xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision 452c1916)
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 <ethdev_driver.h>
19 #include <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 int eth_em_stop(struct rte_eth_dev *dev);
37 static int eth_em_close(struct rte_eth_dev *dev);
38 static int eth_em_promiscuous_enable(struct rte_eth_dev *dev);
39 static int eth_em_promiscuous_disable(struct rte_eth_dev *dev);
40 static int eth_em_allmulticast_enable(struct rte_eth_dev *dev);
41 static int 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 int eth_em_stats_reset(struct rte_eth_dev *dev);
47 static int 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,
93 			struct rte_ether_addr *mac_addr,
94 			uint32_t index, uint32_t pool);
95 static void eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index);
96 static int eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
97 					 struct rte_ether_addr *addr);
98 
99 static int eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
100 				   struct rte_ether_addr *mc_addr_set,
101 				   uint32_t nb_mc_addr);
102 
103 #define EM_FC_PAUSE_TIME 0x0680
104 #define EM_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
105 #define EM_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
106 
107 static enum e1000_fc_mode em_fc_setting = e1000_fc_full;
108 
109 /*
110  * The set of PCI devices this driver supports
111  */
112 static const struct rte_pci_id pci_id_em_map[] = {
113 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82540EM) },
114 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_COPPER) },
115 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82545EM_FIBER) },
116 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_COPPER) },
117 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_FIBER) },
118 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82546EB_QUAD_COPPER) },
119 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_COPPER) },
120 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_FIBER) },
121 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES) },
122 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_DUAL) },
123 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_SERDES_QUAD) },
124 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER) },
125 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571PT_QUAD_COPPER) },
126 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_FIBER) },
127 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82571EB_QUAD_COPPER_LP) },
128 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_COPPER) },
129 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_FIBER) },
130 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI_SERDES) },
131 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82572EI) },
132 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82573L) },
133 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574L) },
134 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82574LA) },
135 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_82583V) },
136 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH2_LV_LM) },
137 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_LM) },
138 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPT_I217_V) },
139 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_LM) },
140 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LPTLP_I218_V) },
141 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM2) },
142 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V2) },
143 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_LM3) },
144 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_I218_V3) },
145 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM) },
146 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V) },
147 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM2) },
148 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V2) },
149 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_LBG_I219_LM3) },
150 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM4) },
151 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V4) },
152 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_LM5) },
153 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_SPT_I219_V5) },
154 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM6) },
155 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V6) },
156 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_LM7) },
157 	{ RTE_PCI_DEVICE(E1000_INTEL_VENDOR_ID, E1000_DEV_ID_PCH_CNP_I219_V7) },
158 	{ .vendor_id = 0, /* sentinel */ },
159 };
160 
161 static const struct eth_dev_ops eth_em_ops = {
162 	.dev_configure        = eth_em_configure,
163 	.dev_start            = eth_em_start,
164 	.dev_stop             = eth_em_stop,
165 	.dev_close            = eth_em_close,
166 	.promiscuous_enable   = eth_em_promiscuous_enable,
167 	.promiscuous_disable  = eth_em_promiscuous_disable,
168 	.allmulticast_enable  = eth_em_allmulticast_enable,
169 	.allmulticast_disable = eth_em_allmulticast_disable,
170 	.link_update          = eth_em_link_update,
171 	.stats_get            = eth_em_stats_get,
172 	.stats_reset          = eth_em_stats_reset,
173 	.dev_infos_get        = eth_em_infos_get,
174 	.mtu_set              = eth_em_mtu_set,
175 	.vlan_filter_set      = eth_em_vlan_filter_set,
176 	.vlan_offload_set     = eth_em_vlan_offload_set,
177 	.rx_queue_setup       = eth_em_rx_queue_setup,
178 	.rx_queue_release     = eth_em_rx_queue_release,
179 	.tx_queue_setup       = eth_em_tx_queue_setup,
180 	.tx_queue_release     = eth_em_tx_queue_release,
181 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
182 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
183 	.dev_led_on           = eth_em_led_on,
184 	.dev_led_off          = eth_em_led_off,
185 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
186 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
187 	.mac_addr_set         = eth_em_default_mac_addr_set,
188 	.mac_addr_add         = eth_em_rar_set,
189 	.mac_addr_remove      = eth_em_rar_clear,
190 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
191 	.rxq_info_get         = em_rxq_info_get,
192 	.txq_info_get         = em_txq_info_get,
193 };
194 
195 
196 /**
197  *  eth_em_dev_is_ich8 - Check for ICH8 device
198  *  @hw: pointer to the HW structure
199  *
200  *  return TRUE for ICH8, otherwise FALSE
201  **/
202 static bool
203 eth_em_dev_is_ich8(struct e1000_hw *hw)
204 {
205 	DEBUGFUNC("eth_em_dev_is_ich8");
206 
207 	switch (hw->device_id) {
208 	case E1000_DEV_ID_PCH2_LV_LM:
209 	case E1000_DEV_ID_PCH_LPT_I217_LM:
210 	case E1000_DEV_ID_PCH_LPT_I217_V:
211 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
212 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
213 	case E1000_DEV_ID_PCH_I218_V2:
214 	case E1000_DEV_ID_PCH_I218_LM2:
215 	case E1000_DEV_ID_PCH_I218_V3:
216 	case E1000_DEV_ID_PCH_I218_LM3:
217 	case E1000_DEV_ID_PCH_SPT_I219_LM:
218 	case E1000_DEV_ID_PCH_SPT_I219_V:
219 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
220 	case E1000_DEV_ID_PCH_SPT_I219_V2:
221 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
222 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
223 	case E1000_DEV_ID_PCH_SPT_I219_V4:
224 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
225 	case E1000_DEV_ID_PCH_SPT_I219_V5:
226 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
227 	case E1000_DEV_ID_PCH_CNP_I219_V6:
228 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
229 	case E1000_DEV_ID_PCH_CNP_I219_V7:
230 		return 1;
231 	default:
232 		return 0;
233 	}
234 }
235 
236 static int
237 eth_em_dev_init(struct rte_eth_dev *eth_dev)
238 {
239 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
240 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
241 	struct e1000_adapter *adapter =
242 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
243 	struct e1000_hw *hw =
244 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
245 	struct e1000_vfta * shadow_vfta =
246 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
247 
248 	eth_dev->dev_ops = &eth_em_ops;
249 	eth_dev->rx_queue_count = eth_em_rx_queue_count;
250 	eth_dev->rx_descriptor_status = eth_em_rx_descriptor_status;
251 	eth_dev->tx_descriptor_status = eth_em_tx_descriptor_status;
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", RTE_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 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
292 		return -ENOMEM;
293 	}
294 
295 	/* Copy the permanent MAC address */
296 	rte_ether_addr_copy((struct rte_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 	PMD_INIT_FUNC_TRACE();
316 
317 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
318 		return 0;
319 
320 	eth_em_close(eth_dev);
321 
322 	return 0;
323 }
324 
325 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
326 	struct rte_pci_device *pci_dev)
327 {
328 	return rte_eth_dev_pci_generic_probe(pci_dev,
329 		sizeof(struct e1000_adapter), eth_em_dev_init);
330 }
331 
332 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
333 {
334 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
335 }
336 
337 static struct rte_pci_driver rte_em_pmd = {
338 	.id_table = pci_id_em_map,
339 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
340 	.probe = eth_em_pci_probe,
341 	.remove = eth_em_pci_remove,
342 };
343 
344 static int
345 em_hw_init(struct e1000_hw *hw)
346 {
347 	int diag;
348 
349 	diag = hw->mac.ops.init_params(hw);
350 	if (diag != 0) {
351 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
352 		return diag;
353 	}
354 	diag = hw->nvm.ops.init_params(hw);
355 	if (diag != 0) {
356 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
357 		return diag;
358 	}
359 	diag = hw->phy.ops.init_params(hw);
360 	if (diag != 0) {
361 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
362 		return diag;
363 	}
364 	(void) e1000_get_bus_info(hw);
365 
366 	hw->mac.autoneg = 1;
367 	hw->phy.autoneg_wait_to_complete = 0;
368 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
369 
370 	e1000_init_script_state_82541(hw, TRUE);
371 	e1000_set_tbi_compatibility_82543(hw, TRUE);
372 
373 	/* Copper options */
374 	if (hw->phy.media_type == e1000_media_type_copper) {
375 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
376 		hw->phy.disable_polarity_correction = 0;
377 		hw->phy.ms_type = e1000_ms_hw_default;
378 	}
379 
380 	/*
381 	 * Start from a known state, this is important in reading the nvm
382 	 * and mac from that.
383 	 */
384 	e1000_reset_hw(hw);
385 
386 	/* Make sure we have a good EEPROM before we read from it */
387 	if (e1000_validate_nvm_checksum(hw) < 0) {
388 		/*
389 		 * Some PCI-E parts fail the first check due to
390 		 * the link being in sleep state, call it again,
391 		 * if it fails a second time its a real issue.
392 		 */
393 		diag = e1000_validate_nvm_checksum(hw);
394 		if (diag < 0) {
395 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
396 			goto error;
397 		}
398 	}
399 
400 	/* Read the permanent MAC address out of the EEPROM */
401 	diag = e1000_read_mac_addr(hw);
402 	if (diag != 0) {
403 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
404 		goto error;
405 	}
406 
407 	/* Now initialize the hardware */
408 	diag = em_hardware_init(hw);
409 	if (diag != 0) {
410 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
411 		goto error;
412 	}
413 
414 	hw->mac.get_link_status = 1;
415 
416 	/* Indicate SOL/IDER usage */
417 	diag = e1000_check_reset_block(hw);
418 	if (diag < 0) {
419 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
420 			"SOL/IDER session");
421 	}
422 	return 0;
423 
424 error:
425 	em_hw_control_release(hw);
426 	return diag;
427 }
428 
429 static int
430 eth_em_configure(struct rte_eth_dev *dev)
431 {
432 	struct e1000_interrupt *intr =
433 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
434 
435 	PMD_INIT_FUNC_TRACE();
436 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
437 
438 	PMD_INIT_FUNC_TRACE();
439 
440 	return 0;
441 }
442 
443 static void
444 em_set_pba(struct e1000_hw *hw)
445 {
446 	uint32_t pba;
447 
448 	/*
449 	 * Packet Buffer Allocation (PBA)
450 	 * Writing PBA sets the receive portion of the buffer
451 	 * the remainder is used for the transmit buffer.
452 	 * Devices before the 82547 had a Packet Buffer of 64K.
453 	 * After the 82547 the buffer was reduced to 40K.
454 	 */
455 	switch (hw->mac.type) {
456 		case e1000_82547:
457 		case e1000_82547_rev_2:
458 		/* 82547: Total Packet Buffer is 40K */
459 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
460 			break;
461 		case e1000_82571:
462 		case e1000_82572:
463 		case e1000_80003es2lan:
464 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
465 			break;
466 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
467 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
468 			break;
469 		case e1000_82574:
470 		case e1000_82583:
471 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
472 			break;
473 		case e1000_ich8lan:
474 			pba = E1000_PBA_8K;
475 			break;
476 		case e1000_ich9lan:
477 		case e1000_ich10lan:
478 			pba = E1000_PBA_10K;
479 			break;
480 		case e1000_pchlan:
481 		case e1000_pch2lan:
482 		case e1000_pch_lpt:
483 		case e1000_pch_spt:
484 		case e1000_pch_cnp:
485 			pba = E1000_PBA_26K;
486 			break;
487 		default:
488 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
489 	}
490 
491 	E1000_WRITE_REG(hw, E1000_PBA, pba);
492 }
493 
494 static void
495 eth_em_rxtx_control(struct rte_eth_dev *dev,
496 		    bool enable)
497 {
498 	struct e1000_hw *hw =
499 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
500 	uint32_t tctl, rctl;
501 
502 	tctl = E1000_READ_REG(hw, E1000_TCTL);
503 	rctl = E1000_READ_REG(hw, E1000_RCTL);
504 	if (enable) {
505 		/* enable Tx/Rx */
506 		tctl |= E1000_TCTL_EN;
507 		rctl |= E1000_RCTL_EN;
508 	} else {
509 		/* disable Tx/Rx */
510 		tctl &= ~E1000_TCTL_EN;
511 		rctl &= ~E1000_RCTL_EN;
512 	}
513 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
514 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
515 	E1000_WRITE_FLUSH(hw);
516 }
517 
518 static int
519 eth_em_start(struct rte_eth_dev *dev)
520 {
521 	struct e1000_adapter *adapter =
522 		E1000_DEV_PRIVATE(dev->data->dev_private);
523 	struct e1000_hw *hw =
524 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
525 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
526 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
527 	int ret, mask;
528 	uint32_t intr_vector = 0;
529 	uint32_t *speeds;
530 	int num_speeds;
531 	bool autoneg;
532 
533 	PMD_INIT_FUNC_TRACE();
534 
535 	ret = eth_em_stop(dev);
536 	if (ret != 0)
537 		return ret;
538 
539 	e1000_power_up_phy(hw);
540 
541 	/* Set default PBA value */
542 	em_set_pba(hw);
543 
544 	/* Put the address into the Receive Address Array */
545 	e1000_rar_set(hw, hw->mac.addr, 0);
546 
547 	/*
548 	 * With the 82571 adapter, RAR[0] may be overwritten
549 	 * when the other port is reset, we make a duplicate
550 	 * in RAR[14] for that eventuality, this assures
551 	 * the interface continues to function.
552 	 */
553 	if (hw->mac.type == e1000_82571) {
554 		e1000_set_laa_state_82571(hw, TRUE);
555 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
556 	}
557 
558 	/* Initialize the hardware */
559 	if (em_hardware_init(hw)) {
560 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
561 		return -EIO;
562 	}
563 
564 	E1000_WRITE_REG(hw, E1000_VET, RTE_ETHER_TYPE_VLAN);
565 
566 	/* Configure for OS presence */
567 	em_init_manageability(hw);
568 
569 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
570 		intr_vector = dev->data->nb_rx_queues;
571 		if (rte_intr_efd_enable(intr_handle, intr_vector))
572 			return -1;
573 	}
574 
575 	if (rte_intr_dp_is_en(intr_handle)) {
576 		if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
577 						   dev->data->nb_rx_queues)) {
578 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
579 				     " intr_vec", dev->data->nb_rx_queues);
580 			return -ENOMEM;
581 		}
582 
583 		/* enable rx interrupt */
584 		em_rxq_intr_enable(hw);
585 	}
586 
587 	eth_em_tx_init(dev);
588 
589 	ret = eth_em_rx_init(dev);
590 	if (ret) {
591 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
592 		em_dev_clear_queues(dev);
593 		return ret;
594 	}
595 
596 	e1000_clear_hw_cntrs_base_generic(hw);
597 
598 	mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
599 			RTE_ETH_VLAN_EXTEND_MASK;
600 	ret = eth_em_vlan_offload_set(dev, mask);
601 	if (ret) {
602 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
603 		em_dev_clear_queues(dev);
604 		return ret;
605 	}
606 
607 	/* Set Interrupt Throttling Rate to maximum allowed value. */
608 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
609 
610 	/* Setup link speed and duplex */
611 	speeds = &dev->data->dev_conf.link_speeds;
612 	if (*speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
613 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
614 		hw->mac.autoneg = 1;
615 	} else {
616 		num_speeds = 0;
617 		autoneg = (*speeds & RTE_ETH_LINK_SPEED_FIXED) == 0;
618 
619 		/* Reset */
620 		hw->phy.autoneg_advertised = 0;
621 
622 		if (*speeds & ~(RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
623 				RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
624 				RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_FIXED)) {
625 			num_speeds = -1;
626 			goto error_invalid_config;
627 		}
628 		if (*speeds & RTE_ETH_LINK_SPEED_10M_HD) {
629 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
630 			num_speeds++;
631 		}
632 		if (*speeds & RTE_ETH_LINK_SPEED_10M) {
633 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
634 			num_speeds++;
635 		}
636 		if (*speeds & RTE_ETH_LINK_SPEED_100M_HD) {
637 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
638 			num_speeds++;
639 		}
640 		if (*speeds & RTE_ETH_LINK_SPEED_100M) {
641 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
642 			num_speeds++;
643 		}
644 		if (*speeds & RTE_ETH_LINK_SPEED_1G) {
645 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
646 			num_speeds++;
647 		}
648 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
649 			goto error_invalid_config;
650 
651 		/* Set/reset the mac.autoneg based on the link speed,
652 		 * fixed or not
653 		 */
654 		if (!autoneg) {
655 			hw->mac.autoneg = 0;
656 			hw->mac.forced_speed_duplex =
657 					hw->phy.autoneg_advertised;
658 		} else {
659 			hw->mac.autoneg = 1;
660 		}
661 	}
662 
663 	e1000_setup_link(hw);
664 
665 	if (rte_intr_allow_others(intr_handle)) {
666 		/* check if lsc interrupt is enabled */
667 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
668 			ret = eth_em_interrupt_setup(dev);
669 			if (ret) {
670 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
671 				em_dev_clear_queues(dev);
672 				return ret;
673 			}
674 		}
675 	} else {
676 		rte_intr_callback_unregister(intr_handle,
677 						eth_em_interrupt_handler,
678 						(void *)dev);
679 		if (dev->data->dev_conf.intr_conf.lsc != 0)
680 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
681 				     " no intr multiplexn");
682 	}
683 	/* check if rxq interrupt is enabled */
684 	if (dev->data->dev_conf.intr_conf.rxq != 0)
685 		eth_em_rxq_interrupt_setup(dev);
686 
687 	rte_intr_enable(intr_handle);
688 
689 	adapter->stopped = 0;
690 
691 	eth_em_rxtx_control(dev, true);
692 	eth_em_link_update(dev, 0);
693 
694 	PMD_INIT_LOG(DEBUG, "<<");
695 
696 	return 0;
697 
698 error_invalid_config:
699 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
700 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
701 	em_dev_clear_queues(dev);
702 	return -EINVAL;
703 }
704 
705 /*********************************************************************
706  *
707  *  This routine disables all traffic on the adapter by issuing a
708  *  global reset on the MAC.
709  *
710  **********************************************************************/
711 static int
712 eth_em_stop(struct rte_eth_dev *dev)
713 {
714 	struct rte_eth_link link;
715 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
716 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
717 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
718 
719 	dev->data->dev_started = 0;
720 
721 	eth_em_rxtx_control(dev, false);
722 	em_rxq_intr_disable(hw);
723 	em_lsc_intr_disable(hw);
724 
725 	e1000_reset_hw(hw);
726 
727 	/* Flush desc rings for i219 */
728 	if (hw->mac.type == e1000_pch_spt || hw->mac.type == e1000_pch_cnp)
729 		em_flush_desc_rings(dev);
730 
731 	if (hw->mac.type >= e1000_82544)
732 		E1000_WRITE_REG(hw, E1000_WUC, 0);
733 
734 	/* Power down the phy. Needed to make the link go down */
735 	e1000_power_down_phy(hw);
736 
737 	em_dev_clear_queues(dev);
738 
739 	/* clear the recorded link status */
740 	memset(&link, 0, sizeof(link));
741 	rte_eth_linkstatus_set(dev, &link);
742 
743 	if (!rte_intr_allow_others(intr_handle))
744 		/* resume to the default handler */
745 		rte_intr_callback_register(intr_handle,
746 					   eth_em_interrupt_handler,
747 					   (void *)dev);
748 
749 	/* Clean datapath event and queue/vec mapping */
750 	rte_intr_efd_disable(intr_handle);
751 	rte_intr_vec_list_free(intr_handle);
752 
753 	return 0;
754 }
755 
756 static int
757 eth_em_close(struct rte_eth_dev *dev)
758 {
759 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
760 	struct e1000_adapter *adapter =
761 		E1000_DEV_PRIVATE(dev->data->dev_private);
762 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
763 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
764 	int ret;
765 
766 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
767 		return 0;
768 
769 	ret = eth_em_stop(dev);
770 	adapter->stopped = 1;
771 	em_dev_free_queues(dev);
772 	e1000_phy_hw_reset(hw);
773 	em_release_manageability(hw);
774 	em_hw_control_release(hw);
775 
776 	/* disable uio intr before callback unregister */
777 	rte_intr_disable(intr_handle);
778 	rte_intr_callback_unregister(intr_handle,
779 				     eth_em_interrupt_handler, dev);
780 
781 	return ret;
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 -
827 		PMD_ROUNDUP(RTE_ETHER_MAX_LEN * 2, 1024);
828 	hw->fc.low_water = hw->fc.high_water - 1500;
829 
830 	if (hw->mac.type == e1000_80003es2lan)
831 		hw->fc.pause_time = UINT16_MAX;
832 	else
833 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
834 
835 	hw->fc.send_xon = 1;
836 
837 	/* Set Flow control, use the tunable location if sane */
838 	if (em_fc_setting <= e1000_fc_full)
839 		hw->fc.requested_mode = em_fc_setting;
840 	else
841 		hw->fc.requested_mode = e1000_fc_none;
842 
843 	/* Workaround: no TX flow ctrl for PCH */
844 	if (hw->mac.type == e1000_pchlan)
845 		hw->fc.requested_mode = e1000_fc_rx_pause;
846 
847 	/* Override - settings for PCH2LAN, ya its magic :) */
848 	if (hw->mac.type == e1000_pch2lan) {
849 		hw->fc.high_water = 0x5C20;
850 		hw->fc.low_water = 0x5048;
851 		hw->fc.pause_time = 0x0650;
852 		hw->fc.refresh_time = 0x0400;
853 	} else if (hw->mac.type == e1000_pch_lpt ||
854 		   hw->mac.type == e1000_pch_spt ||
855 		   hw->mac.type == e1000_pch_cnp) {
856 		hw->fc.requested_mode = e1000_fc_full;
857 	}
858 
859 	diag = e1000_init_hw(hw);
860 	if (diag < 0)
861 		return diag;
862 	e1000_check_for_link(hw);
863 	return 0;
864 }
865 
866 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
867 static int
868 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
869 {
870 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
871 	struct e1000_hw_stats *stats =
872 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
873 	int pause_frames;
874 
875 	if(hw->phy.media_type == e1000_media_type_copper ||
876 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
877 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
878 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
879 	}
880 
881 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
882 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
883 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
884 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
885 
886 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
887 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
888 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
889 	stats->dc += E1000_READ_REG(hw, E1000_DC);
890 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
891 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
892 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
893 
894 	/*
895 	 * For watchdog management we need to know if we have been
896 	 * paused during the last interval, so capture that here.
897 	 */
898 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
899 	stats->xoffrxc += pause_frames;
900 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
901 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
902 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
903 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
904 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
905 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
906 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
907 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
908 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
909 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
910 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
911 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
912 
913 	/*
914 	 * For the 64-bit byte counters the low dword must be read first.
915 	 * Both registers clear on the read of the high dword.
916 	 */
917 
918 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
919 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
920 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
921 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
922 
923 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
924 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
925 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
926 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
927 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
928 
929 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
930 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
931 
932 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
933 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
934 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
935 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
936 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
937 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
938 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
939 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
940 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
941 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
942 
943 	/* Interrupt Counts */
944 
945 	if (hw->mac.type >= e1000_82571) {
946 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
947 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
948 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
949 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
950 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
951 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
952 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
953 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
954 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
955 	}
956 
957 	if (hw->mac.type >= e1000_82543) {
958 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
959 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
960 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
961 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
962 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
963 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
964 	}
965 
966 	if (rte_stats == NULL)
967 		return -EINVAL;
968 
969 	/* Rx Errors */
970 	rte_stats->imissed = stats->mpc;
971 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
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 int
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 	return 0;
997 }
998 
999 static int
1000 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1001 {
1002 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1003 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1004 	struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
1005 
1006 	em_rxq_intr_enable(hw);
1007 	rte_intr_ack(intr_handle);
1008 
1009 	return 0;
1010 }
1011 
1012 static int
1013 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1014 {
1015 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1016 
1017 	em_rxq_intr_disable(hw);
1018 
1019 	return 0;
1020 }
1021 
1022 uint32_t
1023 em_get_max_pktlen(struct rte_eth_dev *dev)
1024 {
1025 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1026 
1027 	switch (hw->mac.type) {
1028 	case e1000_82571:
1029 	case e1000_82572:
1030 	case e1000_ich9lan:
1031 	case e1000_ich10lan:
1032 	case e1000_pch2lan:
1033 	case e1000_pch_lpt:
1034 	case e1000_pch_spt:
1035 	case e1000_pch_cnp:
1036 	case e1000_82574:
1037 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1038 	case e1000_82583:
1039 		return 0x2412;
1040 	case e1000_pchlan:
1041 		return 0x1000;
1042 	/* Adapters that do not support jumbo frames */
1043 	case e1000_ich8lan:
1044 		return RTE_ETHER_MAX_LEN;
1045 	default:
1046 		return MAX_JUMBO_FRAME_SIZE;
1047 	}
1048 }
1049 
1050 static int
1051 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1052 {
1053 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1054 
1055 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1056 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1057 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1058 
1059 	/*
1060 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1061 	 * Unfortunatelly, all these nics have just one TX context.
1062 	 * So we have few choises for TX:
1063 	 * - Use just one TX queue.
1064 	 * - Allow cksum offload only for one TX queue.
1065 	 * - Don't allow TX cksum offload at all.
1066 	 * For now, option #1 was chosen.
1067 	 * To use second RX queue we have to use extended RX descriptor
1068 	 * (Multiple Receive Queues are mutually exclusive with UDP
1069 	 * fragmentation and are not supported when a legacy receive
1070 	 * descriptor format is used).
1071 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1072 	 * don't support extended RXD.
1073 	 * To avoid it we support just one RX queue for now (no RSS).
1074 	 */
1075 
1076 	dev_info->max_rx_queues = 1;
1077 	dev_info->max_tx_queues = 1;
1078 
1079 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa();
1080 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa() |
1081 				    dev_info->rx_queue_offload_capa;
1082 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1083 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1084 				    dev_info->tx_queue_offload_capa;
1085 
1086 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1087 		.nb_max = E1000_MAX_RING_DESC,
1088 		.nb_min = E1000_MIN_RING_DESC,
1089 		.nb_align = EM_RXD_ALIGN,
1090 	};
1091 
1092 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1093 		.nb_max = E1000_MAX_RING_DESC,
1094 		.nb_min = E1000_MIN_RING_DESC,
1095 		.nb_align = EM_TXD_ALIGN,
1096 		.nb_seg_max = EM_TX_MAX_SEG,
1097 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1098 	};
1099 
1100 	dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
1101 			RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
1102 			RTE_ETH_LINK_SPEED_1G;
1103 
1104 	/* Preferred queue parameters */
1105 	dev_info->default_rxportconf.nb_queues = 1;
1106 	dev_info->default_txportconf.nb_queues = 1;
1107 	dev_info->default_txportconf.ring_size = 256;
1108 	dev_info->default_rxportconf.ring_size = 256;
1109 
1110 	return 0;
1111 }
1112 
1113 /* return 0 means link status changed, -1 means not changed */
1114 static int
1115 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1116 {
1117 	struct e1000_hw *hw =
1118 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1119 	struct rte_eth_link link;
1120 	int link_up, count;
1121 
1122 	link_up = 0;
1123 	hw->mac.get_link_status = 1;
1124 
1125 	/* possible wait-to-complete in up to 9 seconds */
1126 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1127 		/* Read the real link status */
1128 		switch (hw->phy.media_type) {
1129 		case e1000_media_type_copper:
1130 			/* Do the work to read phy */
1131 			e1000_check_for_link(hw);
1132 			link_up = !hw->mac.get_link_status;
1133 			break;
1134 
1135 		case e1000_media_type_fiber:
1136 			e1000_check_for_link(hw);
1137 			link_up = (E1000_READ_REG(hw, E1000_STATUS) &
1138 					E1000_STATUS_LU);
1139 			break;
1140 
1141 		case e1000_media_type_internal_serdes:
1142 			e1000_check_for_link(hw);
1143 			link_up = hw->mac.serdes_has_link;
1144 			break;
1145 
1146 		default:
1147 			break;
1148 		}
1149 		if (link_up || wait_to_complete == 0)
1150 			break;
1151 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1152 	}
1153 	memset(&link, 0, sizeof(link));
1154 
1155 	/* Now we check if a transition has happened */
1156 	if (link_up) {
1157 		uint16_t duplex, speed;
1158 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1159 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1160 				RTE_ETH_LINK_FULL_DUPLEX :
1161 				RTE_ETH_LINK_HALF_DUPLEX;
1162 		link.link_speed = speed;
1163 		link.link_status = RTE_ETH_LINK_UP;
1164 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1165 				RTE_ETH_LINK_SPEED_FIXED);
1166 	} else {
1167 		link.link_speed = RTE_ETH_SPEED_NUM_NONE;
1168 		link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
1169 		link.link_status = RTE_ETH_LINK_DOWN;
1170 		link.link_autoneg = RTE_ETH_LINK_FIXED;
1171 	}
1172 
1173 	return rte_eth_linkstatus_set(dev, &link);
1174 }
1175 
1176 /*
1177  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1178  * For ASF and Pass Through versions of f/w this means
1179  * that the driver is loaded. For AMT version type f/w
1180  * this means that the network i/f is open.
1181  */
1182 static void
1183 em_hw_control_acquire(struct e1000_hw *hw)
1184 {
1185 	uint32_t ctrl_ext, swsm;
1186 
1187 	/* Let firmware know the driver has taken over */
1188 	if (hw->mac.type == e1000_82573) {
1189 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1190 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1191 
1192 	} else {
1193 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1194 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1195 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1196 	}
1197 }
1198 
1199 /*
1200  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1201  * For ASF and Pass Through versions of f/w this means that the
1202  * driver is no longer loaded. For AMT versions of the
1203  * f/w this means that the network i/f is closed.
1204  */
1205 static void
1206 em_hw_control_release(struct e1000_hw *hw)
1207 {
1208 	uint32_t ctrl_ext, swsm;
1209 
1210 	/* Let firmware taken over control of h/w */
1211 	if (hw->mac.type == e1000_82573) {
1212 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1213 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1214 	} else {
1215 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1216 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1217 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1218 	}
1219 }
1220 
1221 /*
1222  * Bit of a misnomer, what this really means is
1223  * to enable OS management of the system... aka
1224  * to disable special hardware management features.
1225  */
1226 static void
1227 em_init_manageability(struct e1000_hw *hw)
1228 {
1229 	if (e1000_enable_mng_pass_thru(hw)) {
1230 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1231 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1232 
1233 		/* disable hardware interception of ARP */
1234 		manc &= ~(E1000_MANC_ARP_EN);
1235 
1236 		/* enable receiving management packets to the host */
1237 		manc |= E1000_MANC_EN_MNG2HOST;
1238 		manc2h |= 1 << 5;  /* Mng Port 623 */
1239 		manc2h |= 1 << 6;  /* Mng Port 664 */
1240 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1241 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1242 	}
1243 }
1244 
1245 /*
1246  * Give control back to hardware management
1247  * controller if there is one.
1248  */
1249 static void
1250 em_release_manageability(struct e1000_hw *hw)
1251 {
1252 	uint32_t manc;
1253 
1254 	if (e1000_enable_mng_pass_thru(hw)) {
1255 		manc = E1000_READ_REG(hw, E1000_MANC);
1256 
1257 		/* re-enable hardware interception of ARP */
1258 		manc |= E1000_MANC_ARP_EN;
1259 		manc &= ~E1000_MANC_EN_MNG2HOST;
1260 
1261 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1262 	}
1263 }
1264 
1265 static int
1266 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1267 {
1268 	struct e1000_hw *hw =
1269 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1270 	uint32_t rctl;
1271 
1272 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1273 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1274 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1275 
1276 	return 0;
1277 }
1278 
1279 static int
1280 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1281 {
1282 	struct e1000_hw *hw =
1283 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1284 	uint32_t rctl;
1285 
1286 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1287 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1288 	if (dev->data->all_multicast == 1)
1289 		rctl |= E1000_RCTL_MPE;
1290 	else
1291 		rctl &= (~E1000_RCTL_MPE);
1292 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1293 
1294 	return 0;
1295 }
1296 
1297 static int
1298 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1299 {
1300 	struct e1000_hw *hw =
1301 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1302 	uint32_t rctl;
1303 
1304 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1305 	rctl |= E1000_RCTL_MPE;
1306 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1307 
1308 	return 0;
1309 }
1310 
1311 static int
1312 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1313 {
1314 	struct e1000_hw *hw =
1315 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1316 	uint32_t rctl;
1317 
1318 	if (dev->data->promiscuous == 1)
1319 		return 0; /* must remain in all_multicast mode */
1320 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1321 	rctl &= (~E1000_RCTL_MPE);
1322 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1323 
1324 	return 0;
1325 }
1326 
1327 static int
1328 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1329 {
1330 	struct e1000_hw *hw =
1331 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1332 	struct e1000_vfta * shadow_vfta =
1333 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1334 	uint32_t vfta;
1335 	uint32_t vid_idx;
1336 	uint32_t vid_bit;
1337 
1338 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1339 			      E1000_VFTA_ENTRY_MASK);
1340 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1341 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1342 	if (on)
1343 		vfta |= vid_bit;
1344 	else
1345 		vfta &= ~vid_bit;
1346 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1347 
1348 	/* update local VFTA copy */
1349 	shadow_vfta->vfta[vid_idx] = vfta;
1350 
1351 	return 0;
1352 }
1353 
1354 static void
1355 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1356 {
1357 	struct e1000_hw *hw =
1358 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1359 	uint32_t reg;
1360 
1361 	/* Filter Table Disable */
1362 	reg = E1000_READ_REG(hw, E1000_RCTL);
1363 	reg &= ~E1000_RCTL_CFIEN;
1364 	reg &= ~E1000_RCTL_VFE;
1365 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1366 }
1367 
1368 static void
1369 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1370 {
1371 	struct e1000_hw *hw =
1372 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1373 	struct e1000_vfta * shadow_vfta =
1374 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1375 	uint32_t reg;
1376 	int i;
1377 
1378 	/* Filter Table Enable, CFI not used for packet acceptance */
1379 	reg = E1000_READ_REG(hw, E1000_RCTL);
1380 	reg &= ~E1000_RCTL_CFIEN;
1381 	reg |= E1000_RCTL_VFE;
1382 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1383 
1384 	/* restore vfta from local copy */
1385 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1386 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1387 }
1388 
1389 static void
1390 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1391 {
1392 	struct e1000_hw *hw =
1393 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1394 	uint32_t reg;
1395 
1396 	/* VLAN Mode Disable */
1397 	reg = E1000_READ_REG(hw, E1000_CTRL);
1398 	reg &= ~E1000_CTRL_VME;
1399 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1400 
1401 }
1402 
1403 static void
1404 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1405 {
1406 	struct e1000_hw *hw =
1407 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1408 	uint32_t reg;
1409 
1410 	/* VLAN Mode Enable */
1411 	reg = E1000_READ_REG(hw, E1000_CTRL);
1412 	reg |= E1000_CTRL_VME;
1413 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1414 }
1415 
1416 static int
1417 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1418 {
1419 	struct rte_eth_rxmode *rxmode;
1420 
1421 	rxmode = &dev->data->dev_conf.rxmode;
1422 	if (mask & RTE_ETH_VLAN_STRIP_MASK) {
1423 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
1424 			em_vlan_hw_strip_enable(dev);
1425 		else
1426 			em_vlan_hw_strip_disable(dev);
1427 	}
1428 
1429 	if (mask & RTE_ETH_VLAN_FILTER_MASK) {
1430 		if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
1431 			em_vlan_hw_filter_enable(dev);
1432 		else
1433 			em_vlan_hw_filter_disable(dev);
1434 	}
1435 
1436 	return 0;
1437 }
1438 
1439 /*
1440  * It enables the interrupt mask and then enable the interrupt.
1441  *
1442  * @param dev
1443  *  Pointer to struct rte_eth_dev.
1444  *
1445  * @return
1446  *  - On success, zero.
1447  *  - On failure, a negative value.
1448  */
1449 static int
1450 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1451 {
1452 	uint32_t regval;
1453 	struct e1000_hw *hw =
1454 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1455 
1456 	/* clear interrupt */
1457 	E1000_READ_REG(hw, E1000_ICR);
1458 	regval = E1000_READ_REG(hw, E1000_IMS);
1459 	E1000_WRITE_REG(hw, E1000_IMS,
1460 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1461 	return 0;
1462 }
1463 
1464 /*
1465  * It clears the interrupt causes and enables the interrupt.
1466  * It will be called once only during nic initialized.
1467  *
1468  * @param dev
1469  *  Pointer to struct rte_eth_dev.
1470  *
1471  * @return
1472  *  - On success, zero.
1473  *  - On failure, a negative value.
1474  */
1475 static int
1476 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1477 {
1478 	struct e1000_hw *hw =
1479 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1480 
1481 	E1000_READ_REG(hw, E1000_ICR);
1482 	em_rxq_intr_enable(hw);
1483 	return 0;
1484 }
1485 
1486 /*
1487  * It enable receive packet interrupt.
1488  * @param hw
1489  * Pointer to struct e1000_hw
1490  *
1491  * @return
1492  */
1493 static void
1494 em_rxq_intr_enable(struct e1000_hw *hw)
1495 {
1496 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1497 	E1000_WRITE_FLUSH(hw);
1498 }
1499 
1500 /*
1501  * It disabled lsc interrupt.
1502  * @param hw
1503  * Pointer to struct e1000_hw
1504  *
1505  * @return
1506  */
1507 static void
1508 em_lsc_intr_disable(struct e1000_hw *hw)
1509 {
1510 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1511 	E1000_WRITE_FLUSH(hw);
1512 }
1513 
1514 /*
1515  * It disabled receive packet interrupt.
1516  * @param hw
1517  * Pointer to struct e1000_hw
1518  *
1519  * @return
1520  */
1521 static void
1522 em_rxq_intr_disable(struct e1000_hw *hw)
1523 {
1524 	E1000_READ_REG(hw, E1000_ICR);
1525 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1526 	E1000_WRITE_FLUSH(hw);
1527 }
1528 
1529 /*
1530  * It reads ICR and gets interrupt causes, check it and set a bit flag
1531  * to update link status.
1532  *
1533  * @param dev
1534  *  Pointer to struct rte_eth_dev.
1535  *
1536  * @return
1537  *  - On success, zero.
1538  *  - On failure, a negative value.
1539  */
1540 static int
1541 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1542 {
1543 	uint32_t icr;
1544 	struct e1000_hw *hw =
1545 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1546 	struct e1000_interrupt *intr =
1547 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1548 
1549 	/* read-on-clear nic registers here */
1550 	icr = E1000_READ_REG(hw, E1000_ICR);
1551 	if (icr & E1000_ICR_LSC) {
1552 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1553 	}
1554 
1555 	return 0;
1556 }
1557 
1558 /*
1559  * It executes link_update after knowing an interrupt is prsent.
1560  *
1561  * @param dev
1562  *  Pointer to struct rte_eth_dev.
1563  *
1564  * @return
1565  *  - On success, zero.
1566  *  - On failure, a negative value.
1567  */
1568 static int
1569 eth_em_interrupt_action(struct rte_eth_dev *dev,
1570 			struct rte_intr_handle *intr_handle)
1571 {
1572 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1573 	struct e1000_hw *hw =
1574 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1575 	struct e1000_interrupt *intr =
1576 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1577 	struct rte_eth_link link;
1578 	int ret;
1579 
1580 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1581 		return -1;
1582 
1583 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1584 	rte_intr_ack(intr_handle);
1585 
1586 	/* set get_link_status to check register later */
1587 	hw->mac.get_link_status = 1;
1588 	ret = eth_em_link_update(dev, 0);
1589 
1590 	/* check if link has changed */
1591 	if (ret < 0)
1592 		return 0;
1593 
1594 	rte_eth_linkstatus_get(dev, &link);
1595 
1596 	if (link.link_status) {
1597 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1598 			     dev->data->port_id, link.link_speed,
1599 			     link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
1600 			     "full-duplex" : "half-duplex");
1601 	} else {
1602 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1603 	}
1604 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
1605 		     pci_dev->addr.domain, pci_dev->addr.bus,
1606 		     pci_dev->addr.devid, pci_dev->addr.function);
1607 
1608 	return 0;
1609 }
1610 
1611 /**
1612  * Interrupt handler which shall be registered at first.
1613  *
1614  * @param handle
1615  *  Pointer to interrupt handle.
1616  * @param param
1617  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1618  *
1619  * @return
1620  *  void
1621  */
1622 static void
1623 eth_em_interrupt_handler(void *param)
1624 {
1625 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1626 
1627 	eth_em_interrupt_get_status(dev);
1628 	eth_em_interrupt_action(dev, dev->intr_handle);
1629 	rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1630 }
1631 
1632 static int
1633 eth_em_led_on(struct rte_eth_dev *dev)
1634 {
1635 	struct e1000_hw *hw;
1636 
1637 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1638 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1639 }
1640 
1641 static int
1642 eth_em_led_off(struct rte_eth_dev *dev)
1643 {
1644 	struct e1000_hw *hw;
1645 
1646 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1647 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1648 }
1649 
1650 static int
1651 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1652 {
1653 	struct e1000_hw *hw;
1654 	uint32_t ctrl;
1655 	int tx_pause;
1656 	int rx_pause;
1657 
1658 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1659 	fc_conf->pause_time = hw->fc.pause_time;
1660 	fc_conf->high_water = hw->fc.high_water;
1661 	fc_conf->low_water = hw->fc.low_water;
1662 	fc_conf->send_xon = hw->fc.send_xon;
1663 	fc_conf->autoneg = hw->mac.autoneg;
1664 
1665 	/*
1666 	 * Return rx_pause and tx_pause status according to actual setting of
1667 	 * the TFCE and RFCE bits in the CTRL register.
1668 	 */
1669 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1670 	if (ctrl & E1000_CTRL_TFCE)
1671 		tx_pause = 1;
1672 	else
1673 		tx_pause = 0;
1674 
1675 	if (ctrl & E1000_CTRL_RFCE)
1676 		rx_pause = 1;
1677 	else
1678 		rx_pause = 0;
1679 
1680 	if (rx_pause && tx_pause)
1681 		fc_conf->mode = RTE_ETH_FC_FULL;
1682 	else if (rx_pause)
1683 		fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
1684 	else if (tx_pause)
1685 		fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
1686 	else
1687 		fc_conf->mode = RTE_ETH_FC_NONE;
1688 
1689 	return 0;
1690 }
1691 
1692 static int
1693 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1694 {
1695 	struct e1000_hw *hw;
1696 	int err;
1697 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1698 		e1000_fc_none,
1699 		e1000_fc_rx_pause,
1700 		e1000_fc_tx_pause,
1701 		e1000_fc_full
1702 	};
1703 	uint32_t rx_buf_size;
1704 	uint32_t max_high_water;
1705 	uint32_t rctl;
1706 
1707 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1708 	if (fc_conf->autoneg != hw->mac.autoneg)
1709 		return -ENOTSUP;
1710 	rx_buf_size = em_get_rx_buffer_size(hw);
1711 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1712 
1713 	/* At least reserve one Ethernet frame for watermark */
1714 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
1715 	if ((fc_conf->high_water > max_high_water) ||
1716 	    (fc_conf->high_water < fc_conf->low_water)) {
1717 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1718 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1719 		return -EINVAL;
1720 	}
1721 
1722 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1723 	hw->fc.pause_time     = fc_conf->pause_time;
1724 	hw->fc.high_water     = fc_conf->high_water;
1725 	hw->fc.low_water      = fc_conf->low_water;
1726 	hw->fc.send_xon	      = fc_conf->send_xon;
1727 
1728 	err = e1000_setup_link_generic(hw);
1729 	if (err == E1000_SUCCESS) {
1730 
1731 		/* check if we want to forward MAC frames - driver doesn't have native
1732 		 * capability to do that, so we'll write the registers ourselves */
1733 
1734 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1735 
1736 		/* set or clear MFLCN.PMCF bit depending on configuration */
1737 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1738 			rctl |= E1000_RCTL_PMCF;
1739 		else
1740 			rctl &= ~E1000_RCTL_PMCF;
1741 
1742 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1743 		E1000_WRITE_FLUSH(hw);
1744 
1745 		return 0;
1746 	}
1747 
1748 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1749 	return -EIO;
1750 }
1751 
1752 static int
1753 eth_em_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1754 		uint32_t index, __rte_unused uint32_t pool)
1755 {
1756 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1757 
1758 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1759 }
1760 
1761 static void
1762 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1763 {
1764 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1765 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1766 
1767 	memset(addr, 0, sizeof(addr));
1768 
1769 	e1000_rar_set(hw, addr, index);
1770 }
1771 
1772 static int
1773 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1774 			    struct rte_ether_addr *addr)
1775 {
1776 	eth_em_rar_clear(dev, 0);
1777 
1778 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1779 }
1780 
1781 static int
1782 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1783 {
1784 	struct e1000_hw *hw;
1785 	uint32_t frame_size;
1786 	uint32_t rctl;
1787 
1788 	frame_size = mtu + E1000_ETH_OVERHEAD;
1789 
1790 	/*
1791 	 * If device is started, refuse mtu that requires the support of
1792 	 * scattered packets when this feature has not been enabled before.
1793 	 */
1794 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1795 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1796 		PMD_INIT_LOG(ERR, "Stop port first.");
1797 		return -EINVAL;
1798 	}
1799 
1800 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1801 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1802 
1803 	/* switch to jumbo mode if needed */
1804 	if (mtu > RTE_ETHER_MTU)
1805 		rctl |= E1000_RCTL_LPE;
1806 	else
1807 		rctl &= ~E1000_RCTL_LPE;
1808 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1809 
1810 	return 0;
1811 }
1812 
1813 static int
1814 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1815 			struct rte_ether_addr *mc_addr_set,
1816 			uint32_t nb_mc_addr)
1817 {
1818 	struct e1000_hw *hw;
1819 
1820 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1821 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1822 	return 0;
1823 }
1824 
1825 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1826 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1827 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1828