xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision bb44fb6fe7713ddcd023d5b9bacadf074d68092e)
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 #define VMXNET3_TX_OFFLOAD_CAP		\
46 	(DEV_TX_OFFLOAD_VLAN_INSERT |	\
47 	 DEV_TX_OFFLOAD_IPV4_CKSUM |	\
48 	 DEV_TX_OFFLOAD_TCP_CKSUM |	\
49 	 DEV_TX_OFFLOAD_UDP_CKSUM |	\
50 	 DEV_TX_OFFLOAD_TCP_TSO |	\
51 	 DEV_TX_OFFLOAD_MULTI_SEGS)
52 
53 #define VMXNET3_RX_OFFLOAD_CAP		\
54 	(DEV_RX_OFFLOAD_VLAN_STRIP |	\
55 	 DEV_RX_OFFLOAD_SCATTER |	\
56 	 DEV_RX_OFFLOAD_IPV4_CKSUM |	\
57 	 DEV_RX_OFFLOAD_UDP_CKSUM |	\
58 	 DEV_RX_OFFLOAD_TCP_CKSUM |	\
59 	 DEV_RX_OFFLOAD_TCP_LRO |	\
60 	 DEV_RX_OFFLOAD_JUMBO_FRAME |   \
61 	 DEV_RX_OFFLOAD_CRC_STRIP)
62 
63 static int eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev);
64 static int eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev);
65 static int vmxnet3_dev_configure(struct rte_eth_dev *dev);
66 static int vmxnet3_dev_start(struct rte_eth_dev *dev);
67 static void vmxnet3_dev_stop(struct rte_eth_dev *dev);
68 static void vmxnet3_dev_close(struct rte_eth_dev *dev);
69 static void vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set);
70 static void vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev);
71 static void vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev);
72 static void vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev);
73 static void vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev);
74 static int __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
75 				     int wait_to_complete);
76 static int vmxnet3_dev_link_update(struct rte_eth_dev *dev,
77 				   int wait_to_complete);
78 static void vmxnet3_hw_stats_save(struct vmxnet3_hw *hw);
79 static int vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
80 				  struct rte_eth_stats *stats);
81 static int vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
82 					struct rte_eth_xstat_name *xstats,
83 					unsigned int n);
84 static int vmxnet3_dev_xstats_get(struct rte_eth_dev *dev,
85 				  struct rte_eth_xstat *xstats, unsigned int n);
86 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
87 				 struct rte_eth_dev_info *dev_info);
88 static const uint32_t *
89 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
90 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
91 				       uint16_t vid, int on);
92 static int vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
93 static int vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
94 				 struct ether_addr *mac_addr);
95 static void vmxnet3_interrupt_handler(void *param);
96 
97 int vmxnet3_logtype_init;
98 int vmxnet3_logtype_driver;
99 
100 /*
101  * The set of PCI devices this driver supports
102  */
103 #define VMWARE_PCI_VENDOR_ID 0x15AD
104 #define VMWARE_DEV_ID_VMXNET3 0x07B0
105 static const struct rte_pci_id pci_id_vmxnet3_map[] = {
106 	{ RTE_PCI_DEVICE(VMWARE_PCI_VENDOR_ID, VMWARE_DEV_ID_VMXNET3) },
107 	{ .vendor_id = 0, /* sentinel */ },
108 };
109 
110 static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
111 	.dev_configure        = vmxnet3_dev_configure,
112 	.dev_start            = vmxnet3_dev_start,
113 	.dev_stop             = vmxnet3_dev_stop,
114 	.dev_close            = vmxnet3_dev_close,
115 	.promiscuous_enable   = vmxnet3_dev_promiscuous_enable,
116 	.promiscuous_disable  = vmxnet3_dev_promiscuous_disable,
117 	.allmulticast_enable  = vmxnet3_dev_allmulticast_enable,
118 	.allmulticast_disable = vmxnet3_dev_allmulticast_disable,
119 	.link_update          = vmxnet3_dev_link_update,
120 	.stats_get            = vmxnet3_dev_stats_get,
121 	.xstats_get_names     = vmxnet3_dev_xstats_get_names,
122 	.xstats_get           = vmxnet3_dev_xstats_get,
123 	.mac_addr_set         = vmxnet3_mac_addr_set,
124 	.dev_infos_get        = vmxnet3_dev_info_get,
125 	.dev_supported_ptypes_get = vmxnet3_dev_supported_ptypes_get,
126 	.vlan_filter_set      = vmxnet3_dev_vlan_filter_set,
127 	.vlan_offload_set     = vmxnet3_dev_vlan_offload_set,
128 	.rx_queue_setup       = vmxnet3_dev_rx_queue_setup,
129 	.rx_queue_release     = vmxnet3_dev_rx_queue_release,
130 	.tx_queue_setup       = vmxnet3_dev_tx_queue_setup,
131 	.tx_queue_release     = vmxnet3_dev_tx_queue_release,
132 };
133 
134 struct vmxnet3_xstats_name_off {
135 	char name[RTE_ETH_XSTATS_NAME_SIZE];
136 	unsigned int offset;
137 };
138 
139 /* tx_qX_ is prepended to the name string here */
140 static const struct vmxnet3_xstats_name_off vmxnet3_txq_stat_strings[] = {
141 	{"drop_total",         offsetof(struct vmxnet3_txq_stats, drop_total)},
142 	{"drop_too_many_segs", offsetof(struct vmxnet3_txq_stats, drop_too_many_segs)},
143 	{"drop_tso",           offsetof(struct vmxnet3_txq_stats, drop_tso)},
144 	{"tx_ring_full",       offsetof(struct vmxnet3_txq_stats, tx_ring_full)},
145 };
146 
147 /* rx_qX_ is prepended to the name string here */
148 static const struct vmxnet3_xstats_name_off vmxnet3_rxq_stat_strings[] = {
149 	{"drop_total",           offsetof(struct vmxnet3_rxq_stats, drop_total)},
150 	{"drop_err",             offsetof(struct vmxnet3_rxq_stats, drop_err)},
151 	{"drop_fcs",             offsetof(struct vmxnet3_rxq_stats, drop_fcs)},
152 	{"rx_buf_alloc_failure", offsetof(struct vmxnet3_rxq_stats, rx_buf_alloc_failure)},
153 };
154 
155 static const struct rte_memzone *
156 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
157 		 const char *post_string, int socket_id,
158 		 uint16_t align, bool reuse)
159 {
160 	char z_name[RTE_MEMZONE_NAMESIZE];
161 	const struct rte_memzone *mz;
162 
163 	snprintf(z_name, sizeof(z_name), "%s_%d_%s",
164 		 dev->device->driver->name, dev->data->port_id, post_string);
165 
166 	mz = rte_memzone_lookup(z_name);
167 	if (!reuse) {
168 		if (mz)
169 			rte_memzone_free(mz);
170 		return rte_memzone_reserve_aligned(z_name, size, socket_id,
171 				RTE_MEMZONE_IOVA_CONTIG, align);
172 	}
173 
174 	if (mz)
175 		return mz;
176 
177 	return rte_memzone_reserve_aligned(z_name, size, socket_id,
178 			RTE_MEMZONE_IOVA_CONTIG, align);
179 }
180 
181 /*
182  * This function is based on vmxnet3_disable_intr()
183  */
184 static void
185 vmxnet3_disable_intr(struct vmxnet3_hw *hw)
186 {
187 	int i;
188 
189 	PMD_INIT_FUNC_TRACE();
190 
191 	hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
192 	for (i = 0; i < hw->num_intrs; i++)
193 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
194 }
195 
196 static void
197 vmxnet3_enable_intr(struct vmxnet3_hw *hw)
198 {
199 	int i;
200 
201 	PMD_INIT_FUNC_TRACE();
202 
203 	hw->shared->devRead.intrConf.intrCtrl &= ~VMXNET3_IC_DISABLE_ALL;
204 	for (i = 0; i < hw->num_intrs; i++)
205 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 0);
206 }
207 
208 /*
209  * Gets tx data ring descriptor size.
210  */
211 static uint16_t
212 eth_vmxnet3_txdata_get(struct vmxnet3_hw *hw)
213 {
214 	uint16 txdata_desc_size;
215 
216 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
217 			       VMXNET3_CMD_GET_TXDATA_DESC_SIZE);
218 	txdata_desc_size = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
219 
220 	return (txdata_desc_size < VMXNET3_TXDATA_DESC_MIN_SIZE ||
221 		txdata_desc_size > VMXNET3_TXDATA_DESC_MAX_SIZE ||
222 		txdata_desc_size & VMXNET3_TXDATA_DESC_SIZE_MASK) ?
223 		sizeof(struct Vmxnet3_TxDataDesc) : txdata_desc_size;
224 }
225 
226 /*
227  * It returns 0 on success.
228  */
229 static int
230 eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
231 {
232 	struct rte_pci_device *pci_dev;
233 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
234 	uint32_t mac_hi, mac_lo, ver;
235 	struct rte_eth_link link;
236 
237 	PMD_INIT_FUNC_TRACE();
238 
239 	eth_dev->dev_ops = &vmxnet3_eth_dev_ops;
240 	eth_dev->rx_pkt_burst = &vmxnet3_recv_pkts;
241 	eth_dev->tx_pkt_burst = &vmxnet3_xmit_pkts;
242 	eth_dev->tx_pkt_prepare = vmxnet3_prep_pkts;
243 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
244 
245 	/*
246 	 * for secondary processes, we don't initialize any further as primary
247 	 * has already done this work.
248 	 */
249 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
250 		return 0;
251 
252 	rte_eth_copy_pci_info(eth_dev, pci_dev);
253 
254 	/* Vendor and Device ID need to be set before init of shared code */
255 	hw->device_id = pci_dev->id.device_id;
256 	hw->vendor_id = pci_dev->id.vendor_id;
257 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
258 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
259 
260 	hw->num_rx_queues = 1;
261 	hw->num_tx_queues = 1;
262 	hw->bufs_per_pkt = 1;
263 
264 	/* Check h/w version compatibility with driver. */
265 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
266 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
267 
268 	if (ver & (1 << VMXNET3_REV_3)) {
269 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
270 				       1 << VMXNET3_REV_3);
271 		hw->version = VMXNET3_REV_3 + 1;
272 	} else if (ver & (1 << VMXNET3_REV_2)) {
273 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
274 				       1 << VMXNET3_REV_2);
275 		hw->version = VMXNET3_REV_2 + 1;
276 	} else if (ver & (1 << VMXNET3_REV_1)) {
277 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
278 				       1 << VMXNET3_REV_1);
279 		hw->version = VMXNET3_REV_1 + 1;
280 	} else {
281 		PMD_INIT_LOG(ERR, "Incompatible hardware version: %d", ver);
282 		return -EIO;
283 	}
284 
285 	PMD_INIT_LOG(DEBUG, "Using device version %d\n", hw->version);
286 
287 	/* Check UPT version compatibility with driver. */
288 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
289 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
290 	if (ver & 0x1)
291 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
292 	else {
293 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
294 		return -EIO;
295 	}
296 
297 	/* Getting MAC Address */
298 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
299 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
300 	memcpy(hw->perm_addr, &mac_lo, 4);
301 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
302 
303 	/* Allocate memory for storing MAC addresses */
304 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
305 					       VMXNET3_MAX_MAC_ADDRS, 0);
306 	if (eth_dev->data->mac_addrs == NULL) {
307 		PMD_INIT_LOG(ERR,
308 			     "Failed to allocate %d bytes needed to store MAC addresses",
309 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
310 		return -ENOMEM;
311 	}
312 	/* Copy the permanent MAC address */
313 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
314 			&eth_dev->data->mac_addrs[0]);
315 
316 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
317 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
318 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
319 
320 	/* Put device in Quiesce Mode */
321 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
322 
323 	/* allow untagged pkts */
324 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
325 
326 	hw->txdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
327 		eth_vmxnet3_txdata_get(hw) : sizeof(struct Vmxnet3_TxDataDesc);
328 
329 	hw->rxdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
330 		VMXNET3_DEF_RXDATA_DESC_SIZE : 0;
331 	RTE_ASSERT((hw->rxdata_desc_size & ~VMXNET3_RXDATA_DESC_SIZE_MASK) ==
332 		   hw->rxdata_desc_size);
333 
334 	/* clear shadow stats */
335 	memset(hw->saved_tx_stats, 0, sizeof(hw->saved_tx_stats));
336 	memset(hw->saved_rx_stats, 0, sizeof(hw->saved_rx_stats));
337 
338 	/* set the initial link status */
339 	memset(&link, 0, sizeof(link));
340 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
341 	link.link_speed = ETH_SPEED_NUM_10G;
342 	link.link_autoneg = ETH_LINK_FIXED;
343 	rte_eth_linkstatus_set(eth_dev, &link);
344 
345 	return 0;
346 }
347 
348 static int
349 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
350 {
351 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
352 
353 	PMD_INIT_FUNC_TRACE();
354 
355 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
356 		return 0;
357 
358 	if (hw->adapter_stopped == 0)
359 		vmxnet3_dev_close(eth_dev);
360 
361 	eth_dev->dev_ops = NULL;
362 	eth_dev->rx_pkt_burst = NULL;
363 	eth_dev->tx_pkt_burst = NULL;
364 	eth_dev->tx_pkt_prepare = NULL;
365 
366 	rte_free(eth_dev->data->mac_addrs);
367 	eth_dev->data->mac_addrs = NULL;
368 
369 	return 0;
370 }
371 
372 static int eth_vmxnet3_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
373 	struct rte_pci_device *pci_dev)
374 {
375 	return rte_eth_dev_pci_generic_probe(pci_dev,
376 		sizeof(struct vmxnet3_hw), eth_vmxnet3_dev_init);
377 }
378 
379 static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
380 {
381 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_vmxnet3_dev_uninit);
382 }
383 
384 static struct rte_pci_driver rte_vmxnet3_pmd = {
385 	.id_table = pci_id_vmxnet3_map,
386 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
387 	.probe = eth_vmxnet3_pci_probe,
388 	.remove = eth_vmxnet3_pci_remove,
389 };
390 
391 static int
392 vmxnet3_dev_configure(struct rte_eth_dev *dev)
393 {
394 	const struct rte_memzone *mz;
395 	struct vmxnet3_hw *hw = dev->data->dev_private;
396 	size_t size;
397 
398 	PMD_INIT_FUNC_TRACE();
399 
400 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
401 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
402 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
403 		return -EINVAL;
404 	}
405 
406 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
407 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
408 		return -EINVAL;
409 	}
410 
411 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
412 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
413 
414 	if (size > UINT16_MAX)
415 		return -EINVAL;
416 
417 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
418 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
419 
420 	/*
421 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
422 	 * on current socket
423 	 */
424 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
425 			      "shared", rte_socket_id(), 8, 1);
426 
427 	if (mz == NULL) {
428 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
429 		return -ENOMEM;
430 	}
431 	memset(mz->addr, 0, mz->len);
432 
433 	hw->shared = mz->addr;
434 	hw->sharedPA = mz->iova;
435 
436 	/*
437 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
438 	 * on current socket.
439 	 *
440 	 * We cannot reuse this memzone from previous allocation as its size
441 	 * depends on the number of tx and rx queues, which could be different
442 	 * from one config to another.
443 	 */
444 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
445 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
446 	if (mz == NULL) {
447 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
448 		return -ENOMEM;
449 	}
450 	memset(mz->addr, 0, mz->len);
451 
452 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
453 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
454 
455 	hw->queueDescPA = mz->iova;
456 	hw->queue_desc_len = (uint16_t)size;
457 
458 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
459 		/* Allocate memory structure for UPT1_RSSConf and configure */
460 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
461 				      "rss_conf", rte_socket_id(),
462 				      RTE_CACHE_LINE_SIZE, 1);
463 		if (mz == NULL) {
464 			PMD_INIT_LOG(ERR,
465 				     "ERROR: Creating rss_conf structure zone");
466 			return -ENOMEM;
467 		}
468 		memset(mz->addr, 0, mz->len);
469 
470 		hw->rss_conf = mz->addr;
471 		hw->rss_confPA = mz->iova;
472 	}
473 
474 	return 0;
475 }
476 
477 static void
478 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
479 {
480 	uint32_t val;
481 
482 	PMD_INIT_LOG(DEBUG,
483 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
484 		     addr[0], addr[1], addr[2],
485 		     addr[3], addr[4], addr[5]);
486 
487 	memcpy(&val, addr, 4);
488 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
489 
490 	memcpy(&val, addr + 4, 2);
491 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
492 }
493 
494 static int
495 vmxnet3_dev_setup_memreg(struct rte_eth_dev *dev)
496 {
497 	struct vmxnet3_hw *hw = dev->data->dev_private;
498 	Vmxnet3_DriverShared *shared = hw->shared;
499 	Vmxnet3_CmdInfo *cmdInfo;
500 	struct rte_mempool *mp[VMXNET3_MAX_RX_QUEUES];
501 	uint8_t index[VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES];
502 	uint32_t num, i, j, size;
503 
504 	if (hw->memRegsPA == 0) {
505 		const struct rte_memzone *mz;
506 
507 		size = sizeof(Vmxnet3_MemRegs) +
508 			(VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES) *
509 			sizeof(Vmxnet3_MemoryRegion);
510 
511 		mz = gpa_zone_reserve(dev, size, "memRegs", rte_socket_id(), 8,
512 				      1);
513 		if (mz == NULL) {
514 			PMD_INIT_LOG(ERR, "ERROR: Creating memRegs zone");
515 			return -ENOMEM;
516 		}
517 		memset(mz->addr, 0, mz->len);
518 		hw->memRegs = mz->addr;
519 		hw->memRegsPA = mz->iova;
520 	}
521 
522 	num = hw->num_rx_queues;
523 
524 	for (i = 0; i < num; i++) {
525 		vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
526 
527 		mp[i] = rxq->mp;
528 		index[i] = 1 << i;
529 	}
530 
531 	/*
532 	 * The same mempool could be used by multiple queues. In such a case,
533 	 * remove duplicate mempool entries. Only one entry is kept with
534 	 * bitmask indicating queues that are using this mempool.
535 	 */
536 	for (i = 1; i < num; i++) {
537 		for (j = 0; j < i; j++) {
538 			if (mp[i] == mp[j]) {
539 				mp[i] = NULL;
540 				index[j] |= 1 << i;
541 				break;
542 			}
543 		}
544 	}
545 
546 	j = 0;
547 	for (i = 0; i < num; i++) {
548 		if (mp[i] == NULL)
549 			continue;
550 
551 		Vmxnet3_MemoryRegion *mr = &hw->memRegs->memRegs[j];
552 
553 		mr->startPA =
554 			(uintptr_t)STAILQ_FIRST(&mp[i]->mem_list)->iova;
555 		mr->length = STAILQ_FIRST(&mp[i]->mem_list)->len <= INT32_MAX ?
556 			STAILQ_FIRST(&mp[i]->mem_list)->len : INT32_MAX;
557 		mr->txQueueBits = index[i];
558 		mr->rxQueueBits = index[i];
559 
560 		PMD_INIT_LOG(INFO,
561 			     "index: %u startPA: %" PRIu64 " length: %u, "
562 			     "rxBits: %x",
563 			     j, mr->startPA, mr->length, mr->rxQueueBits);
564 		j++;
565 	}
566 	hw->memRegs->numRegs = j;
567 	PMD_INIT_LOG(INFO, "numRegs: %u", j);
568 
569 	size = sizeof(Vmxnet3_MemRegs) +
570 		(j - 1) * sizeof(Vmxnet3_MemoryRegion);
571 
572 	cmdInfo = &shared->cu.cmdInfo;
573 	cmdInfo->varConf.confVer = 1;
574 	cmdInfo->varConf.confLen = size;
575 	cmdInfo->varConf.confPA = hw->memRegsPA;
576 
577 	return 0;
578 }
579 
580 static int
581 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
582 {
583 	struct rte_eth_conf port_conf = dev->data->dev_conf;
584 	struct vmxnet3_hw *hw = dev->data->dev_private;
585 	uint32_t mtu = dev->data->mtu;
586 	Vmxnet3_DriverShared *shared = hw->shared;
587 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
588 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
589 	uint32_t i;
590 	int ret;
591 
592 	hw->mtu = mtu;
593 
594 	shared->magic = VMXNET3_REV1_MAGIC;
595 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
596 
597 	/* Setting up Guest OS information */
598 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
599 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
600 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
601 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
602 	devRead->misc.driverInfo.uptVerSpt     = 1;
603 
604 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
605 	devRead->misc.queueDescPA  = hw->queueDescPA;
606 	devRead->misc.queueDescLen = hw->queue_desc_len;
607 	devRead->misc.numTxQueues  = hw->num_tx_queues;
608 	devRead->misc.numRxQueues  = hw->num_rx_queues;
609 
610 	/*
611 	 * Set number of interrupts to 1
612 	 * PMD by default disables all the interrupts but this is MUST
613 	 * to activate device. It needs at least one interrupt for
614 	 * link events to handle
615 	 */
616 	hw->num_intrs = devRead->intrConf.numIntrs = 1;
617 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
618 
619 	for (i = 0; i < hw->num_tx_queues; i++) {
620 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
621 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
622 
623 		txq->shared = &hw->tqd_start[i];
624 
625 		tqd->ctrl.txNumDeferred  = 0;
626 		tqd->ctrl.txThreshold    = 1;
627 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
628 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
629 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
630 
631 		tqd->conf.txRingSize   = txq->cmd_ring.size;
632 		tqd->conf.compRingSize = txq->comp_ring.size;
633 		tqd->conf.dataRingSize = txq->data_ring.size;
634 		tqd->conf.txDataRingDescSize = txq->txdata_desc_size;
635 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
636 		tqd->status.stopped    = TRUE;
637 		tqd->status.error      = 0;
638 		memset(&tqd->stats, 0, sizeof(tqd->stats));
639 	}
640 
641 	for (i = 0; i < hw->num_rx_queues; i++) {
642 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
643 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
644 
645 		rxq->shared = &hw->rqd_start[i];
646 
647 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
648 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
649 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
650 
651 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
652 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
653 		rqd->conf.compRingSize    = rxq->comp_ring.size;
654 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
655 		if (VMXNET3_VERSION_GE_3(hw)) {
656 			rqd->conf.rxDataRingBasePA = rxq->data_ring.basePA;
657 			rqd->conf.rxDataRingDescSize = rxq->data_desc_size;
658 		}
659 		rqd->status.stopped       = TRUE;
660 		rqd->status.error         = 0;
661 		memset(&rqd->stats, 0, sizeof(rqd->stats));
662 	}
663 
664 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
665 	devRead->rxFilterConf.rxMode = 0;
666 
667 	/* Setting up feature flags */
668 	if (rx_offloads & DEV_RX_OFFLOAD_CHECKSUM)
669 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
670 
671 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) {
672 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
673 		devRead->misc.maxNumRxSG = 0;
674 	}
675 
676 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
677 		ret = vmxnet3_rss_configure(dev);
678 		if (ret != VMXNET3_SUCCESS)
679 			return ret;
680 
681 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
682 		devRead->rssConfDesc.confVer = 1;
683 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
684 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
685 	}
686 
687 	ret = vmxnet3_dev_vlan_offload_set(dev,
688 			ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
689 	if (ret)
690 		return ret;
691 
692 	vmxnet3_write_mac(hw, dev->data->mac_addrs->addr_bytes);
693 
694 	return VMXNET3_SUCCESS;
695 }
696 
697 /*
698  * Configure device link speed and setup link.
699  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
700  * It returns 0 on success.
701  */
702 static int
703 vmxnet3_dev_start(struct rte_eth_dev *dev)
704 {
705 	int ret;
706 	struct vmxnet3_hw *hw = dev->data->dev_private;
707 
708 	PMD_INIT_FUNC_TRACE();
709 
710 	/* Save stats before it is reset by CMD_ACTIVATE */
711 	vmxnet3_hw_stats_save(hw);
712 
713 	ret = vmxnet3_setup_driver_shared(dev);
714 	if (ret != VMXNET3_SUCCESS)
715 		return ret;
716 
717 	/* check if lsc interrupt feature is enabled */
718 	if (dev->data->dev_conf.intr_conf.lsc) {
719 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
720 
721 		/* Setup interrupt callback  */
722 		rte_intr_callback_register(&pci_dev->intr_handle,
723 					   vmxnet3_interrupt_handler, dev);
724 
725 		if (rte_intr_enable(&pci_dev->intr_handle) < 0) {
726 			PMD_INIT_LOG(ERR, "interrupt enable failed");
727 			return -EIO;
728 		}
729 	}
730 
731 	/* Exchange shared data with device */
732 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
733 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
734 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
735 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
736 
737 	/* Activate device by register write */
738 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
739 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
740 
741 	if (ret != 0) {
742 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
743 		return -EINVAL;
744 	}
745 
746 	/* Setup memory region for rx buffers */
747 	ret = vmxnet3_dev_setup_memreg(dev);
748 	if (ret == 0) {
749 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
750 				       VMXNET3_CMD_REGISTER_MEMREGS);
751 		ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
752 		if (ret != 0)
753 			PMD_INIT_LOG(DEBUG,
754 				     "Failed in setup memory region cmd\n");
755 		ret = 0;
756 	} else {
757 		PMD_INIT_LOG(DEBUG, "Failed to setup memory region\n");
758 	}
759 
760 	/* Disable interrupts */
761 	vmxnet3_disable_intr(hw);
762 
763 	/*
764 	 * Load RX queues with blank mbufs and update next2fill index for device
765 	 * Update RxMode of the device
766 	 */
767 	ret = vmxnet3_dev_rxtx_init(dev);
768 	if (ret != VMXNET3_SUCCESS) {
769 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
770 		return ret;
771 	}
772 
773 	hw->adapter_stopped = FALSE;
774 
775 	/* Setting proper Rx Mode and issue Rx Mode Update command */
776 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
777 
778 	if (dev->data->dev_conf.intr_conf.lsc) {
779 		vmxnet3_enable_intr(hw);
780 
781 		/*
782 		 * Update link state from device since this won't be
783 		 * done upon starting with lsc in use. This is done
784 		 * only after enabling interrupts to avoid any race
785 		 * where the link state could change without an
786 		 * interrupt being fired.
787 		 */
788 		__vmxnet3_dev_link_update(dev, 0);
789 	}
790 
791 	return VMXNET3_SUCCESS;
792 }
793 
794 /*
795  * Stop device: disable rx and tx functions to allow for reconfiguring.
796  */
797 static void
798 vmxnet3_dev_stop(struct rte_eth_dev *dev)
799 {
800 	struct rte_eth_link link;
801 	struct vmxnet3_hw *hw = dev->data->dev_private;
802 
803 	PMD_INIT_FUNC_TRACE();
804 
805 	if (hw->adapter_stopped == 1) {
806 		PMD_INIT_LOG(DEBUG, "Device already closed.");
807 		return;
808 	}
809 
810 	/* disable interrupts */
811 	vmxnet3_disable_intr(hw);
812 
813 	if (dev->data->dev_conf.intr_conf.lsc) {
814 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
815 
816 		rte_intr_disable(&pci_dev->intr_handle);
817 
818 		rte_intr_callback_unregister(&pci_dev->intr_handle,
819 					     vmxnet3_interrupt_handler, dev);
820 	}
821 
822 	/* quiesce the device first */
823 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
824 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
825 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
826 
827 	/* reset the device */
828 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
829 	PMD_INIT_LOG(DEBUG, "Device reset.");
830 	hw->adapter_stopped = 0;
831 
832 	vmxnet3_dev_clear_queues(dev);
833 
834 	/* Clear recorded link status */
835 	memset(&link, 0, sizeof(link));
836 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
837 	link.link_speed = ETH_SPEED_NUM_10G;
838 	link.link_autoneg = ETH_LINK_FIXED;
839 	rte_eth_linkstatus_set(dev, &link);
840 }
841 
842 /*
843  * Reset and stop device.
844  */
845 static void
846 vmxnet3_dev_close(struct rte_eth_dev *dev)
847 {
848 	struct vmxnet3_hw *hw = dev->data->dev_private;
849 
850 	PMD_INIT_FUNC_TRACE();
851 
852 	vmxnet3_dev_stop(dev);
853 	hw->adapter_stopped = 1;
854 }
855 
856 static void
857 vmxnet3_hw_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
858 			struct UPT1_TxStats *res)
859 {
860 #define VMXNET3_UPDATE_TX_STAT(h, i, f, r)		\
861 		((r)->f = (h)->tqd_start[(i)].stats.f +	\
862 			(h)->saved_tx_stats[(i)].f)
863 
864 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastPktsTxOK, res);
865 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastPktsTxOK, res);
866 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastPktsTxOK, res);
867 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastBytesTxOK, res);
868 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastBytesTxOK, res);
869 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastBytesTxOK, res);
870 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxError, res);
871 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxDiscard, res);
872 
873 #undef VMXNET3_UPDATE_TX_STAT
874 }
875 
876 static void
877 vmxnet3_hw_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
878 			struct UPT1_RxStats *res)
879 {
880 #define VMXNET3_UPDATE_RX_STAT(h, i, f, r)		\
881 		((r)->f = (h)->rqd_start[(i)].stats.f +	\
882 			(h)->saved_rx_stats[(i)].f)
883 
884 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastPktsRxOK, res);
885 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastPktsRxOK, res);
886 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastPktsRxOK, res);
887 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastBytesRxOK, res);
888 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastBytesRxOK, res);
889 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastBytesRxOK, res);
890 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxError, res);
891 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxOutOfBuf, res);
892 
893 #undef VMXNET3_UPDATE_RX_STATS
894 }
895 
896 static void
897 vmxnet3_hw_stats_save(struct vmxnet3_hw *hw)
898 {
899 	unsigned int i;
900 
901 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
902 
903 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
904 
905 	for (i = 0; i < hw->num_tx_queues; i++)
906 		vmxnet3_hw_tx_stats_get(hw, i, &hw->saved_tx_stats[i]);
907 	for (i = 0; i < hw->num_rx_queues; i++)
908 		vmxnet3_hw_rx_stats_get(hw, i, &hw->saved_rx_stats[i]);
909 }
910 
911 static int
912 vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
913 			     struct rte_eth_xstat_name *xstats_names,
914 			     unsigned int n)
915 {
916 	unsigned int i, t, count = 0;
917 	unsigned int nstats =
918 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
919 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
920 
921 	if (!xstats_names || n < nstats)
922 		return nstats;
923 
924 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
925 		if (!dev->data->rx_queues[i])
926 			continue;
927 
928 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
929 			snprintf(xstats_names[count].name,
930 				 sizeof(xstats_names[count].name),
931 				 "rx_q%u_%s", i,
932 				 vmxnet3_rxq_stat_strings[t].name);
933 			count++;
934 		}
935 	}
936 
937 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
938 		if (!dev->data->tx_queues[i])
939 			continue;
940 
941 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
942 			snprintf(xstats_names[count].name,
943 				 sizeof(xstats_names[count].name),
944 				 "tx_q%u_%s", i,
945 				 vmxnet3_txq_stat_strings[t].name);
946 			count++;
947 		}
948 	}
949 
950 	return count;
951 }
952 
953 static int
954 vmxnet3_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
955 		       unsigned int n)
956 {
957 	unsigned int i, t, count = 0;
958 	unsigned int nstats =
959 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
960 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
961 
962 	if (n < nstats)
963 		return nstats;
964 
965 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
966 		struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
967 
968 		if (rxq == NULL)
969 			continue;
970 
971 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
972 			xstats[count].value = *(uint64_t *)(((char *)&rxq->stats) +
973 				vmxnet3_rxq_stat_strings[t].offset);
974 			xstats[count].id = count;
975 			count++;
976 		}
977 	}
978 
979 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
980 		struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
981 
982 		if (txq == NULL)
983 			continue;
984 
985 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
986 			xstats[count].value = *(uint64_t *)(((char *)&txq->stats) +
987 				vmxnet3_txq_stat_strings[t].offset);
988 			xstats[count].id = count;
989 			count++;
990 		}
991 	}
992 
993 	return count;
994 }
995 
996 static int
997 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
998 {
999 	unsigned int i;
1000 	struct vmxnet3_hw *hw = dev->data->dev_private;
1001 	struct UPT1_TxStats txStats;
1002 	struct UPT1_RxStats rxStats;
1003 
1004 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
1005 
1006 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
1007 	for (i = 0; i < hw->num_tx_queues; i++) {
1008 		vmxnet3_hw_tx_stats_get(hw, i, &txStats);
1009 
1010 		stats->q_opackets[i] = txStats.ucastPktsTxOK +
1011 			txStats.mcastPktsTxOK +
1012 			txStats.bcastPktsTxOK;
1013 
1014 		stats->q_obytes[i] = txStats.ucastBytesTxOK +
1015 			txStats.mcastBytesTxOK +
1016 			txStats.bcastBytesTxOK;
1017 
1018 		stats->opackets += stats->q_opackets[i];
1019 		stats->obytes += stats->q_obytes[i];
1020 		stats->oerrors += txStats.pktsTxError + txStats.pktsTxDiscard;
1021 	}
1022 
1023 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
1024 	for (i = 0; i < hw->num_rx_queues; i++) {
1025 		vmxnet3_hw_rx_stats_get(hw, i, &rxStats);
1026 
1027 		stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
1028 			rxStats.mcastPktsRxOK +
1029 			rxStats.bcastPktsRxOK;
1030 
1031 		stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
1032 			rxStats.mcastBytesRxOK +
1033 			rxStats.bcastBytesRxOK;
1034 
1035 		stats->ipackets += stats->q_ipackets[i];
1036 		stats->ibytes += stats->q_ibytes[i];
1037 
1038 		stats->q_errors[i] = rxStats.pktsRxError;
1039 		stats->ierrors += rxStats.pktsRxError;
1040 		stats->imissed += rxStats.pktsRxOutOfBuf;
1041 	}
1042 
1043 	return 0;
1044 }
1045 
1046 static void
1047 vmxnet3_dev_info_get(struct rte_eth_dev *dev __rte_unused,
1048 		     struct rte_eth_dev_info *dev_info)
1049 {
1050 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
1051 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
1052 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
1053 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
1054 	dev_info->speed_capa = ETH_LINK_SPEED_10G;
1055 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
1056 
1057 	dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
1058 
1059 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1060 		.nb_max = VMXNET3_RX_RING_MAX_SIZE,
1061 		.nb_min = VMXNET3_DEF_RX_RING_SIZE,
1062 		.nb_align = 1,
1063 	};
1064 
1065 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1066 		.nb_max = VMXNET3_TX_RING_MAX_SIZE,
1067 		.nb_min = VMXNET3_DEF_TX_RING_SIZE,
1068 		.nb_align = 1,
1069 		.nb_seg_max = VMXNET3_TX_MAX_SEG,
1070 		.nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
1071 	};
1072 
1073 	dev_info->rx_offload_capa = VMXNET3_RX_OFFLOAD_CAP;
1074 	dev_info->rx_queue_offload_capa = 0;
1075 	dev_info->tx_offload_capa = VMXNET3_TX_OFFLOAD_CAP;
1076 	dev_info->tx_queue_offload_capa = 0;
1077 }
1078 
1079 static const uint32_t *
1080 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
1081 {
1082 	static const uint32_t ptypes[] = {
1083 		RTE_PTYPE_L3_IPV4_EXT,
1084 		RTE_PTYPE_L3_IPV4,
1085 		RTE_PTYPE_UNKNOWN
1086 	};
1087 
1088 	if (dev->rx_pkt_burst == vmxnet3_recv_pkts)
1089 		return ptypes;
1090 	return NULL;
1091 }
1092 
1093 static int
1094 vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
1095 {
1096 	struct vmxnet3_hw *hw = dev->data->dev_private;
1097 
1098 	ether_addr_copy(mac_addr, (struct ether_addr *)(hw->perm_addr));
1099 	vmxnet3_write_mac(hw, mac_addr->addr_bytes);
1100 	return 0;
1101 }
1102 
1103 /* return 0 means link status changed, -1 means not changed */
1104 static int
1105 __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
1106 			  __rte_unused int wait_to_complete)
1107 {
1108 	struct vmxnet3_hw *hw = dev->data->dev_private;
1109 	struct rte_eth_link link;
1110 	uint32_t ret;
1111 
1112 	memset(&link, 0, sizeof(link));
1113 
1114 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
1115 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
1116 
1117 	if (ret & 0x1)
1118 		link.link_status = ETH_LINK_UP;
1119 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
1120 	link.link_speed = ETH_SPEED_NUM_10G;
1121 	link.link_autoneg = ETH_LINK_FIXED;
1122 
1123 	return rte_eth_linkstatus_set(dev, &link);
1124 }
1125 
1126 static int
1127 vmxnet3_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1128 {
1129 	/* Link status doesn't change for stopped dev */
1130 	if (dev->data->dev_started == 0)
1131 		return -1;
1132 
1133 	return __vmxnet3_dev_link_update(dev, wait_to_complete);
1134 }
1135 
1136 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
1137 static void
1138 vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set)
1139 {
1140 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1141 
1142 	if (set)
1143 		rxConf->rxMode = rxConf->rxMode | feature;
1144 	else
1145 		rxConf->rxMode = rxConf->rxMode & (~feature);
1146 
1147 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_UPDATE_RX_MODE);
1148 }
1149 
1150 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1151 static void
1152 vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev)
1153 {
1154 	struct vmxnet3_hw *hw = dev->data->dev_private;
1155 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1156 
1157 	memset(vf_table, 0, VMXNET3_VFT_TABLE_SIZE);
1158 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 1);
1159 
1160 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1161 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1162 }
1163 
1164 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1165 static void
1166 vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
1167 {
1168 	struct vmxnet3_hw *hw = dev->data->dev_private;
1169 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1170 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
1171 
1172 	if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1173 		memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1174 	else
1175 		memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1176 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
1177 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1178 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1179 }
1180 
1181 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1182 static void
1183 vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev)
1184 {
1185 	struct vmxnet3_hw *hw = dev->data->dev_private;
1186 
1187 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 1);
1188 }
1189 
1190 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1191 static void
1192 vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev)
1193 {
1194 	struct vmxnet3_hw *hw = dev->data->dev_private;
1195 
1196 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 0);
1197 }
1198 
1199 /* Enable/disable filter on vlan */
1200 static int
1201 vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
1202 {
1203 	struct vmxnet3_hw *hw = dev->data->dev_private;
1204 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1205 	uint32_t *vf_table = rxConf->vfTable;
1206 
1207 	/* save state for restore */
1208 	if (on)
1209 		VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1210 	else
1211 		VMXNET3_CLEAR_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1212 
1213 	/* don't change active filter if in promiscuous mode */
1214 	if (rxConf->rxMode & VMXNET3_RXM_PROMISC)
1215 		return 0;
1216 
1217 	/* set in hardware */
1218 	if (on)
1219 		VMXNET3_SET_VFTABLE_ENTRY(vf_table, vid);
1220 	else
1221 		VMXNET3_CLEAR_VFTABLE_ENTRY(vf_table, vid);
1222 
1223 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1224 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1225 	return 0;
1226 }
1227 
1228 static int
1229 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1230 {
1231 	struct vmxnet3_hw *hw = dev->data->dev_private;
1232 	Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
1233 	uint32_t *vf_table = devRead->rxFilterConf.vfTable;
1234 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
1235 
1236 	if (mask & ETH_VLAN_STRIP_MASK) {
1237 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1238 			devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
1239 		else
1240 			devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
1241 
1242 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1243 				       VMXNET3_CMD_UPDATE_FEATURE);
1244 	}
1245 
1246 	if (mask & ETH_VLAN_FILTER_MASK) {
1247 		if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1248 			memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1249 		else
1250 			memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1251 
1252 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1253 				       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1254 	}
1255 
1256 	return 0;
1257 }
1258 
1259 static void
1260 vmxnet3_process_events(struct rte_eth_dev *dev)
1261 {
1262 	struct vmxnet3_hw *hw = dev->data->dev_private;
1263 	uint32_t events = hw->shared->ecr;
1264 
1265 	if (!events)
1266 		return;
1267 
1268 	/*
1269 	 * ECR bits when written with 1b are cleared. Hence write
1270 	 * events back to ECR so that the bits which were set will be reset.
1271 	 */
1272 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
1273 
1274 	/* Check if link state has changed */
1275 	if (events & VMXNET3_ECR_LINK) {
1276 		PMD_DRV_LOG(DEBUG, "Process events: VMXNET3_ECR_LINK event");
1277 		if (vmxnet3_dev_link_update(dev, 0) == 0)
1278 			_rte_eth_dev_callback_process(dev,
1279 						      RTE_ETH_EVENT_INTR_LSC,
1280 						      NULL);
1281 	}
1282 
1283 	/* Check if there is an error on xmit/recv queues */
1284 	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
1285 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1286 				       VMXNET3_CMD_GET_QUEUE_STATUS);
1287 
1288 		if (hw->tqd_start->status.stopped)
1289 			PMD_DRV_LOG(ERR, "tq error 0x%x",
1290 				    hw->tqd_start->status.error);
1291 
1292 		if (hw->rqd_start->status.stopped)
1293 			PMD_DRV_LOG(ERR, "rq error 0x%x",
1294 				     hw->rqd_start->status.error);
1295 
1296 		/* Reset the device */
1297 		/* Have to reset the device */
1298 	}
1299 
1300 	if (events & VMXNET3_ECR_DIC)
1301 		PMD_DRV_LOG(DEBUG, "Device implementation change event.");
1302 
1303 	if (events & VMXNET3_ECR_DEBUG)
1304 		PMD_DRV_LOG(DEBUG, "Debug event generated by device.");
1305 }
1306 
1307 static void
1308 vmxnet3_interrupt_handler(void *param)
1309 {
1310 	struct rte_eth_dev *dev = param;
1311 	struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
1312 
1313 	vmxnet3_process_events(dev);
1314 
1315 	if (rte_intr_enable(&pci_dev->intr_handle) < 0)
1316 		PMD_DRV_LOG(ERR, "interrupt enable failed");
1317 }
1318 
1319 RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd);
1320 RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
1321 RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio-pci");
1322 
1323 RTE_INIT(vmxnet3_init_log)
1324 {
1325 	vmxnet3_logtype_init = rte_log_register("pmd.net.vmxnet3.init");
1326 	if (vmxnet3_logtype_init >= 0)
1327 		rte_log_set_level(vmxnet3_logtype_init, RTE_LOG_NOTICE);
1328 	vmxnet3_logtype_driver = rte_log_register("pmd.net.vmxnet3.driver");
1329 	if (vmxnet3_logtype_driver >= 0)
1330 		rte_log_set_level(vmxnet3_logtype_driver, RTE_LOG_NOTICE);
1331 }
1332