xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision d9aa619c60b67131d0f8d0ad308576ca72ce7c90)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2015 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 <fcntl.h>
13 #include <inttypes.h>
14 #include <rte_byteorder.h>
15 #include <rte_common.h>
16 #include <rte_cycles.h>
17 
18 #include <rte_interrupts.h>
19 #include <rte_log.h>
20 #include <rte_debug.h>
21 #include <rte_pci.h>
22 #include <rte_bus_pci.h>
23 #include <rte_branch_prediction.h>
24 #include <rte_memory.h>
25 #include <rte_memzone.h>
26 #include <rte_eal.h>
27 #include <rte_alarm.h>
28 #include <rte_ether.h>
29 #include <rte_ethdev_driver.h>
30 #include <rte_ethdev_pci.h>
31 #include <rte_string_fns.h>
32 #include <rte_malloc.h>
33 #include <rte_dev.h>
34 
35 #include "base/vmxnet3_defs.h"
36 
37 #include "vmxnet3_ring.h"
38 #include "vmxnet3_logs.h"
39 #include "vmxnet3_ethdev.h"
40 
41 #define PROCESS_SYS_EVENTS 0
42 
43 #define	VMXNET3_TX_MAX_SEG	UINT8_MAX
44 
45 static int eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev);
46 static int eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev);
47 static int vmxnet3_dev_configure(struct rte_eth_dev *dev);
48 static int vmxnet3_dev_start(struct rte_eth_dev *dev);
49 static void vmxnet3_dev_stop(struct rte_eth_dev *dev);
50 static void vmxnet3_dev_close(struct rte_eth_dev *dev);
51 static void vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set);
52 static void vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev);
53 static void vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev);
54 static void vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev);
55 static void vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev);
56 static int __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
57 				     int wait_to_complete);
58 static int vmxnet3_dev_link_update(struct rte_eth_dev *dev,
59 				   int wait_to_complete);
60 static void vmxnet3_hw_stats_save(struct vmxnet3_hw *hw);
61 static int vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
62 				  struct rte_eth_stats *stats);
63 static int vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
64 					struct rte_eth_xstat_name *xstats,
65 					unsigned int n);
66 static int vmxnet3_dev_xstats_get(struct rte_eth_dev *dev,
67 				  struct rte_eth_xstat *xstats, unsigned int n);
68 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
69 				 struct rte_eth_dev_info *dev_info);
70 static const uint32_t *
71 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
72 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
73 				       uint16_t vid, int on);
74 static int vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
75 static int vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
76 				 struct ether_addr *mac_addr);
77 static void vmxnet3_interrupt_handler(void *param);
78 
79 int vmxnet3_logtype_init;
80 int vmxnet3_logtype_driver;
81 
82 /*
83  * The set of PCI devices this driver supports
84  */
85 #define VMWARE_PCI_VENDOR_ID 0x15AD
86 #define VMWARE_DEV_ID_VMXNET3 0x07B0
87 static const struct rte_pci_id pci_id_vmxnet3_map[] = {
88 	{ RTE_PCI_DEVICE(VMWARE_PCI_VENDOR_ID, VMWARE_DEV_ID_VMXNET3) },
89 	{ .vendor_id = 0, /* sentinel */ },
90 };
91 
92 static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
93 	.dev_configure        = vmxnet3_dev_configure,
94 	.dev_start            = vmxnet3_dev_start,
95 	.dev_stop             = vmxnet3_dev_stop,
96 	.dev_close            = vmxnet3_dev_close,
97 	.promiscuous_enable   = vmxnet3_dev_promiscuous_enable,
98 	.promiscuous_disable  = vmxnet3_dev_promiscuous_disable,
99 	.allmulticast_enable  = vmxnet3_dev_allmulticast_enable,
100 	.allmulticast_disable = vmxnet3_dev_allmulticast_disable,
101 	.link_update          = vmxnet3_dev_link_update,
102 	.stats_get            = vmxnet3_dev_stats_get,
103 	.xstats_get_names     = vmxnet3_dev_xstats_get_names,
104 	.xstats_get           = vmxnet3_dev_xstats_get,
105 	.mac_addr_set         = vmxnet3_mac_addr_set,
106 	.dev_infos_get        = vmxnet3_dev_info_get,
107 	.dev_supported_ptypes_get = vmxnet3_dev_supported_ptypes_get,
108 	.vlan_filter_set      = vmxnet3_dev_vlan_filter_set,
109 	.vlan_offload_set     = vmxnet3_dev_vlan_offload_set,
110 	.rx_queue_setup       = vmxnet3_dev_rx_queue_setup,
111 	.rx_queue_release     = vmxnet3_dev_rx_queue_release,
112 	.tx_queue_setup       = vmxnet3_dev_tx_queue_setup,
113 	.tx_queue_release     = vmxnet3_dev_tx_queue_release,
114 };
115 
116 struct vmxnet3_xstats_name_off {
117 	char name[RTE_ETH_XSTATS_NAME_SIZE];
118 	unsigned int offset;
119 };
120 
121 /* tx_qX_ is prepended to the name string here */
122 static const struct vmxnet3_xstats_name_off vmxnet3_txq_stat_strings[] = {
123 	{"drop_total",         offsetof(struct vmxnet3_txq_stats, drop_total)},
124 	{"drop_too_many_segs", offsetof(struct vmxnet3_txq_stats, drop_too_many_segs)},
125 	{"drop_tso",           offsetof(struct vmxnet3_txq_stats, drop_tso)},
126 	{"tx_ring_full",       offsetof(struct vmxnet3_txq_stats, tx_ring_full)},
127 };
128 
129 /* rx_qX_ is prepended to the name string here */
130 static const struct vmxnet3_xstats_name_off vmxnet3_rxq_stat_strings[] = {
131 	{"drop_total",           offsetof(struct vmxnet3_rxq_stats, drop_total)},
132 	{"drop_err",             offsetof(struct vmxnet3_rxq_stats, drop_err)},
133 	{"drop_fcs",             offsetof(struct vmxnet3_rxq_stats, drop_fcs)},
134 	{"rx_buf_alloc_failure", offsetof(struct vmxnet3_rxq_stats, rx_buf_alloc_failure)},
135 };
136 
137 static const struct rte_memzone *
138 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
139 		 const char *post_string, int socket_id,
140 		 uint16_t align, bool reuse)
141 {
142 	char z_name[RTE_MEMZONE_NAMESIZE];
143 	const struct rte_memzone *mz;
144 
145 	snprintf(z_name, sizeof(z_name), "%s_%d_%s",
146 		 dev->device->driver->name, dev->data->port_id, post_string);
147 
148 	mz = rte_memzone_lookup(z_name);
149 	if (!reuse) {
150 		if (mz)
151 			rte_memzone_free(mz);
152 		return rte_memzone_reserve_aligned(z_name, size, socket_id,
153 				RTE_MEMZONE_IOVA_CONTIG, align);
154 	}
155 
156 	if (mz)
157 		return mz;
158 
159 	return rte_memzone_reserve_aligned(z_name, size, socket_id,
160 			RTE_MEMZONE_IOVA_CONTIG, align);
161 }
162 
163 /*
164  * This function is based on vmxnet3_disable_intr()
165  */
166 static void
167 vmxnet3_disable_intr(struct vmxnet3_hw *hw)
168 {
169 	int i;
170 
171 	PMD_INIT_FUNC_TRACE();
172 
173 	hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
174 	for (i = 0; i < hw->num_intrs; i++)
175 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
176 }
177 
178 static void
179 vmxnet3_enable_intr(struct vmxnet3_hw *hw)
180 {
181 	int i;
182 
183 	PMD_INIT_FUNC_TRACE();
184 
185 	hw->shared->devRead.intrConf.intrCtrl &= ~VMXNET3_IC_DISABLE_ALL;
186 	for (i = 0; i < hw->num_intrs; i++)
187 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 0);
188 }
189 
190 /*
191  * Gets tx data ring descriptor size.
192  */
193 static uint16_t
194 eth_vmxnet3_txdata_get(struct vmxnet3_hw *hw)
195 {
196 	uint16 txdata_desc_size;
197 
198 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
199 			       VMXNET3_CMD_GET_TXDATA_DESC_SIZE);
200 	txdata_desc_size = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
201 
202 	return (txdata_desc_size < VMXNET3_TXDATA_DESC_MIN_SIZE ||
203 		txdata_desc_size > VMXNET3_TXDATA_DESC_MAX_SIZE ||
204 		txdata_desc_size & VMXNET3_TXDATA_DESC_SIZE_MASK) ?
205 		sizeof(struct Vmxnet3_TxDataDesc) : txdata_desc_size;
206 }
207 
208 /*
209  * It returns 0 on success.
210  */
211 static int
212 eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
213 {
214 	struct rte_pci_device *pci_dev;
215 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
216 	uint32_t mac_hi, mac_lo, ver;
217 	struct rte_eth_link link;
218 
219 	PMD_INIT_FUNC_TRACE();
220 
221 	eth_dev->dev_ops = &vmxnet3_eth_dev_ops;
222 	eth_dev->rx_pkt_burst = &vmxnet3_recv_pkts;
223 	eth_dev->tx_pkt_burst = &vmxnet3_xmit_pkts;
224 	eth_dev->tx_pkt_prepare = vmxnet3_prep_pkts;
225 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
226 
227 	/*
228 	 * for secondary processes, we don't initialize any further as primary
229 	 * has already done this work.
230 	 */
231 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
232 		return 0;
233 
234 	rte_eth_copy_pci_info(eth_dev, pci_dev);
235 
236 	/* Vendor and Device ID need to be set before init of shared code */
237 	hw->device_id = pci_dev->id.device_id;
238 	hw->vendor_id = pci_dev->id.vendor_id;
239 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
240 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
241 
242 	hw->num_rx_queues = 1;
243 	hw->num_tx_queues = 1;
244 	hw->bufs_per_pkt = 1;
245 
246 	/* Check h/w version compatibility with driver. */
247 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
248 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
249 
250 	if (ver & (1 << VMXNET3_REV_3)) {
251 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
252 				       1 << VMXNET3_REV_3);
253 		hw->version = VMXNET3_REV_3 + 1;
254 	} else if (ver & (1 << VMXNET3_REV_2)) {
255 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
256 				       1 << VMXNET3_REV_2);
257 		hw->version = VMXNET3_REV_2 + 1;
258 	} else if (ver & (1 << VMXNET3_REV_1)) {
259 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
260 				       1 << VMXNET3_REV_1);
261 		hw->version = VMXNET3_REV_1 + 1;
262 	} else {
263 		PMD_INIT_LOG(ERR, "Incompatible hardware version: %d", ver);
264 		return -EIO;
265 	}
266 
267 	PMD_INIT_LOG(DEBUG, "Using device version %d\n", hw->version);
268 
269 	/* Check UPT version compatibility with driver. */
270 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
271 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
272 	if (ver & 0x1)
273 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
274 	else {
275 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
276 		return -EIO;
277 	}
278 
279 	/* Getting MAC Address */
280 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
281 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
282 	memcpy(hw->perm_addr, &mac_lo, 4);
283 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
284 
285 	/* Allocate memory for storing MAC addresses */
286 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
287 					       VMXNET3_MAX_MAC_ADDRS, 0);
288 	if (eth_dev->data->mac_addrs == NULL) {
289 		PMD_INIT_LOG(ERR,
290 			     "Failed to allocate %d bytes needed to store MAC addresses",
291 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
292 		return -ENOMEM;
293 	}
294 	/* Copy the permanent MAC address */
295 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
296 			&eth_dev->data->mac_addrs[0]);
297 
298 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
299 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
300 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
301 
302 	/* Put device in Quiesce Mode */
303 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
304 
305 	/* allow untagged pkts */
306 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
307 
308 	hw->txdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
309 		eth_vmxnet3_txdata_get(hw) : sizeof(struct Vmxnet3_TxDataDesc);
310 
311 	hw->rxdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
312 		VMXNET3_DEF_RXDATA_DESC_SIZE : 0;
313 	RTE_ASSERT((hw->rxdata_desc_size & ~VMXNET3_RXDATA_DESC_SIZE_MASK) ==
314 		   hw->rxdata_desc_size);
315 
316 	/* clear shadow stats */
317 	memset(hw->saved_tx_stats, 0, sizeof(hw->saved_tx_stats));
318 	memset(hw->saved_rx_stats, 0, sizeof(hw->saved_rx_stats));
319 
320 	/* set the initial link status */
321 	memset(&link, 0, sizeof(link));
322 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
323 	link.link_speed = ETH_SPEED_NUM_10G;
324 	link.link_autoneg = ETH_LINK_FIXED;
325 	rte_eth_linkstatus_set(eth_dev, &link);
326 
327 	return 0;
328 }
329 
330 static int
331 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
332 {
333 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
334 
335 	PMD_INIT_FUNC_TRACE();
336 
337 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
338 		return 0;
339 
340 	if (hw->adapter_stopped == 0)
341 		vmxnet3_dev_close(eth_dev);
342 
343 	eth_dev->dev_ops = NULL;
344 	eth_dev->rx_pkt_burst = NULL;
345 	eth_dev->tx_pkt_burst = NULL;
346 	eth_dev->tx_pkt_prepare = NULL;
347 
348 	rte_free(eth_dev->data->mac_addrs);
349 	eth_dev->data->mac_addrs = NULL;
350 
351 	return 0;
352 }
353 
354 static int eth_vmxnet3_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
355 	struct rte_pci_device *pci_dev)
356 {
357 	return rte_eth_dev_pci_generic_probe(pci_dev,
358 		sizeof(struct vmxnet3_hw), eth_vmxnet3_dev_init);
359 }
360 
361 static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
362 {
363 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_vmxnet3_dev_uninit);
364 }
365 
366 static struct rte_pci_driver rte_vmxnet3_pmd = {
367 	.id_table = pci_id_vmxnet3_map,
368 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
369 	.probe = eth_vmxnet3_pci_probe,
370 	.remove = eth_vmxnet3_pci_remove,
371 };
372 
373 static int
374 vmxnet3_dev_configure(struct rte_eth_dev *dev)
375 {
376 	const struct rte_memzone *mz;
377 	struct vmxnet3_hw *hw = dev->data->dev_private;
378 	size_t size;
379 
380 	PMD_INIT_FUNC_TRACE();
381 
382 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
383 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
384 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
385 		return -EINVAL;
386 	}
387 
388 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
389 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
390 		return -EINVAL;
391 	}
392 
393 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
394 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
395 
396 	if (size > UINT16_MAX)
397 		return -EINVAL;
398 
399 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
400 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
401 
402 	/*
403 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
404 	 * on current socket
405 	 */
406 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
407 			      "shared", rte_socket_id(), 8, 1);
408 
409 	if (mz == NULL) {
410 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
411 		return -ENOMEM;
412 	}
413 	memset(mz->addr, 0, mz->len);
414 
415 	hw->shared = mz->addr;
416 	hw->sharedPA = mz->iova;
417 
418 	/*
419 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
420 	 * on current socket.
421 	 *
422 	 * We cannot reuse this memzone from previous allocation as its size
423 	 * depends on the number of tx and rx queues, which could be different
424 	 * from one config to another.
425 	 */
426 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
427 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
428 	if (mz == NULL) {
429 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
430 		return -ENOMEM;
431 	}
432 	memset(mz->addr, 0, mz->len);
433 
434 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
435 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
436 
437 	hw->queueDescPA = mz->iova;
438 	hw->queue_desc_len = (uint16_t)size;
439 
440 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
441 		/* Allocate memory structure for UPT1_RSSConf and configure */
442 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
443 				      "rss_conf", rte_socket_id(),
444 				      RTE_CACHE_LINE_SIZE, 1);
445 		if (mz == NULL) {
446 			PMD_INIT_LOG(ERR,
447 				     "ERROR: Creating rss_conf structure zone");
448 			return -ENOMEM;
449 		}
450 		memset(mz->addr, 0, mz->len);
451 
452 		hw->rss_conf = mz->addr;
453 		hw->rss_confPA = mz->iova;
454 	}
455 
456 	return 0;
457 }
458 
459 static void
460 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
461 {
462 	uint32_t val;
463 
464 	PMD_INIT_LOG(DEBUG,
465 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
466 		     addr[0], addr[1], addr[2],
467 		     addr[3], addr[4], addr[5]);
468 
469 	memcpy(&val, addr, 4);
470 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
471 
472 	memcpy(&val, addr + 4, 2);
473 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
474 }
475 
476 static int
477 vmxnet3_dev_setup_memreg(struct rte_eth_dev *dev)
478 {
479 	struct vmxnet3_hw *hw = dev->data->dev_private;
480 	Vmxnet3_DriverShared *shared = hw->shared;
481 	Vmxnet3_CmdInfo *cmdInfo;
482 	struct rte_mempool *mp[VMXNET3_MAX_RX_QUEUES];
483 	uint8_t index[VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES];
484 	uint32_t num, i, j, size;
485 
486 	if (hw->memRegsPA == 0) {
487 		const struct rte_memzone *mz;
488 
489 		size = sizeof(Vmxnet3_MemRegs) +
490 			(VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES) *
491 			sizeof(Vmxnet3_MemoryRegion);
492 
493 		mz = gpa_zone_reserve(dev, size, "memRegs", rte_socket_id(), 8,
494 				      1);
495 		if (mz == NULL) {
496 			PMD_INIT_LOG(ERR, "ERROR: Creating memRegs zone");
497 			return -ENOMEM;
498 		}
499 		memset(mz->addr, 0, mz->len);
500 		hw->memRegs = mz->addr;
501 		hw->memRegsPA = mz->iova;
502 	}
503 
504 	num = hw->num_rx_queues;
505 
506 	for (i = 0; i < num; i++) {
507 		vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
508 
509 		mp[i] = rxq->mp;
510 		index[i] = 1 << i;
511 	}
512 
513 	/*
514 	 * The same mempool could be used by multiple queues. In such a case,
515 	 * remove duplicate mempool entries. Only one entry is kept with
516 	 * bitmask indicating queues that are using this mempool.
517 	 */
518 	for (i = 1; i < num; i++) {
519 		for (j = 0; j < i; j++) {
520 			if (mp[i] == mp[j]) {
521 				mp[i] = NULL;
522 				index[j] |= 1 << i;
523 				break;
524 			}
525 		}
526 	}
527 
528 	j = 0;
529 	for (i = 0; i < num; i++) {
530 		if (mp[i] == NULL)
531 			continue;
532 
533 		Vmxnet3_MemoryRegion *mr = &hw->memRegs->memRegs[j];
534 
535 		mr->startPA =
536 			(uintptr_t)STAILQ_FIRST(&mp[i]->mem_list)->iova;
537 		mr->length = STAILQ_FIRST(&mp[i]->mem_list)->len <= INT32_MAX ?
538 			STAILQ_FIRST(&mp[i]->mem_list)->len : INT32_MAX;
539 		mr->txQueueBits = index[i];
540 		mr->rxQueueBits = index[i];
541 
542 		PMD_INIT_LOG(INFO,
543 			     "index: %u startPA: %" PRIu64 " length: %u, "
544 			     "rxBits: %x",
545 			     j, mr->startPA, mr->length, mr->rxQueueBits);
546 		j++;
547 	}
548 	hw->memRegs->numRegs = j;
549 	PMD_INIT_LOG(INFO, "numRegs: %u", j);
550 
551 	size = sizeof(Vmxnet3_MemRegs) +
552 		(j - 1) * sizeof(Vmxnet3_MemoryRegion);
553 
554 	cmdInfo = &shared->cu.cmdInfo;
555 	cmdInfo->varConf.confVer = 1;
556 	cmdInfo->varConf.confLen = size;
557 	cmdInfo->varConf.confPA = hw->memRegsPA;
558 
559 	return 0;
560 }
561 
562 static int
563 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
564 {
565 	struct rte_eth_conf port_conf = dev->data->dev_conf;
566 	struct vmxnet3_hw *hw = dev->data->dev_private;
567 	uint32_t mtu = dev->data->mtu;
568 	Vmxnet3_DriverShared *shared = hw->shared;
569 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
570 	uint32_t i;
571 	int ret;
572 
573 	hw->mtu = mtu;
574 
575 	shared->magic = VMXNET3_REV1_MAGIC;
576 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
577 
578 	/* Setting up Guest OS information */
579 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
580 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
581 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
582 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
583 	devRead->misc.driverInfo.uptVerSpt     = 1;
584 
585 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
586 	devRead->misc.queueDescPA  = hw->queueDescPA;
587 	devRead->misc.queueDescLen = hw->queue_desc_len;
588 	devRead->misc.numTxQueues  = hw->num_tx_queues;
589 	devRead->misc.numRxQueues  = hw->num_rx_queues;
590 
591 	/*
592 	 * Set number of interrupts to 1
593 	 * PMD by default disables all the interrupts but this is MUST
594 	 * to activate device. It needs at least one interrupt for
595 	 * link events to handle
596 	 */
597 	hw->num_intrs = devRead->intrConf.numIntrs = 1;
598 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
599 
600 	for (i = 0; i < hw->num_tx_queues; i++) {
601 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
602 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
603 
604 		txq->shared = &hw->tqd_start[i];
605 
606 		tqd->ctrl.txNumDeferred  = 0;
607 		tqd->ctrl.txThreshold    = 1;
608 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
609 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
610 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
611 
612 		tqd->conf.txRingSize   = txq->cmd_ring.size;
613 		tqd->conf.compRingSize = txq->comp_ring.size;
614 		tqd->conf.dataRingSize = txq->data_ring.size;
615 		tqd->conf.txDataRingDescSize = txq->txdata_desc_size;
616 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
617 		tqd->status.stopped    = TRUE;
618 		tqd->status.error      = 0;
619 		memset(&tqd->stats, 0, sizeof(tqd->stats));
620 	}
621 
622 	for (i = 0; i < hw->num_rx_queues; i++) {
623 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
624 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
625 
626 		rxq->shared = &hw->rqd_start[i];
627 
628 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
629 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
630 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
631 
632 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
633 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
634 		rqd->conf.compRingSize    = rxq->comp_ring.size;
635 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
636 		if (VMXNET3_VERSION_GE_3(hw)) {
637 			rqd->conf.rxDataRingBasePA = rxq->data_ring.basePA;
638 			rqd->conf.rxDataRingDescSize = rxq->data_desc_size;
639 		}
640 		rqd->status.stopped       = TRUE;
641 		rqd->status.error         = 0;
642 		memset(&rqd->stats, 0, sizeof(rqd->stats));
643 	}
644 
645 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
646 	devRead->rxFilterConf.rxMode = 0;
647 
648 	/* Setting up feature flags */
649 	if (dev->data->dev_conf.rxmode.hw_ip_checksum)
650 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
651 
652 	if (dev->data->dev_conf.rxmode.enable_lro) {
653 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
654 		devRead->misc.maxNumRxSG = 0;
655 	}
656 
657 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
658 		ret = vmxnet3_rss_configure(dev);
659 		if (ret != VMXNET3_SUCCESS)
660 			return ret;
661 
662 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
663 		devRead->rssConfDesc.confVer = 1;
664 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
665 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
666 	}
667 
668 	ret = vmxnet3_dev_vlan_offload_set(dev,
669 			ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
670 	if (ret)
671 		return ret;
672 
673 	vmxnet3_write_mac(hw, dev->data->mac_addrs->addr_bytes);
674 
675 	return VMXNET3_SUCCESS;
676 }
677 
678 /*
679  * Configure device link speed and setup link.
680  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
681  * It returns 0 on success.
682  */
683 static int
684 vmxnet3_dev_start(struct rte_eth_dev *dev)
685 {
686 	int ret;
687 	struct vmxnet3_hw *hw = dev->data->dev_private;
688 
689 	PMD_INIT_FUNC_TRACE();
690 
691 	/* Save stats before it is reset by CMD_ACTIVATE */
692 	vmxnet3_hw_stats_save(hw);
693 
694 	ret = vmxnet3_setup_driver_shared(dev);
695 	if (ret != VMXNET3_SUCCESS)
696 		return ret;
697 
698 	/* check if lsc interrupt feature is enabled */
699 	if (dev->data->dev_conf.intr_conf.lsc) {
700 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
701 
702 		/* Setup interrupt callback  */
703 		rte_intr_callback_register(&pci_dev->intr_handle,
704 					   vmxnet3_interrupt_handler, dev);
705 
706 		if (rte_intr_enable(&pci_dev->intr_handle) < 0) {
707 			PMD_INIT_LOG(ERR, "interrupt enable failed");
708 			return -EIO;
709 		}
710 	}
711 
712 	/* Exchange shared data with device */
713 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
714 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
715 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
716 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
717 
718 	/* Activate device by register write */
719 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
720 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
721 
722 	if (ret != 0) {
723 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
724 		return -EINVAL;
725 	}
726 
727 	/* Setup memory region for rx buffers */
728 	ret = vmxnet3_dev_setup_memreg(dev);
729 	if (ret == 0) {
730 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
731 				       VMXNET3_CMD_REGISTER_MEMREGS);
732 		ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
733 		if (ret != 0)
734 			PMD_INIT_LOG(DEBUG,
735 				     "Failed in setup memory region cmd\n");
736 		ret = 0;
737 	} else {
738 		PMD_INIT_LOG(DEBUG, "Failed to setup memory region\n");
739 	}
740 
741 	/* Disable interrupts */
742 	vmxnet3_disable_intr(hw);
743 
744 	/*
745 	 * Load RX queues with blank mbufs and update next2fill index for device
746 	 * Update RxMode of the device
747 	 */
748 	ret = vmxnet3_dev_rxtx_init(dev);
749 	if (ret != VMXNET3_SUCCESS) {
750 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
751 		return ret;
752 	}
753 
754 	hw->adapter_stopped = FALSE;
755 
756 	/* Setting proper Rx Mode and issue Rx Mode Update command */
757 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
758 
759 	if (dev->data->dev_conf.intr_conf.lsc) {
760 		vmxnet3_enable_intr(hw);
761 
762 		/*
763 		 * Update link state from device since this won't be
764 		 * done upon starting with lsc in use. This is done
765 		 * only after enabling interrupts to avoid any race
766 		 * where the link state could change without an
767 		 * interrupt being fired.
768 		 */
769 		__vmxnet3_dev_link_update(dev, 0);
770 	}
771 
772 	return VMXNET3_SUCCESS;
773 }
774 
775 /*
776  * Stop device: disable rx and tx functions to allow for reconfiguring.
777  */
778 static void
779 vmxnet3_dev_stop(struct rte_eth_dev *dev)
780 {
781 	struct rte_eth_link link;
782 	struct vmxnet3_hw *hw = dev->data->dev_private;
783 
784 	PMD_INIT_FUNC_TRACE();
785 
786 	if (hw->adapter_stopped == 1) {
787 		PMD_INIT_LOG(DEBUG, "Device already closed.");
788 		return;
789 	}
790 
791 	/* disable interrupts */
792 	vmxnet3_disable_intr(hw);
793 
794 	if (dev->data->dev_conf.intr_conf.lsc) {
795 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
796 
797 		rte_intr_disable(&pci_dev->intr_handle);
798 
799 		rte_intr_callback_unregister(&pci_dev->intr_handle,
800 					     vmxnet3_interrupt_handler, dev);
801 	}
802 
803 	/* quiesce the device first */
804 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
805 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
806 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
807 
808 	/* reset the device */
809 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
810 	PMD_INIT_LOG(DEBUG, "Device reset.");
811 	hw->adapter_stopped = 0;
812 
813 	vmxnet3_dev_clear_queues(dev);
814 
815 	/* Clear recorded link status */
816 	memset(&link, 0, sizeof(link));
817 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
818 	link.link_speed = ETH_SPEED_NUM_10G;
819 	link.link_autoneg = ETH_LINK_FIXED;
820 	rte_eth_linkstatus_set(dev, &link);
821 }
822 
823 /*
824  * Reset and stop device.
825  */
826 static void
827 vmxnet3_dev_close(struct rte_eth_dev *dev)
828 {
829 	struct vmxnet3_hw *hw = dev->data->dev_private;
830 
831 	PMD_INIT_FUNC_TRACE();
832 
833 	vmxnet3_dev_stop(dev);
834 	hw->adapter_stopped = 1;
835 }
836 
837 static void
838 vmxnet3_hw_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
839 			struct UPT1_TxStats *res)
840 {
841 #define VMXNET3_UPDATE_TX_STAT(h, i, f, r)		\
842 		((r)->f = (h)->tqd_start[(i)].stats.f +	\
843 			(h)->saved_tx_stats[(i)].f)
844 
845 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastPktsTxOK, res);
846 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastPktsTxOK, res);
847 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastPktsTxOK, res);
848 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastBytesTxOK, res);
849 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastBytesTxOK, res);
850 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastBytesTxOK, res);
851 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxError, res);
852 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxDiscard, res);
853 
854 #undef VMXNET3_UPDATE_TX_STAT
855 }
856 
857 static void
858 vmxnet3_hw_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
859 			struct UPT1_RxStats *res)
860 {
861 #define VMXNET3_UPDATE_RX_STAT(h, i, f, r)		\
862 		((r)->f = (h)->rqd_start[(i)].stats.f +	\
863 			(h)->saved_rx_stats[(i)].f)
864 
865 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastPktsRxOK, res);
866 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastPktsRxOK, res);
867 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastPktsRxOK, res);
868 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastBytesRxOK, res);
869 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastBytesRxOK, res);
870 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastBytesRxOK, res);
871 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxError, res);
872 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxOutOfBuf, res);
873 
874 #undef VMXNET3_UPDATE_RX_STATS
875 }
876 
877 static void
878 vmxnet3_hw_stats_save(struct vmxnet3_hw *hw)
879 {
880 	unsigned int i;
881 
882 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
883 
884 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
885 
886 	for (i = 0; i < hw->num_tx_queues; i++)
887 		vmxnet3_hw_tx_stats_get(hw, i, &hw->saved_tx_stats[i]);
888 	for (i = 0; i < hw->num_rx_queues; i++)
889 		vmxnet3_hw_rx_stats_get(hw, i, &hw->saved_rx_stats[i]);
890 }
891 
892 static int
893 vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
894 			     struct rte_eth_xstat_name *xstats_names,
895 			     unsigned int n)
896 {
897 	unsigned int i, t, count = 0;
898 	unsigned int nstats =
899 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
900 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
901 
902 	if (!xstats_names || n < nstats)
903 		return nstats;
904 
905 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
906 		if (!dev->data->rx_queues[i])
907 			continue;
908 
909 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
910 			snprintf(xstats_names[count].name,
911 				 sizeof(xstats_names[count].name),
912 				 "rx_q%u_%s", i,
913 				 vmxnet3_rxq_stat_strings[t].name);
914 			count++;
915 		}
916 	}
917 
918 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
919 		if (!dev->data->tx_queues[i])
920 			continue;
921 
922 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
923 			snprintf(xstats_names[count].name,
924 				 sizeof(xstats_names[count].name),
925 				 "tx_q%u_%s", i,
926 				 vmxnet3_txq_stat_strings[t].name);
927 			count++;
928 		}
929 	}
930 
931 	return count;
932 }
933 
934 static int
935 vmxnet3_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
936 		       unsigned int n)
937 {
938 	unsigned int i, t, count = 0;
939 	unsigned int nstats =
940 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
941 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
942 
943 	if (n < nstats)
944 		return nstats;
945 
946 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
947 		struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
948 
949 		if (rxq == NULL)
950 			continue;
951 
952 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
953 			xstats[count].value = *(uint64_t *)(((char *)&rxq->stats) +
954 				vmxnet3_rxq_stat_strings[t].offset);
955 			xstats[count].id = count;
956 			count++;
957 		}
958 	}
959 
960 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
961 		struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
962 
963 		if (txq == NULL)
964 			continue;
965 
966 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
967 			xstats[count].value = *(uint64_t *)(((char *)&txq->stats) +
968 				vmxnet3_txq_stat_strings[t].offset);
969 			xstats[count].id = count;
970 			count++;
971 		}
972 	}
973 
974 	return count;
975 }
976 
977 static int
978 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
979 {
980 	unsigned int i;
981 	struct vmxnet3_hw *hw = dev->data->dev_private;
982 	struct UPT1_TxStats txStats;
983 	struct UPT1_RxStats rxStats;
984 
985 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
986 
987 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
988 	for (i = 0; i < hw->num_tx_queues; i++) {
989 		vmxnet3_hw_tx_stats_get(hw, i, &txStats);
990 
991 		stats->q_opackets[i] = txStats.ucastPktsTxOK +
992 			txStats.mcastPktsTxOK +
993 			txStats.bcastPktsTxOK;
994 
995 		stats->q_obytes[i] = txStats.ucastBytesTxOK +
996 			txStats.mcastBytesTxOK +
997 			txStats.bcastBytesTxOK;
998 
999 		stats->opackets += stats->q_opackets[i];
1000 		stats->obytes += stats->q_obytes[i];
1001 		stats->oerrors += txStats.pktsTxError + txStats.pktsTxDiscard;
1002 	}
1003 
1004 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
1005 	for (i = 0; i < hw->num_rx_queues; i++) {
1006 		vmxnet3_hw_rx_stats_get(hw, i, &rxStats);
1007 
1008 		stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
1009 			rxStats.mcastPktsRxOK +
1010 			rxStats.bcastPktsRxOK;
1011 
1012 		stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
1013 			rxStats.mcastBytesRxOK +
1014 			rxStats.bcastBytesRxOK;
1015 
1016 		stats->ipackets += stats->q_ipackets[i];
1017 		stats->ibytes += stats->q_ibytes[i];
1018 
1019 		stats->q_errors[i] = rxStats.pktsRxError;
1020 		stats->ierrors += rxStats.pktsRxError;
1021 		stats->imissed += rxStats.pktsRxOutOfBuf;
1022 	}
1023 
1024 	return 0;
1025 }
1026 
1027 static void
1028 vmxnet3_dev_info_get(struct rte_eth_dev *dev __rte_unused,
1029 		     struct rte_eth_dev_info *dev_info)
1030 {
1031 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
1032 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
1033 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
1034 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
1035 	dev_info->speed_capa = ETH_LINK_SPEED_10G;
1036 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
1037 
1038 	dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
1039 	dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
1040 
1041 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1042 		.nb_max = VMXNET3_RX_RING_MAX_SIZE,
1043 		.nb_min = VMXNET3_DEF_RX_RING_SIZE,
1044 		.nb_align = 1,
1045 	};
1046 
1047 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1048 		.nb_max = VMXNET3_TX_RING_MAX_SIZE,
1049 		.nb_min = VMXNET3_DEF_TX_RING_SIZE,
1050 		.nb_align = 1,
1051 		.nb_seg_max = VMXNET3_TX_MAX_SEG,
1052 		.nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
1053 	};
1054 
1055 	dev_info->rx_offload_capa =
1056 		DEV_RX_OFFLOAD_VLAN_STRIP |
1057 		DEV_RX_OFFLOAD_UDP_CKSUM |
1058 		DEV_RX_OFFLOAD_TCP_CKSUM |
1059 		DEV_RX_OFFLOAD_TCP_LRO;
1060 
1061 	dev_info->tx_offload_capa =
1062 		DEV_TX_OFFLOAD_VLAN_INSERT |
1063 		DEV_TX_OFFLOAD_TCP_CKSUM |
1064 		DEV_TX_OFFLOAD_UDP_CKSUM |
1065 		DEV_TX_OFFLOAD_TCP_TSO;
1066 }
1067 
1068 static const uint32_t *
1069 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
1070 {
1071 	static const uint32_t ptypes[] = {
1072 		RTE_PTYPE_L3_IPV4_EXT,
1073 		RTE_PTYPE_L3_IPV4,
1074 		RTE_PTYPE_UNKNOWN
1075 	};
1076 
1077 	if (dev->rx_pkt_burst == vmxnet3_recv_pkts)
1078 		return ptypes;
1079 	return NULL;
1080 }
1081 
1082 static int
1083 vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
1084 {
1085 	struct vmxnet3_hw *hw = dev->data->dev_private;
1086 
1087 	ether_addr_copy(mac_addr, (struct ether_addr *)(hw->perm_addr));
1088 	vmxnet3_write_mac(hw, mac_addr->addr_bytes);
1089 	return 0;
1090 }
1091 
1092 /* return 0 means link status changed, -1 means not changed */
1093 static int
1094 __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
1095 			  __rte_unused int wait_to_complete)
1096 {
1097 	struct vmxnet3_hw *hw = dev->data->dev_private;
1098 	struct rte_eth_link link;
1099 	uint32_t ret;
1100 
1101 	memset(&link, 0, sizeof(link));
1102 
1103 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
1104 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
1105 
1106 	if (ret & 0x1)
1107 		link.link_status = ETH_LINK_UP;
1108 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
1109 	link.link_speed = ETH_SPEED_NUM_10G;
1110 	link.link_autoneg = ETH_LINK_FIXED;
1111 
1112 	return rte_eth_linkstatus_set(dev, &link);
1113 }
1114 
1115 static int
1116 vmxnet3_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1117 {
1118 	/* Link status doesn't change for stopped dev */
1119 	if (dev->data->dev_started == 0)
1120 		return -1;
1121 
1122 	return __vmxnet3_dev_link_update(dev, wait_to_complete);
1123 }
1124 
1125 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
1126 static void
1127 vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set)
1128 {
1129 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1130 
1131 	if (set)
1132 		rxConf->rxMode = rxConf->rxMode | feature;
1133 	else
1134 		rxConf->rxMode = rxConf->rxMode & (~feature);
1135 
1136 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_UPDATE_RX_MODE);
1137 }
1138 
1139 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1140 static void
1141 vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev)
1142 {
1143 	struct vmxnet3_hw *hw = dev->data->dev_private;
1144 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1145 
1146 	memset(vf_table, 0, VMXNET3_VFT_TABLE_SIZE);
1147 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 1);
1148 
1149 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1150 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1151 }
1152 
1153 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1154 static void
1155 vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
1156 {
1157 	struct vmxnet3_hw *hw = dev->data->dev_private;
1158 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1159 
1160 	if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1161 		memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1162 	else
1163 		memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1164 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
1165 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1166 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1167 }
1168 
1169 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1170 static void
1171 vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev)
1172 {
1173 	struct vmxnet3_hw *hw = dev->data->dev_private;
1174 
1175 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 1);
1176 }
1177 
1178 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1179 static void
1180 vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev)
1181 {
1182 	struct vmxnet3_hw *hw = dev->data->dev_private;
1183 
1184 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 0);
1185 }
1186 
1187 /* Enable/disable filter on vlan */
1188 static int
1189 vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
1190 {
1191 	struct vmxnet3_hw *hw = dev->data->dev_private;
1192 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1193 	uint32_t *vf_table = rxConf->vfTable;
1194 
1195 	/* save state for restore */
1196 	if (on)
1197 		VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1198 	else
1199 		VMXNET3_CLEAR_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1200 
1201 	/* don't change active filter if in promiscuous mode */
1202 	if (rxConf->rxMode & VMXNET3_RXM_PROMISC)
1203 		return 0;
1204 
1205 	/* set in hardware */
1206 	if (on)
1207 		VMXNET3_SET_VFTABLE_ENTRY(vf_table, vid);
1208 	else
1209 		VMXNET3_CLEAR_VFTABLE_ENTRY(vf_table, vid);
1210 
1211 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1212 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1213 	return 0;
1214 }
1215 
1216 static int
1217 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1218 {
1219 	struct vmxnet3_hw *hw = dev->data->dev_private;
1220 	Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
1221 	uint32_t *vf_table = devRead->rxFilterConf.vfTable;
1222 
1223 	if (mask & ETH_VLAN_STRIP_MASK) {
1224 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1225 			devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
1226 		else
1227 			devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
1228 
1229 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1230 				       VMXNET3_CMD_UPDATE_FEATURE);
1231 	}
1232 
1233 	if (mask & ETH_VLAN_FILTER_MASK) {
1234 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1235 			memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1236 		else
1237 			memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1238 
1239 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1240 				       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1241 	}
1242 
1243 	return 0;
1244 }
1245 
1246 static void
1247 vmxnet3_process_events(struct rte_eth_dev *dev)
1248 {
1249 	struct vmxnet3_hw *hw = dev->data->dev_private;
1250 	uint32_t events = hw->shared->ecr;
1251 
1252 	if (!events)
1253 		return;
1254 
1255 	/*
1256 	 * ECR bits when written with 1b are cleared. Hence write
1257 	 * events back to ECR so that the bits which were set will be reset.
1258 	 */
1259 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
1260 
1261 	/* Check if link state has changed */
1262 	if (events & VMXNET3_ECR_LINK) {
1263 		PMD_DRV_LOG(DEBUG, "Process events: VMXNET3_ECR_LINK event");
1264 		if (vmxnet3_dev_link_update(dev, 0) == 0)
1265 			_rte_eth_dev_callback_process(dev,
1266 						      RTE_ETH_EVENT_INTR_LSC,
1267 						      NULL);
1268 	}
1269 
1270 	/* Check if there is an error on xmit/recv queues */
1271 	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
1272 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1273 				       VMXNET3_CMD_GET_QUEUE_STATUS);
1274 
1275 		if (hw->tqd_start->status.stopped)
1276 			PMD_DRV_LOG(ERR, "tq error 0x%x",
1277 				    hw->tqd_start->status.error);
1278 
1279 		if (hw->rqd_start->status.stopped)
1280 			PMD_DRV_LOG(ERR, "rq error 0x%x",
1281 				     hw->rqd_start->status.error);
1282 
1283 		/* Reset the device */
1284 		/* Have to reset the device */
1285 	}
1286 
1287 	if (events & VMXNET3_ECR_DIC)
1288 		PMD_DRV_LOG(DEBUG, "Device implementation change event.");
1289 
1290 	if (events & VMXNET3_ECR_DEBUG)
1291 		PMD_DRV_LOG(DEBUG, "Debug event generated by device.");
1292 }
1293 
1294 static void
1295 vmxnet3_interrupt_handler(void *param)
1296 {
1297 	struct rte_eth_dev *dev = param;
1298 	struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
1299 
1300 	vmxnet3_process_events(dev);
1301 
1302 	if (rte_intr_enable(&pci_dev->intr_handle) < 0)
1303 		PMD_DRV_LOG(ERR, "interrupt enable failed");
1304 }
1305 
1306 RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd);
1307 RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
1308 RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio-pci");
1309 
1310 RTE_INIT(vmxnet3_init_log);
1311 static void
1312 vmxnet3_init_log(void)
1313 {
1314 	vmxnet3_logtype_init = rte_log_register("pmd.net.vmxnet3.init");
1315 	if (vmxnet3_logtype_init >= 0)
1316 		rte_log_set_level(vmxnet3_logtype_init, RTE_LOG_NOTICE);
1317 	vmxnet3_logtype_driver = rte_log_register("pmd.net.vmxnet3.driver");
1318 	if (vmxnet3_logtype_driver >= 0)
1319 		rte_log_set_level(vmxnet3_logtype_driver, RTE_LOG_NOTICE);
1320 }
1321