xref: /dpdk/drivers/net/e1000/em_ethdev.c (revision ff4e52ef)
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 	eth_dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
268 
269 	hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
270 	hw->device_id = pci_dev->id.device_id;
271 	adapter->stopped = 0;
272 
273 	/* For ICH8 support we'll need to map the flash memory BAR */
274 	if (eth_em_dev_is_ich8(hw))
275 		hw->flash_address = (void *)pci_dev->mem_resource[1].addr;
276 
277 	if (e1000_setup_init_funcs(hw, TRUE) != E1000_SUCCESS ||
278 			em_hw_init(hw) != 0) {
279 		PMD_INIT_LOG(ERR, "port_id %d vendorID=0x%x deviceID=0x%x: "
280 			"failed to init HW",
281 			eth_dev->data->port_id, pci_dev->id.vendor_id,
282 			pci_dev->id.device_id);
283 		return -ENODEV;
284 	}
285 
286 	/* Allocate memory for storing MAC addresses */
287 	eth_dev->data->mac_addrs = rte_zmalloc("e1000", RTE_ETHER_ADDR_LEN *
288 			hw->mac.rar_entry_count, 0);
289 	if (eth_dev->data->mac_addrs == NULL) {
290 		PMD_INIT_LOG(ERR, "Failed to allocate %d bytes needed to "
291 			"store MAC addresses",
292 			RTE_ETHER_ADDR_LEN * hw->mac.rar_entry_count);
293 		return -ENOMEM;
294 	}
295 
296 	/* Copy the permanent MAC address */
297 	rte_ether_addr_copy((struct rte_ether_addr *)hw->mac.addr,
298 		eth_dev->data->mac_addrs);
299 
300 	/* initialize the vfta */
301 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
302 
303 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
304 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
305 		     pci_dev->id.device_id);
306 
307 	rte_intr_callback_register(intr_handle,
308 				   eth_em_interrupt_handler, eth_dev);
309 
310 	return 0;
311 }
312 
313 static int
314 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
315 {
316 	PMD_INIT_FUNC_TRACE();
317 
318 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
319 		return 0;
320 
321 	eth_em_close(eth_dev);
322 
323 	return 0;
324 }
325 
326 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
327 	struct rte_pci_device *pci_dev)
328 {
329 	return rte_eth_dev_pci_generic_probe(pci_dev,
330 		sizeof(struct e1000_adapter), eth_em_dev_init);
331 }
332 
333 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
334 {
335 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
336 }
337 
338 static struct rte_pci_driver rte_em_pmd = {
339 	.id_table = pci_id_em_map,
340 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
341 	.probe = eth_em_pci_probe,
342 	.remove = eth_em_pci_remove,
343 };
344 
345 static int
346 em_hw_init(struct e1000_hw *hw)
347 {
348 	int diag;
349 
350 	diag = hw->mac.ops.init_params(hw);
351 	if (diag != 0) {
352 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
353 		return diag;
354 	}
355 	diag = hw->nvm.ops.init_params(hw);
356 	if (diag != 0) {
357 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
358 		return diag;
359 	}
360 	diag = hw->phy.ops.init_params(hw);
361 	if (diag != 0) {
362 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
363 		return diag;
364 	}
365 	(void) e1000_get_bus_info(hw);
366 
367 	hw->mac.autoneg = 1;
368 	hw->phy.autoneg_wait_to_complete = 0;
369 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
370 
371 	e1000_init_script_state_82541(hw, TRUE);
372 	e1000_set_tbi_compatibility_82543(hw, TRUE);
373 
374 	/* Copper options */
375 	if (hw->phy.media_type == e1000_media_type_copper) {
376 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
377 		hw->phy.disable_polarity_correction = 0;
378 		hw->phy.ms_type = e1000_ms_hw_default;
379 	}
380 
381 	/*
382 	 * Start from a known state, this is important in reading the nvm
383 	 * and mac from that.
384 	 */
385 	e1000_reset_hw(hw);
386 
387 	/* Make sure we have a good EEPROM before we read from it */
388 	if (e1000_validate_nvm_checksum(hw) < 0) {
389 		/*
390 		 * Some PCI-E parts fail the first check due to
391 		 * the link being in sleep state, call it again,
392 		 * if it fails a second time its a real issue.
393 		 */
394 		diag = e1000_validate_nvm_checksum(hw);
395 		if (diag < 0) {
396 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
397 			goto error;
398 		}
399 	}
400 
401 	/* Read the permanent MAC address out of the EEPROM */
402 	diag = e1000_read_mac_addr(hw);
403 	if (diag != 0) {
404 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
405 		goto error;
406 	}
407 
408 	/* Now initialize the hardware */
409 	diag = em_hardware_init(hw);
410 	if (diag != 0) {
411 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
412 		goto error;
413 	}
414 
415 	hw->mac.get_link_status = 1;
416 
417 	/* Indicate SOL/IDER usage */
418 	diag = e1000_check_reset_block(hw);
419 	if (diag < 0) {
420 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
421 			"SOL/IDER session");
422 	}
423 	return 0;
424 
425 error:
426 	em_hw_control_release(hw);
427 	return diag;
428 }
429 
430 static int
431 eth_em_configure(struct rte_eth_dev *dev)
432 {
433 	struct e1000_interrupt *intr =
434 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
435 
436 	PMD_INIT_FUNC_TRACE();
437 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
438 
439 	PMD_INIT_FUNC_TRACE();
440 
441 	return 0;
442 }
443 
444 static void
445 em_set_pba(struct e1000_hw *hw)
446 {
447 	uint32_t pba;
448 
449 	/*
450 	 * Packet Buffer Allocation (PBA)
451 	 * Writing PBA sets the receive portion of the buffer
452 	 * the remainder is used for the transmit buffer.
453 	 * Devices before the 82547 had a Packet Buffer of 64K.
454 	 * After the 82547 the buffer was reduced to 40K.
455 	 */
456 	switch (hw->mac.type) {
457 		case e1000_82547:
458 		case e1000_82547_rev_2:
459 		/* 82547: Total Packet Buffer is 40K */
460 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
461 			break;
462 		case e1000_82571:
463 		case e1000_82572:
464 		case e1000_80003es2lan:
465 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
466 			break;
467 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
468 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
469 			break;
470 		case e1000_82574:
471 		case e1000_82583:
472 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
473 			break;
474 		case e1000_ich8lan:
475 			pba = E1000_PBA_8K;
476 			break;
477 		case e1000_ich9lan:
478 		case e1000_ich10lan:
479 			pba = E1000_PBA_10K;
480 			break;
481 		case e1000_pchlan:
482 		case e1000_pch2lan:
483 		case e1000_pch_lpt:
484 		case e1000_pch_spt:
485 		case e1000_pch_cnp:
486 			pba = E1000_PBA_26K;
487 			break;
488 		default:
489 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
490 	}
491 
492 	E1000_WRITE_REG(hw, E1000_PBA, pba);
493 }
494 
495 static void
496 eth_em_rxtx_control(struct rte_eth_dev *dev,
497 		    bool enable)
498 {
499 	struct e1000_hw *hw =
500 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
501 	uint32_t tctl, rctl;
502 
503 	tctl = E1000_READ_REG(hw, E1000_TCTL);
504 	rctl = E1000_READ_REG(hw, E1000_RCTL);
505 	if (enable) {
506 		/* enable Tx/Rx */
507 		tctl |= E1000_TCTL_EN;
508 		rctl |= E1000_RCTL_EN;
509 	} else {
510 		/* disable Tx/Rx */
511 		tctl &= ~E1000_TCTL_EN;
512 		rctl &= ~E1000_RCTL_EN;
513 	}
514 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
515 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
516 	E1000_WRITE_FLUSH(hw);
517 }
518 
519 static int
520 eth_em_start(struct rte_eth_dev *dev)
521 {
522 	struct e1000_adapter *adapter =
523 		E1000_DEV_PRIVATE(dev->data->dev_private);
524 	struct e1000_hw *hw =
525 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
526 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
527 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
528 	int ret, mask;
529 	uint32_t intr_vector = 0;
530 	uint32_t *speeds;
531 	int num_speeds;
532 	bool autoneg;
533 
534 	PMD_INIT_FUNC_TRACE();
535 
536 	ret = eth_em_stop(dev);
537 	if (ret != 0)
538 		return ret;
539 
540 	e1000_power_up_phy(hw);
541 
542 	/* Set default PBA value */
543 	em_set_pba(hw);
544 
545 	/* Put the address into the Receive Address Array */
546 	e1000_rar_set(hw, hw->mac.addr, 0);
547 
548 	/*
549 	 * With the 82571 adapter, RAR[0] may be overwritten
550 	 * when the other port is reset, we make a duplicate
551 	 * in RAR[14] for that eventuality, this assures
552 	 * the interface continues to function.
553 	 */
554 	if (hw->mac.type == e1000_82571) {
555 		e1000_set_laa_state_82571(hw, TRUE);
556 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
557 	}
558 
559 	/* Initialize the hardware */
560 	if (em_hardware_init(hw)) {
561 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
562 		return -EIO;
563 	}
564 
565 	E1000_WRITE_REG(hw, E1000_VET, RTE_ETHER_TYPE_VLAN);
566 
567 	/* Configure for OS presence */
568 	em_init_manageability(hw);
569 
570 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
571 		intr_vector = dev->data->nb_rx_queues;
572 		if (rte_intr_efd_enable(intr_handle, intr_vector))
573 			return -1;
574 	}
575 
576 	if (rte_intr_dp_is_en(intr_handle)) {
577 		intr_handle->intr_vec =
578 			rte_zmalloc("intr_vec",
579 					dev->data->nb_rx_queues * sizeof(int), 0);
580 		if (intr_handle->intr_vec == NULL) {
581 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
582 						" intr_vec", dev->data->nb_rx_queues);
583 			return -ENOMEM;
584 		}
585 
586 		/* enable rx interrupt */
587 		em_rxq_intr_enable(hw);
588 	}
589 
590 	eth_em_tx_init(dev);
591 
592 	ret = eth_em_rx_init(dev);
593 	if (ret) {
594 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
595 		em_dev_clear_queues(dev);
596 		return ret;
597 	}
598 
599 	e1000_clear_hw_cntrs_base_generic(hw);
600 
601 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
602 			ETH_VLAN_EXTEND_MASK;
603 	ret = eth_em_vlan_offload_set(dev, mask);
604 	if (ret) {
605 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
606 		em_dev_clear_queues(dev);
607 		return ret;
608 	}
609 
610 	/* Set Interrupt Throttling Rate to maximum allowed value. */
611 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
612 
613 	/* Setup link speed and duplex */
614 	speeds = &dev->data->dev_conf.link_speeds;
615 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
616 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
617 		hw->mac.autoneg = 1;
618 	} else {
619 		num_speeds = 0;
620 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
621 
622 		/* Reset */
623 		hw->phy.autoneg_advertised = 0;
624 
625 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
626 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
627 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
628 			num_speeds = -1;
629 			goto error_invalid_config;
630 		}
631 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
632 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
633 			num_speeds++;
634 		}
635 		if (*speeds & ETH_LINK_SPEED_10M) {
636 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
637 			num_speeds++;
638 		}
639 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
640 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
641 			num_speeds++;
642 		}
643 		if (*speeds & ETH_LINK_SPEED_100M) {
644 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
645 			num_speeds++;
646 		}
647 		if (*speeds & ETH_LINK_SPEED_1G) {
648 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
649 			num_speeds++;
650 		}
651 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
652 			goto error_invalid_config;
653 
654 		/* Set/reset the mac.autoneg based on the link speed,
655 		 * fixed or not
656 		 */
657 		if (!autoneg) {
658 			hw->mac.autoneg = 0;
659 			hw->mac.forced_speed_duplex =
660 					hw->phy.autoneg_advertised;
661 		} else {
662 			hw->mac.autoneg = 1;
663 		}
664 	}
665 
666 	e1000_setup_link(hw);
667 
668 	if (rte_intr_allow_others(intr_handle)) {
669 		/* check if lsc interrupt is enabled */
670 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
671 			ret = eth_em_interrupt_setup(dev);
672 			if (ret) {
673 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
674 				em_dev_clear_queues(dev);
675 				return ret;
676 			}
677 		}
678 	} else {
679 		rte_intr_callback_unregister(intr_handle,
680 						eth_em_interrupt_handler,
681 						(void *)dev);
682 		if (dev->data->dev_conf.intr_conf.lsc != 0)
683 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
684 				     " no intr multiplexn");
685 	}
686 	/* check if rxq interrupt is enabled */
687 	if (dev->data->dev_conf.intr_conf.rxq != 0)
688 		eth_em_rxq_interrupt_setup(dev);
689 
690 	rte_intr_enable(intr_handle);
691 
692 	adapter->stopped = 0;
693 
694 	eth_em_rxtx_control(dev, true);
695 	eth_em_link_update(dev, 0);
696 
697 	PMD_INIT_LOG(DEBUG, "<<");
698 
699 	return 0;
700 
701 error_invalid_config:
702 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
703 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
704 	em_dev_clear_queues(dev);
705 	return -EINVAL;
706 }
707 
708 /*********************************************************************
709  *
710  *  This routine disables all traffic on the adapter by issuing a
711  *  global reset on the MAC.
712  *
713  **********************************************************************/
714 static int
715 eth_em_stop(struct rte_eth_dev *dev)
716 {
717 	struct rte_eth_link link;
718 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
719 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
720 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
721 
722 	dev->data->dev_started = 0;
723 
724 	eth_em_rxtx_control(dev, false);
725 	em_rxq_intr_disable(hw);
726 	em_lsc_intr_disable(hw);
727 
728 	e1000_reset_hw(hw);
729 
730 	/* Flush desc rings for i219 */
731 	if (hw->mac.type == e1000_pch_spt || hw->mac.type == e1000_pch_cnp)
732 		em_flush_desc_rings(dev);
733 
734 	if (hw->mac.type >= e1000_82544)
735 		E1000_WRITE_REG(hw, E1000_WUC, 0);
736 
737 	/* Power down the phy. Needed to make the link go down */
738 	e1000_power_down_phy(hw);
739 
740 	em_dev_clear_queues(dev);
741 
742 	/* clear the recorded link status */
743 	memset(&link, 0, sizeof(link));
744 	rte_eth_linkstatus_set(dev, &link);
745 
746 	if (!rte_intr_allow_others(intr_handle))
747 		/* resume to the default handler */
748 		rte_intr_callback_register(intr_handle,
749 					   eth_em_interrupt_handler,
750 					   (void *)dev);
751 
752 	/* Clean datapath event and queue/vec mapping */
753 	rte_intr_efd_disable(intr_handle);
754 	if (intr_handle->intr_vec != NULL) {
755 		rte_free(intr_handle->intr_vec);
756 		intr_handle->intr_vec = NULL;
757 	}
758 
759 	return 0;
760 }
761 
762 static int
763 eth_em_close(struct rte_eth_dev *dev)
764 {
765 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
766 	struct e1000_adapter *adapter =
767 		E1000_DEV_PRIVATE(dev->data->dev_private);
768 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
769 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
770 	int ret;
771 
772 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
773 		return 0;
774 
775 	ret = eth_em_stop(dev);
776 	adapter->stopped = 1;
777 	em_dev_free_queues(dev);
778 	e1000_phy_hw_reset(hw);
779 	em_release_manageability(hw);
780 	em_hw_control_release(hw);
781 
782 	/* disable uio intr before callback unregister */
783 	rte_intr_disable(intr_handle);
784 	rte_intr_callback_unregister(intr_handle,
785 				     eth_em_interrupt_handler, dev);
786 
787 	return ret;
788 }
789 
790 static int
791 em_get_rx_buffer_size(struct e1000_hw *hw)
792 {
793 	uint32_t rx_buf_size;
794 
795 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
796 	return rx_buf_size;
797 }
798 
799 /*********************************************************************
800  *
801  *  Initialize the hardware
802  *
803  **********************************************************************/
804 static int
805 em_hardware_init(struct e1000_hw *hw)
806 {
807 	uint32_t rx_buf_size;
808 	int diag;
809 
810 	/* Issue a global reset */
811 	e1000_reset_hw(hw);
812 
813 	/* Let the firmware know the OS is in control */
814 	em_hw_control_acquire(hw);
815 
816 	/*
817 	 * These parameters control the automatic generation (Tx) and
818 	 * response (Rx) to Ethernet PAUSE frames.
819 	 * - High water mark should allow for at least two standard size (1518)
820 	 *   frames to be received after sending an XOFF.
821 	 * - Low water mark works best when it is very near the high water mark.
822 	 *   This allows the receiver to restart by sending XON when it has
823 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
824 	 *   restart after one full frame is pulled from the buffer. There
825 	 *   could be several smaller frames in the buffer and if so they will
826 	 *   not trigger the XON until their total number reduces the buffer
827 	 *   by 1500.
828 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
829 	 */
830 	rx_buf_size = em_get_rx_buffer_size(hw);
831 
832 	hw->fc.high_water = rx_buf_size -
833 		PMD_ROUNDUP(RTE_ETHER_MAX_LEN * 2, 1024);
834 	hw->fc.low_water = hw->fc.high_water - 1500;
835 
836 	if (hw->mac.type == e1000_80003es2lan)
837 		hw->fc.pause_time = UINT16_MAX;
838 	else
839 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
840 
841 	hw->fc.send_xon = 1;
842 
843 	/* Set Flow control, use the tunable location if sane */
844 	if (em_fc_setting <= e1000_fc_full)
845 		hw->fc.requested_mode = em_fc_setting;
846 	else
847 		hw->fc.requested_mode = e1000_fc_none;
848 
849 	/* Workaround: no TX flow ctrl for PCH */
850 	if (hw->mac.type == e1000_pchlan)
851 		hw->fc.requested_mode = e1000_fc_rx_pause;
852 
853 	/* Override - settings for PCH2LAN, ya its magic :) */
854 	if (hw->mac.type == e1000_pch2lan) {
855 		hw->fc.high_water = 0x5C20;
856 		hw->fc.low_water = 0x5048;
857 		hw->fc.pause_time = 0x0650;
858 		hw->fc.refresh_time = 0x0400;
859 	} else if (hw->mac.type == e1000_pch_lpt ||
860 		   hw->mac.type == e1000_pch_spt ||
861 		   hw->mac.type == e1000_pch_cnp) {
862 		hw->fc.requested_mode = e1000_fc_full;
863 	}
864 
865 	diag = e1000_init_hw(hw);
866 	if (diag < 0)
867 		return diag;
868 	e1000_check_for_link(hw);
869 	return 0;
870 }
871 
872 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
873 static int
874 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
875 {
876 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
877 	struct e1000_hw_stats *stats =
878 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
879 	int pause_frames;
880 
881 	if(hw->phy.media_type == e1000_media_type_copper ||
882 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
883 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
884 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
885 	}
886 
887 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
888 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
889 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
890 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
891 
892 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
893 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
894 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
895 	stats->dc += E1000_READ_REG(hw, E1000_DC);
896 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
897 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
898 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
899 
900 	/*
901 	 * For watchdog management we need to know if we have been
902 	 * paused during the last interval, so capture that here.
903 	 */
904 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
905 	stats->xoffrxc += pause_frames;
906 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
907 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
908 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
909 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
910 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
911 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
912 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
913 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
914 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
915 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
916 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
917 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
918 
919 	/*
920 	 * For the 64-bit byte counters the low dword must be read first.
921 	 * Both registers clear on the read of the high dword.
922 	 */
923 
924 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
925 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
926 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
927 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
928 
929 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
930 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
931 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
932 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
933 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
934 
935 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
936 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
937 
938 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
939 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
940 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
941 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
942 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
943 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
944 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
945 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
946 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
947 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
948 
949 	/* Interrupt Counts */
950 
951 	if (hw->mac.type >= e1000_82571) {
952 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
953 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
954 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
955 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
956 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
957 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
958 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
959 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
960 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
961 	}
962 
963 	if (hw->mac.type >= e1000_82543) {
964 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
965 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
966 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
967 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
968 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
969 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
970 	}
971 
972 	if (rte_stats == NULL)
973 		return -EINVAL;
974 
975 	/* Rx Errors */
976 	rte_stats->imissed = stats->mpc;
977 	rte_stats->ierrors = stats->crcerrs + stats->rlec +
978 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
979 
980 	/* Tx Errors */
981 	rte_stats->oerrors = stats->ecol + stats->latecol;
982 
983 	rte_stats->ipackets = stats->gprc;
984 	rte_stats->opackets = stats->gptc;
985 	rte_stats->ibytes   = stats->gorc;
986 	rte_stats->obytes   = stats->gotc;
987 	return 0;
988 }
989 
990 static int
991 eth_em_stats_reset(struct rte_eth_dev *dev)
992 {
993 	struct e1000_hw_stats *hw_stats =
994 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
995 
996 	/* HW registers are cleared on read */
997 	eth_em_stats_get(dev, NULL);
998 
999 	/* Reset software totals */
1000 	memset(hw_stats, 0, sizeof(*hw_stats));
1001 
1002 	return 0;
1003 }
1004 
1005 static int
1006 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1007 {
1008 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1009 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1010 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1011 
1012 	em_rxq_intr_enable(hw);
1013 	rte_intr_ack(intr_handle);
1014 
1015 	return 0;
1016 }
1017 
1018 static int
1019 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1020 {
1021 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1022 
1023 	em_rxq_intr_disable(hw);
1024 
1025 	return 0;
1026 }
1027 
1028 uint32_t
1029 em_get_max_pktlen(struct rte_eth_dev *dev)
1030 {
1031 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1032 
1033 	switch (hw->mac.type) {
1034 	case e1000_82571:
1035 	case e1000_82572:
1036 	case e1000_ich9lan:
1037 	case e1000_ich10lan:
1038 	case e1000_pch2lan:
1039 	case e1000_pch_lpt:
1040 	case e1000_pch_spt:
1041 	case e1000_pch_cnp:
1042 	case e1000_82574:
1043 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1044 	case e1000_82583:
1045 		return 0x2412;
1046 	case e1000_pchlan:
1047 		return 0x1000;
1048 	/* Adapters that do not support jumbo frames */
1049 	case e1000_ich8lan:
1050 		return RTE_ETHER_MAX_LEN;
1051 	default:
1052 		return MAX_JUMBO_FRAME_SIZE;
1053 	}
1054 }
1055 
1056 static int
1057 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1058 {
1059 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1060 
1061 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1062 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1063 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1064 
1065 	/*
1066 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1067 	 * Unfortunatelly, all these nics have just one TX context.
1068 	 * So we have few choises for TX:
1069 	 * - Use just one TX queue.
1070 	 * - Allow cksum offload only for one TX queue.
1071 	 * - Don't allow TX cksum offload at all.
1072 	 * For now, option #1 was chosen.
1073 	 * To use second RX queue we have to use extended RX descriptor
1074 	 * (Multiple Receive Queues are mutually exclusive with UDP
1075 	 * fragmentation and are not supported when a legacy receive
1076 	 * descriptor format is used).
1077 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1078 	 * don't support extended RXD.
1079 	 * To avoid it we support just one RX queue for now (no RSS).
1080 	 */
1081 
1082 	dev_info->max_rx_queues = 1;
1083 	dev_info->max_tx_queues = 1;
1084 
1085 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1086 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1087 				    dev_info->rx_queue_offload_capa;
1088 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1089 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1090 				    dev_info->tx_queue_offload_capa;
1091 
1092 	dev_info->rx_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_RXD_ALIGN,
1096 	};
1097 
1098 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1099 		.nb_max = E1000_MAX_RING_DESC,
1100 		.nb_min = E1000_MIN_RING_DESC,
1101 		.nb_align = EM_TXD_ALIGN,
1102 		.nb_seg_max = EM_TX_MAX_SEG,
1103 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1104 	};
1105 
1106 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1107 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1108 			ETH_LINK_SPEED_1G;
1109 
1110 	/* Preferred queue parameters */
1111 	dev_info->default_rxportconf.nb_queues = 1;
1112 	dev_info->default_txportconf.nb_queues = 1;
1113 	dev_info->default_txportconf.ring_size = 256;
1114 	dev_info->default_rxportconf.ring_size = 256;
1115 
1116 	return 0;
1117 }
1118 
1119 /* return 0 means link status changed, -1 means not changed */
1120 static int
1121 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1122 {
1123 	struct e1000_hw *hw =
1124 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1125 	struct rte_eth_link link;
1126 	int link_up, count;
1127 
1128 	link_up = 0;
1129 	hw->mac.get_link_status = 1;
1130 
1131 	/* possible wait-to-complete in up to 9 seconds */
1132 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1133 		/* Read the real link status */
1134 		switch (hw->phy.media_type) {
1135 		case e1000_media_type_copper:
1136 			/* Do the work to read phy */
1137 			e1000_check_for_link(hw);
1138 			link_up = !hw->mac.get_link_status;
1139 			break;
1140 
1141 		case e1000_media_type_fiber:
1142 			e1000_check_for_link(hw);
1143 			link_up = (E1000_READ_REG(hw, E1000_STATUS) &
1144 					E1000_STATUS_LU);
1145 			break;
1146 
1147 		case e1000_media_type_internal_serdes:
1148 			e1000_check_for_link(hw);
1149 			link_up = hw->mac.serdes_has_link;
1150 			break;
1151 
1152 		default:
1153 			break;
1154 		}
1155 		if (link_up || wait_to_complete == 0)
1156 			break;
1157 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1158 	}
1159 	memset(&link, 0, sizeof(link));
1160 
1161 	/* Now we check if a transition has happened */
1162 	if (link_up) {
1163 		uint16_t duplex, speed;
1164 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1165 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1166 				ETH_LINK_FULL_DUPLEX :
1167 				ETH_LINK_HALF_DUPLEX;
1168 		link.link_speed = speed;
1169 		link.link_status = ETH_LINK_UP;
1170 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1171 				ETH_LINK_SPEED_FIXED);
1172 	} else {
1173 		link.link_speed = ETH_SPEED_NUM_NONE;
1174 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1175 		link.link_status = ETH_LINK_DOWN;
1176 		link.link_autoneg = ETH_LINK_FIXED;
1177 	}
1178 
1179 	return rte_eth_linkstatus_set(dev, &link);
1180 }
1181 
1182 /*
1183  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1184  * For ASF and Pass Through versions of f/w this means
1185  * that the driver is loaded. For AMT version type f/w
1186  * this means that the network i/f is open.
1187  */
1188 static void
1189 em_hw_control_acquire(struct e1000_hw *hw)
1190 {
1191 	uint32_t ctrl_ext, swsm;
1192 
1193 	/* Let firmware know the driver has taken over */
1194 	if (hw->mac.type == e1000_82573) {
1195 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1196 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1197 
1198 	} else {
1199 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1200 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1201 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1202 	}
1203 }
1204 
1205 /*
1206  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1207  * For ASF and Pass Through versions of f/w this means that the
1208  * driver is no longer loaded. For AMT versions of the
1209  * f/w this means that the network i/f is closed.
1210  */
1211 static void
1212 em_hw_control_release(struct e1000_hw *hw)
1213 {
1214 	uint32_t ctrl_ext, swsm;
1215 
1216 	/* Let firmware taken over control of h/w */
1217 	if (hw->mac.type == e1000_82573) {
1218 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1219 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1220 	} else {
1221 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1222 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1223 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1224 	}
1225 }
1226 
1227 /*
1228  * Bit of a misnomer, what this really means is
1229  * to enable OS management of the system... aka
1230  * to disable special hardware management features.
1231  */
1232 static void
1233 em_init_manageability(struct e1000_hw *hw)
1234 {
1235 	if (e1000_enable_mng_pass_thru(hw)) {
1236 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1237 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1238 
1239 		/* disable hardware interception of ARP */
1240 		manc &= ~(E1000_MANC_ARP_EN);
1241 
1242 		/* enable receiving management packets to the host */
1243 		manc |= E1000_MANC_EN_MNG2HOST;
1244 		manc2h |= 1 << 5;  /* Mng Port 623 */
1245 		manc2h |= 1 << 6;  /* Mng Port 664 */
1246 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1247 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1248 	}
1249 }
1250 
1251 /*
1252  * Give control back to hardware management
1253  * controller if there is one.
1254  */
1255 static void
1256 em_release_manageability(struct e1000_hw *hw)
1257 {
1258 	uint32_t manc;
1259 
1260 	if (e1000_enable_mng_pass_thru(hw)) {
1261 		manc = E1000_READ_REG(hw, E1000_MANC);
1262 
1263 		/* re-enable hardware interception of ARP */
1264 		manc |= E1000_MANC_ARP_EN;
1265 		manc &= ~E1000_MANC_EN_MNG2HOST;
1266 
1267 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1268 	}
1269 }
1270 
1271 static int
1272 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1273 {
1274 	struct e1000_hw *hw =
1275 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1276 	uint32_t rctl;
1277 
1278 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1279 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1280 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1281 
1282 	return 0;
1283 }
1284 
1285 static int
1286 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1287 {
1288 	struct e1000_hw *hw =
1289 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1290 	uint32_t rctl;
1291 
1292 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1293 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1294 	if (dev->data->all_multicast == 1)
1295 		rctl |= E1000_RCTL_MPE;
1296 	else
1297 		rctl &= (~E1000_RCTL_MPE);
1298 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1299 
1300 	return 0;
1301 }
1302 
1303 static int
1304 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1305 {
1306 	struct e1000_hw *hw =
1307 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1308 	uint32_t rctl;
1309 
1310 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1311 	rctl |= E1000_RCTL_MPE;
1312 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1313 
1314 	return 0;
1315 }
1316 
1317 static int
1318 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1319 {
1320 	struct e1000_hw *hw =
1321 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1322 	uint32_t rctl;
1323 
1324 	if (dev->data->promiscuous == 1)
1325 		return 0; /* must remain in all_multicast mode */
1326 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1327 	rctl &= (~E1000_RCTL_MPE);
1328 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1329 
1330 	return 0;
1331 }
1332 
1333 static int
1334 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1335 {
1336 	struct e1000_hw *hw =
1337 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1338 	struct e1000_vfta * shadow_vfta =
1339 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1340 	uint32_t vfta;
1341 	uint32_t vid_idx;
1342 	uint32_t vid_bit;
1343 
1344 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1345 			      E1000_VFTA_ENTRY_MASK);
1346 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1347 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1348 	if (on)
1349 		vfta |= vid_bit;
1350 	else
1351 		vfta &= ~vid_bit;
1352 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1353 
1354 	/* update local VFTA copy */
1355 	shadow_vfta->vfta[vid_idx] = vfta;
1356 
1357 	return 0;
1358 }
1359 
1360 static void
1361 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1362 {
1363 	struct e1000_hw *hw =
1364 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1365 	uint32_t reg;
1366 
1367 	/* Filter Table Disable */
1368 	reg = E1000_READ_REG(hw, E1000_RCTL);
1369 	reg &= ~E1000_RCTL_CFIEN;
1370 	reg &= ~E1000_RCTL_VFE;
1371 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1372 }
1373 
1374 static void
1375 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1376 {
1377 	struct e1000_hw *hw =
1378 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1379 	struct e1000_vfta * shadow_vfta =
1380 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1381 	uint32_t reg;
1382 	int i;
1383 
1384 	/* Filter Table Enable, CFI not used for packet acceptance */
1385 	reg = E1000_READ_REG(hw, E1000_RCTL);
1386 	reg &= ~E1000_RCTL_CFIEN;
1387 	reg |= E1000_RCTL_VFE;
1388 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1389 
1390 	/* restore vfta from local copy */
1391 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1392 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1393 }
1394 
1395 static void
1396 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1397 {
1398 	struct e1000_hw *hw =
1399 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1400 	uint32_t reg;
1401 
1402 	/* VLAN Mode Disable */
1403 	reg = E1000_READ_REG(hw, E1000_CTRL);
1404 	reg &= ~E1000_CTRL_VME;
1405 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1406 
1407 }
1408 
1409 static void
1410 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1411 {
1412 	struct e1000_hw *hw =
1413 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1414 	uint32_t reg;
1415 
1416 	/* VLAN Mode Enable */
1417 	reg = E1000_READ_REG(hw, E1000_CTRL);
1418 	reg |= E1000_CTRL_VME;
1419 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1420 }
1421 
1422 static int
1423 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1424 {
1425 	struct rte_eth_rxmode *rxmode;
1426 
1427 	rxmode = &dev->data->dev_conf.rxmode;
1428 	if(mask & ETH_VLAN_STRIP_MASK){
1429 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1430 			em_vlan_hw_strip_enable(dev);
1431 		else
1432 			em_vlan_hw_strip_disable(dev);
1433 	}
1434 
1435 	if(mask & ETH_VLAN_FILTER_MASK){
1436 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1437 			em_vlan_hw_filter_enable(dev);
1438 		else
1439 			em_vlan_hw_filter_disable(dev);
1440 	}
1441 
1442 	return 0;
1443 }
1444 
1445 /*
1446  * It enables the interrupt mask and then enable the interrupt.
1447  *
1448  * @param dev
1449  *  Pointer to struct rte_eth_dev.
1450  *
1451  * @return
1452  *  - On success, zero.
1453  *  - On failure, a negative value.
1454  */
1455 static int
1456 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1457 {
1458 	uint32_t regval;
1459 	struct e1000_hw *hw =
1460 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1461 
1462 	/* clear interrupt */
1463 	E1000_READ_REG(hw, E1000_ICR);
1464 	regval = E1000_READ_REG(hw, E1000_IMS);
1465 	E1000_WRITE_REG(hw, E1000_IMS,
1466 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1467 	return 0;
1468 }
1469 
1470 /*
1471  * It clears the interrupt causes and enables the interrupt.
1472  * It will be called once only during nic initialized.
1473  *
1474  * @param dev
1475  *  Pointer to struct rte_eth_dev.
1476  *
1477  * @return
1478  *  - On success, zero.
1479  *  - On failure, a negative value.
1480  */
1481 static int
1482 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1483 {
1484 	struct e1000_hw *hw =
1485 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1486 
1487 	E1000_READ_REG(hw, E1000_ICR);
1488 	em_rxq_intr_enable(hw);
1489 	return 0;
1490 }
1491 
1492 /*
1493  * It enable receive packet interrupt.
1494  * @param hw
1495  * Pointer to struct e1000_hw
1496  *
1497  * @return
1498  */
1499 static void
1500 em_rxq_intr_enable(struct e1000_hw *hw)
1501 {
1502 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1503 	E1000_WRITE_FLUSH(hw);
1504 }
1505 
1506 /*
1507  * It disabled lsc interrupt.
1508  * @param hw
1509  * Pointer to struct e1000_hw
1510  *
1511  * @return
1512  */
1513 static void
1514 em_lsc_intr_disable(struct e1000_hw *hw)
1515 {
1516 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1517 	E1000_WRITE_FLUSH(hw);
1518 }
1519 
1520 /*
1521  * It disabled receive packet interrupt.
1522  * @param hw
1523  * Pointer to struct e1000_hw
1524  *
1525  * @return
1526  */
1527 static void
1528 em_rxq_intr_disable(struct e1000_hw *hw)
1529 {
1530 	E1000_READ_REG(hw, E1000_ICR);
1531 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1532 	E1000_WRITE_FLUSH(hw);
1533 }
1534 
1535 /*
1536  * It reads ICR and gets interrupt causes, check it and set a bit flag
1537  * to update link status.
1538  *
1539  * @param dev
1540  *  Pointer to struct rte_eth_dev.
1541  *
1542  * @return
1543  *  - On success, zero.
1544  *  - On failure, a negative value.
1545  */
1546 static int
1547 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1548 {
1549 	uint32_t icr;
1550 	struct e1000_hw *hw =
1551 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1552 	struct e1000_interrupt *intr =
1553 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1554 
1555 	/* read-on-clear nic registers here */
1556 	icr = E1000_READ_REG(hw, E1000_ICR);
1557 	if (icr & E1000_ICR_LSC) {
1558 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1559 	}
1560 
1561 	return 0;
1562 }
1563 
1564 /*
1565  * It executes link_update after knowing an interrupt is prsent.
1566  *
1567  * @param dev
1568  *  Pointer to struct rte_eth_dev.
1569  *
1570  * @return
1571  *  - On success, zero.
1572  *  - On failure, a negative value.
1573  */
1574 static int
1575 eth_em_interrupt_action(struct rte_eth_dev *dev,
1576 			struct rte_intr_handle *intr_handle)
1577 {
1578 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1579 	struct e1000_hw *hw =
1580 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1581 	struct e1000_interrupt *intr =
1582 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1583 	struct rte_eth_link link;
1584 	int ret;
1585 
1586 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1587 		return -1;
1588 
1589 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1590 	rte_intr_ack(intr_handle);
1591 
1592 	/* set get_link_status to check register later */
1593 	hw->mac.get_link_status = 1;
1594 	ret = eth_em_link_update(dev, 0);
1595 
1596 	/* check if link has changed */
1597 	if (ret < 0)
1598 		return 0;
1599 
1600 	rte_eth_linkstatus_get(dev, &link);
1601 
1602 	if (link.link_status) {
1603 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1604 			     dev->data->port_id, link.link_speed,
1605 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1606 			     "full-duplex" : "half-duplex");
1607 	} else {
1608 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1609 	}
1610 	PMD_INIT_LOG(DEBUG, "PCI Address: " PCI_PRI_FMT,
1611 		     pci_dev->addr.domain, pci_dev->addr.bus,
1612 		     pci_dev->addr.devid, pci_dev->addr.function);
1613 
1614 	return 0;
1615 }
1616 
1617 /**
1618  * Interrupt handler which shall be registered at first.
1619  *
1620  * @param handle
1621  *  Pointer to interrupt handle.
1622  * @param param
1623  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1624  *
1625  * @return
1626  *  void
1627  */
1628 static void
1629 eth_em_interrupt_handler(void *param)
1630 {
1631 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1632 
1633 	eth_em_interrupt_get_status(dev);
1634 	eth_em_interrupt_action(dev, dev->intr_handle);
1635 	rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1636 }
1637 
1638 static int
1639 eth_em_led_on(struct rte_eth_dev *dev)
1640 {
1641 	struct e1000_hw *hw;
1642 
1643 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1644 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1645 }
1646 
1647 static int
1648 eth_em_led_off(struct rte_eth_dev *dev)
1649 {
1650 	struct e1000_hw *hw;
1651 
1652 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1653 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1654 }
1655 
1656 static int
1657 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1658 {
1659 	struct e1000_hw *hw;
1660 	uint32_t ctrl;
1661 	int tx_pause;
1662 	int rx_pause;
1663 
1664 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1665 	fc_conf->pause_time = hw->fc.pause_time;
1666 	fc_conf->high_water = hw->fc.high_water;
1667 	fc_conf->low_water = hw->fc.low_water;
1668 	fc_conf->send_xon = hw->fc.send_xon;
1669 	fc_conf->autoneg = hw->mac.autoneg;
1670 
1671 	/*
1672 	 * Return rx_pause and tx_pause status according to actual setting of
1673 	 * the TFCE and RFCE bits in the CTRL register.
1674 	 */
1675 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1676 	if (ctrl & E1000_CTRL_TFCE)
1677 		tx_pause = 1;
1678 	else
1679 		tx_pause = 0;
1680 
1681 	if (ctrl & E1000_CTRL_RFCE)
1682 		rx_pause = 1;
1683 	else
1684 		rx_pause = 0;
1685 
1686 	if (rx_pause && tx_pause)
1687 		fc_conf->mode = RTE_FC_FULL;
1688 	else if (rx_pause)
1689 		fc_conf->mode = RTE_FC_RX_PAUSE;
1690 	else if (tx_pause)
1691 		fc_conf->mode = RTE_FC_TX_PAUSE;
1692 	else
1693 		fc_conf->mode = RTE_FC_NONE;
1694 
1695 	return 0;
1696 }
1697 
1698 static int
1699 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1700 {
1701 	struct e1000_hw *hw;
1702 	int err;
1703 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1704 		e1000_fc_none,
1705 		e1000_fc_rx_pause,
1706 		e1000_fc_tx_pause,
1707 		e1000_fc_full
1708 	};
1709 	uint32_t rx_buf_size;
1710 	uint32_t max_high_water;
1711 	uint32_t rctl;
1712 
1713 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1714 	if (fc_conf->autoneg != hw->mac.autoneg)
1715 		return -ENOTSUP;
1716 	rx_buf_size = em_get_rx_buffer_size(hw);
1717 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1718 
1719 	/* At least reserve one Ethernet frame for watermark */
1720 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
1721 	if ((fc_conf->high_water > max_high_water) ||
1722 	    (fc_conf->high_water < fc_conf->low_water)) {
1723 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1724 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1725 		return -EINVAL;
1726 	}
1727 
1728 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1729 	hw->fc.pause_time     = fc_conf->pause_time;
1730 	hw->fc.high_water     = fc_conf->high_water;
1731 	hw->fc.low_water      = fc_conf->low_water;
1732 	hw->fc.send_xon	      = fc_conf->send_xon;
1733 
1734 	err = e1000_setup_link_generic(hw);
1735 	if (err == E1000_SUCCESS) {
1736 
1737 		/* check if we want to forward MAC frames - driver doesn't have native
1738 		 * capability to do that, so we'll write the registers ourselves */
1739 
1740 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1741 
1742 		/* set or clear MFLCN.PMCF bit depending on configuration */
1743 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1744 			rctl |= E1000_RCTL_PMCF;
1745 		else
1746 			rctl &= ~E1000_RCTL_PMCF;
1747 
1748 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1749 		E1000_WRITE_FLUSH(hw);
1750 
1751 		return 0;
1752 	}
1753 
1754 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1755 	return -EIO;
1756 }
1757 
1758 static int
1759 eth_em_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1760 		uint32_t index, __rte_unused uint32_t pool)
1761 {
1762 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1763 
1764 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1765 }
1766 
1767 static void
1768 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1769 {
1770 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1771 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1772 
1773 	memset(addr, 0, sizeof(addr));
1774 
1775 	e1000_rar_set(hw, addr, index);
1776 }
1777 
1778 static int
1779 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1780 			    struct rte_ether_addr *addr)
1781 {
1782 	eth_em_rar_clear(dev, 0);
1783 
1784 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1785 }
1786 
1787 static int
1788 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1789 {
1790 	struct rte_eth_dev_info dev_info;
1791 	struct e1000_hw *hw;
1792 	uint32_t frame_size;
1793 	uint32_t rctl;
1794 	int ret;
1795 
1796 	ret = eth_em_infos_get(dev, &dev_info);
1797 	if (ret != 0)
1798 		return ret;
1799 
1800 	frame_size = mtu + E1000_ETH_OVERHEAD;
1801 
1802 	/* check that mtu is within the allowed range */
1803 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
1804 		return -EINVAL;
1805 
1806 	/*
1807 	 * If device is started, refuse mtu that requires the support of
1808 	 * scattered packets when this feature has not been enabled before.
1809 	 */
1810 	if (dev->data->dev_started && !dev->data->scattered_rx &&
1811 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
1812 		PMD_INIT_LOG(ERR, "Stop port first.");
1813 		return -EINVAL;
1814 	}
1815 
1816 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1817 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1818 
1819 	/* switch to jumbo mode if needed */
1820 	if (frame_size > E1000_ETH_MAX_LEN) {
1821 		dev->data->dev_conf.rxmode.offloads |=
1822 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1823 		rctl |= E1000_RCTL_LPE;
1824 	} else {
1825 		dev->data->dev_conf.rxmode.offloads &=
1826 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1827 		rctl &= ~E1000_RCTL_LPE;
1828 	}
1829 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1830 
1831 	/* update max frame size */
1832 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1833 	return 0;
1834 }
1835 
1836 static int
1837 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1838 			struct rte_ether_addr *mc_addr_set,
1839 			uint32_t nb_mc_addr)
1840 {
1841 	struct e1000_hw *hw;
1842 
1843 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1844 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1845 	return 0;
1846 }
1847 
1848 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1849 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1850 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1851