xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision 6b72aad61f09b870427f4ebcedae525a985bb37a)
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_atomic.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 void vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
88 				  struct rte_eth_stats *stats);
89 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
90 				 struct rte_eth_dev_info *dev_info);
91 static const uint32_t *
92 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
93 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
94 				       uint16_t vid, int on);
95 static void vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
96 static void vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
97 				 struct ether_addr *mac_addr);
98 
99 #if PROCESS_SYS_EVENTS == 1
100 static void vmxnet3_process_events(struct vmxnet3_hw *);
101 #endif
102 /*
103  * The set of PCI devices this driver supports
104  */
105 #define VMWARE_PCI_VENDOR_ID 0x15AD
106 #define VMWARE_DEV_ID_VMXNET3 0x07B0
107 static const struct rte_pci_id pci_id_vmxnet3_map[] = {
108 	{ RTE_PCI_DEVICE(VMWARE_PCI_VENDOR_ID, VMWARE_DEV_ID_VMXNET3) },
109 	{ .vendor_id = 0, /* sentinel */ },
110 };
111 
112 static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
113 	.dev_configure        = vmxnet3_dev_configure,
114 	.dev_start            = vmxnet3_dev_start,
115 	.dev_stop             = vmxnet3_dev_stop,
116 	.dev_close            = vmxnet3_dev_close,
117 	.promiscuous_enable   = vmxnet3_dev_promiscuous_enable,
118 	.promiscuous_disable  = vmxnet3_dev_promiscuous_disable,
119 	.allmulticast_enable  = vmxnet3_dev_allmulticast_enable,
120 	.allmulticast_disable = vmxnet3_dev_allmulticast_disable,
121 	.link_update          = vmxnet3_dev_link_update,
122 	.stats_get            = vmxnet3_dev_stats_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 static const struct rte_memzone *
135 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
136 		 const char *post_string, int socket_id,
137 		 uint16_t align, bool reuse)
138 {
139 	char z_name[RTE_MEMZONE_NAMESIZE];
140 	const struct rte_memzone *mz;
141 
142 	snprintf(z_name, sizeof(z_name), "%s_%d_%s",
143 		 dev->data->drv_name, dev->data->port_id, post_string);
144 
145 	mz = rte_memzone_lookup(z_name);
146 	if (!reuse) {
147 		if (mz)
148 			rte_memzone_free(mz);
149 		return rte_memzone_reserve_aligned(z_name, size, socket_id,
150 						   0, align);
151 	}
152 
153 	if (mz)
154 		return mz;
155 
156 	return rte_memzone_reserve_aligned(z_name, size, socket_id, 0, align);
157 }
158 
159 /**
160  * Atomically reads the link status information from global
161  * structure rte_eth_dev.
162  *
163  * @param dev
164  *   - Pointer to the structure rte_eth_dev to read from.
165  *   - Pointer to the buffer to be saved with the link status.
166  *
167  * @return
168  *   - On success, zero.
169  *   - On failure, negative value.
170  */
171 
172 static int
173 vmxnet3_dev_atomic_read_link_status(struct rte_eth_dev *dev,
174 				    struct rte_eth_link *link)
175 {
176 	struct rte_eth_link *dst = link;
177 	struct rte_eth_link *src = &(dev->data->dev_link);
178 
179 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
180 				*(uint64_t *)src) == 0)
181 		return -1;
182 
183 	return 0;
184 }
185 
186 /**
187  * Atomically writes the link status information into global
188  * structure rte_eth_dev.
189  *
190  * @param dev
191  *   - Pointer to the structure rte_eth_dev to write to.
192  *   - Pointer to the buffer to be saved with the link status.
193  *
194  * @return
195  *   - On success, zero.
196  *   - On failure, negative value.
197  */
198 static int
199 vmxnet3_dev_atomic_write_link_status(struct rte_eth_dev *dev,
200 				     struct rte_eth_link *link)
201 {
202 	struct rte_eth_link *dst = &(dev->data->dev_link);
203 	struct rte_eth_link *src = link;
204 
205 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
206 				*(uint64_t *)src) == 0)
207 		return -1;
208 
209 	return 0;
210 }
211 
212 /*
213  * This function is based on vmxnet3_disable_intr()
214  */
215 static void
216 vmxnet3_disable_intr(struct vmxnet3_hw *hw)
217 {
218 	int i;
219 
220 	PMD_INIT_FUNC_TRACE();
221 
222 	hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
223 	for (i = 0; i < VMXNET3_MAX_INTRS; i++)
224 		VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
225 }
226 
227 /*
228  * It returns 0 on success.
229  */
230 static int
231 eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
232 {
233 	struct rte_pci_device *pci_dev;
234 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
235 	uint32_t mac_hi, mac_lo, ver;
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_DEV_TO_PCI(eth_dev->device);
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 	eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE;
254 
255 	/* Vendor and Device ID need to be set before init of shared code */
256 	hw->device_id = pci_dev->id.device_id;
257 	hw->vendor_id = pci_dev->id.vendor_id;
258 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
259 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
260 
261 	hw->num_rx_queues = 1;
262 	hw->num_tx_queues = 1;
263 	hw->bufs_per_pkt = 1;
264 
265 	/* Check h/w version compatibility with driver. */
266 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
267 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
268 	if (ver & 0x1)
269 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS, 1);
270 	else {
271 		PMD_INIT_LOG(ERR, "Incompatible h/w version, should be 0x1");
272 		return -EIO;
273 	}
274 
275 	/* Check UPT version compatibility with driver. */
276 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
277 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
278 	if (ver & 0x1)
279 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
280 	else {
281 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
282 		return -EIO;
283 	}
284 
285 	/* Getting MAC Address */
286 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
287 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
288 	memcpy(hw->perm_addr, &mac_lo, 4);
289 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
290 
291 	/* Allocate memory for storing MAC addresses */
292 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
293 					       VMXNET3_MAX_MAC_ADDRS, 0);
294 	if (eth_dev->data->mac_addrs == NULL) {
295 		PMD_INIT_LOG(ERR,
296 			     "Failed to allocate %d bytes needed to store MAC addresses",
297 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
298 		return -ENOMEM;
299 	}
300 	/* Copy the permanent MAC address */
301 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
302 			&eth_dev->data->mac_addrs[0]);
303 
304 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
305 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
306 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
307 
308 	/* Put device in Quiesce Mode */
309 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
310 
311 	/* allow untagged pkts */
312 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
313 
314 	return 0;
315 }
316 
317 static int
318 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
319 {
320 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
321 
322 	PMD_INIT_FUNC_TRACE();
323 
324 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
325 		return 0;
326 
327 	if (hw->adapter_stopped == 0)
328 		vmxnet3_dev_close(eth_dev);
329 
330 	eth_dev->dev_ops = NULL;
331 	eth_dev->rx_pkt_burst = NULL;
332 	eth_dev->tx_pkt_burst = NULL;
333 	eth_dev->tx_pkt_prepare = NULL;
334 
335 	rte_free(eth_dev->data->mac_addrs);
336 	eth_dev->data->mac_addrs = NULL;
337 
338 	return 0;
339 }
340 
341 static struct eth_driver rte_vmxnet3_pmd = {
342 	.pci_drv = {
343 		.id_table = pci_id_vmxnet3_map,
344 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING,
345 		.probe = rte_eth_dev_pci_probe,
346 		.remove = rte_eth_dev_pci_remove,
347 	},
348 	.eth_dev_init = eth_vmxnet3_dev_init,
349 	.eth_dev_uninit = eth_vmxnet3_dev_uninit,
350 	.dev_private_size = sizeof(struct vmxnet3_hw),
351 };
352 
353 static int
354 vmxnet3_dev_configure(struct rte_eth_dev *dev)
355 {
356 	const struct rte_memzone *mz;
357 	struct vmxnet3_hw *hw = dev->data->dev_private;
358 	size_t size;
359 
360 	PMD_INIT_FUNC_TRACE();
361 
362 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
363 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
364 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
365 		return -EINVAL;
366 	}
367 
368 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
369 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
370 		return -EINVAL;
371 	}
372 
373 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
374 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
375 
376 	if (size > UINT16_MAX)
377 		return -EINVAL;
378 
379 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
380 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
381 
382 	/*
383 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
384 	 * on current socket
385 	 */
386 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
387 			      "shared", rte_socket_id(), 8, 1);
388 
389 	if (mz == NULL) {
390 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
391 		return -ENOMEM;
392 	}
393 	memset(mz->addr, 0, mz->len);
394 
395 	hw->shared = mz->addr;
396 	hw->sharedPA = mz->phys_addr;
397 
398 	/*
399 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
400 	 * on current socket.
401 	 *
402 	 * We cannot reuse this memzone from previous allocation as its size
403 	 * depends on the number of tx and rx queues, which could be different
404 	 * from one config to another.
405 	 */
406 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
407 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
408 	if (mz == NULL) {
409 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
410 		return -ENOMEM;
411 	}
412 	memset(mz->addr, 0, mz->len);
413 
414 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
415 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
416 
417 	hw->queueDescPA = mz->phys_addr;
418 	hw->queue_desc_len = (uint16_t)size;
419 
420 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
421 		/* Allocate memory structure for UPT1_RSSConf and configure */
422 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
423 				      "rss_conf", rte_socket_id(),
424 				      RTE_CACHE_LINE_SIZE, 1);
425 		if (mz == NULL) {
426 			PMD_INIT_LOG(ERR,
427 				     "ERROR: Creating rss_conf structure zone");
428 			return -ENOMEM;
429 		}
430 		memset(mz->addr, 0, mz->len);
431 
432 		hw->rss_conf = mz->addr;
433 		hw->rss_confPA = mz->phys_addr;
434 	}
435 
436 	return 0;
437 }
438 
439 static void
440 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
441 {
442 	uint32_t val;
443 
444 	PMD_INIT_LOG(DEBUG,
445 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
446 		     addr[0], addr[1], addr[2],
447 		     addr[3], addr[4], addr[5]);
448 
449 	val = *(const uint32_t *)addr;
450 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
451 
452 	val = (addr[5] << 8) | addr[4];
453 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
454 }
455 
456 static int
457 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
458 {
459 	struct rte_eth_conf port_conf = dev->data->dev_conf;
460 	struct vmxnet3_hw *hw = dev->data->dev_private;
461 	uint32_t mtu = dev->data->mtu;
462 	Vmxnet3_DriverShared *shared = hw->shared;
463 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
464 	uint32_t i;
465 	int ret;
466 
467 	shared->magic = VMXNET3_REV1_MAGIC;
468 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
469 
470 	/* Setting up Guest OS information */
471 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
472 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
473 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
474 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
475 	devRead->misc.driverInfo.uptVerSpt     = 1;
476 
477 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
478 	devRead->misc.queueDescPA  = hw->queueDescPA;
479 	devRead->misc.queueDescLen = hw->queue_desc_len;
480 	devRead->misc.numTxQueues  = hw->num_tx_queues;
481 	devRead->misc.numRxQueues  = hw->num_rx_queues;
482 
483 	/*
484 	 * Set number of interrupts to 1
485 	 * PMD disables all the interrupts but this is MUST to activate device
486 	 * It needs at least one interrupt for link events to handle
487 	 * So we'll disable it later after device activation if needed
488 	 */
489 	devRead->intrConf.numIntrs = 1;
490 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
491 
492 	for (i = 0; i < hw->num_tx_queues; i++) {
493 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
494 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
495 
496 		tqd->ctrl.txNumDeferred  = 0;
497 		tqd->ctrl.txThreshold    = 1;
498 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
499 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
500 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
501 
502 		tqd->conf.txRingSize   = txq->cmd_ring.size;
503 		tqd->conf.compRingSize = txq->comp_ring.size;
504 		tqd->conf.dataRingSize = txq->data_ring.size;
505 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
506 		tqd->status.stopped    = TRUE;
507 		tqd->status.error      = 0;
508 		memset(&tqd->stats, 0, sizeof(tqd->stats));
509 	}
510 
511 	for (i = 0; i < hw->num_rx_queues; i++) {
512 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
513 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
514 
515 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
516 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
517 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
518 
519 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
520 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
521 		rqd->conf.compRingSize    = rxq->comp_ring.size;
522 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
523 		rqd->status.stopped       = TRUE;
524 		rqd->status.error         = 0;
525 		memset(&rqd->stats, 0, sizeof(rqd->stats));
526 	}
527 
528 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
529 	devRead->rxFilterConf.rxMode = 0;
530 
531 	/* Setting up feature flags */
532 	if (dev->data->dev_conf.rxmode.hw_ip_checksum)
533 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
534 
535 	if (dev->data->dev_conf.rxmode.enable_lro) {
536 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
537 		devRead->misc.maxNumRxSG = 0;
538 	}
539 
540 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
541 		ret = vmxnet3_rss_configure(dev);
542 		if (ret != VMXNET3_SUCCESS)
543 			return ret;
544 
545 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
546 		devRead->rssConfDesc.confVer = 1;
547 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
548 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
549 	}
550 
551 	vmxnet3_dev_vlan_offload_set(dev,
552 				     ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
553 
554 	vmxnet3_write_mac(hw, hw->perm_addr);
555 
556 	return VMXNET3_SUCCESS;
557 }
558 
559 /*
560  * Configure device link speed and setup link.
561  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
562  * It returns 0 on success.
563  */
564 static int
565 vmxnet3_dev_start(struct rte_eth_dev *dev)
566 {
567 	int ret;
568 	struct vmxnet3_hw *hw = dev->data->dev_private;
569 
570 	PMD_INIT_FUNC_TRACE();
571 
572 	ret = vmxnet3_setup_driver_shared(dev);
573 	if (ret != VMXNET3_SUCCESS)
574 		return ret;
575 
576 	/* Exchange shared data with device */
577 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
578 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
579 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
580 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
581 
582 	/* Activate device by register write */
583 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
584 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
585 
586 	if (ret != 0) {
587 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
588 		return -EINVAL;
589 	}
590 
591 	/* Disable interrupts */
592 	vmxnet3_disable_intr(hw);
593 
594 	/*
595 	 * Load RX queues with blank mbufs and update next2fill index for device
596 	 * Update RxMode of the device
597 	 */
598 	ret = vmxnet3_dev_rxtx_init(dev);
599 	if (ret != VMXNET3_SUCCESS) {
600 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
601 		return ret;
602 	}
603 
604 	/* Setting proper Rx Mode and issue Rx Mode Update command */
605 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
606 
607 	/*
608 	 * Don't need to handle events for now
609 	 */
610 #if PROCESS_SYS_EVENTS == 1
611 	events = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_ECR);
612 	PMD_INIT_LOG(DEBUG, "Reading events: 0x%X", events);
613 	vmxnet3_process_events(hw);
614 #endif
615 	return VMXNET3_SUCCESS;
616 }
617 
618 /*
619  * Stop device: disable rx and tx functions to allow for reconfiguring.
620  */
621 static void
622 vmxnet3_dev_stop(struct rte_eth_dev *dev)
623 {
624 	struct rte_eth_link link;
625 	struct vmxnet3_hw *hw = dev->data->dev_private;
626 
627 	PMD_INIT_FUNC_TRACE();
628 
629 	if (hw->adapter_stopped == 1) {
630 		PMD_INIT_LOG(DEBUG, "Device already closed.");
631 		return;
632 	}
633 
634 	/* disable interrupts */
635 	vmxnet3_disable_intr(hw);
636 
637 	/* quiesce the device first */
638 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
639 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
640 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
641 
642 	/* reset the device */
643 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
644 	PMD_INIT_LOG(DEBUG, "Device reset.");
645 	hw->adapter_stopped = 0;
646 
647 	vmxnet3_dev_clear_queues(dev);
648 
649 	/* Clear recorded link status */
650 	memset(&link, 0, sizeof(link));
651 	vmxnet3_dev_atomic_write_link_status(dev, &link);
652 }
653 
654 /*
655  * Reset and stop device.
656  */
657 static void
658 vmxnet3_dev_close(struct rte_eth_dev *dev)
659 {
660 	struct vmxnet3_hw *hw = dev->data->dev_private;
661 
662 	PMD_INIT_FUNC_TRACE();
663 
664 	vmxnet3_dev_stop(dev);
665 	hw->adapter_stopped = 1;
666 }
667 
668 static void
669 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
670 {
671 	unsigned int i;
672 	struct vmxnet3_hw *hw = dev->data->dev_private;
673 
674 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
675 
676 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
677 	for (i = 0; i < hw->num_tx_queues; i++) {
678 		struct UPT1_TxStats *txStats = &hw->tqd_start[i].stats;
679 
680 		stats->q_opackets[i] = txStats->ucastPktsTxOK +
681 					txStats->mcastPktsTxOK +
682 					txStats->bcastPktsTxOK;
683 		stats->q_obytes[i] = txStats->ucastBytesTxOK +
684 					txStats->mcastBytesTxOK +
685 					txStats->bcastBytesTxOK;
686 
687 		stats->opackets += stats->q_opackets[i];
688 		stats->obytes += stats->q_obytes[i];
689 		stats->oerrors += txStats->pktsTxError + txStats->pktsTxDiscard;
690 	}
691 
692 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
693 	for (i = 0; i < hw->num_rx_queues; i++) {
694 		struct UPT1_RxStats *rxStats = &hw->rqd_start[i].stats;
695 
696 		stats->q_ipackets[i] = rxStats->ucastPktsRxOK +
697 					rxStats->mcastPktsRxOK +
698 					rxStats->bcastPktsRxOK;
699 
700 		stats->q_ibytes[i] = rxStats->ucastBytesRxOK +
701 					rxStats->mcastBytesRxOK +
702 					rxStats->bcastBytesRxOK;
703 
704 		stats->ipackets += stats->q_ipackets[i];
705 		stats->ibytes += stats->q_ibytes[i];
706 
707 		stats->q_errors[i] = rxStats->pktsRxError;
708 		stats->ierrors += rxStats->pktsRxError;
709 		stats->rx_nombuf += rxStats->pktsRxOutOfBuf;
710 	}
711 }
712 
713 static void
714 vmxnet3_dev_info_get(struct rte_eth_dev *dev,
715 		     struct rte_eth_dev_info *dev_info)
716 {
717 	dev_info->pci_dev = RTE_DEV_TO_PCI(dev->device);
718 
719 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
720 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
721 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
722 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
723 	dev_info->speed_capa = ETH_LINK_SPEED_10G;
724 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
725 
726 	dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
727 	dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
728 
729 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
730 		.nb_max = VMXNET3_RX_RING_MAX_SIZE,
731 		.nb_min = VMXNET3_DEF_RX_RING_SIZE,
732 		.nb_align = 1,
733 	};
734 
735 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
736 		.nb_max = VMXNET3_TX_RING_MAX_SIZE,
737 		.nb_min = VMXNET3_DEF_TX_RING_SIZE,
738 		.nb_align = 1,
739 		.nb_seg_max = VMXNET3_TX_MAX_SEG,
740 		.nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
741 	};
742 
743 	dev_info->rx_offload_capa =
744 		DEV_RX_OFFLOAD_VLAN_STRIP |
745 		DEV_RX_OFFLOAD_UDP_CKSUM |
746 		DEV_RX_OFFLOAD_TCP_CKSUM |
747 		DEV_RX_OFFLOAD_TCP_LRO;
748 
749 	dev_info->tx_offload_capa =
750 		DEV_TX_OFFLOAD_VLAN_INSERT |
751 		DEV_TX_OFFLOAD_TCP_CKSUM |
752 		DEV_TX_OFFLOAD_UDP_CKSUM |
753 		DEV_TX_OFFLOAD_TCP_TSO;
754 }
755 
756 static const uint32_t *
757 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
758 {
759 	static const uint32_t ptypes[] = {
760 		RTE_PTYPE_L3_IPV4_EXT,
761 		RTE_PTYPE_L3_IPV4,
762 		RTE_PTYPE_UNKNOWN
763 	};
764 
765 	if (dev->rx_pkt_burst == vmxnet3_recv_pkts)
766 		return ptypes;
767 	return NULL;
768 }
769 
770 static void
771 vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
772 {
773 	struct vmxnet3_hw *hw = dev->data->dev_private;
774 
775 	vmxnet3_write_mac(hw, mac_addr->addr_bytes);
776 }
777 
778 /* return 0 means link status changed, -1 means not changed */
779 static int
780 vmxnet3_dev_link_update(struct rte_eth_dev *dev,
781 			__rte_unused int wait_to_complete)
782 {
783 	struct vmxnet3_hw *hw = dev->data->dev_private;
784 	struct rte_eth_link old, link;
785 	uint32_t ret;
786 
787 	/* Link status doesn't change for stopped dev */
788 	if (dev->data->dev_started == 0)
789 		return -1;
790 
791 	memset(&link, 0, sizeof(link));
792 	vmxnet3_dev_atomic_read_link_status(dev, &old);
793 
794 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
795 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
796 
797 	if (ret & 0x1) {
798 		link.link_status = ETH_LINK_UP;
799 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
800 		link.link_speed = ETH_SPEED_NUM_10G;
801 		link.link_autoneg = ETH_LINK_SPEED_FIXED;
802 	}
803 
804 	vmxnet3_dev_atomic_write_link_status(dev, &link);
805 
806 	return (old.link_status == link.link_status) ? -1 : 0;
807 }
808 
809 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
810 static void
811 vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set)
812 {
813 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
814 
815 	if (set)
816 		rxConf->rxMode = rxConf->rxMode | feature;
817 	else
818 		rxConf->rxMode = rxConf->rxMode & (~feature);
819 
820 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_UPDATE_RX_MODE);
821 }
822 
823 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
824 static void
825 vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev)
826 {
827 	struct vmxnet3_hw *hw = dev->data->dev_private;
828 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
829 
830 	memset(vf_table, 0, VMXNET3_VFT_TABLE_SIZE);
831 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 1);
832 
833 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
834 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
835 }
836 
837 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
838 static void
839 vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
840 {
841 	struct vmxnet3_hw *hw = dev->data->dev_private;
842 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
843 
844 	memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
845 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
846 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
847 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
848 }
849 
850 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
851 static void
852 vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev)
853 {
854 	struct vmxnet3_hw *hw = dev->data->dev_private;
855 
856 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 1);
857 }
858 
859 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
860 static void
861 vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev)
862 {
863 	struct vmxnet3_hw *hw = dev->data->dev_private;
864 
865 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 0);
866 }
867 
868 /* Enable/disable filter on vlan */
869 static int
870 vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
871 {
872 	struct vmxnet3_hw *hw = dev->data->dev_private;
873 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
874 	uint32_t *vf_table = rxConf->vfTable;
875 
876 	/* save state for restore */
877 	if (on)
878 		VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, vid);
879 	else
880 		VMXNET3_CLEAR_VFTABLE_ENTRY(hw->shadow_vfta, vid);
881 
882 	/* don't change active filter if in promiscuous mode */
883 	if (rxConf->rxMode & VMXNET3_RXM_PROMISC)
884 		return 0;
885 
886 	/* set in hardware */
887 	if (on)
888 		VMXNET3_SET_VFTABLE_ENTRY(vf_table, vid);
889 	else
890 		VMXNET3_CLEAR_VFTABLE_ENTRY(vf_table, vid);
891 
892 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
893 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
894 	return 0;
895 }
896 
897 static void
898 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
899 {
900 	struct vmxnet3_hw *hw = dev->data->dev_private;
901 	Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
902 	uint32_t *vf_table = devRead->rxFilterConf.vfTable;
903 
904 	if (mask & ETH_VLAN_STRIP_MASK) {
905 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
906 			devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
907 		else
908 			devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
909 
910 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
911 				       VMXNET3_CMD_UPDATE_FEATURE);
912 	}
913 
914 	if (mask & ETH_VLAN_FILTER_MASK) {
915 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
916 			memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
917 		else
918 			memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
919 
920 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
921 				       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
922 	}
923 }
924 
925 #if PROCESS_SYS_EVENTS == 1
926 static void
927 vmxnet3_process_events(struct vmxnet3_hw *hw)
928 {
929 	uint32_t events = hw->shared->ecr;
930 
931 	if (!events) {
932 		PMD_INIT_LOG(ERR, "No events to process");
933 		return;
934 	}
935 
936 	/*
937 	 * ECR bits when written with 1b are cleared. Hence write
938 	 * events back to ECR so that the bits which were set will be reset.
939 	 */
940 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
941 
942 	/* Check if link state has changed */
943 	if (events & VMXNET3_ECR_LINK)
944 		PMD_INIT_LOG(ERR,
945 			     "Process events in %s(): VMXNET3_ECR_LINK event",
946 			     __func__);
947 
948 	/* Check if there is an error on xmit/recv queues */
949 	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
950 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
951 				       VMXNET3_CMD_GET_QUEUE_STATUS);
952 
953 		if (hw->tqd_start->status.stopped)
954 			PMD_INIT_LOG(ERR, "tq error 0x%x",
955 				     hw->tqd_start->status.error);
956 
957 		if (hw->rqd_start->status.stopped)
958 			PMD_INIT_LOG(ERR, "rq error 0x%x",
959 				     hw->rqd_start->status.error);
960 
961 		/* Reset the device */
962 		/* Have to reset the device */
963 	}
964 
965 	if (events & VMXNET3_ECR_DIC)
966 		PMD_INIT_LOG(ERR, "Device implementation change event.");
967 
968 	if (events & VMXNET3_ECR_DEBUG)
969 		PMD_INIT_LOG(ERR, "Debug event generated by device.");
970 }
971 #endif
972 
973 RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd.pci_drv);
974 RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
975 RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio");
976