xref: /dpdk/drivers/net/iavf/iavf_ethdev.c (revision 7fa949fd)
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 <ethdev_driver.h>
23 #include <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 #include "rte_pmd_iavf.h"
32 
33 /* devargs */
34 #define IAVF_PROTO_XTR_ARG         "proto_xtr"
35 
36 static const char * const iavf_valid_args[] = {
37 	IAVF_PROTO_XTR_ARG,
38 	NULL
39 };
40 
41 static const struct rte_mbuf_dynfield iavf_proto_xtr_metadata_param = {
42 	.name = "intel_pmd_dynfield_proto_xtr_metadata",
43 	.size = sizeof(uint32_t),
44 	.align = __alignof__(uint32_t),
45 	.flags = 0,
46 };
47 
48 struct iavf_proto_xtr_ol {
49 	const struct rte_mbuf_dynflag param;
50 	uint64_t *ol_flag;
51 	bool required;
52 };
53 
54 static struct iavf_proto_xtr_ol iavf_proto_xtr_params[] = {
55 	[IAVF_PROTO_XTR_VLAN] = {
56 		.param = { .name = "intel_pmd_dynflag_proto_xtr_vlan" },
57 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_vlan_mask },
58 	[IAVF_PROTO_XTR_IPV4] = {
59 		.param = { .name = "intel_pmd_dynflag_proto_xtr_ipv4" },
60 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv4_mask },
61 	[IAVF_PROTO_XTR_IPV6] = {
62 		.param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6" },
63 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_mask },
64 	[IAVF_PROTO_XTR_IPV6_FLOW] = {
65 		.param = { .name = "intel_pmd_dynflag_proto_xtr_ipv6_flow" },
66 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ipv6_flow_mask },
67 	[IAVF_PROTO_XTR_TCP] = {
68 		.param = { .name = "intel_pmd_dynflag_proto_xtr_tcp" },
69 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_tcp_mask },
70 	[IAVF_PROTO_XTR_IP_OFFSET] = {
71 		.param = { .name = "intel_pmd_dynflag_proto_xtr_ip_offset" },
72 		.ol_flag = &rte_pmd_ifd_dynflag_proto_xtr_ip_offset_mask },
73 };
74 
75 static int iavf_dev_configure(struct rte_eth_dev *dev);
76 static int iavf_dev_start(struct rte_eth_dev *dev);
77 static int iavf_dev_stop(struct rte_eth_dev *dev);
78 static int iavf_dev_close(struct rte_eth_dev *dev);
79 static int iavf_dev_reset(struct rte_eth_dev *dev);
80 static int iavf_dev_info_get(struct rte_eth_dev *dev,
81 			     struct rte_eth_dev_info *dev_info);
82 static const uint32_t *iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev);
83 static int iavf_dev_stats_get(struct rte_eth_dev *dev,
84 			     struct rte_eth_stats *stats);
85 static int iavf_dev_stats_reset(struct rte_eth_dev *dev);
86 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
87 				 struct rte_eth_xstat *xstats, unsigned int n);
88 static int iavf_dev_xstats_get_names(struct rte_eth_dev *dev,
89 				       struct rte_eth_xstat_name *xstats_names,
90 				       unsigned int limit);
91 static int iavf_dev_promiscuous_enable(struct rte_eth_dev *dev);
92 static int iavf_dev_promiscuous_disable(struct rte_eth_dev *dev);
93 static int iavf_dev_allmulticast_enable(struct rte_eth_dev *dev);
94 static int iavf_dev_allmulticast_disable(struct rte_eth_dev *dev);
95 static int iavf_dev_add_mac_addr(struct rte_eth_dev *dev,
96 				struct rte_ether_addr *addr,
97 				uint32_t index,
98 				uint32_t pool);
99 static void iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index);
100 static int iavf_dev_vlan_filter_set(struct rte_eth_dev *dev,
101 				   uint16_t vlan_id, int on);
102 static int iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
103 static int iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
104 				   struct rte_eth_rss_reta_entry64 *reta_conf,
105 				   uint16_t reta_size);
106 static int iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
107 				  struct rte_eth_rss_reta_entry64 *reta_conf,
108 				  uint16_t reta_size);
109 static int iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
110 				   struct rte_eth_rss_conf *rss_conf);
111 static int iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
112 				     struct rte_eth_rss_conf *rss_conf);
113 static int iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
114 static int iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
115 					 struct rte_ether_addr *mac_addr);
116 static int iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev,
117 					uint16_t queue_id);
118 static int iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev,
119 					 uint16_t queue_id);
120 static int iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
121 				 const struct rte_flow_ops **ops);
122 static int iavf_set_mc_addr_list(struct rte_eth_dev *dev,
123 			struct rte_ether_addr *mc_addrs,
124 			uint32_t mc_addrs_num);
125 
126 static const struct rte_pci_id pci_id_iavf_map[] = {
127 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_ADAPTIVE_VF) },
128 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF) },
129 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_VF_HV) },
130 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_VF) },
131 	{ RTE_PCI_DEVICE(IAVF_INTEL_VENDOR_ID, IAVF_DEV_ID_X722_A0_VF) },
132 	{ .vendor_id = 0, /* sentinel */ },
133 };
134 
135 struct rte_iavf_xstats_name_off {
136 	char name[RTE_ETH_XSTATS_NAME_SIZE];
137 	unsigned int offset;
138 };
139 
140 static const struct rte_iavf_xstats_name_off rte_iavf_stats_strings[] = {
141 	{"rx_bytes", offsetof(struct iavf_eth_stats, rx_bytes)},
142 	{"rx_unicast_packets", offsetof(struct iavf_eth_stats, rx_unicast)},
143 	{"rx_multicast_packets", offsetof(struct iavf_eth_stats, rx_multicast)},
144 	{"rx_broadcast_packets", offsetof(struct iavf_eth_stats, rx_broadcast)},
145 	{"rx_dropped_packets", offsetof(struct iavf_eth_stats, rx_discards)},
146 	{"rx_unknown_protocol_packets", offsetof(struct iavf_eth_stats,
147 		rx_unknown_protocol)},
148 	{"tx_bytes", offsetof(struct iavf_eth_stats, tx_bytes)},
149 	{"tx_unicast_packets", offsetof(struct iavf_eth_stats, tx_unicast)},
150 	{"tx_multicast_packets", offsetof(struct iavf_eth_stats, tx_multicast)},
151 	{"tx_broadcast_packets", offsetof(struct iavf_eth_stats, tx_broadcast)},
152 	{"tx_dropped_packets", offsetof(struct iavf_eth_stats, tx_discards)},
153 	{"tx_error_packets", offsetof(struct iavf_eth_stats, tx_errors)},
154 };
155 
156 #define IAVF_NB_XSTATS (sizeof(rte_iavf_stats_strings) / \
157 		sizeof(rte_iavf_stats_strings[0]))
158 
159 static const struct eth_dev_ops iavf_eth_dev_ops = {
160 	.dev_configure              = iavf_dev_configure,
161 	.dev_start                  = iavf_dev_start,
162 	.dev_stop                   = iavf_dev_stop,
163 	.dev_close                  = iavf_dev_close,
164 	.dev_reset                  = iavf_dev_reset,
165 	.dev_infos_get              = iavf_dev_info_get,
166 	.dev_supported_ptypes_get   = iavf_dev_supported_ptypes_get,
167 	.link_update                = iavf_dev_link_update,
168 	.stats_get                  = iavf_dev_stats_get,
169 	.stats_reset                = iavf_dev_stats_reset,
170 	.xstats_get                 = iavf_dev_xstats_get,
171 	.xstats_get_names           = iavf_dev_xstats_get_names,
172 	.xstats_reset               = iavf_dev_stats_reset,
173 	.promiscuous_enable         = iavf_dev_promiscuous_enable,
174 	.promiscuous_disable        = iavf_dev_promiscuous_disable,
175 	.allmulticast_enable        = iavf_dev_allmulticast_enable,
176 	.allmulticast_disable       = iavf_dev_allmulticast_disable,
177 	.mac_addr_add               = iavf_dev_add_mac_addr,
178 	.mac_addr_remove            = iavf_dev_del_mac_addr,
179 	.set_mc_addr_list			= iavf_set_mc_addr_list,
180 	.vlan_filter_set            = iavf_dev_vlan_filter_set,
181 	.vlan_offload_set           = iavf_dev_vlan_offload_set,
182 	.rx_queue_start             = iavf_dev_rx_queue_start,
183 	.rx_queue_stop              = iavf_dev_rx_queue_stop,
184 	.tx_queue_start             = iavf_dev_tx_queue_start,
185 	.tx_queue_stop              = iavf_dev_tx_queue_stop,
186 	.rx_queue_setup             = iavf_dev_rx_queue_setup,
187 	.rx_queue_release           = iavf_dev_rx_queue_release,
188 	.tx_queue_setup             = iavf_dev_tx_queue_setup,
189 	.tx_queue_release           = iavf_dev_tx_queue_release,
190 	.mac_addr_set               = iavf_dev_set_default_mac_addr,
191 	.reta_update                = iavf_dev_rss_reta_update,
192 	.reta_query                 = iavf_dev_rss_reta_query,
193 	.rss_hash_update            = iavf_dev_rss_hash_update,
194 	.rss_hash_conf_get          = iavf_dev_rss_hash_conf_get,
195 	.rxq_info_get               = iavf_dev_rxq_info_get,
196 	.txq_info_get               = iavf_dev_txq_info_get,
197 	.mtu_set                    = iavf_dev_mtu_set,
198 	.rx_queue_intr_enable       = iavf_dev_rx_queue_intr_enable,
199 	.rx_queue_intr_disable      = iavf_dev_rx_queue_intr_disable,
200 	.flow_ops_get               = iavf_dev_flow_ops_get,
201 	.tx_done_cleanup	    = iavf_dev_tx_done_cleanup,
202 	.get_monitor_addr           = iavf_get_monitor_addr,
203 };
204 
205 static int
206 iavf_set_mc_addr_list(struct rte_eth_dev *dev,
207 			struct rte_ether_addr *mc_addrs,
208 			uint32_t mc_addrs_num)
209 {
210 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
211 	struct iavf_adapter *adapter =
212 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
213 	int err, ret;
214 
215 	if (mc_addrs_num > IAVF_NUM_MACADDR_MAX) {
216 		PMD_DRV_LOG(ERR,
217 			    "can't add more than a limited number (%u) of addresses.",
218 			    (uint32_t)IAVF_NUM_MACADDR_MAX);
219 		return -EINVAL;
220 	}
221 
222 	/* flush previous addresses */
223 	err = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
224 					false);
225 	if (err)
226 		return err;
227 
228 	/* add new ones */
229 	err = iavf_add_del_mc_addr_list(adapter, mc_addrs, mc_addrs_num, true);
230 
231 	if (err) {
232 		/* if adding mac address list fails, should add the previous
233 		 * addresses back.
234 		 */
235 		ret = iavf_add_del_mc_addr_list(adapter, vf->mc_addrs,
236 						vf->mc_addrs_num, true);
237 		if (ret)
238 			return ret;
239 	} else {
240 		vf->mc_addrs_num = mc_addrs_num;
241 		memcpy(vf->mc_addrs,
242 		       mc_addrs, mc_addrs_num * sizeof(*mc_addrs));
243 	}
244 
245 	return err;
246 }
247 
248 static int
249 iavf_init_rss(struct iavf_adapter *adapter)
250 {
251 	struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(adapter);
252 	struct rte_eth_rss_conf *rss_conf;
253 	uint16_t i, j, nb_q;
254 	int ret;
255 
256 	rss_conf = &adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf;
257 	nb_q = RTE_MIN(adapter->eth_dev->data->nb_rx_queues,
258 		       vf->max_rss_qregion);
259 
260 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF)) {
261 		PMD_DRV_LOG(DEBUG, "RSS is not supported");
262 		return -ENOTSUP;
263 	}
264 	if (adapter->eth_dev->data->dev_conf.rxmode.mq_mode != ETH_MQ_RX_RSS) {
265 		PMD_DRV_LOG(WARNING, "RSS is enabled by PF by default");
266 		/* set all lut items to default queue */
267 		for (i = 0; i < vf->vf_res->rss_lut_size; i++)
268 			vf->rss_lut[i] = 0;
269 		ret = iavf_configure_rss_lut(adapter);
270 		return ret;
271 	}
272 
273 	/* configure RSS key */
274 	if (!rss_conf->rss_key) {
275 		/* Calculate the default hash key */
276 		for (i = 0; i <= vf->vf_res->rss_key_size; i++)
277 			vf->rss_key[i] = (uint8_t)rte_rand();
278 	} else
279 		rte_memcpy(vf->rss_key, rss_conf->rss_key,
280 			   RTE_MIN(rss_conf->rss_key_len,
281 				   vf->vf_res->rss_key_size));
282 
283 	/* init RSS LUT table */
284 	for (i = 0, j = 0; i < vf->vf_res->rss_lut_size; i++, j++) {
285 		if (j >= nb_q)
286 			j = 0;
287 		vf->rss_lut[i] = j;
288 	}
289 	/* send virtchnnl ops to configure rss*/
290 	ret = iavf_configure_rss_lut(adapter);
291 	if (ret)
292 		return ret;
293 	ret = iavf_configure_rss_key(adapter);
294 	if (ret)
295 		return ret;
296 
297 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
298 		/* Set RSS hash configuration based on rss_conf->rss_hf. */
299 		ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
300 		if (ret) {
301 			PMD_DRV_LOG(ERR, "fail to set default RSS");
302 			return ret;
303 		}
304 	}
305 
306 	return 0;
307 }
308 
309 static int
310 iavf_queues_req_reset(struct rte_eth_dev *dev, uint16_t num)
311 {
312 	struct iavf_adapter *ad =
313 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
314 	struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
315 	int ret;
316 
317 	ret = iavf_request_queues(ad, num);
318 	if (ret) {
319 		PMD_DRV_LOG(ERR, "request queues from PF failed");
320 		return ret;
321 	}
322 	PMD_DRV_LOG(INFO, "change queue pairs from %u to %u",
323 			vf->vsi_res->num_queue_pairs, num);
324 
325 	ret = iavf_dev_reset(dev);
326 	if (ret) {
327 		PMD_DRV_LOG(ERR, "vf reset failed");
328 		return ret;
329 	}
330 
331 	return 0;
332 }
333 
334 static int
335 iavf_dev_vlan_insert_set(struct rte_eth_dev *dev)
336 {
337 	struct iavf_adapter *adapter =
338 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
339 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
340 	bool enable;
341 
342 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2))
343 		return 0;
344 
345 	enable = !!(dev->data->dev_conf.txmode.offloads &
346 		    DEV_TX_OFFLOAD_VLAN_INSERT);
347 	iavf_config_vlan_insert_v2(adapter, enable);
348 
349 	return 0;
350 }
351 
352 static int
353 iavf_dev_init_vlan(struct rte_eth_dev *dev)
354 {
355 	int err;
356 
357 	err = iavf_dev_vlan_offload_set(dev,
358 					ETH_VLAN_STRIP_MASK |
359 					ETH_QINQ_STRIP_MASK |
360 					ETH_VLAN_FILTER_MASK |
361 					ETH_VLAN_EXTEND_MASK);
362 	if (err) {
363 		PMD_DRV_LOG(ERR, "Failed to update vlan offload");
364 		return err;
365 	}
366 
367 	err = iavf_dev_vlan_insert_set(dev);
368 	if (err)
369 		PMD_DRV_LOG(ERR, "Failed to update vlan insertion");
370 
371 	return err;
372 }
373 
374 static int
375 iavf_dev_configure(struct rte_eth_dev *dev)
376 {
377 	struct iavf_adapter *ad =
378 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
379 	struct iavf_info *vf =  IAVF_DEV_PRIVATE_TO_VF(ad);
380 	uint16_t num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
381 		dev->data->nb_tx_queues);
382 	int ret;
383 
384 	ad->rx_bulk_alloc_allowed = true;
385 	/* Initialize to TRUE. If any of Rx queues doesn't meet the
386 	 * vector Rx/Tx preconditions, it will be reset.
387 	 */
388 	ad->rx_vec_allowed = true;
389 	ad->tx_vec_allowed = true;
390 
391 	if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
392 		dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
393 
394 	/* Large VF setting */
395 	if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_DFLT) {
396 		if (!(vf->vf_res->vf_cap_flags &
397 				VIRTCHNL_VF_LARGE_NUM_QPAIRS)) {
398 			PMD_DRV_LOG(ERR, "large VF is not supported");
399 			return -1;
400 		}
401 
402 		if (num_queue_pairs > IAVF_MAX_NUM_QUEUES_LV) {
403 			PMD_DRV_LOG(ERR, "queue pairs number cannot be larger than %u",
404 				IAVF_MAX_NUM_QUEUES_LV);
405 			return -1;
406 		}
407 
408 		ret = iavf_queues_req_reset(dev, num_queue_pairs);
409 		if (ret)
410 			return ret;
411 
412 		ret = iavf_get_max_rss_queue_region(ad);
413 		if (ret) {
414 			PMD_INIT_LOG(ERR, "get max rss queue region failed");
415 			return ret;
416 		}
417 
418 		vf->lv_enabled = true;
419 	} else {
420 		/* Check if large VF is already enabled. If so, disable and
421 		 * release redundant queue resource.
422 		 * Or check if enough queue pairs. If not, request them from PF.
423 		 */
424 		if (vf->lv_enabled ||
425 		    num_queue_pairs > vf->vsi_res->num_queue_pairs) {
426 			ret = iavf_queues_req_reset(dev, num_queue_pairs);
427 			if (ret)
428 				return ret;
429 
430 			vf->lv_enabled = false;
431 		}
432 		/* if large VF is not required, use default rss queue region */
433 		vf->max_rss_qregion = IAVF_MAX_NUM_QUEUES_DFLT;
434 	}
435 
436 	ret = iavf_dev_init_vlan(dev);
437 	if (ret)
438 		PMD_DRV_LOG(ERR, "configure VLAN failed: %d", ret);
439 
440 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
441 		if (iavf_init_rss(ad) != 0) {
442 			PMD_DRV_LOG(ERR, "configure rss failed");
443 			return -1;
444 		}
445 	}
446 	return 0;
447 }
448 
449 static int
450 iavf_init_rxq(struct rte_eth_dev *dev, struct iavf_rx_queue *rxq)
451 {
452 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
453 	struct rte_eth_dev_data *dev_data = dev->data;
454 	uint16_t buf_size, max_pkt_len, len;
455 
456 	buf_size = rte_pktmbuf_data_room_size(rxq->mp) - RTE_PKTMBUF_HEADROOM;
457 
458 	/* Calculate the maximum packet length allowed */
459 	len = rxq->rx_buf_len * IAVF_MAX_CHAINED_RX_BUFFERS;
460 	max_pkt_len = RTE_MIN(len, dev->data->dev_conf.rxmode.max_rx_pkt_len);
461 
462 	/* Check if the jumbo frame and maximum packet length are set
463 	 * correctly.
464 	 */
465 	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
466 		if (max_pkt_len <= IAVF_ETH_MAX_LEN ||
467 		    max_pkt_len > IAVF_FRAME_SIZE_MAX) {
468 			PMD_DRV_LOG(ERR, "maximum packet length must be "
469 				    "larger than %u and smaller than %u, "
470 				    "as jumbo frame is enabled",
471 				    (uint32_t)IAVF_ETH_MAX_LEN,
472 				    (uint32_t)IAVF_FRAME_SIZE_MAX);
473 			return -EINVAL;
474 		}
475 	} else {
476 		if (max_pkt_len < RTE_ETHER_MIN_LEN ||
477 		    max_pkt_len > IAVF_ETH_MAX_LEN) {
478 			PMD_DRV_LOG(ERR, "maximum packet length must be "
479 				    "larger than %u and smaller than %u, "
480 				    "as jumbo frame is disabled",
481 				    (uint32_t)RTE_ETHER_MIN_LEN,
482 				    (uint32_t)IAVF_ETH_MAX_LEN);
483 			return -EINVAL;
484 		}
485 	}
486 
487 	rxq->max_pkt_len = max_pkt_len;
488 	if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
489 	    rxq->max_pkt_len > buf_size) {
490 		dev_data->scattered_rx = 1;
491 	}
492 	IAVF_PCI_REG_WRITE(rxq->qrx_tail, rxq->nb_rx_desc - 1);
493 	IAVF_WRITE_FLUSH(hw);
494 
495 	return 0;
496 }
497 
498 static int
499 iavf_init_queues(struct rte_eth_dev *dev)
500 {
501 	struct iavf_rx_queue **rxq =
502 		(struct iavf_rx_queue **)dev->data->rx_queues;
503 	int i, ret = IAVF_SUCCESS;
504 
505 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
506 		if (!rxq[i] || !rxq[i]->q_set)
507 			continue;
508 		ret = iavf_init_rxq(dev, rxq[i]);
509 		if (ret != IAVF_SUCCESS)
510 			break;
511 	}
512 	/* set rx/tx function to vector/scatter/single-segment
513 	 * according to parameters
514 	 */
515 	iavf_set_rx_function(dev);
516 	iavf_set_tx_function(dev);
517 
518 	return ret;
519 }
520 
521 static int iavf_config_rx_queues_irqs(struct rte_eth_dev *dev,
522 				     struct rte_intr_handle *intr_handle)
523 {
524 	struct iavf_adapter *adapter =
525 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
526 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
527 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
528 	struct iavf_qv_map *qv_map;
529 	uint16_t interval, i;
530 	int vec;
531 
532 	if (rte_intr_cap_multiple(intr_handle) &&
533 	    dev->data->dev_conf.intr_conf.rxq) {
534 		if (rte_intr_efd_enable(intr_handle, dev->data->nb_rx_queues))
535 			return -1;
536 	}
537 
538 	if (rte_intr_dp_is_en(intr_handle) && !intr_handle->intr_vec) {
539 		intr_handle->intr_vec =
540 			rte_zmalloc("intr_vec",
541 				    dev->data->nb_rx_queues * sizeof(int), 0);
542 		if (!intr_handle->intr_vec) {
543 			PMD_DRV_LOG(ERR, "Failed to allocate %d rx intr_vec",
544 				    dev->data->nb_rx_queues);
545 			return -1;
546 		}
547 	}
548 
549 	qv_map = rte_zmalloc("qv_map",
550 		dev->data->nb_rx_queues * sizeof(struct iavf_qv_map), 0);
551 	if (!qv_map) {
552 		PMD_DRV_LOG(ERR, "Failed to allocate %d queue-vector map",
553 				dev->data->nb_rx_queues);
554 		return -1;
555 	}
556 
557 	if (!dev->data->dev_conf.intr_conf.rxq ||
558 	    !rte_intr_dp_is_en(intr_handle)) {
559 		/* Rx interrupt disabled, Map interrupt only for writeback */
560 		vf->nb_msix = 1;
561 		if (vf->vf_res->vf_cap_flags &
562 		    VIRTCHNL_VF_OFFLOAD_WB_ON_ITR) {
563 			/* If WB_ON_ITR supports, enable it */
564 			vf->msix_base = IAVF_RX_VEC_START;
565 			/* Set the ITR for index zero, to 2us to make sure that
566 			 * we leave time for aggregation to occur, but don't
567 			 * increase latency dramatically.
568 			 */
569 			IAVF_WRITE_REG(hw,
570 				       IAVF_VFINT_DYN_CTLN1(vf->msix_base - 1),
571 				       (0 << IAVF_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
572 				       IAVF_VFINT_DYN_CTLN1_WB_ON_ITR_MASK |
573 				       (2UL << IAVF_VFINT_DYN_CTLN1_INTERVAL_SHIFT));
574 			/* debug - check for success! the return value
575 			 * should be 2, offset is 0x2800
576 			 */
577 			/* IAVF_READ_REG(hw, IAVF_VFINT_ITRN1(0, 0)); */
578 		} else {
579 			/* If no WB_ON_ITR offload flags, need to set
580 			 * interrupt for descriptor write back.
581 			 */
582 			vf->msix_base = IAVF_MISC_VEC_ID;
583 
584 			/* set ITR to max */
585 			interval = iavf_calc_itr_interval(
586 					IAVF_QUEUE_ITR_INTERVAL_MAX);
587 			IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
588 				       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
589 				       (IAVF_ITR_INDEX_DEFAULT <<
590 					IAVF_VFINT_DYN_CTL01_ITR_INDX_SHIFT) |
591 				       (interval <<
592 					IAVF_VFINT_DYN_CTL01_INTERVAL_SHIFT));
593 		}
594 		IAVF_WRITE_FLUSH(hw);
595 		/* map all queues to the same interrupt */
596 		for (i = 0; i < dev->data->nb_rx_queues; i++) {
597 			qv_map[i].queue_id = i;
598 			qv_map[i].vector_id = vf->msix_base;
599 		}
600 		vf->qv_map = qv_map;
601 	} else {
602 		if (!rte_intr_allow_others(intr_handle)) {
603 			vf->nb_msix = 1;
604 			vf->msix_base = IAVF_MISC_VEC_ID;
605 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
606 				qv_map[i].queue_id = i;
607 				qv_map[i].vector_id = vf->msix_base;
608 				intr_handle->intr_vec[i] = IAVF_MISC_VEC_ID;
609 			}
610 			vf->qv_map = qv_map;
611 			PMD_DRV_LOG(DEBUG,
612 				    "vector %u are mapping to all Rx queues",
613 				    vf->msix_base);
614 		} else {
615 			/* If Rx interrupt is reuquired, and we can use
616 			 * multi interrupts, then the vec is from 1
617 			 */
618 			vf->nb_msix = RTE_MIN(intr_handle->nb_efd,
619 				 (uint16_t)(vf->vf_res->max_vectors - 1));
620 			vf->msix_base = IAVF_RX_VEC_START;
621 			vec = IAVF_RX_VEC_START;
622 			for (i = 0; i < dev->data->nb_rx_queues; i++) {
623 				qv_map[i].queue_id = i;
624 				qv_map[i].vector_id = vec;
625 				intr_handle->intr_vec[i] = vec++;
626 				if (vec >= vf->nb_msix + IAVF_RX_VEC_START)
627 					vec = IAVF_RX_VEC_START;
628 			}
629 			vf->qv_map = qv_map;
630 			PMD_DRV_LOG(DEBUG,
631 				    "%u vectors are mapping to %u Rx queues",
632 				    vf->nb_msix, dev->data->nb_rx_queues);
633 		}
634 	}
635 
636 	if (!vf->lv_enabled) {
637 		if (iavf_config_irq_map(adapter)) {
638 			PMD_DRV_LOG(ERR, "config interrupt mapping failed");
639 			return -1;
640 		}
641 	} else {
642 		uint16_t num_qv_maps = dev->data->nb_rx_queues;
643 		uint16_t index = 0;
644 
645 		while (num_qv_maps > IAVF_IRQ_MAP_NUM_PER_BUF) {
646 			if (iavf_config_irq_map_lv(adapter,
647 					IAVF_IRQ_MAP_NUM_PER_BUF, index)) {
648 				PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
649 				return -1;
650 			}
651 			num_qv_maps -= IAVF_IRQ_MAP_NUM_PER_BUF;
652 			index += IAVF_IRQ_MAP_NUM_PER_BUF;
653 		}
654 
655 		if (iavf_config_irq_map_lv(adapter, num_qv_maps, index)) {
656 			PMD_DRV_LOG(ERR, "config interrupt mapping for large VF failed");
657 			return -1;
658 		}
659 	}
660 	return 0;
661 }
662 
663 static int
664 iavf_start_queues(struct rte_eth_dev *dev)
665 {
666 	struct iavf_rx_queue *rxq;
667 	struct iavf_tx_queue *txq;
668 	int i;
669 
670 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
671 		txq = dev->data->tx_queues[i];
672 		if (txq->tx_deferred_start)
673 			continue;
674 		if (iavf_dev_tx_queue_start(dev, i) != 0) {
675 			PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
676 			return -1;
677 		}
678 	}
679 
680 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
681 		rxq = dev->data->rx_queues[i];
682 		if (rxq->rx_deferred_start)
683 			continue;
684 		if (iavf_dev_rx_queue_start(dev, i) != 0) {
685 			PMD_DRV_LOG(ERR, "Fail to start queue %u", i);
686 			return -1;
687 		}
688 	}
689 
690 	return 0;
691 }
692 
693 static int
694 iavf_dev_start(struct rte_eth_dev *dev)
695 {
696 	struct iavf_adapter *adapter =
697 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
698 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
699 	struct rte_intr_handle *intr_handle = dev->intr_handle;
700 	uint16_t num_queue_pairs;
701 	uint16_t index = 0;
702 
703 	PMD_INIT_FUNC_TRACE();
704 
705 	adapter->stopped = 0;
706 
707 	vf->max_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
708 	vf->num_queue_pairs = RTE_MAX(dev->data->nb_rx_queues,
709 				      dev->data->nb_tx_queues);
710 	num_queue_pairs = vf->num_queue_pairs;
711 
712 	if (iavf_init_queues(dev) != 0) {
713 		PMD_DRV_LOG(ERR, "failed to do Queue init");
714 		return -1;
715 	}
716 
717 	/* If needed, send configure queues msg multiple times to make the
718 	 * adminq buffer length smaller than the 4K limitation.
719 	 */
720 	while (num_queue_pairs > IAVF_CFG_Q_NUM_PER_BUF) {
721 		if (iavf_configure_queues(adapter,
722 				IAVF_CFG_Q_NUM_PER_BUF, index) != 0) {
723 			PMD_DRV_LOG(ERR, "configure queues failed");
724 			goto err_queue;
725 		}
726 		num_queue_pairs -= IAVF_CFG_Q_NUM_PER_BUF;
727 		index += IAVF_CFG_Q_NUM_PER_BUF;
728 	}
729 
730 	if (iavf_configure_queues(adapter, num_queue_pairs, index) != 0) {
731 		PMD_DRV_LOG(ERR, "configure queues failed");
732 		goto err_queue;
733 	}
734 
735 	if (iavf_config_rx_queues_irqs(dev, intr_handle) != 0) {
736 		PMD_DRV_LOG(ERR, "configure irq failed");
737 		goto err_queue;
738 	}
739 	/* re-enable intr again, because efd assign may change */
740 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
741 		rte_intr_disable(intr_handle);
742 		rte_intr_enable(intr_handle);
743 	}
744 
745 	/* Set all mac addrs */
746 	iavf_add_del_all_mac_addr(adapter, true);
747 
748 	/* Set all multicast addresses */
749 	iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
750 				  true);
751 
752 	if (iavf_start_queues(dev) != 0) {
753 		PMD_DRV_LOG(ERR, "enable queues failed");
754 		goto err_mac;
755 	}
756 
757 	return 0;
758 
759 err_mac:
760 	iavf_add_del_all_mac_addr(adapter, false);
761 err_queue:
762 	return -1;
763 }
764 
765 static int
766 iavf_dev_stop(struct rte_eth_dev *dev)
767 {
768 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
769 	struct iavf_adapter *adapter =
770 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
771 	struct rte_intr_handle *intr_handle = dev->intr_handle;
772 
773 	PMD_INIT_FUNC_TRACE();
774 
775 	if (adapter->stopped == 1)
776 		return 0;
777 
778 	iavf_stop_queues(dev);
779 
780 	/* Disable the interrupt for Rx */
781 	rte_intr_efd_disable(intr_handle);
782 	/* Rx interrupt vector mapping free */
783 	if (intr_handle->intr_vec) {
784 		rte_free(intr_handle->intr_vec);
785 		intr_handle->intr_vec = NULL;
786 	}
787 
788 	/* remove all mac addrs */
789 	iavf_add_del_all_mac_addr(adapter, false);
790 
791 	/* remove all multicast addresses */
792 	iavf_add_del_mc_addr_list(adapter, vf->mc_addrs, vf->mc_addrs_num,
793 				  false);
794 
795 	adapter->stopped = 1;
796 	dev->data->dev_started = 0;
797 
798 	return 0;
799 }
800 
801 static int
802 iavf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
803 {
804 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
805 
806 	dev_info->max_rx_queues = IAVF_MAX_NUM_QUEUES_LV;
807 	dev_info->max_tx_queues = IAVF_MAX_NUM_QUEUES_LV;
808 	dev_info->min_rx_bufsize = IAVF_BUF_SIZE_MIN;
809 	dev_info->max_rx_pktlen = IAVF_FRAME_SIZE_MAX;
810 	dev_info->max_mtu = dev_info->max_rx_pktlen - IAVF_ETH_OVERHEAD;
811 	dev_info->min_mtu = RTE_ETHER_MIN_MTU;
812 	dev_info->hash_key_size = vf->vf_res->rss_key_size;
813 	dev_info->reta_size = vf->vf_res->rss_lut_size;
814 	dev_info->flow_type_rss_offloads = IAVF_RSS_OFFLOAD_ALL;
815 	dev_info->max_mac_addrs = IAVF_NUM_MACADDR_MAX;
816 	dev_info->rx_offload_capa =
817 		DEV_RX_OFFLOAD_VLAN_STRIP |
818 		DEV_RX_OFFLOAD_QINQ_STRIP |
819 		DEV_RX_OFFLOAD_IPV4_CKSUM |
820 		DEV_RX_OFFLOAD_UDP_CKSUM |
821 		DEV_RX_OFFLOAD_TCP_CKSUM |
822 		DEV_RX_OFFLOAD_OUTER_IPV4_CKSUM |
823 		DEV_RX_OFFLOAD_SCATTER |
824 		DEV_RX_OFFLOAD_JUMBO_FRAME |
825 		DEV_RX_OFFLOAD_VLAN_FILTER |
826 		DEV_RX_OFFLOAD_RSS_HASH;
827 
828 	dev_info->tx_offload_capa =
829 		DEV_TX_OFFLOAD_VLAN_INSERT |
830 		DEV_TX_OFFLOAD_QINQ_INSERT |
831 		DEV_TX_OFFLOAD_IPV4_CKSUM |
832 		DEV_TX_OFFLOAD_UDP_CKSUM |
833 		DEV_TX_OFFLOAD_TCP_CKSUM |
834 		DEV_TX_OFFLOAD_SCTP_CKSUM |
835 		DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM |
836 		DEV_TX_OFFLOAD_TCP_TSO |
837 		DEV_TX_OFFLOAD_VXLAN_TNL_TSO |
838 		DEV_TX_OFFLOAD_GRE_TNL_TSO |
839 		DEV_TX_OFFLOAD_IPIP_TNL_TSO |
840 		DEV_TX_OFFLOAD_GENEVE_TNL_TSO |
841 		DEV_TX_OFFLOAD_MULTI_SEGS |
842 		DEV_TX_OFFLOAD_MBUF_FAST_FREE;
843 
844 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_CRC)
845 		dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_KEEP_CRC;
846 
847 	dev_info->default_rxconf = (struct rte_eth_rxconf) {
848 		.rx_free_thresh = IAVF_DEFAULT_RX_FREE_THRESH,
849 		.rx_drop_en = 0,
850 		.offloads = 0,
851 	};
852 
853 	dev_info->default_txconf = (struct rte_eth_txconf) {
854 		.tx_free_thresh = IAVF_DEFAULT_TX_FREE_THRESH,
855 		.tx_rs_thresh = IAVF_DEFAULT_TX_RS_THRESH,
856 		.offloads = 0,
857 	};
858 
859 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
860 		.nb_max = IAVF_MAX_RING_DESC,
861 		.nb_min = IAVF_MIN_RING_DESC,
862 		.nb_align = IAVF_ALIGN_RING_DESC,
863 	};
864 
865 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
866 		.nb_max = IAVF_MAX_RING_DESC,
867 		.nb_min = IAVF_MIN_RING_DESC,
868 		.nb_align = IAVF_ALIGN_RING_DESC,
869 	};
870 
871 	return 0;
872 }
873 
874 static const uint32_t *
875 iavf_dev_supported_ptypes_get(struct rte_eth_dev *dev __rte_unused)
876 {
877 	static const uint32_t ptypes[] = {
878 		RTE_PTYPE_L2_ETHER,
879 		RTE_PTYPE_L3_IPV4_EXT_UNKNOWN,
880 		RTE_PTYPE_L4_FRAG,
881 		RTE_PTYPE_L4_ICMP,
882 		RTE_PTYPE_L4_NONFRAG,
883 		RTE_PTYPE_L4_SCTP,
884 		RTE_PTYPE_L4_TCP,
885 		RTE_PTYPE_L4_UDP,
886 		RTE_PTYPE_UNKNOWN
887 	};
888 	return ptypes;
889 }
890 
891 int
892 iavf_dev_link_update(struct rte_eth_dev *dev,
893 		    __rte_unused int wait_to_complete)
894 {
895 	struct rte_eth_link new_link;
896 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
897 
898 	memset(&new_link, 0, sizeof(new_link));
899 
900 	/* Only read status info stored in VF, and the info is updated
901 	 *  when receive LINK_CHANGE evnet from PF by Virtchnnl.
902 	 */
903 	switch (vf->link_speed) {
904 	case 10:
905 		new_link.link_speed = ETH_SPEED_NUM_10M;
906 		break;
907 	case 100:
908 		new_link.link_speed = ETH_SPEED_NUM_100M;
909 		break;
910 	case 1000:
911 		new_link.link_speed = ETH_SPEED_NUM_1G;
912 		break;
913 	case 10000:
914 		new_link.link_speed = ETH_SPEED_NUM_10G;
915 		break;
916 	case 20000:
917 		new_link.link_speed = ETH_SPEED_NUM_20G;
918 		break;
919 	case 25000:
920 		new_link.link_speed = ETH_SPEED_NUM_25G;
921 		break;
922 	case 40000:
923 		new_link.link_speed = ETH_SPEED_NUM_40G;
924 		break;
925 	case 50000:
926 		new_link.link_speed = ETH_SPEED_NUM_50G;
927 		break;
928 	case 100000:
929 		new_link.link_speed = ETH_SPEED_NUM_100G;
930 		break;
931 	default:
932 		new_link.link_speed = ETH_SPEED_NUM_NONE;
933 		break;
934 	}
935 
936 	new_link.link_duplex = ETH_LINK_FULL_DUPLEX;
937 	new_link.link_status = vf->link_up ? ETH_LINK_UP :
938 					     ETH_LINK_DOWN;
939 	new_link.link_autoneg = !(dev->data->dev_conf.link_speeds &
940 				ETH_LINK_SPEED_FIXED);
941 
942 	return rte_eth_linkstatus_set(dev, &new_link);
943 }
944 
945 static int
946 iavf_dev_promiscuous_enable(struct rte_eth_dev *dev)
947 {
948 	struct iavf_adapter *adapter =
949 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
950 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
951 
952 	return iavf_config_promisc(adapter,
953 				  true, vf->promisc_multicast_enabled);
954 }
955 
956 static int
957 iavf_dev_promiscuous_disable(struct rte_eth_dev *dev)
958 {
959 	struct iavf_adapter *adapter =
960 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
961 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
962 
963 	return iavf_config_promisc(adapter,
964 				  false, vf->promisc_multicast_enabled);
965 }
966 
967 static int
968 iavf_dev_allmulticast_enable(struct rte_eth_dev *dev)
969 {
970 	struct iavf_adapter *adapter =
971 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
972 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
973 
974 	return iavf_config_promisc(adapter,
975 				  vf->promisc_unicast_enabled, true);
976 }
977 
978 static int
979 iavf_dev_allmulticast_disable(struct rte_eth_dev *dev)
980 {
981 	struct iavf_adapter *adapter =
982 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
983 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
984 
985 	return iavf_config_promisc(adapter,
986 				  vf->promisc_unicast_enabled, false);
987 }
988 
989 static int
990 iavf_dev_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *addr,
991 		     __rte_unused uint32_t index,
992 		     __rte_unused uint32_t pool)
993 {
994 	struct iavf_adapter *adapter =
995 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
996 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
997 	int err;
998 
999 	if (rte_is_zero_ether_addr(addr)) {
1000 		PMD_DRV_LOG(ERR, "Invalid Ethernet Address");
1001 		return -EINVAL;
1002 	}
1003 
1004 	err = iavf_add_del_eth_addr(adapter, addr, true, VIRTCHNL_ETHER_ADDR_EXTRA);
1005 	if (err) {
1006 		PMD_DRV_LOG(ERR, "fail to add MAC address");
1007 		return -EIO;
1008 	}
1009 
1010 	vf->mac_num++;
1011 
1012 	return 0;
1013 }
1014 
1015 static void
1016 iavf_dev_del_mac_addr(struct rte_eth_dev *dev, uint32_t index)
1017 {
1018 	struct iavf_adapter *adapter =
1019 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1020 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1021 	struct rte_ether_addr *addr;
1022 	int err;
1023 
1024 	addr = &dev->data->mac_addrs[index];
1025 
1026 	err = iavf_add_del_eth_addr(adapter, addr, false, VIRTCHNL_ETHER_ADDR_EXTRA);
1027 	if (err)
1028 		PMD_DRV_LOG(ERR, "fail to delete MAC address");
1029 
1030 	vf->mac_num--;
1031 }
1032 
1033 static int
1034 iavf_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1035 {
1036 	struct iavf_adapter *adapter =
1037 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1038 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1039 	int err;
1040 
1041 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
1042 		err = iavf_add_del_vlan_v2(adapter, vlan_id, on);
1043 		if (err)
1044 			return -EIO;
1045 		return 0;
1046 	}
1047 
1048 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1049 		return -ENOTSUP;
1050 
1051 	err = iavf_add_del_vlan(adapter, vlan_id, on);
1052 	if (err)
1053 		return -EIO;
1054 	return 0;
1055 }
1056 
1057 static void
1058 iavf_iterate_vlan_filters_v2(struct rte_eth_dev *dev, bool enable)
1059 {
1060 	struct rte_vlan_filter_conf *vfc = &dev->data->vlan_filter_conf;
1061 	struct iavf_adapter *adapter =
1062 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1063 	uint32_t i, j;
1064 	uint64_t ids;
1065 
1066 	for (i = 0; i < RTE_DIM(vfc->ids); i++) {
1067 		if (vfc->ids[i] == 0)
1068 			continue;
1069 
1070 		ids = vfc->ids[i];
1071 		for (j = 0; ids != 0 && j < 64; j++, ids >>= 1) {
1072 			if (ids & 1)
1073 				iavf_add_del_vlan_v2(adapter,
1074 						     64 * i + j, enable);
1075 		}
1076 	}
1077 }
1078 
1079 static int
1080 iavf_dev_vlan_offload_set_v2(struct rte_eth_dev *dev, int mask)
1081 {
1082 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
1083 	struct iavf_adapter *adapter =
1084 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1085 	bool enable;
1086 	int err;
1087 
1088 	if (mask & ETH_VLAN_FILTER_MASK) {
1089 		enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER);
1090 
1091 		iavf_iterate_vlan_filters_v2(dev, enable);
1092 	}
1093 
1094 	if (mask & ETH_VLAN_STRIP_MASK) {
1095 		enable = !!(rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP);
1096 
1097 		err = iavf_config_vlan_strip_v2(adapter, enable);
1098 		/* If not support, the stripping is already disabled by PF */
1099 		if (err == -ENOTSUP && !enable)
1100 			err = 0;
1101 		if (err)
1102 			return -EIO;
1103 	}
1104 
1105 	return 0;
1106 }
1107 
1108 static int
1109 iavf_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1110 {
1111 	struct iavf_adapter *adapter =
1112 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1113 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1114 	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
1115 	int err;
1116 
1117 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2)
1118 		return iavf_dev_vlan_offload_set_v2(dev, mask);
1119 
1120 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN))
1121 		return -ENOTSUP;
1122 
1123 	/* Vlan stripping setting */
1124 	if (mask & ETH_VLAN_STRIP_MASK) {
1125 		/* Enable or disable VLAN stripping */
1126 		if (dev_conf->rxmode.offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1127 			err = iavf_enable_vlan_strip(adapter);
1128 		else
1129 			err = iavf_disable_vlan_strip(adapter);
1130 
1131 		if (err)
1132 			return -EIO;
1133 	}
1134 	return 0;
1135 }
1136 
1137 static int
1138 iavf_dev_rss_reta_update(struct rte_eth_dev *dev,
1139 			struct rte_eth_rss_reta_entry64 *reta_conf,
1140 			uint16_t reta_size)
1141 {
1142 	struct iavf_adapter *adapter =
1143 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1144 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1145 	uint8_t *lut;
1146 	uint16_t i, idx, shift;
1147 	int ret;
1148 
1149 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1150 		return -ENOTSUP;
1151 
1152 	if (reta_size != vf->vf_res->rss_lut_size) {
1153 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1154 			"(%d) doesn't match the number of hardware can "
1155 			"support (%d)", reta_size, vf->vf_res->rss_lut_size);
1156 		return -EINVAL;
1157 	}
1158 
1159 	lut = rte_zmalloc("rss_lut", reta_size, 0);
1160 	if (!lut) {
1161 		PMD_DRV_LOG(ERR, "No memory can be allocated");
1162 		return -ENOMEM;
1163 	}
1164 	/* store the old lut table temporarily */
1165 	rte_memcpy(lut, vf->rss_lut, reta_size);
1166 
1167 	for (i = 0; i < reta_size; i++) {
1168 		idx = i / RTE_RETA_GROUP_SIZE;
1169 		shift = i % RTE_RETA_GROUP_SIZE;
1170 		if (reta_conf[idx].mask & (1ULL << shift))
1171 			lut[i] = reta_conf[idx].reta[shift];
1172 	}
1173 
1174 	rte_memcpy(vf->rss_lut, lut, reta_size);
1175 	/* send virtchnnl ops to configure rss*/
1176 	ret = iavf_configure_rss_lut(adapter);
1177 	if (ret) /* revert back */
1178 		rte_memcpy(vf->rss_lut, lut, reta_size);
1179 	rte_free(lut);
1180 
1181 	return ret;
1182 }
1183 
1184 static int
1185 iavf_dev_rss_reta_query(struct rte_eth_dev *dev,
1186 		       struct rte_eth_rss_reta_entry64 *reta_conf,
1187 		       uint16_t reta_size)
1188 {
1189 	struct iavf_adapter *adapter =
1190 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1191 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1192 	uint16_t i, idx, shift;
1193 
1194 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1195 		return -ENOTSUP;
1196 
1197 	if (reta_size != vf->vf_res->rss_lut_size) {
1198 		PMD_DRV_LOG(ERR, "The size of hash lookup table configured "
1199 			"(%d) doesn't match the number of hardware can "
1200 			"support (%d)", reta_size, vf->vf_res->rss_lut_size);
1201 		return -EINVAL;
1202 	}
1203 
1204 	for (i = 0; i < reta_size; i++) {
1205 		idx = i / RTE_RETA_GROUP_SIZE;
1206 		shift = i % RTE_RETA_GROUP_SIZE;
1207 		if (reta_conf[idx].mask & (1ULL << shift))
1208 			reta_conf[idx].reta[shift] = vf->rss_lut[i];
1209 	}
1210 
1211 	return 0;
1212 }
1213 
1214 static int
1215 iavf_set_rss_key(struct iavf_adapter *adapter, uint8_t *key, uint8_t key_len)
1216 {
1217 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1218 
1219 	/* HENA setting, it is enabled by default, no change */
1220 	if (!key || key_len == 0) {
1221 		PMD_DRV_LOG(DEBUG, "No key to be configured");
1222 		return 0;
1223 	} else if (key_len != vf->vf_res->rss_key_size) {
1224 		PMD_DRV_LOG(ERR, "The size of hash key configured "
1225 			"(%d) doesn't match the size of hardware can "
1226 			"support (%d)", key_len,
1227 			vf->vf_res->rss_key_size);
1228 		return -EINVAL;
1229 	}
1230 
1231 	rte_memcpy(vf->rss_key, key, key_len);
1232 
1233 	return iavf_configure_rss_key(adapter);
1234 }
1235 
1236 static int
1237 iavf_dev_rss_hash_update(struct rte_eth_dev *dev,
1238 			struct rte_eth_rss_conf *rss_conf)
1239 {
1240 	struct iavf_adapter *adapter =
1241 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1242 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1243 	int ret;
1244 
1245 	adapter->eth_dev->data->dev_conf.rx_adv_conf.rss_conf = *rss_conf;
1246 
1247 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1248 		return -ENOTSUP;
1249 
1250 	/* Set hash key. */
1251 	ret = iavf_set_rss_key(adapter, rss_conf->rss_key,
1252 			       rss_conf->rss_key_len);
1253 	if (ret)
1254 		return ret;
1255 
1256 	if (rss_conf->rss_hf == 0) {
1257 		vf->rss_hf = 0;
1258 		ret = iavf_set_hena(adapter, 0);
1259 
1260 		/* It is a workaround, temporarily allow error to be returned
1261 		 * due to possible lack of PF handling for hena = 0.
1262 		 */
1263 		if (ret)
1264 			PMD_DRV_LOG(WARNING, "fail to clean existing RSS, lack PF support");
1265 		return 0;
1266 	}
1267 
1268 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
1269 		/* Clear existing RSS. */
1270 		ret = iavf_set_hena(adapter, 0);
1271 
1272 		/* It is a workaround, temporarily allow error to be returned
1273 		 * due to possible lack of PF handling for hena = 0.
1274 		 */
1275 		if (ret)
1276 			PMD_DRV_LOG(WARNING, "fail to clean existing RSS,"
1277 				    "lack PF support");
1278 
1279 		/* Set new RSS configuration. */
1280 		ret = iavf_rss_hash_set(adapter, rss_conf->rss_hf, true);
1281 		if (ret) {
1282 			PMD_DRV_LOG(ERR, "fail to set new RSS");
1283 			return ret;
1284 		}
1285 	}
1286 
1287 	return 0;
1288 }
1289 
1290 static int
1291 iavf_dev_rss_hash_conf_get(struct rte_eth_dev *dev,
1292 			  struct rte_eth_rss_conf *rss_conf)
1293 {
1294 	struct iavf_adapter *adapter =
1295 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1296 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
1297 
1298 	if (!(vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF))
1299 		return -ENOTSUP;
1300 
1301 	rss_conf->rss_hf = vf->rss_hf;
1302 
1303 	if (!rss_conf->rss_key)
1304 		return 0;
1305 
1306 	rss_conf->rss_key_len = vf->vf_res->rss_key_size;
1307 	rte_memcpy(rss_conf->rss_key, vf->rss_key, rss_conf->rss_key_len);
1308 
1309 	return 0;
1310 }
1311 
1312 static int
1313 iavf_dev_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1314 {
1315 	uint32_t frame_size = mtu + IAVF_ETH_OVERHEAD;
1316 	int ret = 0;
1317 
1318 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > IAVF_FRAME_SIZE_MAX)
1319 		return -EINVAL;
1320 
1321 	/* mtu setting is forbidden if port is start */
1322 	if (dev->data->dev_started) {
1323 		PMD_DRV_LOG(ERR, "port must be stopped before configuration");
1324 		return -EBUSY;
1325 	}
1326 
1327 	if (frame_size > IAVF_ETH_MAX_LEN)
1328 		dev->data->dev_conf.rxmode.offloads |=
1329 				DEV_RX_OFFLOAD_JUMBO_FRAME;
1330 	else
1331 		dev->data->dev_conf.rxmode.offloads &=
1332 				~DEV_RX_OFFLOAD_JUMBO_FRAME;
1333 
1334 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1335 
1336 	return ret;
1337 }
1338 
1339 static int
1340 iavf_dev_set_default_mac_addr(struct rte_eth_dev *dev,
1341 			     struct rte_ether_addr *mac_addr)
1342 {
1343 	struct iavf_adapter *adapter =
1344 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1345 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1346 	struct rte_ether_addr *old_addr;
1347 	int ret;
1348 
1349 	old_addr = (struct rte_ether_addr *)hw->mac.addr;
1350 
1351 	if (rte_is_same_ether_addr(old_addr, mac_addr))
1352 		return 0;
1353 
1354 	ret = iavf_add_del_eth_addr(adapter, old_addr, false, VIRTCHNL_ETHER_ADDR_PRIMARY);
1355 	if (ret)
1356 		PMD_DRV_LOG(ERR, "Fail to delete old MAC:"
1357 			    " %02X:%02X:%02X:%02X:%02X:%02X",
1358 			    old_addr->addr_bytes[0],
1359 			    old_addr->addr_bytes[1],
1360 			    old_addr->addr_bytes[2],
1361 			    old_addr->addr_bytes[3],
1362 			    old_addr->addr_bytes[4],
1363 			    old_addr->addr_bytes[5]);
1364 
1365 	ret = iavf_add_del_eth_addr(adapter, mac_addr, true, VIRTCHNL_ETHER_ADDR_PRIMARY);
1366 	if (ret)
1367 		PMD_DRV_LOG(ERR, "Fail to add new MAC:"
1368 			    " %02X:%02X:%02X:%02X:%02X:%02X",
1369 			    mac_addr->addr_bytes[0],
1370 			    mac_addr->addr_bytes[1],
1371 			    mac_addr->addr_bytes[2],
1372 			    mac_addr->addr_bytes[3],
1373 			    mac_addr->addr_bytes[4],
1374 			    mac_addr->addr_bytes[5]);
1375 
1376 	if (ret)
1377 		return -EIO;
1378 
1379 	rte_ether_addr_copy(mac_addr, (struct rte_ether_addr *)hw->mac.addr);
1380 	return 0;
1381 }
1382 
1383 static void
1384 iavf_stat_update_48(uint64_t *offset, uint64_t *stat)
1385 {
1386 	if (*stat >= *offset)
1387 		*stat = *stat - *offset;
1388 	else
1389 		*stat = (uint64_t)((*stat +
1390 			((uint64_t)1 << IAVF_48_BIT_WIDTH)) - *offset);
1391 
1392 	*stat &= IAVF_48_BIT_MASK;
1393 }
1394 
1395 static void
1396 iavf_stat_update_32(uint64_t *offset, uint64_t *stat)
1397 {
1398 	if (*stat >= *offset)
1399 		*stat = (uint64_t)(*stat - *offset);
1400 	else
1401 		*stat = (uint64_t)((*stat +
1402 			((uint64_t)1 << IAVF_32_BIT_WIDTH)) - *offset);
1403 }
1404 
1405 static void
1406 iavf_update_stats(struct iavf_vsi *vsi, struct virtchnl_eth_stats *nes)
1407 {
1408 	struct virtchnl_eth_stats *oes = &vsi->eth_stats_offset;
1409 
1410 	iavf_stat_update_48(&oes->rx_bytes, &nes->rx_bytes);
1411 	iavf_stat_update_48(&oes->rx_unicast, &nes->rx_unicast);
1412 	iavf_stat_update_48(&oes->rx_multicast, &nes->rx_multicast);
1413 	iavf_stat_update_48(&oes->rx_broadcast, &nes->rx_broadcast);
1414 	iavf_stat_update_32(&oes->rx_discards, &nes->rx_discards);
1415 	iavf_stat_update_48(&oes->tx_bytes, &nes->tx_bytes);
1416 	iavf_stat_update_48(&oes->tx_unicast, &nes->tx_unicast);
1417 	iavf_stat_update_48(&oes->tx_multicast, &nes->tx_multicast);
1418 	iavf_stat_update_48(&oes->tx_broadcast, &nes->tx_broadcast);
1419 	iavf_stat_update_32(&oes->tx_errors, &nes->tx_errors);
1420 	iavf_stat_update_32(&oes->tx_discards, &nes->tx_discards);
1421 }
1422 
1423 static int
1424 iavf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1425 {
1426 	struct iavf_adapter *adapter =
1427 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1428 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1429 	struct iavf_vsi *vsi = &vf->vsi;
1430 	struct virtchnl_eth_stats *pstats = NULL;
1431 	int ret;
1432 
1433 	ret = iavf_query_stats(adapter, &pstats);
1434 	if (ret == 0) {
1435 		uint8_t crc_stats_len = (dev->data->dev_conf.rxmode.offloads &
1436 					 DEV_RX_OFFLOAD_KEEP_CRC) ? 0 :
1437 					 RTE_ETHER_CRC_LEN;
1438 		iavf_update_stats(vsi, pstats);
1439 		stats->ipackets = pstats->rx_unicast + pstats->rx_multicast +
1440 				pstats->rx_broadcast - pstats->rx_discards;
1441 		stats->opackets = pstats->tx_broadcast + pstats->tx_multicast +
1442 						pstats->tx_unicast;
1443 		stats->imissed = pstats->rx_discards;
1444 		stats->oerrors = pstats->tx_errors + pstats->tx_discards;
1445 		stats->ibytes = pstats->rx_bytes;
1446 		stats->ibytes -= stats->ipackets * crc_stats_len;
1447 		stats->obytes = pstats->tx_bytes;
1448 	} else {
1449 		PMD_DRV_LOG(ERR, "Get statistics failed");
1450 	}
1451 	return ret;
1452 }
1453 
1454 static int
1455 iavf_dev_stats_reset(struct rte_eth_dev *dev)
1456 {
1457 	int ret;
1458 	struct iavf_adapter *adapter =
1459 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1460 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1461 	struct iavf_vsi *vsi = &vf->vsi;
1462 	struct virtchnl_eth_stats *pstats = NULL;
1463 
1464 	/* read stat values to clear hardware registers */
1465 	ret = iavf_query_stats(adapter, &pstats);
1466 	if (ret != 0)
1467 		return ret;
1468 
1469 	/* set stats offset base on current values */
1470 	vsi->eth_stats_offset = *pstats;
1471 
1472 	return 0;
1473 }
1474 
1475 static int iavf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
1476 				      struct rte_eth_xstat_name *xstats_names,
1477 				      __rte_unused unsigned int limit)
1478 {
1479 	unsigned int i;
1480 
1481 	if (xstats_names != NULL)
1482 		for (i = 0; i < IAVF_NB_XSTATS; i++) {
1483 			snprintf(xstats_names[i].name,
1484 				sizeof(xstats_names[i].name),
1485 				"%s", rte_iavf_stats_strings[i].name);
1486 		}
1487 	return IAVF_NB_XSTATS;
1488 }
1489 
1490 static int iavf_dev_xstats_get(struct rte_eth_dev *dev,
1491 				 struct rte_eth_xstat *xstats, unsigned int n)
1492 {
1493 	int ret;
1494 	unsigned int i;
1495 	struct iavf_adapter *adapter =
1496 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1497 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1498 	struct iavf_vsi *vsi = &vf->vsi;
1499 	struct virtchnl_eth_stats *pstats = NULL;
1500 
1501 	if (n < IAVF_NB_XSTATS)
1502 		return IAVF_NB_XSTATS;
1503 
1504 	ret = iavf_query_stats(adapter, &pstats);
1505 	if (ret != 0)
1506 		return 0;
1507 
1508 	if (!xstats)
1509 		return 0;
1510 
1511 	iavf_update_stats(vsi, pstats);
1512 
1513 	/* loop over xstats array and values from pstats */
1514 	for (i = 0; i < IAVF_NB_XSTATS; i++) {
1515 		xstats[i].id = i;
1516 		xstats[i].value = *(uint64_t *)(((char *)pstats) +
1517 			rte_iavf_stats_strings[i].offset);
1518 	}
1519 
1520 	return IAVF_NB_XSTATS;
1521 }
1522 
1523 
1524 static int
1525 iavf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
1526 {
1527 	struct iavf_adapter *adapter =
1528 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1529 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1530 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
1531 	uint16_t msix_intr;
1532 
1533 	msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1534 	if (msix_intr == IAVF_MISC_VEC_ID) {
1535 		PMD_DRV_LOG(INFO, "MISC is also enabled for control");
1536 		IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
1537 			       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
1538 			       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1539 			       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
1540 	} else {
1541 		IAVF_WRITE_REG(hw,
1542 			       IAVF_VFINT_DYN_CTLN1
1543 				(msix_intr - IAVF_RX_VEC_START),
1544 			       IAVF_VFINT_DYN_CTLN1_INTENA_MASK |
1545 			       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
1546 			       IAVF_VFINT_DYN_CTLN1_ITR_INDX_MASK);
1547 	}
1548 
1549 	IAVF_WRITE_FLUSH(hw);
1550 
1551 	rte_intr_ack(&pci_dev->intr_handle);
1552 
1553 	return 0;
1554 }
1555 
1556 static int
1557 iavf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
1558 {
1559 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1560 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1561 	uint16_t msix_intr;
1562 
1563 	msix_intr = pci_dev->intr_handle.intr_vec[queue_id];
1564 	if (msix_intr == IAVF_MISC_VEC_ID) {
1565 		PMD_DRV_LOG(ERR, "MISC is used for control, cannot disable it");
1566 		return -EIO;
1567 	}
1568 
1569 	IAVF_WRITE_REG(hw,
1570 		      IAVF_VFINT_DYN_CTLN1(msix_intr - IAVF_RX_VEC_START),
1571 		      0);
1572 
1573 	IAVF_WRITE_FLUSH(hw);
1574 	return 0;
1575 }
1576 
1577 static int
1578 iavf_check_vf_reset_done(struct iavf_hw *hw)
1579 {
1580 	int i, reset;
1581 
1582 	for (i = 0; i < IAVF_RESET_WAIT_CNT; i++) {
1583 		reset = IAVF_READ_REG(hw, IAVF_VFGEN_RSTAT) &
1584 			IAVF_VFGEN_RSTAT_VFR_STATE_MASK;
1585 		reset = reset >> IAVF_VFGEN_RSTAT_VFR_STATE_SHIFT;
1586 		if (reset == VIRTCHNL_VFR_VFACTIVE ||
1587 		    reset == VIRTCHNL_VFR_COMPLETED)
1588 			break;
1589 		rte_delay_ms(20);
1590 	}
1591 
1592 	if (i >= IAVF_RESET_WAIT_CNT)
1593 		return -1;
1594 
1595 	return 0;
1596 }
1597 
1598 static int
1599 iavf_lookup_proto_xtr_type(const char *flex_name)
1600 {
1601 	static struct {
1602 		const char *name;
1603 		enum iavf_proto_xtr_type type;
1604 	} xtr_type_map[] = {
1605 		{ "vlan",      IAVF_PROTO_XTR_VLAN      },
1606 		{ "ipv4",      IAVF_PROTO_XTR_IPV4      },
1607 		{ "ipv6",      IAVF_PROTO_XTR_IPV6      },
1608 		{ "ipv6_flow", IAVF_PROTO_XTR_IPV6_FLOW },
1609 		{ "tcp",       IAVF_PROTO_XTR_TCP       },
1610 		{ "ip_offset", IAVF_PROTO_XTR_IP_OFFSET },
1611 	};
1612 	uint32_t i;
1613 
1614 	for (i = 0; i < RTE_DIM(xtr_type_map); i++) {
1615 		if (strcmp(flex_name, xtr_type_map[i].name) == 0)
1616 			return xtr_type_map[i].type;
1617 	}
1618 
1619 	PMD_DRV_LOG(ERR, "wrong proto_xtr type, "
1620 		    "it should be: vlan|ipv4|ipv6|ipv6_flow|tcp|ip_offset");
1621 
1622 	return -1;
1623 }
1624 
1625 /**
1626  * Parse elem, the elem could be single number/range or '(' ')' group
1627  * 1) A single number elem, it's just a simple digit. e.g. 9
1628  * 2) A single range elem, two digits with a '-' between. e.g. 2-6
1629  * 3) A group elem, combines multiple 1) or 2) with '( )'. e.g (0,2-4,6)
1630  *    Within group elem, '-' used for a range separator;
1631  *                       ',' used for a single number.
1632  */
1633 static int
1634 iavf_parse_queue_set(const char *input, int xtr_type,
1635 		     struct iavf_devargs *devargs)
1636 {
1637 	const char *str = input;
1638 	char *end = NULL;
1639 	uint32_t min, max;
1640 	uint32_t idx;
1641 
1642 	while (isblank(*str))
1643 		str++;
1644 
1645 	if (!isdigit(*str) && *str != '(')
1646 		return -1;
1647 
1648 	/* process single number or single range of number */
1649 	if (*str != '(') {
1650 		errno = 0;
1651 		idx = strtoul(str, &end, 10);
1652 		if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1653 			return -1;
1654 
1655 		while (isblank(*end))
1656 			end++;
1657 
1658 		min = idx;
1659 		max = idx;
1660 
1661 		/* process single <number>-<number> */
1662 		if (*end == '-') {
1663 			end++;
1664 			while (isblank(*end))
1665 				end++;
1666 			if (!isdigit(*end))
1667 				return -1;
1668 
1669 			errno = 0;
1670 			idx = strtoul(end, &end, 10);
1671 			if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1672 				return -1;
1673 
1674 			max = idx;
1675 			while (isblank(*end))
1676 				end++;
1677 		}
1678 
1679 		if (*end != ':')
1680 			return -1;
1681 
1682 		for (idx = RTE_MIN(min, max);
1683 		     idx <= RTE_MAX(min, max); idx++)
1684 			devargs->proto_xtr[idx] = xtr_type;
1685 
1686 		return 0;
1687 	}
1688 
1689 	/* process set within bracket */
1690 	str++;
1691 	while (isblank(*str))
1692 		str++;
1693 	if (*str == '\0')
1694 		return -1;
1695 
1696 	min = IAVF_MAX_QUEUE_NUM;
1697 	do {
1698 		/* go ahead to the first digit */
1699 		while (isblank(*str))
1700 			str++;
1701 		if (!isdigit(*str))
1702 			return -1;
1703 
1704 		/* get the digit value */
1705 		errno = 0;
1706 		idx = strtoul(str, &end, 10);
1707 		if (errno || !end || idx >= IAVF_MAX_QUEUE_NUM)
1708 			return -1;
1709 
1710 		/* go ahead to separator '-',',' and ')' */
1711 		while (isblank(*end))
1712 			end++;
1713 		if (*end == '-') {
1714 			if (min == IAVF_MAX_QUEUE_NUM)
1715 				min = idx;
1716 			else /* avoid continuous '-' */
1717 				return -1;
1718 		} else if (*end == ',' || *end == ')') {
1719 			max = idx;
1720 			if (min == IAVF_MAX_QUEUE_NUM)
1721 				min = idx;
1722 
1723 			for (idx = RTE_MIN(min, max);
1724 			     idx <= RTE_MAX(min, max); idx++)
1725 				devargs->proto_xtr[idx] = xtr_type;
1726 
1727 			min = IAVF_MAX_QUEUE_NUM;
1728 		} else {
1729 			return -1;
1730 		}
1731 
1732 		str = end + 1;
1733 	} while (*end != ')' && *end != '\0');
1734 
1735 	return 0;
1736 }
1737 
1738 static int
1739 iavf_parse_queue_proto_xtr(const char *queues, struct iavf_devargs *devargs)
1740 {
1741 	const char *queue_start;
1742 	uint32_t idx;
1743 	int xtr_type;
1744 	char flex_name[32];
1745 
1746 	while (isblank(*queues))
1747 		queues++;
1748 
1749 	if (*queues != '[') {
1750 		xtr_type = iavf_lookup_proto_xtr_type(queues);
1751 		if (xtr_type < 0)
1752 			return -1;
1753 
1754 		devargs->proto_xtr_dflt = xtr_type;
1755 
1756 		return 0;
1757 	}
1758 
1759 	queues++;
1760 	do {
1761 		while (isblank(*queues))
1762 			queues++;
1763 		if (*queues == '\0')
1764 			return -1;
1765 
1766 		queue_start = queues;
1767 
1768 		/* go across a complete bracket */
1769 		if (*queue_start == '(') {
1770 			queues += strcspn(queues, ")");
1771 			if (*queues != ')')
1772 				return -1;
1773 		}
1774 
1775 		/* scan the separator ':' */
1776 		queues += strcspn(queues, ":");
1777 		if (*queues++ != ':')
1778 			return -1;
1779 		while (isblank(*queues))
1780 			queues++;
1781 
1782 		for (idx = 0; ; idx++) {
1783 			if (isblank(queues[idx]) ||
1784 			    queues[idx] == ',' ||
1785 			    queues[idx] == ']' ||
1786 			    queues[idx] == '\0')
1787 				break;
1788 
1789 			if (idx > sizeof(flex_name) - 2)
1790 				return -1;
1791 
1792 			flex_name[idx] = queues[idx];
1793 		}
1794 		flex_name[idx] = '\0';
1795 		xtr_type = iavf_lookup_proto_xtr_type(flex_name);
1796 		if (xtr_type < 0)
1797 			return -1;
1798 
1799 		queues += idx;
1800 
1801 		while (isblank(*queues) || *queues == ',' || *queues == ']')
1802 			queues++;
1803 
1804 		if (iavf_parse_queue_set(queue_start, xtr_type, devargs) < 0)
1805 			return -1;
1806 	} while (*queues != '\0');
1807 
1808 	return 0;
1809 }
1810 
1811 static int
1812 iavf_handle_proto_xtr_arg(__rte_unused const char *key, const char *value,
1813 			  void *extra_args)
1814 {
1815 	struct iavf_devargs *devargs = extra_args;
1816 
1817 	if (!value || !extra_args)
1818 		return -EINVAL;
1819 
1820 	if (iavf_parse_queue_proto_xtr(value, devargs) < 0) {
1821 		PMD_DRV_LOG(ERR, "the proto_xtr's parameter is wrong : '%s'",
1822 			    value);
1823 		return -1;
1824 	}
1825 
1826 	return 0;
1827 }
1828 
1829 static int iavf_parse_devargs(struct rte_eth_dev *dev)
1830 {
1831 	struct iavf_adapter *ad =
1832 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1833 	struct rte_devargs *devargs = dev->device->devargs;
1834 	struct rte_kvargs *kvlist;
1835 	int ret;
1836 
1837 	if (!devargs)
1838 		return 0;
1839 
1840 	kvlist = rte_kvargs_parse(devargs->args, iavf_valid_args);
1841 	if (!kvlist) {
1842 		PMD_INIT_LOG(ERR, "invalid kvargs key\n");
1843 		return -EINVAL;
1844 	}
1845 
1846 	ad->devargs.proto_xtr_dflt = IAVF_PROTO_XTR_NONE;
1847 	memset(ad->devargs.proto_xtr, IAVF_PROTO_XTR_NONE,
1848 	       sizeof(ad->devargs.proto_xtr));
1849 
1850 	ret = rte_kvargs_process(kvlist, IAVF_PROTO_XTR_ARG,
1851 				 &iavf_handle_proto_xtr_arg, &ad->devargs);
1852 	if (ret)
1853 		goto bail;
1854 
1855 bail:
1856 	rte_kvargs_free(kvlist);
1857 	return ret;
1858 }
1859 
1860 static void
1861 iavf_init_proto_xtr(struct rte_eth_dev *dev)
1862 {
1863 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1864 	struct iavf_adapter *ad =
1865 			IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1866 	const struct iavf_proto_xtr_ol *xtr_ol;
1867 	bool proto_xtr_enable = false;
1868 	int offset;
1869 	uint16_t i;
1870 
1871 	vf->proto_xtr = rte_zmalloc("vf proto xtr",
1872 				    vf->vsi_res->num_queue_pairs, 0);
1873 	if (unlikely(!(vf->proto_xtr))) {
1874 		PMD_DRV_LOG(ERR, "no memory for setting up proto_xtr's table");
1875 		return;
1876 	}
1877 
1878 	for (i = 0; i < vf->vsi_res->num_queue_pairs; i++) {
1879 		vf->proto_xtr[i] = ad->devargs.proto_xtr[i] !=
1880 					IAVF_PROTO_XTR_NONE ?
1881 					ad->devargs.proto_xtr[i] :
1882 					ad->devargs.proto_xtr_dflt;
1883 
1884 		if (vf->proto_xtr[i] != IAVF_PROTO_XTR_NONE) {
1885 			uint8_t type = vf->proto_xtr[i];
1886 
1887 			iavf_proto_xtr_params[type].required = true;
1888 			proto_xtr_enable = true;
1889 		}
1890 	}
1891 
1892 	if (likely(!proto_xtr_enable))
1893 		return;
1894 
1895 	offset = rte_mbuf_dynfield_register(&iavf_proto_xtr_metadata_param);
1896 	if (unlikely(offset == -1)) {
1897 		PMD_DRV_LOG(ERR,
1898 			    "failed to extract protocol metadata, error %d",
1899 			    -rte_errno);
1900 		return;
1901 	}
1902 
1903 	PMD_DRV_LOG(DEBUG,
1904 		    "proto_xtr metadata offset in mbuf is : %d",
1905 		    offset);
1906 	rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = offset;
1907 
1908 	for (i = 0; i < RTE_DIM(iavf_proto_xtr_params); i++) {
1909 		xtr_ol = &iavf_proto_xtr_params[i];
1910 
1911 		uint8_t rxdid = iavf_proto_xtr_type_to_rxdid((uint8_t)i);
1912 
1913 		if (!xtr_ol->required)
1914 			continue;
1915 
1916 		if (!(vf->supported_rxdid & BIT(rxdid))) {
1917 			PMD_DRV_LOG(ERR,
1918 				    "rxdid[%u] is not supported in hardware",
1919 				    rxdid);
1920 			rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
1921 			break;
1922 		}
1923 
1924 		offset = rte_mbuf_dynflag_register(&xtr_ol->param);
1925 		if (unlikely(offset == -1)) {
1926 			PMD_DRV_LOG(ERR,
1927 				    "failed to register proto_xtr offload '%s', error %d",
1928 				    xtr_ol->param.name, -rte_errno);
1929 
1930 			rte_pmd_ifd_dynfield_proto_xtr_metadata_offs = -1;
1931 			break;
1932 		}
1933 
1934 		PMD_DRV_LOG(DEBUG,
1935 			    "proto_xtr offload '%s' offset in mbuf is : %d",
1936 			    xtr_ol->param.name, offset);
1937 		*xtr_ol->ol_flag = 1ULL << offset;
1938 	}
1939 }
1940 
1941 static int
1942 iavf_init_vf(struct rte_eth_dev *dev)
1943 {
1944 	int err, bufsz;
1945 	struct iavf_adapter *adapter =
1946 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
1947 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1948 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
1949 
1950 	err = iavf_parse_devargs(dev);
1951 	if (err) {
1952 		PMD_INIT_LOG(ERR, "Failed to parse devargs");
1953 		goto err;
1954 	}
1955 
1956 	err = iavf_set_mac_type(hw);
1957 	if (err) {
1958 		PMD_INIT_LOG(ERR, "set_mac_type failed: %d", err);
1959 		goto err;
1960 	}
1961 
1962 	err = iavf_check_vf_reset_done(hw);
1963 	if (err) {
1964 		PMD_INIT_LOG(ERR, "VF is still resetting");
1965 		goto err;
1966 	}
1967 
1968 	iavf_init_adminq_parameter(hw);
1969 	err = iavf_init_adminq(hw);
1970 	if (err) {
1971 		PMD_INIT_LOG(ERR, "init_adminq failed: %d", err);
1972 		goto err;
1973 	}
1974 
1975 	vf->aq_resp = rte_zmalloc("vf_aq_resp", IAVF_AQ_BUF_SZ, 0);
1976 	if (!vf->aq_resp) {
1977 		PMD_INIT_LOG(ERR, "unable to allocate vf_aq_resp memory");
1978 		goto err_aq;
1979 	}
1980 	if (iavf_check_api_version(adapter) != 0) {
1981 		PMD_INIT_LOG(ERR, "check_api version failed");
1982 		goto err_api;
1983 	}
1984 
1985 	bufsz = sizeof(struct virtchnl_vf_resource) +
1986 		(IAVF_MAX_VF_VSI * sizeof(struct virtchnl_vsi_resource));
1987 	vf->vf_res = rte_zmalloc("vf_res", bufsz, 0);
1988 	if (!vf->vf_res) {
1989 		PMD_INIT_LOG(ERR, "unable to allocate vf_res memory");
1990 		goto err_api;
1991 	}
1992 	if (iavf_get_vf_resource(adapter) != 0) {
1993 		PMD_INIT_LOG(ERR, "iavf_get_vf_config failed");
1994 		goto err_alloc;
1995 	}
1996 	/* Allocate memort for RSS info */
1997 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
1998 		vf->rss_key = rte_zmalloc("rss_key",
1999 					  vf->vf_res->rss_key_size, 0);
2000 		if (!vf->rss_key) {
2001 			PMD_INIT_LOG(ERR, "unable to allocate rss_key memory");
2002 			goto err_rss;
2003 		}
2004 		vf->rss_lut = rte_zmalloc("rss_lut",
2005 					  vf->vf_res->rss_lut_size, 0);
2006 		if (!vf->rss_lut) {
2007 			PMD_INIT_LOG(ERR, "unable to allocate rss_lut memory");
2008 			goto err_rss;
2009 		}
2010 	}
2011 
2012 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RX_FLEX_DESC) {
2013 		if (iavf_get_supported_rxdid(adapter) != 0) {
2014 			PMD_INIT_LOG(ERR, "failed to do get supported rxdid");
2015 			goto err_rss;
2016 		}
2017 	}
2018 
2019 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_VLAN_V2) {
2020 		if (iavf_get_vlan_offload_caps_v2(adapter) != 0) {
2021 			PMD_INIT_LOG(ERR, "failed to do get VLAN offload v2 capabilities");
2022 			goto err_rss;
2023 		}
2024 	}
2025 
2026 	iavf_init_proto_xtr(dev);
2027 
2028 	return 0;
2029 err_rss:
2030 	rte_free(vf->rss_key);
2031 	rte_free(vf->rss_lut);
2032 err_alloc:
2033 	rte_free(vf->vf_res);
2034 	vf->vsi_res = NULL;
2035 err_api:
2036 	rte_free(vf->aq_resp);
2037 err_aq:
2038 	iavf_shutdown_adminq(hw);
2039 err:
2040 	return -1;
2041 }
2042 
2043 /* Enable default admin queue interrupt setting */
2044 static inline void
2045 iavf_enable_irq0(struct iavf_hw *hw)
2046 {
2047 	/* Enable admin queue interrupt trigger */
2048 	IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1,
2049 		       IAVF_VFINT_ICR0_ENA1_ADMINQ_MASK);
2050 
2051 	IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2052 		       IAVF_VFINT_DYN_CTL01_INTENA_MASK |
2053 		       IAVF_VFINT_DYN_CTL01_CLEARPBA_MASK |
2054 		       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2055 
2056 	IAVF_WRITE_FLUSH(hw);
2057 }
2058 
2059 static inline void
2060 iavf_disable_irq0(struct iavf_hw *hw)
2061 {
2062 	/* Disable all interrupt types */
2063 	IAVF_WRITE_REG(hw, IAVF_VFINT_ICR0_ENA1, 0);
2064 	IAVF_WRITE_REG(hw, IAVF_VFINT_DYN_CTL01,
2065 		       IAVF_VFINT_DYN_CTL01_ITR_INDX_MASK);
2066 	IAVF_WRITE_FLUSH(hw);
2067 }
2068 
2069 static void
2070 iavf_dev_interrupt_handler(void *param)
2071 {
2072 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
2073 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2074 
2075 	iavf_disable_irq0(hw);
2076 
2077 	iavf_handle_virtchnl_msg(dev);
2078 
2079 	iavf_enable_irq0(hw);
2080 }
2081 
2082 static int
2083 iavf_dev_flow_ops_get(struct rte_eth_dev *dev,
2084 		      const struct rte_flow_ops **ops)
2085 {
2086 	if (!dev)
2087 		return -EINVAL;
2088 
2089 	*ops = &iavf_flow_ops;
2090 	return 0;
2091 }
2092 
2093 static void
2094 iavf_default_rss_disable(struct iavf_adapter *adapter)
2095 {
2096 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(adapter);
2097 	int ret = 0;
2098 
2099 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_ADV_RSS_PF) {
2100 		/* Set hena = 0 to ask PF to cleanup all existing RSS. */
2101 		ret = iavf_set_hena(adapter, 0);
2102 		if (ret)
2103 			/* It is a workaround, temporarily allow error to be
2104 			 * returned due to possible lack of PF handling for
2105 			 * hena = 0.
2106 			 */
2107 			PMD_INIT_LOG(WARNING, "fail to disable default RSS,"
2108 				    "lack PF support");
2109 	}
2110 }
2111 
2112 static int
2113 iavf_dev_init(struct rte_eth_dev *eth_dev)
2114 {
2115 	struct iavf_adapter *adapter =
2116 		IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2117 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(adapter);
2118 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
2119 	int ret = 0;
2120 
2121 	PMD_INIT_FUNC_TRACE();
2122 
2123 	/* assign ops func pointer */
2124 	eth_dev->dev_ops = &iavf_eth_dev_ops;
2125 	eth_dev->rx_queue_count = iavf_dev_rxq_count;
2126 	eth_dev->rx_descriptor_status = iavf_dev_rx_desc_status;
2127 	eth_dev->tx_descriptor_status = iavf_dev_tx_desc_status;
2128 	eth_dev->rx_pkt_burst = &iavf_recv_pkts;
2129 	eth_dev->tx_pkt_burst = &iavf_xmit_pkts;
2130 	eth_dev->tx_pkt_prepare = &iavf_prep_pkts;
2131 
2132 	/* For secondary processes, we don't initialise any further as primary
2133 	 * has already done this work. Only check if we need a different RX
2134 	 * and TX function.
2135 	 */
2136 	if (rte_eal_process_type() != RTE_PROC_PRIMARY) {
2137 		iavf_set_rx_function(eth_dev);
2138 		iavf_set_tx_function(eth_dev);
2139 		return 0;
2140 	}
2141 	rte_eth_copy_pci_info(eth_dev, pci_dev);
2142 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
2143 
2144 	hw->vendor_id = pci_dev->id.vendor_id;
2145 	hw->device_id = pci_dev->id.device_id;
2146 	hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
2147 	hw->subsystem_device_id = pci_dev->id.subsystem_device_id;
2148 	hw->bus.bus_id = pci_dev->addr.bus;
2149 	hw->bus.device = pci_dev->addr.devid;
2150 	hw->bus.func = pci_dev->addr.function;
2151 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
2152 	hw->back = IAVF_DEV_PRIVATE_TO_ADAPTER(eth_dev->data->dev_private);
2153 	adapter->eth_dev = eth_dev;
2154 	adapter->stopped = 1;
2155 
2156 	if (iavf_init_vf(eth_dev) != 0) {
2157 		PMD_INIT_LOG(ERR, "Init vf failed");
2158 		return -1;
2159 	}
2160 
2161 	/* set default ptype table */
2162 	adapter->ptype_tbl = iavf_get_default_ptype_table();
2163 
2164 	/* copy mac addr */
2165 	eth_dev->data->mac_addrs = rte_zmalloc(
2166 		"iavf_mac", RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX, 0);
2167 	if (!eth_dev->data->mac_addrs) {
2168 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to"
2169 			     " store MAC addresses",
2170 			     RTE_ETHER_ADDR_LEN * IAVF_NUM_MACADDR_MAX);
2171 		return -ENOMEM;
2172 	}
2173 	/* If the MAC address is not configured by host,
2174 	 * generate a random one.
2175 	 */
2176 	if (!rte_is_valid_assigned_ether_addr(
2177 			(struct rte_ether_addr *)hw->mac.addr))
2178 		rte_eth_random_addr(hw->mac.addr);
2179 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
2180 			&eth_dev->data->mac_addrs[0]);
2181 
2182 	/* register callback func to eal lib */
2183 	rte_intr_callback_register(&pci_dev->intr_handle,
2184 				   iavf_dev_interrupt_handler,
2185 				   (void *)eth_dev);
2186 
2187 	/* enable uio intr after callback register */
2188 	rte_intr_enable(&pci_dev->intr_handle);
2189 
2190 	/* configure and enable device interrupt */
2191 	iavf_enable_irq0(hw);
2192 
2193 	ret = iavf_flow_init(adapter);
2194 	if (ret) {
2195 		PMD_INIT_LOG(ERR, "Failed to initialize flow");
2196 		return ret;
2197 	}
2198 
2199 	iavf_default_rss_disable(adapter);
2200 
2201 	return 0;
2202 }
2203 
2204 static int
2205 iavf_dev_close(struct rte_eth_dev *dev)
2206 {
2207 	struct iavf_hw *hw = IAVF_DEV_PRIVATE_TO_HW(dev->data->dev_private);
2208 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
2209 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
2210 	struct iavf_adapter *adapter =
2211 		IAVF_DEV_PRIVATE_TO_ADAPTER(dev->data->dev_private);
2212 	struct iavf_info *vf = IAVF_DEV_PRIVATE_TO_VF(dev->data->dev_private);
2213 	int ret;
2214 
2215 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2216 		return 0;
2217 
2218 	ret = iavf_dev_stop(dev);
2219 
2220 	iavf_flow_flush(dev, NULL);
2221 	iavf_flow_uninit(adapter);
2222 
2223 	/*
2224 	 * disable promiscuous mode before reset vf
2225 	 * it is a workaround solution when work with kernel driver
2226 	 * and it is not the normal way
2227 	 */
2228 	if (vf->promisc_unicast_enabled || vf->promisc_multicast_enabled)
2229 		iavf_config_promisc(adapter, false, false);
2230 
2231 	iavf_shutdown_adminq(hw);
2232 	/* disable uio intr before callback unregister */
2233 	rte_intr_disable(intr_handle);
2234 
2235 	/* unregister callback func from eal lib */
2236 	rte_intr_callback_unregister(intr_handle,
2237 				     iavf_dev_interrupt_handler, dev);
2238 	iavf_disable_irq0(hw);
2239 
2240 	if (vf->vf_res->vf_cap_flags & VIRTCHNL_VF_OFFLOAD_RSS_PF) {
2241 		if (vf->rss_lut) {
2242 			rte_free(vf->rss_lut);
2243 			vf->rss_lut = NULL;
2244 		}
2245 		if (vf->rss_key) {
2246 			rte_free(vf->rss_key);
2247 			vf->rss_key = NULL;
2248 		}
2249 	}
2250 
2251 	rte_free(vf->vf_res);
2252 	vf->vsi_res = NULL;
2253 	vf->vf_res = NULL;
2254 
2255 	rte_free(vf->aq_resp);
2256 	vf->aq_resp = NULL;
2257 
2258 	vf->vf_reset = false;
2259 
2260 	return ret;
2261 }
2262 
2263 static int
2264 iavf_dev_uninit(struct rte_eth_dev *dev)
2265 {
2266 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
2267 		return -EPERM;
2268 
2269 	iavf_dev_close(dev);
2270 
2271 	return 0;
2272 }
2273 
2274 /*
2275  * Reset VF device only to re-initialize resources in PMD layer
2276  */
2277 static int
2278 iavf_dev_reset(struct rte_eth_dev *dev)
2279 {
2280 	int ret;
2281 
2282 	ret = iavf_dev_uninit(dev);
2283 	if (ret)
2284 		return ret;
2285 
2286 	return iavf_dev_init(dev);
2287 }
2288 
2289 static int
2290 iavf_dcf_cap_check_handler(__rte_unused const char *key,
2291 			   const char *value, __rte_unused void *opaque)
2292 {
2293 	if (strcmp(value, "dcf"))
2294 		return -1;
2295 
2296 	return 0;
2297 }
2298 
2299 static int
2300 iavf_dcf_cap_selected(struct rte_devargs *devargs)
2301 {
2302 	struct rte_kvargs *kvlist;
2303 	const char *key = "cap";
2304 	int ret = 0;
2305 
2306 	if (devargs == NULL)
2307 		return 0;
2308 
2309 	kvlist = rte_kvargs_parse(devargs->args, NULL);
2310 	if (kvlist == NULL)
2311 		return 0;
2312 
2313 	if (!rte_kvargs_count(kvlist, key))
2314 		goto exit;
2315 
2316 	/* dcf capability selected when there's a key-value pair: cap=dcf */
2317 	if (rte_kvargs_process(kvlist, key,
2318 			       iavf_dcf_cap_check_handler, NULL) < 0)
2319 		goto exit;
2320 
2321 	ret = 1;
2322 
2323 exit:
2324 	rte_kvargs_free(kvlist);
2325 	return ret;
2326 }
2327 
2328 static int
2329 iavf_drv_i40evf_check_handler(__rte_unused const char *key,
2330 			      const char *value, __rte_unused void *opaque)
2331 {
2332 	if (strcmp(value, "i40evf"))
2333 		return -1;
2334 
2335 	return 0;
2336 }
2337 
2338 static int
2339 iavf_drv_i40evf_selected(struct rte_devargs *devargs, uint16_t device_id)
2340 {
2341 	struct rte_kvargs *kvlist;
2342 	const char *key = "driver";
2343 	int ret = 0;
2344 
2345 	if (device_id != IAVF_DEV_ID_VF &&
2346 	    device_id != IAVF_DEV_ID_VF_HV &&
2347 	    device_id != IAVF_DEV_ID_X722_VF &&
2348 	    device_id != IAVF_DEV_ID_X722_A0_VF)
2349 		return 0;
2350 
2351 	if (devargs == NULL)
2352 		return 0;
2353 
2354 	kvlist = rte_kvargs_parse(devargs->args, NULL);
2355 	if (kvlist == NULL)
2356 		return 0;
2357 
2358 	if (!rte_kvargs_count(kvlist, key))
2359 		goto exit;
2360 
2361 	/* i40evf driver selected when there's a key-value pair:
2362 	 * driver=i40evf
2363 	 */
2364 	if (rte_kvargs_process(kvlist, key,
2365 			       iavf_drv_i40evf_check_handler, NULL) < 0)
2366 		goto exit;
2367 
2368 	ret = 1;
2369 
2370 exit:
2371 	rte_kvargs_free(kvlist);
2372 	return ret;
2373 }
2374 
2375 static int eth_iavf_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
2376 			     struct rte_pci_device *pci_dev)
2377 {
2378 	if (iavf_dcf_cap_selected(pci_dev->device.devargs) ||
2379 	    iavf_drv_i40evf_selected(pci_dev->device.devargs,
2380 				     pci_dev->id.device_id))
2381 		return 1;
2382 
2383 	return rte_eth_dev_pci_generic_probe(pci_dev,
2384 		sizeof(struct iavf_adapter), iavf_dev_init);
2385 }
2386 
2387 static int eth_iavf_pci_remove(struct rte_pci_device *pci_dev)
2388 {
2389 	return rte_eth_dev_pci_generic_remove(pci_dev, iavf_dev_uninit);
2390 }
2391 
2392 /* Adaptive virtual function driver struct */
2393 static struct rte_pci_driver rte_iavf_pmd = {
2394 	.id_table = pci_id_iavf_map,
2395 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
2396 	.probe = eth_iavf_pci_probe,
2397 	.remove = eth_iavf_pci_remove,
2398 };
2399 
2400 RTE_PMD_REGISTER_PCI(net_iavf, rte_iavf_pmd);
2401 RTE_PMD_REGISTER_PCI_TABLE(net_iavf, pci_id_iavf_map);
2402 RTE_PMD_REGISTER_KMOD_DEP(net_iavf, "* igb_uio | vfio-pci");
2403 RTE_PMD_REGISTER_PARAM_STRING(net_iavf, "cap=dcf driver=i40evf");
2404 RTE_LOG_REGISTER(iavf_logtype_init, pmd.net.iavf.init, NOTICE);
2405 RTE_LOG_REGISTER(iavf_logtype_driver, pmd.net.iavf.driver, NOTICE);
2406 #ifdef RTE_ETHDEV_DEBUG_RX
2407 RTE_LOG_REGISTER(iavf_logtype_rx, pmd.net.iavf.rx, DEBUG);
2408 #endif
2409 #ifdef RTE_ETHDEV_DEBUG_TX
2410 RTE_LOG_REGISTER(iavf_logtype_tx, pmd.net.iavf.tx, DEBUG);
2411 #endif
2412