xref: /f-stack/dpdk/drivers/net/e1000/em_ethdev.c (revision a9643ea8)
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 <stdarg.h>
39 
40 #include <rte_common.h>
41 #include <rte_interrupts.h>
42 #include <rte_byteorder.h>
43 #include <rte_log.h>
44 #include <rte_debug.h>
45 #include <rte_pci.h>
46 #include <rte_ether.h>
47 #include <rte_ethdev.h>
48 #include <rte_memory.h>
49 #include <rte_memzone.h>
50 #include <rte_eal.h>
51 #include <rte_atomic.h>
52 #include <rte_malloc.h>
53 #include <rte_dev.h>
54 
55 #include "e1000_logs.h"
56 #include "base/e1000_api.h"
57 #include "e1000_ethdev.h"
58 
59 #define EM_EIAC			0x000DC
60 
61 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
62 
63 
64 static int eth_em_configure(struct rte_eth_dev *dev);
65 static int eth_em_start(struct rte_eth_dev *dev);
66 static void eth_em_stop(struct rte_eth_dev *dev);
67 static void eth_em_close(struct rte_eth_dev *dev);
68 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
69 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
70 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
71 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
72 static int eth_em_link_update(struct rte_eth_dev *dev,
73 				int wait_to_complete);
74 static void eth_em_stats_get(struct rte_eth_dev *dev,
75 				struct rte_eth_stats *rte_stats);
76 static void eth_em_stats_reset(struct rte_eth_dev *dev);
77 static void eth_em_infos_get(struct rte_eth_dev *dev,
78 				struct rte_eth_dev_info *dev_info);
79 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
80 				struct rte_eth_fc_conf *fc_conf);
81 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
82 				struct rte_eth_fc_conf *fc_conf);
83 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
84 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
85 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
86 static int eth_em_interrupt_action(struct rte_eth_dev *dev);
87 static void eth_em_interrupt_handler(struct rte_intr_handle *handle,
88 							void *param);
89 
90 static int em_hw_init(struct e1000_hw *hw);
91 static int em_hardware_init(struct e1000_hw *hw);
92 static void em_hw_control_acquire(struct e1000_hw *hw);
93 static void em_hw_control_release(struct e1000_hw *hw);
94 static void em_init_manageability(struct e1000_hw *hw);
95 static void em_release_manageability(struct e1000_hw *hw);
96 
97 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
98 
99 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
100 		uint16_t vlan_id, int on);
101 static void eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
102 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
103 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
104 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
105 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
106 
107 /*
108 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
109 					uint16_t vlan_id, int on);
110 */
111 
112 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
113 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
114 static void em_lsc_intr_disable(struct e1000_hw *hw);
115 static void em_rxq_intr_enable(struct e1000_hw *hw);
116 static void em_rxq_intr_disable(struct e1000_hw *hw);
117 
118 static int eth_em_led_on(struct rte_eth_dev *dev);
119 static int eth_em_led_off(struct rte_eth_dev *dev);
120 
121 static int em_get_rx_buffer_size(struct e1000_hw *hw);
122 static void eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
123 		uint32_t index, uint32_t pool);
124 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
125 
126 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
127 				   struct ether_addr *mc_addr_set,
128 				   uint32_t nb_mc_addr);
129 
130 #define EM_FC_PAUSE_TIME 0x0680
131 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
132 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
133 
134 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
135 
136 /*
137  * The set of PCI devices this driver supports
138  */
139 static const struct rte_pci_id pci_id_em_map[] = {
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
158 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
159 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
160 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
161 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
162 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
163 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
164 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
165 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
166 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
167 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
168 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
169 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
170 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
171 	{ .vendor_id = 0, /* sentinel */ },
172 };
173 
174 static const struct eth_dev_ops eth_em_ops = {
175 	.dev_configure        = eth_em_configure,
176 	.dev_start            = eth_em_start,
177 	.dev_stop             = eth_em_stop,
178 	.dev_close            = eth_em_close,
179 	.promiscuous_enable   = eth_em_promiscuous_enable,
180 	.promiscuous_disable  = eth_em_promiscuous_disable,
181 	.allmulticast_enable  = eth_em_allmulticast_enable,
182 	.allmulticast_disable = eth_em_allmulticast_disable,
183 	.link_update          = eth_em_link_update,
184 	.stats_get            = eth_em_stats_get,
185 	.stats_reset          = eth_em_stats_reset,
186 	.dev_infos_get        = eth_em_infos_get,
187 	.mtu_set              = eth_em_mtu_set,
188 	.vlan_filter_set      = eth_em_vlan_filter_set,
189 	.vlan_offload_set     = eth_em_vlan_offload_set,
190 	.rx_queue_setup       = eth_em_rx_queue_setup,
191 	.rx_queue_release     = eth_em_rx_queue_release,
192 	.rx_queue_count       = eth_em_rx_queue_count,
193 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
194 	.tx_queue_setup       = eth_em_tx_queue_setup,
195 	.tx_queue_release     = eth_em_tx_queue_release,
196 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
197 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
198 	.dev_led_on           = eth_em_led_on,
199 	.dev_led_off          = eth_em_led_off,
200 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
201 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
202 	.mac_addr_add         = eth_em_rar_set,
203 	.mac_addr_remove      = eth_em_rar_clear,
204 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
205 	.rxq_info_get         = em_rxq_info_get,
206 	.txq_info_get         = em_txq_info_get,
207 };
208 
209 /**
210  * Atomically reads the link status information from global
211  * structure rte_eth_dev.
212  *
213  * @param dev
214  *   - Pointer to the structure rte_eth_dev to read from.
215  *   - Pointer to the buffer to be saved with the link status.
216  *
217  * @return
218  *   - On success, zero.
219  *   - On failure, negative value.
220  */
221 static inline int
222 rte_em_dev_atomic_read_link_status(struct rte_eth_dev *dev,
223 				struct rte_eth_link *link)
224 {
225 	struct rte_eth_link *dst = link;
226 	struct rte_eth_link *src = &(dev->data->dev_link);
227 
228 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
229 					*(uint64_t *)src) == 0)
230 		return -1;
231 
232 	return 0;
233 }
234 
235 /**
236  * Atomically writes the link status information into global
237  * structure rte_eth_dev.
238  *
239  * @param dev
240  *   - Pointer to the structure rte_eth_dev to read from.
241  *   - Pointer to the buffer to be saved with the link status.
242  *
243  * @return
244  *   - On success, zero.
245  *   - On failure, negative value.
246  */
247 static inline int
248 rte_em_dev_atomic_write_link_status(struct rte_eth_dev *dev,
249 				struct rte_eth_link *link)
250 {
251 	struct rte_eth_link *dst = &(dev->data->dev_link);
252 	struct rte_eth_link *src = link;
253 
254 	if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
255 					*(uint64_t *)src) == 0)
256 		return -1;
257 
258 	return 0;
259 }
260 
261 /**
262  *  eth_em_dev_is_ich8 - Check for ICH8 device
263  *  @hw: pointer to the HW structure
264  *
265  *  return TRUE for ICH8, otherwise FALSE
266  **/
267 static bool
268 eth_em_dev_is_ich8(struct e1000_hw *hw)
269 {
270 	DEBUGFUNC("eth_em_dev_is_ich8");
271 
272 	switch (hw->device_id) {
273 	case E1000_DEV_ID_PCH_LPT_I217_LM:
274 	case E1000_DEV_ID_PCH_LPT_I217_V:
275 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
276 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
277 	case E1000_DEV_ID_PCH_I218_V2:
278 	case E1000_DEV_ID_PCH_I218_LM2:
279 	case E1000_DEV_ID_PCH_I218_V3:
280 	case E1000_DEV_ID_PCH_I218_LM3:
281 		return 1;
282 	default:
283 		return 0;
284 	}
285 }
286 
287 static int
288 eth_em_dev_init(struct rte_eth_dev *eth_dev)
289 {
290 	struct rte_pci_device *pci_dev;
291 	struct e1000_adapter *adapter =
292 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
293 	struct e1000_hw *hw =
294 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
295 	struct e1000_vfta * shadow_vfta =
296 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
297 
298 	pci_dev = eth_dev->pci_dev;
299 
300 	eth_dev->dev_ops = &eth_em_ops;
301 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
302 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
303 
304 	/* for secondary processes, we don't initialise any further as primary
305 	 * has already done this work. Only check we don't need a different
306 	 * RX function */
307 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
308 		if (eth_dev->data->scattered_rx)
309 			eth_dev->rx_pkt_burst =
310 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
311 		return 0;
312 	}
313 
314 	rte_eth_copy_pci_info(eth_dev, pci_dev);
315 
316 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
317 	hw->device_id = pci_dev->id.device_id;
318 	adapter->stopped = 0;
319 
320 	/* For ICH8 support we'll need to map the flash memory BAR */
321 	if (eth_em_dev_is_ich8(hw))
322 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
323 
324 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
325 			em_hw_init(hw) != 0) {
326 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
327 			"failed to init HW",
328 			eth_dev->data->port_id, pci_dev->id.vendor_id,
329 			pci_dev->id.device_id);
330 		return -ENODEV;
331 	}
332 
333 	/* Allocate memory for storing MAC addresses */
334 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
335 			hw->mac.rar_entry_count, 0);
336 	if (eth_dev->data->mac_addrs == NULL) {
337 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
338 			"store MAC addresses",
339 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
340 		return -ENOMEM;
341 	}
342 
343 	/* Copy the permanent MAC address */
344 	ether_addr_copy((struct ether_addr *) hw->mac.addr,
345 		eth_dev->data->mac_addrs);
346 
347 	/* initialize the vfta */
348 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
349 
350 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
351 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
352 		     pci_dev->id.device_id);
353 
354 	rte_intr_callback_register(&(pci_dev->intr_handle),
355 		eth_em_interrupt_handler, (void *)eth_dev);
356 
357 	return 0;
358 }
359 
360 static int
361 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
362 {
363 	struct rte_pci_device *pci_dev;
364 	struct e1000_adapter *adapter =
365 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
366 
367 	PMD_INIT_FUNC_TRACE();
368 
369 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
370 		return -EPERM;
371 
372 	pci_dev = eth_dev->pci_dev;
373 
374 	if (adapter->stopped == 0)
375 		eth_em_close(eth_dev);
376 
377 	eth_dev->dev_ops = NULL;
378 	eth_dev->rx_pkt_burst = NULL;
379 	eth_dev->tx_pkt_burst = NULL;
380 
381 	rte_free(eth_dev->data->mac_addrs);
382 	eth_dev->data->mac_addrs = NULL;
383 
384 	/* disable uio intr before callback unregister */
385 	rte_intr_disable(&(pci_dev->intr_handle));
386 	rte_intr_callback_unregister(&(pci_dev->intr_handle),
387 		eth_em_interrupt_handler, (void *)eth_dev);
388 
389 	return 0;
390 }
391 
392 static struct eth_driver rte_em_pmd = {
393 	.pci_drv = {
394 		.name = "rte_em_pmd",
395 		.id_table = pci_id_em_map,
396 		.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
397 			RTE_PCI_DRV_DETACHABLE,
398 	},
399 	.eth_dev_init = eth_em_dev_init,
400 	.eth_dev_uninit = eth_em_dev_uninit,
401 	.dev_private_size = sizeof(struct e1000_adapter),
402 };
403 
404 static int
405 rte_em_pmd_init(const char *name __rte_unused, const char *params __rte_unused)
406 {
407 	rte_eth_driver_register(&rte_em_pmd);
408 	return 0;
409 }
410 
411 static int
412 em_hw_init(struct e1000_hw *hw)
413 {
414 	int diag;
415 
416 	diag = hw->mac.ops.init_params(hw);
417 	if (diag != 0) {
418 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
419 		return diag;
420 	}
421 	diag = hw->nvm.ops.init_params(hw);
422 	if (diag != 0) {
423 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
424 		return diag;
425 	}
426 	diag = hw->phy.ops.init_params(hw);
427 	if (diag != 0) {
428 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
429 		return diag;
430 	}
431 	(void) e1000_get_bus_info(hw);
432 
433 	hw->mac.autoneg = 1;
434 	hw->phy.autoneg_wait_to_complete = 0;
435 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
436 
437 	e1000_init_script_state_82541(hw, TRUE);
438 	e1000_set_tbi_compatibility_82543(hw, TRUE);
439 
440 	/* Copper options */
441 	if (hw->phy.media_type == e1000_media_type_copper) {
442 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
443 		hw->phy.disable_polarity_correction = 0;
444 		hw->phy.ms_type = e1000_ms_hw_default;
445 	}
446 
447 	/*
448 	 * Start from a known state, this is important in reading the nvm
449 	 * and mac from that.
450 	 */
451 	e1000_reset_hw(hw);
452 
453 	/* Make sure we have a good EEPROM before we read from it */
454 	if (e1000_validate_nvm_checksum(hw) < 0) {
455 		/*
456 		 * Some PCI-E parts fail the first check due to
457 		 * the link being in sleep state, call it again,
458 		 * if it fails a second time its a real issue.
459 		 */
460 		diag = e1000_validate_nvm_checksum(hw);
461 		if (diag < 0) {
462 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
463 			goto error;
464 		}
465 	}
466 
467 	/* Read the permanent MAC address out of the EEPROM */
468 	diag = e1000_read_mac_addr(hw);
469 	if (diag != 0) {
470 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
471 		goto error;
472 	}
473 
474 	/* Now initialize the hardware */
475 	diag = em_hardware_init(hw);
476 	if (diag != 0) {
477 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
478 		goto error;
479 	}
480 
481 	hw->mac.get_link_status = 1;
482 
483 	/* Indicate SOL/IDER usage */
484 	diag = e1000_check_reset_block(hw);
485 	if (diag < 0) {
486 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
487 			"SOL/IDER session");
488 	}
489 	return 0;
490 
491 error:
492 	em_hw_control_release(hw);
493 	return diag;
494 }
495 
496 static int
497 eth_em_configure(struct rte_eth_dev *dev)
498 {
499 	struct e1000_interrupt *intr =
500 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
501 
502 	PMD_INIT_FUNC_TRACE();
503 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
504 	PMD_INIT_FUNC_TRACE();
505 
506 	return 0;
507 }
508 
509 static void
510 em_set_pba(struct e1000_hw *hw)
511 {
512 	uint32_t pba;
513 
514 	/*
515 	 * Packet Buffer Allocation (PBA)
516 	 * Writing PBA sets the receive portion of the buffer
517 	 * the remainder is used for the transmit buffer.
518 	 * Devices before the 82547 had a Packet Buffer of 64K.
519 	 * After the 82547 the buffer was reduced to 40K.
520 	 */
521 	switch (hw->mac.type) {
522 		case e1000_82547:
523 		case e1000_82547_rev_2:
524 		/* 82547: Total Packet Buffer is 40K */
525 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
526 			break;
527 		case e1000_82571:
528 		case e1000_82572:
529 		case e1000_80003es2lan:
530 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
531 			break;
532 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
533 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
534 			break;
535 		case e1000_82574:
536 		case e1000_82583:
537 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
538 			break;
539 		case e1000_ich8lan:
540 			pba = E1000_PBA_8K;
541 			break;
542 		case e1000_ich9lan:
543 		case e1000_ich10lan:
544 			pba = E1000_PBA_10K;
545 			break;
546 		case e1000_pchlan:
547 		case e1000_pch2lan:
548 		case e1000_pch_lpt:
549 			pba = E1000_PBA_26K;
550 			break;
551 		default:
552 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
553 	}
554 
555 	E1000_WRITE_REG(hw, E1000_PBA, pba);
556 }
557 
558 static int
559 eth_em_start(struct rte_eth_dev *dev)
560 {
561 	struct e1000_adapter *adapter =
562 		E1000_DEV_PRIVATE(dev->data->dev_private);
563 	struct e1000_hw *hw =
564 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
565 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
566 	int ret, mask;
567 	uint32_t intr_vector = 0;
568 	uint32_t *speeds;
569 	int num_speeds;
570 	bool autoneg;
571 
572 	PMD_INIT_FUNC_TRACE();
573 
574 	eth_em_stop(dev);
575 
576 	e1000_power_up_phy(hw);
577 
578 	/* Set default PBA value */
579 	em_set_pba(hw);
580 
581 	/* Put the address into the Receive Address Array */
582 	e1000_rar_set(hw, hw->mac.addr, 0);
583 
584 	/*
585 	 * With the 82571 adapter, RAR[0] may be overwritten
586 	 * when the other port is reset, we make a duplicate
587 	 * in RAR[14] for that eventuality, this assures
588 	 * the interface continues to function.
589 	 */
590 	if (hw->mac.type == e1000_82571) {
591 		e1000_set_laa_state_82571(hw, TRUE);
592 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
593 	}
594 
595 	/* Initialize the hardware */
596 	if (em_hardware_init(hw)) {
597 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
598 		return -EIO;
599 	}
600 
601 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
602 
603 	/* Configure for OS presence */
604 	em_init_manageability(hw);
605 
606 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
607 		intr_vector = dev->data->nb_rx_queues;
608 		if (rte_intr_efd_enable(intr_handle, intr_vector))
609 			return -1;
610 	}
611 
612 	if (rte_intr_dp_is_en(intr_handle)) {
613 		intr_handle->intr_vec =
614 			rte_zmalloc("intr_vec",
615 					dev->data->nb_rx_queues * sizeof(int), 0);
616 		if (intr_handle->intr_vec == NULL) {
617 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
618 						" intr_vec\n", dev->data->nb_rx_queues);
619 			return -ENOMEM;
620 		}
621 
622 		/* enable rx interrupt */
623 		em_rxq_intr_enable(hw);
624 	}
625 
626 	eth_em_tx_init(dev);
627 
628 	ret = eth_em_rx_init(dev);
629 	if (ret) {
630 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
631 		em_dev_clear_queues(dev);
632 		return ret;
633 	}
634 
635 	e1000_clear_hw_cntrs_base_generic(hw);
636 
637 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
638 			ETH_VLAN_EXTEND_MASK;
639 	eth_em_vlan_offload_set(dev, mask);
640 
641 	/* Set Interrupt Throttling Rate to maximum allowed value. */
642 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
643 
644 	/* Setup link speed and duplex */
645 	speeds = &dev->data->dev_conf.link_speeds;
646 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
647 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
648 	} else {
649 		num_speeds = 0;
650 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
651 
652 		/* Reset */
653 		hw->phy.autoneg_advertised = 0;
654 
655 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
656 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
657 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
658 			num_speeds = -1;
659 			goto error_invalid_config;
660 		}
661 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
662 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
663 			num_speeds++;
664 		}
665 		if (*speeds & ETH_LINK_SPEED_10M) {
666 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
667 			num_speeds++;
668 		}
669 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
670 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
671 			num_speeds++;
672 		}
673 		if (*speeds & ETH_LINK_SPEED_100M) {
674 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
675 			num_speeds++;
676 		}
677 		if (*speeds & ETH_LINK_SPEED_1G) {
678 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
679 			num_speeds++;
680 		}
681 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
682 			goto error_invalid_config;
683 	}
684 
685 	e1000_setup_link(hw);
686 
687 	if (rte_intr_allow_others(intr_handle)) {
688 		/* check if lsc interrupt is enabled */
689 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
690 			ret = eth_em_interrupt_setup(dev);
691 			if (ret) {
692 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
693 				em_dev_clear_queues(dev);
694 				return ret;
695 			}
696 		}
697 	} else {
698 		rte_intr_callback_unregister(intr_handle,
699 						eth_em_interrupt_handler,
700 						(void *)dev);
701 		if (dev->data->dev_conf.intr_conf.lsc != 0)
702 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
703 				     " no intr multiplex\n");
704 	}
705 	/* check if rxq interrupt is enabled */
706 	if (dev->data->dev_conf.intr_conf.rxq != 0)
707 		eth_em_rxq_interrupt_setup(dev);
708 
709 	rte_intr_enable(intr_handle);
710 
711 	adapter->stopped = 0;
712 
713 	PMD_INIT_LOG(DEBUG, "<<");
714 
715 	return 0;
716 
717 error_invalid_config:
718 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
719 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
720 	em_dev_clear_queues(dev);
721 	return -EINVAL;
722 }
723 
724 /*********************************************************************
725  *
726  *  This routine disables all traffic on the adapter by issuing a
727  *  global reset on the MAC.
728  *
729  **********************************************************************/
730 static void
731 eth_em_stop(struct rte_eth_dev *dev)
732 {
733 	struct rte_eth_link link;
734 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
735 	struct rte_intr_handle *intr_handle = &dev->pci_dev->intr_handle;
736 
737 	em_rxq_intr_disable(hw);
738 	em_lsc_intr_disable(hw);
739 
740 	e1000_reset_hw(hw);
741 	if (hw->mac.type >= e1000_82544)
742 		E1000_WRITE_REG(hw, E1000_WUC, 0);
743 
744 	/* Power down the phy. Needed to make the link go down */
745 	e1000_power_down_phy(hw);
746 
747 	em_dev_clear_queues(dev);
748 
749 	/* clear the recorded link status */
750 	memset(&link, 0, sizeof(link));
751 	rte_em_dev_atomic_write_link_status(dev, &link);
752 
753 	if (!rte_intr_allow_others(intr_handle))
754 		/* resume to the default handler */
755 		rte_intr_callback_register(intr_handle,
756 					   eth_em_interrupt_handler,
757 					   (void *)dev);
758 
759 	/* Clean datapath event and queue/vec mapping */
760 	rte_intr_efd_disable(intr_handle);
761 	if (intr_handle->intr_vec != NULL) {
762 		rte_free(intr_handle->intr_vec);
763 		intr_handle->intr_vec = NULL;
764 	}
765 }
766 
767 static void
768 eth_em_close(struct rte_eth_dev *dev)
769 {
770 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
771 	struct e1000_adapter *adapter =
772 		E1000_DEV_PRIVATE(dev->data->dev_private);
773 
774 	eth_em_stop(dev);
775 	adapter->stopped = 1;
776 	em_dev_free_queues(dev);
777 	e1000_phy_hw_reset(hw);
778 	em_release_manageability(hw);
779 	em_hw_control_release(hw);
780 }
781 
782 static int
783 em_get_rx_buffer_size(struct e1000_hw *hw)
784 {
785 	uint32_t rx_buf_size;
786 
787 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
788 	return rx_buf_size;
789 }
790 
791 /*********************************************************************
792  *
793  *  Initialize the hardware
794  *
795  **********************************************************************/
796 static int
797 em_hardware_init(struct e1000_hw *hw)
798 {
799 	uint32_t rx_buf_size;
800 	int diag;
801 
802 	/* Issue a global reset */
803 	e1000_reset_hw(hw);
804 
805 	/* Let the firmware know the OS is in control */
806 	em_hw_control_acquire(hw);
807 
808 	/*
809 	 * These parameters control the automatic generation (Tx) and
810 	 * response (Rx) to Ethernet PAUSE frames.
811 	 * - High water mark should allow for at least two standard size (1518)
812 	 *   frames to be received after sending an XOFF.
813 	 * - Low water mark works best when it is very near the high water mark.
814 	 *   This allows the receiver to restart by sending XON when it has
815 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
816 	 *   restart after one full frame is pulled from the buffer. There
817 	 *   could be several smaller frames in the buffer and if so they will
818 	 *   not trigger the XON until their total number reduces the buffer
819 	 *   by 1500.
820 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
821 	 */
822 	rx_buf_size = em_get_rx_buffer_size(hw);
823 
824 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
825 	hw->fc.low_water = hw->fc.high_water - 1500;
826 
827 	if (hw->mac.type == e1000_80003es2lan)
828 		hw->fc.pause_time = UINT16_MAX;
829 	else
830 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
831 
832 	hw->fc.send_xon = 1;
833 
834 	/* Set Flow control, use the tunable location if sane */
835 	if (em_fc_setting <= e1000_fc_full)
836 		hw->fc.requested_mode = em_fc_setting;
837 	else
838 		hw->fc.requested_mode = e1000_fc_none;
839 
840 	/* Workaround: no TX flow ctrl for PCH */
841 	if (hw->mac.type == e1000_pchlan)
842 		hw->fc.requested_mode = e1000_fc_rx_pause;
843 
844 	/* Override - settings for PCH2LAN, ya its magic :) */
845 	if (hw->mac.type == e1000_pch2lan) {
846 		hw->fc.high_water = 0x5C20;
847 		hw->fc.low_water = 0x5048;
848 		hw->fc.pause_time = 0x0650;
849 		hw->fc.refresh_time = 0x0400;
850 	} else if (hw->mac.type == e1000_pch_lpt) {
851 		hw->fc.requested_mode = e1000_fc_full;
852 	}
853 
854 	diag = e1000_init_hw(hw);
855 	if (diag < 0)
856 		return diag;
857 	e1000_check_for_link(hw);
858 	return 0;
859 }
860 
861 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
862 static void
863 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
864 {
865 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
866 	struct e1000_hw_stats *stats =
867 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
868 	int pause_frames;
869 
870 	if(hw->phy.media_type == e1000_media_type_copper ||
871 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
872 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
873 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
874 	}
875 
876 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
877 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
878 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
879 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
880 
881 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
882 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
883 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
884 	stats->dc += E1000_READ_REG(hw, E1000_DC);
885 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
886 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
887 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
888 
889 	/*
890 	 * For watchdog management we need to know if we have been
891 	 * paused during the last interval, so capture that here.
892 	 */
893 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
894 	stats->xoffrxc += pause_frames;
895 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
896 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
897 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
898 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
899 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
900 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
901 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
902 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
903 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
904 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
905 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
906 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
907 
908 	/*
909 	 * For the 64-bit byte counters the low dword must be read first.
910 	 * Both registers clear on the read of the high dword.
911 	 */
912 
913 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
914 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
915 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
916 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
917 
918 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
919 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
920 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
921 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
922 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
923 
924 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
925 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
926 
927 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
928 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
929 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
930 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
931 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
932 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
933 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
934 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
935 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
936 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
937 
938 	/* Interrupt Counts */
939 
940 	if (hw->mac.type >= e1000_82571) {
941 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
942 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
943 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
944 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
945 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
946 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
947 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
948 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
949 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
950 	}
951 
952 	if (hw->mac.type >= e1000_82543) {
953 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
954 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
955 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
956 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
957 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
958 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
959 	}
960 
961 	if (rte_stats == NULL)
962 		return;
963 
964 	/* Rx Errors */
965 	rte_stats->imissed = stats->mpc;
966 	rte_stats->ierrors = stats->crcerrs +
967 	                     stats->rlec + stats->ruc + stats->roc +
968 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
969 
970 	/* Tx Errors */
971 	rte_stats->oerrors = stats->ecol + stats->latecol;
972 
973 	rte_stats->ipackets = stats->gprc;
974 	rte_stats->opackets = stats->gptc;
975 	rte_stats->ibytes   = stats->gorc;
976 	rte_stats->obytes   = stats->gotc;
977 }
978 
979 static void
980 eth_em_stats_reset(struct rte_eth_dev *dev)
981 {
982 	struct e1000_hw_stats *hw_stats =
983 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
984 
985 	/* HW registers are cleared on read */
986 	eth_em_stats_get(dev, NULL);
987 
988 	/* Reset software totals */
989 	memset(hw_stats, 0, sizeof(*hw_stats));
990 }
991 
992 static int
993 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
994 {
995 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
996 
997 	em_rxq_intr_enable(hw);
998 	rte_intr_enable(&dev->pci_dev->intr_handle);
999 
1000 	return 0;
1001 }
1002 
1003 static int
1004 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1005 {
1006 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1007 
1008 	em_rxq_intr_disable(hw);
1009 
1010 	return 0;
1011 }
1012 
1013 static uint32_t
1014 em_get_max_pktlen(const struct e1000_hw *hw)
1015 {
1016 	switch (hw->mac.type) {
1017 	case e1000_82571:
1018 	case e1000_82572:
1019 	case e1000_ich9lan:
1020 	case e1000_ich10lan:
1021 	case e1000_pch2lan:
1022 	case e1000_pch_lpt:
1023 	case e1000_82574:
1024 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1025 	case e1000_82583:
1026 		return 0x2412;
1027 	case e1000_pchlan:
1028 		return 0x1000;
1029 	/* Adapters that do not support jumbo frames */
1030 	case e1000_ich8lan:
1031 		return ETHER_MAX_LEN;
1032 	default:
1033 		return MAX_JUMBO_FRAME_SIZE;
1034 	}
1035 }
1036 
1037 static void
1038 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1039 {
1040 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1041 
1042 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1043 	dev_info->max_rx_pktlen = em_get_max_pktlen(hw);
1044 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1045 
1046 	/*
1047 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1048 	 * Unfortunatelly, all these nics have just one TX context.
1049 	 * So we have few choises for TX:
1050 	 * - Use just one TX queue.
1051 	 * - Allow cksum offload only for one TX queue.
1052 	 * - Don't allow TX cksum offload at all.
1053 	 * For now, option #1 was chosen.
1054 	 * To use second RX queue we have to use extended RX descriptor
1055 	 * (Multiple Receive Queues are mutually exclusive with UDP
1056 	 * fragmentation and are not supported when a legacy receive
1057 	 * descriptor format is used).
1058 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1059 	 * don't support extended RXD.
1060 	 * To avoid it we support just one RX queue for now (no RSS).
1061 	 */
1062 
1063 	dev_info->max_rx_queues = 1;
1064 	dev_info->max_tx_queues = 1;
1065 
1066 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1067 		.nb_max = E1000_MAX_RING_DESC,
1068 		.nb_min = E1000_MIN_RING_DESC,
1069 		.nb_align = EM_RXD_ALIGN,
1070 	};
1071 
1072 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1073 		.nb_max = E1000_MAX_RING_DESC,
1074 		.nb_min = E1000_MIN_RING_DESC,
1075 		.nb_align = EM_TXD_ALIGN,
1076 	};
1077 
1078 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1079 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1080 			ETH_LINK_SPEED_1G;
1081 }
1082 
1083 /* return 0 means link status changed, -1 means not changed */
1084 static int
1085 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1086 {
1087 	struct e1000_hw *hw =
1088 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1089 	struct rte_eth_link link, old;
1090 	int link_check, count;
1091 
1092 	link_check = 0;
1093 	hw->mac.get_link_status = 1;
1094 
1095 	/* possible wait-to-complete in up to 9 seconds */
1096 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1097 		/* Read the real link status */
1098 		switch (hw->phy.media_type) {
1099 		case e1000_media_type_copper:
1100 			/* Do the work to read phy */
1101 			e1000_check_for_link(hw);
1102 			link_check = !hw->mac.get_link_status;
1103 			break;
1104 
1105 		case e1000_media_type_fiber:
1106 			e1000_check_for_link(hw);
1107 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1108 					E1000_STATUS_LU);
1109 			break;
1110 
1111 		case e1000_media_type_internal_serdes:
1112 			e1000_check_for_link(hw);
1113 			link_check = hw->mac.serdes_has_link;
1114 			break;
1115 
1116 		default:
1117 			break;
1118 		}
1119 		if (link_check || wait_to_complete == 0)
1120 			break;
1121 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1122 	}
1123 	memset(&link, 0, sizeof(link));
1124 	rte_em_dev_atomic_read_link_status(dev, &link);
1125 	old = link;
1126 
1127 	/* Now we check if a transition has happened */
1128 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1129 		uint16_t duplex, speed;
1130 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1131 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1132 				ETH_LINK_FULL_DUPLEX :
1133 				ETH_LINK_HALF_DUPLEX;
1134 		link.link_speed = speed;
1135 		link.link_status = ETH_LINK_UP;
1136 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1137 				ETH_LINK_SPEED_FIXED);
1138 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1139 		link.link_speed = 0;
1140 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1141 		link.link_status = ETH_LINK_DOWN;
1142 		link.link_autoneg = ETH_LINK_SPEED_FIXED;
1143 	}
1144 	rte_em_dev_atomic_write_link_status(dev, &link);
1145 
1146 	/* not changed */
1147 	if (old.link_status == link.link_status)
1148 		return -1;
1149 
1150 	/* changed */
1151 	return 0;
1152 }
1153 
1154 /*
1155  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1156  * For ASF and Pass Through versions of f/w this means
1157  * that the driver is loaded. For AMT version type f/w
1158  * this means that the network i/f is open.
1159  */
1160 static void
1161 em_hw_control_acquire(struct e1000_hw *hw)
1162 {
1163 	uint32_t ctrl_ext, swsm;
1164 
1165 	/* Let firmware know the driver has taken over */
1166 	if (hw->mac.type == e1000_82573) {
1167 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1168 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1169 
1170 	} else {
1171 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1172 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1173 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1174 	}
1175 }
1176 
1177 /*
1178  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1179  * For ASF and Pass Through versions of f/w this means that the
1180  * driver is no longer loaded. For AMT versions of the
1181  * f/w this means that the network i/f is closed.
1182  */
1183 static void
1184 em_hw_control_release(struct e1000_hw *hw)
1185 {
1186 	uint32_t ctrl_ext, swsm;
1187 
1188 	/* Let firmware taken over control of h/w */
1189 	if (hw->mac.type == e1000_82573) {
1190 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1191 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1192 	} else {
1193 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1194 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1195 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1196 	}
1197 }
1198 
1199 /*
1200  * Bit of a misnomer, what this really means is
1201  * to enable OS management of the system... aka
1202  * to disable special hardware management features.
1203  */
1204 static void
1205 em_init_manageability(struct e1000_hw *hw)
1206 {
1207 	if (e1000_enable_mng_pass_thru(hw)) {
1208 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1209 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1210 
1211 		/* disable hardware interception of ARP */
1212 		manc &= ~(E1000_MANC_ARP_EN);
1213 
1214 		/* enable receiving management packets to the host */
1215 		manc |= E1000_MANC_EN_MNG2HOST;
1216 		manc2h |= 1 << 5;  /* Mng Port 623 */
1217 		manc2h |= 1 << 6;  /* Mng Port 664 */
1218 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1219 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1220 	}
1221 }
1222 
1223 /*
1224  * Give control back to hardware management
1225  * controller if there is one.
1226  */
1227 static void
1228 em_release_manageability(struct e1000_hw *hw)
1229 {
1230 	uint32_t manc;
1231 
1232 	if (e1000_enable_mng_pass_thru(hw)) {
1233 		manc = E1000_READ_REG(hw, E1000_MANC);
1234 
1235 		/* re-enable hardware interception of ARP */
1236 		manc |= E1000_MANC_ARP_EN;
1237 		manc &= ~E1000_MANC_EN_MNG2HOST;
1238 
1239 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1240 	}
1241 }
1242 
1243 static void
1244 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1245 {
1246 	struct e1000_hw *hw =
1247 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1248 	uint32_t rctl;
1249 
1250 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1251 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1252 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1253 }
1254 
1255 static void
1256 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1257 {
1258 	struct e1000_hw *hw =
1259 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1260 	uint32_t rctl;
1261 
1262 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1263 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1264 	if (dev->data->all_multicast == 1)
1265 		rctl |= E1000_RCTL_MPE;
1266 	else
1267 		rctl &= (~E1000_RCTL_MPE);
1268 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1269 }
1270 
1271 static void
1272 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1273 {
1274 	struct e1000_hw *hw =
1275 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1276 	uint32_t rctl;
1277 
1278 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1279 	rctl |= E1000_RCTL_MPE;
1280 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1281 }
1282 
1283 static void
1284 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1285 {
1286 	struct e1000_hw *hw =
1287 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1288 	uint32_t rctl;
1289 
1290 	if (dev->data->promiscuous == 1)
1291 		return; /* must remain in all_multicast mode */
1292 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1293 	rctl &= (~E1000_RCTL_MPE);
1294 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1295 }
1296 
1297 static int
1298 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1299 {
1300 	struct e1000_hw *hw =
1301 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1302 	struct e1000_vfta * shadow_vfta =
1303 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1304 	uint32_t vfta;
1305 	uint32_t vid_idx;
1306 	uint32_t vid_bit;
1307 
1308 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1309 			      E1000_VFTA_ENTRY_MASK);
1310 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1311 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1312 	if (on)
1313 		vfta |= vid_bit;
1314 	else
1315 		vfta &= ~vid_bit;
1316 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1317 
1318 	/* update local VFTA copy */
1319 	shadow_vfta->vfta[vid_idx] = vfta;
1320 
1321 	return 0;
1322 }
1323 
1324 static void
1325 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1326 {
1327 	struct e1000_hw *hw =
1328 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1329 	uint32_t reg;
1330 
1331 	/* Filter Table Disable */
1332 	reg = E1000_READ_REG(hw, E1000_RCTL);
1333 	reg &= ~E1000_RCTL_CFIEN;
1334 	reg &= ~E1000_RCTL_VFE;
1335 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1336 }
1337 
1338 static void
1339 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1340 {
1341 	struct e1000_hw *hw =
1342 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1343 	struct e1000_vfta * shadow_vfta =
1344 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1345 	uint32_t reg;
1346 	int i;
1347 
1348 	/* Filter Table Enable, CFI not used for packet acceptance */
1349 	reg = E1000_READ_REG(hw, E1000_RCTL);
1350 	reg &= ~E1000_RCTL_CFIEN;
1351 	reg |= E1000_RCTL_VFE;
1352 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1353 
1354 	/* restore vfta from local copy */
1355 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1356 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1357 }
1358 
1359 static void
1360 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1361 {
1362 	struct e1000_hw *hw =
1363 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1364 	uint32_t reg;
1365 
1366 	/* VLAN Mode Disable */
1367 	reg = E1000_READ_REG(hw, E1000_CTRL);
1368 	reg &= ~E1000_CTRL_VME;
1369 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1370 
1371 }
1372 
1373 static void
1374 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1375 {
1376 	struct e1000_hw *hw =
1377 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1378 	uint32_t reg;
1379 
1380 	/* VLAN Mode Enable */
1381 	reg = E1000_READ_REG(hw, E1000_CTRL);
1382 	reg |= E1000_CTRL_VME;
1383 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1384 }
1385 
1386 static void
1387 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1388 {
1389 	if(mask & ETH_VLAN_STRIP_MASK){
1390 		if (dev->data->dev_conf.rxmode.hw_vlan_strip)
1391 			em_vlan_hw_strip_enable(dev);
1392 		else
1393 			em_vlan_hw_strip_disable(dev);
1394 	}
1395 
1396 	if(mask & ETH_VLAN_FILTER_MASK){
1397 		if (dev->data->dev_conf.rxmode.hw_vlan_filter)
1398 			em_vlan_hw_filter_enable(dev);
1399 		else
1400 			em_vlan_hw_filter_disable(dev);
1401 	}
1402 }
1403 
1404 /*
1405  * It enables the interrupt mask and then enable the interrupt.
1406  *
1407  * @param dev
1408  *  Pointer to struct rte_eth_dev.
1409  *
1410  * @return
1411  *  - On success, zero.
1412  *  - On failure, a negative value.
1413  */
1414 static int
1415 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1416 {
1417 	uint32_t regval;
1418 	struct e1000_hw *hw =
1419 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1420 
1421 	/* clear interrupt */
1422 	E1000_READ_REG(hw, E1000_ICR);
1423 	regval = E1000_READ_REG(hw, E1000_IMS);
1424 	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1425 	return 0;
1426 }
1427 
1428 /*
1429  * It clears the interrupt causes and enables the interrupt.
1430  * It will be called once only during nic initialized.
1431  *
1432  * @param dev
1433  *  Pointer to struct rte_eth_dev.
1434  *
1435  * @return
1436  *  - On success, zero.
1437  *  - On failure, a negative value.
1438  */
1439 static int
1440 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1441 {
1442 	struct e1000_hw *hw =
1443 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1444 
1445 	E1000_READ_REG(hw, E1000_ICR);
1446 	em_rxq_intr_enable(hw);
1447 	return 0;
1448 }
1449 
1450 /*
1451  * It enable receive packet interrupt.
1452  * @param hw
1453  * Pointer to struct e1000_hw
1454  *
1455  * @return
1456  */
1457 static void
1458 em_rxq_intr_enable(struct e1000_hw *hw)
1459 {
1460 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1461 	E1000_WRITE_FLUSH(hw);
1462 }
1463 
1464 /*
1465  * It disabled lsc interrupt.
1466  * @param hw
1467  * Pointer to struct e1000_hw
1468  *
1469  * @return
1470  */
1471 static void
1472 em_lsc_intr_disable(struct e1000_hw *hw)
1473 {
1474 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1475 	E1000_WRITE_FLUSH(hw);
1476 }
1477 
1478 /*
1479  * It disabled receive packet interrupt.
1480  * @param hw
1481  * Pointer to struct e1000_hw
1482  *
1483  * @return
1484  */
1485 static void
1486 em_rxq_intr_disable(struct e1000_hw *hw)
1487 {
1488 	E1000_READ_REG(hw, E1000_ICR);
1489 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1490 	E1000_WRITE_FLUSH(hw);
1491 }
1492 
1493 /*
1494  * It reads ICR and gets interrupt causes, check it and set a bit flag
1495  * to update link status.
1496  *
1497  * @param dev
1498  *  Pointer to struct rte_eth_dev.
1499  *
1500  * @return
1501  *  - On success, zero.
1502  *  - On failure, a negative value.
1503  */
1504 static int
1505 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1506 {
1507 	uint32_t icr;
1508 	struct e1000_hw *hw =
1509 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1510 	struct e1000_interrupt *intr =
1511 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1512 
1513 	/* read-on-clear nic registers here */
1514 	icr = E1000_READ_REG(hw, E1000_ICR);
1515 	if (icr & E1000_ICR_LSC) {
1516 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1517 	}
1518 
1519 	return 0;
1520 }
1521 
1522 /*
1523  * It executes link_update after knowing an interrupt is prsent.
1524  *
1525  * @param dev
1526  *  Pointer to struct rte_eth_dev.
1527  *
1528  * @return
1529  *  - On success, zero.
1530  *  - On failure, a negative value.
1531  */
1532 static int
1533 eth_em_interrupt_action(struct rte_eth_dev *dev)
1534 {
1535 	struct e1000_hw *hw =
1536 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1537 	struct e1000_interrupt *intr =
1538 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1539 	uint32_t tctl, rctl;
1540 	struct rte_eth_link link;
1541 	int ret;
1542 
1543 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1544 		return -1;
1545 
1546 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1547 	rte_intr_enable(&(dev->pci_dev->intr_handle));
1548 
1549 	/* set get_link_status to check register later */
1550 	hw->mac.get_link_status = 1;
1551 	ret = eth_em_link_update(dev, 0);
1552 
1553 	/* check if link has changed */
1554 	if (ret < 0)
1555 		return 0;
1556 
1557 	memset(&link, 0, sizeof(link));
1558 	rte_em_dev_atomic_read_link_status(dev, &link);
1559 	if (link.link_status) {
1560 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1561 			     dev->data->port_id, (unsigned)link.link_speed,
1562 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1563 			     "full-duplex" : "half-duplex");
1564 	} else {
1565 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1566 	}
1567 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1568 		     dev->pci_dev->addr.domain, dev->pci_dev->addr.bus,
1569 		     dev->pci_dev->addr.devid, dev->pci_dev->addr.function);
1570 
1571 	tctl = E1000_READ_REG(hw, E1000_TCTL);
1572 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1573 	if (link.link_status) {
1574 		/* enable Tx/Rx */
1575 		tctl |= E1000_TCTL_EN;
1576 		rctl |= E1000_RCTL_EN;
1577 	} else {
1578 		/* disable Tx/Rx */
1579 		tctl &= ~E1000_TCTL_EN;
1580 		rctl &= ~E1000_RCTL_EN;
1581 	}
1582 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
1583 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1584 	E1000_WRITE_FLUSH(hw);
1585 
1586 	return 0;
1587 }
1588 
1589 /**
1590  * Interrupt handler which shall be registered at first.
1591  *
1592  * @param handle
1593  *  Pointer to interrupt handle.
1594  * @param param
1595  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1596  *
1597  * @return
1598  *  void
1599  */
1600 static void
1601 eth_em_interrupt_handler(__rte_unused struct rte_intr_handle *handle,
1602 							void *param)
1603 {
1604 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1605 
1606 	eth_em_interrupt_get_status(dev);
1607 	eth_em_interrupt_action(dev);
1608 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC);
1609 }
1610 
1611 static int
1612 eth_em_led_on(struct rte_eth_dev *dev)
1613 {
1614 	struct e1000_hw *hw;
1615 
1616 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1617 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1618 }
1619 
1620 static int
1621 eth_em_led_off(struct rte_eth_dev *dev)
1622 {
1623 	struct e1000_hw *hw;
1624 
1625 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1626 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1627 }
1628 
1629 static int
1630 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1631 {
1632 	struct e1000_hw *hw;
1633 	uint32_t ctrl;
1634 	int tx_pause;
1635 	int rx_pause;
1636 
1637 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1638 	fc_conf->pause_time = hw->fc.pause_time;
1639 	fc_conf->high_water = hw->fc.high_water;
1640 	fc_conf->low_water = hw->fc.low_water;
1641 	fc_conf->send_xon = hw->fc.send_xon;
1642 	fc_conf->autoneg = hw->mac.autoneg;
1643 
1644 	/*
1645 	 * Return rx_pause and tx_pause status according to actual setting of
1646 	 * the TFCE and RFCE bits in the CTRL register.
1647 	 */
1648 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1649 	if (ctrl & E1000_CTRL_TFCE)
1650 		tx_pause = 1;
1651 	else
1652 		tx_pause = 0;
1653 
1654 	if (ctrl & E1000_CTRL_RFCE)
1655 		rx_pause = 1;
1656 	else
1657 		rx_pause = 0;
1658 
1659 	if (rx_pause && tx_pause)
1660 		fc_conf->mode = RTE_FC_FULL;
1661 	else if (rx_pause)
1662 		fc_conf->mode = RTE_FC_RX_PAUSE;
1663 	else if (tx_pause)
1664 		fc_conf->mode = RTE_FC_TX_PAUSE;
1665 	else
1666 		fc_conf->mode = RTE_FC_NONE;
1667 
1668 	return 0;
1669 }
1670 
1671 static int
1672 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1673 {
1674 	struct e1000_hw *hw;
1675 	int err;
1676 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1677 		e1000_fc_none,
1678 		e1000_fc_rx_pause,
1679 		e1000_fc_tx_pause,
1680 		e1000_fc_full
1681 	};
1682 	uint32_t rx_buf_size;
1683 	uint32_t max_high_water;
1684 	uint32_t rctl;
1685 
1686 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1687 	if (fc_conf->autoneg != hw->mac.autoneg)
1688 		return -ENOTSUP;
1689 	rx_buf_size = em_get_rx_buffer_size(hw);
1690 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1691 
1692 	/* At least reserve one Ethernet frame for watermark */
1693 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1694 	if ((fc_conf->high_water > max_high_water) ||
1695 	    (fc_conf->high_water < fc_conf->low_water)) {
1696 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1697 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1698 		return -EINVAL;
1699 	}
1700 
1701 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1702 	hw->fc.pause_time     = fc_conf->pause_time;
1703 	hw->fc.high_water     = fc_conf->high_water;
1704 	hw->fc.low_water      = fc_conf->low_water;
1705 	hw->fc.send_xon	      = fc_conf->send_xon;
1706 
1707 	err = e1000_setup_link_generic(hw);
1708 	if (err == E1000_SUCCESS) {
1709 
1710 		/* check if we want to forward MAC frames - driver doesn't have native
1711 		 * capability to do that, so we'll write the registers ourselves */
1712 
1713 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1714 
1715 		/* set or clear MFLCN.PMCF bit depending on configuration */
1716 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1717 			rctl |= E1000_RCTL_PMCF;
1718 		else
1719 			rctl &= ~E1000_RCTL_PMCF;
1720 
1721 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1722 		E1000_WRITE_FLUSH(hw);
1723 
1724 		return 0;
1725 	}
1726 
1727 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1728 	return -EIO;
1729 }
1730 
1731 static void
1732 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1733 		uint32_t index, __rte_unused uint32_t pool)
1734 {
1735 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1736 
1737 	e1000_rar_set(hw, mac_addr->addr_bytes, index);
1738 }
1739 
1740 static void
1741 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1742 {
1743 	uint8_t addr[ETHER_ADDR_LEN];
1744 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1745 
1746 	memset(addr, 0, sizeof(addr));
1747 
1748 	e1000_rar_set(hw, addr, index);
1749 }
1750 
1751 static int
1752 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1753 {
1754 	struct rte_eth_dev_info dev_info;
1755 	struct e1000_hw *hw;
1756 	uint32_t frame_size;
1757 	uint32_t rctl;
1758 
1759 	eth_em_infos_get(dev, &dev_info);
1760 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1761 
1762 	/* check that mtu is within the allowed range */
1763 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1764 		return -EINVAL;
1765 
1766 	/* refuse mtu that requires the support of scattered packets when this
1767 	 * feature has not been enabled before. */
1768 	if (!dev->data->scattered_rx &&
1769 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1770 		return -EINVAL;
1771 
1772 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1773 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1774 
1775 	/* switch to jumbo mode if needed */
1776 	if (frame_size > ETHER_MAX_LEN) {
1777 		dev->data->dev_conf.rxmode.jumbo_frame = 1;
1778 		rctl |= E1000_RCTL_LPE;
1779 	} else {
1780 		dev->data->dev_conf.rxmode.jumbo_frame = 0;
1781 		rctl &= ~E1000_RCTL_LPE;
1782 	}
1783 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1784 
1785 	/* update max frame size */
1786 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1787 	return 0;
1788 }
1789 
1790 static int
1791 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1792 			struct ether_addr *mc_addr_set,
1793 			uint32_t nb_mc_addr)
1794 {
1795 	struct e1000_hw *hw;
1796 
1797 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1798 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1799 	return 0;
1800 }
1801 
1802 struct rte_driver em_pmd_drv = {
1803 	.type = PMD_PDEV,
1804 	.init = rte_em_pmd_init,
1805 };
1806 
1807 PMD_REGISTER_DRIVER(em_pmd_drv, em);
1808 DRIVER_REGISTER_PCI_TABLE(em, pci_id_em_map);
1809