xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision a4f56123)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2016 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdarg.h>
10 
11 #include <rte_common.h>
12 #include <rte_interrupts.h>
13 #include <rte_byteorder.h>
14 #include <rte_debug.h>
15 #include <rte_pci.h>
16 #include <rte_bus_pci.h>
17 #include <rte_ether.h>
18 #include <rte_ethdev_driver.h>
19 #include <rte_ethdev_pci.h>
20 #include <rte_memory.h>
21 #include <rte_eal.h>
22 #include <rte_malloc.h>
23 #include <rte_dev.h>
24 
25 #include "e1000_logs.h"
26 #include "base/e1000_api.h"
27 #include "e1000_ethdev.h"
28 
29 #define EM_EIAC			0x000DC
30 
31 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
32 
33 
34 static int eth_em_configure(struct rte_eth_dev *dev);
35 static int eth_em_start(struct rte_eth_dev *dev);
36 static void eth_em_stop(struct rte_eth_dev *dev);
37 static void eth_em_close(struct rte_eth_dev *dev);
38 static void eth_em_promiscuous_enable(struct rte_eth_dev *dev);
39 static void eth_em_promiscuous_disable(struct rte_eth_dev *dev);
40 static void eth_em_allmulticast_enable(struct rte_eth_dev *dev);
41 static void eth_em_allmulticast_disable(struct rte_eth_dev *dev);
42 static int eth_em_link_update(struct rte_eth_dev *dev,
43 				int wait_to_complete);
44 static int eth_em_stats_get(struct rte_eth_dev *dev,
45 				struct rte_eth_stats *rte_stats);
46 static void eth_em_stats_reset(struct rte_eth_dev *dev);
47 static void eth_em_infos_get(struct rte_eth_dev *dev,
48 				struct rte_eth_dev_info *dev_info);
49 static int eth_em_flow_ctrl_get(struct rte_eth_dev *dev,
50 				struct rte_eth_fc_conf *fc_conf);
51 static int eth_em_flow_ctrl_set(struct rte_eth_dev *dev,
52 				struct rte_eth_fc_conf *fc_conf);
53 static int eth_em_interrupt_setup(struct rte_eth_dev *dev);
54 static int eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev);
55 static int eth_em_interrupt_get_status(struct rte_eth_dev *dev);
56 static int eth_em_interrupt_action(struct rte_eth_dev *dev,
57 				   struct rte_intr_handle *handle);
58 static void eth_em_interrupt_handler(void *param);
59 
60 static int em_hw_init(struct e1000_hw *hw);
61 static int em_hardware_init(struct e1000_hw *hw);
62 static void em_hw_control_acquire(struct e1000_hw *hw);
63 static void em_hw_control_release(struct e1000_hw *hw);
64 static void em_init_manageability(struct e1000_hw *hw);
65 static void em_release_manageability(struct e1000_hw *hw);
66 
67 static int eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
68 
69 static int eth_em_vlan_filter_set(struct rte_eth_dev *dev,
70 		uint16_t vlan_id, int on);
71 static int eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask);
72 static void em_vlan_hw_filter_enable(struct rte_eth_dev *dev);
73 static void em_vlan_hw_filter_disable(struct rte_eth_dev *dev);
74 static void em_vlan_hw_strip_enable(struct rte_eth_dev *dev);
75 static void em_vlan_hw_strip_disable(struct rte_eth_dev *dev);
76 
77 /*
78 static void eth_em_vlan_filter_set(struct rte_eth_dev *dev,
79 					uint16_t vlan_id, int on);
80 */
81 
82 static int eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
83 static int eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
84 static void em_lsc_intr_disable(struct e1000_hw *hw);
85 static void em_rxq_intr_enable(struct e1000_hw *hw);
86 static void em_rxq_intr_disable(struct e1000_hw *hw);
87 
88 static int eth_em_led_on(struct rte_eth_dev *dev);
89 static int eth_em_led_off(struct rte_eth_dev *dev);
90 
91 static int em_get_rx_buffer_size(struct e1000_hw *hw);
92 static int eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
93 			  uint32_t index, uint32_t pool);
94 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
95 static int eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
96 					 struct ether_addr *addr);
97 
98 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
99 				   struct ether_addr *mc_addr_set,
100 				   uint32_t nb_mc_addr);
101 
102 #define EM_FC_PAUSE_TIME 0x0680
103 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
104 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
105 
106 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
107 
108 /*
109  * The set of PCI devices this driver supports
110  */
111 static const struct rte_pci_id pci_id_em_map[] = {
112 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
113 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
114 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
115 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
116 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
117 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
118 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
119 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
120 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
121 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
122 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
123 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
124 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
125 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
126 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
127 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
128 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
129 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
130 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
131 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
132 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
133 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
134 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
135 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH2_LV_LM) },
136 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
137 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
138 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
139 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
157 	{ .vendor_id = 0, /* sentinel */ },
158 };
159 
160 static const struct eth_dev_ops eth_em_ops = {
161 	.dev_configure        = eth_em_configure,
162 	.dev_start            = eth_em_start,
163 	.dev_stop             = eth_em_stop,
164 	.dev_close            = eth_em_close,
165 	.promiscuous_enable   = eth_em_promiscuous_enable,
166 	.promiscuous_disable  = eth_em_promiscuous_disable,
167 	.allmulticast_enable  = eth_em_allmulticast_enable,
168 	.allmulticast_disable = eth_em_allmulticast_disable,
169 	.link_update          = eth_em_link_update,
170 	.stats_get            = eth_em_stats_get,
171 	.stats_reset          = eth_em_stats_reset,
172 	.dev_infos_get        = eth_em_infos_get,
173 	.mtu_set              = eth_em_mtu_set,
174 	.vlan_filter_set      = eth_em_vlan_filter_set,
175 	.vlan_offload_set     = eth_em_vlan_offload_set,
176 	.rx_queue_setup       = eth_em_rx_queue_setup,
177 	.rx_queue_release     = eth_em_rx_queue_release,
178 	.rx_queue_count       = eth_em_rx_queue_count,
179 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
180 	.rx_descriptor_status = eth_em_rx_descriptor_status,
181 	.tx_descriptor_status = eth_em_tx_descriptor_status,
182 	.tx_queue_setup       = eth_em_tx_queue_setup,
183 	.tx_queue_release     = eth_em_tx_queue_release,
184 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
185 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
186 	.dev_led_on           = eth_em_led_on,
187 	.dev_led_off          = eth_em_led_off,
188 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
189 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
190 	.mac_addr_set         = eth_em_default_mac_addr_set,
191 	.mac_addr_add         = eth_em_rar_set,
192 	.mac_addr_remove      = eth_em_rar_clear,
193 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
194 	.rxq_info_get         = em_rxq_info_get,
195 	.txq_info_get         = em_txq_info_get,
196 };
197 
198 
199 /**
200  *  eth_em_dev_is_ich8 - Check for ICH8 device
201  *  @hw: pointer to the HW structure
202  *
203  *  return TRUE for ICH8, otherwise FALSE
204  **/
205 static bool
206 eth_em_dev_is_ich8(struct e1000_hw *hw)
207 {
208 	DEBUGFUNC("eth_em_dev_is_ich8");
209 
210 	switch (hw->device_id) {
211 	case E1000_DEV_ID_PCH2_LV_LM:
212 	case E1000_DEV_ID_PCH_LPT_I217_LM:
213 	case E1000_DEV_ID_PCH_LPT_I217_V:
214 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
215 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
216 	case E1000_DEV_ID_PCH_I218_V2:
217 	case E1000_DEV_ID_PCH_I218_LM2:
218 	case E1000_DEV_ID_PCH_I218_V3:
219 	case E1000_DEV_ID_PCH_I218_LM3:
220 	case E1000_DEV_ID_PCH_SPT_I219_LM:
221 	case E1000_DEV_ID_PCH_SPT_I219_V:
222 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
223 	case E1000_DEV_ID_PCH_SPT_I219_V2:
224 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
225 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
226 	case E1000_DEV_ID_PCH_SPT_I219_V4:
227 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
228 	case E1000_DEV_ID_PCH_SPT_I219_V5:
229 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
230 	case E1000_DEV_ID_PCH_CNP_I219_V6:
231 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
232 	case E1000_DEV_ID_PCH_CNP_I219_V7:
233 		return 1;
234 	default:
235 		return 0;
236 	}
237 }
238 
239 static int
240 eth_em_dev_init(struct rte_eth_dev *eth_dev)
241 {
242 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
243 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
244 	struct e1000_adapter *adapter =
245 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
246 	struct e1000_hw *hw =
247 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
248 	struct e1000_vfta * shadow_vfta =
249 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
250 
251 	eth_dev->dev_ops = &eth_em_ops;
252 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
253 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
254 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
255 
256 	/* for secondary processes, we don't initialise any further as primary
257 	 * has already done this work. Only check we don't need a different
258 	 * RX function */
259 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
260 		if (eth_dev->data->scattered_rx)
261 			eth_dev->rx_pkt_burst =
262 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
263 		return 0;
264 	}
265 
266 	rte_eth_copy_pci_info(eth_dev, pci_dev);
267 
268 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
269 	hw->device_id = pci_dev->id.device_id;
270 	adapter->stopped = 0;
271 
272 	/* For ICH8 support we'll need to map the flash memory BAR */
273 	if (eth_em_dev_is_ich8(hw))
274 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
275 
276 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
277 			em_hw_init(hw) != 0) {
278 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
279 			"failed to init HW",
280 			eth_dev->data->port_id, pci_dev->id.vendor_id,
281 			pci_dev->id.device_id);
282 		return -ENODEV;
283 	}
284 
285 	/* Allocate memory for storing MAC addresses */
286 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", ETHER_ADDR_LEN *
287 			hw->mac.rar_entry_count, 0);
288 	if (eth_dev->data->mac_addrs == NULL) {
289 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
290 			"store MAC addresses",
291 			ETHER_ADDR_LEN * hw->mac.rar_entry_count);
292 		return -ENOMEM;
293 	}
294 
295 	/* Copy the permanent MAC address */
296 	ether_addr_copy((struct ether_addr *) hw->mac.addr,
297 		eth_dev->data->mac_addrs);
298 
299 	/* initialize the vfta */
300 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
301 
302 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
303 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
304 		     pci_dev->id.device_id);
305 
306 	rte_intr_callback_register(intr_handle,
307 				   eth_em_interrupt_handler, eth_dev);
308 
309 	return 0;
310 }
311 
312 static int
313 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
314 {
315 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
316 	struct e1000_adapter *adapter =
317 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
318 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
319 
320 	PMD_INIT_FUNC_TRACE();
321 
322 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
323 		return -EPERM;
324 
325 	if (adapter->stopped == 0)
326 		eth_em_close(eth_dev);
327 
328 	eth_dev->dev_ops = NULL;
329 	eth_dev->rx_pkt_burst = NULL;
330 	eth_dev->tx_pkt_burst = NULL;
331 
332 	rte_free(eth_dev->data->mac_addrs);
333 	eth_dev->data->mac_addrs = NULL;
334 
335 	/* disable uio intr before callback unregister */
336 	rte_intr_disable(intr_handle);
337 	rte_intr_callback_unregister(intr_handle,
338 				     eth_em_interrupt_handler, eth_dev);
339 
340 	return 0;
341 }
342 
343 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
344 	struct rte_pci_device *pci_dev)
345 {
346 	return rte_eth_dev_pci_generic_probe(pci_dev,
347 		sizeof(struct e1000_adapter), eth_em_dev_init);
348 }
349 
350 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
351 {
352 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
353 }
354 
355 static struct rte_pci_driver rte_em_pmd = {
356 	.id_table = pci_id_em_map,
357 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
358 		     RTE_PCI_DRV_IOVA_AS_VA,
359 	.probe = eth_em_pci_probe,
360 	.remove = eth_em_pci_remove,
361 };
362 
363 static int
364 em_hw_init(struct e1000_hw *hw)
365 {
366 	int diag;
367 
368 	diag = hw->mac.ops.init_params(hw);
369 	if (diag != 0) {
370 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
371 		return diag;
372 	}
373 	diag = hw->nvm.ops.init_params(hw);
374 	if (diag != 0) {
375 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
376 		return diag;
377 	}
378 	diag = hw->phy.ops.init_params(hw);
379 	if (diag != 0) {
380 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
381 		return diag;
382 	}
383 	(void) e1000_get_bus_info(hw);
384 
385 	hw->mac.autoneg = 1;
386 	hw->phy.autoneg_wait_to_complete = 0;
387 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
388 
389 	e1000_init_script_state_82541(hw, TRUE);
390 	e1000_set_tbi_compatibility_82543(hw, TRUE);
391 
392 	/* Copper options */
393 	if (hw->phy.media_type == e1000_media_type_copper) {
394 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
395 		hw->phy.disable_polarity_correction = 0;
396 		hw->phy.ms_type = e1000_ms_hw_default;
397 	}
398 
399 	/*
400 	 * Start from a known state, this is important in reading the nvm
401 	 * and mac from that.
402 	 */
403 	e1000_reset_hw(hw);
404 
405 	/* Make sure we have a good EEPROM before we read from it */
406 	if (e1000_validate_nvm_checksum(hw) < 0) {
407 		/*
408 		 * Some PCI-E parts fail the first check due to
409 		 * the link being in sleep state, call it again,
410 		 * if it fails a second time its a real issue.
411 		 */
412 		diag = e1000_validate_nvm_checksum(hw);
413 		if (diag < 0) {
414 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
415 			goto error;
416 		}
417 	}
418 
419 	/* Read the permanent MAC address out of the EEPROM */
420 	diag = e1000_read_mac_addr(hw);
421 	if (diag != 0) {
422 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
423 		goto error;
424 	}
425 
426 	/* Now initialize the hardware */
427 	diag = em_hardware_init(hw);
428 	if (diag != 0) {
429 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
430 		goto error;
431 	}
432 
433 	hw->mac.get_link_status = 1;
434 
435 	/* Indicate SOL/IDER usage */
436 	diag = e1000_check_reset_block(hw);
437 	if (diag < 0) {
438 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
439 			"SOL/IDER session");
440 	}
441 	return 0;
442 
443 error:
444 	em_hw_control_release(hw);
445 	return diag;
446 }
447 
448 static int
449 eth_em_configure(struct rte_eth_dev *dev)
450 {
451 	struct e1000_interrupt *intr =
452 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
453 
454 	PMD_INIT_FUNC_TRACE();
455 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
456 
457 	PMD_INIT_FUNC_TRACE();
458 
459 	return 0;
460 }
461 
462 static void
463 em_set_pba(struct e1000_hw *hw)
464 {
465 	uint32_t pba;
466 
467 	/*
468 	 * Packet Buffer Allocation (PBA)
469 	 * Writing PBA sets the receive portion of the buffer
470 	 * the remainder is used for the transmit buffer.
471 	 * Devices before the 82547 had a Packet Buffer of 64K.
472 	 * After the 82547 the buffer was reduced to 40K.
473 	 */
474 	switch (hw->mac.type) {
475 		case e1000_82547:
476 		case e1000_82547_rev_2:
477 		/* 82547: Total Packet Buffer is 40K */
478 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
479 			break;
480 		case e1000_82571:
481 		case e1000_82572:
482 		case e1000_80003es2lan:
483 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
484 			break;
485 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
486 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
487 			break;
488 		case e1000_82574:
489 		case e1000_82583:
490 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
491 			break;
492 		case e1000_ich8lan:
493 			pba = E1000_PBA_8K;
494 			break;
495 		case e1000_ich9lan:
496 		case e1000_ich10lan:
497 			pba = E1000_PBA_10K;
498 			break;
499 		case e1000_pchlan:
500 		case e1000_pch2lan:
501 		case e1000_pch_lpt:
502 		case e1000_pch_spt:
503 		case e1000_pch_cnp:
504 			pba = E1000_PBA_26K;
505 			break;
506 		default:
507 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
508 	}
509 
510 	E1000_WRITE_REG(hw, E1000_PBA, pba);
511 }
512 
513 static void
514 eth_em_rxtx_control(struct rte_eth_dev *dev,
515 		    bool enable)
516 {
517 	struct e1000_hw *hw =
518 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
519 	uint32_t tctl, rctl;
520 
521 	tctl = E1000_READ_REG(hw, E1000_TCTL);
522 	rctl = E1000_READ_REG(hw, E1000_RCTL);
523 	if (enable) {
524 		/* enable Tx/Rx */
525 		tctl |= E1000_TCTL_EN;
526 		rctl |= E1000_RCTL_EN;
527 	} else {
528 		/* disable Tx/Rx */
529 		tctl &= ~E1000_TCTL_EN;
530 		rctl &= ~E1000_RCTL_EN;
531 	}
532 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
533 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
534 	E1000_WRITE_FLUSH(hw);
535 }
536 
537 static int
538 eth_em_start(struct rte_eth_dev *dev)
539 {
540 	struct e1000_adapter *adapter =
541 		E1000_DEV_PRIVATE(dev->data->dev_private);
542 	struct e1000_hw *hw =
543 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
544 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
545 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
546 	int ret, mask;
547 	uint32_t intr_vector = 0;
548 	uint32_t *speeds;
549 	int num_speeds;
550 	bool autoneg;
551 
552 	PMD_INIT_FUNC_TRACE();
553 
554 	eth_em_stop(dev);
555 
556 	e1000_power_up_phy(hw);
557 
558 	/* Set default PBA value */
559 	em_set_pba(hw);
560 
561 	/* Put the address into the Receive Address Array */
562 	e1000_rar_set(hw, hw->mac.addr, 0);
563 
564 	/*
565 	 * With the 82571 adapter, RAR[0] may be overwritten
566 	 * when the other port is reset, we make a duplicate
567 	 * in RAR[14] for that eventuality, this assures
568 	 * the interface continues to function.
569 	 */
570 	if (hw->mac.type == e1000_82571) {
571 		e1000_set_laa_state_82571(hw, TRUE);
572 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
573 	}
574 
575 	/* Initialize the hardware */
576 	if (em_hardware_init(hw)) {
577 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
578 		return -EIO;
579 	}
580 
581 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
582 
583 	/* Configure for OS presence */
584 	em_init_manageability(hw);
585 
586 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
587 		intr_vector = dev->data->nb_rx_queues;
588 		if (rte_intr_efd_enable(intr_handle, intr_vector))
589 			return -1;
590 	}
591 
592 	if (rte_intr_dp_is_en(intr_handle)) {
593 		intr_handle->intr_vec =
594 			rte_zmalloc("intr_vec",
595 					dev->data->nb_rx_queues * sizeof(int), 0);
596 		if (intr_handle->intr_vec == NULL) {
597 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
598 						" intr_vec", dev->data->nb_rx_queues);
599 			return -ENOMEM;
600 		}
601 
602 		/* enable rx interrupt */
603 		em_rxq_intr_enable(hw);
604 	}
605 
606 	eth_em_tx_init(dev);
607 
608 	ret = eth_em_rx_init(dev);
609 	if (ret) {
610 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
611 		em_dev_clear_queues(dev);
612 		return ret;
613 	}
614 
615 	e1000_clear_hw_cntrs_base_generic(hw);
616 
617 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
618 			ETH_VLAN_EXTEND_MASK;
619 	ret = eth_em_vlan_offload_set(dev, mask);
620 	if (ret) {
621 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
622 		em_dev_clear_queues(dev);
623 		return ret;
624 	}
625 
626 	/* Set Interrupt Throttling Rate to maximum allowed value. */
627 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
628 
629 	/* Setup link speed and duplex */
630 	speeds = &dev->data->dev_conf.link_speeds;
631 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
632 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
633 		hw->mac.autoneg = 1;
634 	} else {
635 		num_speeds = 0;
636 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
637 
638 		/* Reset */
639 		hw->phy.autoneg_advertised = 0;
640 
641 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
642 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
643 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
644 			num_speeds = -1;
645 			goto error_invalid_config;
646 		}
647 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
648 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
649 			num_speeds++;
650 		}
651 		if (*speeds & ETH_LINK_SPEED_10M) {
652 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
653 			num_speeds++;
654 		}
655 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
656 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
657 			num_speeds++;
658 		}
659 		if (*speeds & ETH_LINK_SPEED_100M) {
660 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
661 			num_speeds++;
662 		}
663 		if (*speeds & ETH_LINK_SPEED_1G) {
664 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
665 			num_speeds++;
666 		}
667 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
668 			goto error_invalid_config;
669 
670 		/* Set/reset the mac.autoneg based on the link speed,
671 		 * fixed or not
672 		 */
673 		if (!autoneg) {
674 			hw->mac.autoneg = 0;
675 			hw->mac.forced_speed_duplex =
676 					hw->phy.autoneg_advertised;
677 		} else {
678 			hw->mac.autoneg = 1;
679 		}
680 	}
681 
682 	e1000_setup_link(hw);
683 
684 	if (rte_intr_allow_others(intr_handle)) {
685 		/* check if lsc interrupt is enabled */
686 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
687 			ret = eth_em_interrupt_setup(dev);
688 			if (ret) {
689 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
690 				em_dev_clear_queues(dev);
691 				return ret;
692 			}
693 		}
694 	} else {
695 		rte_intr_callback_unregister(intr_handle,
696 						eth_em_interrupt_handler,
697 						(void *)dev);
698 		if (dev->data->dev_conf.intr_conf.lsc != 0)
699 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
700 				     " no intr multiplexn");
701 	}
702 	/* check if rxq interrupt is enabled */
703 	if (dev->data->dev_conf.intr_conf.rxq != 0)
704 		eth_em_rxq_interrupt_setup(dev);
705 
706 	rte_intr_enable(intr_handle);
707 
708 	adapter->stopped = 0;
709 
710 	eth_em_rxtx_control(dev, true);
711 	eth_em_link_update(dev, 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_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
736 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
737 
738 	eth_em_rxtx_control(dev, false);
739 	em_rxq_intr_disable(hw);
740 	em_lsc_intr_disable(hw);
741 
742 	e1000_reset_hw(hw);
743 	if (hw->mac.type >= e1000_82544)
744 		E1000_WRITE_REG(hw, E1000_WUC, 0);
745 
746 	/* Power down the phy. Needed to make the link go down */
747 	e1000_power_down_phy(hw);
748 
749 	em_dev_clear_queues(dev);
750 
751 	/* clear the recorded link status */
752 	memset(&link, 0, sizeof(link));
753 	rte_eth_linkstatus_set(dev, &link);
754 
755 	if (!rte_intr_allow_others(intr_handle))
756 		/* resume to the default handler */
757 		rte_intr_callback_register(intr_handle,
758 					   eth_em_interrupt_handler,
759 					   (void *)dev);
760 
761 	/* Clean datapath event and queue/vec mapping */
762 	rte_intr_efd_disable(intr_handle);
763 	if (intr_handle->intr_vec != NULL) {
764 		rte_free(intr_handle->intr_vec);
765 		intr_handle->intr_vec = NULL;
766 	}
767 }
768 
769 static void
770 eth_em_close(struct rte_eth_dev *dev)
771 {
772 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
773 	struct e1000_adapter *adapter =
774 		E1000_DEV_PRIVATE(dev->data->dev_private);
775 
776 	eth_em_stop(dev);
777 	adapter->stopped = 1;
778 	em_dev_free_queues(dev);
779 	e1000_phy_hw_reset(hw);
780 	em_release_manageability(hw);
781 	em_hw_control_release(hw);
782 }
783 
784 static int
785 em_get_rx_buffer_size(struct e1000_hw *hw)
786 {
787 	uint32_t rx_buf_size;
788 
789 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
790 	return rx_buf_size;
791 }
792 
793 /*********************************************************************
794  *
795  *  Initialize the hardware
796  *
797  **********************************************************************/
798 static int
799 em_hardware_init(struct e1000_hw *hw)
800 {
801 	uint32_t rx_buf_size;
802 	int diag;
803 
804 	/* Issue a global reset */
805 	e1000_reset_hw(hw);
806 
807 	/* Let the firmware know the OS is in control */
808 	em_hw_control_acquire(hw);
809 
810 	/*
811 	 * These parameters control the automatic generation (Tx) and
812 	 * response (Rx) to Ethernet PAUSE frames.
813 	 * - High water mark should allow for at least two standard size (1518)
814 	 *   frames to be received after sending an XOFF.
815 	 * - Low water mark works best when it is very near the high water mark.
816 	 *   This allows the receiver to restart by sending XON when it has
817 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
818 	 *   restart after one full frame is pulled from the buffer. There
819 	 *   could be several smaller frames in the buffer and if so they will
820 	 *   not trigger the XON until their total number reduces the buffer
821 	 *   by 1500.
822 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
823 	 */
824 	rx_buf_size = em_get_rx_buffer_size(hw);
825 
826 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
827 	hw->fc.low_water = hw->fc.high_water - 1500;
828 
829 	if (hw->mac.type == e1000_80003es2lan)
830 		hw->fc.pause_time = UINT16_MAX;
831 	else
832 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
833 
834 	hw->fc.send_xon = 1;
835 
836 	/* Set Flow control, use the tunable location if sane */
837 	if (em_fc_setting <= e1000_fc_full)
838 		hw->fc.requested_mode = em_fc_setting;
839 	else
840 		hw->fc.requested_mode = e1000_fc_none;
841 
842 	/* Workaround: no TX flow ctrl for PCH */
843 	if (hw->mac.type == e1000_pchlan)
844 		hw->fc.requested_mode = e1000_fc_rx_pause;
845 
846 	/* Override - settings for PCH2LAN, ya its magic :) */
847 	if (hw->mac.type == e1000_pch2lan) {
848 		hw->fc.high_water = 0x5C20;
849 		hw->fc.low_water = 0x5048;
850 		hw->fc.pause_time = 0x0650;
851 		hw->fc.refresh_time = 0x0400;
852 	} else if (hw->mac.type == e1000_pch_lpt ||
853 		   hw->mac.type == e1000_pch_spt ||
854 		   hw->mac.type == e1000_pch_cnp) {
855 		hw->fc.requested_mode = e1000_fc_full;
856 	}
857 
858 	diag = e1000_init_hw(hw);
859 	if (diag < 0)
860 		return diag;
861 	e1000_check_for_link(hw);
862 	return 0;
863 }
864 
865 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
866 static int
867 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
868 {
869 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
870 	struct e1000_hw_stats *stats =
871 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
872 	int pause_frames;
873 
874 	if(hw->phy.media_type == e1000_media_type_copper ||
875 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
876 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
877 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
878 	}
879 
880 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
881 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
882 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
883 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
884 
885 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
886 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
887 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
888 	stats->dc += E1000_READ_REG(hw, E1000_DC);
889 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
890 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
891 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
892 
893 	/*
894 	 * For watchdog management we need to know if we have been
895 	 * paused during the last interval, so capture that here.
896 	 */
897 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
898 	stats->xoffrxc += pause_frames;
899 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
900 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
901 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
902 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
903 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
904 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
905 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
906 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
907 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
908 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
909 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
910 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
911 
912 	/*
913 	 * For the 64-bit byte counters the low dword must be read first.
914 	 * Both registers clear on the read of the high dword.
915 	 */
916 
917 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
918 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
919 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
920 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
921 
922 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
923 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
924 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
925 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
926 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
927 
928 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
929 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
930 
931 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
932 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
933 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
934 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
935 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
936 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
937 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
938 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
939 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
940 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
941 
942 	/* Interrupt Counts */
943 
944 	if (hw->mac.type >= e1000_82571) {
945 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
946 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
947 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
948 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
949 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
950 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
951 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
952 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
953 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
954 	}
955 
956 	if (hw->mac.type >= e1000_82543) {
957 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
958 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
959 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
960 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
961 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
962 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
963 	}
964 
965 	if (rte_stats == NULL)
966 		return -EINVAL;
967 
968 	/* Rx Errors */
969 	rte_stats->imissed = stats->mpc;
970 	rte_stats->ierrors = stats->crcerrs +
971 	                     stats->rlec + stats->ruc + stats->roc +
972 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
973 
974 	/* Tx Errors */
975 	rte_stats->oerrors = stats->ecol + stats->latecol;
976 
977 	rte_stats->ipackets = stats->gprc;
978 	rte_stats->opackets = stats->gptc;
979 	rte_stats->ibytes   = stats->gorc;
980 	rte_stats->obytes   = stats->gotc;
981 	return 0;
982 }
983 
984 static void
985 eth_em_stats_reset(struct rte_eth_dev *dev)
986 {
987 	struct e1000_hw_stats *hw_stats =
988 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
989 
990 	/* HW registers are cleared on read */
991 	eth_em_stats_get(dev, NULL);
992 
993 	/* Reset software totals */
994 	memset(hw_stats, 0, sizeof(*hw_stats));
995 }
996 
997 static int
998 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
999 {
1000 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1001 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1002 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1003 
1004 	em_rxq_intr_enable(hw);
1005 	rte_intr_enable(intr_handle);
1006 
1007 	return 0;
1008 }
1009 
1010 static int
1011 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1012 {
1013 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1014 
1015 	em_rxq_intr_disable(hw);
1016 
1017 	return 0;
1018 }
1019 
1020 uint32_t
1021 em_get_max_pktlen(struct rte_eth_dev *dev)
1022 {
1023 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1024 
1025 	switch (hw->mac.type) {
1026 	case e1000_82571:
1027 	case e1000_82572:
1028 	case e1000_ich9lan:
1029 	case e1000_ich10lan:
1030 	case e1000_pch2lan:
1031 	case e1000_pch_lpt:
1032 	case e1000_pch_spt:
1033 	case e1000_pch_cnp:
1034 	case e1000_82574:
1035 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1036 	case e1000_82583:
1037 		return 0x2412;
1038 	case e1000_pchlan:
1039 		return 0x1000;
1040 	/* Adapters that do not support jumbo frames */
1041 	case e1000_ich8lan:
1042 		return ETHER_MAX_LEN;
1043 	default:
1044 		return MAX_JUMBO_FRAME_SIZE;
1045 	}
1046 }
1047 
1048 static void
1049 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1050 {
1051 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1052 
1053 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1054 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1055 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1056 
1057 	/*
1058 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1059 	 * Unfortunatelly, all these nics have just one TX context.
1060 	 * So we have few choises for TX:
1061 	 * - Use just one TX queue.
1062 	 * - Allow cksum offload only for one TX queue.
1063 	 * - Don't allow TX cksum offload at all.
1064 	 * For now, option #1 was chosen.
1065 	 * To use second RX queue we have to use extended RX descriptor
1066 	 * (Multiple Receive Queues are mutually exclusive with UDP
1067 	 * fragmentation and are not supported when a legacy receive
1068 	 * descriptor format is used).
1069 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1070 	 * don't support extended RXD.
1071 	 * To avoid it we support just one RX queue for now (no RSS).
1072 	 */
1073 
1074 	dev_info->max_rx_queues = 1;
1075 	dev_info->max_tx_queues = 1;
1076 
1077 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1078 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1079 				    dev_info->rx_queue_offload_capa;
1080 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1081 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1082 				    dev_info->tx_queue_offload_capa;
1083 
1084 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1085 		.nb_max = E1000_MAX_RING_DESC,
1086 		.nb_min = E1000_MIN_RING_DESC,
1087 		.nb_align = EM_RXD_ALIGN,
1088 	};
1089 
1090 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1091 		.nb_max = E1000_MAX_RING_DESC,
1092 		.nb_min = E1000_MIN_RING_DESC,
1093 		.nb_align = EM_TXD_ALIGN,
1094 		.nb_seg_max = EM_TX_MAX_SEG,
1095 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1096 	};
1097 
1098 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1099 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1100 			ETH_LINK_SPEED_1G;
1101 
1102 	/* Preferred queue parameters */
1103 	dev_info->default_rxportconf.nb_queues = 1;
1104 	dev_info->default_txportconf.nb_queues = 1;
1105 	dev_info->default_txportconf.ring_size = 256;
1106 	dev_info->default_rxportconf.ring_size = 256;
1107 }
1108 
1109 /* return 0 means link status changed, -1 means not changed */
1110 static int
1111 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1112 {
1113 	struct e1000_hw *hw =
1114 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1115 	struct rte_eth_link link;
1116 	int link_check, count;
1117 
1118 	link_check = 0;
1119 	hw->mac.get_link_status = 1;
1120 
1121 	/* possible wait-to-complete in up to 9 seconds */
1122 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1123 		/* Read the real link status */
1124 		switch (hw->phy.media_type) {
1125 		case e1000_media_type_copper:
1126 			/* Do the work to read phy */
1127 			e1000_check_for_link(hw);
1128 			link_check = !hw->mac.get_link_status;
1129 			break;
1130 
1131 		case e1000_media_type_fiber:
1132 			e1000_check_for_link(hw);
1133 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1134 					E1000_STATUS_LU);
1135 			break;
1136 
1137 		case e1000_media_type_internal_serdes:
1138 			e1000_check_for_link(hw);
1139 			link_check = hw->mac.serdes_has_link;
1140 			break;
1141 
1142 		default:
1143 			break;
1144 		}
1145 		if (link_check || wait_to_complete == 0)
1146 			break;
1147 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1148 	}
1149 	memset(&link, 0, sizeof(link));
1150 
1151 	/* Now we check if a transition has happened */
1152 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1153 		uint16_t duplex, speed;
1154 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1155 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1156 				ETH_LINK_FULL_DUPLEX :
1157 				ETH_LINK_HALF_DUPLEX;
1158 		link.link_speed = speed;
1159 		link.link_status = ETH_LINK_UP;
1160 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1161 				ETH_LINK_SPEED_FIXED);
1162 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1163 		link.link_speed = ETH_SPEED_NUM_NONE;
1164 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1165 		link.link_status = ETH_LINK_DOWN;
1166 		link.link_autoneg = ETH_LINK_FIXED;
1167 	}
1168 
1169 	return rte_eth_linkstatus_set(dev, &link);
1170 }
1171 
1172 /*
1173  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1174  * For ASF and Pass Through versions of f/w this means
1175  * that the driver is loaded. For AMT version type f/w
1176  * this means that the network i/f is open.
1177  */
1178 static void
1179 em_hw_control_acquire(struct e1000_hw *hw)
1180 {
1181 	uint32_t ctrl_ext, swsm;
1182 
1183 	/* Let firmware know the driver has taken over */
1184 	if (hw->mac.type == e1000_82573) {
1185 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1186 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1187 
1188 	} else {
1189 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1190 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1191 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1192 	}
1193 }
1194 
1195 /*
1196  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1197  * For ASF and Pass Through versions of f/w this means that the
1198  * driver is no longer loaded. For AMT versions of the
1199  * f/w this means that the network i/f is closed.
1200  */
1201 static void
1202 em_hw_control_release(struct e1000_hw *hw)
1203 {
1204 	uint32_t ctrl_ext, swsm;
1205 
1206 	/* Let firmware taken over control of h/w */
1207 	if (hw->mac.type == e1000_82573) {
1208 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1209 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1210 	} else {
1211 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1212 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1213 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1214 	}
1215 }
1216 
1217 /*
1218  * Bit of a misnomer, what this really means is
1219  * to enable OS management of the system... aka
1220  * to disable special hardware management features.
1221  */
1222 static void
1223 em_init_manageability(struct e1000_hw *hw)
1224 {
1225 	if (e1000_enable_mng_pass_thru(hw)) {
1226 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1227 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1228 
1229 		/* disable hardware interception of ARP */
1230 		manc &= ~(E1000_MANC_ARP_EN);
1231 
1232 		/* enable receiving management packets to the host */
1233 		manc |= E1000_MANC_EN_MNG2HOST;
1234 		manc2h |= 1 << 5;  /* Mng Port 623 */
1235 		manc2h |= 1 << 6;  /* Mng Port 664 */
1236 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1237 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1238 	}
1239 }
1240 
1241 /*
1242  * Give control back to hardware management
1243  * controller if there is one.
1244  */
1245 static void
1246 em_release_manageability(struct e1000_hw *hw)
1247 {
1248 	uint32_t manc;
1249 
1250 	if (e1000_enable_mng_pass_thru(hw)) {
1251 		manc = E1000_READ_REG(hw, E1000_MANC);
1252 
1253 		/* re-enable hardware interception of ARP */
1254 		manc |= E1000_MANC_ARP_EN;
1255 		manc &= ~E1000_MANC_EN_MNG2HOST;
1256 
1257 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1258 	}
1259 }
1260 
1261 static void
1262 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1263 {
1264 	struct e1000_hw *hw =
1265 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1266 	uint32_t rctl;
1267 
1268 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1269 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1270 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1271 }
1272 
1273 static void
1274 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1275 {
1276 	struct e1000_hw *hw =
1277 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1278 	uint32_t rctl;
1279 
1280 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1281 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1282 	if (dev->data->all_multicast == 1)
1283 		rctl |= E1000_RCTL_MPE;
1284 	else
1285 		rctl &= (~E1000_RCTL_MPE);
1286 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1287 }
1288 
1289 static void
1290 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1291 {
1292 	struct e1000_hw *hw =
1293 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1294 	uint32_t rctl;
1295 
1296 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1297 	rctl |= E1000_RCTL_MPE;
1298 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1299 }
1300 
1301 static void
1302 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1303 {
1304 	struct e1000_hw *hw =
1305 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1306 	uint32_t rctl;
1307 
1308 	if (dev->data->promiscuous == 1)
1309 		return; /* must remain in all_multicast mode */
1310 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1311 	rctl &= (~E1000_RCTL_MPE);
1312 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1313 }
1314 
1315 static int
1316 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1317 {
1318 	struct e1000_hw *hw =
1319 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1320 	struct e1000_vfta * shadow_vfta =
1321 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1322 	uint32_t vfta;
1323 	uint32_t vid_idx;
1324 	uint32_t vid_bit;
1325 
1326 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1327 			      E1000_VFTA_ENTRY_MASK);
1328 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1329 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1330 	if (on)
1331 		vfta |= vid_bit;
1332 	else
1333 		vfta &= ~vid_bit;
1334 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1335 
1336 	/* update local VFTA copy */
1337 	shadow_vfta->vfta[vid_idx] = vfta;
1338 
1339 	return 0;
1340 }
1341 
1342 static void
1343 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1344 {
1345 	struct e1000_hw *hw =
1346 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1347 	uint32_t reg;
1348 
1349 	/* Filter Table Disable */
1350 	reg = E1000_READ_REG(hw, E1000_RCTL);
1351 	reg &= ~E1000_RCTL_CFIEN;
1352 	reg &= ~E1000_RCTL_VFE;
1353 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1354 }
1355 
1356 static void
1357 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1358 {
1359 	struct e1000_hw *hw =
1360 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1361 	struct e1000_vfta * shadow_vfta =
1362 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1363 	uint32_t reg;
1364 	int i;
1365 
1366 	/* Filter Table Enable, CFI not used for packet acceptance */
1367 	reg = E1000_READ_REG(hw, E1000_RCTL);
1368 	reg &= ~E1000_RCTL_CFIEN;
1369 	reg |= E1000_RCTL_VFE;
1370 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1371 
1372 	/* restore vfta from local copy */
1373 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1374 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1375 }
1376 
1377 static void
1378 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1379 {
1380 	struct e1000_hw *hw =
1381 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1382 	uint32_t reg;
1383 
1384 	/* VLAN Mode Disable */
1385 	reg = E1000_READ_REG(hw, E1000_CTRL);
1386 	reg &= ~E1000_CTRL_VME;
1387 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1388 
1389 }
1390 
1391 static void
1392 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1393 {
1394 	struct e1000_hw *hw =
1395 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1396 	uint32_t reg;
1397 
1398 	/* VLAN Mode Enable */
1399 	reg = E1000_READ_REG(hw, E1000_CTRL);
1400 	reg |= E1000_CTRL_VME;
1401 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1402 }
1403 
1404 static int
1405 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1406 {
1407 	struct rte_eth_rxmode *rxmode;
1408 
1409 	rxmode = &dev->data->dev_conf.rxmode;
1410 	if(mask & ETH_VLAN_STRIP_MASK){
1411 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1412 			em_vlan_hw_strip_enable(dev);
1413 		else
1414 			em_vlan_hw_strip_disable(dev);
1415 	}
1416 
1417 	if(mask & ETH_VLAN_FILTER_MASK){
1418 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1419 			em_vlan_hw_filter_enable(dev);
1420 		else
1421 			em_vlan_hw_filter_disable(dev);
1422 	}
1423 
1424 	return 0;
1425 }
1426 
1427 /*
1428  * It enables the interrupt mask and then enable the interrupt.
1429  *
1430  * @param dev
1431  *  Pointer to struct rte_eth_dev.
1432  *
1433  * @return
1434  *  - On success, zero.
1435  *  - On failure, a negative value.
1436  */
1437 static int
1438 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1439 {
1440 	uint32_t regval;
1441 	struct e1000_hw *hw =
1442 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1443 
1444 	/* clear interrupt */
1445 	E1000_READ_REG(hw, E1000_ICR);
1446 	regval = E1000_READ_REG(hw, E1000_IMS);
1447 	E1000_WRITE_REG(hw, E1000_IMS, regval | E1000_ICR_LSC);
1448 	return 0;
1449 }
1450 
1451 /*
1452  * It clears the interrupt causes and enables the interrupt.
1453  * It will be called once only during nic initialized.
1454  *
1455  * @param dev
1456  *  Pointer to struct rte_eth_dev.
1457  *
1458  * @return
1459  *  - On success, zero.
1460  *  - On failure, a negative value.
1461  */
1462 static int
1463 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1464 {
1465 	struct e1000_hw *hw =
1466 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1467 
1468 	E1000_READ_REG(hw, E1000_ICR);
1469 	em_rxq_intr_enable(hw);
1470 	return 0;
1471 }
1472 
1473 /*
1474  * It enable receive packet interrupt.
1475  * @param hw
1476  * Pointer to struct e1000_hw
1477  *
1478  * @return
1479  */
1480 static void
1481 em_rxq_intr_enable(struct e1000_hw *hw)
1482 {
1483 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1484 	E1000_WRITE_FLUSH(hw);
1485 }
1486 
1487 /*
1488  * It disabled lsc interrupt.
1489  * @param hw
1490  * Pointer to struct e1000_hw
1491  *
1492  * @return
1493  */
1494 static void
1495 em_lsc_intr_disable(struct e1000_hw *hw)
1496 {
1497 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC);
1498 	E1000_WRITE_FLUSH(hw);
1499 }
1500 
1501 /*
1502  * It disabled receive packet interrupt.
1503  * @param hw
1504  * Pointer to struct e1000_hw
1505  *
1506  * @return
1507  */
1508 static void
1509 em_rxq_intr_disable(struct e1000_hw *hw)
1510 {
1511 	E1000_READ_REG(hw, E1000_ICR);
1512 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1513 	E1000_WRITE_FLUSH(hw);
1514 }
1515 
1516 /*
1517  * It reads ICR and gets interrupt causes, check it and set a bit flag
1518  * to update link status.
1519  *
1520  * @param dev
1521  *  Pointer to struct rte_eth_dev.
1522  *
1523  * @return
1524  *  - On success, zero.
1525  *  - On failure, a negative value.
1526  */
1527 static int
1528 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1529 {
1530 	uint32_t icr;
1531 	struct e1000_hw *hw =
1532 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1533 	struct e1000_interrupt *intr =
1534 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1535 
1536 	/* read-on-clear nic registers here */
1537 	icr = E1000_READ_REG(hw, E1000_ICR);
1538 	if (icr & E1000_ICR_LSC) {
1539 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1540 	}
1541 
1542 	return 0;
1543 }
1544 
1545 /*
1546  * It executes link_update after knowing an interrupt is prsent.
1547  *
1548  * @param dev
1549  *  Pointer to struct rte_eth_dev.
1550  *
1551  * @return
1552  *  - On success, zero.
1553  *  - On failure, a negative value.
1554  */
1555 static int
1556 eth_em_interrupt_action(struct rte_eth_dev *dev,
1557 			struct rte_intr_handle *intr_handle)
1558 {
1559 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1560 	struct e1000_hw *hw =
1561 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1562 	struct e1000_interrupt *intr =
1563 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1564 	struct rte_eth_link link;
1565 	int ret;
1566 
1567 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1568 		return -1;
1569 
1570 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1571 	rte_intr_enable(intr_handle);
1572 
1573 	/* set get_link_status to check register later */
1574 	hw->mac.get_link_status = 1;
1575 	ret = eth_em_link_update(dev, 0);
1576 
1577 	/* check if link has changed */
1578 	if (ret < 0)
1579 		return 0;
1580 
1581 	rte_eth_linkstatus_get(dev, &link);
1582 
1583 	if (link.link_status) {
1584 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1585 			     dev->data->port_id, link.link_speed,
1586 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1587 			     "full-duplex" : "half-duplex");
1588 	} else {
1589 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1590 	}
1591 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1592 		     pci_dev->addr.domain, pci_dev->addr.bus,
1593 		     pci_dev->addr.devid, pci_dev->addr.function);
1594 
1595 	return 0;
1596 }
1597 
1598 /**
1599  * Interrupt handler which shall be registered at first.
1600  *
1601  * @param handle
1602  *  Pointer to interrupt handle.
1603  * @param param
1604  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1605  *
1606  * @return
1607  *  void
1608  */
1609 static void
1610 eth_em_interrupt_handler(void *param)
1611 {
1612 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1613 
1614 	eth_em_interrupt_get_status(dev);
1615 	eth_em_interrupt_action(dev, dev->intr_handle);
1616 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1617 }
1618 
1619 static int
1620 eth_em_led_on(struct rte_eth_dev *dev)
1621 {
1622 	struct e1000_hw *hw;
1623 
1624 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1625 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1626 }
1627 
1628 static int
1629 eth_em_led_off(struct rte_eth_dev *dev)
1630 {
1631 	struct e1000_hw *hw;
1632 
1633 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1634 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1635 }
1636 
1637 static int
1638 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1639 {
1640 	struct e1000_hw *hw;
1641 	uint32_t ctrl;
1642 	int tx_pause;
1643 	int rx_pause;
1644 
1645 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1646 	fc_conf->pause_time = hw->fc.pause_time;
1647 	fc_conf->high_water = hw->fc.high_water;
1648 	fc_conf->low_water = hw->fc.low_water;
1649 	fc_conf->send_xon = hw->fc.send_xon;
1650 	fc_conf->autoneg = hw->mac.autoneg;
1651 
1652 	/*
1653 	 * Return rx_pause and tx_pause status according to actual setting of
1654 	 * the TFCE and RFCE bits in the CTRL register.
1655 	 */
1656 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1657 	if (ctrl & E1000_CTRL_TFCE)
1658 		tx_pause = 1;
1659 	else
1660 		tx_pause = 0;
1661 
1662 	if (ctrl & E1000_CTRL_RFCE)
1663 		rx_pause = 1;
1664 	else
1665 		rx_pause = 0;
1666 
1667 	if (rx_pause && tx_pause)
1668 		fc_conf->mode = RTE_FC_FULL;
1669 	else if (rx_pause)
1670 		fc_conf->mode = RTE_FC_RX_PAUSE;
1671 	else if (tx_pause)
1672 		fc_conf->mode = RTE_FC_TX_PAUSE;
1673 	else
1674 		fc_conf->mode = RTE_FC_NONE;
1675 
1676 	return 0;
1677 }
1678 
1679 static int
1680 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1681 {
1682 	struct e1000_hw *hw;
1683 	int err;
1684 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1685 		e1000_fc_none,
1686 		e1000_fc_rx_pause,
1687 		e1000_fc_tx_pause,
1688 		e1000_fc_full
1689 	};
1690 	uint32_t rx_buf_size;
1691 	uint32_t max_high_water;
1692 	uint32_t rctl;
1693 
1694 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1695 	if (fc_conf->autoneg != hw->mac.autoneg)
1696 		return -ENOTSUP;
1697 	rx_buf_size = em_get_rx_buffer_size(hw);
1698 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1699 
1700 	/* At least reserve one Ethernet frame for watermark */
1701 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1702 	if ((fc_conf->high_water > max_high_water) ||
1703 	    (fc_conf->high_water < fc_conf->low_water)) {
1704 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1705 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1706 		return -EINVAL;
1707 	}
1708 
1709 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1710 	hw->fc.pause_time     = fc_conf->pause_time;
1711 	hw->fc.high_water     = fc_conf->high_water;
1712 	hw->fc.low_water      = fc_conf->low_water;
1713 	hw->fc.send_xon	      = fc_conf->send_xon;
1714 
1715 	err = e1000_setup_link_generic(hw);
1716 	if (err == E1000_SUCCESS) {
1717 
1718 		/* check if we want to forward MAC frames - driver doesn't have native
1719 		 * capability to do that, so we'll write the registers ourselves */
1720 
1721 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1722 
1723 		/* set or clear MFLCN.PMCF bit depending on configuration */
1724 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1725 			rctl |= E1000_RCTL_PMCF;
1726 		else
1727 			rctl &= ~E1000_RCTL_PMCF;
1728 
1729 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1730 		E1000_WRITE_FLUSH(hw);
1731 
1732 		return 0;
1733 	}
1734 
1735 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1736 	return -EIO;
1737 }
1738 
1739 static int
1740 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1741 		uint32_t index, __rte_unused uint32_t pool)
1742 {
1743 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1744 
1745 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1746 }
1747 
1748 static void
1749 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1750 {
1751 	uint8_t addr[ETHER_ADDR_LEN];
1752 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1753 
1754 	memset(addr, 0, sizeof(addr));
1755 
1756 	e1000_rar_set(hw, addr, index);
1757 }
1758 
1759 static int
1760 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1761 			    struct ether_addr *addr)
1762 {
1763 	eth_em_rar_clear(dev, 0);
1764 
1765 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1766 }
1767 
1768 static int
1769 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1770 {
1771 	struct rte_eth_dev_info dev_info;
1772 	struct e1000_hw *hw;
1773 	uint32_t frame_size;
1774 	uint32_t rctl;
1775 
1776 	eth_em_infos_get(dev, &dev_info);
1777 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1778 
1779 	/* check that mtu is within the allowed range */
1780 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1781 		return -EINVAL;
1782 
1783 	/* refuse mtu that requires the support of scattered packets when this
1784 	 * feature has not been enabled before. */
1785 	if (!dev->data->scattered_rx &&
1786 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1787 		return -EINVAL;
1788 
1789 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1790 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1791 
1792 	/* switch to jumbo mode if needed */
1793 	if (frame_size > ETHER_MAX_LEN) {
1794 		dev->data->dev_conf.rxmode.offloads |=
1795 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1796 		rctl |= E1000_RCTL_LPE;
1797 	} else {
1798 		dev->data->dev_conf.rxmode.offloads &=
1799 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1800 		rctl &= ~E1000_RCTL_LPE;
1801 	}
1802 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1803 
1804 	/* update max frame size */
1805 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1806 	return 0;
1807 }
1808 
1809 static int
1810 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1811 			struct ether_addr *mc_addr_set,
1812 			uint32_t nb_mc_addr)
1813 {
1814 	struct e1000_hw *hw;
1815 
1816 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1817 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1818 	return 0;
1819 }
1820 
1821 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1822 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1823 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1824 
1825 /* see e1000_logs.c */
1826 RTE_INIT(igb_init_log)
1827 {
1828 	e1000_igb_init_log();
1829 }
1830