                                 vec--;
7782                         }
7783 #endif
7784                         break;
7785                 }
7786         }
7787
7788         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
7789             (pf->num_vmdq_msix == 0)) {
7790                 dev_info(&pf->pdev->dev, "VMDq disabled, not enough MSI-X vectors\n");
7791                 pf->flags &= ~I40E_FLAG_VMDQ_ENABLED;
7792         }
7793
7794         if ((pf->flags & I40E_FLAG_IWARP_ENABLED) &&
7795             (pf->num_iwarp_msix == 0)) {
7796                 dev_info(&pf->pdev->dev, "IWARP disabled, not enough MSI-X vectors\n");
7797                 pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
7798         }
7799 #ifdef I40E_FCOE
7800
7801         if ((pf->flags & I40E_FLAG_FCOE_ENABLED) && (pf->num_fcoe_msix == 0)) {
7802                 dev_info(&pf->pdev->dev, "FCOE disabled, not enough MSI-X vectors\n");
7803                 pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
7804         }
7805 #endif
7806         return v_actual;
7807 }
7808
7809 /**
7810  * i40e_vsi_alloc_q_vector - Allocate memory for a single interrupt vector
7811  * @vsi: the VSI being configured
7812  * @v_idx: index of the vector in the vsi struct
7813  * @cpu: cpu to be used on affinity_mask
7814  *
7815  * We allocate one q_vector.  If allocation fails we return -ENOMEM.
7816  **/
7817 static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
7818 {
7819         struct i40e_q_vector *q_vector;
7820
7821         /* allocate q_vector */
7822         q_vector = kzalloc(sizeof(struct i40e_q_vector), GFP_KERNEL);
7823         if (!q_vector)
7824                 return -ENOMEM;
7825
7826         q_vector->vsi = vsi;
7827         q_vector->v_idx = v_idx;
7828         cpumask_set_cpu(cpu, &q_vector->affinity_mask);
7829
7830         if (vsi->netdev)
7831                 netif_napi_add(vsi->netdev, &q_vector->napi,
7832                                i40e_napi_poll, NAPI_POLL_WEIGHT);
7833
7834         q_vector->rx.latency_range = I40E_LOW_LATENCY;
7835         q_vector->tx.latency_range = I40E_LOW_LATENCY;
7836
7837         /* tie q_vector and vsi together */
7838         vsi->q_vectors[v_idx] = q_vector;
7839
7840         return 0;
7841 }
7842
7843 /**
7844  * i40e_vsi_alloc_q_vectors - Allocate memory for interrupt vectors
7845  * @vsi: the VSI being configured
7846  *
7847  * We allocate one q_vector per queue interrupt.  If allocation fails we
7848  * return -ENOMEM.
7849  **/
7850 static int i40e_vsi_alloc_q_vectors(struct i40e_vsi *vsi)
7851 {
7852         struct i40e_pf *pf = vsi->back;
7853         int err, v_idx, num_q_vectors, current_cpu;
7854
7855         /* if not MSIX, give the one vector only to the LAN VSI */
7856         if (pf->flags & I40E_FLAG_MSIX_ENABLED)
7857                 num_q_vectors = vsi->num_q_vectors;
7858         else if (vsi == pf->vsi[pf->lan_vsi])
7859                 num_q_vectors = 1;
7860         else
7861                 return -EINVAL;
7862
7863         current_cpu = cpumask_first(cpu_online_mask);
7864
7865         for (v_idx = 0; v_idx < num_q_vectors; v_idx++) {
7866                 err = i40e_vsi_alloc_q_vector(vsi, v_idx, current_cpu);
7867                 if (err)
7868                         goto err_out;
7869                 current_cpu = cpumask_next(current_cpu, cpu_online_mask);
7870                 if (unlikely(current_cpu >= nr_cpu_ids))
7871                         current_cpu = cpumask_first(cpu_online_mask);
7872         }
7873
7874         return 0;
7875
7876 err_out:
7877         while (v_idx--)
7878                 i40e_free_q_vector(vsi, v_idx);
7879
7880         return err;
7881 }
7882
7883 /**
7884  * i40e_init_interrupt_scheme - Determine proper interrupt scheme
7885  * @pf: board private structure to initialize
7886  **/
7887 static int i40e_init_interrupt_scheme(struct i40e_pf *pf)
7888 {
7889         int vectors = 0;
7890         ssize_t size;
7891
7892         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
7893                 vectors = i40e_init_msix(pf);
7894                 if (vectors < 0) {
7895                         pf->flags &= ~(I40E_FLAG_MSIX_ENABLED   |
7896                                        I40E_FLAG_IWARP_ENABLED  |
7897 #ifdef I40E_FCOE
7898                                        I40E_FLAG_FCOE_ENABLED   |
7899 #endif
7900                                        I40E_FLAG_RSS_ENABLED    |
7901                                        I40E_FLAG_DCB_CAPABLE    |
7902                                        I40E_FLAG_DCB_ENABLED    |
7903                                        I40E_FLAG_SRIOV_ENABLED  |
7904                                        I40E_FLAG_FD_SB_ENABLED  |
7905                                        I40E_FLAG_FD_ATR_ENABLED |
7906                                        I40E_FLAG_VMDQ_ENABLED);
7907
7908                         /* rework the queue expectations without MSIX */
7909                         i40e_determine_queue_usage(pf);
7910                 }
7911         }
7912
7913         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED) &&
7914             (pf->flags & I40E_FLAG_MSI_ENABLED)) {
7915                 dev_info(&pf->pdev->dev, "MSI-X not available, trying MSI\n");
7916                 vectors = pci_enable_msi(pf->pdev);
7917                 if (vectors < 0) {
7918                         dev_info(&pf->pdev->dev, "MSI init failed - %d\n",
7919                                  vectors);
7920                         pf->flags &= ~I40E_FLAG_MSI_ENABLED;
7921                 }
7922                 vectors = 1;  /* one MSI or Legacy vector */
7923         }
7924
7925         if (!(pf->flags & (I40E_FLAG_MSIX_ENABLED | I40E_FLAG_MSI_ENABLED)))
7926                 dev_info(&pf->pdev->dev, "MSI-X and MSI not available, falling back to Legacy IRQ\n");
7927
7928         /* set up vector assignment tracking */
7929         size = sizeof(struct i40e_lump_tracking) + (sizeof(u16) * vectors);
7930         pf->irq_pile = kzalloc(size, GFP_KERNEL);
7931         if (!pf->irq_pile) {
7932                 dev_err(&pf->pdev->dev, "error allocating irq_pile memory\n");
7933                 return -ENOMEM;
7934         }
7935         pf->irq_pile->num_entries = vectors;
7936         pf->irq_pile->search_hint = 0;
7937
7938         /* track first vector for misc interrupts, ignore return */
7939         (void)i40e_get_lump(pf, pf->irq_pile, 1, I40E_PILE_VALID_BIT - 1);
7940
7941         return 0;
7942 }
7943
7944 /**
7945  * i40e_setup_misc_vector - Setup the misc vector to handle non queue events
7946  * @pf: board private structure
7947  *
7948  * This sets up the handler for MSIX 0, which is used to manage the
7949  * non-queue interrupts, e.g. AdminQ and errors.  This is not used
7950  * when in MSI or Legacy interrupt mode.
7951  **/
7952 static int i40e_setup_misc_vector(struct i40e_pf *pf)
7953 {
7954         struct i40e_hw *hw = &pf->hw;
7955         int err = 0;
7956
7957         /* Only request the irq if this is the first time through, and
7958          * not when we're rebuilding after a Reset
7959          */
7960         if (!test_bit(__I40E_RESET_RECOVERY_PENDING, &pf->state)) {
7961                 err = request_irq(pf->msix_entries[0].vector,
7962                                   i40e_intr, 0, pf->int_name, pf);
7963                 if (err) {
7964                         dev_info(&pf->pdev->dev,
7965                                  "request_irq for %s failed: %d\n",
7966                                  pf->int_name, err);
7967                         return -EFAULT;
7968                 }
7969         }
7970
7971         i40e_enable_misc_int_causes(pf);
7972
7973         /* associate no queues to the misc vector */
7974         wr32(hw, I40E_PFINT_LNKLST0, I40E_QUEUE_END_OF_LIST);
7975         wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), I40E_ITR_8K);
7976
7977         i40e_flush(hw);
7978
7979         i40e_irq_dynamic_enable_icr0(pf, true);
7980
7981         return err;
7982 }
7983
7984 /**
7985  * i40e_config_rss_aq - Prepare for RSS using AQ commands
7986  * @vsi: vsi structure
7987  * @seed: RSS hash seed
7988  **/
7989 static int i40e_config_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
7990                               u8 *lut, u16 lut_size)
7991 {
7992         struct i40e_pf *pf = vsi->back;
7993         struct i40e_hw *hw = &pf->hw;
7994         int ret = 0;
7995
7996         if (seed) {
7997                 struct i40e_aqc_get_set_rss_key_data *seed_dw =
7998                         (struct i40e_aqc_get_set_rss_key_data *)seed;
7999                 ret = i40e_aq_set_rss_key(hw, vsi->id, seed_dw);
8000                 if (ret) {
8001                         dev_info(&pf->pdev->dev,
8002                                  "Cannot set RSS key, err %s aq_err %s\n",
8003                                  i40e_stat_str(hw, ret),
8004                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8005                         return ret;
8006                 }
8007         }
8008         if (lut) {
8009                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8010
8011                 ret = i40e_aq_set_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8012                 if (ret) {
8013                         dev_info(&pf->pdev->dev,
8014                                  "Cannot set RSS lut, err %s aq_err %s\n",
8015                                  i40e_stat_str(hw, ret),
8016                                  i40e_aq_str(hw, hw->aq.asq_last_status));
8017                         return ret;
8018                 }
8019         }
8020         return ret;
8021 }
8022
8023 /**
8024  * i40e_get_rss_aq - Get RSS keys and lut by using AQ commands
8025  * @vsi: Pointer to vsi structure
8026  * @seed: Buffter to store the hash keys
8027  * @lut: Buffer to store the lookup table entries
8028  * @lut_size: Size of buffer to store the lookup table entries
8029  *
8030  * Return 0 on success, negative on failure
8031  */
8032 static int i40e_get_rss_aq(struct i40e_vsi *vsi, const u8 *seed,
8033                            u8 *lut, u16 lut_size)
8034 {
8035         struct i40e_pf *pf = vsi->back;
8036         struct i40e_hw *hw = &pf->hw;
8037         int ret = 0;
8038
8039         if (seed) {
8040                 ret = i40e_aq_get_rss_key(hw, vsi->id,
8041                         (struct i40e_aqc_get_set_rss_key_data *)seed);
8042                 if (ret) {
8043                         dev_info(&pf->pdev->dev,
8044                                  "Cannot get RSS key, err %s aq_err %s\n",
8045                                  i40e_stat_str(&pf->hw, ret),
8046                                  i40e_aq_str(&pf->hw,
8047                                              pf->hw.aq.asq_last_status));
8048                         return ret;
8049                 }
8050         }
8051
8052         if (lut) {
8053                 bool pf_lut = vsi->type == I40E_VSI_MAIN ? true : false;
8054
8055                 ret = i40e_aq_get_rss_lut(hw, vsi->id, pf_lut, lut, lut_size);
8056                 if (ret) {
8057                         dev_info(&pf->pdev->dev,
8058                                  "Cannot get RSS lut, err %s aq_err %s\n",
8059                                  i40e_stat_str(&pf->hw, ret),
8060                                  i40e_aq_str(&pf->hw,
8061                                              pf->hw.aq.asq_last_status));
8062                         return ret;
8063                 }
8064         }
8065
8066         return ret;
8067 }
8068
8069 /**
8070  * i40e_vsi_config_rss - Prepare for VSI(VMDq) RSS if used
8071  * @vsi: VSI structure
8072  **/
8073 static int i40e_vsi_config_rss(struct i40e_vsi *vsi)
8074 {
8075         u8 seed[I40E_HKEY_ARRAY_SIZE];
8076         struct i40e_pf *pf = vsi->back;
8077         u8 *lut;
8078         int ret;
8079
8080         if (!(pf->flags & I40E_FLAG_RSS_AQ_CAPABLE))
8081                 return 0;
8082
8083         if (!vsi->rss_size)
8084                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8085                                       vsi->num_queue_pairs);
8086         if (!vsi->rss_size)
8087                 return -EINVAL;
8088
8089         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8090         if (!lut)
8091                 return -ENOMEM;
8092         /* Use the user configured hash keys and lookup table if there is one,
8093          * otherwise use default
8094          */
8095         if (vsi->rss_lut_user)
8096                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8097         else
8098                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8099         if (vsi->rss_hkey_user)
8100                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8101         else
8102                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8103         ret = i40e_config_rss_aq(vsi, seed, lut, vsi->rss_table_size);
8104         kfree(lut);
8105
8106         return ret;
8107 }
8108
8109 /**
8110  * i40e_config_rss_reg - Configure RSS keys and lut by writing registers
8111  * @vsi: Pointer to vsi structure
8112  * @seed: RSS hash seed
8113  * @lut: Lookup table
8114  * @lut_size: Lookup table size
8115  *
8116  * Returns 0 on success, negative on failure
8117  **/
8118 static int i40e_config_rss_reg(struct i40e_vsi *vsi, const u8 *seed,
8119                                const u8 *lut, u16 lut_size)
8120 {
8121         struct i40e_pf *pf = vsi->back;
8122         struct i40e_hw *hw = &pf->hw;
8123         u16 vf_id = vsi->vf_id;
8124         u8 i;
8125
8126         /* Fill out hash function seed */
8127         if (seed) {
8128                 u32 *seed_dw = (u32 *)seed;
8129
8130                 if (vsi->type == I40E_VSI_MAIN) {
8131                         for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8132                                 i40e_write_rx_ctl(hw, I40E_PFQF_HKEY(i),
8133                                                   seed_dw[i]);
8134                 } else if (vsi->type == I40E_VSI_SRIOV) {
8135                         for (i = 0; i <= I40E_VFQF_HKEY1_MAX_INDEX; i++)
8136                                 i40e_write_rx_ctl(hw,
8137                                                   I40E_VFQF_HKEY1(i, vf_id),
8138                                                   seed_dw[i]);
8139                 } else {
8140                         dev_err(&pf->pdev->dev, "Cannot set RSS seed - invalid VSI type\n");
8141                 }
8142         }
8143
8144         if (lut) {
8145                 u32 *lut_dw = (u32 *)lut;
8146
8147                 if (vsi->type == I40E_VSI_MAIN) {
8148                         if (lut_size != I40E_HLUT_ARRAY_SIZE)
8149                                 return -EINVAL;
8150                         for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8151                                 wr32(hw, I40E_PFQF_HLUT(i), lut_dw[i]);
8152                 } else if (vsi->type == I40E_VSI_SRIOV) {
8153                         if (lut_size != I40E_VF_HLUT_ARRAY_SIZE)
8154                                 return -EINVAL;
8155                         for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8156                                 i40e_write_rx_ctl(hw,
8157                                                   I40E_VFQF_HLUT1(i, vf_id),
8158                                                   lut_dw[i]);
8159                 } else {
8160                         dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8161                 }
8162         }
8163         i40e_flush(hw);
8164
8165         return 0;
8166 }
8167
8168 /**
8169  * i40e_get_rss_reg - Get the RSS keys and lut by reading registers
8170  * @vsi: Pointer to VSI structure
8171  * @seed: Buffer to store the keys
8172  * @lut: Buffer to store the lookup table entries
8173  * @lut_size: Size of buffer to store the lookup table entries
8174  *
8175  * Returns 0 on success, negative on failure
8176  */
8177 static int i40e_get_rss_reg(struct i40e_vsi *vsi, u8 *seed,
8178                             u8 *lut, u16 lut_size)
8179 {
8180         struct i40e_pf *pf = vsi->back;
8181         struct i40e_hw *hw = &pf->hw;
8182         u16 i;
8183
8184         if (seed) {
8185                 u32 *seed_dw = (u32 *)seed;
8186
8187                 for (i = 0; i <= I40E_PFQF_HKEY_MAX_INDEX; i++)
8188                         seed_dw[i] = i40e_read_rx_ctl(hw, I40E_PFQF_HKEY(i));
8189         }
8190         if (lut) {
8191                 u32 *lut_dw = (u32 *)lut;
8192
8193                 if (lut_size != I40E_HLUT_ARRAY_SIZE)
8194                         return -EINVAL;
8195                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8196                         lut_dw[i] = rd32(hw, I40E_PFQF_HLUT(i));
8197         }
8198
8199         return 0;
8200 }
8201
8202 /**
8203  * i40e_config_rss - Configure RSS keys and lut
8204  * @vsi: Pointer to VSI structure
8205  * @seed: RSS hash seed
8206  * @lut: Lookup table
8207  * @lut_size: Lookup table size
8208  *
8209  * Returns 0 on success, negative on failure
8210  */
8211 int i40e_config_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8212 {
8213         struct i40e_pf *pf = vsi->back;
8214
8215         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8216                 return i40e_config_rss_aq(vsi, seed, lut, lut_size);
8217         else
8218                 return i40e_config_rss_reg(vsi, seed, lut, lut_size);
8219 }
8220
8221 /**
8222  * i40e_get_rss - Get RSS keys and lut
8223  * @vsi: Pointer to VSI structure
8224  * @seed: Buffer to store the keys
8225  * @lut: Buffer to store the lookup table entries
8226  * lut_size: Size of buffer to store the lookup table entries
8227  *
8228  * Returns 0 on success, negative on failure
8229  */
8230 int i40e_get_rss(struct i40e_vsi *vsi, u8 *seed, u8 *lut, u16 lut_size)
8231 {
8232         struct i40e_pf *pf = vsi->back;
8233
8234         if (pf->flags & I40E_FLAG_RSS_AQ_CAPABLE)
8235                 return i40e_get_rss_aq(vsi, seed, lut, lut_size);
8236         else
8237                 return i40e_get_rss_reg(vsi, seed, lut, lut_size);
8238 }
8239
8240 /**
8241  * i40e_fill_rss_lut - Fill the RSS lookup table with default values
8242  * @pf: Pointer to board private structure
8243  * @lut: Lookup table
8244  * @rss_table_size: Lookup table size
8245  * @rss_size: Range of queue number for hashing
8246  */
8247 static void i40e_fill_rss_lut(struct i40e_pf *pf, u8 *lut,
8248                               u16 rss_table_size, u16 rss_size)
8249 {
8250         u16 i;
8251
8252         for (i = 0; i < rss_table_size; i++)
8253                 lut[i] = i % rss_size;
8254 }
8255
8256 /**
8257  * i40e_pf_config_rss - Prepare for RSS if used
8258  * @pf: board private structure
8259  **/
8260 static int i40e_pf_config_rss(struct i40e_pf *pf)
8261 {
8262         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8263         u8 seed[I40E_HKEY_ARRAY_SIZE];
8264         u8 *lut;
8265         struct i40e_hw *hw = &pf->hw;
8266         u32 reg_val;
8267         u64 hena;
8268         int ret;
8269
8270         /* By default we enable TCP/UDP with IPv4/IPv6 ptypes */
8271         hena = (u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(0)) |
8272                 ((u64)i40e_read_rx_ctl(hw, I40E_PFQF_HENA(1)) << 32);
8273         hena |= i40e_pf_get_default_rss_hena(pf);
8274
8275         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), (u32)hena);
8276         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), (u32)(hena >> 32));
8277
8278         /* Determine the RSS table size based on the hardware capabilities */
8279         reg_val = i40e_read_rx_ctl(hw, I40E_PFQF_CTL_0);
8280         reg_val = (pf->rss_table_size == 512) ?
8281                         (reg_val | I40E_PFQF_CTL_0_HASHLUTSIZE_512) :
8282                         (reg_val & ~I40E_PFQF_CTL_0_HASHLUTSIZE_512);
8283         i40e_write_rx_ctl(hw, I40E_PFQF_CTL_0, reg_val);
8284
8285         /* Determine the RSS size of the VSI */
8286         if (!vsi->rss_size)
8287                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8288                                       vsi->num_queue_pairs);
8289
8290         lut = kzalloc(vsi->rss_table_size, GFP_KERNEL);
8291         if (!lut)
8292                 return -ENOMEM;
8293
8294         /* Use user configured lut if there is one, otherwise use default */
8295         if (vsi->rss_lut_user)
8296                 memcpy(lut, vsi->rss_lut_user, vsi->rss_table_size);
8297         else
8298                 i40e_fill_rss_lut(pf, lut, vsi->rss_table_size, vsi->rss_size);
8299
8300         /* Use user configured hash key if there is one, otherwise
8301          * use default.
8302          */
8303         if (vsi->rss_hkey_user)
8304                 memcpy(seed, vsi->rss_hkey_user, I40E_HKEY_ARRAY_SIZE);
8305         else
8306                 netdev_rss_key_fill((void *)seed, I40E_HKEY_ARRAY_SIZE);
8307         ret = i40e_config_rss(vsi, seed, lut, vsi->rss_table_size);
8308         kfree(lut);
8309
8310         return ret;
8311 }
8312
8313 /**
8314  * i40e_reconfig_rss_queues - change number of queues for rss and rebuild
8315  * @pf: board private structure
8316  * @queue_count: the requested queue count for rss.
8317  *
8318  * returns 0 if rss is not enabled, if enabled returns the final rss queue
8319  * count which may be different from the requested queue count.
8320  **/
8321 int i40e_reconfig_rss_queues(struct i40e_pf *pf, int queue_count)
8322 {
8323         struct i40e_vsi *vsi = pf->vsi[pf->lan_vsi];
8324         int new_rss_size;
8325
8326         if (!(pf->flags & I40E_FLAG_RSS_ENABLED))
8327                 return 0;
8328
8329         new_rss_size = min_t(int, queue_count, pf->rss_size_max);
8330
8331         if (queue_count != vsi->num_queue_pairs) {
8332                 vsi->req_queue_pairs = queue_count;
8333                 i40e_prep_for_reset(pf);
8334
8335                 pf->alloc_rss_size = new_rss_size;
8336
8337                 i40e_reset_and_rebuild(pf, true);
8338
8339                 /* Discard the user configured hash keys and lut, if less
8340                  * queues are enabled.
8341                  */
8342                 if (queue_count < vsi->rss_size) {
8343                         i40e_clear_rss_config_user(vsi);
8344                         dev_dbg(&pf->pdev->dev,
8345                                 "discard user configured hash keys and lut\n");
8346                 }
8347
8348                 /* Reset vsi->rss_size, as number of enabled queues changed */
8349                 vsi->rss_size = min_t(int, pf->alloc_rss_size,
8350                                       vsi->num_queue_pairs);
8351
8352                 i40e_pf_config_rss(pf);
8353         }
8354         dev_info(&pf->pdev->dev, "RSS count/HW max RSS count:  %d/%d\n",
8355                  pf->alloc_rss_size, pf->rss_size_max);
8356         return pf->alloc_rss_size;
8357 }
8358
8359 /**
8360  * i40e_get_npar_bw_setting - Retrieve BW settings for this PF partition
8361  * @pf: board private structure
8362  **/
8363 i40e_status i40e_get_npar_bw_setting(struct i40e_pf *pf)
8364 {
8365         i40e_status status;
8366         bool min_valid, max_valid;
8367         u32 max_bw, min_bw;
8368
8369         status = i40e_read_bw_from_alt_ram(&pf->hw, &max_bw, &min_bw,
8370                                            &min_valid, &max_valid);
8371
8372         if (!status) {
8373                 if (min_valid)
8374                         pf->npar_min_bw = min_bw;
8375                 if (max_valid)
8376                         pf->npar_max_bw = max_bw;
8377         }
8378
8379         return status;
8380 }
8381
8382 /**
8383  * i40e_set_npar_bw_setting - Set BW settings for this PF partition
8384  * @pf: board private structure
8385  **/
8386 i40e_status i40e_set_npar_bw_setting(struct i40e_pf *pf)
8387 {
8388         struct i40e_aqc_configure_partition_bw_data bw_data;
8389         i40e_status status;
8390
8391         /* Set the valid bit for this PF */
8392         bw_data.pf_valid_bits = cpu_to_le16(BIT(pf->hw.pf_id));
8393         bw_data.max_bw[pf->hw.pf_id] = pf->npar_max_bw & I40E_ALT_BW_VALUE_MASK;
8394         bw_data.min_bw[pf->hw.pf_id] = pf->npar_min_bw & I40E_ALT_BW_VALUE_MASK;
8395
8396         /* Set the new bandwidths */
8397         status = i40e_aq_configure_partition_bw(&pf->hw, &bw_data, NULL);
8398
8399         return status;
8400 }
8401
8402 /**
8403  * i40e_commit_npar_bw_setting - Commit BW settings for this PF partition
8404  * @pf: board private structure
8405  **/
8406 i40e_status i40e_commit_npar_bw_setting(struct i40e_pf *pf)
8407 {
8408         /* Commit temporary BW setting to permanent NVM image */
8409         enum i40e_admin_queue_err last_aq_status;
8410         i40e_status ret;
8411         u16 nvm_word;
8412
8413         if (pf->hw.partition_id != 1) {
8414                 dev_info(&pf->pdev->dev,
8415                          "Commit BW only works on partition 1! This is partition %d",
8416                          pf->hw.partition_id);
8417                 ret = I40E_NOT_SUPPORTED;
8418                 goto bw_commit_out;
8419         }
8420
8421         /* Acquire NVM for read access */
8422         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_READ);
8423         last_aq_status = pf->hw.aq.asq_last_status;
8424         if (ret) {
8425                 dev_info(&pf->pdev->dev,
8426                          "Cannot acquire NVM for read access, err %s aq_err %s\n",
8427                          i40e_stat_str(&pf->hw, ret),
8428                          i40e_aq_str(&pf->hw, last_aq_status));
8429                 goto bw_commit_out;
8430         }
8431
8432         /* Read word 0x10 of NVM - SW compatibility word 1 */
8433         ret = i40e_aq_read_nvm(&pf->hw,
8434                                I40E_SR_NVM_CONTROL_WORD,
8435                                0x10, sizeof(nvm_word), &nvm_word,
8436                                false, NULL);
8437         /* Save off last admin queue command status before releasing
8438          * the NVM
8439          */
8440         last_aq_status = pf->hw.aq.asq_last_status;
8441         i40e_release_nvm(&pf->hw);
8442         if (ret) {
8443                 dev_info(&pf->pdev->dev, "NVM read error, err %s aq_err %s\n",
8444                          i40e_stat_str(&pf->hw, ret),
8445                          i40e_aq_str(&pf->hw, last_aq_status));
8446                 goto bw_commit_out;
8447         }
8448
8449         /* Wait a bit for NVM release to complete */
8450         msleep(50);
8451
8452         /* Acquire NVM for write access */
8453         ret = i40e_acquire_nvm(&pf->hw, I40E_RESOURCE_WRITE);
8454         last_aq_status = pf->hw.aq.asq_last_status;
8455         if (ret) {
8456                 dev_info(&pf->pdev->dev,
8457                          "Cannot acquire NVM for write access, err %s aq_err %s\n",
8458                          i40e_stat_str(&pf->hw, ret),
8459                          i40e_aq_str(&pf->hw, last_aq_status));
8460                 goto bw_commit_out;
8461         }
8462         /* Write it back out unchanged to initiate update NVM,
8463          * which will force a write of the shadow (alt) RAM to
8464          * the NVM - thus storing the bandwidth values permanently.
8465          */
8466         ret = i40e_aq_update_nvm(&pf->hw,
8467                                  I40E_SR_NVM_CONTROL_WORD,
8468                                  0x10, sizeof(nvm_word),
8469                                  &nvm_word, true, NULL);
8470         /* Save off last admin queue command status before releasing
8471          * the NVM
8472          */
8473         last_aq_status = pf->hw.aq.asq_last_status;
8474         i40e_release_nvm(&pf->hw);
8475         if (ret)
8476                 dev_info(&pf->pdev->dev,
8477                          "BW settings NOT SAVED, err %s aq_err %s\n",
8478                          i40e_stat_str(&pf->hw, ret),
8479                          i40e_aq_str(&pf->hw, last_aq_status));
8480 bw_commit_out:
8481
8482         return ret;
8483 }
8484
8485 /**
8486  * i40e_sw_init - Initialize general software structures (struct i40e_pf)
8487  * @pf: board private structure to initialize
8488  *
8489  * i40e_sw_init initializes the Adapter private data structure.
8490  * Fields are initialized based on PCI device information and
8491  * OS network device settings (MTU size).
8492  **/
8493 static int i40e_sw_init(struct i40e_pf *pf)
8494 {
8495         int err = 0;
8496         int size;
8497
8498         pf->msg_enable = netif_msg_init(I40E_DEFAULT_MSG_ENABLE,
8499                                 (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK));
8500         if (debug != -1 && debug != I40E_DEFAULT_MSG_ENABLE) {
8501                 if (I40E_DEBUG_USER & debug)
8502                         pf->hw.debug_mask = debug;
8503                 pf->msg_enable = netif_msg_init((debug & ~I40E_DEBUG_USER),
8504                                                 I40E_DEFAULT_MSG_ENABLE);
8505         }
8506
8507         /* Set default capability flags */
8508         pf->flags = I40E_FLAG_RX_CSUM_ENABLED |
8509                     I40E_FLAG_MSI_ENABLED     |
8510                     I40E_FLAG_MSIX_ENABLED;
8511
8512         /* Set default ITR */
8513         pf->rx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_RX_DEF;
8514         pf->tx_itr_default = I40E_ITR_DYNAMIC | I40E_ITR_TX_DEF;
8515
8516         /* Depending on PF configurations, it is possible that the RSS
8517          * maximum might end up larger than the available queues
8518          */
8519         pf->rss_size_max = BIT(pf->hw.func_caps.rss_table_entry_width);
8520         pf->alloc_rss_size = 1;
8521         pf->rss_table_size = pf->hw.func_caps.rss_table_size;
8522         pf->rss_size_max = min_t(int, pf->rss_size_max,
8523                                  pf->hw.func_caps.num_tx_qp);
8524         if (pf->hw.func_caps.rss) {
8525                 pf->flags |= I40E_FLAG_RSS_ENABLED;
8526                 pf->alloc_rss_size = min_t(int, pf->rss_size_max,
8527                                            num_online_cpus());
8528         }
8529
8530         /* MFP mode enabled */
8531         if (pf->hw.func_caps.npar_enable || pf->hw.func_caps.flex10_enable) {
8532                 pf->flags |= I40E_FLAG_MFP_ENABLED;
8533                 dev_info(&pf->pdev->dev, "MFP mode Enabled\n");
8534                 if (i40e_get_npar_bw_setting(pf))
8535                         dev_warn(&pf->pdev->dev,
8536                                  "Could not get NPAR bw settings\n");
8537                 else
8538                         dev_info(&pf->pdev->dev,
8539                                  "Min BW = %8.8x, Max BW = %8.8x\n",
8540                                  pf->npar_min_bw, pf->npar_max_bw);
8541         }
8542
8543         /* FW/NVM is not yet fixed in this regard */
8544         if ((pf->hw.func_caps.fd_filters_guaranteed > 0) ||
8545             (pf->hw.func_caps.fd_filters_best_effort > 0)) {
8546                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8547                 pf->atr_sample_rate = I40E_DEFAULT_ATR_SAMPLE_RATE;
8548                 if (pf->flags & I40E_FLAG_MFP_ENABLED &&
8549                     pf->hw.num_partitions > 1)
8550                         dev_info(&pf->pdev->dev,
8551                                  "Flow Director Sideband mode Disabled in MFP mode\n");
8552                 else
8553                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8554                 pf->fdir_pf_filter_count =
8555                                  pf->hw.func_caps.fd_filters_guaranteed;
8556                 pf->hw.fdir_shared_filter_count =
8557                                  pf->hw.func_caps.fd_filters_best_effort;
8558         }
8559
8560         if (i40e_is_mac_710(&pf->hw) &&
8561             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 33)) ||
8562             (pf->hw.aq.fw_maj_ver < 4))) {
8563                 pf->flags |= I40E_FLAG_RESTART_AUTONEG;
8564                 /* No DCB support  for FW < v4.33 */
8565                 pf->flags |= I40E_FLAG_NO_DCB_SUPPORT;
8566         }
8567
8568         /* Disable FW LLDP if FW < v4.3 */
8569         if (i40e_is_mac_710(&pf->hw) &&
8570             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver < 3)) ||
8571             (pf->hw.aq.fw_maj_ver < 4)))
8572                 pf->flags |= I40E_FLAG_STOP_FW_LLDP;
8573
8574         /* Use the FW Set LLDP MIB API if FW > v4.40 */
8575         if (i40e_is_mac_710(&pf->hw) &&
8576             (((pf->hw.aq.fw_maj_ver == 4) && (pf->hw.aq.fw_min_ver >= 40)) ||
8577             (pf->hw.aq.fw_maj_ver >= 5)))
8578                 pf->flags |= I40E_FLAG_USE_SET_LLDP_MIB;
8579
8580         if (pf->hw.func_caps.vmdq) {
8581                 pf->num_vmdq_vsis = I40E_DEFAULT_NUM_VMDQ_VSI;
8582                 pf->flags |= I40E_FLAG_VMDQ_ENABLED;
8583                 pf->num_vmdq_qps = i40e_default_queues_per_vmdq(pf);
8584         }
8585
8586         if (pf->hw.func_caps.iwarp) {
8587                 pf->flags |= I40E_FLAG_IWARP_ENABLED;
8588                 /* IWARP needs one extra vector for CQP just like MISC.*/
8589                 pf->num_iwarp_msix = (int)num_online_cpus() + 1;
8590         }
8591
8592 #ifdef I40E_FCOE
8593         i40e_init_pf_fcoe(pf);
8594
8595 #endif /* I40E_FCOE */
8596 #ifdef CONFIG_PCI_IOV
8597         if (pf->hw.func_caps.num_vfs && pf->hw.partition_id == 1) {
8598                 pf->num_vf_qps = I40E_DEFAULT_QUEUES_PER_VF;
8599                 pf->flags |= I40E_FLAG_SRIOV_ENABLED;
8600                 pf->num_req_vfs = min_t(int,
8601                                         pf->hw.func_caps.num_vfs,
8602                                         I40E_MAX_VF_COUNT);
8603         }
8604 #endif /* CONFIG_PCI_IOV */
8605         if (pf->hw.mac.type == I40E_MAC_X722) {
8606                 pf->flags |= I40E_FLAG_RSS_AQ_CAPABLE |
8607                              I40E_FLAG_128_QP_RSS_CAPABLE |
8608                              I40E_FLAG_HW_ATR_EVICT_CAPABLE |
8609                              I40E_FLAG_OUTER_UDP_CSUM_CAPABLE |
8610                              I40E_FLAG_WB_ON_ITR_CAPABLE |
8611                              I40E_FLAG_MULTIPLE_TCP_UDP_RSS_PCTYPE |
8612                              I40E_FLAG_NO_PCI_LINK_CHECK |
8613                              I40E_FLAG_100M_SGMII_CAPABLE |
8614                              I40E_FLAG_USE_SET_LLDP_MIB |
8615                              I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8616         } else if ((pf->hw.aq.api_maj_ver > 1) ||
8617                    ((pf->hw.aq.api_maj_ver == 1) &&
8618                     (pf->hw.aq.api_min_ver > 4))) {
8619                 /* Supported in FW API version higher than 1.4 */
8620                 pf->flags |= I40E_FLAG_GENEVE_OFFLOAD_CAPABLE;
8621                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8622         } else {
8623                 pf->auto_disable_flags = I40E_FLAG_HW_ATR_EVICT_CAPABLE;
8624         }
8625
8626         pf->eeprom_version = 0xDEAD;
8627         pf->lan_veb = I40E_NO_VEB;
8628         pf->lan_vsi = I40E_NO_VSI;
8629
8630         /* By default FW has this off for performance reasons */
8631         pf->flags &= ~I40E_FLAG_VEB_STATS_ENABLED;
8632
8633         /* set up queue assignment tracking */
8634         size = sizeof(struct i40e_lump_tracking)
8635                 + (sizeof(u16) * pf->hw.func_caps.num_tx_qp);
8636         pf->qp_pile = kzalloc(size, GFP_KERNEL);
8637         if (!pf->qp_pile) {
8638                 err = -ENOMEM;
8639                 goto sw_init_done;
8640         }
8641         pf->qp_pile->num_entries = pf->hw.func_caps.num_tx_qp;
8642         pf->qp_pile->search_hint = 0;
8643
8644         pf->tx_timeout_recovery_level = 1;
8645
8646         mutex_init(&pf->switch_mutex);
8647
8648         /* If NPAR is enabled nudge the Tx scheduler */
8649         if (pf->hw.func_caps.npar_enable && (!i40e_get_npar_bw_setting(pf)))
8650                 i40e_set_npar_bw_setting(pf);
8651
8652 sw_init_done:
8653         return err;
8654 }
8655
8656 /**
8657  * i40e_set_ntuple - set the ntuple feature flag and take action
8658  * @pf: board private structure to initialize
8659  * @features: the feature set that the stack is suggesting
8660  *
8661  * returns a bool to indicate if reset needs to happen
8662  **/
8663 bool i40e_set_ntuple(struct i40e_pf *pf, netdev_features_t features)
8664 {
8665         bool need_reset = false;
8666
8667         /* Check if Flow Director n-tuple support was enabled or disabled.  If
8668          * the state changed, we need to reset.
8669          */
8670         if (features & NETIF_F_NTUPLE) {
8671                 /* Enable filters and mark for reset */
8672                 if (!(pf->flags & I40E_FLAG_FD_SB_ENABLED))
8673                         need_reset = true;
8674                 /* enable FD_SB only if there is MSI-X vector */
8675                 if (pf->num_fdsb_msix > 0)
8676                         pf->flags |= I40E_FLAG_FD_SB_ENABLED;
8677         } else {
8678                 /* turn off filters, mark for reset and clear SW filter list */
8679                 if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
8680                         need_reset = true;
8681                         i40e_fdir_filter_exit(pf);
8682                 }
8683                 pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
8684                 pf->auto_disable_flags &= ~I40E_FLAG_FD_SB_ENABLED;
8685                 /* reset fd counters */
8686                 pf->fd_add_err = pf->fd_atr_cnt = pf->fd_tcp_rule = 0;
8687                 pf->fdir_pf_active_filters = 0;
8688                 pf->flags |= I40E_FLAG_FD_ATR_ENABLED;
8689                 if (I40E_DEBUG_FD & pf->hw.debug_mask)
8690                         dev_info(&pf->pdev->dev, "ATR re-enabled.\n");
8691                 /* if ATR was auto disabled it can be re-enabled. */
8692                 if ((pf->flags & I40E_FLAG_FD_ATR_ENABLED) &&
8693                     (pf->auto_disable_flags & I40E_FLAG_FD_ATR_ENABLED))
8694                         pf->auto_disable_flags &= ~I40E_FLAG_FD_ATR_ENABLED;
8695         }
8696         return need_reset;
8697 }
8698
8699 /**
8700  * i40e_clear_rss_lut - clear the rx hash lookup table
8701  * @vsi: the VSI being configured
8702  **/
8703 static void i40e_clear_rss_lut(struct i40e_vsi *vsi)
8704 {
8705         struct i40e_pf *pf = vsi->back;
8706         struct i40e_hw *hw = &pf->hw;
8707         u16 vf_id = vsi->vf_id;
8708         u8 i;
8709
8710         if (vsi->type == I40E_VSI_MAIN) {
8711                 for (i = 0; i <= I40E_PFQF_HLUT_MAX_INDEX; i++)
8712                         wr32(hw, I40E_PFQF_HLUT(i), 0);
8713         } else if (vsi->type == I40E_VSI_SRIOV) {
8714                 for (i = 0; i <= I40E_VFQF_HLUT_MAX_INDEX; i++)
8715                         i40e_write_rx_ctl(hw, I40E_VFQF_HLUT1(i, vf_id), 0);
8716         } else {
8717                 dev_err(&pf->pdev->dev, "Cannot set RSS LUT - invalid VSI type\n");
8718         }
8719 }
8720
8721 /**
8722  * i40e_set_features - set the netdev feature flags
8723  * @netdev: ptr to the netdev being adjusted
8724  * @features: the feature set that the stack is suggesting
8725  **/
8726 static int i40e_set_features(struct net_device *netdev,
8727                              netdev_features_t features)
8728 {
8729         struct i40e_netdev_priv *np = netdev_priv(netdev);
8730         struct i40e_vsi *vsi = np->vsi;
8731         struct i40e_pf *pf = vsi->back;
8732         bool need_reset;
8733
8734         if (features & NETIF_F_RXHASH && !(netdev->features & NETIF_F_RXHASH))
8735                 i40e_pf_config_rss(pf);
8736         else if (!(features & NETIF_F_RXHASH) &&
8737                  netdev->features & NETIF_F_RXHASH)
8738                 i40e_clear_rss_lut(vsi);
8739
8740         if (features & NETIF_F_HW_VLAN_CTAG_RX)
8741                 i40e_vlan_stripping_enable(vsi);
8742         else
8743                 i40e_vlan_stripping_disable(vsi);
8744
8745         need_reset = i40e_set_ntuple(pf, features);
8746
8747         if (need_reset)
8748                 i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
8749
8750         return 0;
8751 }
8752
8753 /**
8754  * i40e_get_udp_port_idx - Lookup a possibly offloaded for Rx UDP port
8755  * @pf: board private structure
8756  * @port: The UDP port to look up
8757  *
8758  * Returns the index number or I40E_MAX_PF_UDP_OFFLOAD_PORTS if port not found
8759  **/
8760 static u8 i40e_get_udp_port_idx(struct i40e_pf *pf, __be16 port)
8761 {
8762         u8 i;
8763
8764         for (i = 0; i < I40E_MAX_PF_UDP_OFFLOAD_PORTS; i++) {
8765                 if (pf->udp_ports[i].index == port)
8766                         return i;
8767         }
8768
8769         return i;
8770 }
8771
8772 /**
8773  * i40e_udp_tunnel_add - Get notifications about UDP tunnel ports that come up
8774  * @netdev: This physical port's netdev
8775  * @ti: Tunnel endpoint information
8776  **/
8777 static void i40e_udp_tunnel_add(struct net_device *netdev,
8778                                 struct udp_tunnel_info *ti)
8779 {
8780         struct i40e_netdev_priv *np = netdev_priv(netdev);
8781         struct i40e_vsi *vsi = np->vsi;
8782         struct i40e_pf *pf = vsi->back;
8783         __be16 port = ti->port;
8784         u8 next_idx;
8785         u8 idx;
8786
8787         idx = i40e_get_udp_port_idx(pf, port);
8788
8789         /* Check if port already exists */
8790         if (idx < I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8791                 netdev_info(netdev, "port %d already offloaded\n",
8792                             ntohs(port));
8793                 return;
8794         }
8795
8796         /* Now check if there is space to add the new port */
8797         next_idx = i40e_get_udp_port_idx(pf, 0);
8798
8799         if (next_idx == I40E_MAX_PF_UDP_OFFLOAD_PORTS) {
8800                 netdev_info(netdev, "maximum number of offloaded UDP ports reached, not adding port %d\n",
8801                             ntohs(port));
8802                 return;
8803         }
8804
8805         switch (ti->type) {
8806         case UDP_TUNNEL_TYPE_VXLAN:
8807                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_VXLAN;
8808                 break;
8809         case UDP_TUNNEL_TYPE_GENEVE:
8810                 if (!(pf->flags & I40E_FLAG_GENEVE_OFFLOAD_CAPABLE))
8811                         return;
8812                 pf->udp_ports[next_idx].type = I40E_AQC_TUNNEL_TYPE_NGE;
8813                 break;
8814         default:
8815                 return;
8816         }
8817
8818         /* New port: add it and mark its index in the bitmap */
8819         pf->udp_ports[next_idx].index = port;
8820         pf->pending_udp_bitmap |= BIT_ULL(next_idx);
8821         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8822 }
8823
8824 /**
8825  * i40e_udp_tunnel_del - Get notifications about UDP tunnel ports that go away
8826  * @netdev: This physical port's netdev
8827  * @ti: Tunnel endpoint information
8828  **/
8829 static void i40e_udp_tunnel_del(struct net_device *netdev,
8830                                 struct udp_tunnel_info *ti)
8831 {
8832         struct i40e_netdev_priv *np = netdev_priv(netdev);
8833         struct i40e_vsi *vsi = np->vsi;
8834         struct i40e_pf *pf = vsi->back;
8835         __be16 port = ti->port;
8836         u8 idx;
8837
8838         idx = i40e_get_udp_port_idx(pf, port);
8839
8840         /* Check if port already exists */
8841         if (idx >= I40E_MAX_PF_UDP_OFFLOAD_PORTS)
8842                 goto not_found;
8843
8844         switch (ti->type) {
8845         case UDP_TUNNEL_TYPE_VXLAN:
8846                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_VXLAN)
8847                         goto not_found;
8848                 break;
8849         case UDP_TUNNEL_TYPE_GENEVE:
8850                 if (pf->udp_ports[idx].type != I40E_AQC_TUNNEL_TYPE_NGE)
8851                         goto not_found;
8852                 break;
8853         default:
8854                 goto not_found;
8855         }
8856
8857         /* if port exists, set it to 0 (mark for deletion)
8858          * and make it pending
8859          */
8860         pf->udp_ports[idx].index = 0;
8861         pf->pending_udp_bitmap |= BIT_ULL(idx);
8862         pf->flags |= I40E_FLAG_UDP_FILTER_SYNC;
8863
8864         return;
8865 not_found:
8866         netdev_warn(netdev, "UDP port %d was not found, not deleting\n",
8867                     ntohs(port));
8868 }
8869
8870 static int i40e_get_phys_port_id(struct net_device *netdev,
8871                                  struct netdev_phys_item_id *ppid)
8872 {
8873         struct i40e_netdev_priv *np = netdev_priv(netdev);
8874         struct i40e_pf *pf = np->vsi->back;
8875         struct i40e_hw *hw = &pf->hw;
8876
8877         if (!(pf->flags & I40E_FLAG_PORT_ID_VALID))
8878                 return -EOPNOTSUPP;
8879
8880         ppid->id_len = min_t(int, sizeof(hw->mac.port_addr), sizeof(ppid->id));
8881         memcpy(ppid->id, hw->mac.port_addr, ppid->id_len);
8882
8883         return 0;
8884 }
8885
8886 /**
8887  * i40e_ndo_fdb_add - add an entry to the hardware database
8888  * @ndm: the input from the stack
8889  * @tb: pointer to array of nladdr (unused)
8890  * @dev: the net device pointer
8891  * @addr: the MAC address entry being added
8892  * @flags: instructions from stack about fdb operation
8893  */
8894 static int i40e_ndo_fdb_add(struct ndmsg *ndm, struct nlattr *tb[],
8895                             struct net_device *dev,
8896                             const unsigned char *addr, u16 vid,
8897                             u16 flags)
8898 {
8899         struct i40e_netdev_priv *np = netdev_priv(dev);
8900         struct i40e_pf *pf = np->vsi->back;
8901         int err = 0;
8902
8903         if (!(pf->flags & I40E_FLAG_SRIOV_ENABLED))
8904                 return -EOPNOTSUPP;
8905
8906         if (vid) {
8907                 pr_info("%s: vlans aren't supported yet for dev_uc|mc_add()\n", dev->name);
8908                 return -EINVAL;
8909         }
8910
8911         /* Hardware does not support aging addresses so if a
8912          * ndm_state is given only allow permanent addresses
8913          */
8914         if (ndm->ndm_state && !(ndm->ndm_state & NUD_PERMANENT)) {
8915                 netdev_info(dev, "FDB only supports static addresses\n");
8916                 return -EINVAL;
8917         }
8918
8919         if (is_unicast_ether_addr(addr) || is_link_local_ether_addr(addr))
8920                 err = dev_uc_add_excl(dev, addr);
8921         else if (is_multicast_ether_addr(addr))
8922                 err = dev_mc_add_excl(dev, addr);
8923         else
8924                 err = -EINVAL;
8925
8926         /* Only return duplicate errors if NLM_F_EXCL is set */
8927         if (err == -EEXIST && !(flags & NLM_F_EXCL))
8928                 err = 0;
8929
8930         return err;
8931 }
8932
8933 /**
8934  * i40e_ndo_bridge_setlink - Set the hardware bridge mode
8935  * @dev: the netdev being configured
8936  * @nlh: RTNL message
8937  *
8938  * Inserts a new hardware bridge if not already created and
8939  * enables the bridging mode requested (VEB or VEPA). If the
8940  * hardware bridge has already been inserted and the request
8941  * is to change the mode then that requires a PF reset to
8942  * allow rebuild of the components with required hardware
8943  * bridge mode enabled.
8944  **/
8945 static int i40e_ndo_bridge_setlink(struct net_device *dev,
8946                                    struct nlmsghdr *nlh,
8947                                    u16 flags)
8948 {
8949         struct i40e_netdev_priv *np = netdev_priv(dev);
8950         struct i40e_vsi *vsi = np->vsi;
8951         struct i40e_pf *pf = vsi->back;
8952         struct i40e_veb *veb = NULL;
8953         struct nlattr *attr, *br_spec;
8954         int i, rem;
8955
8956         /* Only for PF VSI for now */
8957         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
8958                 return -EOPNOTSUPP;
8959
8960         /* Find the HW bridge for PF VSI */
8961         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
8962                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
8963                         veb = pf->veb[i];
8964         }
8965
8966         br_spec = nlmsg_find_attr(nlh, sizeof(struct ifinfomsg), IFLA_AF_SPEC);
8967
8968         nla_for_each_nested(attr, br_spec, rem) {
8969                 __u16 mode;
8970
8971                 if (nla_type(attr) != IFLA_BRIDGE_MODE)
8972                         continue;
8973
8974                 mode = nla_get_u16(attr);
8975                 if ((mode != BRIDGE_MODE_VEPA) &&
8976                     (mode != BRIDGE_MODE_VEB))
8977                         return -EINVAL;
8978
8979                 /* Insert a new HW bridge */
8980                 if (!veb) {
8981                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
8982                                              vsi->tc_config.enabled_tc);
8983                         if (veb) {
8984                                 veb->bridge_mode = mode;
8985                                 i40e_config_bridge_mode(veb);
8986                         } else {
8987                                 /* No Bridge HW offload available */
8988                                 return -ENOENT;
8989                         }
8990                         break;
8991                 } else if (mode != veb->bridge_mode) {
8992                         /* Existing HW bridge but different mode needs reset */
8993                         veb->bridge_mode = mode;
8994                         /* TODO: If no VFs or VMDq VSIs, disallow VEB mode */
8995                         if (mode == BRIDGE_MODE_VEB)
8996                                 pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
8997                         else
8998                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
8999                         i40e_do_reset(pf, BIT_ULL(__I40E_PF_RESET_REQUESTED));
9000                         break;
9001                 }
9002         }
9003
9004         return 0;
9005 }
9006
9007 /**
9008  * i40e_ndo_bridge_getlink - Get the hardware bridge mode
9009  * @skb: skb buff
9010  * @pid: process id
9011  * @seq: RTNL message seq #
9012  * @dev: the netdev being configured
9013  * @filter_mask: unused
9014  * @nlflags: netlink flags passed in
9015  *
9016  * Return the mode in which the hardware bridge is operating in
9017  * i.e VEB or VEPA.
9018  **/
9019 static int i40e_ndo_bridge_getlink(struct sk_buff *skb, u32 pid, u32 seq,
9020                                    struct net_device *dev,
9021                                    u32 __always_unused filter_mask,
9022                                    int nlflags)
9023 {
9024         struct i40e_netdev_priv *np = netdev_priv(dev);
9025         struct i40e_vsi *vsi = np->vsi;
9026         struct i40e_pf *pf = vsi->back;
9027         struct i40e_veb *veb = NULL;
9028         int i;
9029
9030         /* Only for PF VSI for now */
9031         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid)
9032                 return -EOPNOTSUPP;
9033
9034         /* Find the HW bridge for the PF VSI */
9035         for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9036                 if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9037                         veb = pf->veb[i];
9038         }
9039
9040         if (!veb)
9041                 return 0;
9042
9043         return ndo_dflt_bridge_getlink(skb, pid, seq, dev, veb->bridge_mode,
9044                                        nlflags, 0, 0, filter_mask, NULL);
9045 }
9046
9047 /* Hardware supports L4 tunnel length of 128B (=2^7) which includes
9048  * inner mac plus all inner ethertypes.
9049  */
9050 #define I40E_MAX_TUNNEL_HDR_LEN 128
9051 /**
9052  * i40e_features_check - Validate encapsulated packet conforms to limits
9053  * @skb: skb buff
9054  * @dev: This physical port's netdev
9055  * @features: Offload features that the stack believes apply
9056  **/
9057 static netdev_features_t i40e_features_check(struct sk_buff *skb,
9058                                              struct net_device *dev,
9059                                              netdev_features_t features)
9060 {
9061         if (skb->encapsulation &&
9062             ((skb_inner_network_header(skb) - skb_transport_header(skb)) >
9063              I40E_MAX_TUNNEL_HDR_LEN))
9064                 return features & ~(NETIF_F_CSUM_MASK | NETIF_F_GSO_MASK);
9065
9066         return features;
9067 }
9068
9069 static const struct net_device_ops i40e_netdev_ops = {
9070         .ndo_open               = i40e_open,
9071         .ndo_stop               = i40e_close,
9072         .ndo_start_xmit         = i40e_lan_xmit_frame,
9073         .ndo_get_stats64        = i40e_get_netdev_stats_struct,
9074         .ndo_set_rx_mode        = i40e_set_rx_mode,
9075         .ndo_validate_addr      = eth_validate_addr,
9076         .ndo_set_mac_address    = i40e_set_mac,
9077         .ndo_change_mtu         = i40e_change_mtu,
9078         .ndo_do_ioctl           = i40e_ioctl,
9079         .ndo_tx_timeout         = i40e_tx_timeout,
9080         .ndo_vlan_rx_add_vid    = i40e_vlan_rx_add_vid,
9081         .ndo_vlan_rx_kill_vid   = i40e_vlan_rx_kill_vid,
9082 #ifdef CONFIG_NET_POLL_CONTROLLER
9083         .ndo_poll_controller    = i40e_netpoll,
9084 #endif
9085         .ndo_setup_tc           = __i40e_setup_tc,
9086 #ifdef I40E_FCOE
9087         .ndo_fcoe_enable        = i40e_fcoe_enable,
9088         .ndo_fcoe_disable       = i40e_fcoe_disable,
9089 #endif
9090         .ndo_set_features       = i40e_set_features,
9091         .ndo_set_vf_mac         = i40e_ndo_set_vf_mac,
9092         .ndo_set_vf_vlan        = i40e_ndo_set_vf_port_vlan,
9093         .ndo_set_vf_rate        = i40e_ndo_set_vf_bw,
9094         .ndo_get_vf_config      = i40e_ndo_get_vf_config,
9095         .ndo_set_vf_link_state  = i40e_ndo_set_vf_link_state,
9096         .ndo_set_vf_spoofchk    = i40e_ndo_set_vf_spoofchk,
9097         .ndo_set_vf_trust       = i40e_ndo_set_vf_trust,
9098         .ndo_udp_tunnel_add     = i40e_udp_tunnel_add,
9099         .ndo_udp_tunnel_del     = i40e_udp_tunnel_del,
9100         .ndo_get_phys_port_id   = i40e_get_phys_port_id,
9101         .ndo_fdb_add            = i40e_ndo_fdb_add,
9102         .ndo_features_check     = i40e_features_check,
9103         .ndo_bridge_getlink     = i40e_ndo_bridge_getlink,
9104         .ndo_bridge_setlink     = i40e_ndo_bridge_setlink,
9105 };
9106
9107 /**
9108  * i40e_config_netdev - Setup the netdev flags
9109  * @vsi: the VSI being configured
9110  *
9111  * Returns 0 on success, negative value on failure
9112  **/
9113 static int i40e_config_netdev(struct i40e_vsi *vsi)
9114 {
9115         struct i40e_pf *pf = vsi->back;
9116         struct i40e_hw *hw = &pf->hw;
9117         struct i40e_netdev_priv *np;
9118         struct net_device *netdev;
9119         u8 mac_addr[ETH_ALEN];
9120         int etherdev_size;
9121
9122         etherdev_size = sizeof(struct i40e_netdev_priv);
9123         netdev = alloc_etherdev_mq(etherdev_size, vsi->alloc_queue_pairs);
9124         if (!netdev)
9125                 return -ENOMEM;
9126
9127         vsi->netdev = netdev;
9128         np = netdev_priv(netdev);
9129         np->vsi = vsi;
9130
9131         netdev->hw_enc_features |= NETIF_F_SG                   |
9132                                    NETIF_F_IP_CSUM              |
9133                                    NETIF_F_IPV6_CSUM            |
9134                                    NETIF_F_HIGHDMA              |
9135                                    NETIF_F_SOFT_FEATURES        |
9136                                    NETIF_F_TSO                  |
9137                                    NETIF_F_TSO_ECN              |
9138                                    NETIF_F_TSO6                 |
9139                                    NETIF_F_GSO_GRE              |
9140                                    NETIF_F_GSO_GRE_CSUM         |
9141                                    NETIF_F_GSO_IPXIP4           |
9142                                    NETIF_F_GSO_IPXIP6           |
9143                                    NETIF_F_GSO_UDP_TUNNEL       |
9144                                    NETIF_F_GSO_UDP_TUNNEL_CSUM  |
9145                                    NETIF_F_GSO_PARTIAL          |
9146                                    NETIF_F_SCTP_CRC             |
9147                                    NETIF_F_RXHASH               |
9148                                    NETIF_F_RXCSUM               |
9149                                    0;
9150
9151         if (!(pf->flags & I40E_FLAG_OUTER_UDP_CSUM_CAPABLE))
9152                 netdev->gso_partial_features |= NETIF_F_GSO_UDP_TUNNEL_CSUM;
9153
9154         netdev->gso_partial_features |= NETIF_F_GSO_GRE_CSUM;
9155
9156         /* record features VLANs can make use of */
9157         netdev->vlan_features |= netdev->hw_enc_features |
9158                                  NETIF_F_TSO_MANGLEID;
9159
9160         if (!(pf->flags & I40E_FLAG_MFP_ENABLED))
9161                 netdev->hw_features |= NETIF_F_NTUPLE;
9162
9163         netdev->hw_features |= netdev->hw_enc_features  |
9164                                NETIF_F_HW_VLAN_CTAG_TX  |
9165                                NETIF_F_HW_VLAN_CTAG_RX;
9166
9167         netdev->features |= netdev->hw_features | NETIF_F_HW_VLAN_CTAG_FILTER;
9168         netdev->hw_enc_features |= NETIF_F_TSO_MANGLEID;
9169
9170         if (vsi->type == I40E_VSI_MAIN) {
9171                 SET_NETDEV_DEV(netdev, &pf->pdev->dev);
9172                 ether_addr_copy(mac_addr, hw->mac.perm_addr);
9173                 /* The following steps are necessary to prevent reception
9174                  * of tagged packets - some older NVM configurations load a
9175                  * default a MAC-VLAN filter that accepts any tagged packet
9176                  * which must be replaced by a normal filter.
9177                  */
9178                 i40e_rm_default_mac_filter(vsi, mac_addr);
9179                 spin_lock_bh(&vsi->mac_filter_list_lock);
9180                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, true);
9181                 spin_unlock_bh(&vsi->mac_filter_list_lock);
9182         } else {
9183                 /* relate the VSI_VMDQ name to the VSI_MAIN name */
9184                 snprintf(netdev->name, IFNAMSIZ, "%sv%%d",
9185                          pf->vsi[pf->lan_vsi]->netdev->name);
9186                 random_ether_addr(mac_addr);
9187
9188                 spin_lock_bh(&vsi->mac_filter_list_lock);
9189                 i40e_add_filter(vsi, mac_addr, I40E_VLAN_ANY, false, false);
9190                 spin_unlock_bh(&vsi->mac_filter_list_lock);
9191         }
9192
9193         ether_addr_copy(netdev->dev_addr, mac_addr);
9194         ether_addr_copy(netdev->perm_addr, mac_addr);
9195
9196         netdev->priv_flags |= IFF_UNICAST_FLT;
9197         netdev->priv_flags |= IFF_SUPP_NOFCS;
9198         /* Setup netdev TC information */
9199         i40e_vsi_config_netdev_tc(vsi, vsi->tc_config.enabled_tc);
9200
9201         netdev->netdev_ops = &i40e_netdev_ops;
9202         netdev->watchdog_timeo = 5 * HZ;
9203         i40e_set_ethtool_ops(netdev);
9204 #ifdef I40E_FCOE
9205         i40e_fcoe_config_netdev(netdev, vsi);
9206 #endif
9207
9208         return 0;
9209 }
9210
9211 /**
9212  * i40e_vsi_delete - Delete a VSI from the switch
9213  * @vsi: the VSI being removed
9214  *
9215  * Returns 0 on success, negative value on failure
9216  **/
9217 static void i40e_vsi_delete(struct i40e_vsi *vsi)
9218 {
9219         /* remove default VSI is not allowed */
9220         if (vsi == vsi->back->vsi[vsi->back->lan_vsi])
9221                 return;
9222
9223         i40e_aq_delete_element(&vsi->back->hw, vsi->seid, NULL);
9224 }
9225
9226 /**
9227  * i40e_is_vsi_uplink_mode_veb - Check if the VSI's uplink bridge mode is VEB
9228  * @vsi: the VSI being queried
9229  *
9230  * Returns 1 if HW bridge mode is VEB and return 0 in case of VEPA mode
9231  **/
9232 int i40e_is_vsi_uplink_mode_veb(struct i40e_vsi *vsi)
9233 {
9234         struct i40e_veb *veb;
9235         struct i40e_pf *pf = vsi->back;
9236
9237         /* Uplink is not a bridge so default to VEB */
9238         if (vsi->veb_idx == I40E_NO_VEB)
9239                 return 1;
9240
9241         veb = pf->veb[vsi->veb_idx];
9242         if (!veb) {
9243                 dev_info(&pf->pdev->dev,
9244                          "There is no veb associated with the bridge\n");
9245                 return -ENOENT;
9246         }
9247
9248         /* Uplink is a bridge in VEPA mode */
9249         if (veb->bridge_mode & BRIDGE_MODE_VEPA) {
9250                 return 0;
9251         } else {
9252                 /* Uplink is a bridge in VEB mode */
9253                 return 1;
9254         }
9255
9256         /* VEPA is now default bridge, so return 0 */
9257         return 0;
9258 }
9259
9260 /**
9261  * i40e_add_vsi - Add a VSI to the switch
9262  * @vsi: the VSI being configured
9263  *
9264  * This initializes a VSI context depending on the VSI type to be added and
9265  * passes it down to the add_vsi aq command.
9266  **/
9267 static int i40e_add_vsi(struct i40e_vsi *vsi)
9268 {
9269         int ret = -ENODEV;
9270         i40e_status aq_ret = 0;
9271         struct i40e_pf *pf = vsi->back;
9272         struct i40e_hw *hw = &pf->hw;
9273         struct i40e_vsi_context ctxt;
9274         struct i40e_mac_filter *f, *ftmp;
9275
9276         u8 enabled_tc = 0x1; /* TC0 enabled */
9277         int f_count = 0;
9278
9279         memset(&ctxt, 0, sizeof(ctxt));
9280         switch (vsi->type) {
9281         case I40E_VSI_MAIN:
9282                 /* The PF's main VSI is already setup as part of the
9283                  * device initialization, so we'll not bother with
9284                  * the add_vsi call, but we will retrieve the current
9285                  * VSI context.
9286                  */
9287                 ctxt.seid = pf->main_vsi_seid;
9288                 ctxt.pf_num = pf->hw.pf_id;
9289                 ctxt.vf_num = 0;
9290                 ret = i40e_aq_get_vsi_params(&pf->hw, &ctxt, NULL);
9291                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9292                 if (ret) {
9293                         dev_info(&pf->pdev->dev,
9294                                  "couldn't get PF vsi config, err %s aq_err %s\n",
9295                                  i40e_stat_str(&pf->hw, ret),
9296                                  i40e_aq_str(&pf->hw,
9297                                              pf->hw.aq.asq_last_status));
9298                         return -ENOENT;
9299                 }
9300                 vsi->info = ctxt.info;
9301                 vsi->info.valid_sections = 0;
9302
9303                 vsi->seid = ctxt.seid;
9304                 vsi->id = ctxt.vsi_number;
9305
9306                 enabled_tc = i40e_pf_get_tc_map(pf);
9307
9308                 /* MFP mode setup queue map and update VSI */
9309                 if ((pf->flags & I40E_FLAG_MFP_ENABLED) &&
9310                     !(pf->hw.func_caps.iscsi)) { /* NIC type PF */
9311                         memset(&ctxt, 0, sizeof(ctxt));
9312                         ctxt.seid = pf->main_vsi_seid;
9313                         ctxt.pf_num = pf->hw.pf_id;
9314                         ctxt.vf_num = 0;
9315                         i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, false);
9316                         ret = i40e_aq_update_vsi_params(hw, &ctxt, NULL);
9317                         if (ret) {
9318                                 dev_info(&pf->pdev->dev,
9319                                          "update vsi failed, err %s aq_err %s\n",
9320                                          i40e_stat_str(&pf->hw, ret),
9321                                          i40e_aq_str(&pf->hw,
9322                                                     pf->hw.aq.asq_last_status));
9323                                 ret = -ENOENT;
9324                                 goto err;
9325                         }
9326                         /* update the local VSI info queue map */
9327                         i40e_vsi_update_queue_map(vsi, &ctxt);
9328                         vsi->info.valid_sections = 0;
9329                 } else {
9330                         /* Default/Main VSI is only enabled for TC0
9331                          * reconfigure it to enable all TCs that are
9332                          * available on the port in SFP mode.
9333                          * For MFP case the iSCSI PF would use this
9334                          * flow to enable LAN+iSCSI TC.
9335                          */
9336                         ret = i40e_vsi_config_tc(vsi, enabled_tc);
9337                         if (ret) {
9338                                 dev_info(&pf->pdev->dev,
9339                                          "failed to configure TCs for main VSI tc_map 0x%08x, err %s aq_err %s\n",
9340                                          enabled_tc,
9341                                          i40e_stat_str(&pf->hw, ret),
9342                                          i40e_aq_str(&pf->hw,
9343                                                     pf->hw.aq.asq_last_status));
9344                                 ret = -ENOENT;
9345                         }
9346                 }
9347                 break;
9348
9349         case I40E_VSI_FDIR:
9350                 ctxt.pf_num = hw->pf_id;
9351                 ctxt.vf_num = 0;
9352                 ctxt.uplink_seid = vsi->uplink_seid;
9353                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9354                 ctxt.flags = I40E_AQ_VSI_TYPE_PF;
9355                 if ((pf->flags & I40E_FLAG_VEB_MODE_ENABLED) &&
9356                     (i40e_is_vsi_uplink_mode_veb(vsi))) {
9357                         ctxt.info.valid_sections |=
9358                              cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9359                         ctxt.info.switch_id =
9360                            cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9361                 }
9362                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9363                 break;
9364
9365         case I40E_VSI_VMDQ2:
9366                 ctxt.pf_num = hw->pf_id;
9367                 ctxt.vf_num = 0;
9368                 ctxt.uplink_seid = vsi->uplink_seid;
9369                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9370                 ctxt.flags = I40E_AQ_VSI_TYPE_VMDQ2;
9371
9372                 /* This VSI is connected to VEB so the switch_id
9373                  * should be set to zero by default.
9374                  */
9375                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9376                         ctxt.info.valid_sections |=
9377                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9378                         ctxt.info.switch_id =
9379                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9380                 }
9381
9382                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9383                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9384                 break;
9385
9386         case I40E_VSI_SRIOV:
9387                 ctxt.pf_num = hw->pf_id;
9388                 ctxt.vf_num = vsi->vf_id + hw->func_caps.vf_base_id;
9389                 ctxt.uplink_seid = vsi->uplink_seid;
9390                 ctxt.connection_type = I40E_AQ_VSI_CONN_TYPE_NORMAL;
9391                 ctxt.flags = I40E_AQ_VSI_TYPE_VF;
9392
9393                 /* This VSI is connected to VEB so the switch_id
9394                  * should be set to zero by default.
9395                  */
9396                 if (i40e_is_vsi_uplink_mode_veb(vsi)) {
9397                         ctxt.info.valid_sections |=
9398                                 cpu_to_le16(I40E_AQ_VSI_PROP_SWITCH_VALID);
9399                         ctxt.info.switch_id =
9400                                 cpu_to_le16(I40E_AQ_VSI_SW_ID_FLAG_ALLOW_LB);
9401                 }
9402
9403                 if (vsi->back->flags & I40E_FLAG_IWARP_ENABLED) {
9404                         ctxt.info.valid_sections |=
9405                                 cpu_to_le16(I40E_AQ_VSI_PROP_QUEUE_OPT_VALID);
9406                         ctxt.info.queueing_opt_flags |=
9407                                 (I40E_AQ_VSI_QUE_OPT_TCP_ENA |
9408                                  I40E_AQ_VSI_QUE_OPT_RSS_LUT_VSI);
9409                 }
9410
9411                 ctxt.info.valid_sections |= cpu_to_le16(I40E_AQ_VSI_PROP_VLAN_VALID);
9412                 ctxt.info.port_vlan_flags |= I40E_AQ_VSI_PVLAN_MODE_ALL;
9413                 if (pf->vf[vsi->vf_id].spoofchk) {
9414                         ctxt.info.valid_sections |=
9415                                 cpu_to_le16(I40E_AQ_VSI_PROP_SECURITY_VALID);
9416                         ctxt.info.sec_flags |=
9417                                 (I40E_AQ_VSI_SEC_FLAG_ENABLE_VLAN_CHK |
9418                                  I40E_AQ_VSI_SEC_FLAG_ENABLE_MAC_CHK);
9419                 }
9420                 /* Setup the VSI tx/rx queue map for TC0 only for now */
9421                 i40e_vsi_setup_queue_map(vsi, &ctxt, enabled_tc, true);
9422                 break;
9423
9424 #ifdef I40E_FCOE
9425         case I40E_VSI_FCOE:
9426                 ret = i40e_fcoe_vsi_init(vsi, &ctxt);
9427                 if (ret) {
9428                         dev_info(&pf->pdev->dev, "failed to initialize FCoE VSI\n");
9429                         return ret;
9430                 }
9431                 break;
9432
9433 #endif /* I40E_FCOE */
9434         case I40E_VSI_IWARP:
9435                 /* send down message to iWARP */
9436                 break;
9437
9438         default:
9439                 return -ENODEV;
9440         }
9441
9442         if (vsi->type != I40E_VSI_MAIN) {
9443                 ret = i40e_aq_add_vsi(hw, &ctxt, NULL);
9444                 if (ret) {
9445                         dev_info(&vsi->back->pdev->dev,
9446                                  "add vsi failed, err %s aq_err %s\n",
9447                                  i40e_stat_str(&pf->hw, ret),
9448                                  i40e_aq_str(&pf->hw,
9449                                              pf->hw.aq.asq_last_status));
9450                         ret = -ENOENT;
9451                         goto err;
9452                 }
9453                 vsi->info = ctxt.info;
9454                 vsi->info.valid_sections = 0;
9455                 vsi->seid = ctxt.seid;
9456                 vsi->id = ctxt.vsi_number;
9457         }
9458         /* Except FDIR VSI, for all othet VSI set the broadcast filter */
9459         if (vsi->type != I40E_VSI_FDIR) {
9460                 aq_ret = i40e_aq_set_vsi_broadcast(hw, vsi->seid, true, NULL);
9461                 if (aq_ret) {
9462                         ret = i40e_aq_rc_to_posix(aq_ret,
9463                                                   hw->aq.asq_last_status);
9464                         dev_info(&pf->pdev->dev,
9465                                  "set brdcast promisc failed, err %s, aq_err %s\n",
9466                                  i40e_stat_str(hw, aq_ret),
9467                                  i40e_aq_str(hw, hw->aq.asq_last_status));
9468                 }
9469         }
9470
9471         vsi->active_filters = 0;
9472         clear_bit(__I40E_FILTER_OVERFLOW_PROMISC, &vsi->state);
9473         spin_lock_bh(&vsi->mac_filter_list_lock);
9474         /* If macvlan filters already exist, force them to get loaded */
9475         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list) {
9476                 f->state = I40E_FILTER_NEW;
9477                 f_count++;
9478         }
9479         spin_unlock_bh(&vsi->mac_filter_list_lock);
9480
9481         if (f_count) {
9482                 vsi->flags |= I40E_VSI_FLAG_FILTER_CHANGED;
9483                 pf->flags |= I40E_FLAG_FILTER_SYNC;
9484         }
9485
9486         /* Update VSI BW information */
9487         ret = i40e_vsi_get_bw_info(vsi);
9488         if (ret) {
9489                 dev_info(&pf->pdev->dev,
9490                          "couldn't get vsi bw info, err %s aq_err %s\n",
9491                          i40e_stat_str(&pf->hw, ret),
9492                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
9493                 /* VSI is already added so not tearing that up */
9494                 ret = 0;
9495         }
9496
9497 err:
9498         return ret;
9499 }
9500
9501 /**
9502  * i40e_vsi_release - Delete a VSI and free its resources
9503  * @vsi: the VSI being removed
9504  *
9505  * Returns 0 on success or < 0 on error
9506  **/
9507 int i40e_vsi_release(struct i40e_vsi *vsi)
9508 {
9509         struct i40e_mac_filter *f, *ftmp;
9510         struct i40e_veb *veb = NULL;
9511         struct i40e_pf *pf;
9512         u16 uplink_seid;
9513         int i, n;
9514
9515         pf = vsi->back;
9516
9517         /* release of a VEB-owner or last VSI is not allowed */
9518         if (vsi->flags & I40E_VSI_FLAG_VEB_OWNER) {
9519                 dev_info(&pf->pdev->dev, "VSI %d has existing VEB %d\n",
9520                          vsi->seid, vsi->uplink_seid);
9521                 return -ENODEV;
9522         }
9523         if (vsi == pf->vsi[pf->lan_vsi] &&
9524             !test_bit(__I40E_DOWN, &pf->state)) {
9525                 dev_info(&pf->pdev->dev, "Can't remove PF VSI\n");
9526                 return -ENODEV;
9527         }
9528
9529         uplink_seid = vsi->uplink_seid;
9530         if (vsi->type != I40E_VSI_SRIOV) {
9531                 if (vsi->netdev_registered) {
9532                         vsi->netdev_registered = false;
9533                         if (vsi->netdev) {
9534                                 /* results in a call to i40e_close() */
9535                                 unregister_netdev(vsi->netdev);
9536                         }
9537                 } else {
9538                         i40e_vsi_close(vsi);
9539                 }
9540                 i40e_vsi_disable_irq(vsi);
9541         }
9542
9543         spin_lock_bh(&vsi->mac_filter_list_lock);
9544         list_for_each_entry_safe(f, ftmp, &vsi->mac_filter_list, list)
9545                 i40e_del_filter(vsi, f->macaddr, f->vlan,
9546                                 f->is_vf, f->is_netdev);
9547         spin_unlock_bh(&vsi->mac_filter_list_lock);
9548
9549         i40e_sync_vsi_filters(vsi);
9550
9551         i40e_vsi_delete(vsi);
9552         i40e_vsi_free_q_vectors(vsi);
9553         if (vsi->netdev) {
9554                 free_netdev(vsi->netdev);
9555                 vsi->netdev = NULL;
9556         }
9557         i40e_vsi_clear_rings(vsi);
9558         i40e_vsi_clear(vsi);
9559
9560         /* If this was the last thing on the VEB, except for the
9561          * controlling VSI, remove the VEB, which puts the controlling
9562          * VSI onto the next level down in the switch.
9563          *
9564          * Well, okay, there's one more exception here: don't remove
9565          * the orphan VEBs yet.  We'll wait for an explicit remove request
9566          * from up the network stack.
9567          */
9568         for (n = 0, i = 0; i < pf->num_alloc_vsi; i++) {
9569                 if (pf->vsi[i] &&
9570                     pf->vsi[i]->uplink_seid == uplink_seid &&
9571                     (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
9572                         n++;      /* count the VSIs */
9573                 }
9574         }
9575         for (i = 0; i < I40E_MAX_VEB; i++) {
9576                 if (!pf->veb[i])
9577                         continue;
9578                 if (pf->veb[i]->uplink_seid == uplink_seid)
9579                         n++;     /* count the VEBs */
9580                 if (pf->veb[i]->seid == uplink_seid)
9581                         veb = pf->veb[i];
9582         }
9583         if (n == 0 && veb && veb->uplink_seid != 0)
9584                 i40e_veb_release(veb);
9585
9586         return 0;
9587 }
9588
9589 /**
9590  * i40e_vsi_setup_vectors - Set up the q_vectors for the given VSI
9591  * @vsi: ptr to the VSI
9592  *
9593  * This should only be called after i40e_vsi_mem_alloc() which allocates the
9594  * corresponding SW VSI structure and initializes num_queue_pairs for the
9595  * newly allocated VSI.
9596  *
9597  * Returns 0 on success or negative on failure
9598  **/
9599 static int i40e_vsi_setup_vectors(struct i40e_vsi *vsi)
9600 {
9601         int ret = -ENOENT;
9602         struct i40e_pf *pf = vsi->back;
9603
9604         if (vsi->q_vectors[0]) {
9605                 dev_info(&pf->pdev->dev, "VSI %d has existing q_vectors\n",
9606                          vsi->seid);
9607                 return -EEXIST;
9608         }
9609
9610         if (vsi->base_vector) {
9611                 dev_info(&pf->pdev->dev, "VSI %d has non-zero base vector %d\n",
9612                          vsi->seid, vsi->base_vector);
9613                 return -EEXIST;
9614         }
9615
9616         ret = i40e_vsi_alloc_q_vectors(vsi);
9617         if (ret) {
9618                 dev_info(&pf->pdev->dev,
9619                          "failed to allocate %d q_vector for VSI %d, ret=%d\n",
9620                          vsi->num_q_vectors, vsi->seid, ret);
9621                 vsi->num_q_vectors = 0;
9622                 goto vector_setup_out;
9623         }
9624
9625         /* In Legacy mode, we do not have to get any other vector since we
9626          * piggyback on the misc/ICR0 for queue interrupts.
9627         */
9628         if (!(pf->flags & I40E_FLAG_MSIX_ENABLED))
9629                 return ret;
9630         if (vsi->num_q_vectors)
9631                 vsi->base_vector = i40e_get_lump(pf, pf->irq_pile,
9632                                                  vsi->num_q_vectors, vsi->idx);
9633         if (vsi->base_vector < 0) {
9634                 dev_info(&pf->pdev->dev,
9635                          "failed to get tracking for %d vectors for VSI %d, err=%d\n",
9636                          vsi->num_q_vectors, vsi->seid, vsi->base_vector);
9637                 i40e_vsi_free_q_vectors(vsi);
9638                 ret = -ENOENT;
9639                 goto vector_setup_out;
9640         }
9641
9642 vector_setup_out:
9643         return ret;
9644 }
9645
9646 /**
9647  * i40e_vsi_reinit_setup - return and reallocate resources for a VSI
9648  * @vsi: pointer to the vsi.
9649  *
9650  * This re-allocates a vsi's queue resources.
9651  *
9652  * Returns pointer to the successfully allocated and configured VSI sw struct
9653  * on success, otherwise returns NULL on failure.
9654  **/
9655 static struct i40e_vsi *i40e_vsi_reinit_setup(struct i40e_vsi *vsi)
9656 {
9657         struct i40e_pf *pf;
9658         u8 enabled_tc;
9659         int ret;
9660
9661         if (!vsi)
9662                 return NULL;
9663
9664         pf = vsi->back;
9665
9666         i40e_put_lump(pf->qp_pile, vsi->base_queue, vsi->idx);
9667         i40e_vsi_clear_rings(vsi);
9668
9669         i40e_vsi_free_arrays(vsi, false);
9670         i40e_set_num_rings_in_vsi(vsi);
9671         ret = i40e_vsi_alloc_arrays(vsi, false);
9672         if (ret)
9673                 goto err_vsi;
9674
9675         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs, vsi->idx);
9676         if (ret < 0) {
9677                 dev_info(&pf->pdev->dev,
9678                          "failed to get tracking for %d queues for VSI %d err %d\n",
9679                          vsi->alloc_queue_pairs, vsi->seid, ret);
9680                 goto err_vsi;
9681         }
9682         vsi->base_queue = ret;
9683
9684         /* Update the FW view of the VSI. Force a reset of TC and queue
9685          * layout configurations.
9686          */
9687         enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
9688         pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
9689         pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
9690         i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
9691         if (vsi->type == I40E_VSI_MAIN)
9692                 i40e_rm_default_mac_filter(vsi, pf->hw.mac.perm_addr);
9693
9694         /* assign it some queues */
9695         ret = i40e_alloc_rings(vsi);
9696         if (ret)
9697                 goto err_rings;
9698
9699         /* map all of the rings to the q_vectors */
9700         i40e_vsi_map_rings_to_vectors(vsi);
9701         return vsi;
9702
9703 err_rings:
9704         i40e_vsi_free_q_vectors(vsi);
9705         if (vsi->netdev_registered) {
9706                 vsi->netdev_registered = false;
9707                 unregister_netdev(vsi->netdev);
9708                 free_netdev(vsi->netdev);
9709                 vsi->netdev = NULL;
9710         }
9711         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9712 err_vsi:
9713         i40e_vsi_clear(vsi);
9714         return NULL;
9715 }
9716
9717 /**
9718  * i40e_vsi_setup - Set up a VSI by a given type
9719  * @pf: board private structure
9720  * @type: VSI type
9721  * @uplink_seid: the switch element to link to
9722  * @param1: usage depends upon VSI type. For VF types, indicates VF id
9723  *
9724  * This allocates the sw VSI structure and its queue resources, then add a VSI
9725  * to the identified VEB.
9726  *
9727  * Returns pointer to the successfully allocated and configure VSI sw struct on
9728  * success, otherwise returns NULL on failure.
9729  **/
9730 struct i40e_vsi *i40e_vsi_setup(struct i40e_pf *pf, u8 type,
9731                                 u16 uplink_seid, u32 param1)
9732 {
9733         struct i40e_vsi *vsi = NULL;
9734         struct i40e_veb *veb = NULL;
9735         int ret, i;
9736         int v_idx;
9737
9738         /* The requested uplink_seid must be either
9739          *     - the PF's port seid
9740          *              no VEB is needed because this is the PF
9741          *              or this is a Flow Director special case VSI
9742          *     - seid of an existing VEB
9743          *     - seid of a VSI that owns an existing VEB
9744          *     - seid of a VSI that doesn't own a VEB
9745          *              a new VEB is created and the VSI becomes the owner
9746          *     - seid of the PF VSI, which is what creates the first VEB
9747          *              this is a special case of the previous
9748          *
9749          * Find which uplink_seid we were given and create a new VEB if needed
9750          */
9751         for (i = 0; i < I40E_MAX_VEB; i++) {
9752                 if (pf->veb[i] && pf->veb[i]->seid == uplink_seid) {
9753                         veb = pf->veb[i];
9754                         break;
9755                 }
9756         }
9757
9758         if (!veb && uplink_seid != pf->mac_seid) {
9759
9760                 for (i = 0; i < pf->num_alloc_vsi; i++) {
9761                         if (pf->vsi[i] && pf->vsi[i]->seid == uplink_seid) {
9762                                 vsi = pf->vsi[i];
9763                                 break;
9764                         }
9765                 }
9766                 if (!vsi) {
9767                         dev_info(&pf->pdev->dev, "no such uplink_seid %d\n",
9768                                  uplink_seid);
9769                         return NULL;
9770                 }
9771
9772                 if (vsi->uplink_seid == pf->mac_seid)
9773                         veb = i40e_veb_setup(pf, 0, pf->mac_seid, vsi->seid,
9774                                              vsi->tc_config.enabled_tc);
9775                 else if ((vsi->flags & I40E_VSI_FLAG_VEB_OWNER) == 0)
9776                         veb = i40e_veb_setup(pf, 0, vsi->uplink_seid, vsi->seid,
9777                                              vsi->tc_config.enabled_tc);
9778                 if (veb) {
9779                         if (vsi->seid != pf->vsi[pf->lan_vsi]->seid) {
9780                                 dev_info(&vsi->back->pdev->dev,
9781                                          "New VSI creation error, uplink seid of LAN VSI expected.\n");
9782                                 return NULL;
9783                         }
9784                         /* We come up by default in VEPA mode if SRIOV is not
9785                          * already enabled, in which case we can't force VEPA
9786                          * mode.
9787                          */
9788                         if (!(pf->flags & I40E_FLAG_VEB_MODE_ENABLED)) {
9789                                 veb->bridge_mode = BRIDGE_MODE_VEPA;
9790                                 pf->flags &= ~I40E_FLAG_VEB_MODE_ENABLED;
9791                         }
9792                         i40e_config_bridge_mode(veb);
9793                 }
9794                 for (i = 0; i < I40E_MAX_VEB && !veb; i++) {
9795                         if (pf->veb[i] && pf->veb[i]->seid == vsi->uplink_seid)
9796                                 veb = pf->veb[i];
9797                 }
9798                 if (!veb) {
9799                         dev_info(&pf->pdev->dev, "couldn't add VEB\n");
9800                         return NULL;
9801                 }
9802
9803                 vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
9804                 uplink_seid = veb->seid;
9805         }
9806
9807         /* get vsi sw struct */
9808         v_idx = i40e_vsi_mem_alloc(pf, type);
9809         if (v_idx < 0)
9810                 goto err_alloc;
9811         vsi = pf->vsi[v_idx];
9812         if (!vsi)
9813                 goto err_alloc;
9814         vsi->type = type;
9815         vsi->veb_idx = (veb ? veb->idx : I40E_NO_VEB);
9816
9817         if (type == I40E_VSI_MAIN)
9818                 pf->lan_vsi = v_idx;
9819         else if (type == I40E_VSI_SRIOV)
9820                 vsi->vf_id = param1;
9821         /* assign it some queues */
9822         ret = i40e_get_lump(pf, pf->qp_pile, vsi->alloc_queue_pairs,
9823                                 vsi->idx);
9824         if (ret < 0) {
9825                 dev_info(&pf->pdev->dev,
9826                          "failed to get tracking for %d queues for VSI %d err=%d\n",
9827                          vsi->alloc_queue_pairs, vsi->seid, ret);
9828                 goto err_vsi;
9829         }
9830         vsi->base_queue = ret;
9831
9832         /* get a VSI from the hardware */
9833         vsi->uplink_seid = uplink_seid;
9834         ret = i40e_add_vsi(vsi);
9835         if (ret)
9836                 goto err_vsi;
9837
9838         switch (vsi->type) {
9839         /* setup the netdev if needed */
9840         case I40E_VSI_MAIN:
9841                 /* Apply relevant filters if a platform-specific mac
9842                  * address was selected.
9843                  */
9844                 if (!!(pf->flags & I40E_FLAG_PF_MAC)) {
9845                         ret = i40e_macaddr_init(vsi, pf->hw.mac.addr);
9846                         if (ret) {
9847                                 dev_warn(&pf->pdev->dev,
9848                                          "could not set up macaddr; err %d\n",
9849                                          ret);
9850                         }
9851                 }
9852         case I40E_VSI_VMDQ2:
9853         case I40E_VSI_FCOE:
9854                 ret = i40e_config_netdev(vsi);
9855                 if (ret)
9856                         goto err_netdev;
9857                 ret = register_netdev(vsi->netdev);
9858                 if (ret)
9859                         goto err_netdev;
9860                 vsi->netdev_registered = true;
9861                 netif_carrier_off(vsi->netdev);
9862 #ifdef CONFIG_I40E_DCB
9863                 /* Setup DCB netlink interface */
9864                 i40e_dcbnl_setup(vsi);
9865 #endif /* CONFIG_I40E_DCB */
9866                 /* fall through */
9867
9868         case I40E_VSI_FDIR:
9869                 /* set up vectors and rings if needed */
9870                 ret = i40e_vsi_setup_vectors(vsi);
9871                 if (ret)
9872                         goto err_msix;
9873
9874                 ret = i40e_alloc_rings(vsi);
9875                 if (ret)
9876                         goto err_rings;
9877
9878                 /* map all of the rings to the q_vectors */
9879                 i40e_vsi_map_rings_to_vectors(vsi);
9880
9881                 i40e_vsi_reset_stats(vsi);
9882                 break;
9883
9884         default:
9885                 /* no netdev or rings for the other VSI types */
9886                 break;
9887         }
9888
9889         if ((pf->flags & I40E_FLAG_RSS_AQ_CAPABLE) &&
9890             (vsi->type == I40E_VSI_VMDQ2)) {
9891                 ret = i40e_vsi_config_rss(vsi);
9892         }
9893         return vsi;
9894
9895 err_rings:
9896         i40e_vsi_free_q_vectors(vsi);
9897 err_msix:
9898         if (vsi->netdev_registered) {
9899                 vsi->netdev_registered = false;
9900                 unregister_netdev(vsi->netdev);
9901                 free_netdev(vsi->netdev);
9902                 vsi->netdev = NULL;
9903         }
9904 err_netdev:
9905         i40e_aq_delete_element(&pf->hw, vsi->seid, NULL);
9906 err_vsi:
9907         i40e_vsi_clear(vsi);
9908 err_alloc:
9909         return NULL;
9910 }
9911
9912 /**
9913  * i40e_veb_get_bw_info - Query VEB BW information
9914  * @veb: the veb to query
9915  *
9916  * Query the Tx scheduler BW configuration data for given VEB
9917  **/
9918 static int i40e_veb_get_bw_info(struct i40e_veb *veb)
9919 {
9920         struct i40e_aqc_query_switching_comp_ets_config_resp ets_data;
9921         struct i40e_aqc_query_switching_comp_bw_config_resp bw_data;
9922         struct i40e_pf *pf = veb->pf;
9923         struct i40e_hw *hw = &pf->hw;
9924         u32 tc_bw_max;
9925         int ret = 0;
9926         int i;
9927
9928         ret = i40e_aq_query_switch_comp_bw_config(hw, veb->seid,
9929                                                   &bw_data, NULL);
9930         if (ret) {
9931                 dev_info(&pf->pdev->dev,
9932                          "query veb bw config failed, err %s aq_err %s\n",
9933                          i40e_stat_str(&pf->hw, ret),
9934                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9935                 goto out;
9936         }
9937
9938         ret = i40e_aq_query_switch_comp_ets_config(hw, veb->seid,
9939                                                    &ets_data, NULL);
9940         if (ret) {
9941                 dev_info(&pf->pdev->dev,
9942                          "query veb bw ets config failed, err %s aq_err %s\n",
9943                          i40e_stat_str(&pf->hw, ret),
9944                          i40e_aq_str(&pf->hw, hw->aq.asq_last_status));
9945                 goto out;
9946         }
9947
9948         veb->bw_limit = le16_to_cpu(ets_data.port_bw_limit);
9949         veb->bw_max_quanta = ets_data.tc_bw_max;
9950         veb->is_abs_credits = bw_data.absolute_credits_enable;
9951         veb->enabled_tc = ets_data.tc_valid_bits;
9952         tc_bw_max = le16_to_cpu(bw_data.tc_bw_max[0]) |
9953                     (le16_to_cpu(bw_data.tc_bw_max[1]) << 16);
9954         for (i = 0; i < I40E_MAX_TRAFFIC_CLASS; i++) {
9955                 veb->bw_tc_share_credits[i] = bw_data.tc_bw_share_credits[i];
9956                 veb->bw_tc_limit_credits[i] =
9957                                         le16_to_cpu(bw_data.tc_bw_limits[i]);
9958                 veb->bw_tc_max_quanta[i] = ((tc_bw_max >> (i*4)) & 0x7);
9959         }
9960
9961 out:
9962         return ret;
9963 }
9964
9965 /**
9966  * i40e_veb_mem_alloc - Allocates the next available struct veb in the PF
9967  * @pf: board private structure
9968  *
9969  * On error: returns error code (negative)
9970  * On success: returns vsi index in PF (positive)
9971  **/
9972 static int i40e_veb_mem_alloc(struct i40e_pf *pf)
9973 {
9974         int ret = -ENOENT;
9975         struct i40e_veb *veb;
9976         int i;
9977
9978         /* Need to protect the allocation of switch elements at the PF level */
9979         mutex_lock(&pf->switch_mutex);
9980
9981         /* VEB list may be fragmented if VEB creation/destruction has
9982          * been happening.  We can afford to do a quick scan to look
9983          * for any free slots in the list.
9984          *
9985          * find next empty veb slot, looping back around if necessary
9986          */
9987         i = 0;
9988         while ((i < I40E_MAX_VEB) && (pf->veb[i] != NULL))
9989                 i++;
9990         if (i >= I40E_MAX_VEB) {
9991                 ret = -ENOMEM;
9992                 goto err_alloc_veb;  /* out of VEB slots! */
9993         }
9994
9995         veb = kzalloc(sizeof(*veb), GFP_KERNEL);
9996         if (!veb) {
9997                 ret = -ENOMEM;
9998                 goto err_alloc_veb;
9999         }
10000         veb->pf = pf;
10001         veb->idx = i;
10002         veb->enabled_tc = 1;
10003
10004         pf->veb[i] = veb;
10005         ret = i;
10006 err_alloc_veb:
10007         mutex_unlock(&pf->switch_mutex);
10008         return ret;
10009 }
10010
10011 /**
10012  * i40e_switch_branch_release - Delete a branch of the switch tree
10013  * @branch: where to start deleting
10014  *
10015  * This uses recursion to find the tips of the branch to be
10016  * removed, deleting until we get back to and can delete this VEB.
10017  **/
10018 static void i40e_switch_branch_release(struct i40e_veb *branch)
10019 {
10020         struct i40e_pf *pf = branch->pf;
10021         u16 branch_seid = branch->seid;
10022         u16 veb_idx = branch->idx;
10023         int i;
10024
10025         /* release any VEBs on this VEB - RECURSION */
10026         for (i = 0; i < I40E_MAX_VEB; i++) {
10027                 if (!pf->veb[i])
10028                         continue;
10029                 if (pf->veb[i]->uplink_seid == branch->seid)
10030                         i40e_switch_branch_release(pf->veb[i]);
10031         }
10032
10033         /* Release the VSIs on this VEB, but not the owner VSI.
10034          *
10035          * NOTE: Removing the last VSI on a VEB has the SIDE EFFECT of removing
10036          *       the VEB itself, so don't use (*branch) after this loop.
10037          */
10038         for (i = 0; i < pf->num_alloc_vsi; i++) {
10039                 if (!pf->vsi[i])
10040                         continue;
10041                 if (pf->vsi[i]->uplink_seid == branch_seid &&
10042                    (pf->vsi[i]->flags & I40E_VSI_FLAG_VEB_OWNER) == 0) {
10043                         i40e_vsi_release(pf->vsi[i]);
10044                 }
10045         }
10046
10047         /* There's one corner case where the VEB might not have been
10048          * removed, so double check it here and remove it if needed.
10049          * This case happens if the veb was created from the debugfs
10050          * commands and no VSIs were added to it.
10051          */
10052         if (pf->veb[veb_idx])
10053                 i40e_veb_release(pf->veb[veb_idx]);
10054 }
10055
10056 /**
10057  * i40e_veb_clear - remove veb struct
10058  * @veb: the veb to remove
10059  **/
10060 static void i40e_veb_clear(struct i40e_veb *veb)
10061 {
10062         if (!veb)
10063                 return;
10064
10065         if (veb->pf) {
10066                 struct i40e_pf *pf = veb->pf;
10067
10068                 mutex_lock(&pf->switch_mutex);
10069                 if (pf->veb[veb->idx] == veb)
10070                         pf->veb[veb->idx] = NULL;
10071                 mutex_unlock(&pf->switch_mutex);
10072         }
10073
10074         kfree(veb);
10075 }
10076
10077 /**
10078  * i40e_veb_release - Delete a VEB and free its resources
10079  * @veb: the VEB being removed
10080  **/
10081 void i40e_veb_release(struct i40e_veb *veb)
10082 {
10083         struct i40e_vsi *vsi = NULL;
10084         struct i40e_pf *pf;
10085         int i, n = 0;
10086
10087         pf = veb->pf;
10088
10089         /* find the remaining VSI and check for extras */
10090         for (i = 0; i < pf->num_alloc_vsi; i++) {
10091                 if (pf->vsi[i] && pf->vsi[i]->uplink_seid == veb->seid) {
10092                         n++;
10093                         vsi = pf->vsi[i];
10094                 }
10095         }
10096         if (n != 1) {
10097                 dev_info(&pf->pdev->dev,
10098                          "can't remove VEB %d with %d VSIs left\n",
10099                          veb->seid, n);
10100                 return;
10101         }
10102
10103         /* move the remaining VSI to uplink veb */
10104         vsi->flags &= ~I40E_VSI_FLAG_VEB_OWNER;
10105         if (veb->uplink_seid) {
10106                 vsi->uplink_seid = veb->uplink_seid;
10107                 if (veb->uplink_seid == pf->mac_seid)
10108                         vsi->veb_idx = I40E_NO_VEB;
10109                 else
10110                         vsi->veb_idx = veb->veb_idx;
10111         } else {
10112                 /* floating VEB */
10113                 vsi->uplink_seid = pf->vsi[pf->lan_vsi]->uplink_seid;
10114                 vsi->veb_idx = pf->vsi[pf->lan_vsi]->veb_idx;
10115         }
10116
10117         i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10118         i40e_veb_clear(veb);
10119 }
10120
10121 /**
10122  * i40e_add_veb - create the VEB in the switch
10123  * @veb: the VEB to be instantiated
10124  * @vsi: the controlling VSI
10125  **/
10126 static int i40e_add_veb(struct i40e_veb *veb, struct i40e_vsi *vsi)
10127 {
10128         struct i40e_pf *pf = veb->pf;
10129         bool enable_stats = !!(pf->flags & I40E_FLAG_VEB_STATS_ENABLED);
10130         int ret;
10131
10132         ret = i40e_aq_add_veb(&pf->hw, veb->uplink_seid, vsi->seid,
10133                               veb->enabled_tc, false,
10134                               &veb->seid, enable_stats, NULL);
10135
10136         /* get a VEB from the hardware */
10137         if (ret) {
10138                 dev_info(&pf->pdev->dev,
10139                          "couldn't add VEB, err %s aq_err %s\n",
10140                          i40e_stat_str(&pf->hw, ret),
10141                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10142                 return -EPERM;
10143         }
10144
10145         /* get statistics counter */
10146         ret = i40e_aq_get_veb_parameters(&pf->hw, veb->seid, NULL, NULL,
10147                                          &veb->stats_idx, NULL, NULL, NULL);
10148         if (ret) {
10149                 dev_info(&pf->pdev->dev,
10150                          "couldn't get VEB statistics idx, err %s aq_err %s\n",
10151                          i40e_stat_str(&pf->hw, ret),
10152                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10153                 return -EPERM;
10154         }
10155         ret = i40e_veb_get_bw_info(veb);
10156         if (ret) {
10157                 dev_info(&pf->pdev->dev,
10158                          "couldn't get VEB bw info, err %s aq_err %s\n",
10159                          i40e_stat_str(&pf->hw, ret),
10160                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10161                 i40e_aq_delete_element(&pf->hw, veb->seid, NULL);
10162                 return -ENOENT;
10163         }
10164
10165         vsi->uplink_seid = veb->seid;
10166         vsi->veb_idx = veb->idx;
10167         vsi->flags |= I40E_VSI_FLAG_VEB_OWNER;
10168
10169         return 0;
10170 }
10171
10172 /**
10173  * i40e_veb_setup - Set up a VEB
10174  * @pf: board private structure
10175  * @flags: VEB setup flags
10176  * @uplink_seid: the switch element to link to
10177  * @vsi_seid: the initial VSI seid
10178  * @enabled_tc: Enabled TC bit-map
10179  *
10180  * This allocates the sw VEB structure and links it into the switch
10181  * It is possible and legal for this to be a duplicate of an already
10182  * existing VEB.  It is also possible for both uplink and vsi seids
10183  * to be zero, in order to create a floating VEB.
10184  *
10185  * Returns pointer to the successfully allocated VEB sw struct on
10186  * success, otherwise returns NULL on failure.
10187  **/
10188 struct i40e_veb *i40e_veb_setup(struct i40e_pf *pf, u16 flags,
10189                                 u16 uplink_seid, u16 vsi_seid,
10190                                 u8 enabled_tc)
10191 {
10192         struct i40e_veb *veb, *uplink_veb = NULL;
10193         int vsi_idx, veb_idx;
10194         int ret;
10195
10196         /* if one seid is 0, the other must be 0 to create a floating relay */
10197         if ((uplink_seid == 0 || vsi_seid == 0) &&
10198             (uplink_seid + vsi_seid != 0)) {
10199                 dev_info(&pf->pdev->dev,
10200                          "one, not both seid's are 0: uplink=%d vsi=%d\n",
10201                          uplink_seid, vsi_seid);
10202                 return NULL;
10203         }
10204
10205         /* make sure there is such a vsi and uplink */
10206         for (vsi_idx = 0; vsi_idx < pf->num_alloc_vsi; vsi_idx++)
10207                 if (pf->vsi[vsi_idx] && pf->vsi[vsi_idx]->seid == vsi_seid)
10208                         break;
10209         if (vsi_idx >= pf->num_alloc_vsi && vsi_seid != 0) {
10210                 dev_info(&pf->pdev->dev, "vsi seid %d not found\n",
10211                          vsi_seid);
10212                 return NULL;
10213         }
10214
10215         if (uplink_seid && uplink_seid != pf->mac_seid) {
10216                 for (veb_idx = 0; veb_idx < I40E_MAX_VEB; veb_idx++) {
10217                         if (pf->veb[veb_idx] &&
10218                             pf->veb[veb_idx]->seid == uplink_seid) {
10219                                 uplink_veb = pf->veb[veb_idx];
10220                                 break;
10221                         }
10222                 }
10223                 if (!uplink_veb) {
10224                         dev_info(&pf->pdev->dev,
10225                                  "uplink seid %d not found\n", uplink_seid);
10226                         return NULL;
10227                 }
10228         }
10229
10230         /* get veb sw struct */
10231         veb_idx = i40e_veb_mem_alloc(pf);
10232         if (veb_idx < 0)
10233                 goto err_alloc;
10234         veb = pf->veb[veb_idx];
10235         veb->flags = flags;
10236         veb->uplink_seid = uplink_seid;
10237         veb->veb_idx = (uplink_veb ? uplink_veb->idx : I40E_NO_VEB);
10238         veb->enabled_tc = (enabled_tc ? enabled_tc : 0x1);
10239
10240         /* create the VEB in the switch */
10241         ret = i40e_add_veb(veb, pf->vsi[vsi_idx]);
10242         if (ret)
10243                 goto err_veb;
10244         if (vsi_idx == pf->lan_vsi)
10245                 pf->lan_veb = veb->idx;
10246
10247         return veb;
10248
10249 err_veb:
10250         i40e_veb_clear(veb);
10251 err_alloc:
10252         return NULL;
10253 }
10254
10255 /**
10256  * i40e_setup_pf_switch_element - set PF vars based on switch type
10257  * @pf: board private structure
10258  * @ele: element we are building info from
10259  * @num_reported: total number of elements
10260  * @printconfig: should we print the contents
10261  *
10262  * helper function to assist in extracting a few useful SEID values.
10263  **/
10264 static void i40e_setup_pf_switch_element(struct i40e_pf *pf,
10265                                 struct i40e_aqc_switch_config_element_resp *ele,
10266                                 u16 num_reported, bool printconfig)
10267 {
10268         u16 downlink_seid = le16_to_cpu(ele->downlink_seid);
10269         u16 uplink_seid = le16_to_cpu(ele->uplink_seid);
10270         u8 element_type = ele->element_type;
10271         u16 seid = le16_to_cpu(ele->seid);
10272
10273         if (printconfig)
10274                 dev_info(&pf->pdev->dev,
10275                          "type=%d seid=%d uplink=%d downlink=%d\n",
10276                          element_type, seid, uplink_seid, downlink_seid);
10277
10278         switch (element_type) {
10279         case I40E_SWITCH_ELEMENT_TYPE_MAC:
10280                 pf->mac_seid = seid;
10281                 break;
10282         case I40E_SWITCH_ELEMENT_TYPE_VEB:
10283                 /* Main VEB? */
10284                 if (uplink_seid != pf->mac_seid)
10285                         break;
10286                 if (pf->lan_veb == I40E_NO_VEB) {
10287                         int v;
10288
10289                         /* find existing or else empty VEB */
10290                         for (v = 0; v < I40E_MAX_VEB; v++) {
10291                                 if (pf->veb[v] && (pf->veb[v]->seid == seid)) {
10292                                         pf->lan_veb = v;
10293                                         break;
10294                                 }
10295                         }
10296                         if (pf->lan_veb == I40E_NO_VEB) {
10297                                 v = i40e_veb_mem_alloc(pf);
10298                                 if (v < 0)
10299                                         break;
10300                                 pf->lan_veb = v;
10301                         }
10302                 }
10303
10304                 pf->veb[pf->lan_veb]->seid = seid;
10305                 pf->veb[pf->lan_veb]->uplink_seid = pf->mac_seid;
10306                 pf->veb[pf->lan_veb]->pf = pf;
10307                 pf->veb[pf->lan_veb]->veb_idx = I40E_NO_VEB;
10308                 break;
10309         case I40E_SWITCH_ELEMENT_TYPE_VSI:
10310                 if (num_reported != 1)
10311                         break;
10312                 /* This is immediately after a reset so we can assume this is
10313                  * the PF's VSI
10314                  */
10315                 pf->mac_seid = uplink_seid;
10316                 pf->pf_seid = downlink_seid;
10317                 pf->main_vsi_seid = seid;
10318                 if (printconfig)
10319                         dev_info(&pf->pdev->dev,
10320                                  "pf_seid=%d main_vsi_seid=%d\n",
10321                                  pf->pf_seid, pf->main_vsi_seid);
10322                 break;
10323         case I40E_SWITCH_ELEMENT_TYPE_PF:
10324         case I40E_SWITCH_ELEMENT_TYPE_VF:
10325         case I40E_SWITCH_ELEMENT_TYPE_EMP:
10326         case I40E_SWITCH_ELEMENT_TYPE_BMC:
10327         case I40E_SWITCH_ELEMENT_TYPE_PE:
10328         case I40E_SWITCH_ELEMENT_TYPE_PA:
10329                 /* ignore these for now */
10330                 break;
10331         default:
10332                 dev_info(&pf->pdev->dev, "unknown element type=%d seid=%d\n",
10333                          element_type, seid);
10334                 break;
10335         }
10336 }
10337
10338 /**
10339  * i40e_fetch_switch_configuration - Get switch config from firmware
10340  * @pf: board private structure
10341  * @printconfig: should we print the contents
10342  *
10343  * Get the current switch configuration from the device and
10344  * extract a few useful SEID values.
10345  **/
10346 int i40e_fetch_switch_configuration(struct i40e_pf *pf, bool printconfig)
10347 {
10348         struct i40e_aqc_get_switch_config_resp *sw_config;
10349         u16 next_seid = 0;
10350         int ret = 0;
10351         u8 *aq_buf;
10352         int i;
10353
10354         aq_buf = kzalloc(I40E_AQ_LARGE_BUF, GFP_KERNEL);
10355         if (!aq_buf)
10356                 return -ENOMEM;
10357
10358         sw_config = (struct i40e_aqc_get_switch_config_resp *)aq_buf;
10359         do {
10360                 u16 num_reported, num_total;
10361
10362                 ret = i40e_aq_get_switch_config(&pf->hw, sw_config,
10363                                                 I40E_AQ_LARGE_BUF,
10364                                                 &next_seid, NULL);
10365                 if (ret) {
10366                         dev_info(&pf->pdev->dev,
10367                                  "get switch config failed err %s aq_err %s\n",
10368                                  i40e_stat_str(&pf->hw, ret),
10369                                  i40e_aq_str(&pf->hw,
10370                                              pf->hw.aq.asq_last_status));
10371                         kfree(aq_buf);
10372                         return -ENOENT;
10373                 }
10374
10375                 num_reported = le16_to_cpu(sw_config->header.num_reported);
10376                 num_total = le16_to_cpu(sw_config->header.num_total);
10377
10378                 if (printconfig)
10379                         dev_info(&pf->pdev->dev,
10380                                  "header: %d reported %d total\n",
10381                                  num_reported, num_total);
10382
10383                 for (i = 0; i < num_reported; i++) {
10384                         struct i40e_aqc_switch_config_element_resp *ele =
10385                                 &sw_config->element[i];
10386
10387                         i40e_setup_pf_switch_element(pf, ele, num_reported,
10388                                                      printconfig);
10389                 }
10390         } while (next_seid != 0);
10391
10392         kfree(aq_buf);
10393         return ret;
10394 }
10395
10396 /**
10397  * i40e_setup_pf_switch - Setup the HW switch on startup or after reset
10398  * @pf: board private structure
10399  * @reinit: if the Main VSI needs to re-initialized.
10400  *
10401  * Returns 0 on success, negative value on failure
10402  **/
10403 static int i40e_setup_pf_switch(struct i40e_pf *pf, bool reinit)
10404 {
10405         u16 flags = 0;
10406         int ret;
10407
10408         /* find out what's out there already */
10409         ret = i40e_fetch_switch_configuration(pf, false);
10410         if (ret) {
10411                 dev_info(&pf->pdev->dev,
10412                          "couldn't fetch switch config, err %s aq_err %s\n",
10413                          i40e_stat_str(&pf->hw, ret),
10414                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
10415                 return ret;
10416         }
10417         i40e_pf_reset_stats(pf);
10418
10419         /* set the switch config bit for the whole device to
10420          * support limited promisc or true promisc
10421          * when user requests promisc. The default is limited
10422          * promisc.
10423         */
10424
10425         if ((pf->hw.pf_id == 0) &&
10426             !(pf->flags & I40E_FLAG_TRUE_PROMISC_SUPPORT))
10427                 flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10428
10429         if (pf->hw.pf_id == 0) {
10430                 u16 valid_flags;
10431
10432                 valid_flags = I40E_AQ_SET_SWITCH_CFG_PROMISC;
10433                 ret = i40e_aq_set_switch_config(&pf->hw, flags, valid_flags,
10434                                                 NULL);
10435                 if (ret && pf->hw.aq.asq_last_status != I40E_AQ_RC_ESRCH) {
10436                         dev_info(&pf->pdev->dev,
10437                                  "couldn't set switch config bits, err %s aq_err %s\n",
10438                                  i40e_stat_str(&pf->hw, ret),
10439                                  i40e_aq_str(&pf->hw,
10440                                              pf->hw.aq.asq_last_status));
10441                         /* not a fatal problem, just keep going */
10442                 }
10443         }
10444
10445         /* first time setup */
10446         if (pf->lan_vsi == I40E_NO_VSI || reinit) {
10447                 struct i40e_vsi *vsi = NULL;
10448                 u16 uplink_seid;
10449
10450                 /* Set up the PF VSI associated with the PF's main VSI
10451                  * that is already in the HW switch
10452                  */
10453                 if (pf->lan_veb != I40E_NO_VEB && pf->veb[pf->lan_veb])
10454                         uplink_seid = pf->veb[pf->lan_veb]->seid;
10455                 else
10456                         uplink_seid = pf->mac_seid;
10457                 if (pf->lan_vsi == I40E_NO_VSI)
10458                         vsi = i40e_vsi_setup(pf, I40E_VSI_MAIN, uplink_seid, 0);
10459                 else if (reinit)
10460                         vsi = i40e_vsi_reinit_setup(pf->vsi[pf->lan_vsi]);
10461                 if (!vsi) {
10462                         dev_info(&pf->pdev->dev, "setup of MAIN VSI failed\n");
10463                         i40e_fdir_teardown(pf);
10464                         return -EAGAIN;
10465                 }
10466         } else {
10467                 /* force a reset of TC and queue layout configurations */
10468                 u8 enabled_tc = pf->vsi[pf->lan_vsi]->tc_config.enabled_tc;
10469
10470                 pf->vsi[pf->lan_vsi]->tc_config.enabled_tc = 0;
10471                 pf->vsi[pf->lan_vsi]->seid = pf->main_vsi_seid;
10472                 i40e_vsi_config_tc(pf->vsi[pf->lan_vsi], enabled_tc);
10473         }
10474         i40e_vlan_stripping_disable(pf->vsi[pf->lan_vsi]);
10475
10476         i40e_fdir_sb_setup(pf);
10477
10478         /* Setup static PF queue filter control settings */
10479         ret = i40e_setup_pf_filter_control(pf);
10480         if (ret) {
10481                 dev_info(&pf->pdev->dev, "setup_pf_filter_control failed: %d\n",
10482                          ret);
10483                 /* Failure here should not stop continuing other steps */
10484         }
10485
10486         /* enable RSS in the HW, even for only one queue, as the stack can use
10487          * the hash
10488          */
10489         if ((pf->flags & I40E_FLAG_RSS_ENABLED))
10490                 i40e_pf_config_rss(pf);
10491
10492         /* fill in link information and enable LSE reporting */
10493         i40e_update_link_info(&pf->hw);
10494         i40e_link_event(pf);
10495
10496         /* Initialize user-specific link properties */
10497         pf->fc_autoneg_status = ((pf->hw.phy.link_info.an_info &
10498                                   I40E_AQ_AN_COMPLETED) ? true : false);
10499
10500         i40e_ptp_init(pf);
10501
10502         return ret;
10503 }
10504
10505 /**
10506  * i40e_determine_queue_usage - Work out queue distribution
10507  * @pf: board private structure
10508  **/
10509 static void i40e_determine_queue_usage(struct i40e_pf *pf)
10510 {
10511         int queues_left;
10512
10513         pf->num_lan_qps = 0;
10514 #ifdef I40E_FCOE
10515         pf->num_fcoe_qps = 0;
10516 #endif
10517
10518         /* Find the max queues to be put into basic use.  We'll always be
10519          * using TC0, whether or not DCB is running, and TC0 will get the
10520          * big RSS set.
10521          */
10522         queues_left = pf->hw.func_caps.num_tx_qp;
10523
10524         if ((queues_left == 1) ||
10525             !(pf->flags & I40E_FLAG_MSIX_ENABLED)) {
10526                 /* one qp for PF, no queues for anything else */
10527                 queues_left = 0;
10528                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10529
10530                 /* make sure all the fancies are disabled */
10531                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10532                                I40E_FLAG_IWARP_ENABLED  |
10533 #ifdef I40E_FCOE
10534                                I40E_FLAG_FCOE_ENABLED   |
10535 #endif
10536                                I40E_FLAG_FD_SB_ENABLED  |
10537                                I40E_FLAG_FD_ATR_ENABLED |
10538                                I40E_FLAG_DCB_CAPABLE    |
10539                                I40E_FLAG_DCB_ENABLED    |
10540                                I40E_FLAG_SRIOV_ENABLED  |
10541                                I40E_FLAG_VMDQ_ENABLED);
10542         } else if (!(pf->flags & (I40E_FLAG_RSS_ENABLED |
10543                                   I40E_FLAG_FD_SB_ENABLED |
10544                                   I40E_FLAG_FD_ATR_ENABLED |
10545                                   I40E_FLAG_DCB_CAPABLE))) {
10546                 /* one qp for PF */
10547                 pf->alloc_rss_size = pf->num_lan_qps = 1;
10548                 queues_left -= pf->num_lan_qps;
10549
10550                 pf->flags &= ~(I40E_FLAG_RSS_ENABLED    |
10551                                I40E_FLAG_IWARP_ENABLED  |
10552 #ifdef I40E_FCOE
10553                                I40E_FLAG_FCOE_ENABLED   |
10554 #endif
10555                                I40E_FLAG_FD_SB_ENABLED  |
10556                                I40E_FLAG_FD_ATR_ENABLED |
10557                                I40E_FLAG_DCB_ENABLED    |
10558                                I40E_FLAG_VMDQ_ENABLED);
10559         } else {
10560                 /* Not enough queues for all TCs */
10561                 if ((pf->flags & I40E_FLAG_DCB_CAPABLE) &&
10562                     (queues_left < I40E_MAX_TRAFFIC_CLASS)) {
10563                         pf->flags &= ~(I40E_FLAG_DCB_CAPABLE |
10564                                         I40E_FLAG_DCB_ENABLED);
10565                         dev_info(&pf->pdev->dev, "not enough queues for DCB. DCB is disabled.\n");
10566                 }
10567                 pf->num_lan_qps = max_t(int, pf->rss_size_max,
10568                                         num_online_cpus());
10569                 pf->num_lan_qps = min_t(int, pf->num_lan_qps,
10570                                         pf->hw.func_caps.num_tx_qp);
10571
10572                 queues_left -= pf->num_lan_qps;
10573         }
10574
10575 #ifdef I40E_FCOE
10576         if (pf->flags & I40E_FLAG_FCOE_ENABLED) {
10577                 if (I40E_DEFAULT_FCOE <= queues_left) {
10578                         pf->num_fcoe_qps = I40E_DEFAULT_FCOE;
10579                 } else if (I40E_MINIMUM_FCOE <= queues_left) {
10580                         pf->num_fcoe_qps = I40E_MINIMUM_FCOE;
10581                 } else {
10582                         pf->num_fcoe_qps = 0;
10583                         pf->flags &= ~I40E_FLAG_FCOE_ENABLED;
10584                         dev_info(&pf->pdev->dev, "not enough queues for FCoE. FCoE feature will be disabled\n");
10585                 }
10586
10587                 queues_left -= pf->num_fcoe_qps;
10588         }
10589
10590 #endif
10591         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10592                 if (queues_left > 1) {
10593                         queues_left -= 1; /* save 1 queue for FD */
10594                 } else {
10595                         pf->flags &= ~I40E_FLAG_FD_SB_ENABLED;
10596                         dev_info(&pf->pdev->dev, "not enough queues for Flow Director. Flow Director feature is disabled\n");
10597                 }
10598         }
10599
10600         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
10601             pf->num_vf_qps && pf->num_req_vfs && queues_left) {
10602                 pf->num_req_vfs = min_t(int, pf->num_req_vfs,
10603                                         (queues_left / pf->num_vf_qps));
10604                 queues_left -= (pf->num_req_vfs * pf->num_vf_qps);
10605         }
10606
10607         if ((pf->flags & I40E_FLAG_VMDQ_ENABLED) &&
10608             pf->num_vmdq_vsis && pf->num_vmdq_qps && queues_left) {
10609                 pf->num_vmdq_vsis = min_t(int, pf->num_vmdq_vsis,
10610                                           (queues_left / pf->num_vmdq_qps));
10611                 queues_left -= (pf->num_vmdq_vsis * pf->num_vmdq_qps);
10612         }
10613
10614         pf->queues_left = queues_left;
10615         dev_dbg(&pf->pdev->dev,
10616                 "qs_avail=%d FD SB=%d lan_qs=%d lan_tc0=%d vf=%d*%d vmdq=%d*%d, remaining=%d\n",
10617                 pf->hw.func_caps.num_tx_qp,
10618                 !!(pf->flags & I40E_FLAG_FD_SB_ENABLED),
10619                 pf->num_lan_qps, pf->alloc_rss_size, pf->num_req_vfs,
10620                 pf->num_vf_qps, pf->num_vmdq_vsis, pf->num_vmdq_qps,
10621                 queues_left);
10622 #ifdef I40E_FCOE
10623         dev_dbg(&pf->pdev->dev, "fcoe queues = %d\n", pf->num_fcoe_qps);
10624 #endif
10625 }
10626
10627 /**
10628  * i40e_setup_pf_filter_control - Setup PF static filter control
10629  * @pf: PF to be setup
10630  *
10631  * i40e_setup_pf_filter_control sets up a PF's initial filter control
10632  * settings. If PE/FCoE are enabled then it will also set the per PF
10633  * based filter sizes required for them. It also enables Flow director,
10634  * ethertype and macvlan type filter settings for the pf.
10635  *
10636  * Returns 0 on success, negative on failure
10637  **/
10638 static int i40e_setup_pf_filter_control(struct i40e_pf *pf)
10639 {
10640         struct i40e_filter_control_settings *settings = &pf->filter_settings;
10641
10642         settings->hash_lut_size = I40E_HASH_LUT_SIZE_128;
10643
10644         /* Flow Director is enabled */
10645         if (pf->flags & (I40E_FLAG_FD_SB_ENABLED | I40E_FLAG_FD_ATR_ENABLED))
10646                 settings->enable_fdir = true;
10647
10648         /* Ethtype and MACVLAN filters enabled for PF */
10649         settings->enable_ethtype = true;
10650         settings->enable_macvlan = true;
10651
10652         if (i40e_set_filter_control(&pf->hw, settings))
10653                 return -ENOENT;
10654
10655         return 0;
10656 }
10657
10658 #define INFO_STRING_LEN 255
10659 #define REMAIN(__x) (INFO_STRING_LEN - (__x))
10660 static void i40e_print_features(struct i40e_pf *pf)
10661 {
10662         struct i40e_hw *hw = &pf->hw;
10663         char *buf;
10664         int i;
10665
10666         buf = kmalloc(INFO_STRING_LEN, GFP_KERNEL);
10667         if (!buf)
10668                 return;
10669
10670         i = snprintf(buf, INFO_STRING_LEN, "Features: PF-id[%d]", hw->pf_id);
10671 #ifdef CONFIG_PCI_IOV
10672         i += snprintf(&buf[i], REMAIN(i), " VFs: %d", pf->num_req_vfs);
10673 #endif
10674         i += snprintf(&buf[i], REMAIN(i), " VSIs: %d QP: %d",
10675                       pf->hw.func_caps.num_vsis,
10676                       pf->vsi[pf->lan_vsi]->num_queue_pairs);
10677         if (pf->flags & I40E_FLAG_RSS_ENABLED)
10678                 i += snprintf(&buf[i], REMAIN(i), " RSS");
10679         if (pf->flags & I40E_FLAG_FD_ATR_ENABLED)
10680                 i += snprintf(&buf[i], REMAIN(i), " FD_ATR");
10681         if (pf->flags & I40E_FLAG_FD_SB_ENABLED) {
10682                 i += snprintf(&buf[i], REMAIN(i), " FD_SB");
10683                 i += snprintf(&buf[i], REMAIN(i), " NTUPLE");
10684         }
10685         if (pf->flags & I40E_FLAG_DCB_CAPABLE)
10686                 i += snprintf(&buf[i], REMAIN(i), " DCB");
10687         i += snprintf(&buf[i], REMAIN(i), " VxLAN");
10688         i += snprintf(&buf[i], REMAIN(i), " Geneve");
10689         if (pf->flags & I40E_FLAG_PTP)
10690                 i += snprintf(&buf[i], REMAIN(i), " PTP");
10691 #ifdef I40E_FCOE
10692         if (pf->flags & I40E_FLAG_FCOE_ENABLED)
10693                 i += snprintf(&buf[i], REMAIN(i), " FCOE");
10694 #endif
10695         if (pf->flags & I40E_FLAG_VEB_MODE_ENABLED)
10696                 i += snprintf(&buf[i], REMAIN(i), " VEB");
10697         else
10698                 i += snprintf(&buf[i], REMAIN(i), " VEPA");
10699
10700         dev_info(&pf->pdev->dev, "%s\n", buf);
10701         kfree(buf);
10702         WARN_ON(i > INFO_STRING_LEN);
10703 }
10704
10705 /**
10706  * i40e_get_platform_mac_addr - get platform-specific MAC address
10707  *
10708  * @pdev: PCI device information struct
10709  * @pf: board private structure
10710  *
10711  * Look up the MAC address in Open Firmware  on systems that support it,
10712  * and use IDPROM on SPARC if no OF address is found. On return, the
10713  * I40E_FLAG_PF_MAC will be wset in pf->flags if a platform-specific value
10714  * has been selected.
10715  **/
10716 static void i40e_get_platform_mac_addr(struct pci_dev *pdev, struct i40e_pf *pf)
10717 {
10718         pf->flags &= ~I40E_FLAG_PF_MAC;
10719         if (!eth_platform_get_mac_address(&pdev->dev, pf->hw.mac.addr))
10720                 pf->flags |= I40E_FLAG_PF_MAC;
10721 }
10722
10723 /**
10724  * i40e_probe - Device initialization routine
10725  * @pdev: PCI device information struct
10726  * @ent: entry in i40e_pci_tbl
10727  *
10728  * i40e_probe initializes a PF identified by a pci_dev structure.
10729  * The OS initialization, configuring of the PF private structure,
10730  * and a hardware reset occur.
10731  *
10732  * Returns 0 on success, negative on failure
10733  **/
10734 static int i40e_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
10735 {
10736         struct i40e_aq_get_phy_abilities_resp abilities;
10737         struct i40e_pf *pf;
10738         struct i40e_hw *hw;
10739         static u16 pfs_found;
10740         u16 wol_nvm_bits;
10741         u16 link_status;
10742         int err;
10743         u32 val;
10744         u32 i;
10745         u8 set_fc_aq_fail;
10746
10747         err = pci_enable_device_mem(pdev);
10748         if (err)
10749                 return err;
10750
10751         /* set up for high or low dma */
10752         err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
10753         if (err) {
10754                 err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
10755                 if (err) {
10756                         dev_err(&pdev->dev,
10757                                 "DMA configuration failed: 0x%x\n", err);
10758                         goto err_dma;
10759                 }
10760         }
10761
10762         /* set up pci connections */
10763         err = pci_request_mem_regions(pdev, i40e_driver_name);
10764         if (err) {
10765                 dev_info(&pdev->dev,
10766                          "pci_request_selected_regions failed %d\n", err);
10767                 goto err_pci_reg;
10768         }
10769
10770         pci_enable_pcie_error_reporting(pdev);
10771         pci_set_master(pdev);
10772
10773         /* Now that we have a PCI connection, we need to do the
10774          * low level device setup.  This is primarily setting up
10775          * the Admin Queue structures and then querying for the
10776          * device's current profile information.
10777          */
10778         pf = kzalloc(sizeof(*pf), GFP_KERNEL);
10779         if (!pf) {
10780                 err = -ENOMEM;
10781                 goto err_pf_alloc;
10782         }
10783         pf->next_vsi = 0;
10784         pf->pdev = pdev;
10785         set_bit(__I40E_DOWN, &pf->state);
10786
10787         hw = &pf->hw;
10788         hw->back = pf;
10789
10790         pf->ioremap_len = min_t(int, pci_resource_len(pdev, 0),
10791                                 I40E_MAX_CSR_SPACE);
10792
10793         hw->hw_addr = ioremap(pci_resource_start(pdev, 0), pf->ioremap_len);
10794         if (!hw->hw_addr) {
10795                 err = -EIO;
10796                 dev_info(&pdev->dev, "ioremap(0x%04x, 0x%04x) failed: 0x%x\n",
10797                          (unsigned int)pci_resource_start(pdev, 0),
10798                          pf->ioremap_len, err);
10799                 goto err_ioremap;
10800         }
10801         hw->vendor_id = pdev->vendor;
10802         hw->device_id = pdev->device;
10803         pci_read_config_byte(pdev, PCI_REVISION_ID, &hw->revision_id);
10804         hw->subsystem_vendor_id = pdev->subsystem_vendor;
10805         hw->subsystem_device_id = pdev->subsystem_device;
10806         hw->bus.device = PCI_SLOT(pdev->devfn);
10807         hw->bus.func = PCI_FUNC(pdev->devfn);
10808         pf->instance = pfs_found;
10809
10810         /* set up the locks for the AQ, do this only once in probe
10811          * and destroy them only once in remove
10812          */
10813         mutex_init(&hw->aq.asq_mutex);
10814         mutex_init(&hw->aq.arq_mutex);
10815
10816         if (debug != -1) {
10817                 pf->msg_enable = pf->hw.debug_mask;
10818                 pf->msg_enable = debug;
10819         }
10820
10821         /* do a special CORER for clearing PXE mode once at init */
10822         if (hw->revision_id == 0 &&
10823             (rd32(hw, I40E_GLLAN_RCTL_0) & I40E_GLLAN_RCTL_0_PXE_MODE_MASK)) {
10824                 wr32(hw, I40E_GLGEN_RTRIG, I40E_GLGEN_RTRIG_CORER_MASK);
10825                 i40e_flush(hw);
10826                 msleep(200);
10827                 pf->corer_count++;
10828
10829                 i40e_clear_pxe_mode(hw);
10830         }
10831
10832         /* Reset here to make sure all is clean and to define PF 'n' */
10833         i40e_clear_hw(hw);
10834         err = i40e_pf_reset(hw);
10835         if (err) {
10836                 dev_info(&pdev->dev, "Initial pf_reset failed: %d\n", err);
10837                 goto err_pf_reset;
10838         }
10839         pf->pfr_count++;
10840
10841         hw->aq.num_arq_entries = I40E_AQ_LEN;
10842         hw->aq.num_asq_entries = I40E_AQ_LEN;
10843         hw->aq.arq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10844         hw->aq.asq_buf_size = I40E_MAX_AQ_BUF_SIZE;
10845         pf->adminq_work_limit = I40E_AQ_WORK_LIMIT;
10846
10847         snprintf(pf->int_name, sizeof(pf->int_name) - 1,
10848                  "%s-%s:misc",
10849                  dev_driver_string(&pf->pdev->dev), dev_name(&pdev->dev));
10850
10851         err = i40e_init_shared_code(hw);
10852         if (err) {
10853                 dev_warn(&pdev->dev, "unidentified MAC or BLANK NVM: %d\n",
10854                          err);
10855                 goto err_pf_reset;
10856         }
10857
10858         /* set up a default setting for link flow control */
10859         pf->hw.fc.requested_mode = I40E_FC_NONE;
10860
10861         err = i40e_init_adminq(hw);
10862         if (err) {
10863                 if (err == I40E_ERR_FIRMWARE_API_VERSION)
10864                         dev_info(&pdev->dev,
10865                                  "The driver for the device stopped because the NVM image is newer than expected. You must install the most recent version of the network driver.\n");
10866                 else
10867                         dev_info(&pdev->dev,
10868                                  "The driver for the device stopped because the device firmware failed to init. Try updating your NVM image.\n");
10869
10870                 goto err_pf_reset;
10871         }
10872
10873         /* provide nvm, fw, api versions */
10874         dev_info(&pdev->dev, "fw %d.%d.%05d api %d.%d nvm %s\n",
10875                  hw->aq.fw_maj_ver, hw->aq.fw_min_ver, hw->aq.fw_build,
10876                  hw->aq.api_maj_ver, hw->aq.api_min_ver,
10877                  i40e_nvm_version_str(hw));
10878
10879         if (hw->aq.api_maj_ver == I40E_FW_API_VERSION_MAJOR &&
10880             hw->aq.api_min_ver > I40E_FW_API_VERSION_MINOR)
10881                 dev_info(&pdev->dev,
10882                          "The driver for the device detected a newer version of the NVM image than expected. Please install the most recent version of the network driver.\n");
10883         else if (hw->aq.api_maj_ver < I40E_FW_API_VERSION_MAJOR ||
10884                  hw->aq.api_min_ver < (I40E_FW_API_VERSION_MINOR - 1))
10885                 dev_info(&pdev->dev,
10886                          "The driver for the device detected an older version of the NVM image than expected. Please update the NVM image.\n");
10887
10888         i40e_verify_eeprom(pf);
10889
10890         /* Rev 0 hardware was never productized */
10891         if (hw->revision_id < 1)
10892                 dev_warn(&pdev->dev, "This device is a pre-production adapter/LOM. Please be aware there may be issues with your hardware. If you are experiencing problems please contact your Intel or hardware representative who provided you with this hardware.\n");
10893
10894         i40e_clear_pxe_mode(hw);
10895         err = i40e_get_capabilities(pf);
10896         if (err)
10897                 goto err_adminq_setup;
10898
10899         err = i40e_sw_init(pf);
10900         if (err) {
10901                 dev_info(&pdev->dev, "sw_init failed: %d\n", err);
10902                 goto err_sw_init;
10903         }
10904
10905         err = i40e_init_lan_hmc(hw, hw->func_caps.num_tx_qp,
10906                                 hw->func_caps.num_rx_qp,
10907                                 pf->fcoe_hmc_cntx_num, pf->fcoe_hmc_filt_num);
10908         if (err) {
10909                 dev_info(&pdev->dev, "init_lan_hmc failed: %d\n", err);
10910                 goto err_init_lan_hmc;
10911         }
10912
10913         err = i40e_configure_lan_hmc(hw, I40E_HMC_MODEL_DIRECT_ONLY);
10914         if (err) {
10915                 dev_info(&pdev->dev, "configure_lan_hmc failed: %d\n", err);
10916                 err = -ENOENT;
10917                 goto err_configure_lan_hmc;
10918         }
10919
10920         /* Disable LLDP for NICs that have firmware versions lower than v4.3.
10921          * Ignore error return codes because if it was already disabled via
10922          * hardware settings this will fail
10923          */
10924         if (pf->flags & I40E_FLAG_STOP_FW_LLDP) {
10925                 dev_info(&pdev->dev, "Stopping firmware LLDP agent.\n");
10926                 i40e_aq_stop_lldp(hw, true, NULL);
10927         }
10928
10929         i40e_get_mac_addr(hw, hw->mac.addr);
10930         /* allow a platform config to override the HW addr */
10931         i40e_get_platform_mac_addr(pdev, pf);
10932         if (!is_valid_ether_addr(hw->mac.addr)) {
10933                 dev_info(&pdev->dev, "invalid MAC address %pM\n", hw->mac.addr);
10934                 err = -EIO;
10935                 goto err_mac_addr;
10936         }
10937         dev_info(&pdev->dev, "MAC address: %pM\n", hw->mac.addr);
10938         ether_addr_copy(hw->mac.perm_addr, hw->mac.addr);
10939         i40e_get_port_mac_addr(hw, hw->mac.port_addr);
10940         if (is_valid_ether_addr(hw->mac.port_addr))
10941                 pf->flags |= I40E_FLAG_PORT_ID_VALID;
10942 #ifdef I40E_FCOE
10943         err = i40e_get_san_mac_addr(hw, hw->mac.san_addr);
10944         if (err)
10945                 dev_info(&pdev->dev,
10946                          "(non-fatal) SAN MAC retrieval failed: %d\n", err);
10947         if (!is_valid_ether_addr(hw->mac.san_addr)) {
10948                 dev_warn(&pdev->dev, "invalid SAN MAC address %pM, falling back to LAN MAC\n",
10949                          hw->mac.san_addr);
10950                 ether_addr_copy(hw->mac.san_addr, hw->mac.addr);
10951         }
10952         dev_info(&pf->pdev->dev, "SAN MAC: %pM\n", hw->mac.san_addr);
10953 #endif /* I40E_FCOE */
10954
10955         pci_set_drvdata(pdev, pf);
10956         pci_save_state(pdev);
10957 #ifdef CONFIG_I40E_DCB
10958         err = i40e_init_pf_dcb(pf);
10959         if (err) {
10960                 dev_info(&pdev->dev, "DCB init failed %d, disabled\n", err);
10961                 pf->flags &= ~(I40E_FLAG_DCB_CAPABLE & I40E_FLAG_DCB_ENABLED);
10962                 /* Continue without DCB enabled */
10963         }
10964 #endif /* CONFIG_I40E_DCB */
10965
10966         /* set up periodic task facility */
10967         setup_timer(&pf->service_timer, i40e_service_timer, (unsigned long)pf);
10968         pf->service_timer_period = HZ;
10969
10970         INIT_WORK(&pf->service_task, i40e_service_task);
10971         clear_bit(__I40E_SERVICE_SCHED, &pf->state);
10972         pf->flags |= I40E_FLAG_NEED_LINK_UPDATE;
10973
10974         /* NVM bit on means WoL disabled for the port */
10975         i40e_read_nvm_word(hw, I40E_SR_NVM_WAKE_ON_LAN, &wol_nvm_bits);
10976         if (BIT (hw->port) & wol_nvm_bits || hw->partition_id != 1)
10977                 pf->wol_en = false;
10978         else
10979                 pf->wol_en = true;
10980         device_set_wakeup_enable(&pf->pdev->dev, pf->wol_en);
10981
10982         /* set up the main switch operations */
10983         i40e_determine_queue_usage(pf);
10984         err = i40e_init_interrupt_scheme(pf);
10985         if (err)
10986                 goto err_switch_setup;
10987
10988         /* The number of VSIs reported by the FW is the minimum guaranteed
10989          * to us; HW supports far more and we share the remaining pool with
10990          * the other PFs. We allocate space for more than the guarantee with
10991          * the understanding that we might not get them all later.
10992          */
10993         if (pf->hw.func_caps.num_vsis < I40E_MIN_VSI_ALLOC)
10994                 pf->num_alloc_vsi = I40E_MIN_VSI_ALLOC;
10995         else
10996                 pf->num_alloc_vsi = pf->hw.func_caps.num_vsis;
10997
10998         /* Set up the *vsi struct and our local tracking of the MAIN PF vsi. */
10999         pf->vsi = kcalloc(pf->num_alloc_vsi, sizeof(struct i40e_vsi *),
11000                           GFP_KERNEL);
11001         if (!pf->vsi) {
11002                 err = -ENOMEM;
11003                 goto err_switch_setup;
11004         }
11005
11006 #ifdef CONFIG_PCI_IOV
11007         /* prep for VF support */
11008         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11009             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11010             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11011                 if (pci_num_vf(pdev))
11012                         pf->flags |= I40E_FLAG_VEB_MODE_ENABLED;
11013         }
11014 #endif
11015         err = i40e_setup_pf_switch(pf, false);
11016         if (err) {
11017                 dev_info(&pdev->dev, "setup_pf_switch failed: %d\n", err);
11018                 goto err_vsis;
11019         }
11020
11021         /* Make sure flow control is set according to current settings */
11022         err = i40e_set_fc(hw, &set_fc_aq_fail, true);
11023         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_GET)
11024                 dev_dbg(&pf->pdev->dev,
11025                         "Set fc with err %s aq_err %s on get_phy_cap\n",
11026                         i40e_stat_str(hw, err),
11027                         i40e_aq_str(hw, hw->aq.asq_last_status));
11028         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_SET)
11029                 dev_dbg(&pf->pdev->dev,
11030                         "Set fc with err %s aq_err %s on set_phy_config\n",
11031                         i40e_stat_str(hw, err),
11032                         i40e_aq_str(hw, hw->aq.asq_last_status));
11033         if (set_fc_aq_fail & I40E_SET_FC_AQ_FAIL_UPDATE)
11034                 dev_dbg(&pf->pdev->dev,
11035                         "Set fc with err %s aq_err %s on get_link_info\n",
11036                         i40e_stat_str(hw, err),
11037                         i40e_aq_str(hw, hw->aq.asq_last_status));
11038
11039         /* if FDIR VSI was set up, start it now */
11040         for (i = 0; i < pf->num_alloc_vsi; i++) {
11041                 if (pf->vsi[i] && pf->vsi[i]->type == I40E_VSI_FDIR) {
11042                         i40e_vsi_open(pf->vsi[i]);
11043                         break;
11044                 }
11045         }
11046
11047         /* The driver only wants link up/down and module qualification
11048          * reports from firmware.  Note the negative logic.
11049          */
11050         err = i40e_aq_set_phy_int_mask(&pf->hw,
11051                                        ~(I40E_AQ_EVENT_LINK_UPDOWN |
11052                                          I40E_AQ_EVENT_MEDIA_NA |
11053                                          I40E_AQ_EVENT_MODULE_QUAL_FAIL), NULL);
11054         if (err)
11055                 dev_info(&pf->pdev->dev, "set phy mask fail, err %s aq_err %s\n",
11056                          i40e_stat_str(&pf->hw, err),
11057                          i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11058
11059         /* Reconfigure hardware for allowing smaller MSS in the case
11060          * of TSO, so that we avoid the MDD being fired and causing
11061          * a reset in the case of small MSS+TSO.
11062          */
11063         val = rd32(hw, I40E_REG_MSS);
11064         if ((val & I40E_REG_MSS_MIN_MASK) > I40E_64BYTE_MSS) {
11065                 val &= ~I40E_REG_MSS_MIN_MASK;
11066                 val |= I40E_64BYTE_MSS;
11067                 wr32(hw, I40E_REG_MSS, val);
11068         }
11069
11070         if (pf->flags & I40E_FLAG_RESTART_AUTONEG) {
11071                 msleep(75);
11072                 err = i40e_aq_set_link_restart_an(&pf->hw, true, NULL);
11073                 if (err)
11074                         dev_info(&pf->pdev->dev, "link restart failed, err %s aq_err %s\n",
11075                                  i40e_stat_str(&pf->hw, err),
11076                                  i40e_aq_str(&pf->hw,
11077                                              pf->hw.aq.asq_last_status));
11078         }
11079         /* The main driver is (mostly) up and happy. We need to set this state
11080          * before setting up the misc vector or we get a race and the vector
11081          * ends up disabled forever.
11082          */
11083         clear_bit(__I40E_DOWN, &pf->state);
11084
11085         /* In case of MSIX we are going to setup the misc vector right here
11086          * to handle admin queue events etc. In case of legacy and MSI
11087          * the misc functionality and queue processing is combined in
11088          * the same vector and that gets setup at open.
11089          */
11090         if (pf->flags & I40E_FLAG_MSIX_ENABLED) {
11091                 err = i40e_setup_misc_vector(pf);
11092                 if (err) {
11093                         dev_info(&pdev->dev,
11094                                  "setup of misc vector failed: %d\n", err);
11095                         goto err_vsis;
11096                 }
11097         }
11098
11099 #ifdef CONFIG_PCI_IOV
11100         /* prep for VF support */
11101         if ((pf->flags & I40E_FLAG_SRIOV_ENABLED) &&
11102             (pf->flags & I40E_FLAG_MSIX_ENABLED) &&
11103             !test_bit(__I40E_BAD_EEPROM, &pf->state)) {
11104                 /* disable link interrupts for VFs */
11105                 val = rd32(hw, I40E_PFGEN_PORTMDIO_NUM);
11106                 val &= ~I40E_PFGEN_PORTMDIO_NUM_VFLINK_STAT_ENA_MASK;
11107                 wr32(hw, I40E_PFGEN_PORTMDIO_NUM, val);
11108                 i40e_flush(hw);
11109
11110                 if (pci_num_vf(pdev)) {
11111                         dev_info(&pdev->dev,
11112                                  "Active VFs found, allocating resources.\n");
11113                         err = i40e_alloc_vfs(pf, pci_num_vf(pdev));
11114                         if (err)
11115                                 dev_info(&pdev->dev,
11116                                          "Error %d allocating resources for existing VFs\n",
11117                                          err);
11118                 }
11119         }
11120 #endif /* CONFIG_PCI_IOV */
11121
11122         if (pf->flags & I40E_FLAG_IWARP_ENABLED) {
11123                 pf->iwarp_base_vector = i40e_get_lump(pf, pf->irq_pile,
11124                                                       pf->num_iwarp_msix,
11125                                                       I40E_IWARP_IRQ_PILE_ID);
11126                 if (pf->iwarp_base_vector < 0) {
11127                         dev_info(&pdev->dev,
11128                                  "failed to get tracking for %d vectors for IWARP err=%d\n",
11129                                  pf->num_iwarp_msix, pf->iwarp_base_vector);
11130                         pf->flags &= ~I40E_FLAG_IWARP_ENABLED;
11131                 }
11132         }
11133
11134         i40e_dbg_pf_init(pf);
11135
11136         /* tell the firmware that we're starting */
11137         i40e_send_version(pf);
11138
11139         /* since everything's happy, start the service_task timer */
11140         mod_timer(&pf->service_timer,
11141                   round_jiffies(jiffies + pf->service_timer_period));
11142
11143         /* add this PF to client device list and launch a client service task */
11144         err = i40e_lan_add_device(pf);
11145         if (err)
11146                 dev_info(&pdev->dev, "Failed to add PF to client API service list: %d\n",
11147                          err);
11148
11149 #ifdef I40E_FCOE
11150         /* create FCoE interface */
11151         i40e_fcoe_vsi_setup(pf);
11152
11153 #endif
11154 #define PCI_SPEED_SIZE 8
11155 #define PCI_WIDTH_SIZE 8
11156         /* Devices on the IOSF bus do not have this information
11157          * and will report PCI Gen 1 x 1 by default so don't bother
11158          * checking them.
11159          */
11160         if (!(pf->flags & I40E_FLAG_NO_PCI_LINK_CHECK)) {
11161                 char speed[PCI_SPEED_SIZE] = "Unknown";
11162                 char width[PCI_WIDTH_SIZE] = "Unknown";
11163
11164                 /* Get the negotiated link width and speed from PCI config
11165                  * space
11166                  */
11167                 pcie_capability_read_word(pf->pdev, PCI_EXP_LNKSTA,
11168                                           &link_status);
11169
11170                 i40e_set_pci_config_data(hw, link_status);
11171
11172                 switch (hw->bus.speed) {
11173                 case i40e_bus_speed_8000:
11174                         strncpy(speed, "8.0", PCI_SPEED_SIZE); break;
11175                 case i40e_bus_speed_5000:
11176                         strncpy(speed, "5.0", PCI_SPEED_SIZE); break;
11177                 case i40e_bus_speed_2500:
11178                         strncpy(speed, "2.5", PCI_SPEED_SIZE); break;
11179                 default:
11180                         break;
11181                 }
11182                 switch (hw->bus.width) {
11183                 case i40e_bus_width_pcie_x8:
11184                         strncpy(width, "8", PCI_WIDTH_SIZE); break;
11185                 case i40e_bus_width_pcie_x4:
11186                         strncpy(width, "4", PCI_WIDTH_SIZE); break;
11187                 case i40e_bus_width_pcie_x2:
11188                         strncpy(width, "2", PCI_WIDTH_SIZE); break;
11189                 case i40e_bus_width_pcie_x1:
11190                         strncpy(width, "1", PCI_WIDTH_SIZE); break;
11191                 default:
11192                         break;
11193                 }
11194
11195                 dev_info(&pdev->dev, "PCI-Express: Speed %sGT/s Width x%s\n",
11196                          speed, width);
11197
11198                 if (hw->bus.width < i40e_bus_width_pcie_x8 ||
11199                     hw->bus.speed < i40e_bus_speed_8000) {
11200                         dev_warn(&pdev->dev, "PCI-Express bandwidth available for this device may be insufficient for optimal performance.\n");
11201                         dev_warn(&pdev->dev, "Please move the device to a different PCI-e link with more lanes and/or higher transfer rate.\n");
11202                 }
11203         }
11204
11205         /* get the requested speeds from the fw */
11206         err = i40e_aq_get_phy_capabilities(hw, false, false, &abilities, NULL);
11207         if (err)
11208                 dev_dbg(&pf->pdev->dev, "get requested speeds ret =  %s last_status =  %s\n",
11209                         i40e_stat_str(&pf->hw, err),
11210                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11211         pf->hw.phy.link_info.requested_speeds = abilities.link_speed;
11212
11213         /* get the supported phy types from the fw */
11214         err = i40e_aq_get_phy_capabilities(hw, false, true, &abilities, NULL);
11215         if (err)
11216                 dev_dbg(&pf->pdev->dev, "get supported phy types ret =  %s last_status =  %s\n",
11217                         i40e_stat_str(&pf->hw, err),
11218                         i40e_aq_str(&pf->hw, pf->hw.aq.asq_last_status));
11219         pf->hw.phy.phy_types = le32_to_cpu(abilities.phy_type);
11220
11221         /* Add a filter to drop all Flow control frames from any VSI from being
11222          * transmitted. By doing so we stop a malicious VF from sending out
11223          * PAUSE or PFC frames and potentially controlling traffic for other
11224          * PF/VF VSIs.
11225          * The FW can still send Flow control frames if enabled.
11226          */
11227         i40e_add_filter_to_drop_tx_flow_control_frames(&pf->hw,
11228                                                        pf->main_vsi_seid);
11229
11230         if ((pf->hw.device_id == I40E_DEV_ID_10G_BASE_T) ||
11231             (pf->hw.device_id == I40E_DEV_ID_10G_BASE_T4))
11232                 pf->flags |= I40E_FLAG_HAVE_10GBASET_PHY;
11233
11234         /* print a string summarizing features */
11235         i40e_print_features(pf);
11236
11237         return 0;
11238
11239         /* Unwind what we've done if something failed in the setup */
11240 err_vsis:
11241         set_bit(__I40E_DOWN, &pf->state);
11242         i40e_clear_interrupt_scheme(pf);
11243         kfree(pf->vsi);
11244 err_switch_setup:
11245         i40e_reset_interrupt_capability(pf);
11246         del_timer_sync(&pf->service_timer);
11247 err_mac_addr:
11248 err_configure_lan_hmc:
11249         (void)i40e_shutdown_lan_hmc(hw);
11250 err_init_lan_hmc:
11251         kfree(pf->qp_pile);
11252 err_sw_init:
11253 err_adminq_setup:
11254 err_pf_reset:
11255         iounmap(hw->hw_addr);
11256 err_ioremap:
11257         kfree(pf);
11258 err_pf_alloc:
11259         pci_disable_pcie_error_reporting(pdev);
11260         pci_release_mem_regions(pdev);
11261 err_pci_reg:
11262 err_dma:
11263         pci_disable_device(pdev);
11264         return err;
11265 }
11266
11267 /**
11268  * i40e_remove - Device removal routine
11269  * @pdev: PCI device information struct
11270  *
11271  * i40e_remove is called by the PCI subsystem to alert the driver
11272  * that is should release a PCI device.  This could be caused by a
11273  * Hot-Plug event, or because the driver is going to be removed from
11274  * memory.
11275  **/
11276 static void i40e_remove(struct pci_dev *pdev)
11277 {
11278         struct i40e_pf *pf = pci_get_drvdata(pdev);
11279         struct i40e_hw *hw = &pf->hw;
11280         i40e_status ret_code;
11281         int i;
11282
11283         i40e_dbg_pf_exit(pf);
11284
11285         i40e_ptp_stop(pf);
11286
11287         /* Disable RSS in hw */
11288         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(0), 0);
11289         i40e_write_rx_ctl(hw, I40E_PFQF_HENA(1), 0);
11290
11291         /* no more scheduling of any task */
11292         set_bit(__I40E_SUSPENDED, &pf->state);
11293         set_bit(__I40E_DOWN, &pf->state);
11294         if (pf->service_timer.data)
11295                 del_timer_sync(&pf->service_timer);
11296         if (pf->service_task.func)
11297                 cancel_work_sync(&pf->service_task);
11298
11299         if (pf->flags & I40E_FLAG_SRIOV_ENABLED) {
11300                 i40e_free_vfs(pf);
11301                 pf->flags &= ~I40E_FLAG_SRIOV_ENABLED;
11302         }
11303
11304         i40e_fdir_teardown(pf);
11305
11306         /* If there is a switch structure or any orphans, remove them.
11307          * This will leave only the PF's VSI remaining.
11308          */
11309         for (i = 0; i < I40E_MAX_VEB; i++) {
11310                 if (!pf->veb[i])
11311                         continue;
11312
11313                 if (pf->veb[i]->uplink_seid == pf->mac_seid ||
11314                     pf->veb[i]->uplink_seid == 0)
11315                         i40e_switch_branch_release(pf->veb[i]);
11316         }
11317
11318         /* Now we can shutdown the PF's VSI, just before we kill
11319          * adminq and hmc.
11320          */
11321         if (pf->vsi[pf->lan_vsi])
11322                 i40e_vsi_release(pf->vsi[pf->lan_vsi]);
11323
11324         /* remove attached clients */
11325         ret_code = i40e_lan_del_device(pf);
11326         if (ret_code) {
11327                 dev_warn(&pdev->dev, "Failed to delete client device: %d\n",
11328                          ret_code);
11329         }
11330
11331         /* shutdown and destroy the HMC */
11332         if (hw->hmc.hmc_obj) {
11333                 ret_code = i40e_shutdown_lan_hmc(hw);
11334                 if (ret_code)
11335                         dev_warn(&pdev->dev,
11336                                  "Failed to destroy the HMC resources: %d\n",
11337                                  ret_code);
11338         }
11339
11340         /* shutdown the adminq */
11341         ret_code = i40e_shutdown_adminq(hw);
11342         if (ret_code)
11343                 dev_warn(&pdev->dev,
11344                          "Failed to destroy the Admin Queue resources: %d\n",
11345                          ret_code);
11346
11347         /* destroy the locks only once, here */
11348         mutex_destroy(&hw->aq.arq_mutex);
11349         mutex_destroy(&hw->aq.asq_mutex);
11350
11351         /* Clear all dynamic memory lists of rings, q_vectors, and VSIs */
11352         i40e_clear_interrupt_scheme(pf);
11353         for (i = 0; i < pf->num_alloc_vsi; i++) {
11354                 if (pf->vsi[i]) {
11355                         i40e_vsi_clear_rings(pf->vsi[i]);
11356                         i40e_vsi_clear(pf->vsi[i]);
11357                         pf->vsi[i] = NULL;
11358                 }
11359         }
11360
11361         for (i = 0; i < I40E_MAX_VEB; i++) {
11362                 kfree(pf->veb[i]);
11363                 pf->veb[i] = NULL;
11364         }
11365
11366         kfree(pf->qp_pile);
11367         kfree(pf->vsi);
11368
11369         iounmap(hw->hw_addr);
11370         kfree(pf);
11371         pci_release_mem_regions(pdev);
11372
11373         pci_disable_pcie_error_reporting(pdev);
11374         pci_disable_device(pdev);
11375 }
11376
11377 /**
11378  * i40e_pci_error_detected - warning that something funky happened in PCI land
11379  * @pdev: PCI device information struct
11380  *
11381  * Called to warn that something happened and the error handling steps
11382  * are in progress.  Allows the driver to quiesce things, be ready for
11383  * remediation.
11384  **/
11385 static pci_ers_result_t i40e_pci_error_detected(struct pci_dev *pdev,
11386                                                 enum pci_channel_state error)
11387 {
11388         struct i40e_pf *pf = pci_get_drvdata(pdev);
11389
11390         dev_info(&pdev->dev, "%s: error %d\n", __func__, error);
11391
11392         /* shutdown all operations */
11393         if (!test_bit(__I40E_SUSPENDED, &pf->state)) {
11394                 rtnl_lock();
11395                 i40e_prep_for_reset(pf);
11396                 rtnl_unlock();
11397         }
11398
11399         /* Request a slot reset */
11400         return PCI_ERS_RESULT_NEED_RESET;
11401 }
11402
11403 /**
11404  * i40e_pci_error_slot_reset - a PCI slot reset just happened
11405  * @pdev: PCI device information struct
11406  *
11407  * Called to find if the driver can work with the device now that
11408  * the pci slot has been reset.  If a basic connection seems good
11409  * (registers are readable and have sane content) then return a
11410  * happy little PCI_ERS_RESULT_xxx.
11411  **/
11412 static pci_ers_result_t i40e_pci_error_slot_reset(struct pci_dev *pdev)
11413 {
11414         struct i40e_pf *pf = pci_get_drvdata(pdev);
11415         pci_ers_result_t result;
11416         int err;
11417         u32 reg;
11418
11419         dev_dbg(&pdev->dev, "%s\n", __func__);
11420         if (pci_enable_device_mem(pdev)) {
11421                 dev_info(&pdev->dev,
11422                          "Cannot re-enable PCI device after reset.\n");
11423                 result = PCI_ERS_RESULT_DISCONNECT;
11424         } else {
11425                 pci_set_master(pdev);
11426                 pci_restore_state(pdev);
11427                 pci_save_state(pdev);
11428                 pci_wake_from_d3(pdev, false);
11429
11430                 reg = rd32(&pf->hw, I40E_GLGEN_RTRIG);
11431                 if (reg == 0)
11432                         result = PCI_ERS_RESULT_RECOVERED;
11433                 else
11434                         result = PCI_ERS_RESULT_DISCONNECT;
11435         }
11436
11437         err = pci_cleanup_aer_uncorrect_error_status(pdev);
11438         if (err) {
11439                 dev_info(&pdev->dev,
11440                          "pci_cleanup_aer_uncorrect_error_status failed 0x%0x\n",
11441                          err);
11442                 /* non-fatal, continue */
11443         }
11444
11445         return result;
11446 }
11447
11448 /**
11449  * i40e_pci_error_resume - restart operations after PCI error recovery
11450  * @pdev: PCI device information struct
11451  *
11452  * Called to allow the driver to bring things back up after PCI error
11453  * and/or reset recovery has finished.
11454  **/
11455 static void i40e_pci_error_resume(struct pci_dev *pdev)
11456 {
11457         struct i40e_pf *pf = pci_get_drvdata(pdev);
11458
11459         dev_dbg(&pdev->dev, "%s\n", __func__);
11460         if (test_bit(__I40E_SUSPENDED, &pf->state))
11461                 return;
11462
11463         rtnl_lock();
11464         i40e_handle_reset_warning(pf);
11465         rtnl_unlock();
11466 }
11467
11468 /**
11469  * i40e_shutdown - PCI callback for shutting down
11470  * @pdev: PCI device information struct
11471  **/
11472 static void i40e_shutdown(struct pci_dev *pdev)
11473 {
11474         struct i40e_pf *pf = pci_get_drvdata(pdev);
11475         struct i40e_hw *hw = &pf->hw;
11476
11477         set_bit(__I40E_SUSPENDED, &pf->state);
11478         set_bit(__I40E_DOWN, &pf->state);
11479         rtnl_lock();
11480         i40e_prep_for_reset(pf);
11481         rtnl_unlock();
11482
11483         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11484         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11485
11486         del_timer_sync(&pf->service_timer);
11487         cancel_work_sync(&pf->service_task);
11488         i40e_fdir_teardown(pf);
11489
11490         rtnl_lock();
11491         i40e_prep_for_reset(pf);
11492         rtnl_unlock();
11493
11494         wr32(hw, I40E_PFPM_APM,
11495              (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11496         wr32(hw, I40E_PFPM_WUFC,
11497              (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11498
11499         i40e_clear_interrupt_scheme(pf);
11500
11501         if (system_state == SYSTEM_POWER_OFF) {
11502                 pci_wake_from_d3(pdev, pf->wol_en);
11503                 pci_set_power_state(pdev, PCI_D3hot);
11504         }
11505 }
11506
11507 #ifdef CONFIG_PM
11508 /**
11509  * i40e_suspend - PCI callback for moving to D3
11510  * @pdev: PCI device information struct
11511  **/
11512 static int i40e_suspend(struct pci_dev *pdev, pm_message_t state)
11513 {
11514         struct i40e_pf *pf = pci_get_drvdata(pdev);
11515         struct i40e_hw *hw = &pf->hw;
11516         int retval = 0;
11517
11518         set_bit(__I40E_SUSPENDED, &pf->state);
11519         set_bit(__I40E_DOWN, &pf->state);
11520
11521         rtnl_lock();
11522         i40e_prep_for_reset(pf);
11523         rtnl_unlock();
11524
11525         wr32(hw, I40E_PFPM_APM, (pf->wol_en ? I40E_PFPM_APM_APME_MASK : 0));
11526         wr32(hw, I40E_PFPM_WUFC, (pf->wol_en ? I40E_PFPM_WUFC_MAG_MASK : 0));
11527
11528         i40e_stop_misc_vector(pf);
11529
11530         retval = pci_save_state(pdev);
11531         if (retval)
11532                 return retval;
11533
11534         pci_wake_from_d3(pdev, pf->wol_en);
11535         pci_set_power_state(pdev, PCI_D3hot);
11536
11537         return retval;
11538 }
11539
11540 /**
11541  * i40e_resume - PCI callback for waking up from D3
11542  * @pdev: PCI device information struct
11543  **/
11544 static int i40e_resume(struct pci_dev *pdev)
11545 {
11546         struct i40e_pf *pf = pci_get_drvdata(pdev);
11547         u32 err;
11548
11549         pci_set_power_state(pdev, PCI_D0);
11550         pci_restore_state(pdev);
11551         /* pci_restore_state() clears dev->state_saves, so
11552          * call pci_save_state() again to restore it.
11553          */
11554         pci_save_state(pdev);
11555
11556         err = pci_enable_device_mem(pdev);
11557         if (err) {
11558                 dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
11559                 return err;
11560         }
11561         pci_set_master(pdev);
11562
11563         /* no wakeup events while running */
11564         pci_wake_from_d3(pdev, false);
11565
11566         /* handling the reset will rebuild the device state */
11567         if (test_and_clear_bit(__I40E_SUSPENDED, &pf->state)) {
11568                 clear_bit(__I40E_DOWN, &pf->state);
11569                 rtnl_lock();
11570                 i40e_reset_and_rebuild(pf, false);
11571                 rtnl_unlock();
11572         }
11573
11574         return 0;
11575 }
11576
11577 #endif
11578 static const struct pci_error_handlers i40e_err_handler = {
11579         .error_detected = i40e_pci_error_detected,
11580         .slot_reset = i40e_pci_error_slot_reset,
11581         .resume = i40e_pci_error_resume,
11582 };
11583
11584 static struct pci_driver i40e_driver = {
11585         .name     = i40e_driver_name,
11586         .id_table = i40e_pci_tbl,
11587         .probe    = i40e_probe,
11588         .remove   = i40e_remove,
11589 #ifdef CONFIG_PM
11590         .suspend  = i40e_suspend,
11591         .resume   = i40e_resume,
11592 #endif
11593         .shutdown = i40e_shutdown,
11594         .err_handler = &i40e_err_handler,
11595         .sriov_configure = i40e_pci_sriov_configure,
11596 };
11597
11598 /**
11599  * i40e_init_module - Driver registration routine
11600  *
11601  * i40e_init_module is the first routine called when the driver is
11602  * loaded. All it does is register with the PCI subsystem.
11603  **/
11604 static int __init i40e_init_module(void)
11605 {
11606         pr_info("%s: %s - version %s\n", i40e_driver_name,
11607                 i40e_driver_string, i40e_driver_version_str);
11608         pr_info("%s: %s\n", i40e_driver_name, i40e_copyright);
11609
11610         /* we will see if single thread per module is enough for now,
11611          * it can't be any worse than using the system workqueue which
11612          * was already single threaded
11613          */
11614         i40e_wq = alloc_workqueue("%s", WQ_UNBOUND | WQ_MEM_RECLAIM, 1,
11615                                   i40e_driver_name);
11616         if (!i40e_wq) {
11617                 pr_err("%s: Failed to create workqueue\n", i40e_driver_name);
11618                 return -ENOMEM;
11619         }
11620
11621         i40e_dbg_init();
11622         return pci_register_driver(&i40e_driver);
11623 }
11624 module_init(i40e_init_module);
11625
11626 /**
11627  * i40e_exit_module - Driver exit cleanup routine
11628  *
11629  * i40e_exit_module is called just before the driver is removed
11630  * from memory.
11631  **/
11632 static void __exit i40e_exit_module(void)
11633 {
11634         pci_unregister_driver(&i40e_driver);
11635         destroy_workqueue(i40e_wq);
11636         i40e_dbg_exit();
11637 }
11638 module_exit(i40e_exit_module);