xref: /dpdk/drivers/net/vmxnet3/vmxnet3_ethdev.c (revision c6dab2a873f65c5a4ea9735aa24d9539426adba4)
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 
254 	/* Vendor and Device ID need to be set before init of shared code */
255 	hw->device_id = pci_dev->id.device_id;
256 	hw->vendor_id = pci_dev->id.vendor_id;
257 	hw->hw_addr0 = (void *)pci_dev->mem_resource[0].addr;
258 	hw->hw_addr1 = (void *)pci_dev->mem_resource[1].addr;
259 
260 	hw->num_rx_queues = 1;
261 	hw->num_tx_queues = 1;
262 	hw->bufs_per_pkt = 1;
263 
264 	/* Check h/w version compatibility with driver. */
265 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
266 	PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
267 	if (ver & 0x1)
268 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS, 1);
269 	else {
270 		PMD_INIT_LOG(ERR, "Incompatible h/w version, should be 0x1");
271 		return -EIO;
272 	}
273 
274 	/* Check UPT version compatibility with driver. */
275 	ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
276 	PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
277 	if (ver & 0x1)
278 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_UVRS, 1);
279 	else {
280 		PMD_INIT_LOG(ERR, "Incompatible UPT version.");
281 		return -EIO;
282 	}
283 
284 	/* Getting MAC Address */
285 	mac_lo = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACL);
286 	mac_hi = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_MACH);
287 	memcpy(hw->perm_addr, &mac_lo, 4);
288 	memcpy(hw->perm_addr + 4, &mac_hi, 2);
289 
290 	/* Allocate memory for storing MAC addresses */
291 	eth_dev->data->mac_addrs = rte_zmalloc("vmxnet3", ETHER_ADDR_LEN *
292 					       VMXNET3_MAX_MAC_ADDRS, 0);
293 	if (eth_dev->data->mac_addrs == NULL) {
294 		PMD_INIT_LOG(ERR,
295 			     "Failed to allocate %d bytes needed to store MAC addresses",
296 			     ETHER_ADDR_LEN * VMXNET3_MAX_MAC_ADDRS);
297 		return -ENOMEM;
298 	}
299 	/* Copy the permanent MAC address */
300 	ether_addr_copy((struct ether_addr *) hw->perm_addr,
301 			&eth_dev->data->mac_addrs[0]);
302 
303 	PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
304 		     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
305 		     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
306 
307 	/* Put device in Quiesce Mode */
308 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
309 
310 	/* allow untagged pkts */
311 	VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
312 
313 	return 0;
314 }
315 
316 static int
317 eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
318 {
319 	struct vmxnet3_hw *hw = eth_dev->data->dev_private;
320 
321 	PMD_INIT_FUNC_TRACE();
322 
323 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
324 		return 0;
325 
326 	if (hw->adapter_stopped == 0)
327 		vmxnet3_dev_close(eth_dev);
328 
329 	eth_dev->dev_ops = NULL;
330 	eth_dev->rx_pkt_burst = NULL;
331 	eth_dev->tx_pkt_burst = NULL;
332 	eth_dev->tx_pkt_prepare = NULL;
333 
334 	rte_free(eth_dev->data->mac_addrs);
335 	eth_dev->data->mac_addrs = NULL;
336 
337 	return 0;
338 }
339 
340 static struct eth_driver rte_vmxnet3_pmd = {
341 	.pci_drv = {
342 		.id_table = pci_id_vmxnet3_map,
343 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
344 		.probe = rte_eth_dev_pci_probe,
345 		.remove = rte_eth_dev_pci_remove,
346 	},
347 	.eth_dev_init = eth_vmxnet3_dev_init,
348 	.eth_dev_uninit = eth_vmxnet3_dev_uninit,
349 	.dev_private_size = sizeof(struct vmxnet3_hw),
350 };
351 
352 static int
353 vmxnet3_dev_configure(struct rte_eth_dev *dev)
354 {
355 	const struct rte_memzone *mz;
356 	struct vmxnet3_hw *hw = dev->data->dev_private;
357 	size_t size;
358 
359 	PMD_INIT_FUNC_TRACE();
360 
361 	if (dev->data->nb_tx_queues > VMXNET3_MAX_TX_QUEUES ||
362 	    dev->data->nb_rx_queues > VMXNET3_MAX_RX_QUEUES) {
363 		PMD_INIT_LOG(ERR, "ERROR: Number of queues not supported");
364 		return -EINVAL;
365 	}
366 
367 	if (!rte_is_power_of_2(dev->data->nb_rx_queues)) {
368 		PMD_INIT_LOG(ERR, "ERROR: Number of rx queues not power of 2");
369 		return -EINVAL;
370 	}
371 
372 	size = dev->data->nb_rx_queues * sizeof(struct Vmxnet3_TxQueueDesc) +
373 		dev->data->nb_tx_queues * sizeof(struct Vmxnet3_RxQueueDesc);
374 
375 	if (size > UINT16_MAX)
376 		return -EINVAL;
377 
378 	hw->num_rx_queues = (uint8_t)dev->data->nb_rx_queues;
379 	hw->num_tx_queues = (uint8_t)dev->data->nb_tx_queues;
380 
381 	/*
382 	 * Allocate a memzone for Vmxnet3_DriverShared - Vmxnet3_DSDevRead
383 	 * on current socket
384 	 */
385 	mz = gpa_zone_reserve(dev, sizeof(struct Vmxnet3_DriverShared),
386 			      "shared", rte_socket_id(), 8, 1);
387 
388 	if (mz == NULL) {
389 		PMD_INIT_LOG(ERR, "ERROR: Creating shared zone");
390 		return -ENOMEM;
391 	}
392 	memset(mz->addr, 0, mz->len);
393 
394 	hw->shared = mz->addr;
395 	hw->sharedPA = mz->phys_addr;
396 
397 	/*
398 	 * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
399 	 * on current socket.
400 	 *
401 	 * We cannot reuse this memzone from previous allocation as its size
402 	 * depends on the number of tx and rx queues, which could be different
403 	 * from one config to another.
404 	 */
405 	mz = gpa_zone_reserve(dev, size, "queuedesc", rte_socket_id(),
406 			      VMXNET3_QUEUE_DESC_ALIGN, 0);
407 	if (mz == NULL) {
408 		PMD_INIT_LOG(ERR, "ERROR: Creating queue descriptors zone");
409 		return -ENOMEM;
410 	}
411 	memset(mz->addr, 0, mz->len);
412 
413 	hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
414 	hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
415 
416 	hw->queueDescPA = mz->phys_addr;
417 	hw->queue_desc_len = (uint16_t)size;
418 
419 	if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
420 		/* Allocate memory structure for UPT1_RSSConf and configure */
421 		mz = gpa_zone_reserve(dev, sizeof(struct VMXNET3_RSSConf),
422 				      "rss_conf", rte_socket_id(),
423 				      RTE_CACHE_LINE_SIZE, 1);
424 		if (mz == NULL) {
425 			PMD_INIT_LOG(ERR,
426 				     "ERROR: Creating rss_conf structure zone");
427 			return -ENOMEM;
428 		}
429 		memset(mz->addr, 0, mz->len);
430 
431 		hw->rss_conf = mz->addr;
432 		hw->rss_confPA = mz->phys_addr;
433 	}
434 
435 	return 0;
436 }
437 
438 static void
439 vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
440 {
441 	uint32_t val;
442 
443 	PMD_INIT_LOG(DEBUG,
444 		     "Writing MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
445 		     addr[0], addr[1], addr[2],
446 		     addr[3], addr[4], addr[5]);
447 
448 	val = *(const uint32_t *)addr;
449 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
450 
451 	val = (addr[5] << 8) | addr[4];
452 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
453 }
454 
455 static int
456 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
457 {
458 	struct rte_eth_conf port_conf = dev->data->dev_conf;
459 	struct vmxnet3_hw *hw = dev->data->dev_private;
460 	uint32_t mtu = dev->data->mtu;
461 	Vmxnet3_DriverShared *shared = hw->shared;
462 	Vmxnet3_DSDevRead *devRead = &shared->devRead;
463 	uint32_t i;
464 	int ret;
465 
466 	shared->magic = VMXNET3_REV1_MAGIC;
467 	devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
468 
469 	/* Setting up Guest OS information */
470 	devRead->misc.driverInfo.gos.gosBits   = sizeof(void *) == 4 ?
471 		VMXNET3_GOS_BITS_32 : VMXNET3_GOS_BITS_64;
472 	devRead->misc.driverInfo.gos.gosType   = VMXNET3_GOS_TYPE_LINUX;
473 	devRead->misc.driverInfo.vmxnet3RevSpt = 1;
474 	devRead->misc.driverInfo.uptVerSpt     = 1;
475 
476 	devRead->misc.mtu = rte_le_to_cpu_32(mtu);
477 	devRead->misc.queueDescPA  = hw->queueDescPA;
478 	devRead->misc.queueDescLen = hw->queue_desc_len;
479 	devRead->misc.numTxQueues  = hw->num_tx_queues;
480 	devRead->misc.numRxQueues  = hw->num_rx_queues;
481 
482 	/*
483 	 * Set number of interrupts to 1
484 	 * PMD disables all the interrupts but this is MUST to activate device
485 	 * It needs at least one interrupt for link events to handle
486 	 * So we'll disable it later after device activation if needed
487 	 */
488 	devRead->intrConf.numIntrs = 1;
489 	devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
490 
491 	for (i = 0; i < hw->num_tx_queues; i++) {
492 		Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
493 		vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
494 
495 		tqd->ctrl.txNumDeferred  = 0;
496 		tqd->ctrl.txThreshold    = 1;
497 		tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
498 		tqd->conf.compRingBasePA = txq->comp_ring.basePA;
499 		tqd->conf.dataRingBasePA = txq->data_ring.basePA;
500 
501 		tqd->conf.txRingSize   = txq->cmd_ring.size;
502 		tqd->conf.compRingSize = txq->comp_ring.size;
503 		tqd->conf.dataRingSize = txq->data_ring.size;
504 		tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
505 		tqd->status.stopped    = TRUE;
506 		tqd->status.error      = 0;
507 		memset(&tqd->stats, 0, sizeof(tqd->stats));
508 	}
509 
510 	for (i = 0; i < hw->num_rx_queues; i++) {
511 		Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
512 		vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
513 
514 		rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
515 		rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
516 		rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
517 
518 		rqd->conf.rxRingSize[0]   = rxq->cmd_ring[0].size;
519 		rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
520 		rqd->conf.compRingSize    = rxq->comp_ring.size;
521 		rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
522 		rqd->status.stopped       = TRUE;
523 		rqd->status.error         = 0;
524 		memset(&rqd->stats, 0, sizeof(rqd->stats));
525 	}
526 
527 	/* RxMode set to 0 of VMXNET3_RXM_xxx */
528 	devRead->rxFilterConf.rxMode = 0;
529 
530 	/* Setting up feature flags */
531 	if (dev->data->dev_conf.rxmode.hw_ip_checksum)
532 		devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
533 
534 	if (dev->data->dev_conf.rxmode.enable_lro) {
535 		devRead->misc.uptFeatures |= VMXNET3_F_LRO;
536 		devRead->misc.maxNumRxSG = 0;
537 	}
538 
539 	if (port_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
540 		ret = vmxnet3_rss_configure(dev);
541 		if (ret != VMXNET3_SUCCESS)
542 			return ret;
543 
544 		devRead->misc.uptFeatures |= VMXNET3_F_RSS;
545 		devRead->rssConfDesc.confVer = 1;
546 		devRead->rssConfDesc.confLen = sizeof(struct VMXNET3_RSSConf);
547 		devRead->rssConfDesc.confPA  = hw->rss_confPA;
548 	}
549 
550 	vmxnet3_dev_vlan_offload_set(dev,
551 				     ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
552 
553 	vmxnet3_write_mac(hw, hw->perm_addr);
554 
555 	return VMXNET3_SUCCESS;
556 }
557 
558 /*
559  * Configure device link speed and setup link.
560  * Must be called after eth_vmxnet3_dev_init. Other wise it might fail
561  * It returns 0 on success.
562  */
563 static int
564 vmxnet3_dev_start(struct rte_eth_dev *dev)
565 {
566 	int ret;
567 	struct vmxnet3_hw *hw = dev->data->dev_private;
568 
569 	PMD_INIT_FUNC_TRACE();
570 
571 	ret = vmxnet3_setup_driver_shared(dev);
572 	if (ret != VMXNET3_SUCCESS)
573 		return ret;
574 
575 	/* Exchange shared data with device */
576 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
577 			       VMXNET3_GET_ADDR_LO(hw->sharedPA));
578 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH,
579 			       VMXNET3_GET_ADDR_HI(hw->sharedPA));
580 
581 	/* Activate device by register write */
582 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_ACTIVATE_DEV);
583 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
584 
585 	if (ret != 0) {
586 		PMD_INIT_LOG(ERR, "Device activation: UNSUCCESSFUL");
587 		return -EINVAL;
588 	}
589 
590 	/* Disable interrupts */
591 	vmxnet3_disable_intr(hw);
592 
593 	/*
594 	 * Load RX queues with blank mbufs and update next2fill index for device
595 	 * Update RxMode of the device
596 	 */
597 	ret = vmxnet3_dev_rxtx_init(dev);
598 	if (ret != VMXNET3_SUCCESS) {
599 		PMD_INIT_LOG(ERR, "Device queue init: UNSUCCESSFUL");
600 		return ret;
601 	}
602 
603 	/* Setting proper Rx Mode and issue Rx Mode Update command */
604 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
605 
606 	/*
607 	 * Don't need to handle events for now
608 	 */
609 #if PROCESS_SYS_EVENTS == 1
610 	events = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_ECR);
611 	PMD_INIT_LOG(DEBUG, "Reading events: 0x%X", events);
612 	vmxnet3_process_events(hw);
613 #endif
614 	return VMXNET3_SUCCESS;
615 }
616 
617 /*
618  * Stop device: disable rx and tx functions to allow for reconfiguring.
619  */
620 static void
621 vmxnet3_dev_stop(struct rte_eth_dev *dev)
622 {
623 	struct rte_eth_link link;
624 	struct vmxnet3_hw *hw = dev->data->dev_private;
625 
626 	PMD_INIT_FUNC_TRACE();
627 
628 	if (hw->adapter_stopped == 1) {
629 		PMD_INIT_LOG(DEBUG, "Device already closed.");
630 		return;
631 	}
632 
633 	/* disable interrupts */
634 	vmxnet3_disable_intr(hw);
635 
636 	/* quiesce the device first */
637 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
638 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
639 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAH, 0);
640 
641 	/* reset the device */
642 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
643 	PMD_INIT_LOG(DEBUG, "Device reset.");
644 	hw->adapter_stopped = 0;
645 
646 	vmxnet3_dev_clear_queues(dev);
647 
648 	/* Clear recorded link status */
649 	memset(&link, 0, sizeof(link));
650 	vmxnet3_dev_atomic_write_link_status(dev, &link);
651 }
652 
653 /*
654  * Reset and stop device.
655  */
656 static void
657 vmxnet3_dev_close(struct rte_eth_dev *dev)
658 {
659 	struct vmxnet3_hw *hw = dev->data->dev_private;
660 
661 	PMD_INIT_FUNC_TRACE();
662 
663 	vmxnet3_dev_stop(dev);
664 	hw->adapter_stopped = 1;
665 }
666 
667 static void
668 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
669 {
670 	unsigned int i;
671 	struct vmxnet3_hw *hw = dev->data->dev_private;
672 
673 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
674 
675 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
676 	for (i = 0; i < hw->num_tx_queues; i++) {
677 		struct UPT1_TxStats *txStats = &hw->tqd_start[i].stats;
678 
679 		stats->q_opackets[i] = txStats->ucastPktsTxOK +
680 					txStats->mcastPktsTxOK +
681 					txStats->bcastPktsTxOK;
682 		stats->q_obytes[i] = txStats->ucastBytesTxOK +
683 					txStats->mcastBytesTxOK +
684 					txStats->bcastBytesTxOK;
685 
686 		stats->opackets += stats->q_opackets[i];
687 		stats->obytes += stats->q_obytes[i];
688 		stats->oerrors += txStats->pktsTxError + txStats->pktsTxDiscard;
689 	}
690 
691 	RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
692 	for (i = 0; i < hw->num_rx_queues; i++) {
693 		struct UPT1_RxStats *rxStats = &hw->rqd_start[i].stats;
694 
695 		stats->q_ipackets[i] = rxStats->ucastPktsRxOK +
696 					rxStats->mcastPktsRxOK +
697 					rxStats->bcastPktsRxOK;
698 
699 		stats->q_ibytes[i] = rxStats->ucastBytesRxOK +
700 					rxStats->mcastBytesRxOK +
701 					rxStats->bcastBytesRxOK;
702 
703 		stats->ipackets += stats->q_ipackets[i];
704 		stats->ibytes += stats->q_ibytes[i];
705 
706 		stats->q_errors[i] = rxStats->pktsRxError;
707 		stats->ierrors += rxStats->pktsRxError;
708 		stats->rx_nombuf += rxStats->pktsRxOutOfBuf;
709 	}
710 }
711 
712 static void
713 vmxnet3_dev_info_get(struct rte_eth_dev *dev,
714 		     struct rte_eth_dev_info *dev_info)
715 {
716 	dev_info->pci_dev = RTE_DEV_TO_PCI(dev->device);
717 
718 	dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
719 	dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
720 	dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
721 	dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
722 	dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
723 
724 	dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
725 	dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
726 
727 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
728 		.nb_max = VMXNET3_RX_RING_MAX_SIZE,
729 		.nb_min = VMXNET3_DEF_RX_RING_SIZE,
730 		.nb_align = 1,
731 	};
732 
733 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
734 		.nb_max = VMXNET3_TX_RING_MAX_SIZE,
735 		.nb_min = VMXNET3_DEF_TX_RING_SIZE,
736 		.nb_align = 1,
737 		.nb_seg_max = VMXNET3_TX_MAX_SEG,
738 		.nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
739 	};
740 
741 	dev_info->rx_offload_capa =
742 		DEV_RX_OFFLOAD_VLAN_STRIP |
743 		DEV_RX_OFFLOAD_UDP_CKSUM |
744 		DEV_RX_OFFLOAD_TCP_CKSUM |
745 		DEV_RX_OFFLOAD_TCP_LRO;
746 
747 	dev_info->tx_offload_capa =
748 		DEV_TX_OFFLOAD_VLAN_INSERT |
749 		DEV_TX_OFFLOAD_TCP_CKSUM |
750 		DEV_TX_OFFLOAD_UDP_CKSUM |
751 		DEV_TX_OFFLOAD_TCP_TSO;
752 }
753 
754 static const uint32_t *
755 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
756 {
757 	static const uint32_t ptypes[] = {
758 		RTE_PTYPE_L3_IPV4_EXT,
759 		RTE_PTYPE_L3_IPV4,
760 		RTE_PTYPE_UNKNOWN
761 	};
762 
763 	if (dev->rx_pkt_burst == vmxnet3_recv_pkts)
764 		return ptypes;
765 	return NULL;
766 }
767 
768 static void
769 vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
770 {
771 	struct vmxnet3_hw *hw = dev->data->dev_private;
772 
773 	vmxnet3_write_mac(hw, mac_addr->addr_bytes);
774 }
775 
776 /* return 0 means link status changed, -1 means not changed */
777 static int
778 vmxnet3_dev_link_update(struct rte_eth_dev *dev,
779 			__rte_unused int wait_to_complete)
780 {
781 	struct vmxnet3_hw *hw = dev->data->dev_private;
782 	struct rte_eth_link old, link;
783 	uint32_t ret;
784 
785 	/* Link status doesn't change for stopped dev */
786 	if (dev->data->dev_started == 0)
787 		return -1;
788 
789 	memset(&link, 0, sizeof(link));
790 	vmxnet3_dev_atomic_read_link_status(dev, &old);
791 
792 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
793 	ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
794 
795 	if (ret & 0x1) {
796 		link.link_status = ETH_LINK_UP;
797 		link.link_duplex = ETH_LINK_FULL_DUPLEX;
798 		link.link_speed = ETH_SPEED_NUM_10G;
799 		link.link_autoneg = ETH_LINK_SPEED_FIXED;
800 	}
801 
802 	vmxnet3_dev_atomic_write_link_status(dev, &link);
803 
804 	return (old.link_status == link.link_status) ? -1 : 0;
805 }
806 
807 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
808 static void
809 vmxnet3_dev_set_rxmode(struct vmxnet3_hw *hw, uint32_t feature, int set)
810 {
811 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
812 
813 	if (set)
814 		rxConf->rxMode = rxConf->rxMode | feature;
815 	else
816 		rxConf->rxMode = rxConf->rxMode & (~feature);
817 
818 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_UPDATE_RX_MODE);
819 }
820 
821 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
822 static void
823 vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev)
824 {
825 	struct vmxnet3_hw *hw = dev->data->dev_private;
826 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
827 
828 	memset(vf_table, 0, VMXNET3_VFT_TABLE_SIZE);
829 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 1);
830 
831 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
832 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
833 }
834 
835 /* Promiscuous supported only if Vmxnet3_DriverShared is initialized in adapter */
836 static void
837 vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
838 {
839 	struct vmxnet3_hw *hw = dev->data->dev_private;
840 	uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
841 
842 	memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
843 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
844 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
845 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
846 }
847 
848 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
849 static void
850 vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev)
851 {
852 	struct vmxnet3_hw *hw = dev->data->dev_private;
853 
854 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 1);
855 }
856 
857 /* Allmulticast supported only if Vmxnet3_DriverShared is initialized in adapter */
858 static void
859 vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev)
860 {
861 	struct vmxnet3_hw *hw = dev->data->dev_private;
862 
863 	vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_ALL_MULTI, 0);
864 }
865 
866 /* Enable/disable filter on vlan */
867 static int
868 vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
869 {
870 	struct vmxnet3_hw *hw = dev->data->dev_private;
871 	struct Vmxnet3_RxFilterConf *rxConf = &hw->shared->devRead.rxFilterConf;
872 	uint32_t *vf_table = rxConf->vfTable;
873 
874 	/* save state for restore */
875 	if (on)
876 		VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, vid);
877 	else
878 		VMXNET3_CLEAR_VFTABLE_ENTRY(hw->shadow_vfta, vid);
879 
880 	/* don't change active filter if in promiscuous mode */
881 	if (rxConf->rxMode & VMXNET3_RXM_PROMISC)
882 		return 0;
883 
884 	/* set in hardware */
885 	if (on)
886 		VMXNET3_SET_VFTABLE_ENTRY(vf_table, vid);
887 	else
888 		VMXNET3_CLEAR_VFTABLE_ENTRY(vf_table, vid);
889 
890 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
891 			       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
892 	return 0;
893 }
894 
895 static void
896 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
897 {
898 	struct vmxnet3_hw *hw = dev->data->dev_private;
899 	Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
900 	uint32_t *vf_table = devRead->rxFilterConf.vfTable;
901 
902 	if (mask & ETH_VLAN_STRIP_MASK) {
903 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
904 			devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
905 		else
906 			devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
907 
908 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
909 				       VMXNET3_CMD_UPDATE_FEATURE);
910 	}
911 
912 	if (mask & ETH_VLAN_FILTER_MASK) {
913 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
914 			memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
915 		else
916 			memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
917 
918 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
919 				       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
920 	}
921 }
922 
923 #if PROCESS_SYS_EVENTS == 1
924 static void
925 vmxnet3_process_events(struct vmxnet3_hw *hw)
926 {
927 	uint32_t events = hw->shared->ecr;
928 
929 	if (!events) {
930 		PMD_INIT_LOG(ERR, "No events to process");
931 		return;
932 	}
933 
934 	/*
935 	 * ECR bits when written with 1b are cleared. Hence write
936 	 * events back to ECR so that the bits which were set will be reset.
937 	 */
938 	VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
939 
940 	/* Check if link state has changed */
941 	if (events & VMXNET3_ECR_LINK)
942 		PMD_INIT_LOG(ERR,
943 			     "Process events in %s(): VMXNET3_ECR_LINK event",
944 			     __func__);
945 
946 	/* Check if there is an error on xmit/recv queues */
947 	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
948 		VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
949 				       VMXNET3_CMD_GET_QUEUE_STATUS);
950 
951 		if (hw->tqd_start->status.stopped)
952 			PMD_INIT_LOG(ERR, "tq error 0x%x",
953 				     hw->tqd_start->status.error);
954 
955 		if (hw->rqd_start->status.stopped)
956 			PMD_INIT_LOG(ERR, "rq error 0x%x",
957 				     hw->rqd_start->status.error);
958 
959 		/* Reset the device */
960 		/* Have to reset the device */
961 	}
962 
963 	if (events & VMXNET3_ECR_DIC)
964 		PMD_INIT_LOG(ERR, "Device implementation change event.");
965 
966 	if (events & VMXNET3_ECR_DEBUG)
967 		PMD_INIT_LOG(ERR, "Debug event generated by device.");
968 }
969 #endif
970 
971 RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd.pci_drv);
972 RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
973 RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio");
974