xref: /f-stack/dpdk/drivers/net/e1000/em_ethdev.c (revision 552bc48c)
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 	/* disable uio intr before callback unregister */
333 	rte_intr_disable(intr_handle);
334 	rte_intr_callback_unregister(intr_handle,
335 				     eth_em_interrupt_handler, eth_dev);
336 
337 	return 0;
338 }
339 
340 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
341 	struct rte_pci_device *pci_dev)
342 {
343 	return rte_eth_dev_pci_generic_probe(pci_dev,
344 		sizeof(struct e1000_adapter), eth_em_dev_init);
345 }
346 
347 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
348 {
349 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
350 }
351 
352 static struct rte_pci_driver rte_em_pmd = {
353 	.id_table = pci_id_em_map,
354 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC |
355 		     RTE_PCI_DRV_IOVA_AS_VA,
356 	.probe = eth_em_pci_probe,
357 	.remove = eth_em_pci_remove,
358 };
359 
360 static int
361 em_hw_init(struct e1000_hw *hw)
362 {
363 	int diag;
364 
365 	diag = hw->mac.ops.init_params(hw);
366 	if (diag != 0) {
367 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
368 		return diag;
369 	}
370 	diag = hw->nvm.ops.init_params(hw);
371 	if (diag != 0) {
372 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
373 		return diag;
374 	}
375 	diag = hw->phy.ops.init_params(hw);
376 	if (diag != 0) {
377 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
378 		return diag;
379 	}
380 	(void) e1000_get_bus_info(hw);
381 
382 	hw->mac.autoneg = 1;
383 	hw->phy.autoneg_wait_to_complete = 0;
384 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
385 
386 	e1000_init_script_state_82541(hw, TRUE);
387 	e1000_set_tbi_compatibility_82543(hw, TRUE);
388 
389 	/* Copper options */
390 	if (hw->phy.media_type == e1000_media_type_copper) {
391 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
392 		hw->phy.disable_polarity_correction = 0;
393 		hw->phy.ms_type = e1000_ms_hw_default;
394 	}
395 
396 	/*
397 	 * Start from a known state, this is important in reading the nvm
398 	 * and mac from that.
399 	 */
400 	e1000_reset_hw(hw);
401 
402 	/* Make sure we have a good EEPROM before we read from it */
403 	if (e1000_validate_nvm_checksum(hw) < 0) {
404 		/*
405 		 * Some PCI-E parts fail the first check due to
406 		 * the link being in sleep state, call it again,
407 		 * if it fails a second time its a real issue.
408 		 */
409 		diag = e1000_validate_nvm_checksum(hw);
410 		if (diag < 0) {
411 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
412 			goto error;
413 		}
414 	}
415 
416 	/* Read the permanent MAC address out of the EEPROM */
417 	diag = e1000_read_mac_addr(hw);
418 	if (diag != 0) {
419 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
420 		goto error;
421 	}
422 
423 	/* Now initialize the hardware */
424 	diag = em_hardware_init(hw);
425 	if (diag != 0) {
426 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
427 		goto error;
428 	}
429 
430 	hw->mac.get_link_status = 1;
431 
432 	/* Indicate SOL/IDER usage */
433 	diag = e1000_check_reset_block(hw);
434 	if (diag < 0) {
435 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
436 			"SOL/IDER session");
437 	}
438 	return 0;
439 
440 error:
441 	em_hw_control_release(hw);
442 	return diag;
443 }
444 
445 static int
446 eth_em_configure(struct rte_eth_dev *dev)
447 {
448 	struct e1000_interrupt *intr =
449 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
450 
451 	PMD_INIT_FUNC_TRACE();
452 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
453 
454 	PMD_INIT_FUNC_TRACE();
455 
456 	return 0;
457 }
458 
459 static void
460 em_set_pba(struct e1000_hw *hw)
461 {
462 	uint32_t pba;
463 
464 	/*
465 	 * Packet Buffer Allocation (PBA)
466 	 * Writing PBA sets the receive portion of the buffer
467 	 * the remainder is used for the transmit buffer.
468 	 * Devices before the 82547 had a Packet Buffer of 64K.
469 	 * After the 82547 the buffer was reduced to 40K.
470 	 */
471 	switch (hw->mac.type) {
472 		case e1000_82547:
473 		case e1000_82547_rev_2:
474 		/* 82547: Total Packet Buffer is 40K */
475 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
476 			break;
477 		case e1000_82571:
478 		case e1000_82572:
479 		case e1000_80003es2lan:
480 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
481 			break;
482 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
483 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
484 			break;
485 		case e1000_82574:
486 		case e1000_82583:
487 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
488 			break;
489 		case e1000_ich8lan:
490 			pba = E1000_PBA_8K;
491 			break;
492 		case e1000_ich9lan:
493 		case e1000_ich10lan:
494 			pba = E1000_PBA_10K;
495 			break;
496 		case e1000_pchlan:
497 		case e1000_pch2lan:
498 		case e1000_pch_lpt:
499 		case e1000_pch_spt:
500 		case e1000_pch_cnp:
501 			pba = E1000_PBA_26K;
502 			break;
503 		default:
504 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
505 	}
506 
507 	E1000_WRITE_REG(hw, E1000_PBA, pba);
508 }
509 
510 static void
511 eth_em_rxtx_control(struct rte_eth_dev *dev,
512 		    bool enable)
513 {
514 	struct e1000_hw *hw =
515 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
516 	uint32_t tctl, rctl;
517 
518 	tctl = E1000_READ_REG(hw, E1000_TCTL);
519 	rctl = E1000_READ_REG(hw, E1000_RCTL);
520 	if (enable) {
521 		/* enable Tx/Rx */
522 		tctl |= E1000_TCTL_EN;
523 		rctl |= E1000_RCTL_EN;
524 	} else {
525 		/* disable Tx/Rx */
526 		tctl &= ~E1000_TCTL_EN;
527 		rctl &= ~E1000_RCTL_EN;
528 	}
529 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
530 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
531 	E1000_WRITE_FLUSH(hw);
532 }
533 
534 static int
535 eth_em_start(struct rte_eth_dev *dev)
536 {
537 	struct e1000_adapter *adapter =
538 		E1000_DEV_PRIVATE(dev->data->dev_private);
539 	struct e1000_hw *hw =
540 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
541 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
542 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
543 	int ret, mask;
544 	uint32_t intr_vector = 0;
545 	uint32_t *speeds;
546 	int num_speeds;
547 	bool autoneg;
548 
549 	PMD_INIT_FUNC_TRACE();
550 
551 	eth_em_stop(dev);
552 
553 	e1000_power_up_phy(hw);
554 
555 	/* Set default PBA value */
556 	em_set_pba(hw);
557 
558 	/* Put the address into the Receive Address Array */
559 	e1000_rar_set(hw, hw->mac.addr, 0);
560 
561 	/*
562 	 * With the 82571 adapter, RAR[0] may be overwritten
563 	 * when the other port is reset, we make a duplicate
564 	 * in RAR[14] for that eventuality, this assures
565 	 * the interface continues to function.
566 	 */
567 	if (hw->mac.type == e1000_82571) {
568 		e1000_set_laa_state_82571(hw, TRUE);
569 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
570 	}
571 
572 	/* Initialize the hardware */
573 	if (em_hardware_init(hw)) {
574 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
575 		return -EIO;
576 	}
577 
578 	E1000_WRITE_REG(hw, E1000_VET, ETHER_TYPE_VLAN);
579 
580 	/* Configure for OS presence */
581 	em_init_manageability(hw);
582 
583 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
584 		intr_vector = dev->data->nb_rx_queues;
585 		if (rte_intr_efd_enable(intr_handle, intr_vector))
586 			return -1;
587 	}
588 
589 	if (rte_intr_dp_is_en(intr_handle)) {
590 		intr_handle->intr_vec =
591 			rte_zmalloc("intr_vec",
592 					dev->data->nb_rx_queues * sizeof(int), 0);
593 		if (intr_handle->intr_vec == NULL) {
594 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
595 						" intr_vec", dev->data->nb_rx_queues);
596 			return -ENOMEM;
597 		}
598 
599 		/* enable rx interrupt */
600 		em_rxq_intr_enable(hw);
601 	}
602 
603 	eth_em_tx_init(dev);
604 
605 	ret = eth_em_rx_init(dev);
606 	if (ret) {
607 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
608 		em_dev_clear_queues(dev);
609 		return ret;
610 	}
611 
612 	e1000_clear_hw_cntrs_base_generic(hw);
613 
614 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
615 			ETH_VLAN_EXTEND_MASK;
616 	ret = eth_em_vlan_offload_set(dev, mask);
617 	if (ret) {
618 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
619 		em_dev_clear_queues(dev);
620 		return ret;
621 	}
622 
623 	/* Set Interrupt Throttling Rate to maximum allowed value. */
624 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
625 
626 	/* Setup link speed and duplex */
627 	speeds = &dev->data->dev_conf.link_speeds;
628 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
629 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
630 		hw->mac.autoneg = 1;
631 	} else {
632 		num_speeds = 0;
633 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
634 
635 		/* Reset */
636 		hw->phy.autoneg_advertised = 0;
637 
638 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
639 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
640 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
641 			num_speeds = -1;
642 			goto error_invalid_config;
643 		}
644 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
645 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
646 			num_speeds++;
647 		}
648 		if (*speeds & ETH_LINK_SPEED_10M) {
649 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
650 			num_speeds++;
651 		}
652 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
653 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
654 			num_speeds++;
655 		}
656 		if (*speeds & ETH_LINK_SPEED_100M) {
657 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
658 			num_speeds++;
659 		}
660 		if (*speeds & ETH_LINK_SPEED_1G) {
661 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
662 			num_speeds++;
663 		}
664 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
665 			goto error_invalid_config;
666 
667 		/* Set/reset the mac.autoneg based on the link speed,
668 		 * fixed or not
669 		 */
670 		if (!autoneg) {
671 			hw->mac.autoneg = 0;
672 			hw->mac.forced_speed_duplex =
673 					hw->phy.autoneg_advertised;
674 		} else {
675 			hw->mac.autoneg = 1;
676 		}
677 	}
678 
679 	e1000_setup_link(hw);
680 
681 	if (rte_intr_allow_others(intr_handle)) {
682 		/* check if lsc interrupt is enabled */
683 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
684 			ret = eth_em_interrupt_setup(dev);
685 			if (ret) {
686 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
687 				em_dev_clear_queues(dev);
688 				return ret;
689 			}
690 		}
691 	} else {
692 		rte_intr_callback_unregister(intr_handle,
693 						eth_em_interrupt_handler,
694 						(void *)dev);
695 		if (dev->data->dev_conf.intr_conf.lsc != 0)
696 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
697 				     " no intr multiplexn");
698 	}
699 	/* check if rxq interrupt is enabled */
700 	if (dev->data->dev_conf.intr_conf.rxq != 0)
701 		eth_em_rxq_interrupt_setup(dev);
702 
703 	rte_intr_enable(intr_handle);
704 
705 	adapter->stopped = 0;
706 
707 	eth_em_rxtx_control(dev, true);
708 	eth_em_link_update(dev, 0);
709 
710 	PMD_INIT_LOG(DEBUG, "<<");
711 
712 	return 0;
713 
714 error_invalid_config:
715 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
716 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
717 	em_dev_clear_queues(dev);
718 	return -EINVAL;
719 }
720 
721 /*********************************************************************
722  *
723  *  This routine disables all traffic on the adapter by issuing a
724  *  global reset on the MAC.
725  *
726  **********************************************************************/
727 static void
728 eth_em_stop(struct rte_eth_dev *dev)
729 {
730 	struct rte_eth_link link;
731 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
732 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
733 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
734 
735 	eth_em_rxtx_control(dev, false);
736 	em_rxq_intr_disable(hw);
737 	em_lsc_intr_disable(hw);
738 
739 	e1000_reset_hw(hw);
740 	if (hw->mac.type >= e1000_82544)
741 		E1000_WRITE_REG(hw, E1000_WUC, 0);
742 
743 	/* Power down the phy. Needed to make the link go down */
744 	e1000_power_down_phy(hw);
745 
746 	em_dev_clear_queues(dev);
747 
748 	/* clear the recorded link status */
749 	memset(&link, 0, sizeof(link));
750 	rte_eth_linkstatus_set(dev, &link);
751 
752 	if (!rte_intr_allow_others(intr_handle))
753 		/* resume to the default handler */
754 		rte_intr_callback_register(intr_handle,
755 					   eth_em_interrupt_handler,
756 					   (void *)dev);
757 
758 	/* Clean datapath event and queue/vec mapping */
759 	rte_intr_efd_disable(intr_handle);
760 	if (intr_handle->intr_vec != NULL) {
761 		rte_free(intr_handle->intr_vec);
762 		intr_handle->intr_vec = NULL;
763 	}
764 }
765 
766 static void
767 eth_em_close(struct rte_eth_dev *dev)
768 {
769 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
770 	struct e1000_adapter *adapter =
771 		E1000_DEV_PRIVATE(dev->data->dev_private);
772 
773 	eth_em_stop(dev);
774 	adapter->stopped = 1;
775 	em_dev_free_queues(dev);
776 	e1000_phy_hw_reset(hw);
777 	em_release_manageability(hw);
778 	em_hw_control_release(hw);
779 }
780 
781 static int
782 em_get_rx_buffer_size(struct e1000_hw *hw)
783 {
784 	uint32_t rx_buf_size;
785 
786 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
787 	return rx_buf_size;
788 }
789 
790 /*********************************************************************
791  *
792  *  Initialize the hardware
793  *
794  **********************************************************************/
795 static int
796 em_hardware_init(struct e1000_hw *hw)
797 {
798 	uint32_t rx_buf_size;
799 	int diag;
800 
801 	/* Issue a global reset */
802 	e1000_reset_hw(hw);
803 
804 	/* Let the firmware know the OS is in control */
805 	em_hw_control_acquire(hw);
806 
807 	/*
808 	 * These parameters control the automatic generation (Tx) and
809 	 * response (Rx) to Ethernet PAUSE frames.
810 	 * - High water mark should allow for at least two standard size (1518)
811 	 *   frames to be received after sending an XOFF.
812 	 * - Low water mark works best when it is very near the high water mark.
813 	 *   This allows the receiver to restart by sending XON when it has
814 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
815 	 *   restart after one full frame is pulled from the buffer. There
816 	 *   could be several smaller frames in the buffer and if so they will
817 	 *   not trigger the XON until their total number reduces the buffer
818 	 *   by 1500.
819 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
820 	 */
821 	rx_buf_size = em_get_rx_buffer_size(hw);
822 
823 	hw->fc.high_water = rx_buf_size - PMD_ROUNDUP(ETHER_MAX_LEN * 2, 1024);
824 	hw->fc.low_water = hw->fc.high_water - 1500;
825 
826 	if (hw->mac.type == e1000_80003es2lan)
827 		hw->fc.pause_time = UINT16_MAX;
828 	else
829 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
830 
831 	hw->fc.send_xon = 1;
832 
833 	/* Set Flow control, use the tunable location if sane */
834 	if (em_fc_setting <= e1000_fc_full)
835 		hw->fc.requested_mode = em_fc_setting;
836 	else
837 		hw->fc.requested_mode = e1000_fc_none;
838 
839 	/* Workaround: no TX flow ctrl for PCH */
840 	if (hw->mac.type == e1000_pchlan)
841 		hw->fc.requested_mode = e1000_fc_rx_pause;
842 
843 	/* Override - settings for PCH2LAN, ya its magic :) */
844 	if (hw->mac.type == e1000_pch2lan) {
845 		hw->fc.high_water = 0x5C20;
846 		hw->fc.low_water = 0x5048;
847 		hw->fc.pause_time = 0x0650;
848 		hw->fc.refresh_time = 0x0400;
849 	} else if (hw->mac.type == e1000_pch_lpt ||
850 		   hw->mac.type == e1000_pch_spt ||
851 		   hw->mac.type == e1000_pch_cnp) {
852 		hw->fc.requested_mode = e1000_fc_full;
853 	}
854 
855 	diag = e1000_init_hw(hw);
856 	if (diag < 0)
857 		return diag;
858 	e1000_check_for_link(hw);
859 	return 0;
860 }
861 
862 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
863 static int
864 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
865 {
866 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
867 	struct e1000_hw_stats *stats =
868 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
869 	int pause_frames;
870 
871 	if(hw->phy.media_type == e1000_media_type_copper ||
872 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
873 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
874 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
875 	}
876 
877 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
878 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
879 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
880 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
881 
882 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
883 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
884 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
885 	stats->dc += E1000_READ_REG(hw, E1000_DC);
886 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
887 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
888 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
889 
890 	/*
891 	 * For watchdog management we need to know if we have been
892 	 * paused during the last interval, so capture that here.
893 	 */
894 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
895 	stats->xoffrxc += pause_frames;
896 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
897 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
898 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
899 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
900 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
901 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
902 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
903 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
904 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
905 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
906 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
907 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
908 
909 	/*
910 	 * For the 64-bit byte counters the low dword must be read first.
911 	 * Both registers clear on the read of the high dword.
912 	 */
913 
914 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
915 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
916 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
917 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
918 
919 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
920 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
921 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
922 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
923 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
924 
925 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
926 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
927 
928 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
929 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
930 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
931 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
932 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
933 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
934 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
935 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
936 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
937 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
938 
939 	/* Interrupt Counts */
940 
941 	if (hw->mac.type >= e1000_82571) {
942 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
943 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
944 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
945 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
946 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
947 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
948 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
949 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
950 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
951 	}
952 
953 	if (hw->mac.type >= e1000_82543) {
954 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
955 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
956 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
957 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
958 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
959 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
960 	}
961 
962 	if (rte_stats == NULL)
963 		return -EINVAL;
964 
965 	/* Rx Errors */
966 	rte_stats->imissed = stats->mpc;
967 	rte_stats->ierrors = stats->crcerrs +
968 	                     stats->rlec + stats->ruc + stats->roc +
969 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
970 
971 	/* Tx Errors */
972 	rte_stats->oerrors = stats->ecol + stats->latecol;
973 
974 	rte_stats->ipackets = stats->gprc;
975 	rte_stats->opackets = stats->gptc;
976 	rte_stats->ibytes   = stats->gorc;
977 	rte_stats->obytes   = stats->gotc;
978 	return 0;
979 }
980 
981 static void
982 eth_em_stats_reset(struct rte_eth_dev *dev)
983 {
984 	struct e1000_hw_stats *hw_stats =
985 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
986 
987 	/* HW registers are cleared on read */
988 	eth_em_stats_get(dev, NULL);
989 
990 	/* Reset software totals */
991 	memset(hw_stats, 0, sizeof(*hw_stats));
992 }
993 
994 static int
995 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
996 {
997 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
998 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
999 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1000 
1001 	em_rxq_intr_enable(hw);
1002 	rte_intr_enable(intr_handle);
1003 
1004 	return 0;
1005 }
1006 
1007 static int
1008 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1009 {
1010 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1011 
1012 	em_rxq_intr_disable(hw);
1013 
1014 	return 0;
1015 }
1016 
1017 uint32_t
1018 em_get_max_pktlen(struct rte_eth_dev *dev)
1019 {
1020 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1021 
1022 	switch (hw->mac.type) {
1023 	case e1000_82571:
1024 	case e1000_82572:
1025 	case e1000_ich9lan:
1026 	case e1000_ich10lan:
1027 	case e1000_pch2lan:
1028 	case e1000_pch_lpt:
1029 	case e1000_pch_spt:
1030 	case e1000_pch_cnp:
1031 	case e1000_82574:
1032 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1033 	case e1000_82583:
1034 		return 0x2412;
1035 	case e1000_pchlan:
1036 		return 0x1000;
1037 	/* Adapters that do not support jumbo frames */
1038 	case e1000_ich8lan:
1039 		return ETHER_MAX_LEN;
1040 	default:
1041 		return MAX_JUMBO_FRAME_SIZE;
1042 	}
1043 }
1044 
1045 static void
1046 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1047 {
1048 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1049 
1050 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1051 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1052 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1053 
1054 	/*
1055 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1056 	 * Unfortunatelly, all these nics have just one TX context.
1057 	 * So we have few choises for TX:
1058 	 * - Use just one TX queue.
1059 	 * - Allow cksum offload only for one TX queue.
1060 	 * - Don't allow TX cksum offload at all.
1061 	 * For now, option #1 was chosen.
1062 	 * To use second RX queue we have to use extended RX descriptor
1063 	 * (Multiple Receive Queues are mutually exclusive with UDP
1064 	 * fragmentation and are not supported when a legacy receive
1065 	 * descriptor format is used).
1066 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1067 	 * don't support extended RXD.
1068 	 * To avoid it we support just one RX queue for now (no RSS).
1069 	 */
1070 
1071 	dev_info->max_rx_queues = 1;
1072 	dev_info->max_tx_queues = 1;
1073 
1074 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1075 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1076 				    dev_info->rx_queue_offload_capa;
1077 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1078 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1079 				    dev_info->tx_queue_offload_capa;
1080 
1081 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1082 		.nb_max = E1000_MAX_RING_DESC,
1083 		.nb_min = E1000_MIN_RING_DESC,
1084 		.nb_align = EM_RXD_ALIGN,
1085 	};
1086 
1087 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1088 		.nb_max = E1000_MAX_RING_DESC,
1089 		.nb_min = E1000_MIN_RING_DESC,
1090 		.nb_align = EM_TXD_ALIGN,
1091 		.nb_seg_max = EM_TX_MAX_SEG,
1092 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1093 	};
1094 
1095 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1096 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1097 			ETH_LINK_SPEED_1G;
1098 
1099 	/* Preferred queue parameters */
1100 	dev_info->default_rxportconf.nb_queues = 1;
1101 	dev_info->default_txportconf.nb_queues = 1;
1102 	dev_info->default_txportconf.ring_size = 256;
1103 	dev_info->default_rxportconf.ring_size = 256;
1104 }
1105 
1106 /* return 0 means link status changed, -1 means not changed */
1107 static int
1108 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1109 {
1110 	struct e1000_hw *hw =
1111 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1112 	struct rte_eth_link link;
1113 	int link_check, count;
1114 
1115 	link_check = 0;
1116 	hw->mac.get_link_status = 1;
1117 
1118 	/* possible wait-to-complete in up to 9 seconds */
1119 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1120 		/* Read the real link status */
1121 		switch (hw->phy.media_type) {
1122 		case e1000_media_type_copper:
1123 			/* Do the work to read phy */
1124 			e1000_check_for_link(hw);
1125 			link_check = !hw->mac.get_link_status;
1126 			break;
1127 
1128 		case e1000_media_type_fiber:
1129 			e1000_check_for_link(hw);
1130 			link_check = (E1000_READ_REG(hw, E1000_STATUS) &
1131 					E1000_STATUS_LU);
1132 			break;
1133 
1134 		case e1000_media_type_internal_serdes:
1135 			e1000_check_for_link(hw);
1136 			link_check = hw->mac.serdes_has_link;
1137 			break;
1138 
1139 		default:
1140 			break;
1141 		}
1142 		if (link_check || wait_to_complete == 0)
1143 			break;
1144 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1145 	}
1146 	memset(&link, 0, sizeof(link));
1147 
1148 	/* Now we check if a transition has happened */
1149 	if (link_check && (link.link_status == ETH_LINK_DOWN)) {
1150 		uint16_t duplex, speed;
1151 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1152 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1153 				ETH_LINK_FULL_DUPLEX :
1154 				ETH_LINK_HALF_DUPLEX;
1155 		link.link_speed = speed;
1156 		link.link_status = ETH_LINK_UP;
1157 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1158 				ETH_LINK_SPEED_FIXED);
1159 	} else if (!link_check && (link.link_status == ETH_LINK_UP)) {
1160 		link.link_speed = ETH_SPEED_NUM_NONE;
1161 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1162 		link.link_status = ETH_LINK_DOWN;
1163 		link.link_autoneg = ETH_LINK_FIXED;
1164 	}
1165 
1166 	return rte_eth_linkstatus_set(dev, &link);
1167 }
1168 
1169 /*
1170  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1171  * For ASF and Pass Through versions of f/w this means
1172  * that the driver is loaded. For AMT version type f/w
1173  * this means that the network i/f is open.
1174  */
1175 static void
1176 em_hw_control_acquire(struct e1000_hw *hw)
1177 {
1178 	uint32_t ctrl_ext, swsm;
1179 
1180 	/* Let firmware know the driver has taken over */
1181 	if (hw->mac.type == e1000_82573) {
1182 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1183 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1184 
1185 	} else {
1186 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1187 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1188 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1189 	}
1190 }
1191 
1192 /*
1193  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1194  * For ASF and Pass Through versions of f/w this means that the
1195  * driver is no longer loaded. For AMT versions of the
1196  * f/w this means that the network i/f is closed.
1197  */
1198 static void
1199 em_hw_control_release(struct e1000_hw *hw)
1200 {
1201 	uint32_t ctrl_ext, swsm;
1202 
1203 	/* Let firmware taken over control of h/w */
1204 	if (hw->mac.type == e1000_82573) {
1205 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1206 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1207 	} else {
1208 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1209 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1210 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1211 	}
1212 }
1213 
1214 /*
1215  * Bit of a misnomer, what this really means is
1216  * to enable OS management of the system... aka
1217  * to disable special hardware management features.
1218  */
1219 static void
1220 em_init_manageability(struct e1000_hw *hw)
1221 {
1222 	if (e1000_enable_mng_pass_thru(hw)) {
1223 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1224 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1225 
1226 		/* disable hardware interception of ARP */
1227 		manc &= ~(E1000_MANC_ARP_EN);
1228 
1229 		/* enable receiving management packets to the host */
1230 		manc |= E1000_MANC_EN_MNG2HOST;
1231 		manc2h |= 1 << 5;  /* Mng Port 623 */
1232 		manc2h |= 1 << 6;  /* Mng Port 664 */
1233 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1234 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1235 	}
1236 }
1237 
1238 /*
1239  * Give control back to hardware management
1240  * controller if there is one.
1241  */
1242 static void
1243 em_release_manageability(struct e1000_hw *hw)
1244 {
1245 	uint32_t manc;
1246 
1247 	if (e1000_enable_mng_pass_thru(hw)) {
1248 		manc = E1000_READ_REG(hw, E1000_MANC);
1249 
1250 		/* re-enable hardware interception of ARP */
1251 		manc |= E1000_MANC_ARP_EN;
1252 		manc &= ~E1000_MANC_EN_MNG2HOST;
1253 
1254 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1255 	}
1256 }
1257 
1258 static void
1259 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1260 {
1261 	struct e1000_hw *hw =
1262 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1263 	uint32_t rctl;
1264 
1265 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1266 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1267 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1268 }
1269 
1270 static void
1271 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1272 {
1273 	struct e1000_hw *hw =
1274 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1275 	uint32_t rctl;
1276 
1277 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1278 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1279 	if (dev->data->all_multicast == 1)
1280 		rctl |= E1000_RCTL_MPE;
1281 	else
1282 		rctl &= (~E1000_RCTL_MPE);
1283 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1284 }
1285 
1286 static void
1287 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1288 {
1289 	struct e1000_hw *hw =
1290 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1291 	uint32_t rctl;
1292 
1293 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1294 	rctl |= E1000_RCTL_MPE;
1295 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1296 }
1297 
1298 static void
1299 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1300 {
1301 	struct e1000_hw *hw =
1302 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1303 	uint32_t rctl;
1304 
1305 	if (dev->data->promiscuous == 1)
1306 		return; /* must remain in all_multicast mode */
1307 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1308 	rctl &= (~E1000_RCTL_MPE);
1309 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1310 }
1311 
1312 static int
1313 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1314 {
1315 	struct e1000_hw *hw =
1316 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1317 	struct e1000_vfta * shadow_vfta =
1318 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1319 	uint32_t vfta;
1320 	uint32_t vid_idx;
1321 	uint32_t vid_bit;
1322 
1323 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1324 			      E1000_VFTA_ENTRY_MASK);
1325 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1326 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1327 	if (on)
1328 		vfta |= vid_bit;
1329 	else
1330 		vfta &= ~vid_bit;
1331 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1332 
1333 	/* update local VFTA copy */
1334 	shadow_vfta->vfta[vid_idx] = vfta;
1335 
1336 	return 0;
1337 }
1338 
1339 static void
1340 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1341 {
1342 	struct e1000_hw *hw =
1343 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1344 	uint32_t reg;
1345 
1346 	/* Filter Table Disable */
1347 	reg = E1000_READ_REG(hw, E1000_RCTL);
1348 	reg &= ~E1000_RCTL_CFIEN;
1349 	reg &= ~E1000_RCTL_VFE;
1350 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1351 }
1352 
1353 static void
1354 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1355 {
1356 	struct e1000_hw *hw =
1357 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1358 	struct e1000_vfta * shadow_vfta =
1359 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1360 	uint32_t reg;
1361 	int i;
1362 
1363 	/* Filter Table Enable, CFI not used for packet acceptance */
1364 	reg = E1000_READ_REG(hw, E1000_RCTL);
1365 	reg &= ~E1000_RCTL_CFIEN;
1366 	reg |= E1000_RCTL_VFE;
1367 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1368 
1369 	/* restore vfta from local copy */
1370 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1371 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1372 }
1373 
1374 static void
1375 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1376 {
1377 	struct e1000_hw *hw =
1378 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1379 	uint32_t reg;
1380 
1381 	/* VLAN Mode Disable */
1382 	reg = E1000_READ_REG(hw, E1000_CTRL);
1383 	reg &= ~E1000_CTRL_VME;
1384 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1385 
1386 }
1387 
1388 static void
1389 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1390 {
1391 	struct e1000_hw *hw =
1392 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1393 	uint32_t reg;
1394 
1395 	/* VLAN Mode Enable */
1396 	reg = E1000_READ_REG(hw, E1000_CTRL);
1397 	reg |= E1000_CTRL_VME;
1398 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1399 }
1400 
1401 static int
1402 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1403 {
1404 	struct rte_eth_rxmode *rxmode;
1405 
1406 	rxmode = &dev->data->dev_conf.rxmode;
1407 	if(mask & ETH_VLAN_STRIP_MASK){
1408 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1409 			em_vlan_hw_strip_enable(dev);
1410 		else
1411 			em_vlan_hw_strip_disable(dev);
1412 	}
1413 
1414 	if(mask & ETH_VLAN_FILTER_MASK){
1415 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1416 			em_vlan_hw_filter_enable(dev);
1417 		else
1418 			em_vlan_hw_filter_disable(dev);
1419 	}
1420 
1421 	return 0;
1422 }
1423 
1424 /*
1425  * It enables the interrupt mask and then enable the interrupt.
1426  *
1427  * @param dev
1428  *  Pointer to struct rte_eth_dev.
1429  *
1430  * @return
1431  *  - On success, zero.
1432  *  - On failure, a negative value.
1433  */
1434 static int
1435 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1436 {
1437 	uint32_t regval;
1438 	struct e1000_hw *hw =
1439 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1440 
1441 	/* clear interrupt */
1442 	E1000_READ_REG(hw, E1000_ICR);
1443 	regval = E1000_READ_REG(hw, E1000_IMS);
1444 	E1000_WRITE_REG(hw, E1000_IMS,
1445 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1446 	return 0;
1447 }
1448 
1449 /*
1450  * It clears the interrupt causes and enables the interrupt.
1451  * It will be called once only during nic initialized.
1452  *
1453  * @param dev
1454  *  Pointer to struct rte_eth_dev.
1455  *
1456  * @return
1457  *  - On success, zero.
1458  *  - On failure, a negative value.
1459  */
1460 static int
1461 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1462 {
1463 	struct e1000_hw *hw =
1464 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1465 
1466 	E1000_READ_REG(hw, E1000_ICR);
1467 	em_rxq_intr_enable(hw);
1468 	return 0;
1469 }
1470 
1471 /*
1472  * It enable receive packet interrupt.
1473  * @param hw
1474  * Pointer to struct e1000_hw
1475  *
1476  * @return
1477  */
1478 static void
1479 em_rxq_intr_enable(struct e1000_hw *hw)
1480 {
1481 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1482 	E1000_WRITE_FLUSH(hw);
1483 }
1484 
1485 /*
1486  * It disabled lsc interrupt.
1487  * @param hw
1488  * Pointer to struct e1000_hw
1489  *
1490  * @return
1491  */
1492 static void
1493 em_lsc_intr_disable(struct e1000_hw *hw)
1494 {
1495 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1496 	E1000_WRITE_FLUSH(hw);
1497 }
1498 
1499 /*
1500  * It disabled receive packet interrupt.
1501  * @param hw
1502  * Pointer to struct e1000_hw
1503  *
1504  * @return
1505  */
1506 static void
1507 em_rxq_intr_disable(struct e1000_hw *hw)
1508 {
1509 	E1000_READ_REG(hw, E1000_ICR);
1510 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1511 	E1000_WRITE_FLUSH(hw);
1512 }
1513 
1514 /*
1515  * It reads ICR and gets interrupt causes, check it and set a bit flag
1516  * to update link status.
1517  *
1518  * @param dev
1519  *  Pointer to struct rte_eth_dev.
1520  *
1521  * @return
1522  *  - On success, zero.
1523  *  - On failure, a negative value.
1524  */
1525 static int
1526 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1527 {
1528 	uint32_t icr;
1529 	struct e1000_hw *hw =
1530 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1531 	struct e1000_interrupt *intr =
1532 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1533 
1534 	/* read-on-clear nic registers here */
1535 	icr = E1000_READ_REG(hw, E1000_ICR);
1536 	if (icr & E1000_ICR_LSC) {
1537 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1538 	}
1539 
1540 	return 0;
1541 }
1542 
1543 /*
1544  * It executes link_update after knowing an interrupt is prsent.
1545  *
1546  * @param dev
1547  *  Pointer to struct rte_eth_dev.
1548  *
1549  * @return
1550  *  - On success, zero.
1551  *  - On failure, a negative value.
1552  */
1553 static int
1554 eth_em_interrupt_action(struct rte_eth_dev *dev,
1555 			struct rte_intr_handle *intr_handle)
1556 {
1557 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1558 	struct e1000_hw *hw =
1559 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1560 	struct e1000_interrupt *intr =
1561 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1562 	struct rte_eth_link link;
1563 	int ret;
1564 
1565 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1566 		return -1;
1567 
1568 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1569 	rte_intr_enable(intr_handle);
1570 
1571 	/* set get_link_status to check register later */
1572 	hw->mac.get_link_status = 1;
1573 	ret = eth_em_link_update(dev, 0);
1574 
1575 	/* check if link has changed */
1576 	if (ret < 0)
1577 		return 0;
1578 
1579 	rte_eth_linkstatus_get(dev, &link);
1580 
1581 	if (link.link_status) {
1582 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1583 			     dev->data->port_id, link.link_speed,
1584 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1585 			     "full-duplex" : "half-duplex");
1586 	} else {
1587 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1588 	}
1589 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1590 		     pci_dev->addr.domain, pci_dev->addr.bus,
1591 		     pci_dev->addr.devid, pci_dev->addr.function);
1592 
1593 	return 0;
1594 }
1595 
1596 /**
1597  * Interrupt handler which shall be registered at first.
1598  *
1599  * @param handle
1600  *  Pointer to interrupt handle.
1601  * @param param
1602  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1603  *
1604  * @return
1605  *  void
1606  */
1607 static void
1608 eth_em_interrupt_handler(void *param)
1609 {
1610 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1611 
1612 	eth_em_interrupt_get_status(dev);
1613 	eth_em_interrupt_action(dev, dev->intr_handle);
1614 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1615 }
1616 
1617 static int
1618 eth_em_led_on(struct rte_eth_dev *dev)
1619 {
1620 	struct e1000_hw *hw;
1621 
1622 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1623 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1624 }
1625 
1626 static int
1627 eth_em_led_off(struct rte_eth_dev *dev)
1628 {
1629 	struct e1000_hw *hw;
1630 
1631 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1632 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1633 }
1634 
1635 static int
1636 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1637 {
1638 	struct e1000_hw *hw;
1639 	uint32_t ctrl;
1640 	int tx_pause;
1641 	int rx_pause;
1642 
1643 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1644 	fc_conf->pause_time = hw->fc.pause_time;
1645 	fc_conf->high_water = hw->fc.high_water;
1646 	fc_conf->low_water = hw->fc.low_water;
1647 	fc_conf->send_xon = hw->fc.send_xon;
1648 	fc_conf->autoneg = hw->mac.autoneg;
1649 
1650 	/*
1651 	 * Return rx_pause and tx_pause status according to actual setting of
1652 	 * the TFCE and RFCE bits in the CTRL register.
1653 	 */
1654 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1655 	if (ctrl & E1000_CTRL_TFCE)
1656 		tx_pause = 1;
1657 	else
1658 		tx_pause = 0;
1659 
1660 	if (ctrl & E1000_CTRL_RFCE)
1661 		rx_pause = 1;
1662 	else
1663 		rx_pause = 0;
1664 
1665 	if (rx_pause && tx_pause)
1666 		fc_conf->mode = RTE_FC_FULL;
1667 	else if (rx_pause)
1668 		fc_conf->mode = RTE_FC_RX_PAUSE;
1669 	else if (tx_pause)
1670 		fc_conf->mode = RTE_FC_TX_PAUSE;
1671 	else
1672 		fc_conf->mode = RTE_FC_NONE;
1673 
1674 	return 0;
1675 }
1676 
1677 static int
1678 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1679 {
1680 	struct e1000_hw *hw;
1681 	int err;
1682 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1683 		e1000_fc_none,
1684 		e1000_fc_rx_pause,
1685 		e1000_fc_tx_pause,
1686 		e1000_fc_full
1687 	};
1688 	uint32_t rx_buf_size;
1689 	uint32_t max_high_water;
1690 	uint32_t rctl;
1691 
1692 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1693 	if (fc_conf->autoneg != hw->mac.autoneg)
1694 		return -ENOTSUP;
1695 	rx_buf_size = em_get_rx_buffer_size(hw);
1696 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1697 
1698 	/* At least reserve one Ethernet frame for watermark */
1699 	max_high_water = rx_buf_size - ETHER_MAX_LEN;
1700 	if ((fc_conf->high_water > max_high_water) ||
1701 	    (fc_conf->high_water < fc_conf->low_water)) {
1702 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1703 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1704 		return -EINVAL;
1705 	}
1706 
1707 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1708 	hw->fc.pause_time     = fc_conf->pause_time;
1709 	hw->fc.high_water     = fc_conf->high_water;
1710 	hw->fc.low_water      = fc_conf->low_water;
1711 	hw->fc.send_xon	      = fc_conf->send_xon;
1712 
1713 	err = e1000_setup_link_generic(hw);
1714 	if (err == E1000_SUCCESS) {
1715 
1716 		/* check if we want to forward MAC frames - driver doesn't have native
1717 		 * capability to do that, so we'll write the registers ourselves */
1718 
1719 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1720 
1721 		/* set or clear MFLCN.PMCF bit depending on configuration */
1722 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1723 			rctl |= E1000_RCTL_PMCF;
1724 		else
1725 			rctl &= ~E1000_RCTL_PMCF;
1726 
1727 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1728 		E1000_WRITE_FLUSH(hw);
1729 
1730 		return 0;
1731 	}
1732 
1733 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1734 	return -EIO;
1735 }
1736 
1737 static int
1738 eth_em_rar_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr,
1739 		uint32_t index, __rte_unused uint32_t pool)
1740 {
1741 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1742 
1743 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1744 }
1745 
1746 static void
1747 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1748 {
1749 	uint8_t addr[ETHER_ADDR_LEN];
1750 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1751 
1752 	memset(addr, 0, sizeof(addr));
1753 
1754 	e1000_rar_set(hw, addr, index);
1755 }
1756 
1757 static int
1758 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1759 			    struct ether_addr *addr)
1760 {
1761 	eth_em_rar_clear(dev, 0);
1762 
1763 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1764 }
1765 
1766 static int
1767 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1768 {
1769 	struct rte_eth_dev_info dev_info;
1770 	struct e1000_hw *hw;
1771 	uint32_t frame_size;
1772 	uint32_t rctl;
1773 
1774 	eth_em_infos_get(dev, &dev_info);
1775 	frame_size = mtu + ETHER_HDR_LEN + ETHER_CRC_LEN + VLAN_TAG_SIZE;
1776 
1777 	/* check that mtu is within the allowed range */
1778 	if ((mtu < ETHER_MIN_MTU) || (frame_size > dev_info.max_rx_pktlen))
1779 		return -EINVAL;
1780 
1781 	/* refuse mtu that requires the support of scattered packets when this
1782 	 * feature has not been enabled before. */
1783 	if (!dev->data->scattered_rx &&
1784 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1785 		return -EINVAL;
1786 
1787 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1788 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1789 
1790 	/* switch to jumbo mode if needed */
1791 	if (frame_size > ETHER_MAX_LEN) {
1792 		dev->data->dev_conf.rxmode.offloads |=
1793 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1794 		rctl |= E1000_RCTL_LPE;
1795 	} else {
1796 		dev->data->dev_conf.rxmode.offloads &=
1797 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1798 		rctl &= ~E1000_RCTL_LPE;
1799 	}
1800 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1801 
1802 	/* update max frame size */
1803 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1804 	return 0;
1805 }
1806 
1807 static int
1808 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1809 			struct ether_addr *mc_addr_set,
1810 			uint32_t nb_mc_addr)
1811 {
1812 	struct e1000_hw *hw;
1813 
1814 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1815 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1816 	return 0;
1817 }
1818 
1819 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1820 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1821 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1822 
1823 /* see e1000_logs.c */
1824 RTE_INIT(igb_init_log)
1825 {
1826 	e1000_igb_init_log();
1827 }
1828