xref: /dpdk/drivers/net/iavf/iavf_ethdev.c (revision 0d09cbc7)
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 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
77 			struct rte_ether_addr *mc_addrs,
78 			uint32_t mc_addrs_num);
79 
80 static const struct rte_pci_id pci_id_iavf_map[] = {
81 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
82 	{ .vendor_id = 0, /* sentinel */ },
83 };
84 
85 static const struct eth_dev_ops iavf_eth_dev_ops = {
86 	.dev_configure              = iavf_dev_configure,
87 	.dev_start                  = iavf_dev_start,
88 	.dev_stop                   = iavf_dev_stop,
89 	.dev_close                  = iavf_dev_close,
90 	.dev_reset                  = iavf_dev_reset,
91 	.dev_infos_get              = iavf_dev_info_get,
92 	.dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
93 	.link_update                = iavf_dev_link_update,
94 	.stats_get                  = iavf_dev_stats_get,
95 	.stats_reset                = iavf_dev_stats_reset,
96 	.promiscuous_enable         = iavf_dev_promiscuous_enable,
97 	.promiscuous_disable        = iavf_dev_promiscuous_disable,
98 	.allmulticast_enable        = iavf_dev_allmulticast_enable,
99 	.allmulticast_disable       = iavf_dev_allmulticast_disable,
100 	.mac_addr_add               = iavf_dev_add_mac_addr,
101 	.mac_addr_remove            = iavf_dev_del_mac_addr,
102 	.set_mc_addr_list			= iavf_set_mc_addr_list,
103 	.vlan_filter_set            = iavf_dev_vlan_filter_set,
104 	.vlan_offload_set           = iavf_dev_vlan_offload_set,
105 	.rx_queue_start             = iavf_dev_rx_queue_start,
106 	.rx_queue_stop              = iavf_dev_rx_queue_stop,
107 	.tx_queue_start             = iavf_dev_tx_queue_start,
108 	.tx_queue_stop              = iavf_dev_tx_queue_stop,
109 	.rx_queue_setup             = iavf_dev_rx_queue_setup,
110 	.rx_queue_release           = iavf_dev_rx_queue_release,
111 	.tx_queue_setup             = iavf_dev_tx_queue_setup,
112 	.tx_queue_release           = iavf_dev_tx_queue_release,
113 	.mac_addr_set               = iavf_dev_set_default_mac_addr,
114 	.reta_update                = iavf_dev_rss_reta_update,
115 	.reta_query                 = iavf_dev_rss_reta_query,
116 	.rss_hash_update            = iavf_dev_rss_hash_update,
117 	.rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
118 	.rxq_info_get               = iavf_dev_rxq_info_get,
119 	.txq_info_get               = iavf_dev_txq_info_get,
120 	.mtu_set                    = iavf_dev_mtu_set,
121 	.rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
122 	.rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
123 	.filter_ctrl                = iavf_dev_filter_ctrl,
124 };
125 
126 static int
127 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
128 			struct rte_ether_addr *mc_addrs,
129 			uint32_t mc_addrs_num)
130 {
131 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
132 	struct iavf_adapter *adapter =
133 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
134 	int err;
135 
136 	/* flush previous addresses */
137 	err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
138 					false);
139 	if (err)
140 		return err;
141 
142 	vf->mc_addrs_num = 0;
143 
144 	/* add new ones */
145 	err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
146 	if (err)
147 		return err;
148 
149 	vf->mc_addrs_num = mc_addrs_num;
150 	memcpy(vf->mc_addrs, mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
151 
152 	return 0;
153 }
154 
155 static int
156 iavf_init_rss(struct iavf_adapter *adapter)
157 {
158 	struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
159 	struct rte_eth_rss_conf *rss_conf;
160 	uint16_t i, j, nb_q;
161 	int ret;
162 
163 	rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
164 	nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
165 		       IAVF_MAX_NUM_QUEUES);
166 
167 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
168 		PMD_DRV_LOG(DEBUG, "RSS is not supported");
169 		return -ENOTSUP;
170 	}
171 	if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
172 		PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
173 		/* set all lut items to default queue */
174 		for (i = 0; i < vf->vf_res->rss_lut_size; i++)
175 			vf->rss_lut[i] = 0;
176 		ret = iavf_configure_rss_lut(adapter);
177 		return ret;
178 	}
179 
180 	/* In IAVF, RSS enablement is set by PF driver. It is not supported
181 	 * to set based on rss_conf->rss_hf.
182 	 */
183 
184 	/* configure RSS key */
185 	if (!rss_conf->rss_key) {
186 		/* Calculate the default hash key */
187 		for (i = 0; i <= vf->vf_res->rss_key_size; i++)
188 			vf->rss_key[i] = (uint8_t)rte_rand();
189 	} else
190 		rte_memcpy(vf->rss_key, rss_conf->rss_key,
191 			   RTE_MIN(rss_conf->rss_key_len,
192 				   vf->vf_res->rss_key_size));
193 
194 	/* init RSS LUT table */
195 	for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
196 		if (j >= nb_q)
197 			j = 0;
198 		vf->rss_lut[i] = j;
199 	}
200 	/* send virtchnnl ops to configure rss*/
201 	ret = iavf_configure_rss_lut(adapter);
202 	if (ret)
203 		return ret;
204 	ret = iavf_configure_rss_key(adapter);
205 	if (ret)
206 		return ret;
207 
208 	return 0;
209 }
210 
211 static int
212 iavf_dev_configure(struct rte_eth_dev *dev)
213 {
214 	struct iavf_adapter *ad =
215 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
216 	struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
217 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
218 
219 	ad->rx_bulk_alloc_allowed = true;
220 	/* Initialize to TRUE. If any of Rx queues doesn't meet the
221 	 * vector Rx/Tx preconditions, it will be reset.
222 	 */
223 	ad->rx_vec_allowed = true;
224 	ad->tx_vec_allowed = true;
225 
226 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
227 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
228 
229 	/* Vlan stripping setting */
230 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN) {
231 		if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
232 			iavf_enable_vlan_strip(ad);
233 		else
234 			iavf_disable_vlan_strip(ad);
235 	}
236 
237 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
238 		if (iavf_init_rss(ad) != 0) {
239 			PMD_DRV_LOG(ERR, "configure rss failed");
240 			return -1;
241 		}
242 	}
243 	return 0;
244 }
245 
246 static int
247 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
248 {
249 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
250 	struct rte_eth_dev_data *dev_data = dev->data;
251 	uint16_t buf_size, max_pkt_len, len;
252 
253 	buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
254 
255 	/* Calculate the maximum packet length allowed */
256 	len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
257 	max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
258 
259 	/* Check if the jumbo frame and maximum packet length are set
260 	 * correctly.
261 	 */
262 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
263 		if (max_pkt_len <= RTE_ETHER_MAX_LEN ||
264 		    max_pkt_len > IAVF_FRAME_SIZE_MAX) {
265 			PMD_DRV_LOG(ERR, "maximum packet length must be "
266 				    "larger than %u and smaller than %u, "
267 				    "as jumbo frame is enabled",
268 				    (uint32_t)RTE_ETHER_MAX_LEN,
269 				    (uint32_t)IAVF_FRAME_SIZE_MAX);
270 			return -EINVAL;
271 		}
272 	} else {
273 		if (max_pkt_len < RTE_ETHER_MIN_LEN ||
274 		    max_pkt_len > RTE_ETHER_MAX_LEN) {
275 			PMD_DRV_LOG(ERR, "maximum packet length must be "
276 				    "larger than %u and smaller than %u, "
277 				    "as jumbo frame is disabled",
278 				    (uint32_t)RTE_ETHER_MIN_LEN,
279 				    (uint32_t)RTE_ETHER_MAX_LEN);
280 			return -EINVAL;
281 		}
282 	}
283 
284 	rxq->max_pkt_len = max_pkt_len;
285 	if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
286 	    rxq->max_pkt_len > buf_size) {
287 		dev_data->scattered_rx = 1;
288 	}
289 	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
290 	IAVF_WRITE_FLUSH(hw);
291 
292 	return 0;
293 }
294 
295 static int
296 iavf_init_queues(struct rte_eth_dev *dev)
297 {
298 	struct iavf_rx_queue **rxq =
299 		(struct iavf_rx_queue **)dev->data->rx_queues;
300 	int i, ret = IAVF_SUCCESS;
301 
302 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
303 		if (!rxq[i] || !rxq[i]->q_set)
304 			continue;
305 		ret = iavf_init_rxq(dev, rxq[i]);
306 		if (ret != IAVF_SUCCESS)
307 			break;
308 	}
309 	/* set rx/tx function to vector/scatter/single-segment
310 	 * according to parameters
311 	 */
312 	iavf_set_rx_function(dev);
313 	iavf_set_tx_function(dev);
314 
315 	return ret;
316 }
317 
318 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
319 				     struct rte_intr_handle *intr_handle)
320 {
321 	struct iavf_adapter *adapter =
322 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
323 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
324 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
325 	uint16_t interval, i;
326 	int vec;
327 
328 	if (rte_intr_cap_multiple(intr_handle) &&
329 	    dev->data->dev_conf.intr_conf.rxq) {
330 		if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
331 			return -1;
332 	}
333 
334 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
335 		intr_handle->intr_vec =
336 			rte_zmalloc("intr_vec",
337 				    dev->data->nb_rx_queues * sizeof(int), 0);
338 		if (!intr_handle->intr_vec) {
339 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
340 				    dev->data->nb_rx_queues);
341 			return -1;
342 		}
343 	}
344 
345 	if (!dev->data->dev_conf.intr_conf.rxq ||
346 	    !rte_intr_dp_is_en(intr_handle)) {
347 		/* Rx interrupt disabled, Map interrupt only for writeback */
348 		vf->nb_msix = 1;
349 		if (vf->vf_res->vf_cap_flags &
350 		    VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
351 			/* If WB_ON_ITR supports, enable it */
352 			vf->msix_base = IAVF_RX_VEC_START;
353 			IAVF_WRITE_REG(hw,
354 				       IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
355 				       IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK |
356 				       IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK);
357 		} else {
358 			/* If no WB_ON_ITR offload flags, need to set
359 			 * interrupt for descriptor write back.
360 			 */
361 			vf->msix_base = IAVF_MISC_VEC_ID;
362 
363 			/* set ITR to max */
364 			interval = iavf_calc_itr_interval(
365 					IAVF_QUEUE_ITR_INTERVAL_MAX);
366 			IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
367 				       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
368 				       (IAVF_ITR_INDEX_DEFAULT <<
369 					IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
370 				       (interval <<
371 					IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
372 		}
373 		IAVF_WRITE_FLUSH(hw);
374 		/* map all queues to the same interrupt */
375 		for (i = 0; i < dev->data->nb_rx_queues; i++)
376 			vf->rxq_map[vf->msix_base] |= 1 << i;
377 	} else {
378 		if (!rte_intr_allow_others(intr_handle)) {
379 			vf->nb_msix = 1;
380 			vf->msix_base = IAVF_MISC_VEC_ID;
381 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
382 				vf->rxq_map[vf->msix_base] |= 1 << i;
383 				intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
384 			}
385 			PMD_DRV_LOG(DEBUG,
386 				    "vector %u are mapping to all Rx queues",
387 				    vf->msix_base);
388 		} else {
389 			/* If Rx interrupt is reuquired, and we can use
390 			 * multi interrupts, then the vec is from 1
391 			 */
392 			vf->nb_msix = RTE_MIN(vf->vf_res->max_vectors,
393 					      intr_handle->nb_efd);
394 			vf->msix_base = IAVF_RX_VEC_START;
395 			vec = IAVF_RX_VEC_START;
396 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
397 				vf->rxq_map[vec] |= 1 << i;
398 				intr_handle->intr_vec[i] = vec++;
399 				if (vec >= vf->nb_msix)
400 					vec = IAVF_RX_VEC_START;
401 			}
402 			PMD_DRV_LOG(DEBUG,
403 				    "%u vectors are mapping to %u Rx queues",
404 				    vf->nb_msix, dev->data->nb_rx_queues);
405 		}
406 	}
407 
408 	if (iavf_config_irq_map(adapter)) {
409 		PMD_DRV_LOG(ERR, "config interrupt mapping failed");
410 		return -1;
411 	}
412 	return 0;
413 }
414 
415 static int
416 iavf_start_queues(struct rte_eth_dev *dev)
417 {
418 	struct iavf_rx_queue *rxq;
419 	struct iavf_tx_queue *txq;
420 	int i;
421 
422 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
423 		txq = dev->data->tx_queues[i];
424 		if (txq->tx_deferred_start)
425 			continue;
426 		if (iavf_dev_tx_queue_start(dev, i) != 0) {
427 			PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
428 			return -1;
429 		}
430 	}
431 
432 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
433 		rxq = dev->data->rx_queues[i];
434 		if (rxq->rx_deferred_start)
435 			continue;
436 		if (iavf_dev_rx_queue_start(dev, i) != 0) {
437 			PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
438 			return -1;
439 		}
440 	}
441 
442 	return 0;
443 }
444 
445 static int
446 iavf_dev_start(struct rte_eth_dev *dev)
447 {
448 	struct iavf_adapter *adapter =
449 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
450 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
451 	struct rte_intr_handle *intr_handle = dev->intr_handle;
452 
453 	PMD_INIT_FUNC_TRACE();
454 
455 	adapter->stopped = 0;
456 
457 	vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
458 	vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
459 				      dev->data->nb_tx_queues);
460 
461 	if (iavf_init_queues(dev) != 0) {
462 		PMD_DRV_LOG(ERR, "failed to do Queue init");
463 		return -1;
464 	}
465 
466 	if (iavf_configure_queues(adapter) != 0) {
467 		PMD_DRV_LOG(ERR, "configure queues failed");
468 		goto err_queue;
469 	}
470 
471 	if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
472 		PMD_DRV_LOG(ERR, "configure irq failed");
473 		goto err_queue;
474 	}
475 	/* re-enable intr again, because efd assign may change */
476 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
477 		rte_intr_disable(intr_handle);
478 		rte_intr_enable(intr_handle);
479 	}
480 
481 	/* Set all mac addrs */
482 	iavf_add_del_all_mac_addr(adapter, true);
483 
484 	/* Set all multicast addresses */
485 	iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
486 				  true);
487 
488 	if (iavf_start_queues(dev) != 0) {
489 		PMD_DRV_LOG(ERR, "enable queues failed");
490 		goto err_mac;
491 	}
492 
493 	return 0;
494 
495 err_mac:
496 	iavf_add_del_all_mac_addr(adapter, false);
497 err_queue:
498 	return -1;
499 }
500 
501 static void
502 iavf_dev_stop(struct rte_eth_dev *dev)
503 {
504 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
505 	struct iavf_adapter *adapter =
506 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
507 	struct rte_intr_handle *intr_handle = dev->intr_handle;
508 
509 	PMD_INIT_FUNC_TRACE();
510 
511 	if (adapter->stopped == 1)
512 		return;
513 
514 	iavf_stop_queues(dev);
515 
516 	/* Disable the interrupt for Rx */
517 	rte_intr_efd_disable(intr_handle);
518 	/* Rx interrupt vector mapping free */
519 	if (intr_handle->intr_vec) {
520 		rte_free(intr_handle->intr_vec);
521 		intr_handle->intr_vec = NULL;
522 	}
523 
524 	/* remove all mac addrs */
525 	iavf_add_del_all_mac_addr(adapter, false);
526 
527 	/* remove all multicast addresses */
528 	iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
529 				  false);
530 
531 	adapter->stopped = 1;
532 }
533 
534 static int
535 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
536 {
537 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
538 
539 	dev_info->max_rx_queues = vf->vsi_res->num_queue_pairs;
540 	dev_info->max_tx_queues = vf->vsi_res->num_queue_pairs;
541 	dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
542 	dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
543 	dev_info->hash_key_size = vf->vf_res->rss_key_size;
544 	dev_info->reta_size = vf->vf_res->rss_lut_size;
545 	dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
546 	dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
547 	dev_info->rx_offload_capa =
548 		DEV_RX_OFFLOAD_VLAN_STRIP |
549 		DEV_RX_OFFLOAD_QINQ_STRIP |
550 		DEV_RX_OFFLOAD_IPV4_CKSUM |
551 		DEV_RX_OFFLOAD_UDP_CKSUM |
552 		DEV_RX_OFFLOAD_TCP_CKSUM |
553 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
554 		DEV_RX_OFFLOAD_SCATTER |
555 		DEV_RX_OFFLOAD_JUMBO_FRAME |
556 		DEV_RX_OFFLOAD_VLAN_FILTER |
557 		DEV_RX_OFFLOAD_RSS_HASH;
558 	dev_info->tx_offload_capa =
559 		DEV_TX_OFFLOAD_VLAN_INSERT |
560 		DEV_TX_OFFLOAD_QINQ_INSERT |
561 		DEV_TX_OFFLOAD_IPV4_CKSUM |
562 		DEV_TX_OFFLOAD_UDP_CKSUM |
563 		DEV_TX_OFFLOAD_TCP_CKSUM |
564 		DEV_TX_OFFLOAD_SCTP_CKSUM |
565 		DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
566 		DEV_TX_OFFLOAD_TCP_TSO |
567 		DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
568 		DEV_TX_OFFLOAD_GRE_TNL_TSO |
569 		DEV_TX_OFFLOAD_IPIP_TNL_TSO |
570 		DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
571 		DEV_TX_OFFLOAD_MULTI_SEGS;
572 
573 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
574 		.rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
575 		.rx_drop_en = 0,
576 		.offloads = 0,
577 	};
578 
579 	dev_info->default_txconf = (struct rte_eth_txconf) {
580 		.tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
581 		.tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
582 		.offloads = 0,
583 	};
584 
585 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
586 		.nb_max = IAVF_MAX_RING_DESC,
587 		.nb_min = IAVF_MIN_RING_DESC,
588 		.nb_align = IAVF_ALIGN_RING_DESC,
589 	};
590 
591 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
592 		.nb_max = IAVF_MAX_RING_DESC,
593 		.nb_min = IAVF_MIN_RING_DESC,
594 		.nb_align = IAVF_ALIGN_RING_DESC,
595 	};
596 
597 	return 0;
598 }
599 
600 static const uint32_t *
601 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
602 {
603 	static const uint32_t ptypes[] = {
604 		RTE_PTYPE_L2_ETHER,
605 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
606 		RTE_PTYPE_L4_FRAG,
607 		RTE_PTYPE_L4_ICMP,
608 		RTE_PTYPE_L4_NONFRAG,
609 		RTE_PTYPE_L4_SCTP,
610 		RTE_PTYPE_L4_TCP,
611 		RTE_PTYPE_L4_UDP,
612 		RTE_PTYPE_UNKNOWN
613 	};
614 	return ptypes;
615 }
616 
617 int
618 iavf_dev_link_update(struct rte_eth_dev *dev,
619 		    __rte_unused int wait_to_complete)
620 {
621 	struct rte_eth_link new_link;
622 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
623 
624 	memset(&new_link, 0, sizeof(new_link));
625 
626 	/* Only read status info stored in VF, and the info is updated
627 	 *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
628 	 */
629 	switch (vf->link_speed) {
630 	case 10:
631 		new_link.link_speed = ETH_SPEED_NUM_10M;
632 		break;
633 	case 100:
634 		new_link.link_speed = ETH_SPEED_NUM_100M;
635 		break;
636 	case 1000:
637 		new_link.link_speed = ETH_SPEED_NUM_1G;
638 		break;
639 	case 10000:
640 		new_link.link_speed = ETH_SPEED_NUM_10G;
641 		break;
642 	case 20000:
643 		new_link.link_speed = ETH_SPEED_NUM_20G;
644 		break;
645 	case 25000:
646 		new_link.link_speed = ETH_SPEED_NUM_25G;
647 		break;
648 	case 40000:
649 		new_link.link_speed = ETH_SPEED_NUM_40G;
650 		break;
651 	case 50000:
652 		new_link.link_speed = ETH_SPEED_NUM_50G;
653 		break;
654 	case 100000:
655 		new_link.link_speed = ETH_SPEED_NUM_100G;
656 		break;
657 	default:
658 		new_link.link_speed = ETH_SPEED_NUM_NONE;
659 		break;
660 	}
661 
662 	new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
663 	new_link.link_status = vf->link_up ? ETH_LINK_UP :
664 					     ETH_LINK_DOWN;
665 	new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
666 				ETH_LINK_SPEED_FIXED);
667 
668 	if (rte_atomic64_cmpset((uint64_t *)&dev->data->dev_link,
669 				*(uint64_t *)&dev->data->dev_link,
670 				*(uint64_t *)&new_link) == 0)
671 		return -1;
672 
673 	return 0;
674 }
675 
676 static int
677 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
678 {
679 	struct iavf_adapter *adapter =
680 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
681 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
682 	int ret;
683 
684 	if (vf->promisc_unicast_enabled)
685 		return 0;
686 
687 	ret = iavf_config_promisc(adapter, true, vf->promisc_multicast_enabled);
688 	if (!ret)
689 		vf->promisc_unicast_enabled = true;
690 	else if (ret == IAVF_NOT_SUPPORTED)
691 		ret = -ENOTSUP;
692 	else
693 		ret = -EAGAIN;
694 
695 	return ret;
696 }
697 
698 static int
699 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
700 {
701 	struct iavf_adapter *adapter =
702 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
703 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
704 	int ret;
705 
706 	if (!vf->promisc_unicast_enabled)
707 		return 0;
708 
709 	ret = iavf_config_promisc(adapter, false,
710 				  vf->promisc_multicast_enabled);
711 	if (!ret)
712 		vf->promisc_unicast_enabled = false;
713 	else if (ret == IAVF_NOT_SUPPORTED)
714 		ret = -ENOTSUP;
715 	else
716 		ret = -EAGAIN;
717 
718 	return ret;
719 }
720 
721 static int
722 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
723 {
724 	struct iavf_adapter *adapter =
725 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
726 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
727 	int ret;
728 
729 	if (vf->promisc_multicast_enabled)
730 		return 0;
731 
732 	ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, true);
733 	if (!ret)
734 		vf->promisc_multicast_enabled = true;
735 	else if (ret == IAVF_NOT_SUPPORTED)
736 		ret = -ENOTSUP;
737 	else
738 		ret = -EAGAIN;
739 
740 	return ret;
741 }
742 
743 static int
744 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
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 	int ret;
750 
751 	if (!vf->promisc_multicast_enabled)
752 		return 0;
753 
754 	ret = iavf_config_promisc(adapter, vf->promisc_unicast_enabled, false);
755 	if (!ret)
756 		vf->promisc_multicast_enabled = false;
757 	else if (ret == IAVF_NOT_SUPPORTED)
758 		ret = -ENOTSUP;
759 	else
760 		ret = -EAGAIN;
761 
762 	return ret;
763 }
764 
765 static int
766 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
767 		     __rte_unused uint32_t index,
768 		     __rte_unused uint32_t pool)
769 {
770 	struct iavf_adapter *adapter =
771 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
772 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
773 	int err;
774 
775 	if (rte_is_zero_ether_addr(addr)) {
776 		PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
777 		return -EINVAL;
778 	}
779 
780 	err = iavf_add_del_eth_addr(adapter, addr, true);
781 	if (err) {
782 		PMD_DRV_LOG(ERR, "fail to add MAC address");
783 		return -EIO;
784 	}
785 
786 	vf->mac_num++;
787 
788 	return 0;
789 }
790 
791 static void
792 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
793 {
794 	struct iavf_adapter *adapter =
795 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
796 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
797 	struct rte_ether_addr *addr;
798 	int err;
799 
800 	addr = &dev->data->mac_addrs[index];
801 
802 	err = iavf_add_del_eth_addr(adapter, addr, false);
803 	if (err)
804 		PMD_DRV_LOG(ERR, "fail to delete MAC address");
805 
806 	vf->mac_num--;
807 }
808 
809 static int
810 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
811 {
812 	struct iavf_adapter *adapter =
813 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
814 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
815 	int err;
816 
817 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
818 		return -ENOTSUP;
819 
820 	err = iavf_add_del_vlan(adapter, vlan_id, on);
821 	if (err)
822 		return -EIO;
823 	return 0;
824 }
825 
826 static int
827 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
828 {
829 	struct iavf_adapter *adapter =
830 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
831 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
832 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
833 	int err;
834 
835 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
836 		return -ENOTSUP;
837 
838 	/* Vlan stripping setting */
839 	if (mask & ETH_VLAN_STRIP_MASK) {
840 		/* Enable or disable VLAN stripping */
841 		if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
842 			err = iavf_enable_vlan_strip(adapter);
843 		else
844 			err = iavf_disable_vlan_strip(adapter);
845 
846 		if (err)
847 			return -EIO;
848 	}
849 	return 0;
850 }
851 
852 static int
853 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
854 			struct rte_eth_rss_reta_entry64 *reta_conf,
855 			uint16_t reta_size)
856 {
857 	struct iavf_adapter *adapter =
858 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
859 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
860 	uint8_t *lut;
861 	uint16_t i, idx, shift;
862 	int ret;
863 
864 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
865 		return -ENOTSUP;
866 
867 	if (reta_size != vf->vf_res->rss_lut_size) {
868 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
869 			"(%d) doesn't match the number of hardware can "
870 			"support (%d)", reta_size, vf->vf_res->rss_lut_size);
871 		return -EINVAL;
872 	}
873 
874 	lut = rte_zmalloc("rss_lut", reta_size, 0);
875 	if (!lut) {
876 		PMD_DRV_LOG(ERR, "No memory can be allocated");
877 		return -ENOMEM;
878 	}
879 	/* store the old lut table temporarily */
880 	rte_memcpy(lut, vf->rss_lut, reta_size);
881 
882 	for (i = 0; i < reta_size; i++) {
883 		idx = i / RTE_RETA_GROUP_SIZE;
884 		shift = i % RTE_RETA_GROUP_SIZE;
885 		if (reta_conf[idx].mask & (1ULL << shift))
886 			lut[i] = reta_conf[idx].reta[shift];
887 	}
888 
889 	rte_memcpy(vf->rss_lut, lut, reta_size);
890 	/* send virtchnnl ops to configure rss*/
891 	ret = iavf_configure_rss_lut(adapter);
892 	if (ret) /* revert back */
893 		rte_memcpy(vf->rss_lut, lut, reta_size);
894 	rte_free(lut);
895 
896 	return ret;
897 }
898 
899 static int
900 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
901 		       struct rte_eth_rss_reta_entry64 *reta_conf,
902 		       uint16_t reta_size)
903 {
904 	struct iavf_adapter *adapter =
905 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
906 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
907 	uint16_t i, idx, shift;
908 
909 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
910 		return -ENOTSUP;
911 
912 	if (reta_size != vf->vf_res->rss_lut_size) {
913 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
914 			"(%d) doesn't match the number of hardware can "
915 			"support (%d)", reta_size, vf->vf_res->rss_lut_size);
916 		return -EINVAL;
917 	}
918 
919 	for (i = 0; i < reta_size; i++) {
920 		idx = i / RTE_RETA_GROUP_SIZE;
921 		shift = i % RTE_RETA_GROUP_SIZE;
922 		if (reta_conf[idx].mask & (1ULL << shift))
923 			reta_conf[idx].reta[shift] = vf->rss_lut[i];
924 	}
925 
926 	return 0;
927 }
928 
929 static int
930 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
931 			struct rte_eth_rss_conf *rss_conf)
932 {
933 	struct iavf_adapter *adapter =
934 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
935 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
936 
937 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
938 		return -ENOTSUP;
939 
940 	/* HENA setting, it is enabled by default, no change */
941 	if (!rss_conf->rss_key || rss_conf->rss_key_len == 0) {
942 		PMD_DRV_LOG(DEBUG, "No key to be configured");
943 		return 0;
944 	} else if (rss_conf->rss_key_len != vf->vf_res->rss_key_size) {
945 		PMD_DRV_LOG(ERR, "The size of hash key configured "
946 			"(%d) doesn't match the size of hardware can "
947 			"support (%d)", rss_conf->rss_key_len,
948 			vf->vf_res->rss_key_size);
949 		return -EINVAL;
950 	}
951 
952 	rte_memcpy(vf->rss_key, rss_conf->rss_key, rss_conf->rss_key_len);
953 
954 	return iavf_configure_rss_key(adapter);
955 }
956 
957 static int
958 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
959 			  struct rte_eth_rss_conf *rss_conf)
960 {
961 	struct iavf_adapter *adapter =
962 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
963 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
964 
965 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
966 		return -ENOTSUP;
967 
968 	 /* Just set it to default value now. */
969 	rss_conf->rss_hf = IAVF_RSS_OFFLOAD_ALL;
970 
971 	if (!rss_conf->rss_key)
972 		return 0;
973 
974 	rss_conf->rss_key_len = vf->vf_res->rss_key_size;
975 	rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
976 
977 	return 0;
978 }
979 
980 static int
981 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
982 {
983 	uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
984 	int ret = 0;
985 
986 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
987 		return -EINVAL;
988 
989 	/* mtu setting is forbidden if port is start */
990 	if (dev->data->dev_started) {
991 		PMD_DRV_LOG(ERR, "port must be stopped before configuration");
992 		return -EBUSY;
993 	}
994 
995 	if (frame_size > RTE_ETHER_MAX_LEN)
996 		dev->data->dev_conf.rxmode.offloads |=
997 				DEV_RX_OFFLOAD_JUMBO_FRAME;
998 	else
999 		dev->data->dev_conf.rxmode.offloads &=
1000 				~DEV_RX_OFFLOAD_JUMBO_FRAME;
1001 
1002 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1003 
1004 	return ret;
1005 }
1006 
1007 static int
1008 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1009 			     struct rte_ether_addr *mac_addr)
1010 {
1011 	struct iavf_adapter *adapter =
1012 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1013 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1014 	struct rte_ether_addr *perm_addr, *old_addr;
1015 	int ret;
1016 
1017 	old_addr = (struct rte_ether_addr *)hw->mac.addr;
1018 	perm_addr = (struct rte_ether_addr *)hw->mac.perm_addr;
1019 
1020 	/* If the MAC address is configured by host, skip the setting */
1021 	if (rte_is_valid_assigned_ether_addr(perm_addr))
1022 		return -EPERM;
1023 
1024 	ret = iavf_add_del_eth_addr(adapter, old_addr, false);
1025 	if (ret)
1026 		PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1027 			    " %02X:%02X:%02X:%02X:%02X:%02X",
1028 			    old_addr->addr_bytes[0],
1029 			    old_addr->addr_bytes[1],
1030 			    old_addr->addr_bytes[2],
1031 			    old_addr->addr_bytes[3],
1032 			    old_addr->addr_bytes[4],
1033 			    old_addr->addr_bytes[5]);
1034 
1035 	ret = iavf_add_del_eth_addr(adapter, mac_addr, true);
1036 	if (ret)
1037 		PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1038 			    " %02X:%02X:%02X:%02X:%02X:%02X",
1039 			    mac_addr->addr_bytes[0],
1040 			    mac_addr->addr_bytes[1],
1041 			    mac_addr->addr_bytes[2],
1042 			    mac_addr->addr_bytes[3],
1043 			    mac_addr->addr_bytes[4],
1044 			    mac_addr->addr_bytes[5]);
1045 
1046 	if (ret)
1047 		return -EIO;
1048 
1049 	rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1050 	return 0;
1051 }
1052 
1053 static void
1054 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1055 {
1056 	if (*stat >= *offset)
1057 		*stat = *stat - *offset;
1058 	else
1059 		*stat = (uint64_t)((*stat +
1060 			((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1061 
1062 	*stat &= IAVF_48_BIT_MASK;
1063 }
1064 
1065 static void
1066 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1067 {
1068 	if (*stat >= *offset)
1069 		*stat = (uint64_t)(*stat - *offset);
1070 	else
1071 		*stat = (uint64_t)((*stat +
1072 			((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1073 }
1074 
1075 static void
1076 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1077 {
1078 	struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1079 
1080 	iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1081 	iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1082 	iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1083 	iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1084 	iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1085 	iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1086 	iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1087 	iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1088 	iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1089 	iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1090 	iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1091 }
1092 
1093 static int
1094 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1095 {
1096 	struct iavf_adapter *adapter =
1097 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1098 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1099 	struct iavf_vsi *vsi = &vf->vsi;
1100 	struct virtchnl_eth_stats *pstats = NULL;
1101 	int ret;
1102 
1103 	ret = iavf_query_stats(adapter, &pstats);
1104 	if (ret == 0) {
1105 		iavf_update_stats(vsi, pstats);
1106 		stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1107 				pstats->rx_broadcast - pstats->rx_discards;
1108 		stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1109 						pstats->tx_unicast;
1110 		stats->imissed = pstats->rx_discards;
1111 		stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1112 		stats->ibytes = pstats->rx_bytes;
1113 		stats->ibytes -= stats->ipackets * RTE_ETHER_CRC_LEN;
1114 		stats->obytes = pstats->tx_bytes;
1115 	} else {
1116 		PMD_DRV_LOG(ERR, "Get statistics failed");
1117 	}
1118 	return ret;
1119 }
1120 
1121 static int
1122 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1123 {
1124 	int ret;
1125 	struct iavf_adapter *adapter =
1126 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1127 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1128 	struct iavf_vsi *vsi = &vf->vsi;
1129 	struct virtchnl_eth_stats *pstats = NULL;
1130 
1131 	/* read stat values to clear hardware registers */
1132 	ret = iavf_query_stats(adapter, &pstats);
1133 	if (ret != 0)
1134 		return ret;
1135 
1136 	/* set stats offset base on current values */
1137 	vsi->eth_stats_offset = *pstats;
1138 
1139 	return 0;
1140 }
1141 
1142 static int
1143 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1144 {
1145 	struct iavf_adapter *adapter =
1146 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1147 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1148 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1149 	uint16_t msix_intr;
1150 
1151 	msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1152 	if (msix_intr == IAVF_MISC_VEC_ID) {
1153 		PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1154 		IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1155 			       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1156 			       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1157 			       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1158 	} else {
1159 		IAVF_WRITE_REG(hw,
1160 			       IAVF_VFINT_DYN_CTLN1
1161 				(msix_intr - IAVF_RX_VEC_START),
1162 			       IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1163 			       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1164 			       IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1165 	}
1166 
1167 	IAVF_WRITE_FLUSH(hw);
1168 
1169 	rte_intr_ack(&pci_dev->intr_handle);
1170 
1171 	return 0;
1172 }
1173 
1174 static int
1175 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1176 {
1177 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1178 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1179 	uint16_t msix_intr;
1180 
1181 	msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1182 	if (msix_intr == IAVF_MISC_VEC_ID) {
1183 		PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1184 		return -EIO;
1185 	}
1186 
1187 	IAVF_WRITE_REG(hw,
1188 		      IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1189 		      0);
1190 
1191 	IAVF_WRITE_FLUSH(hw);
1192 	return 0;
1193 }
1194 
1195 static int
1196 iavf_check_vf_reset_done(struct iavf_hw *hw)
1197 {
1198 	int i, reset;
1199 
1200 	for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1201 		reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1202 			IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1203 		reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1204 		if (reset == VIRTCHNL_VFR_VFACTIVE ||
1205 		    reset == VIRTCHNL_VFR_COMPLETED)
1206 			break;
1207 		rte_delay_ms(20);
1208 	}
1209 
1210 	if (i >= IAVF_RESET_WAIT_CNT)
1211 		return -1;
1212 
1213 	return 0;
1214 }
1215 
1216 static int
1217 iavf_init_vf(struct rte_eth_dev *dev)
1218 {
1219 	int err, bufsz;
1220 	struct iavf_adapter *adapter =
1221 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1222 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1223 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1224 
1225 	err = iavf_set_mac_type(hw);
1226 	if (err) {
1227 		PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1228 		goto err;
1229 	}
1230 
1231 	err = iavf_check_vf_reset_done(hw);
1232 	if (err) {
1233 		PMD_INIT_LOG(ERR, "VF is still resetting");
1234 		goto err;
1235 	}
1236 
1237 	iavf_init_adminq_parameter(hw);
1238 	err = iavf_init_adminq(hw);
1239 	if (err) {
1240 		PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1241 		goto err;
1242 	}
1243 
1244 	vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1245 	if (!vf->aq_resp) {
1246 		PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1247 		goto err_aq;
1248 	}
1249 	if (iavf_check_api_version(adapter) != 0) {
1250 		PMD_INIT_LOG(ERR, "check_api version failed");
1251 		goto err_api;
1252 	}
1253 
1254 	bufsz = sizeof(struct virtchnl_vf_resource) +
1255 		(IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1256 	vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1257 	if (!vf->vf_res) {
1258 		PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1259 		goto err_api;
1260 	}
1261 	if (iavf_get_vf_resource(adapter) != 0) {
1262 		PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1263 		goto err_alloc;
1264 	}
1265 	/* Allocate memort for RSS info */
1266 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1267 		vf->rss_key = rte_zmalloc("rss_key",
1268 					  vf->vf_res->rss_key_size, 0);
1269 		if (!vf->rss_key) {
1270 			PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
1271 			goto err_rss;
1272 		}
1273 		vf->rss_lut = rte_zmalloc("rss_lut",
1274 					  vf->vf_res->rss_lut_size, 0);
1275 		if (!vf->rss_lut) {
1276 			PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
1277 			goto err_rss;
1278 		}
1279 	}
1280 
1281 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
1282 		if (iavf_get_supported_rxdid(adapter) != 0) {
1283 			PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
1284 			goto err_rss;
1285 		}
1286 	}
1287 
1288 	vf->vf_reset = false;
1289 
1290 	return 0;
1291 err_rss:
1292 	rte_free(vf->rss_key);
1293 	rte_free(vf->rss_lut);
1294 err_alloc:
1295 	rte_free(vf->vf_res);
1296 	vf->vsi_res = NULL;
1297 err_api:
1298 	rte_free(vf->aq_resp);
1299 err_aq:
1300 	iavf_shutdown_adminq(hw);
1301 err:
1302 	return -1;
1303 }
1304 
1305 /* Enable default admin queue interrupt setting */
1306 static inline void
1307 iavf_enable_irq0(struct iavf_hw *hw)
1308 {
1309 	/* Enable admin queue interrupt trigger */
1310 	IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
1311 		       IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
1312 
1313 	IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1314 		       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1315 		       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1316 		       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1317 
1318 	IAVF_WRITE_FLUSH(hw);
1319 }
1320 
1321 static inline void
1322 iavf_disable_irq0(struct iavf_hw *hw)
1323 {
1324 	/* Disable all interrupt types */
1325 	IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
1326 	IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1327 		       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1328 	IAVF_WRITE_FLUSH(hw);
1329 }
1330 
1331 static void
1332 iavf_dev_interrupt_handler(void *param)
1333 {
1334 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1335 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1336 
1337 	iavf_disable_irq0(hw);
1338 
1339 	iavf_handle_virtchnl_msg(dev);
1340 
1341 	iavf_enable_irq0(hw);
1342 }
1343 
1344 static int
1345 iavf_dev_filter_ctrl(struct rte_eth_dev *dev,
1346 		     enum rte_filter_type filter_type,
1347 		     enum rte_filter_op filter_op,
1348 		     void *arg)
1349 {
1350 	int ret = 0;
1351 
1352 	if (!dev)
1353 		return -EINVAL;
1354 
1355 	switch (filter_type) {
1356 	case RTE_ETH_FILTER_GENERIC:
1357 		if (filter_op != RTE_ETH_FILTER_GET)
1358 			return -EINVAL;
1359 		*(const void **)arg = &iavf_flow_ops;
1360 		break;
1361 	default:
1362 		PMD_DRV_LOG(WARNING, "Filter type (%d) not supported",
1363 			    filter_type);
1364 		ret = -EINVAL;
1365 		break;
1366 	}
1367 
1368 	return ret;
1369 }
1370 
1371 
1372 static int
1373 iavf_dev_init(struct rte_eth_dev *eth_dev)
1374 {
1375 	struct iavf_adapter *adapter =
1376 		IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1377 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1378 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
1379 	int ret = 0;
1380 
1381 	PMD_INIT_FUNC_TRACE();
1382 
1383 	/* assign ops func pointer */
1384 	eth_dev->dev_ops = &iavf_eth_dev_ops;
1385 	eth_dev->rx_queue_count = iavf_dev_rxq_count;
1386 	eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
1387 	eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
1388 	eth_dev->rx_pkt_burst = &iavf_recv_pkts;
1389 	eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
1390 	eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
1391 
1392 	/* For secondary processes, we don't initialise any further as primary
1393 	 * has already done this work. Only check if we need a different RX
1394 	 * and TX function.
1395 	 */
1396 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
1397 		iavf_set_rx_function(eth_dev);
1398 		iavf_set_tx_function(eth_dev);
1399 		return 0;
1400 	}
1401 	rte_eth_copy_pci_info(eth_dev, pci_dev);
1402 
1403 	hw->vendor_id = pci_dev->id.vendor_id;
1404 	hw->device_id = pci_dev->id.device_id;
1405 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
1406 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
1407 	hw->bus.bus_id = pci_dev->addr.bus;
1408 	hw->bus.device = pci_dev->addr.devid;
1409 	hw->bus.func = pci_dev->addr.function;
1410 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
1411 	hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
1412 	adapter->eth_dev = eth_dev;
1413 	adapter->stopped = 1;
1414 
1415 	/* Pass the information to the rte_eth_dev_close() that it should also
1416 	 * release the private port resources.
1417 	 */
1418 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
1419 
1420 	if (iavf_init_vf(eth_dev) != 0) {
1421 		PMD_INIT_LOG(ERR, "Init vf failed");
1422 		return -1;
1423 	}
1424 
1425 	/* set default ptype table */
1426 	adapter->ptype_tbl = iavf_get_default_ptype_table();
1427 
1428 	/* copy mac addr */
1429 	eth_dev->data->mac_addrs = rte_zmalloc(
1430 		"iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
1431 	if (!eth_dev->data->mac_addrs) {
1432 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
1433 			     " store MAC addresses",
1434 			     RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
1435 		return -ENOMEM;
1436 	}
1437 	/* If the MAC address is not configured by host,
1438 	 * generate a random one.
1439 	 */
1440 	if (!rte_is_valid_assigned_ether_addr(
1441 			(struct rte_ether_addr *)hw->mac.addr))
1442 		rte_eth_random_addr(hw->mac.addr);
1443 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
1444 			&eth_dev->data->mac_addrs[0]);
1445 
1446 	/* register callback func to eal lib */
1447 	rte_intr_callback_register(&pci_dev->intr_handle,
1448 				   iavf_dev_interrupt_handler,
1449 				   (void *)eth_dev);
1450 
1451 	/* enable uio intr after callback register */
1452 	rte_intr_enable(&pci_dev->intr_handle);
1453 
1454 	/* configure and enable device interrupt */
1455 	iavf_enable_irq0(hw);
1456 
1457 	ret = iavf_flow_init(adapter);
1458 	if (ret) {
1459 		PMD_INIT_LOG(ERR, "Failed to initialize flow");
1460 		return ret;
1461 	}
1462 
1463 	return 0;
1464 }
1465 
1466 static void
1467 iavf_dev_close(struct rte_eth_dev *dev)
1468 {
1469 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1470 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1471 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1472 	struct iavf_adapter *adapter =
1473 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1474 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1475 
1476 	iavf_dev_stop(dev);
1477 	iavf_flow_flush(dev, NULL);
1478 	iavf_flow_uninit(adapter);
1479 	iavf_shutdown_adminq(hw);
1480 	/* disable uio intr before callback unregister */
1481 	rte_intr_disable(intr_handle);
1482 
1483 	/* unregister callback func from eal lib */
1484 	rte_intr_callback_unregister(intr_handle,
1485 				     iavf_dev_interrupt_handler, dev);
1486 	iavf_disable_irq0(hw);
1487 
1488 	dev->dev_ops = NULL;
1489 	dev->rx_pkt_burst = NULL;
1490 	dev->tx_pkt_burst = NULL;
1491 
1492 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1493 		if (vf->rss_lut) {
1494 			rte_free(vf->rss_lut);
1495 			vf->rss_lut = NULL;
1496 		}
1497 		if (vf->rss_key) {
1498 			rte_free(vf->rss_key);
1499 			vf->rss_key = NULL;
1500 		}
1501 	}
1502 
1503 	rte_free(vf->vf_res);
1504 	vf->vsi_res = NULL;
1505 	vf->vf_res = NULL;
1506 
1507 	rte_free(vf->aq_resp);
1508 	vf->aq_resp = NULL;
1509 }
1510 
1511 static int
1512 iavf_dev_uninit(struct rte_eth_dev *dev)
1513 {
1514 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
1515 		return -EPERM;
1516 
1517 	iavf_dev_close(dev);
1518 
1519 	return 0;
1520 }
1521 
1522 /*
1523  * Reset VF device only to re-initialize resources in PMD layer
1524  */
1525 static int
1526 iavf_dev_reset(struct rte_eth_dev *dev)
1527 {
1528 	int ret;
1529 
1530 	ret = iavf_dev_uninit(dev);
1531 	if (ret)
1532 		return ret;
1533 
1534 	return iavf_dev_init(dev);
1535 }
1536 
1537 static int
1538 iavf_dcf_cap_check_handler(__rte_unused const char *key,
1539 			   const char *value, __rte_unused void *opaque)
1540 {
1541 	if (strcmp(value, "dcf"))
1542 		return -1;
1543 
1544 	return 0;
1545 }
1546 
1547 static int
1548 iavf_dcf_cap_selected(struct rte_devargs *devargs)
1549 {
1550 	struct rte_kvargs *kvlist;
1551 	const char *key = "cap";
1552 	int ret = 0;
1553 
1554 	if (devargs == NULL)
1555 		return 0;
1556 
1557 	kvlist = rte_kvargs_parse(devargs->args, NULL);
1558 	if (kvlist == NULL)
1559 		return 0;
1560 
1561 	if (!rte_kvargs_count(kvlist, key))
1562 		goto exit;
1563 
1564 	/* dcf capability selected when there's a key-value pair: cap=dcf */
1565 	if (rte_kvargs_process(kvlist, key,
1566 			       iavf_dcf_cap_check_handler, NULL) < 0)
1567 		goto exit;
1568 
1569 	ret = 1;
1570 
1571 exit:
1572 	rte_kvargs_free(kvlist);
1573 	return ret;
1574 }
1575 
1576 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
1577 			     struct rte_pci_device *pci_dev)
1578 {
1579 	if (iavf_dcf_cap_selected(pci_dev->device.devargs))
1580 		return 1;
1581 
1582 	return rte_eth_dev_pci_generic_probe(pci_dev,
1583 		sizeof(struct iavf_adapter), iavf_dev_init);
1584 }
1585 
1586 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
1587 {
1588 	return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
1589 }
1590 
1591 /* Adaptive virtual function driver struct */
1592 static struct rte_pci_driver rte_iavf_pmd = {
1593 	.id_table = pci_id_iavf_map,
1594 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
1595 	.probe = eth_iavf_pci_probe,
1596 	.remove = eth_iavf_pci_remove,
1597 };
1598 
1599 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
1600 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
1601 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
1602 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf");
1603 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
1604 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
1605 #ifdef RTE_LIBRTE_IAVF_DEBUG_RX
1606 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
1607 #endif
1608 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX
1609 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
1610 #endif
1611 #ifdef RTE_LIBRTE_IAVF_DEBUG_TX_FREE
1612 RTE_LOG_REGISTER(iavf_logtype_tx_free, pmd.net.iavf.tx_free, DEBUG);
1613 #endif
1614