xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision 5a4806d304e084573eb2193341add736ef0af50f)
1 /*-
2  *   BSD LICENSE
3  *
4  *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
5  *   All rights reserved.
6  *
7  *   Redistribution and use in source and binary forms, with or without
8  *   modification, are permitted provided that the following conditions
9  *   are met:
10  *
11  *     * Redistributions of source code must retain the above copyright
12  *       notice, this list of conditions and the following disclaimer.
13  *     * Redistributions in binary form must reproduce the above copyright
14  *       notice, this list of conditions and the following disclaimer in
15  *       the documentation and/or other materials provided with the
16  *       distribution.
17  *     * Neither the name of Intel Corporation nor the names of its
18  *       contributors may be used to endorse or promote products derived
19  *       from this software without specific prior written permission.
20  *
21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
32  */
33 
34 #include <sys/queue.h>
35 #include <stdio.h>
36 #include <errno.h>
37 #include <stdint.h>
38 #include <string.h>
39 #include <unistd.h>
40 #include <stdarg.h>
41 #include <fcntl.h>
42 #include <inttypes.h>
43 #include <rte_byteorder.h>
44 #include <rte_common.h>
45 #include <rte_cycles.h>
46 
47 #include <rte_interrupts.h>
48 #include <rte_log.h>
49 #include <rte_debug.h>
50 #include <rte_pci.h>
51 #include <rte_atomic.h>
52 #include <rte_branch_prediction.h>
53 #include <rte_memory.h>
54 #include <rte_memzone.h>
55 #include <rte_eal.h>
56 #include <rte_alarm.h>
57 #include <rte_ether.h>
58 #include <rte_ethdev.h>
59 #include <rte_ethdev_pci.h>
60 #include <rte_string_fns.h>
61 #include <rte_malloc.h>
62 #include <rte_dev.h>
63 
64 #include "base/vmxnet3_defs.h"
65 
66 #include "vmxnet3_ring.h"
67 #include "vmxnet3_logs.h"
68 #include "vmxnet3_ethdev.h"
69 
70 #define PROCESS_SYS_EVENTS 0
71 
72 #define	VMXNET3_TX_MAX_SEG	UINT8_MAX
73 
74 static int eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev);
75 static int eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev);
76 static int vmxnet3_dev_configure(struct rte_eth_dev *dev);
77 static int vmxnet3_dev_start(struct rte_eth_dev *dev);
78 static void vmxnet3_dev_stop(struct rte_eth_dev *dev);
79 static void vmxnet3_dev_close(struct rte_eth_dev *dev);
80 static void vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set);
81 static void vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev);
82 static void vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev);
83 static void vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev);
84 static void vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev);
85 static int __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
86 				     int wait_to_complete);
87 static int vmxnet3_dev_link_update(struct rte_eth_dev *dev,
88 				   int wait_to_complete);
89 static void vmxnet3_hw_stats_save(struct vmxnet3_hw *hw);
90 static void vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
91 				  struct rte_eth_stats *stats);
92 static int vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
93 					struct rte_eth_xstat_name *xstats,
94 					unsigned int n);
95 static int vmxnet3_dev_xstats_get(struct rte_eth_dev *dev,
96 				  struct rte_eth_xstat *xstats, unsigned int n);
97 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
98 				 struct rte_eth_dev_info *dev_info);
99 static const uint32_t *
100 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
101 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
102 				       uint16_t vid, int on);
103 static void vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
104 static void vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
105 				 struct ether_addr *mac_addr);
106 static void vmxnet3_interrupt_handler(void *param);
107 
108 /*
109  * The set of PCI devices this driver supports
110  */
111 #define VMWARE_PCI_VENDOR_ID 0x15AD
112 #define VMWARE_DEV_ID_VMXNET3 0x07B0
113 static const struct rte_pci_id pci_id_vmxnet3_map[] = {
114 	{ RTE_PCI_DEVICE(VMWARE_PCI_VENDOR_ID, VMWARE_DEV_ID_VMXNET3) },
115 	{ .vendor_id = 0, /* sentinel */ },
116 };
117 
118 static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
119 	.dev_configure        = vmxnet3_dev_configure,
120 	.dev_start            = vmxnet3_dev_start,
121 	.dev_stop             = vmxnet3_dev_stop,
122 	.dev_close            = vmxnet3_dev_close,
123 	.promiscuous_enable   = vmxnet3_dev_promiscuous_enable,
124 	.promiscuous_disable  = vmxnet3_dev_promiscuous_disable,
125 	.allmulticast_enable  = vmxnet3_dev_allmulticast_enable,
126 	.allmulticast_disable = vmxnet3_dev_allmulticast_disable,
127 	.link_update          = vmxnet3_dev_link_update,
128 	.stats_get            = vmxnet3_dev_stats_get,
129 	.xstats_get_names     = vmxnet3_dev_xstats_get_names,
130 	.xstats_get           = vmxnet3_dev_xstats_get,
131 	.mac_addr_set         = vmxnet3_mac_addr_set,
132 	.dev_infos_get        = vmxnet3_dev_info_get,
133 	.dev_supported_ptypes_get = vmxnet3_dev_supported_ptypes_get,
134 	.vlan_filter_set      = vmxnet3_dev_vlan_filter_set,
135 	.vlan_offload_set     = vmxnet3_dev_vlan_offload_set,
136 	.rx_queue_setup       = vmxnet3_dev_rx_queue_setup,
137 	.rx_queue_release     = vmxnet3_dev_rx_queue_release,
138 	.tx_queue_setup       = vmxnet3_dev_tx_queue_setup,
139 	.tx_queue_release     = vmxnet3_dev_tx_queue_release,
140 };
141 
142 struct vmxnet3_xstats_name_off {
143 	char name[RTE_ETH_XSTATS_NAME_SIZE];
144 	unsigned int offset;
145 };
146 
147 /* tx_qX_ is prepended to the name string here */
148 static const struct vmxnet3_xstats_name_off vmxnet3_txq_stat_strings[] = {
149 	{"drop_total",         offsetof(struct vmxnet3_txq_stats, drop_total)},
150 	{"drop_too_many_segs", offsetof(struct vmxnet3_txq_stats, drop_too_many_segs)},
151 	{"drop_tso",           offsetof(struct vmxnet3_txq_stats, drop_tso)},
152 	{"tx_ring_full",       offsetof(struct vmxnet3_txq_stats, tx_ring_full)},
153 };
154 
155 /* rx_qX_ is prepended to the name string here */
156 static const struct vmxnet3_xstats_name_off vmxnet3_rxq_stat_strings[] = {
157 	{"drop_total",           offsetof(struct vmxnet3_rxq_stats, drop_total)},
158 	{"drop_err",             offsetof(struct vmxnet3_rxq_stats, drop_err)},
159 	{"drop_fcs",             offsetof(struct vmxnet3_rxq_stats, drop_fcs)},
160 	{"rx_buf_alloc_failure", offsetof(struct vmxnet3_rxq_stats, rx_buf_alloc_failure)},
161 };
162 
163 static const struct rte_memzone *
164 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
165 		 const char *post_string, int socket_id,
166 		 uint16_t align, bool reuse)
167 {
168 	char z_name[RTE_MEMZONE_NAMESIZE];
169 	const struct rte_memzone *mz;
170 
171 	snprintf(z_name, sizeof(z_name), "%s_%d_%s",
172 		 dev->device->driver->name, dev->data->port_id, post_string);
173 
174 	mz = rte_memzone_lookup(z_name);
175 	if (!reuse) {
176 		if (mz)
177 			rte_memzone_free(mz);
178 		return rte_memzone_reserve_aligned(z_name, size, socket_id,
179 						   0, align);
180 	}
181 
182 	if (mz)
183 		return mz;
184 
185 	return rte_memzone_reserve_aligned(z_name, size, socket_id, 0, align);
186 }
187 
188 /**
189  * Atomically reads the link status information from global
190  * structure rte_eth_dev.
191  *
192  * @param dev
193  *   - Pointer to the structure rte_eth_dev to read from.
194  *   - Pointer to the buffer to be saved with the link status.
195  *
196  * @return
197  *   - On success, zero.
198  *   - On failure, negative value.
199  */
200 
201 static int
202 vmxnet3_dev_atomic_read_link_status(struct rte_eth_dev *dev,
203 				    struct rte_eth_link *link)
204 {
205 	struct rte_eth_link *dst = link;
206 	struct rte_eth_link *src = &(dev->data->dev_link);
207 
208 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
209 				*(uint64_t *)src) == 0)
210 		return -1;
211 
212 	return 0;
213 }
214 
215 /**
216  * Atomically writes the link status information into global
217  * structure rte_eth_dev.
218  *
219  * @param dev
220  *   - Pointer to the structure rte_eth_dev to write to.
221  *   - Pointer to the buffer to be saved with the link status.
222  *
223  * @return
224  *   - On success, zero.
225  *   - On failure, negative value.
226  */
227 static int
228 vmxnet3_dev_atomic_write_link_status(struct rte_eth_dev *dev,
229 				     struct rte_eth_link *link)
230 {
231 	struct rte_eth_link *dst = &(dev->data->dev_link);
232 	struct rte_eth_link *src = link;
233 
234 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
235 				*(uint64_t *)src) == 0)
236 		return -1;
237 
238 	return 0;
239 }
240 
241 /*
242  * This function is based on vmxnet3_disable_intr()
243  */
244 static void
245 vmxnet3_disable_intr(struct vmxnet3_hw *hw)
246 {
247 	int i;
248 
249 	PMD_INIT_FUNC_TRACE();
250 
251 	hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
252 	for (i = 0; i < hw->num_intrs; i++)
253 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
254 }
255 
256 static void
257 vmxnet3_enable_intr(struct vmxnet3_hw *hw)
258 {
259 	int i;
260 
261 	PMD_INIT_FUNC_TRACE();
262 
263 	hw->shared->devRead.intrConf.intrCtrl &= ~VMXNET3_IC_DISABLE_ALL;
264 	for (i = 0; i < hw->num_intrs; i++)
265 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 0);
266 }
267 
268 /*
269  * Gets tx data ring descriptor size.
270  */
271 static uint16_t
272 eth_vmxnet3_txdata_get(struct vmxnet3_hw *hw)
273 {
274 	uint16 txdata_desc_size;
275 
276 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
277 			       VMXNET3_CMD_GET_TXDATA_DESC_SIZE);
278 	txdata_desc_size = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
279 
280 	return (txdata_desc_size < VMXNET3_TXDATA_DESC_MIN_SIZE ||
281 		txdata_desc_size > VMXNET3_TXDATA_DESC_MAX_SIZE ||
282 		txdata_desc_size & VMXNET3_TXDATA_DESC_SIZE_MASK) ?
283 		sizeof(struct Vmxnet3_TxDataDesc) : txdata_desc_size;
284 }
285 
286 /*
287  * It returns 0 on success.
288  */
289 static int
290 eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
291 {
292 	struct rte_pci_device *pci_dev;
293 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
294 	uint32_t mac_hi, mac_lo, ver;
295 
296 	PMD_INIT_FUNC_TRACE();
297 
298 	eth_dev->dev_ops = &vmxnet3_eth_dev_ops;
299 	eth_dev->rx_pkt_burst = &vmxnet3_recv_pkts;
300 	eth_dev->tx_pkt_burst = &vmxnet3_xmit_pkts;
301 	eth_dev->tx_pkt_prepare = vmxnet3_prep_pkts;
302 	pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
303 
304 	/*
305 	 * for secondary processes, we don't initialize any further as primary
306 	 * has already done this work.
307 	 */
308 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
309 		return 0;
310 
311 	rte_eth_copy_pci_info(eth_dev, pci_dev);
312 	eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
313 
314 	/* Vendor and Device ID need to be set before init of shared code */
315 	hw->device_id = pci_dev->id.device_id;
316 	hw->vendor_id = pci_dev->id.vendor_id;
317 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
318 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
319 
320 	hw->num_rx_queues = 1;
321 	hw->num_tx_queues = 1;
322 	hw->bufs_per_pkt = 1;
323 
324 	/* Check h/w version compatibility with driver. */
325 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
326 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
327 
328 	if (ver & (1 << VMXNET3_REV_3)) {
329 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
330 				       1 << VMXNET3_REV_3);
331 		hw->version = VMXNET3_REV_3 + 1;
332 	} else if (ver & (1 << VMXNET3_REV_2)) {
333 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
334 				       1 << VMXNET3_REV_2);
335 		hw->version = VMXNET3_REV_2 + 1;
336 	} else if (ver & (1 << VMXNET3_REV_1)) {
337 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
338 				       1 << VMXNET3_REV_1);
339 		hw->version = VMXNET3_REV_1 + 1;
340 	} else {
341 		PMD_INIT_LOG(ERR, "Incompatible hardware version: %d", ver);
342 		return -EIO;
343 	}
344 
345 	PMD_INIT_LOG(DEBUG, "Using device version %d\n", hw->version);
346 
347 	/* Check UPT version compatibility with driver. */
348 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
349 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
350 	if (ver & 0x1)
351 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
352 	else {
353 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
354 		return -EIO;
355 	}
356 
357 	/* Getting MAC Address */
358 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
359 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
360 	memcpy(hw->perm_addr, &mac_lo, 4);
361 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
362 
363 	/* Allocate memory for storing MAC addresses */
364 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
365 					       VMXNET3_MAX_MAC_ADDRS, 0);
366 	if (eth_dev->data->mac_addrs == NULL) {
367 		PMD_INIT_LOG(ERR,
368 			     "Failed to allocate %d bytes needed to store MAC addresses",
369 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
370 		return -ENOMEM;
371 	}
372 	/* Copy the permanent MAC address */
373 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
374 			&eth_dev->data->mac_addrs[0]);
375 
376 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
377 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
378 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
379 
380 	/* Put device in Quiesce Mode */
381 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
382 
383 	/* allow untagged pkts */
384 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
385 
386 	hw->txdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
387 		eth_vmxnet3_txdata_get(hw) : sizeof(struct Vmxnet3_TxDataDesc);
388 
389 	hw->rxdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
390 		VMXNET3_DEF_RXDATA_DESC_SIZE : 0;
391 	RTE_ASSERT((hw->rxdata_desc_size & ~VMXNET3_RXDATA_DESC_SIZE_MASK) ==
392 		   hw->rxdata_desc_size);
393 
394 	/* clear shadow stats */
395 	memset(hw->saved_tx_stats, 0, sizeof(hw->saved_tx_stats));
396 	memset(hw->saved_rx_stats, 0, sizeof(hw->saved_rx_stats));
397 
398 	return 0;
399 }
400 
401 static int
402 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
403 {
404 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
405 
406 	PMD_INIT_FUNC_TRACE();
407 
408 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
409 		return 0;
410 
411 	if (hw->adapter_stopped == 0)
412 		vmxnet3_dev_close(eth_dev);
413 
414 	eth_dev->dev_ops = NULL;
415 	eth_dev->rx_pkt_burst = NULL;
416 	eth_dev->tx_pkt_burst = NULL;
417 	eth_dev->tx_pkt_prepare = NULL;
418 
419 	rte_free(eth_dev->data->mac_addrs);
420 	eth_dev->data->mac_addrs = NULL;
421 
422 	return 0;
423 }
424 
425 static int eth_vmxnet3_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
426 	struct rte_pci_device *pci_dev)
427 {
428 	return rte_eth_dev_pci_generic_probe(pci_dev,
429 		sizeof(struct vmxnet3_hw), eth_vmxnet3_dev_init);
430 }
431 
432 static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
433 {
434 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_vmxnet3_dev_uninit);
435 }
436 
437 static struct rte_pci_driver rte_vmxnet3_pmd = {
438 	.id_table = pci_id_vmxnet3_map,
439 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
440 	.probe = eth_vmxnet3_pci_probe,
441 	.remove = eth_vmxnet3_pci_remove,
442 };
443 
444 static int
445 vmxnet3_dev_configure(struct rte_eth_dev *dev)
446 {
447 	const struct rte_memzone *mz;
448 	struct vmxnet3_hw *hw = dev->data->dev_private;
449 	size_t size;
450 
451 	PMD_INIT_FUNC_TRACE();
452 
453 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
454 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
455 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
456 		return -EINVAL;
457 	}
458 
459 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
460 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
461 		return -EINVAL;
462 	}
463 
464 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
465 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
466 
467 	if (size > UINT16_MAX)
468 		return -EINVAL;
469 
470 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
471 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
472 
473 	/*
474 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
475 	 * on current socket
476 	 */
477 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
478 			      "shared", rte_socket_id(), 8, 1);
479 
480 	if (mz == NULL) {
481 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
482 		return -ENOMEM;
483 	}
484 	memset(mz->addr, 0, mz->len);
485 
486 	hw->shared = mz->addr;
487 	hw->sharedPA = mz->phys_addr;
488 
489 	/*
490 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
491 	 * on current socket.
492 	 *
493 	 * We cannot reuse this memzone from previous allocation as its size
494 	 * depends on the number of tx and rx queues, which could be different
495 	 * from one config to another.
496 	 */
497 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
498 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
499 	if (mz == NULL) {
500 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
501 		return -ENOMEM;
502 	}
503 	memset(mz->addr, 0, mz->len);
504 
505 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
506 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
507 
508 	hw->queueDescPA = mz->phys_addr;
509 	hw->queue_desc_len = (uint16_t)size;
510 
511 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
512 		/* Allocate memory structure for UPT1_RSSConf and configure */
513 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
514 				      "rss_conf", rte_socket_id(),
515 				      RTE_CACHE_LINE_SIZE, 1);
516 		if (mz == NULL) {
517 			PMD_INIT_LOG(ERR,
518 				     "ERROR: Creating rss_conf structure zone");
519 			return -ENOMEM;
520 		}
521 		memset(mz->addr, 0, mz->len);
522 
523 		hw->rss_conf = mz->addr;
524 		hw->rss_confPA = mz->phys_addr;
525 	}
526 
527 	return 0;
528 }
529 
530 static void
531 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
532 {
533 	uint32_t val;
534 
535 	PMD_INIT_LOG(DEBUG,
536 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
537 		     addr[0], addr[1], addr[2],
538 		     addr[3], addr[4], addr[5]);
539 
540 	memcpy(&val, addr, 4);
541 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
542 
543 	memcpy(&val, addr + 4, 2);
544 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
545 }
546 
547 static int
548 vmxnet3_dev_setup_memreg(struct rte_eth_dev *dev)
549 {
550 	struct vmxnet3_hw *hw = dev->data->dev_private;
551 	Vmxnet3_DriverShared *shared = hw->shared;
552 	Vmxnet3_CmdInfo *cmdInfo;
553 	struct rte_mempool *mp[VMXNET3_MAX_RX_QUEUES];
554 	uint8_t index[VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES];
555 	uint32_t num, i, j, size;
556 
557 	if (hw->memRegsPA == 0) {
558 		const struct rte_memzone *mz;
559 
560 		size = sizeof(Vmxnet3_MemRegs) +
561 			(VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES) *
562 			sizeof(Vmxnet3_MemoryRegion);
563 
564 		mz = gpa_zone_reserve(dev, size, "memRegs", rte_socket_id(), 8,
565 				      1);
566 		if (mz == NULL) {
567 			PMD_INIT_LOG(ERR, "ERROR: Creating memRegs zone");
568 			return -ENOMEM;
569 		}
570 		memset(mz->addr, 0, mz->len);
571 		hw->memRegs = mz->addr;
572 		hw->memRegsPA = mz->phys_addr;
573 	}
574 
575 	num = hw->num_rx_queues;
576 
577 	for (i = 0; i < num; i++) {
578 		vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
579 
580 		mp[i] = rxq->mp;
581 		index[i] = 1 << i;
582 	}
583 
584 	/*
585 	 * The same mempool could be used by multiple queues. In such a case,
586 	 * remove duplicate mempool entries. Only one entry is kept with
587 	 * bitmask indicating queues that are using this mempool.
588 	 */
589 	for (i = 1; i < num; i++) {
590 		for (j = 0; j < i; j++) {
591 			if (mp[i] == mp[j]) {
592 				mp[i] = NULL;
593 				index[j] |= 1 << i;
594 				break;
595 			}
596 		}
597 	}
598 
599 	j = 0;
600 	for (i = 0; i < num; i++) {
601 		if (mp[i] == NULL)
602 			continue;
603 
604 		Vmxnet3_MemoryRegion *mr = &hw->memRegs->memRegs[j];
605 
606 		mr->startPA =
607 			(uintptr_t)STAILQ_FIRST(&mp[i]->mem_list)->phys_addr;
608 		mr->length = STAILQ_FIRST(&mp[i]->mem_list)->len <= INT32_MAX ?
609 			STAILQ_FIRST(&mp[i]->mem_list)->len : INT32_MAX;
610 		mr->txQueueBits = index[i];
611 		mr->rxQueueBits = index[i];
612 
613 		PMD_INIT_LOG(INFO,
614 			     "index: %u startPA: %" PRIu64 " length: %u, "
615 			     "rxBits: %x",
616 			     j, mr->startPA, mr->length, mr->rxQueueBits);
617 		j++;
618 	}
619 	hw->memRegs->numRegs = j;
620 	PMD_INIT_LOG(INFO, "numRegs: %u", j);
621 
622 	size = sizeof(Vmxnet3_MemRegs) +
623 		(j - 1) * sizeof(Vmxnet3_MemoryRegion);
624 
625 	cmdInfo = &shared->cu.cmdInfo;
626 	cmdInfo->varConf.confVer = 1;
627 	cmdInfo->varConf.confLen = size;
628 	cmdInfo->varConf.confPA = hw->memRegsPA;
629 
630 	return 0;
631 }
632 
633 static int
634 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
635 {
636 	struct rte_eth_conf port_conf = dev->data->dev_conf;
637 	struct vmxnet3_hw *hw = dev->data->dev_private;
638 	uint32_t mtu = dev->data->mtu;
639 	Vmxnet3_DriverShared *shared = hw->shared;
640 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
641 	uint32_t i;
642 	int ret;
643 
644 	shared->magic = VMXNET3_REV1_MAGIC;
645 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
646 
647 	/* Setting up Guest OS information */
648 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
649 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
650 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
651 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
652 	devRead->misc.driverInfo.uptVerSpt     = 1;
653 
654 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
655 	devRead->misc.queueDescPA  = hw->queueDescPA;
656 	devRead->misc.queueDescLen = hw->queue_desc_len;
657 	devRead->misc.numTxQueues  = hw->num_tx_queues;
658 	devRead->misc.numRxQueues  = hw->num_rx_queues;
659 
660 	/*
661 	 * Set number of interrupts to 1
662 	 * PMD by default disables all the interrupts but this is MUST
663 	 * to activate device. It needs at least one interrupt for
664 	 * link events to handle
665 	 */
666 	hw->num_intrs = devRead->intrConf.numIntrs = 1;
667 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
668 
669 	for (i = 0; i < hw->num_tx_queues; i++) {
670 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
671 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
672 
673 		tqd->ctrl.txNumDeferred  = 0;
674 		tqd->ctrl.txThreshold    = 1;
675 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
676 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
677 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
678 
679 		tqd->conf.txRingSize   = txq->cmd_ring.size;
680 		tqd->conf.compRingSize = txq->comp_ring.size;
681 		tqd->conf.dataRingSize = txq->data_ring.size;
682 		tqd->conf.txDataRingDescSize = txq->txdata_desc_size;
683 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
684 		tqd->status.stopped    = TRUE;
685 		tqd->status.error      = 0;
686 		memset(&tqd->stats, 0, sizeof(tqd->stats));
687 	}
688 
689 	for (i = 0; i < hw->num_rx_queues; i++) {
690 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
691 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
692 
693 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
694 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
695 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
696 
697 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
698 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
699 		rqd->conf.compRingSize    = rxq->comp_ring.size;
700 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
701 		if (VMXNET3_VERSION_GE_3(hw)) {
702 			rqd->conf.rxDataRingBasePA = rxq->data_ring.basePA;
703 			rqd->conf.rxDataRingDescSize = rxq->data_desc_size;
704 		}
705 		rqd->status.stopped       = TRUE;
706 		rqd->status.error         = 0;
707 		memset(&rqd->stats, 0, sizeof(rqd->stats));
708 	}
709 
710 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
711 	devRead->rxFilterConf.rxMode = 0;
712 
713 	/* Setting up feature flags */
714 	if (dev->data->dev_conf.rxmode.hw_ip_checksum)
715 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
716 
717 	if (dev->data->dev_conf.rxmode.enable_lro) {
718 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
719 		devRead->misc.maxNumRxSG = 0;
720 	}
721 
722 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
723 		ret = vmxnet3_rss_configure(dev);
724 		if (ret != VMXNET3_SUCCESS)
725 			return ret;
726 
727 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
728 		devRead->rssConfDesc.confVer = 1;
729 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
730 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
731 	}
732 
733 	vmxnet3_dev_vlan_offload_set(dev,
734 				     ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
735 
736 	vmxnet3_write_mac(hw, dev->data->mac_addrs->addr_bytes);
737 
738 	return VMXNET3_SUCCESS;
739 }
740 
741 /*
742  * Configure device link speed and setup link.
743  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
744  * It returns 0 on success.
745  */
746 static int
747 vmxnet3_dev_start(struct rte_eth_dev *dev)
748 {
749 	int ret;
750 	struct vmxnet3_hw *hw = dev->data->dev_private;
751 
752 	PMD_INIT_FUNC_TRACE();
753 
754 	/* Save stats before it is reset by CMD_ACTIVATE */
755 	vmxnet3_hw_stats_save(hw);
756 
757 	ret = vmxnet3_setup_driver_shared(dev);
758 	if (ret != VMXNET3_SUCCESS)
759 		return ret;
760 
761 	/* check if lsc interrupt feature is enabled */
762 	if (dev->data->dev_conf.intr_conf.lsc) {
763 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
764 
765 		/* Setup interrupt callback  */
766 		rte_intr_callback_register(&pci_dev->intr_handle,
767 					   vmxnet3_interrupt_handler, dev);
768 
769 		if (rte_intr_enable(&pci_dev->intr_handle) < 0) {
770 			PMD_INIT_LOG(ERR, "interrupt enable failed");
771 			return -EIO;
772 		}
773 	}
774 
775 	/* Exchange shared data with device */
776 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
777 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
778 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
779 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
780 
781 	/* Activate device by register write */
782 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
783 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
784 
785 	if (ret != 0) {
786 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
787 		return -EINVAL;
788 	}
789 
790 	/* Setup memory region for rx buffers */
791 	ret = vmxnet3_dev_setup_memreg(dev);
792 	if (ret == 0) {
793 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
794 				       VMXNET3_CMD_REGISTER_MEMREGS);
795 		ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
796 		if (ret != 0)
797 			PMD_INIT_LOG(DEBUG,
798 				     "Failed in setup memory region cmd\n");
799 		ret = 0;
800 	} else {
801 		PMD_INIT_LOG(DEBUG, "Failed to setup memory region\n");
802 	}
803 
804 	/* Disable interrupts */
805 	vmxnet3_disable_intr(hw);
806 
807 	/*
808 	 * Load RX queues with blank mbufs and update next2fill index for device
809 	 * Update RxMode of the device
810 	 */
811 	ret = vmxnet3_dev_rxtx_init(dev);
812 	if (ret != VMXNET3_SUCCESS) {
813 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
814 		return ret;
815 	}
816 
817 	hw->adapter_stopped = FALSE;
818 
819 	/* Setting proper Rx Mode and issue Rx Mode Update command */
820 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
821 
822 	if (dev->data->dev_conf.intr_conf.lsc) {
823 		vmxnet3_enable_intr(hw);
824 
825 		/*
826 		 * Update link state from device since this won't be
827 		 * done upon starting with lsc in use. This is done
828 		 * only after enabling interrupts to avoid any race
829 		 * where the link state could change without an
830 		 * interrupt being fired.
831 		 */
832 		__vmxnet3_dev_link_update(dev, 0);
833 	}
834 
835 	return VMXNET3_SUCCESS;
836 }
837 
838 /*
839  * Stop device: disable rx and tx functions to allow for reconfiguring.
840  */
841 static void
842 vmxnet3_dev_stop(struct rte_eth_dev *dev)
843 {
844 	struct rte_eth_link link;
845 	struct vmxnet3_hw *hw = dev->data->dev_private;
846 
847 	PMD_INIT_FUNC_TRACE();
848 
849 	if (hw->adapter_stopped == 1) {
850 		PMD_INIT_LOG(DEBUG, "Device already closed.");
851 		return;
852 	}
853 
854 	/* disable interrupts */
855 	vmxnet3_disable_intr(hw);
856 
857 	if (dev->data->dev_conf.intr_conf.lsc) {
858 		struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
859 
860 		rte_intr_disable(&pci_dev->intr_handle);
861 
862 		rte_intr_callback_unregister(&pci_dev->intr_handle,
863 					     vmxnet3_interrupt_handler, dev);
864 	}
865 
866 	/* quiesce the device first */
867 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
868 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
869 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
870 
871 	/* reset the device */
872 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
873 	PMD_INIT_LOG(DEBUG, "Device reset.");
874 	hw->adapter_stopped = 0;
875 
876 	vmxnet3_dev_clear_queues(dev);
877 
878 	/* Clear recorded link status */
879 	memset(&link, 0, sizeof(link));
880 	vmxnet3_dev_atomic_write_link_status(dev, &link);
881 }
882 
883 /*
884  * Reset and stop device.
885  */
886 static void
887 vmxnet3_dev_close(struct rte_eth_dev *dev)
888 {
889 	struct vmxnet3_hw *hw = dev->data->dev_private;
890 
891 	PMD_INIT_FUNC_TRACE();
892 
893 	vmxnet3_dev_stop(dev);
894 	hw->adapter_stopped = 1;
895 }
896 
897 static void
898 vmxnet3_hw_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
899 			struct UPT1_TxStats *res)
900 {
901 #define VMXNET3_UPDATE_TX_STAT(h, i, f, r)		\
902 		((r)->f = (h)->tqd_start[(i)].stats.f +	\
903 			(h)->saved_tx_stats[(i)].f)
904 
905 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastPktsTxOK, res);
906 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastPktsTxOK, res);
907 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastPktsTxOK, res);
908 	VMXNET3_UPDATE_TX_STAT(hw, q, ucastBytesTxOK, res);
909 	VMXNET3_UPDATE_TX_STAT(hw, q, mcastBytesTxOK, res);
910 	VMXNET3_UPDATE_TX_STAT(hw, q, bcastBytesTxOK, res);
911 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxError, res);
912 	VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxDiscard, res);
913 
914 #undef VMXNET3_UPDATE_TX_STAT
915 }
916 
917 static void
918 vmxnet3_hw_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
919 			struct UPT1_RxStats *res)
920 {
921 #define VMXNET3_UPDATE_RX_STAT(h, i, f, r)		\
922 		((r)->f = (h)->rqd_start[(i)].stats.f +	\
923 			(h)->saved_rx_stats[(i)].f)
924 
925 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastPktsRxOK, res);
926 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastPktsRxOK, res);
927 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastPktsRxOK, res);
928 	VMXNET3_UPDATE_RX_STAT(hw, q, ucastBytesRxOK, res);
929 	VMXNET3_UPDATE_RX_STAT(hw, q, mcastBytesRxOK, res);
930 	VMXNET3_UPDATE_RX_STAT(hw, q, bcastBytesRxOK, res);
931 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxError, res);
932 	VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxOutOfBuf, res);
933 
934 #undef VMXNET3_UPDATE_RX_STATS
935 }
936 
937 static void
938 vmxnet3_hw_stats_save(struct vmxnet3_hw *hw)
939 {
940 	unsigned int i;
941 
942 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
943 
944 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
945 
946 	for (i = 0; i < hw->num_tx_queues; i++)
947 		vmxnet3_hw_tx_stats_get(hw, i, &hw->saved_tx_stats[i]);
948 	for (i = 0; i < hw->num_rx_queues; i++)
949 		vmxnet3_hw_rx_stats_get(hw, i, &hw->saved_rx_stats[i]);
950 }
951 
952 static int
953 vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
954 			     struct rte_eth_xstat_name *xstats_names,
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 (!xstats_names || n < nstats)
963 		return nstats;
964 
965 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
966 		if (!dev->data->rx_queues[i])
967 			continue;
968 
969 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
970 			snprintf(xstats_names[count].name,
971 				 sizeof(xstats_names[count].name),
972 				 "rx_q%u_%s", i,
973 				 vmxnet3_rxq_stat_strings[t].name);
974 			count++;
975 		}
976 	}
977 
978 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
979 		if (!dev->data->tx_queues[i])
980 			continue;
981 
982 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
983 			snprintf(xstats_names[count].name,
984 				 sizeof(xstats_names[count].name),
985 				 "tx_q%u_%s", i,
986 				 vmxnet3_txq_stat_strings[t].name);
987 			count++;
988 		}
989 	}
990 
991 	return count;
992 }
993 
994 static int
995 vmxnet3_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
996 		       unsigned int n)
997 {
998 	unsigned int i, t, count = 0;
999 	unsigned int nstats =
1000 		dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
1001 		dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
1002 
1003 	if (n < nstats)
1004 		return nstats;
1005 
1006 	for (i = 0; i < dev->data->nb_rx_queues; i++) {
1007 		struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
1008 
1009 		if (rxq == NULL)
1010 			continue;
1011 
1012 		for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
1013 			xstats[count].value = *(uint64_t *)(((char *)&rxq->stats) +
1014 				vmxnet3_rxq_stat_strings[t].offset);
1015 			xstats[count].id = count;
1016 			count++;
1017 		}
1018 	}
1019 
1020 	for (i = 0; i < dev->data->nb_tx_queues; i++) {
1021 		struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
1022 
1023 		if (txq == NULL)
1024 			continue;
1025 
1026 		for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
1027 			xstats[count].value = *(uint64_t *)(((char *)&txq->stats) +
1028 				vmxnet3_txq_stat_strings[t].offset);
1029 			xstats[count].id = count;
1030 			count++;
1031 		}
1032 	}
1033 
1034 	return count;
1035 }
1036 
1037 static void
1038 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
1039 {
1040 	unsigned int i;
1041 	struct vmxnet3_hw *hw = dev->data->dev_private;
1042 	struct UPT1_TxStats txStats;
1043 	struct UPT1_RxStats rxStats;
1044 
1045 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
1046 
1047 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
1048 	for (i = 0; i < hw->num_tx_queues; i++) {
1049 		vmxnet3_hw_tx_stats_get(hw, i, &txStats);
1050 
1051 		stats->q_opackets[i] = txStats.ucastPktsTxOK +
1052 			txStats.mcastPktsTxOK +
1053 			txStats.bcastPktsTxOK;
1054 
1055 		stats->q_obytes[i] = txStats.ucastBytesTxOK +
1056 			txStats.mcastBytesTxOK +
1057 			txStats.bcastBytesTxOK;
1058 
1059 		stats->opackets += stats->q_opackets[i];
1060 		stats->obytes += stats->q_obytes[i];
1061 		stats->oerrors += txStats.pktsTxError + txStats.pktsTxDiscard;
1062 	}
1063 
1064 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
1065 	for (i = 0; i < hw->num_rx_queues; i++) {
1066 		vmxnet3_hw_rx_stats_get(hw, i, &rxStats);
1067 
1068 		stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
1069 			rxStats.mcastPktsRxOK +
1070 			rxStats.bcastPktsRxOK;
1071 
1072 		stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
1073 			rxStats.mcastBytesRxOK +
1074 			rxStats.bcastBytesRxOK;
1075 
1076 		stats->ipackets += stats->q_ipackets[i];
1077 		stats->ibytes += stats->q_ibytes[i];
1078 
1079 		stats->q_errors[i] = rxStats.pktsRxError;
1080 		stats->ierrors += rxStats.pktsRxError;
1081 		stats->rx_nombuf += rxStats.pktsRxOutOfBuf;
1082 	}
1083 }
1084 
1085 static void
1086 vmxnet3_dev_info_get(struct rte_eth_dev *dev,
1087 		     struct rte_eth_dev_info *dev_info)
1088 {
1089 	dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1090 
1091 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
1092 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
1093 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
1094 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
1095 	dev_info->speed_capa = ETH_LINK_SPEED_10G;
1096 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
1097 
1098 	dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
1099 	dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
1100 
1101 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1102 		.nb_max = VMXNET3_RX_RING_MAX_SIZE,
1103 		.nb_min = VMXNET3_DEF_RX_RING_SIZE,
1104 		.nb_align = 1,
1105 	};
1106 
1107 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1108 		.nb_max = VMXNET3_TX_RING_MAX_SIZE,
1109 		.nb_min = VMXNET3_DEF_TX_RING_SIZE,
1110 		.nb_align = 1,
1111 		.nb_seg_max = VMXNET3_TX_MAX_SEG,
1112 		.nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
1113 	};
1114 
1115 	dev_info->rx_offload_capa =
1116 		DEV_RX_OFFLOAD_VLAN_STRIP |
1117 		DEV_RX_OFFLOAD_UDP_CKSUM |
1118 		DEV_RX_OFFLOAD_TCP_CKSUM |
1119 		DEV_RX_OFFLOAD_TCP_LRO;
1120 
1121 	dev_info->tx_offload_capa =
1122 		DEV_TX_OFFLOAD_VLAN_INSERT |
1123 		DEV_TX_OFFLOAD_TCP_CKSUM |
1124 		DEV_TX_OFFLOAD_UDP_CKSUM |
1125 		DEV_TX_OFFLOAD_TCP_TSO;
1126 }
1127 
1128 static const uint32_t *
1129 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
1130 {
1131 	static const uint32_t ptypes[] = {
1132 		RTE_PTYPE_L3_IPV4_EXT,
1133 		RTE_PTYPE_L3_IPV4,
1134 		RTE_PTYPE_UNKNOWN
1135 	};
1136 
1137 	if (dev->rx_pkt_burst == vmxnet3_recv_pkts)
1138 		return ptypes;
1139 	return NULL;
1140 }
1141 
1142 static void
1143 vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
1144 {
1145 	struct vmxnet3_hw *hw = dev->data->dev_private;
1146 
1147 	ether_addr_copy(mac_addr, (struct ether_addr *)(hw->perm_addr));
1148 	ether_addr_copy(mac_addr, &dev->data->mac_addrs[0]);
1149 	vmxnet3_write_mac(hw, mac_addr->addr_bytes);
1150 }
1151 
1152 /* return 0 means link status changed, -1 means not changed */
1153 static int
1154 __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
1155 			  __rte_unused int wait_to_complete)
1156 {
1157 	struct vmxnet3_hw *hw = dev->data->dev_private;
1158 	struct rte_eth_link old = { 0 }, link;
1159 	uint32_t ret;
1160 
1161 	memset(&link, 0, sizeof(link));
1162 	vmxnet3_dev_atomic_read_link_status(dev, &old);
1163 
1164 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
1165 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
1166 
1167 	if (ret & 0x1) {
1168 		link.link_status = ETH_LINK_UP;
1169 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
1170 		link.link_speed = ETH_SPEED_NUM_10G;
1171 		link.link_autoneg = ETH_LINK_SPEED_FIXED;
1172 	}
1173 
1174 	vmxnet3_dev_atomic_write_link_status(dev, &link);
1175 
1176 	return (old.link_status == link.link_status) ? -1 : 0;
1177 }
1178 
1179 static int
1180 vmxnet3_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1181 {
1182 	/* Link status doesn't change for stopped dev */
1183 	if (dev->data->dev_started == 0)
1184 		return -1;
1185 
1186 	return __vmxnet3_dev_link_update(dev, wait_to_complete);
1187 }
1188 
1189 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
1190 static void
1191 vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set)
1192 {
1193 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1194 
1195 	if (set)
1196 		rxConf->rxMode = rxConf->rxMode | feature;
1197 	else
1198 		rxConf->rxMode = rxConf->rxMode & (~feature);
1199 
1200 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_UPDATE_RX_MODE);
1201 }
1202 
1203 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1204 static void
1205 vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev)
1206 {
1207 	struct vmxnet3_hw *hw = dev->data->dev_private;
1208 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1209 
1210 	memset(vf_table, 0, VMXNET3_VFT_TABLE_SIZE);
1211 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 1);
1212 
1213 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1214 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1215 }
1216 
1217 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
1218 static void
1219 vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
1220 {
1221 	struct vmxnet3_hw *hw = dev->data->dev_private;
1222 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
1223 
1224 	if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1225 		memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1226 	else
1227 		memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1228 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
1229 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1230 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1231 }
1232 
1233 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1234 static void
1235 vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev)
1236 {
1237 	struct vmxnet3_hw *hw = dev->data->dev_private;
1238 
1239 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 1);
1240 }
1241 
1242 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
1243 static void
1244 vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev)
1245 {
1246 	struct vmxnet3_hw *hw = dev->data->dev_private;
1247 
1248 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 0);
1249 }
1250 
1251 /* Enable/disable filter on vlan */
1252 static int
1253 vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
1254 {
1255 	struct vmxnet3_hw *hw = dev->data->dev_private;
1256 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
1257 	uint32_t *vf_table = rxConf->vfTable;
1258 
1259 	/* save state for restore */
1260 	if (on)
1261 		VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1262 	else
1263 		VMXNET3_CLEAR_VFTABLE_ENTRY(hw->shadow_vfta, vid);
1264 
1265 	/* don't change active filter if in promiscuous mode */
1266 	if (rxConf->rxMode & VMXNET3_RXM_PROMISC)
1267 		return 0;
1268 
1269 	/* set in hardware */
1270 	if (on)
1271 		VMXNET3_SET_VFTABLE_ENTRY(vf_table, vid);
1272 	else
1273 		VMXNET3_CLEAR_VFTABLE_ENTRY(vf_table, vid);
1274 
1275 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1276 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1277 	return 0;
1278 }
1279 
1280 static void
1281 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1282 {
1283 	struct vmxnet3_hw *hw = dev->data->dev_private;
1284 	Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
1285 	uint32_t *vf_table = devRead->rxFilterConf.vfTable;
1286 
1287 	if (mask & ETH_VLAN_STRIP_MASK) {
1288 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1289 			devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
1290 		else
1291 			devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
1292 
1293 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1294 				       VMXNET3_CMD_UPDATE_FEATURE);
1295 	}
1296 
1297 	if (mask & ETH_VLAN_FILTER_MASK) {
1298 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1299 			memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
1300 		else
1301 			memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
1302 
1303 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1304 				       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
1305 	}
1306 }
1307 
1308 static void
1309 vmxnet3_process_events(struct rte_eth_dev *dev)
1310 {
1311 	struct vmxnet3_hw *hw = dev->data->dev_private;
1312 	uint32_t events = hw->shared->ecr;
1313 
1314 	if (!events)
1315 		return;
1316 
1317 	/*
1318 	 * ECR bits when written with 1b are cleared. Hence write
1319 	 * events back to ECR so that the bits which were set will be reset.
1320 	 */
1321 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
1322 
1323 	/* Check if link state has changed */
1324 	if (events & VMXNET3_ECR_LINK) {
1325 		PMD_DRV_LOG(DEBUG, "Process events: VMXNET3_ECR_LINK event");
1326 		if (vmxnet3_dev_link_update(dev, 0) == 0)
1327 			_rte_eth_dev_callback_process(dev,
1328 						      RTE_ETH_EVENT_INTR_LSC,
1329 						      NULL, NULL);
1330 	}
1331 
1332 	/* Check if there is an error on xmit/recv queues */
1333 	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
1334 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
1335 				       VMXNET3_CMD_GET_QUEUE_STATUS);
1336 
1337 		if (hw->tqd_start->status.stopped)
1338 			PMD_DRV_LOG(ERR, "tq error 0x%x",
1339 				    hw->tqd_start->status.error);
1340 
1341 		if (hw->rqd_start->status.stopped)
1342 			PMD_DRV_LOG(ERR, "rq error 0x%x",
1343 				     hw->rqd_start->status.error);
1344 
1345 		/* Reset the device */
1346 		/* Have to reset the device */
1347 	}
1348 
1349 	if (events & VMXNET3_ECR_DIC)
1350 		PMD_DRV_LOG(DEBUG, "Device implementation change event.");
1351 
1352 	if (events & VMXNET3_ECR_DEBUG)
1353 		PMD_DRV_LOG(DEBUG, "Debug event generated by device.");
1354 }
1355 
1356 static void
1357 vmxnet3_interrupt_handler(void *param)
1358 {
1359 	struct rte_eth_dev *dev = param;
1360 	struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
1361 
1362 	vmxnet3_process_events(dev);
1363 
1364 	if (rte_intr_enable(&pci_dev->intr_handle) < 0)
1365 		PMD_DRV_LOG(ERR, "interrupt enable failed");
1366 }
1367 
1368 RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd);
1369 RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
1370 RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio-pci");
1371