xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision 08aa6271c86a561b66c6dd91f9a54fa2f12bc859)
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 
62 static int eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev);
63 static int eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev);
64 static int vmxnet3_dev_configure(struct rte_eth_dev *dev);
65 static int vmxnet3_dev_start(struct rte_eth_dev *dev);
66 static void vmxnet3_dev_stop(struct rte_eth_dev *dev);
67 static void vmxnet3_dev_close(struct rte_eth_dev *dev);
68 static void vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set);
69 static void vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev);
70 static void vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev);
71 static void vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev);
72 static void vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev);
73 static int __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
74 				     int wait_to_complete);
75 static int vmxnet3_dev_link_update(struct rte_eth_dev *dev,
76 				   int wait_to_complete);
77 static void vmxnet3_hw_stats_save(struct vmxnet3_hw *hw);
78 static int vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
79 				  struct rte_eth_stats *stats);
80 static int vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
81 					struct rte_eth_xstat_name *xstats,
82 					unsigned int n);
83 static int vmxnet3_dev_xstats_get(struct rte_eth_dev *dev,
84 				  struct rte_eth_xstat *xstats, unsigned int n);
85 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
86 				 struct rte_eth_dev_info *dev_info);
87 static const uint32_t *
88 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
89 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
90 				       uint16_t vid, int on);
91 static int vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
92 static int vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
93 				 struct ether_addr *mac_addr);
94 static void vmxnet3_interrupt_handler(void *param);
95 
96 int vmxnet3_logtype_init;
97 int vmxnet3_logtype_driver;
98 
99 /*
100  * The set of PCI devices this driver supports
101  */
102 #define VMWARE_PCI_VENDOR_ID 0x15AD
103 #define VMWARE_DEV_ID_VMXNET3 0x07B0
104 static const struct rte_pci_id pci_id_vmxnet3_map[] = {
105 	{ RTE_PCI_DEVICE(VMWARE_PCI_VENDOR_ID, VMWARE_DEV_ID_VMXNET3) },
106 	{ .vendor_id = 0, /* sentinel */ },
107 };
108 
109 static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
110 	.dev_configure        = vmxnet3_dev_configure,
111 	.dev_start            = vmxnet3_dev_start,
112 	.dev_stop             = vmxnet3_dev_stop,
113 	.dev_close            = vmxnet3_dev_close,
114 	.promiscuous_enable   = vmxnet3_dev_promiscuous_enable,
115 	.promiscuous_disable  = vmxnet3_dev_promiscuous_disable,
116 	.allmulticast_enable  = vmxnet3_dev_allmulticast_enable,
117 	.allmulticast_disable = vmxnet3_dev_allmulticast_disable,
118 	.link_update          = vmxnet3_dev_link_update,
119 	.stats_get            = vmxnet3_dev_stats_get,
120 	.xstats_get_names     = vmxnet3_dev_xstats_get_names,
121 	.xstats_get           = vmxnet3_dev_xstats_get,
122 	.mac_addr_set         = vmxnet3_mac_addr_set,
123 	.dev_infos_get        = vmxnet3_dev_info_get,
124 	.dev_supported_ptypes_get = vmxnet3_dev_supported_ptypes_get,
125 	.vlan_filter_set      = vmxnet3_dev_vlan_filter_set,
126 	.vlan_offload_set     = vmxnet3_dev_vlan_offload_set,
127 	.rx_queue_setup       = vmxnet3_dev_rx_queue_setup,
128 	.rx_queue_release     = vmxnet3_dev_rx_queue_release,
129 	.tx_queue_setup       = vmxnet3_dev_tx_queue_setup,
130 	.tx_queue_release     = vmxnet3_dev_tx_queue_release,
131 };
132 
133 struct vmxnet3_xstats_name_off {
134 	char name[RTE_ETH_XSTATS_NAME_SIZE];
135 	unsigned int offset;
136 };
137 
138 /* tx_qX_ is prepended to the name string here */
139 static const struct vmxnet3_xstats_name_off vmxnet3_txq_stat_strings[] = {
140 	{"drop_total",         offsetof(struct vmxnet3_txq_stats, drop_total)},
141 	{"drop_too_many_segs", offsetof(struct vmxnet3_txq_stats, drop_too_many_segs)},
142 	{"drop_tso",           offsetof(struct vmxnet3_txq_stats, drop_tso)},
143 	{"tx_ring_full",       offsetof(struct vmxnet3_txq_stats, tx_ring_full)},
144 };
145 
146 /* rx_qX_ is prepended to the name string here */
147 static const struct vmxnet3_xstats_name_off vmxnet3_rxq_stat_strings[] = {
148 	{"drop_total",           offsetof(struct vmxnet3_rxq_stats, drop_total)},
149 	{"drop_err",             offsetof(struct vmxnet3_rxq_stats, drop_err)},
150 	{"drop_fcs",             offsetof(struct vmxnet3_rxq_stats, drop_fcs)},
151 	{"rx_buf_alloc_failure", offsetof(struct vmxnet3_rxq_stats, rx_buf_alloc_failure)},
152 };
153 
154 static const struct rte_memzone *
155 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
156 		 const char *post_string, int socket_id,
157 		 uint16_t align, bool reuse)
158 {
159 	char z_name[RTE_MEMZONE_NAMESIZE];
160 	const struct rte_memzone *mz;
161 
162 	snprintf(z_name, sizeof(z_name), "%s_%d_%s",
163 		 dev->device->driver->name, dev->data->port_id, post_string);
164 
165 	mz = rte_memzone_lookup(z_name);
166 	if (!reuse) {
167 		if (mz)
168 			rte_memzone_free(mz);
169 		return rte_memzone_reserve_aligned(z_name, size, socket_id,
170 				RTE_MEMZONE_IOVA_CONTIG, align);
171 	}
172 
173 	if (mz)
174 		return mz;
175 
176 	return rte_memzone_reserve_aligned(z_name, size, socket_id,
177 			RTE_MEMZONE_IOVA_CONTIG, align);
178 }
179 
180 /*
181  * This function is based on vmxnet3_disable_intr()
182  */
183 static void
184 vmxnet3_disable_intr(struct vmxnet3_hw *hw)
185 {
186 	int i;
187 
188 	PMD_INIT_FUNC_TRACE();
189 
190 	hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
191 	for (i = 0; i < hw->num_intrs; i++)
192 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
193 }
194 
195 static void
196 vmxnet3_enable_intr(struct vmxnet3_hw *hw)
197 {
198 	int i;
199 
200 	PMD_INIT_FUNC_TRACE();
201 
202 	hw->shared->devRead.intrConf.intrCtrl &= ~VMXNET3_IC_DISABLE_ALL;
203 	for (i = 0; i < hw->num_intrs; i++)
204 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 0);
205 }
206 
207 /*
208  * Gets tx data ring descriptor size.
209  */
210 static uint16_t
211 eth_vmxnet3_txdata_get(struct vmxnet3_hw *hw)
212 {
213 	uint16 txdata_desc_size;
214 
215 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
216 			       VMXNET3_CMD_GET_TXDATA_DESC_SIZE);
217 	txdata_desc_size = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
218 
219 	return (txdata_desc_size < VMXNET3_TXDATA_DESC_MIN_SIZE ||
220 		txdata_desc_size > VMXNET3_TXDATA_DESC_MAX_SIZE ||
221 		txdata_desc_size & VMXNET3_TXDATA_DESC_SIZE_MASK) ?
222 		sizeof(struct Vmxnet3_TxDataDesc) : txdata_desc_size;
223 }
224 
225 /*
226  * It returns 0 on success.
227  */
228 static int
229 eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
230 {
231 	struct rte_pci_device *pci_dev;
232 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
233 	uint32_t mac_hi, mac_lo, ver;
234 	struct rte_eth_link link;
235 
236 	PMD_INIT_FUNC_TRACE();
237 
238 	eth_dev->dev_ops = &vmxnet3_eth_dev_ops;
239 	eth_dev->rx_pkt_burst = &vmxnet3_recv_pkts;
240 	eth_dev->tx_pkt_burst = &vmxnet3_xmit_pkts;
241 	eth_dev->tx_pkt_prepare = vmxnet3_prep_pkts;
242 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
243 
244 	/*
245 	 * for secondary processes, we don't initialize any further as primary
246 	 * has already done this work.
247 	 */
248 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
249 		return 0;
250 
251 	rte_eth_copy_pci_info(eth_dev, pci_dev);
252 
253 	/* Vendor and Device ID need to be set before init of shared code */
254 	hw->device_id = pci_dev->id.device_id;
255 	hw->vendor_id = pci_dev->id.vendor_id;
256 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
257 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
258 
259 	hw->num_rx_queues = 1;
260 	hw->num_tx_queues = 1;
261 	hw->bufs_per_pkt = 1;
262 
263 	/* Check h/w version compatibility with driver. */
264 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
265 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
266 
267 	if (ver & (1 << VMXNET3_REV_3)) {
268 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
269 				       1 << VMXNET3_REV_3);
270 		hw->version = VMXNET3_REV_3 + 1;
271 	} else if (ver & (1 << VMXNET3_REV_2)) {
272 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
273 				       1 << VMXNET3_REV_2);
274 		hw->version = VMXNET3_REV_2 + 1;
275 	} else if (ver & (1 << VMXNET3_REV_1)) {
276 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
277 				       1 << VMXNET3_REV_1);
278 		hw->version = VMXNET3_REV_1 + 1;
279 	} else {
280 		PMD_INIT_LOG(ERR, "Incompatible hardware version: %d", ver);
281 		return -EIO;
282 	}
283 
284 	PMD_INIT_LOG(DEBUG, "Using device version %d\n", hw->version);
285 
286 	/* Check UPT version compatibility with driver. */
287 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
288 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
289 	if (ver & 0x1)
290 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
291 	else {
292 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
293 		return -EIO;
294 	}
295 
296 	/* Getting MAC Address */
297 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
298 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
299 	memcpy(hw->perm_addr, &mac_lo, 4);
300 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
301 
302 	/* Allocate memory for storing MAC addresses */
303 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
304 					       VMXNET3_MAX_MAC_ADDRS, 0);
305 	if (eth_dev->data->mac_addrs == NULL) {
306 		PMD_INIT_LOG(ERR,
307 			     "Failed to allocate %d bytes needed to store MAC addresses",
308 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
309 		return -ENOMEM;
310 	}
311 	/* Copy the permanent MAC address */
312 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
313 			&eth_dev->data->mac_addrs[0]);
314 
315 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
316 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
317 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
318 
319 	/* Put device in Quiesce Mode */
320 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
321 
322 	/* allow untagged pkts */
323 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
324 
325 	hw->txdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
326 		eth_vmxnet3_txdata_get(hw) : sizeof(struct Vmxnet3_TxDataDesc);
327 
328 	hw->rxdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
329 		VMXNET3_DEF_RXDATA_DESC_SIZE : 0;
330 	RTE_ASSERT((hw->rxdata_desc_size & ~VMXNET3_RXDATA_DESC_SIZE_MASK) ==
331 		   hw->rxdata_desc_size);
332 
333 	/* clear shadow stats */
334 	memset(hw->saved_tx_stats, 0, sizeof(hw->saved_tx_stats));
335 	memset(hw->saved_rx_stats, 0, sizeof(hw->saved_rx_stats));
336 
337 	/* set the initial link status */
338 	memset(&link, 0, sizeof(link));
339 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
340 	link.link_speed = ETH_SPEED_NUM_10G;
341 	link.link_autoneg = ETH_LINK_FIXED;
342 	rte_eth_linkstatus_set(eth_dev, &link);
343 
344 	return 0;
345 }
346 
347 static int
348 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
349 {
350 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
351 
352 	PMD_INIT_FUNC_TRACE();
353 
354 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
355 		return 0;
356 
357 	if (hw->adapter_stopped == 0)
358 		vmxnet3_dev_close(eth_dev);
359 
360 	eth_dev->dev_ops = NULL;
361 	eth_dev->rx_pkt_burst = NULL;
362 	eth_dev->tx_pkt_burst = NULL;
363 	eth_dev->tx_pkt_prepare = NULL;
364 
365 	rte_free(eth_dev->data->mac_addrs);
366 	eth_dev->data->mac_addrs = NULL;
367 
368 	return 0;
369 }
370 
371 static int eth_vmxnet3_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
372 	struct rte_pci_device *pci_dev)
373 {
374 	return rte_eth_dev_pci_generic_probe(pci_dev,
375 		sizeof(struct vmxnet3_hw), eth_vmxnet3_dev_init);
376 }
377 
378 static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
379 {
380 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_vmxnet3_dev_uninit);
381 }
382 
383 static struct rte_pci_driver rte_vmxnet3_pmd = {
384 	.id_table = pci_id_vmxnet3_map,
385 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
386 	.probe = eth_vmxnet3_pci_probe,
387 	.remove = eth_vmxnet3_pci_remove,
388 };
389 
390 static int
391 vmxnet3_dev_configure(struct rte_eth_dev *dev)
392 {
393 	const struct rte_memzone *mz;
394 	struct vmxnet3_hw *hw = dev->data->dev_private;
395 	size_t size;
396 
397 	PMD_INIT_FUNC_TRACE();
398 
399 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
400 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
401 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
402 		return -EINVAL;
403 	}
404 
405 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
406 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
407 		return -EINVAL;
408 	}
409 
410 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
411 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
412 
413 	if (size > UINT16_MAX)
414 		return -EINVAL;
415 
416 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
417 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
418 
419 	/*
420 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
421 	 * on current socket
422 	 */
423 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
424 			      "shared", rte_socket_id(), 8, 1);
425 
426 	if (mz == NULL) {
427 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
428 		return -ENOMEM;
429 	}
430 	memset(mz->addr, 0, mz->len);
431 
432 	hw->shared = mz->addr;
433 	hw->sharedPA = mz->iova;
434 
435 	/*
436 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
437 	 * on current socket.
438 	 *
439 	 * We cannot reuse this memzone from previous allocation as its size
440 	 * depends on the number of tx and rx queues, which could be different
441 	 * from one config to another.
442 	 */
443 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
444 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
445 	if (mz == NULL) {
446 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
447 		return -ENOMEM;
448 	}
449 	memset(mz->addr, 0, mz->len);
450 
451 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
452 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
453 
454 	hw->queueDescPA = mz->iova;
455 	hw->queue_desc_len = (uint16_t)size;
456 
457 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
458 		/* Allocate memory structure for UPT1_RSSConf and configure */
459 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
460 				      "rss_conf", rte_socket_id(),
461 				      RTE_CACHE_LINE_SIZE, 1);
462 		if (mz == NULL) {
463 			PMD_INIT_LOG(ERR,
464 				     "ERROR: Creating rss_conf structure zone");
465 			return -ENOMEM;
466 		}
467 		memset(mz->addr, 0, mz->len);
468 
469 		hw->rss_conf = mz->addr;
470 		hw->rss_confPA = mz->iova;
471 	}
472 
473 	return 0;
474 }
475 
476 static void
477 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
478 {
479 	uint32_t val;
480 
481 	PMD_INIT_LOG(DEBUG,
482 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
483 		     addr[0], addr[1], addr[2],
484 		     addr[3], addr[4], addr[5]);
485 
486 	memcpy(&val, addr, 4);
487 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
488 
489 	memcpy(&val, addr + 4, 2);
490 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
491 }
492 
493 static int
494 vmxnet3_dev_setup_memreg(struct rte_eth_dev *dev)
495 {
496 	struct vmxnet3_hw *hw = dev->data->dev_private;
497 	Vmxnet3_DriverShared *shared = hw->shared;
498 	Vmxnet3_CmdInfo *cmdInfo;
499 	struct rte_mempool *mp[VMXNET3_MAX_RX_QUEUES];
500 	uint8_t index[VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES];
501 	uint32_t num, i, j, size;
502 
503 	if (hw->memRegsPA == 0) {
504 		const struct rte_memzone *mz;
505 
506 		size = sizeof(Vmxnet3_MemRegs) +
507 			(VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES) *
508 			sizeof(Vmxnet3_MemoryRegion);
509 
510 		mz = gpa_zone_reserve(dev, size, "memRegs", rte_socket_id(), 8,
511 				      1);
512 		if (mz == NULL) {
513 			PMD_INIT_LOG(ERR, "ERROR: Creating memRegs zone");
514 			return -ENOMEM;
515 		}
516 		memset(mz->addr, 0, mz->len);
517 		hw->memRegs = mz->addr;
518 		hw->memRegsPA = mz->iova;
519 	}
520 
521 	num = hw->num_rx_queues;
522 
523 	for (i = 0; i < num; i++) {
524 		vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
525 
526 		mp[i] = rxq->mp;
527 		index[i] = 1 << i;
528 	}
529 
530 	/*
531 	 * The same mempool could be used by multiple queues. In such a case,
532 	 * remove duplicate mempool entries. Only one entry is kept with
533 	 * bitmask indicating queues that are using this mempool.
534 	 */
535 	for (i = 1; i < num; i++) {
536 		for (j = 0; j < i; j++) {
537 			if (mp[i] == mp[j]) {
538 				mp[i] = NULL;
539 				index[j] |= 1 << i;
540 				break;
541 			}
542 		}
543 	}
544 
545 	j = 0;
546 	for (i = 0; i < num; i++) {
547 		if (mp[i] == NULL)
548 			continue;
549 
550 		Vmxnet3_MemoryRegion *mr = &hw->memRegs->memRegs[j];
551 
552 		mr->startPA =
553 			(uintptr_t)STAILQ_FIRST(&mp[i]->mem_list)->iova;
554 		mr->length = STAILQ_FIRST(&mp[i]->mem_list)->len <= INT32_MAX ?
555 			STAILQ_FIRST(&mp[i]->mem_list)->len : INT32_MAX;
556 		mr->txQueueBits = index[i];
557 		mr->rxQueueBits = index[i];
558 
559 		PMD_INIT_LOG(INFO,
560 			     "index: %u startPA: %" PRIu64 " length: %u, "
561 			     "rxBits: %x",
562 			     j, mr->startPA, mr->length, mr->rxQueueBits);
563 		j++;
564 	}
565 	hw->memRegs->numRegs = j;
566 	PMD_INIT_LOG(INFO, "numRegs: %u", j);
567 
568 	size = sizeof(Vmxnet3_MemRegs) +
569 		(j - 1) * sizeof(Vmxnet3_MemoryRegion);
570 
571 	cmdInfo = &shared->cu.cmdInfo;
572 	cmdInfo->varConf.confVer = 1;
573 	cmdInfo->varConf.confLen = size;
574 	cmdInfo->varConf.confPA = hw->memRegsPA;
575 
576 	return 0;
577 }
578 
579 static int
580 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
581 {
582 	struct rte_eth_conf port_conf = dev->data->dev_conf;
583 	struct vmxnet3_hw *hw = dev->data->dev_private;
584 	uint32_t mtu = dev->data->mtu;
585 	Vmxnet3_DriverShared *shared = hw->shared;
586 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
587 	uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
588 	uint32_t i;
589 	int ret;
590 
591 	hw->mtu = mtu;
592 
593 	shared->magic = VMXNET3_REV1_MAGIC;
594 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
595 
596 	/* Setting up Guest OS information */
597 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
598 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
599 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
600 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
601 	devRead->misc.driverInfo.uptVerSpt     = 1;
602 
603 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
604 	devRead->misc.queueDescPA  = hw->queueDescPA;
605 	devRead->misc.queueDescLen = hw->queue_desc_len;
606 	devRead->misc.numTxQueues  = hw->num_tx_queues;
607 	devRead->misc.numRxQueues  = hw->num_rx_queues;
608 
609 	/*
610 	 * Set number of interrupts to 1
611 	 * PMD by default disables all the interrupts but this is MUST
612 	 * to activate device. It needs at least one interrupt for
613 	 * link events to handle
614 	 */
615 	hw->num_intrs = devRead->intrConf.numIntrs = 1;
616 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
617 
618 	for (i = 0; i < hw->num_tx_queues; i++) {
619 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
620 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
621 
622 		txq->shared = &hw->tqd_start[i];
623 
624 		tqd->ctrl.txNumDeferred  = 0;
625 		tqd->ctrl.txThreshold    = 1;
626 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
627 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
628 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
629 
630 		tqd->conf.txRingSize   = txq->cmd_ring.size;
631 		tqd->conf.compRingSize = txq->comp_ring.size;
632 		tqd->conf.dataRingSize = txq->data_ring.size;
633 		tqd->conf.txDataRingDescSize = txq->txdata_desc_size;
634 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
635 		tqd->status.stopped    = TRUE;
636 		tqd->status.error      = 0;
637 		memset(&tqd->stats, 0, sizeof(tqd->stats));
638 	}
639 
640 	for (i = 0; i < hw->num_rx_queues; i++) {
641 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
642 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
643 
644 		rxq->shared = &hw->rqd_start[i];
645 
646 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
647 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
648 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
649 
650 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
651 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
652 		rqd->conf.compRingSize    = rxq->comp_ring.size;
653 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
654 		if (VMXNET3_VERSION_GE_3(hw)) {
655 			rqd->conf.rxDataRingBasePA = rxq->data_ring.basePA;
656 			rqd->conf.rxDataRingDescSize = rxq->data_desc_size;
657 		}
658 		rqd->status.stopped       = TRUE;
659 		rqd->status.error         = 0;
660 		memset(&rqd->stats, 0, sizeof(rqd->stats));
661 	}
662 
663 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
664 	devRead->rxFilterConf.rxMode = 0;
665 
666 	/* Setting up feature flags */
667 	if (rx_offloads & DEV_RX_OFFLOAD_CHECKSUM)
668 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
669 
670 	if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) {
671 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
672 		devRead->misc.maxNumRxSG = 0;
673 	}
674 
675 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
676 		ret = vmxnet3_rss_configure(dev);
677 		if (ret != VMXNET3_SUCCESS)
678 			return ret;
679 
680 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
681 		devRead->rssConfDesc.confVer = 1;
682 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
683 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
684 	}
685 
686 	ret = vmxnet3_dev_vlan_offload_set(dev,
687 			ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
688 	if (ret)
689 		return ret;
690 
691 	vmxnet3_write_mac(hw, dev->data->mac_addrs->addr_bytes);
692 
693 	return VMXNET3_SUCCESS;
694 }
695 
696 /*
697  * Configure device link speed and setup link.
698  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
699  * It returns 0 on success.
700  */
701 static int
702 vmxnet3_dev_start(struct rte_eth_dev *dev)
703 {
704 	int ret;
705 	struct vmxnet3_hw *hw = dev->data->dev_private;
706 
707 	PMD_INIT_FUNC_TRACE();
708 
709 	/* Save stats before it is reset by CMD_ACTIVATE */
710 	vmxnet3_hw_stats_save(hw);
711 
712 	ret = vmxnet3_setup_driver_shared(dev);
713 	if (ret != VMXNET3_SUCCESS)
714 		return ret;
715 
716 	/* check if lsc interrupt feature is enabled */
717 	if (dev->data->dev_conf.intr_conf.lsc) {
718 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
719 
720 		/* Setup interrupt callback  */
721 		rte_intr_callback_register(&pci_dev->intr_handle,
722 					   vmxnet3_interrupt_handler, dev);
723 
724 		if (rte_intr_enable(&pci_dev->intr_handle) < 0) {
725 			PMD_INIT_LOG(ERR, "interrupt enable failed");
726 			return -EIO;
727 		}
728 	}
729 
730 	/* Exchange shared data with device */
731 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
732 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
733 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
734 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
735 
736 	/* Activate device by register write */
737 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
738 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
739 
740 	if (ret != 0) {
741 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
742 		return -EINVAL;
743 	}
744 
745 	/* Setup memory region for rx buffers */
746 	ret = vmxnet3_dev_setup_memreg(dev);
747 	if (ret == 0) {
748 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
749 				       VMXNET3_CMD_REGISTER_MEMREGS);
750 		ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
751 		if (ret != 0)
752 			PMD_INIT_LOG(DEBUG,
753 				     "Failed in setup memory region cmd\n");
754 		ret = 0;
755 	} else {
756 		PMD_INIT_LOG(DEBUG, "Failed to setup memory region\n");
757 	}
758 
759 	/* Disable interrupts */
760 	vmxnet3_disable_intr(hw);
761 
762 	/*
763 	 * Load RX queues with blank mbufs and update next2fill index for device
764 	 * Update RxMode of the device
765 	 */
766 	ret = vmxnet3_dev_rxtx_init(dev);
767 	if (ret != VMXNET3_SUCCESS) {
768 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
769 		return ret;
770 	}
771 
772 	hw->adapter_stopped = FALSE;
773 
774 	/* Setting proper Rx Mode and issue Rx Mode Update command */
775 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
776 
777 	if (dev->data->dev_conf.intr_conf.lsc) {
778 		vmxnet3_enable_intr(hw);
779 
780 		/*
781 		 * Update link state from device since this won't be
782 		 * done upon starting with lsc in use. This is done
783 		 * only after enabling interrupts to avoid any race
784 		 * where the link state could change without an
785 		 * interrupt being fired.
786 		 */
787 		__vmxnet3_dev_link_update(dev, 0);
788 	}
789 
790 	return VMXNET3_SUCCESS;
791 }
792 
793 /*
794  * Stop device: disable rx and tx functions to allow for reconfiguring.
795  */
796 static void
797 vmxnet3_dev_stop(struct rte_eth_dev *dev)
798 {
799 	struct rte_eth_link link;
800 	struct vmxnet3_hw *hw = dev->data->dev_private;
801 
802 	PMD_INIT_FUNC_TRACE();
803 
804 	if (hw->adapter_stopped == 1) {
805 		PMD_INIT_LOG(DEBUG, "Device already closed.");
806 		return;
807 	}
808 
809 	/* disable interrupts */
810 	vmxnet3_disable_intr(hw);
811 
812 	if (dev->data->dev_conf.intr_conf.lsc) {
813 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
814 
815 		rte_intr_disable(&pci_dev->intr_handle);
816 
817 		rte_intr_callback_unregister(&pci_dev->intr_handle,
818 					     vmxnet3_interrupt_handler, dev);
819 	}
820 
821 	/* quiesce the device first */
822 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
823 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
824 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
825 
826 	/* reset the device */
827 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
828 	PMD_INIT_LOG(DEBUG, "Device reset.");
829 	hw->adapter_stopped = 0;
830 
831 	vmxnet3_dev_clear_queues(dev);
832 
833 	/* Clear recorded link status */
834 	memset(&link, 0, sizeof(link));
835 	link.link_duplex = ETH_LINK_FULL_DUPLEX;
836 	link.link_speed = ETH_SPEED_NUM_10G;
837 	link.link_autoneg = ETH_LINK_FIXED;
838 	rte_eth_linkstatus_set(dev, &link);
839 }
840 
841 /*
842  * Reset and stop device.
843  */
844 static void
845 vmxnet3_dev_close(struct rte_eth_dev *dev)
846 {
847 	struct vmxnet3_hw *hw = dev->data->dev_private;
848 
849 	PMD_INIT_FUNC_TRACE();
850 
851 	vmxnet3_dev_stop(dev);
852 	hw->adapter_stopped = 1;
853 }
854 
855 static void
856 vmxnet3_hw_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
857 			struct UPT1_TxStats *res)
858 {
859 #define VMXNET3_UPDATE_TX_STAT(h, i, f, r)		\
860 		((r)->f = (h)->tqd_start[(i)].stats.f +	\
861 			(h)->saved_tx_stats[(i)].f)
862 
863 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastPktsTxOK, res);
864 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastPktsTxOK, res);
865 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastPktsTxOK, res);
866 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastBytesTxOK, res);
867 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastBytesTxOK, res);
868 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastBytesTxOK, res);
869 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxError, res);
870 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxDiscard, res);
871 
872 #undef VMXNET3_UPDATE_TX_STAT
873 }
874 
875 static void
876 vmxnet3_hw_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
877 			struct UPT1_RxStats *res)
878 {
879 #define VMXNET3_UPDATE_RX_STAT(h, i, f, r)		\
880 		((r)->f = (h)->rqd_start[(i)].stats.f +	\
881 			(h)->saved_rx_stats[(i)].f)
882 
883 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastPktsRxOK, res);
884 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastPktsRxOK, res);
885 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastPktsRxOK, res);
886 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastBytesRxOK, res);
887 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastBytesRxOK, res);
888 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastBytesRxOK, res);
889 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxError, res);
890 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxOutOfBuf, res);
891 
892 #undef VMXNET3_UPDATE_RX_STATS
893 }
894 
895 static void
896 vmxnet3_hw_stats_save(struct vmxnet3_hw *hw)
897 {
898 	unsigned int i;
899 
900 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
901 
902 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
903 
904 	for (i = 0; i < hw->num_tx_queues; i++)
905 		vmxnet3_hw_tx_stats_get(hw, i, &hw->saved_tx_stats[i]);
906 	for (i = 0; i < hw->num_rx_queues; i++)
907 		vmxnet3_hw_rx_stats_get(hw, i, &hw->saved_rx_stats[i]);
908 }
909 
910 static int
911 vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
912 			     struct rte_eth_xstat_name *xstats_names,
913 			     unsigned int n)
914 {
915 	unsigned int i, t, count = 0;
916 	unsigned int nstats =
917 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
918 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
919 
920 	if (!xstats_names || n < nstats)
921 		return nstats;
922 
923 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
924 		if (!dev->data->rx_queues[i])
925 			continue;
926 
927 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
928 			snprintf(xstats_names[count].name,
929 				 sizeof(xstats_names[count].name),
930 				 "rx_q%u_%s", i,
931 				 vmxnet3_rxq_stat_strings[t].name);
932 			count++;
933 		}
934 	}
935 
936 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
937 		if (!dev->data->tx_queues[i])
938 			continue;
939 
940 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
941 			snprintf(xstats_names[count].name,
942 				 sizeof(xstats_names[count].name),
943 				 "tx_q%u_%s", i,
944 				 vmxnet3_txq_stat_strings[t].name);
945 			count++;
946 		}
947 	}
948 
949 	return count;
950 }
951 
952 static int
953 vmxnet3_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
954 		       unsigned int n)
955 {
956 	unsigned int i, t, count = 0;
957 	unsigned int nstats =
958 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
959 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
960 
961 	if (n < nstats)
962 		return nstats;
963 
964 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
965 		struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
966 
967 		if (rxq == NULL)
968 			continue;
969 
970 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
971 			xstats[count].value = *(uint64_t *)(((char *)&rxq->stats) +
972 				vmxnet3_rxq_stat_strings[t].offset);
973 			xstats[count].id = count;
974 			count++;
975 		}
976 	}
977 
978 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
979 		struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
980 
981 		if (txq == NULL)
982 			continue;
983 
984 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
985 			xstats[count].value = *(uint64_t *)(((char *)&txq->stats) +
986 				vmxnet3_txq_stat_strings[t].offset);
987 			xstats[count].id = count;
988 			count++;
989 		}
990 	}
991 
992 	return count;
993 }
994 
995 static int
996 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
997 {
998 	unsigned int i;
999 	struct vmxnet3_hw *hw = dev->data->dev_private;
1000 	struct UPT1_TxStats txStats;
1001 	struct UPT1_RxStats rxStats;
1002 
1003 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
1004 
1005 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
1006 	for (i = 0; i < hw->num_tx_queues; i++) {
1007 		vmxnet3_hw_tx_stats_get(hw, i, &txStats);
1008 
1009 		stats->q_opackets[i] = txStats.ucastPktsTxOK +
1010 			txStats.mcastPktsTxOK +
1011 			txStats.bcastPktsTxOK;
1012 
1013 		stats->q_obytes[i] = txStats.ucastBytesTxOK +
1014 			txStats.mcastBytesTxOK +
1015 			txStats.bcastBytesTxOK;
1016 
1017 		stats->opackets += stats->q_opackets[i];
1018 		stats->obytes += stats->q_obytes[i];
1019 		stats->oerrors += txStats.pktsTxError + txStats.pktsTxDiscard;
1020 	}
1021 
1022 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
1023 	for (i = 0; i < hw->num_rx_queues; i++) {
1024 		vmxnet3_hw_rx_stats_get(hw, i, &rxStats);
1025 
1026 		stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
1027 			rxStats.mcastPktsRxOK +
1028 			rxStats.bcastPktsRxOK;
1029 
1030 		stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
1031 			rxStats.mcastBytesRxOK +
1032 			rxStats.bcastBytesRxOK;
1033 
1034 		stats->ipackets += stats->q_ipackets[i];
1035 		stats->ibytes += stats->q_ibytes[i];
1036 
1037 		stats->q_errors[i] = rxStats.pktsRxError;
1038 		stats->ierrors += rxStats.pktsRxError;
1039 		stats->imissed += rxStats.pktsRxOutOfBuf;
1040 	}
1041 
1042 	return 0;
1043 }
1044 
1045 static void
1046 vmxnet3_dev_info_get(struct rte_eth_dev *dev __rte_unused,
1047 		     struct rte_eth_dev_info *dev_info)
1048 {
1049 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
1050 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
1051 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
1052 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
1053 	dev_info->speed_capa = ETH_LINK_SPEED_10G;
1054 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
1055 
1056 	dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
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 static void
1325 vmxnet3_init_log(void)
1326 {
1327 	vmxnet3_logtype_init = rte_log_register("pmd.net.vmxnet3.init");
1328 	if (vmxnet3_logtype_init >= 0)
1329 		rte_log_set_level(vmxnet3_logtype_init, RTE_LOG_NOTICE);
1330 	vmxnet3_logtype_driver = rte_log_register("pmd.net.vmxnet3.driver");
1331 	if (vmxnet3_logtype_driver >= 0)
1332 		rte_log_set_level(vmxnet3_logtype_driver, RTE_LOG_NOTICE);
1333 }
1334