drivers/net: fix exposing internal headers
[dpdk.git] / drivers / net / iavf / iavf_ethdev.c
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2017 Intel Corporation
3  */
4
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <string.h>
10 #include <unistd.h>
11 #include <stdarg.h>
12 #include <inttypes.h>
13 #include <rte_byteorder.h>
14 #include <rte_common.h>
15
16 #include <rte_interrupts.h>
17 #include <rte_debug.h>
18 #include <rte_pci.h>
19 #include <rte_atomic.h>
20 #include <rte_eal.h>
21 #include <rte_ether.h>
22 #include <rte_ethdev_driver.h>
23 #include <rte_ethdev_pci.h>
24 #include <rte_malloc.h>
25 #include <rte_memzone.h>
26 #include <rte_dev.h>
27
28 #include "iavf.h"
29 #include "iavf_rxtx.h"
30 #include "iavf_generic_flow.h"
31
32 static int iavf_dev_configure(struct rte_eth_dev *dev);
33 static int iavf_dev_start(struct rte_eth_dev *dev);
34 static void iavf_dev_stop(struct rte_eth_dev *dev);
35 static void iavf_dev_close(struct rte_eth_dev *dev);
36 static int iavf_dev_reset(struct rte_eth_dev *dev);
37 static int iavf_dev_info_get(struct rte_eth_dev *dev,
38                              struct rte_eth_dev_info *dev_info);
39 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
40 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
41                              struct rte_eth_stats *stats);
42 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
43 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
44 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
45 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
46 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
47 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
48                                 struct rte_ether_addr *addr,
49                                 uint32_t index,
50                                 uint32_t pool);
51 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
52 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
53                                    uint16_t vlan_id, int on);
54 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
55 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
56                                    struct rte_eth_rss_reta_entry64 *reta_conf,
57                                    uint16_t reta_size);
58 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
59                                   struct rte_eth_rss_reta_entry64 *reta_conf,
60                                   uint16_t reta_size);
61 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
62                                    struct rte_eth_rss_conf *rss_conf);
63 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
64                                      struct rte_eth_rss_conf *rss_conf);
65 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
66 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
67                                          struct rte_ether_addr *mac_addr);
68 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
69                                         uint16_t queue_id);
70 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
71                                          uint16_t queue_id);
72 static int iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
73                      enum rte_filter_type filter_type,
74                      enum rte_filter_op filter_op,
75                      void *arg);
76
77 static const struct rte_pci_id pci_id_iavf_map[] = {
78         { RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
79         { .vendor_id = 0, /* sentinel */ },
80 };
81
82 static const struct eth_dev_ops iavf_eth_dev_ops = {
83         .dev_configure              = iavf_dev_configure,
84         .dev_start                  = iavf_dev_start,
85         .dev_stop                   = iavf_dev_stop,
86         .dev_close                  = iavf_dev_close,
87         .dev_reset                  = iavf_dev_reset,
88         .dev_infos_get              = iavf_dev_info_get,
89         .dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
90         .link_update                = iavf_dev_link_update,
91         .stats_get                  = iavf_dev_stats_get,
92         .stats_reset                = iavf_dev_stats_reset,
93         .promiscuous_enable         = iavf_dev_promiscuous_enable,
94         .promiscuous_disable        = iavf_dev_promiscuous_disable,
95         .allmulticast_enable        = iavf_dev_allmulticast_enable,
96         .allmulticast_disable       = iavf_dev_allmulticast_disable,
97         .mac_addr_add               = iavf_dev_add_mac_addr,
98         .mac_addr_remove            = iavf_dev_del_mac_addr,
99         .vlan_filter_set            = iavf_dev_vlan_filter_set,
100         .vlan_offload_set           = iavf_dev_vlan_offload_set,
101         .rx_queue_start             = iavf_dev_rx_queue_start,
102         .rx_queue_stop              = iavf_dev_rx_queue_stop,
103         .tx_queue_start             = iavf_dev_tx_queue_start,
104         .tx_queue_stop              = iavf_dev_tx_queue_stop,
105         .rx_queue_setup             = iavf_dev_rx_queue_setup,
106         .rx_queue_release           = iavf_dev_rx_queue_release,
107         .tx_queue_setup             = iavf_dev_tx_queue_setup,
108         .tx_queue_release           = iavf_dev_tx_queue_release,
109         .mac_addr_set               = iavf_dev_set_default_mac_addr,
110         .reta_update                = iavf_dev_rss_reta_update,
111         .reta_query                 = iavf_dev_rss_reta_query,
112         .rss_hash_update            = iavf_dev_rss_hash_update,
113         .rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
114         .rxq_info_get               = iavf_dev_rxq_info_get,
115         .txq_info_get               = iavf_dev_txq_info_get,
116         .rx_queue_count             = iavf_dev_rxq_count,
117         .rx_descriptor_status       = iavf_dev_rx_desc_status,
118         .tx_descriptor_status       = iavf_dev_tx_desc_status,
119         .mtu_set                    = iavf_dev_mtu_set,
120         .rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
121         .rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
122         .filter_ctrl                = iavf_dev_filter_ctrl,
123 };
124
125 static int
126 iavf_init_rss(struct iavf_adapter *adapter)
127 {
128         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
129         struct rte_eth_rss_conf *rss_conf;
130         uint8_t i, j, nb_q;
131         int ret;
132
133         rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
134         nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
135                        IAVF_MAX_NUM_QUEUES);
136
137         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
138                 PMD_DRV_LOG(DEBUG, "RSS is not supported");
139                 return -ENOTSUP;
140         }
141         if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
142                 PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
143                 /* set all lut items to default queue */
144                 for (i = 0; i < vf->vf_res->rss_lut_size; i++)
145                         vf->rss_lut[i] = 0;
146                 ret = iavf_configure_rss_lut(adapter);
147                 return ret;
148         }
149
150         /* In IAVF, RSS enablement is set by PF driver. It is not supported
151          * to set based on rss_conf->rss_hf.
152          */
153
154         /* configure RSS key */
155         if (!rss_conf->rss_key) {
156                 /* Calculate the default hash key */
157                 for (i = 0; i <= vf->vf_res->rss_key_size; i++)
158                         vf->rss_key[i] = (uint8_t)rte_rand();
159         } else
160                 rte_memcpy(vf->rss_key, rss_conf->rss_key,
161                            RTE_MIN(rss_conf->rss_key_len,
162                                    vf->vf_res->rss_key_size));
163
164         /* init RSS LUT table */
165         for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
166                 if (j >= nb_q)
167                         j = 0;
168                 vf->rss_lut[i] = j;
169         }
170         /* send virtchnnl ops to configure rss*/
171         ret = iavf_configure_rss_lut(adapter);
172         if (ret)
173                 return ret;
174         ret = iavf_configure_rss_key(adapter);
175         if (ret)
176                 return ret;
177
178         return 0;
179 }
180
181 static int
182 iavf_dev_configure(struct rte_eth_dev *dev)
183 {
184         struct iavf_adapter *ad =
185                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
186         struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
187         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
188
189         ad->rx_bulk_alloc_allowed = true;
190         /* Initialize to TRUE. If any of Rx queues doesn't meet the
191          * vector Rx/Tx preconditions, it will be reset.
192          */
193         ad->rx_vec_allowed = true;
194         ad->tx_vec_allowed = true;
195
196         if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
197                 dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
198
199         /* Vlan stripping setting */
200         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
201                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
202                         iavf_enable_vlan_strip(ad);
203                 else
204                         iavf_disable_vlan_strip(ad);
205         }
206
207         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
208                 if (iavf_init_rss(ad) != 0) {
209                         PMD_DRV_LOG(ERR, "configure rss failed");
210                         return -1;
211                 }
212         }
213         return 0;
214 }
215
216 static int
217 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
218 {
219         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
220         struct rte_eth_dev_data *dev_data = dev->data;
221         uint16_t buf_size, max_pkt_len, len;
222
223         buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
224
225         /* Calculate the maximum packet length allowed */
226         len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
227         max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
228
229         /* Check if the jumbo frame and maximum packet length are set
230          * correctly.
231          */
232         if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
233                 if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
234                     max_pkt_len > IAVF_FRAME_SIZE_MAX) {
235                         PMD_DRV_LOG(ERR, "maximum packet length must be "
236                                     "larger than %u and smaller than %u, "
237                                     "as jumbo frame is enabled",
238                                     (uint32_t)RTE_ETHER_MAX_LEN,
239                                     (uint32_t)IAVF_FRAME_SIZE_MAX);
240                         return -EINVAL;
241                 }
242         } else {
243                 if (max_pkt_len < RTE_ETHER_MIN_LEN ||
244                     max_pkt_len > RTE_ETHER_MAX_LEN) {
245                         PMD_DRV_LOG(ERR, "maximum packet length must be "
246                                     "larger than %u and smaller than %u, "
247                                     "as jumbo frame is disabled",
248                                     (uint32_t)RTE_ETHER_MIN_LEN,
249                                     (uint32_t)RTE_ETHER_MAX_LEN);
250                         return -EINVAL;
251                 }
252         }
253
254         rxq->max_pkt_len = max_pkt_len;
255         if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
256             (rxq->max_pkt_len + 2 * IAVF_VLAN_TAG_SIZE) > buf_size) {
257                 dev_data->scattered_rx = 1;
258         }
259         IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
260         IAVF_WRITE_FLUSH(hw);
261
262         return 0;
263 }
264
265 static int
266 iavf_init_queues(struct rte_eth_dev *dev)
267 {
268         struct iavf_rx_queue **rxq =
269                 (struct iavf_rx_queue **)dev->data->rx_queues;
270         int i, ret = IAVF_SUCCESS;
271
272         for (i = 0; i < dev->data->nb_rx_queues; i++) {
273                 if (!rxq[i] || !rxq[i]->q_set)
274                         continue;
275                 ret = iavf_init_rxq(dev, rxq[i]);
276                 if (ret != IAVF_SUCCESS)
277                         break;
278         }
279         /* set rx/tx function to vector/scatter/single-segment
280          * according to parameters
281          */
282         iavf_set_rx_function(dev);
283         iavf_set_tx_function(dev);
284
285         return ret;
286 }
287
288 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
289                                      struct rte_intr_handle *intr_handle)
290 {
291         struct iavf_adapter *adapter =
292                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
293         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
294         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
295         uint16_t interval, i;
296         int vec;
297
298         if (rte_intr_cap_multiple(intr_handle) &&
299             dev->data->dev_conf.intr_conf.rxq) {
300                 if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
301                         return -1;
302         }
303
304         if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
305                 intr_handle->intr_vec =
306                         rte_zmalloc("intr_vec",
307                                     dev->data->nb_rx_queues * sizeof(int), 0);
308                 if (!intr_handle->intr_vec) {
309                         PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
310                                     dev->data->nb_rx_queues);
311                         return -1;
312                 }
313         }
314
315         if (!dev->data->dev_conf.intr_conf.rxq ||
316             !rte_intr_dp_is_en(intr_handle)) {
317                 /* Rx interrupt disabled, Map interrupt only for writeback */
318                 vf->nb_msix = 1;
319                 if (vf->vf_res->vf_cap_flags &
320                     VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
321                         /* If WB_ON_ITR supports, enable it */
322                         vf->msix_base = IAVF_RX_VEC_START;
323                         IAVF_WRITE_REG(hw,
324                                        IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
325                                        IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
326                                        IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
327                 } else {
328                         /* If no WB_ON_ITR offload flags, need to set
329                          * interrupt for descriptor write back.
330                          */
331                         vf->msix_base = IAVF_MISC_VEC_ID;
332
333                         /* set ITR to max */
334                         interval = iavf_calc_itr_interval(
335                                         IAVF_QUEUE_ITR_INTERVAL_MAX);
336                         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
337                                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
338                                        (IAVF_ITR_INDEX_DEFAULT <<
339                                         IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
340                                        (interval <<
341                                         IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
342                 }
343                 IAVF_WRITE_FLUSH(hw);
344                 /* map all queues to the same interrupt */
345                 for (i = 0; i < dev->data->nb_rx_queues; i++)
346                         vf->rxq_map[vf->msix_base] |= 1 << i;
347         } else {
348                 if (!rte_intr_allow_others(intr_handle)) {
349                         vf->nb_msix = 1;
350                         vf->msix_base = IAVF_MISC_VEC_ID;
351                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
352                                 vf->rxq_map[vf->msix_base] |= 1 << i;
353                                 intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
354                         }
355                         PMD_DRV_LOG(DEBUG,
356                                     "vector %u are mapping to all Rx queues",
357                                     vf->msix_base);
358                 } else {
359                         /* If Rx interrupt is reuquired, and we can use
360                          * multi interrupts, then the vec is from 1
361                          */
362                         vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
363                                               intr_handle->nb_efd);
364                         vf->msix_base = IAVF_RX_VEC_START;
365                         vec = IAVF_RX_VEC_START;
366                         for (i = 0; i < dev->data->nb_rx_queues; i++) {
367                                 vf->rxq_map[vec] |= 1 << i;
368                                 intr_handle->intr_vec[i] = vec++;
369                                 if (vec >= vf->nb_msix)
370                                         vec = IAVF_RX_VEC_START;
371                         }
372                         PMD_DRV_LOG(DEBUG,
373                                     "%u vectors are mapping to %u Rx queues",
374                                     vf->nb_msix, dev->data->nb_rx_queues);
375                 }
376         }
377
378         if (iavf_config_irq_map(adapter)) {
379                 PMD_DRV_LOG(ERR, "config interrupt mapping failed");
380                 return -1;
381         }
382         return 0;
383 }
384
385 static int
386 iavf_start_queues(struct rte_eth_dev *dev)
387 {
388         struct iavf_rx_queue *rxq;
389         struct iavf_tx_queue *txq;
390         int i;
391
392         for (i = 0; i < dev->data->nb_tx_queues; i++) {
393                 txq = dev->data->tx_queues[i];
394                 if (txq->tx_deferred_start)
395                         continue;
396                 if (iavf_dev_tx_queue_start(dev, i) != 0) {
397                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
398                         return -1;
399                 }
400         }
401
402         for (i = 0; i < dev->data->nb_rx_queues; i++) {
403                 rxq = dev->data->rx_queues[i];
404                 if (rxq->rx_deferred_start)
405                         continue;
406                 if (iavf_dev_rx_queue_start(dev, i) != 0) {
407                         PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
408                         return -1;
409                 }
410         }
411
412         return 0;
413 }
414
415 static int
416 iavf_dev_start(struct rte_eth_dev *dev)
417 {
418         struct iavf_adapter *adapter =
419                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
420         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
421         struct rte_intr_handle *intr_handle = dev->intr_handle;
422
423         PMD_INIT_FUNC_TRACE();
424
425         adapter->stopped = 0;
426
427         vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
428         vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
429                                       dev->data->nb_tx_queues);
430
431         if (iavf_init_queues(dev) != 0) {
432                 PMD_DRV_LOG(ERR, "failed to do Queue init");
433                 return -1;
434         }
435
436         if (iavf_configure_queues(adapter) != 0) {
437                 PMD_DRV_LOG(ERR, "configure queues failed");
438                 goto err_queue;
439         }
440
441         if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
442                 PMD_DRV_LOG(ERR, "configure irq failed");
443                 goto err_queue;
444         }
445         /* re-enable intr again, because efd assign may change */
446         if (dev->data->dev_conf.intr_conf.rxq != 0) {
447                 rte_intr_disable(intr_handle);
448                 rte_intr_enable(intr_handle);
449         }
450
451         /* Set all mac addrs */
452         iavf_add_del_all_mac_addr(adapter, true);
453
454         if (iavf_start_queues(dev) != 0) {
455                 PMD_DRV_LOG(ERR, "enable queues failed");
456                 goto err_mac;
457         }
458
459         return 0;
460
461 err_mac:
462         iavf_add_del_all_mac_addr(adapter, false);
463 err_queue:
464         return -1;
465 }
466
467 static void
468 iavf_dev_stop(struct rte_eth_dev *dev)
469 {
470         struct iavf_adapter *adapter =
471                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
472         struct rte_intr_handle *intr_handle = dev->intr_handle;
473
474         PMD_INIT_FUNC_TRACE();
475
476         if (adapter->stopped == 1)
477                 return;
478
479         iavf_stop_queues(dev);
480
481         /* Disable the interrupt for Rx */
482         rte_intr_efd_disable(intr_handle);
483         /* Rx interrupt vector mapping free */
484         if (intr_handle->intr_vec) {
485                 rte_free(intr_handle->intr_vec);
486                 intr_handle->intr_vec = NULL;
487         }
488
489         /* remove all mac addrs */
490         iavf_add_del_all_mac_addr(adapter, false);
491         adapter->stopped = 1;
492 }
493
494 static int
495 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
496 {
497         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
498
499         dev_info->max_rx_queues = vf->vsi_res->num_queue_pairs;
500         dev_info->max_tx_queues = vf->vsi_res->num_queue_pairs;
501         dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
502         dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
503         dev_info->hash_key_size = vf->vf_res->rss_key_size;
504         dev_info->reta_size = vf->vf_res->rss_lut_size;
505         dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
506         dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
507         dev_info->rx_offload_capa =
508                 DEV_RX_OFFLOAD_VLAN_STRIP |
509                 DEV_RX_OFFLOAD_QINQ_STRIP |
510                 DEV_RX_OFFLOAD_IPV4_CKSUM |
511                 DEV_RX_OFFLOAD_UDP_CKSUM |
512                 DEV_RX_OFFLOAD_TCP_CKSUM |
513                 DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
514                 DEV_RX_OFFLOAD_SCATTER |
515                 DEV_RX_OFFLOAD_JUMBO_FRAME |
516                 DEV_RX_OFFLOAD_VLAN_FILTER |
517                 DEV_RX_OFFLOAD_RSS_HASH;
518         dev_info->tx_offload_capa =
519                 DEV_TX_OFFLOAD_VLAN_INSERT |
520                 DEV_TX_OFFLOAD_QINQ_INSERT |
521                 DEV_TX_OFFLOAD_IPV4_CKSUM |
522                 DEV_TX_OFFLOAD_UDP_CKSUM |
523                 DEV_TX_OFFLOAD_TCP_CKSUM |
524                 DEV_TX_OFFLOAD_SCTP_CKSUM |
525                 DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
526                 DEV_TX_OFFLOAD_TCP_TSO |
527                 DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
528                 DEV_TX_OFFLOAD_GRE_TNL_TSO |
529                 DEV_TX_OFFLOAD_IPIP_TNL_TSO |
530                 DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
531                 DEV_TX_OFFLOAD_MULTI_SEGS;
532
533         dev_info->default_rxconf = (struct rte_eth_rxconf) {
534                 .rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
535                 .rx_drop_en = 0,
536                 .offloads = 0,
537         };
538
539         dev_info->default_txconf = (struct rte_eth_txconf) {
540                 .tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
541                 .tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
542                 .offloads = 0,
543         };
544
545         dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
546                 .nb_max = IAVF_MAX_RING_DESC,
547                 .nb_min = IAVF_MIN_RING_DESC,
548                 .nb_align = IAVF_ALIGN_RING_DESC,
549         };
550
551         dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
552                 .nb_max = IAVF_MAX_RING_DESC,
553                 .nb_min = IAVF_MIN_RING_DESC,
554                 .nb_align = IAVF_ALIGN_RING_DESC,
555         };
556
557         return 0;
558 }
559
560 static const uint32_t *
561 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
562 {
563         static const uint32_t ptypes[] = {
564                 RTE_PTYPE_L2_ETHER,
565                 RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
566                 RTE_PTYPE_L4_FRAG,
567                 RTE_PTYPE_L4_ICMP,
568                 RTE_PTYPE_L4_NONFRAG,
569                 RTE_PTYPE_L4_SCTP,
570                 RTE_PTYPE_L4_TCP,
571                 RTE_PTYPE_L4_UDP,
572                 RTE_PTYPE_UNKNOWN
573         };
574         return ptypes;
575 }
576
577 int
578 iavf_dev_link_update(struct rte_eth_dev *dev,
579                     __rte_unused int wait_to_complete)
580 {
581         struct rte_eth_link new_link;
582         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
583
584         memset(&new_link, 0, sizeof(new_link));
585
586         /* Only read status info stored in VF, and the info is updated
587          *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
588          */
589         switch (vf->link_speed) {
590         case 10:
591                 new_link.link_speed = ETH_SPEED_NUM_10M;
592                 break;
593         case 100:
594                 new_link.link_speed = ETH_SPEED_NUM_100M;
595                 break;
596         case 1000:
597                 new_link.link_speed = ETH_SPEED_NUM_1G;
598                 break;
599         case 10000:
600                 new_link.link_speed = ETH_SPEED_NUM_10G;
601                 break;
602         case 20000:
603                 new_link.link_speed = ETH_SPEED_NUM_20G;
604                 break;
605         case 25000:
606                 new_link.link_speed = ETH_SPEED_NUM_25G;
607                 break;
608         case 40000:
609                 new_link.link_speed = ETH_SPEED_NUM_40G;
610                 break;
611         case 50000:
612                 new_link.link_speed = ETH_SPEED_NUM_50G;
613                 break;
614         case 100000:
615                 new_link.link_speed = ETH_SPEED_NUM_100G;
616                 break;
617         default:
618                 new_link.link_speed = ETH_SPEED_NUM_NONE;
619                 break;
620         }
621
622         new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
623         new_link.link_status = vf->link_up ? ETH_LINK_UP :
624                                              ETH_LINK_DOWN;
625         new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
626                                 ETH_LINK_SPEED_FIXED);
627
628         if (rte_atomic64_cmpset((uint64_t *)&dev->data->dev_link,
629                                 *(uint64_t *)&dev->data->dev_link,
630                                 *(uint64_t *)&new_link) == 0)
631                 return -1;
632
633         return 0;
634 }
635
636 static int
637 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
638 {
639         struct iavf_adapter *adapter =
640                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
641         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
642         int ret;
643
644         if (vf->promisc_unicast_enabled)
645                 return 0;
646
647         ret = iavf_config_promisc(adapter, true, vf->promisc_multicast_enabled);
648         if (!ret)
649                 vf->promisc_unicast_enabled = true;
650         else
651                 ret = -EAGAIN;
652
653         return ret;
654 }
655
656 static int
657 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
658 {
659         struct iavf_adapter *adapter =
660                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
661         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
662         int ret;
663
664         if (!vf->promisc_unicast_enabled)
665                 return 0;
666
667         ret = iavf_config_promisc(adapter, false,
668                                   vf->promisc_multicast_enabled);
669         if (!ret)
670                 vf->promisc_unicast_enabled = false;
671         else
672                 ret = -EAGAIN;
673
674         return ret;
675 }
676
677 static int
678 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
679 {
680         struct iavf_adapter *adapter =
681                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
682         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
683         int ret;
684
685         if (vf->promisc_multicast_enabled)
686                 return 0;
687
688         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, true);
689         if (!ret)
690                 vf->promisc_multicast_enabled = true;
691         else
692                 ret = -EAGAIN;
693
694         return ret;
695 }
696
697 static int
698 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
699 {
700         struct iavf_adapter *adapter =
701                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
702         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
703         int ret;
704
705         if (!vf->promisc_multicast_enabled)
706                 return 0;
707
708         ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, false);
709         if (!ret)
710                 vf->promisc_multicast_enabled = false;
711         else
712                 ret = -EAGAIN;
713
714         return ret;
715 }
716
717 static int
718 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
719                      __rte_unused uint32_t index,
720                      __rte_unused uint32_t pool)
721 {
722         struct iavf_adapter *adapter =
723                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
724         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
725         int err;
726
727         if (rte_is_zero_ether_addr(addr)) {
728                 PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
729                 return -EINVAL;
730         }
731
732         err = iavf_add_del_eth_addr(adapter, addr, true);
733         if (err) {
734                 PMD_DRV_LOG(ERR, "fail to add MAC address");
735                 return -EIO;
736         }
737
738         vf->mac_num++;
739
740         return 0;
741 }
742
743 static void
744 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
745 {
746         struct iavf_adapter *adapter =
747                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
748         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
749         struct rte_ether_addr *addr;
750         int err;
751
752         addr = &dev->data->mac_addrs[index];
753
754         err = iavf_add_del_eth_addr(adapter, addr, false);
755         if (err)
756                 PMD_DRV_LOG(ERR, "fail to delete MAC address");
757
758         vf->mac_num--;
759 }
760
761 static int
762 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
763 {
764         struct iavf_adapter *adapter =
765                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
766         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
767         int err;
768
769         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
770                 return -ENOTSUP;
771
772         err = iavf_add_del_vlan(adapter, vlan_id, on);
773         if (err)
774                 return -EIO;
775         return 0;
776 }
777
778 static int
779 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
780 {
781         struct iavf_adapter *adapter =
782                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
783         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
784         struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
785         int err;
786
787         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
788                 return -ENOTSUP;
789
790         /* Vlan stripping setting */
791         if (mask & ETH_VLAN_STRIP_MASK) {
792                 /* Enable or disable VLAN stripping */
793                 if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
794                         err = iavf_enable_vlan_strip(adapter);
795                 else
796                         err = iavf_disable_vlan_strip(adapter);
797
798                 if (err)
799                         return -EIO;
800         }
801         return 0;
802 }
803
804 static int
805 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
806                         struct rte_eth_rss_reta_entry64 *reta_conf,
807                         uint16_t reta_size)
808 {
809         struct iavf_adapter *adapter =
810                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
811         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
812         uint8_t *lut;
813         uint16_t i, idx, shift;
814         int ret;
815
816         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
817                 return -ENOTSUP;
818
819         if (reta_size != vf->vf_res->rss_lut_size) {
820                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
821                         "(%d) doesn't match the number of hardware can "
822                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
823                 return -EINVAL;
824         }
825
826         lut = rte_zmalloc("rss_lut", reta_size, 0);
827         if (!lut) {
828                 PMD_DRV_LOG(ERR, "No memory can be allocated");
829                 return -ENOMEM;
830         }
831         /* store the old lut table temporarily */
832         rte_memcpy(lut, vf->rss_lut, reta_size);
833
834         for (i = 0; i < reta_size; i++) {
835                 idx = i / RTE_RETA_GROUP_SIZE;
836                 shift = i % RTE_RETA_GROUP_SIZE;
837                 if (reta_conf[idx].mask & (1ULL << shift))
838                         lut[i] = reta_conf[idx].reta[shift];
839         }
840
841         rte_memcpy(vf->rss_lut, lut, reta_size);
842         /* send virtchnnl ops to configure rss*/
843         ret = iavf_configure_rss_lut(adapter);
844         if (ret) /* revert back */
845                 rte_memcpy(vf->rss_lut, lut, reta_size);
846         rte_free(lut);
847
848         return ret;
849 }
850
851 static int
852 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
853                        struct rte_eth_rss_reta_entry64 *reta_conf,
854                        uint16_t reta_size)
855 {
856         struct iavf_adapter *adapter =
857                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
858         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
859         uint16_t i, idx, shift;
860
861         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
862                 return -ENOTSUP;
863
864         if (reta_size != vf->vf_res->rss_lut_size) {
865                 PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
866                         "(%d) doesn't match the number of hardware can "
867                         "support (%d)", reta_size, vf->vf_res->rss_lut_size);
868                 return -EINVAL;
869         }
870
871         for (i = 0; i < reta_size; i++) {
872                 idx = i / RTE_RETA_GROUP_SIZE;
873                 shift = i % RTE_RETA_GROUP_SIZE;
874                 if (reta_conf[idx].mask & (1ULL << shift))
875                         reta_conf[idx].reta[shift] = vf->rss_lut[i];
876         }
877
878         return 0;
879 }
880
881 static int
882 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
883                         struct rte_eth_rss_conf *rss_conf)
884 {
885         struct iavf_adapter *adapter =
886                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
887         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
888
889         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
890                 return -ENOTSUP;
891
892         /* HENA setting, it is enabled by default, no change */
893         if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
894                 PMD_DRV_LOG(DEBUG, "No key to be configured");
895                 return 0;
896         } else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
897                 PMD_DRV_LOG(ERR, "The size of hash key configured "
898                         "(%d) doesn't match the size of hardware can "
899                         "support (%d)", rss_conf->rss_key_len,
900                         vf->vf_res->rss_key_size);
901                 return -EINVAL;
902         }
903
904         rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
905
906         return iavf_configure_rss_key(adapter);
907 }
908
909 static int
910 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
911                           struct rte_eth_rss_conf *rss_conf)
912 {
913         struct iavf_adapter *adapter =
914                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
915         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
916
917         if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
918                 return -ENOTSUP;
919
920          /* Just set it to default value now. */
921         rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
922
923         if (!rss_conf->rss_key)
924                 return 0;
925
926         rss_conf->rss_key_len = vf->vf_res->rss_key_size;
927         rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
928
929         return 0;
930 }
931
932 static int
933 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
934 {
935         uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
936         int ret = 0;
937
938         if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
939                 return -EINVAL;
940
941         /* mtu setting is forbidden if port is start */
942         if (dev->data->dev_started) {
943                 PMD_DRV_LOG(ERR, "port must be stopped before configuration");
944                 return -EBUSY;
945         }
946
947         if (frame_size > RTE_ETHER_MAX_LEN)
948                 dev->data->dev_conf.rxmode.offloads |=
949                                 DEV_RX_OFFLOAD_JUMBO_FRAME;
950         else
951                 dev->data->dev_conf.rxmode.offloads &=
952                                 ~DEV_RX_OFFLOAD_JUMBO_FRAME;
953
954         dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
955
956         return ret;
957 }
958
959 static int
960 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
961                              struct rte_ether_addr *mac_addr)
962 {
963         struct iavf_adapter *adapter =
964                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
965         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
966         struct rte_ether_addr *perm_addr, *old_addr;
967         int ret;
968
969         old_addr = (struct rte_ether_addr *)hw->mac.addr;
970         perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
971
972         if (rte_is_same_ether_addr(mac_addr, old_addr))
973                 return 0;
974
975         /* If the MAC address is configured by host, skip the setting */
976         if (rte_is_valid_assigned_ether_addr(perm_addr))
977                 return -EPERM;
978
979         ret = iavf_add_del_eth_addr(adapter, old_addr, false);
980         if (ret)
981                 PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
982                             " %02X:%02X:%02X:%02X:%02X:%02X",
983                             old_addr->addr_bytes[0],
984                             old_addr->addr_bytes[1],
985                             old_addr->addr_bytes[2],
986                             old_addr->addr_bytes[3],
987                             old_addr->addr_bytes[4],
988                             old_addr->addr_bytes[5]);
989
990         ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
991         if (ret)
992                 PMD_DRV_LOG(ERR, "Fail to add new MAC:"
993                             " %02X:%02X:%02X:%02X:%02X:%02X",
994                             mac_addr->addr_bytes[0],
995                             mac_addr->addr_bytes[1],
996                             mac_addr->addr_bytes[2],
997                             mac_addr->addr_bytes[3],
998                             mac_addr->addr_bytes[4],
999                             mac_addr->addr_bytes[5]);
1000
1001         if (ret)
1002                 return -EIO;
1003
1004         rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1005         return 0;
1006 }
1007
1008 static void
1009 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1010 {
1011         if (*stat >= *offset)
1012                 *stat = *stat - *offset;
1013         else
1014                 *stat = (uint64_t)((*stat +
1015                         ((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1016
1017         *stat &= IAVF_48_BIT_MASK;
1018 }
1019
1020 static void
1021 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1022 {
1023         if (*stat >= *offset)
1024                 *stat = (uint64_t)(*stat - *offset);
1025         else
1026                 *stat = (uint64_t)((*stat +
1027                         ((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1028 }
1029
1030 static void
1031 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1032 {
1033         struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1034
1035         iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1036         iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1037         iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1038         iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1039         iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1040         iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1041         iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1042         iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1043         iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1044         iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1045         iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1046 }
1047
1048 static int
1049 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1050 {
1051         struct iavf_adapter *adapter =
1052                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1053         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1054         struct iavf_vsi *vsi = &vf->vsi;
1055         struct virtchnl_eth_stats *pstats = NULL;
1056         int ret;
1057
1058         ret = iavf_query_stats(adapter, &pstats);
1059         if (ret == 0) {
1060                 iavf_update_stats(vsi, pstats);
1061                 stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1062                                 pstats->rx_broadcast - pstats->rx_discards;
1063                 stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1064                                                 pstats->tx_unicast;
1065                 stats->imissed = pstats->rx_discards;
1066                 stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1067                 stats->ibytes = pstats->rx_bytes;
1068                 stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1069                 stats->obytes = pstats->tx_bytes;
1070         } else {
1071                 PMD_DRV_LOG(ERR, "Get statistics failed");
1072         }
1073         return ret;
1074 }
1075
1076 static int
1077 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1078 {
1079         int ret;
1080         struct iavf_adapter *adapter =
1081                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1082         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1083         struct iavf_vsi *vsi = &vf->vsi;
1084         struct virtchnl_eth_stats *pstats = NULL;
1085
1086         /* read stat values to clear hardware registers */
1087         ret = iavf_query_stats(adapter, &pstats);
1088         if (ret != 0)
1089                 return ret;
1090
1091         /* set stats offset base on current values */
1092         vsi->eth_stats_offset = *pstats;
1093
1094         return 0;
1095 }
1096
1097 static int
1098 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1099 {
1100         struct iavf_adapter *adapter =
1101                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1102         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1103         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1104         uint16_t msix_intr;
1105
1106         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1107         if (msix_intr == IAVF_MISC_VEC_ID) {
1108                 PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1109                 IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1110                                IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1111                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1112                                IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1113         } else {
1114                 IAVF_WRITE_REG(hw,
1115                                IAVF_VFINT_DYN_CTLN1
1116                                 (msix_intr - IAVF_RX_VEC_START),
1117                                IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1118                                IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1119                                IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1120         }
1121
1122         IAVF_WRITE_FLUSH(hw);
1123
1124         rte_intr_ack(&pci_dev->intr_handle);
1125
1126         return 0;
1127 }
1128
1129 static int
1130 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1131 {
1132         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1133         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1134         uint16_t msix_intr;
1135
1136         msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1137         if (msix_intr == IAVF_MISC_VEC_ID) {
1138                 PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1139                 return -EIO;
1140         }
1141
1142         IAVF_WRITE_REG(hw,
1143                       IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1144                       0);
1145
1146         IAVF_WRITE_FLUSH(hw);
1147         return 0;
1148 }
1149
1150 static int
1151 iavf_check_vf_reset_done(struct iavf_hw *hw)
1152 {
1153         int i, reset;
1154
1155         for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1156                 reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1157                         IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1158                 reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1159                 if (reset == VIRTCHNL_VFR_VFACTIVE ||
1160                     reset == VIRTCHNL_VFR_COMPLETED)
1161                         break;
1162                 rte_delay_ms(20);
1163         }
1164
1165         if (i >= IAVF_RESET_WAIT_CNT)
1166                 return -1;
1167
1168         return 0;
1169 }
1170
1171 static int
1172 iavf_init_vf(struct rte_eth_dev *dev)
1173 {
1174         int err, bufsz;
1175         struct iavf_adapter *adapter =
1176                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1177         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1178         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1179
1180         err = iavf_set_mac_type(hw);
1181         if (err) {
1182                 PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1183                 goto err;
1184         }
1185
1186         err = iavf_check_vf_reset_done(hw);
1187         if (err) {
1188                 PMD_INIT_LOG(ERR, "VF is still resetting");
1189                 goto err;
1190         }
1191
1192         iavf_init_adminq_parameter(hw);
1193         err = iavf_init_adminq(hw);
1194         if (err) {
1195                 PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1196                 goto err;
1197         }
1198
1199         vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1200         if (!vf->aq_resp) {
1201                 PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1202                 goto err_aq;
1203         }
1204         if (iavf_check_api_version(adapter) != 0) {
1205                 PMD_INIT_LOG(ERR, "check_api version failed");
1206                 goto err_api;
1207         }
1208
1209         bufsz = sizeof(struct virtchnl_vf_resource) +
1210                 (IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1211         vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1212         if (!vf->vf_res) {
1213                 PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1214                 goto err_api;
1215         }
1216         if (iavf_get_vf_resource(adapter) != 0) {
1217                 PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1218                 goto err_alloc;
1219         }
1220         /* Allocate memort for RSS info */
1221         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1222                 vf->rss_key = rte_zmalloc("rss_key",
1223                                           vf->vf_res->rss_key_size, 0);
1224                 if (!vf->rss_key) {
1225                         PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1226                         goto err_rss;
1227                 }
1228                 vf->rss_lut = rte_zmalloc("rss_lut",
1229                                           vf->vf_res->rss_lut_size, 0);
1230                 if (!vf->rss_lut) {
1231                         PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1232                         goto err_rss;
1233                 }
1234         }
1235
1236         if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1237                 if (iavf_get_supported_rxdid(adapter) != 0) {
1238                         PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1239                         goto err_rss;
1240                 }
1241         }
1242
1243         return 0;
1244 err_rss:
1245         rte_free(vf->rss_key);
1246         rte_free(vf->rss_lut);
1247 err_alloc:
1248         rte_free(vf->vf_res);
1249         vf->vsi_res = NULL;
1250 err_api:
1251         rte_free(vf->aq_resp);
1252 err_aq:
1253         iavf_shutdown_adminq(hw);
1254 err:
1255         return -1;
1256 }
1257
1258 /* Enable default admin queue interrupt setting */
1259 static inline void
1260 iavf_enable_irq0(struct iavf_hw *hw)
1261 {
1262         /* Enable admin queue interrupt trigger */
1263         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1264                        IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1265
1266         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1267                        IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1268                        IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1269                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1270
1271         IAVF_WRITE_FLUSH(hw);
1272 }
1273
1274 static inline void
1275 iavf_disable_irq0(struct iavf_hw *hw)
1276 {
1277         /* Disable all interrupt types */
1278         IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1279         IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1280                        IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1281         IAVF_WRITE_FLUSH(hw);
1282 }
1283
1284 static void
1285 iavf_dev_interrupt_handler(void *param)
1286 {
1287         struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1288         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1289
1290         iavf_disable_irq0(hw);
1291
1292         iavf_handle_virtchnl_msg(dev);
1293
1294         iavf_enable_irq0(hw);
1295 }
1296
1297 static int
1298 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1299                      enum rte_filter_type filter_type,
1300                      enum rte_filter_op filter_op,
1301                      void *arg)
1302 {
1303         int ret = 0;
1304
1305         if (!dev)
1306                 return -EINVAL;
1307
1308         switch (filter_type) {
1309         case RTE_ETH_FILTER_GENERIC:
1310                 if (filter_op != RTE_ETH_FILTER_GET)
1311                         return -EINVAL;
1312                 *(const void **)arg = &iavf_flow_ops;
1313                 break;
1314         default:
1315                 PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1316                             filter_type);
1317                 ret = -EINVAL;
1318                 break;
1319         }
1320
1321         return ret;
1322 }
1323
1324
1325 static int
1326 iavf_dev_init(struct rte_eth_dev *eth_dev)
1327 {
1328         struct iavf_adapter *adapter =
1329                 IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1330         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1331         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1332         int ret = 0;
1333
1334         PMD_INIT_FUNC_TRACE();
1335
1336         /* assign ops func pointer */
1337         eth_dev->dev_ops = &iavf_eth_dev_ops;
1338         eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1339         eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1340         eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1341
1342         /* For secondary processes, we don't initialise any further as primary
1343          * has already done this work. Only check if we need a different RX
1344          * and TX function.
1345          */
1346         if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1347                 iavf_set_rx_function(eth_dev);
1348                 iavf_set_tx_function(eth_dev);
1349                 return 0;
1350         }
1351         rte_eth_copy_pci_info(eth_dev, pci_dev);
1352
1353         hw->vendor_id = pci_dev->id.vendor_id;
1354         hw->device_id = pci_dev->id.device_id;
1355         hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1356         hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1357         hw->bus.bus_id = pci_dev->addr.bus;
1358         hw->bus.device = pci_dev->addr.devid;
1359         hw->bus.func = pci_dev->addr.function;
1360         hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1361         hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1362         adapter->eth_dev = eth_dev;
1363         adapter->stopped = 1;
1364
1365         if (iavf_init_vf(eth_dev) != 0) {
1366                 PMD_INIT_LOG(ERR, "Init vf failed");
1367                 return -1;
1368         }
1369
1370         /* set default ptype table */
1371         adapter->ptype_tbl = iavf_get_default_ptype_table();
1372
1373         /* copy mac addr */
1374         eth_dev->data->mac_addrs = rte_zmalloc(
1375                 "iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1376         if (!eth_dev->data->mac_addrs) {
1377                 PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1378                              " store MAC addresses",
1379                              RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1380                 return -ENOMEM;
1381         }
1382         /* If the MAC address is not configured by host,
1383          * generate a random one.
1384          */
1385         if (!rte_is_valid_assigned_ether_addr(
1386                         (struct rte_ether_addr *)hw->mac.addr))
1387                 rte_eth_random_addr(hw->mac.addr);
1388         rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1389                         &eth_dev->data->mac_addrs[0]);
1390
1391         /* register callback func to eal lib */
1392         rte_intr_callback_register(&pci_dev->intr_handle,
1393                                    iavf_dev_interrupt_handler,
1394                                    (void *)eth_dev);
1395
1396         /* enable uio intr after callback register */
1397         rte_intr_enable(&pci_dev->intr_handle);
1398
1399         /* configure and enable device interrupt */
1400         iavf_enable_irq0(hw);
1401
1402         ret = iavf_flow_init(adapter);
1403         if (ret) {
1404                 PMD_INIT_LOG(ERR, "Failed to initialize flow");
1405                 return ret;
1406         }
1407
1408         return 0;
1409 }
1410
1411 static void
1412 iavf_dev_close(struct rte_eth_dev *dev)
1413 {
1414         struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1415         struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1416         struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1417         struct iavf_adapter *adapter =
1418                 IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1419
1420         iavf_dev_stop(dev);
1421         iavf_flow_flush(dev, NULL);
1422         iavf_flow_uninit(adapter);
1423         iavf_shutdown_adminq(hw);
1424         /* disable uio intr before callback unregister */
1425         rte_intr_disable(intr_handle);
1426
1427         /* unregister callback func from eal lib */
1428         rte_intr_callback_unregister(intr_handle,
1429                                      iavf_dev_interrupt_handler, dev);
1430         iavf_disable_irq0(hw);
1431 }
1432
1433 static int
1434 iavf_dev_uninit(struct rte_eth_dev *dev)
1435 {
1436         struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1437
1438         if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1439                 return -EPERM;
1440
1441         dev->dev_ops = NULL;
1442         dev->rx_pkt_burst = NULL;
1443         dev->tx_pkt_burst = NULL;
1444         iavf_dev_close(dev);
1445
1446         rte_free(vf->vf_res);
1447         vf->vsi_res = NULL;
1448         vf->vf_res = NULL;
1449
1450         rte_free(vf->aq_resp);
1451         vf->aq_resp = NULL;
1452
1453         if (vf->rss_lut) {
1454                 rte_free(vf->rss_lut);
1455                 vf->rss_lut = NULL;
1456         }
1457         if (vf->rss_key) {
1458                 rte_free(vf->rss_key);
1459                 vf->rss_key = NULL;
1460         }
1461
1462         return 0;
1463 }
1464
1465 /*
1466  * Reset VF device only to re-initialize resources in PMD layer
1467  */
1468 static int
1469 iavf_dev_reset(struct rte_eth_dev *dev)
1470 {
1471         int ret;
1472
1473         ret = iavf_dev_uninit(dev);
1474         if (ret)
1475                 return ret;
1476
1477         return iavf_dev_init(dev);
1478 }
1479
1480 static int
1481 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1482                            const char *value, __rte_unused void *opaque)
1483 {
1484         if (strcmp(value, "dcf"))
1485                 return -1;
1486
1487         return 0;
1488 }
1489
1490 static int
1491 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1492 {
1493         struct rte_kvargs *kvlist;
1494         const char *key = "cap";
1495         int ret = 0;
1496
1497         if (devargs == NULL)
1498                 return 0;
1499
1500         kvlist = rte_kvargs_parse(devargs->args, NULL);
1501         if (kvlist == NULL)
1502                 return 0;
1503
1504         if (!rte_kvargs_count(kvlist, key))
1505                 goto exit;
1506
1507         /* dcf capability selected when there's a key-value pair: cap=dcf */
1508         if (rte_kvargs_process(kvlist, key,
1509                                iavf_dcf_cap_check_handler, NULL) < 0)
1510                 goto exit;
1511
1512         ret = 1;
1513
1514 exit:
1515         rte_kvargs_free(kvlist);
1516         return ret;
1517 }
1518
1519 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1520                              struct rte_pci_device *pci_dev)
1521 {
1522         if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1523                 return 1;
1524
1525         return rte_eth_dev_pci_generic_probe(pci_dev,
1526                 sizeof(struct iavf_adapter), iavf_dev_init);
1527 }
1528
1529 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1530 {
1531         return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1532 }
1533
1534 /* Adaptive virtual function driver struct */
1535 static struct rte_pci_driver rte_iavf_pmd = {
1536         .id_table = pci_id_iavf_map,
1537         .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1538         .probe = eth_iavf_pci_probe,
1539         .remove = eth_iavf_pci_remove,
1540 };
1541
1542 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1543 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1544 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1545 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1546 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1547 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1548 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1549 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1550 #endif
1551 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1552 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1553 #endif
1554 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1555 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1556 #endif