xref: /f-stack/dpdk/drivers/net/e1000/em_ethdev.c (revision ebf5cedb)
1 /* SPDX-License-Identifier: BSD-3-Clause
2  * Copyright(c) 2010-2016 Intel Corporation
3  */
4 
5 #include <sys/queue.h>
6 #include <stdio.h>
7 #include <errno.h>
8 #include <stdint.h>
9 #include <stdarg.h>
10 
11 #include <rte_common.h>
12 #include <rte_interrupts.h>
13 #include <rte_byteorder.h>
14 #include <rte_debug.h>
15 #include <rte_pci.h>
16 #include <rte_bus_pci.h>
17 #include <rte_ether.h>
18 #include <rte_ethdev_driver.h>
19 #include <rte_ethdev_pci.h>
20 #include <rte_memory.h>
21 #include <rte_eal.h>
22 #include <rte_malloc.h>
23 #include <rte_dev.h>
24 
25 #include "e1000_logs.h"
26 #include "base/e1000_api.h"
27 #include "e1000_ethdev.h"
28 
29 #define EM_EIAC			0x000DC
30 
31 #define PMD_ROUNDUP(x,y)	(((x) + (y) - 1)/(y) * (y))
32 
33 
34 static int eth_em_configure(struct rte_eth_dev *dev);
35 static int eth_em_start(struct rte_eth_dev *dev);
36 static void eth_em_stop(struct rte_eth_dev *dev);
37 static void eth_em_close(struct rte_eth_dev *dev);
38 static 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 	.rx_queue_count       = eth_em_rx_queue_count,
180 	.rx_descriptor_done   = eth_em_rx_descriptor_done,
181 	.rx_descriptor_status = eth_em_rx_descriptor_status,
182 	.tx_descriptor_status = eth_em_tx_descriptor_status,
183 	.tx_queue_setup       = eth_em_tx_queue_setup,
184 	.tx_queue_release     = eth_em_tx_queue_release,
185 	.rx_queue_intr_enable = eth_em_rx_queue_intr_enable,
186 	.rx_queue_intr_disable = eth_em_rx_queue_intr_disable,
187 	.dev_led_on           = eth_em_led_on,
188 	.dev_led_off          = eth_em_led_off,
189 	.flow_ctrl_get        = eth_em_flow_ctrl_get,
190 	.flow_ctrl_set        = eth_em_flow_ctrl_set,
191 	.mac_addr_set         = eth_em_default_mac_addr_set,
192 	.mac_addr_add         = eth_em_rar_set,
193 	.mac_addr_remove      = eth_em_rar_clear,
194 	.set_mc_addr_list     = eth_em_set_mc_addr_list,
195 	.rxq_info_get         = em_rxq_info_get,
196 	.txq_info_get         = em_txq_info_get,
197 };
198 
199 
200 /**
201  *  eth_em_dev_is_ich8 - Check for ICH8 device
202  *  @hw: pointer to the HW structure
203  *
204  *  return TRUE for ICH8, otherwise FALSE
205  **/
206 static bool
207 eth_em_dev_is_ich8(struct e1000_hw *hw)
208 {
209 	DEBUGFUNC("eth_em_dev_is_ich8");
210 
211 	switch (hw->device_id) {
212 	case E1000_DEV_ID_PCH2_LV_LM:
213 	case E1000_DEV_ID_PCH_LPT_I217_LM:
214 	case E1000_DEV_ID_PCH_LPT_I217_V:
215 	case E1000_DEV_ID_PCH_LPTLP_I218_LM:
216 	case E1000_DEV_ID_PCH_LPTLP_I218_V:
217 	case E1000_DEV_ID_PCH_I218_V2:
218 	case E1000_DEV_ID_PCH_I218_LM2:
219 	case E1000_DEV_ID_PCH_I218_V3:
220 	case E1000_DEV_ID_PCH_I218_LM3:
221 	case E1000_DEV_ID_PCH_SPT_I219_LM:
222 	case E1000_DEV_ID_PCH_SPT_I219_V:
223 	case E1000_DEV_ID_PCH_SPT_I219_LM2:
224 	case E1000_DEV_ID_PCH_SPT_I219_V2:
225 	case E1000_DEV_ID_PCH_LBG_I219_LM3:
226 	case E1000_DEV_ID_PCH_SPT_I219_LM4:
227 	case E1000_DEV_ID_PCH_SPT_I219_V4:
228 	case E1000_DEV_ID_PCH_SPT_I219_LM5:
229 	case E1000_DEV_ID_PCH_SPT_I219_V5:
230 	case E1000_DEV_ID_PCH_CNP_I219_LM6:
231 	case E1000_DEV_ID_PCH_CNP_I219_V6:
232 	case E1000_DEV_ID_PCH_CNP_I219_LM7:
233 	case E1000_DEV_ID_PCH_CNP_I219_V7:
234 		return 1;
235 	default:
236 		return 0;
237 	}
238 }
239 
240 static int
241 eth_em_dev_init(struct rte_eth_dev *eth_dev)
242 {
243 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
244 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
245 	struct e1000_adapter *adapter =
246 		E1000_DEV_PRIVATE(eth_dev->data->dev_private);
247 	struct e1000_hw *hw =
248 		E1000_DEV_PRIVATE_TO_HW(eth_dev->data->dev_private);
249 	struct e1000_vfta * shadow_vfta =
250 		E1000_DEV_PRIVATE_TO_VFTA(eth_dev->data->dev_private);
251 
252 	eth_dev->dev_ops = &eth_em_ops;
253 	eth_dev->rx_pkt_burst = (eth_rx_burst_t)&eth_em_recv_pkts;
254 	eth_dev->tx_pkt_burst = (eth_tx_burst_t)&eth_em_xmit_pkts;
255 	eth_dev->tx_pkt_prepare = (eth_tx_prep_t)&eth_em_prep_pkts;
256 
257 	/* for secondary processes, we don't initialise any further as primary
258 	 * has already done this work. Only check we don't need a different
259 	 * RX function */
260 	if (rte_eal_process_type() != RTE_PROC_PRIMARY){
261 		if (eth_dev->data->scattered_rx)
262 			eth_dev->rx_pkt_burst =
263 				(eth_rx_burst_t)&eth_em_recv_scattered_pkts;
264 		return 0;
265 	}
266 
267 	rte_eth_copy_pci_info(eth_dev, pci_dev);
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 	/* Pass the information to the rte_eth_dev_close() that it should also
301 	 * release the private port resources.
302 	 */
303 	eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
304 
305 	/* initialize the vfta */
306 	memset(shadow_vfta, 0, sizeof(*shadow_vfta));
307 
308 	PMD_INIT_LOG(DEBUG, "port_id %d vendorID=0x%x deviceID=0x%x",
309 		     eth_dev->data->port_id, pci_dev->id.vendor_id,
310 		     pci_dev->id.device_id);
311 
312 	rte_intr_callback_register(intr_handle,
313 				   eth_em_interrupt_handler, eth_dev);
314 
315 	return 0;
316 }
317 
318 static int
319 eth_em_dev_uninit(struct rte_eth_dev *eth_dev)
320 {
321 	PMD_INIT_FUNC_TRACE();
322 
323 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
324 		return -EPERM;
325 
326 	eth_em_close(eth_dev);
327 
328 	return 0;
329 }
330 
331 static int eth_em_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
332 	struct rte_pci_device *pci_dev)
333 {
334 	return rte_eth_dev_pci_generic_probe(pci_dev,
335 		sizeof(struct e1000_adapter), eth_em_dev_init);
336 }
337 
338 static int eth_em_pci_remove(struct rte_pci_device *pci_dev)
339 {
340 	return rte_eth_dev_pci_generic_remove(pci_dev, eth_em_dev_uninit);
341 }
342 
343 static struct rte_pci_driver rte_em_pmd = {
344 	.id_table = pci_id_em_map,
345 	.drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
346 	.probe = eth_em_pci_probe,
347 	.remove = eth_em_pci_remove,
348 };
349 
350 static int
351 em_hw_init(struct e1000_hw *hw)
352 {
353 	int diag;
354 
355 	diag = hw->mac.ops.init_params(hw);
356 	if (diag != 0) {
357 		PMD_INIT_LOG(ERR, "MAC Initialization Error");
358 		return diag;
359 	}
360 	diag = hw->nvm.ops.init_params(hw);
361 	if (diag != 0) {
362 		PMD_INIT_LOG(ERR, "NVM Initialization Error");
363 		return diag;
364 	}
365 	diag = hw->phy.ops.init_params(hw);
366 	if (diag != 0) {
367 		PMD_INIT_LOG(ERR, "PHY Initialization Error");
368 		return diag;
369 	}
370 	(void) e1000_get_bus_info(hw);
371 
372 	hw->mac.autoneg = 1;
373 	hw->phy.autoneg_wait_to_complete = 0;
374 	hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
375 
376 	e1000_init_script_state_82541(hw, TRUE);
377 	e1000_set_tbi_compatibility_82543(hw, TRUE);
378 
379 	/* Copper options */
380 	if (hw->phy.media_type == e1000_media_type_copper) {
381 		hw->phy.mdix = 0; /* AUTO_ALL_MODES */
382 		hw->phy.disable_polarity_correction = 0;
383 		hw->phy.ms_type = e1000_ms_hw_default;
384 	}
385 
386 	/*
387 	 * Start from a known state, this is important in reading the nvm
388 	 * and mac from that.
389 	 */
390 	e1000_reset_hw(hw);
391 
392 	/* Make sure we have a good EEPROM before we read from it */
393 	if (e1000_validate_nvm_checksum(hw) < 0) {
394 		/*
395 		 * Some PCI-E parts fail the first check due to
396 		 * the link being in sleep state, call it again,
397 		 * if it fails a second time its a real issue.
398 		 */
399 		diag = e1000_validate_nvm_checksum(hw);
400 		if (diag < 0) {
401 			PMD_INIT_LOG(ERR, "EEPROM checksum invalid");
402 			goto error;
403 		}
404 	}
405 
406 	/* Read the permanent MAC address out of the EEPROM */
407 	diag = e1000_read_mac_addr(hw);
408 	if (diag != 0) {
409 		PMD_INIT_LOG(ERR, "EEPROM error while reading MAC address");
410 		goto error;
411 	}
412 
413 	/* Now initialize the hardware */
414 	diag = em_hardware_init(hw);
415 	if (diag != 0) {
416 		PMD_INIT_LOG(ERR, "Hardware initialization failed");
417 		goto error;
418 	}
419 
420 	hw->mac.get_link_status = 1;
421 
422 	/* Indicate SOL/IDER usage */
423 	diag = e1000_check_reset_block(hw);
424 	if (diag < 0) {
425 		PMD_INIT_LOG(ERR, "PHY reset is blocked due to "
426 			"SOL/IDER session");
427 	}
428 	return 0;
429 
430 error:
431 	em_hw_control_release(hw);
432 	return diag;
433 }
434 
435 static int
436 eth_em_configure(struct rte_eth_dev *dev)
437 {
438 	struct e1000_interrupt *intr =
439 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
440 
441 	PMD_INIT_FUNC_TRACE();
442 	intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
443 
444 	PMD_INIT_FUNC_TRACE();
445 
446 	return 0;
447 }
448 
449 static void
450 em_set_pba(struct e1000_hw *hw)
451 {
452 	uint32_t pba;
453 
454 	/*
455 	 * Packet Buffer Allocation (PBA)
456 	 * Writing PBA sets the receive portion of the buffer
457 	 * the remainder is used for the transmit buffer.
458 	 * Devices before the 82547 had a Packet Buffer of 64K.
459 	 * After the 82547 the buffer was reduced to 40K.
460 	 */
461 	switch (hw->mac.type) {
462 		case e1000_82547:
463 		case e1000_82547_rev_2:
464 		/* 82547: Total Packet Buffer is 40K */
465 			pba = E1000_PBA_22K; /* 22K for Rx, 18K for Tx */
466 			break;
467 		case e1000_82571:
468 		case e1000_82572:
469 		case e1000_80003es2lan:
470 			pba = E1000_PBA_32K; /* 32K for Rx, 16K for Tx */
471 			break;
472 		case e1000_82573: /* 82573: Total Packet Buffer is 32K */
473 			pba = E1000_PBA_12K; /* 12K for Rx, 20K for Tx */
474 			break;
475 		case e1000_82574:
476 		case e1000_82583:
477 			pba = E1000_PBA_20K; /* 20K for Rx, 20K for Tx */
478 			break;
479 		case e1000_ich8lan:
480 			pba = E1000_PBA_8K;
481 			break;
482 		case e1000_ich9lan:
483 		case e1000_ich10lan:
484 			pba = E1000_PBA_10K;
485 			break;
486 		case e1000_pchlan:
487 		case e1000_pch2lan:
488 		case e1000_pch_lpt:
489 		case e1000_pch_spt:
490 		case e1000_pch_cnp:
491 			pba = E1000_PBA_26K;
492 			break;
493 		default:
494 			pba = E1000_PBA_40K; /* 40K for Rx, 24K for Tx */
495 	}
496 
497 	E1000_WRITE_REG(hw, E1000_PBA, pba);
498 }
499 
500 static void
501 eth_em_rxtx_control(struct rte_eth_dev *dev,
502 		    bool enable)
503 {
504 	struct e1000_hw *hw =
505 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
506 	uint32_t tctl, rctl;
507 
508 	tctl = E1000_READ_REG(hw, E1000_TCTL);
509 	rctl = E1000_READ_REG(hw, E1000_RCTL);
510 	if (enable) {
511 		/* enable Tx/Rx */
512 		tctl |= E1000_TCTL_EN;
513 		rctl |= E1000_RCTL_EN;
514 	} else {
515 		/* disable Tx/Rx */
516 		tctl &= ~E1000_TCTL_EN;
517 		rctl &= ~E1000_RCTL_EN;
518 	}
519 	E1000_WRITE_REG(hw, E1000_TCTL, tctl);
520 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
521 	E1000_WRITE_FLUSH(hw);
522 }
523 
524 static int
525 eth_em_start(struct rte_eth_dev *dev)
526 {
527 	struct e1000_adapter *adapter =
528 		E1000_DEV_PRIVATE(dev->data->dev_private);
529 	struct e1000_hw *hw =
530 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
531 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
532 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
533 	int ret, mask;
534 	uint32_t intr_vector = 0;
535 	uint32_t *speeds;
536 	int num_speeds;
537 	bool autoneg;
538 
539 	PMD_INIT_FUNC_TRACE();
540 
541 	eth_em_stop(dev);
542 
543 	e1000_power_up_phy(hw);
544 
545 	/* Set default PBA value */
546 	em_set_pba(hw);
547 
548 	/* Put the address into the Receive Address Array */
549 	e1000_rar_set(hw, hw->mac.addr, 0);
550 
551 	/*
552 	 * With the 82571 adapter, RAR[0] may be overwritten
553 	 * when the other port is reset, we make a duplicate
554 	 * in RAR[14] for that eventuality, this assures
555 	 * the interface continues to function.
556 	 */
557 	if (hw->mac.type == e1000_82571) {
558 		e1000_set_laa_state_82571(hw, TRUE);
559 		e1000_rar_set(hw, hw->mac.addr, E1000_RAR_ENTRIES - 1);
560 	}
561 
562 	/* Initialize the hardware */
563 	if (em_hardware_init(hw)) {
564 		PMD_INIT_LOG(ERR, "Unable to initialize the hardware");
565 		return -EIO;
566 	}
567 
568 	E1000_WRITE_REG(hw, E1000_VET, RTE_ETHER_TYPE_VLAN);
569 
570 	/* Configure for OS presence */
571 	em_init_manageability(hw);
572 
573 	if (dev->data->dev_conf.intr_conf.rxq != 0) {
574 		intr_vector = dev->data->nb_rx_queues;
575 		if (rte_intr_efd_enable(intr_handle, intr_vector))
576 			return -1;
577 	}
578 
579 	if (rte_intr_dp_is_en(intr_handle)) {
580 		intr_handle->intr_vec =
581 			rte_zmalloc("intr_vec",
582 					dev->data->nb_rx_queues * sizeof(int), 0);
583 		if (intr_handle->intr_vec == NULL) {
584 			PMD_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
585 						" intr_vec", dev->data->nb_rx_queues);
586 			return -ENOMEM;
587 		}
588 
589 		/* enable rx interrupt */
590 		em_rxq_intr_enable(hw);
591 	}
592 
593 	eth_em_tx_init(dev);
594 
595 	ret = eth_em_rx_init(dev);
596 	if (ret) {
597 		PMD_INIT_LOG(ERR, "Unable to initialize RX hardware");
598 		em_dev_clear_queues(dev);
599 		return ret;
600 	}
601 
602 	e1000_clear_hw_cntrs_base_generic(hw);
603 
604 	mask = ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK | \
605 			ETH_VLAN_EXTEND_MASK;
606 	ret = eth_em_vlan_offload_set(dev, mask);
607 	if (ret) {
608 		PMD_INIT_LOG(ERR, "Unable to update vlan offload");
609 		em_dev_clear_queues(dev);
610 		return ret;
611 	}
612 
613 	/* Set Interrupt Throttling Rate to maximum allowed value. */
614 	E1000_WRITE_REG(hw, E1000_ITR, UINT16_MAX);
615 
616 	/* Setup link speed and duplex */
617 	speeds = &dev->data->dev_conf.link_speeds;
618 	if (*speeds == ETH_LINK_SPEED_AUTONEG) {
619 		hw->phy.autoneg_advertised = E1000_ALL_SPEED_DUPLEX;
620 		hw->mac.autoneg = 1;
621 	} else {
622 		num_speeds = 0;
623 		autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
624 
625 		/* Reset */
626 		hw->phy.autoneg_advertised = 0;
627 
628 		if (*speeds & ~(ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
629 				ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
630 				ETH_LINK_SPEED_1G | ETH_LINK_SPEED_FIXED)) {
631 			num_speeds = -1;
632 			goto error_invalid_config;
633 		}
634 		if (*speeds & ETH_LINK_SPEED_10M_HD) {
635 			hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
636 			num_speeds++;
637 		}
638 		if (*speeds & ETH_LINK_SPEED_10M) {
639 			hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
640 			num_speeds++;
641 		}
642 		if (*speeds & ETH_LINK_SPEED_100M_HD) {
643 			hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
644 			num_speeds++;
645 		}
646 		if (*speeds & ETH_LINK_SPEED_100M) {
647 			hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
648 			num_speeds++;
649 		}
650 		if (*speeds & ETH_LINK_SPEED_1G) {
651 			hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
652 			num_speeds++;
653 		}
654 		if (num_speeds == 0 || (!autoneg && (num_speeds > 1)))
655 			goto error_invalid_config;
656 
657 		/* Set/reset the mac.autoneg based on the link speed,
658 		 * fixed or not
659 		 */
660 		if (!autoneg) {
661 			hw->mac.autoneg = 0;
662 			hw->mac.forced_speed_duplex =
663 					hw->phy.autoneg_advertised;
664 		} else {
665 			hw->mac.autoneg = 1;
666 		}
667 	}
668 
669 	e1000_setup_link(hw);
670 
671 	if (rte_intr_allow_others(intr_handle)) {
672 		/* check if lsc interrupt is enabled */
673 		if (dev->data->dev_conf.intr_conf.lsc != 0) {
674 			ret = eth_em_interrupt_setup(dev);
675 			if (ret) {
676 				PMD_INIT_LOG(ERR, "Unable to setup interrupts");
677 				em_dev_clear_queues(dev);
678 				return ret;
679 			}
680 		}
681 	} else {
682 		rte_intr_callback_unregister(intr_handle,
683 						eth_em_interrupt_handler,
684 						(void *)dev);
685 		if (dev->data->dev_conf.intr_conf.lsc != 0)
686 			PMD_INIT_LOG(INFO, "lsc won't enable because of"
687 				     " no intr multiplexn");
688 	}
689 	/* check if rxq interrupt is enabled */
690 	if (dev->data->dev_conf.intr_conf.rxq != 0)
691 		eth_em_rxq_interrupt_setup(dev);
692 
693 	rte_intr_enable(intr_handle);
694 
695 	adapter->stopped = 0;
696 
697 	eth_em_rxtx_control(dev, true);
698 	eth_em_link_update(dev, 0);
699 
700 	PMD_INIT_LOG(DEBUG, "<<");
701 
702 	return 0;
703 
704 error_invalid_config:
705 	PMD_INIT_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
706 		     dev->data->dev_conf.link_speeds, dev->data->port_id);
707 	em_dev_clear_queues(dev);
708 	return -EINVAL;
709 }
710 
711 /*********************************************************************
712  *
713  *  This routine disables all traffic on the adapter by issuing a
714  *  global reset on the MAC.
715  *
716  **********************************************************************/
717 static void
718 eth_em_stop(struct rte_eth_dev *dev)
719 {
720 	struct rte_eth_link link;
721 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
722 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
723 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
724 
725 	eth_em_rxtx_control(dev, false);
726 	em_rxq_intr_disable(hw);
727 	em_lsc_intr_disable(hw);
728 
729 	e1000_reset_hw(hw);
730 
731 	/* Flush desc rings for i219 */
732 	if (hw->mac.type == e1000_pch_spt || hw->mac.type == e1000_pch_cnp)
733 		em_flush_desc_rings(dev);
734 
735 	if (hw->mac.type >= e1000_82544)
736 		E1000_WRITE_REG(hw, E1000_WUC, 0);
737 
738 	/* Power down the phy. Needed to make the link go down */
739 	e1000_power_down_phy(hw);
740 
741 	em_dev_clear_queues(dev);
742 
743 	/* clear the recorded link status */
744 	memset(&link, 0, sizeof(link));
745 	rte_eth_linkstatus_set(dev, &link);
746 
747 	if (!rte_intr_allow_others(intr_handle))
748 		/* resume to the default handler */
749 		rte_intr_callback_register(intr_handle,
750 					   eth_em_interrupt_handler,
751 					   (void *)dev);
752 
753 	/* Clean datapath event and queue/vec mapping */
754 	rte_intr_efd_disable(intr_handle);
755 	if (intr_handle->intr_vec != NULL) {
756 		rte_free(intr_handle->intr_vec);
757 		intr_handle->intr_vec = NULL;
758 	}
759 }
760 
761 static void
762 eth_em_close(struct rte_eth_dev *dev)
763 {
764 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
765 	struct e1000_adapter *adapter =
766 		E1000_DEV_PRIVATE(dev->data->dev_private);
767 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
768 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
769 
770 	eth_em_stop(dev);
771 	adapter->stopped = 1;
772 	em_dev_free_queues(dev);
773 	e1000_phy_hw_reset(hw);
774 	em_release_manageability(hw);
775 	em_hw_control_release(hw);
776 
777 	dev->dev_ops = NULL;
778 	dev->rx_pkt_burst = NULL;
779 	dev->tx_pkt_burst = NULL;
780 
781 	/* disable uio intr before callback unregister */
782 	rte_intr_disable(intr_handle);
783 	rte_intr_callback_unregister(intr_handle,
784 				     eth_em_interrupt_handler, dev);
785 }
786 
787 static int
788 em_get_rx_buffer_size(struct e1000_hw *hw)
789 {
790 	uint32_t rx_buf_size;
791 
792 	rx_buf_size = ((E1000_READ_REG(hw, E1000_PBA) & UINT16_MAX) << 10);
793 	return rx_buf_size;
794 }
795 
796 /*********************************************************************
797  *
798  *  Initialize the hardware
799  *
800  **********************************************************************/
801 static int
802 em_hardware_init(struct e1000_hw *hw)
803 {
804 	uint32_t rx_buf_size;
805 	int diag;
806 
807 	/* Issue a global reset */
808 	e1000_reset_hw(hw);
809 
810 	/* Let the firmware know the OS is in control */
811 	em_hw_control_acquire(hw);
812 
813 	/*
814 	 * These parameters control the automatic generation (Tx) and
815 	 * response (Rx) to Ethernet PAUSE frames.
816 	 * - High water mark should allow for at least two standard size (1518)
817 	 *   frames to be received after sending an XOFF.
818 	 * - Low water mark works best when it is very near the high water mark.
819 	 *   This allows the receiver to restart by sending XON when it has
820 	 *   drained a bit. Here we use an arbitrary value of 1500 which will
821 	 *   restart after one full frame is pulled from the buffer. There
822 	 *   could be several smaller frames in the buffer and if so they will
823 	 *   not trigger the XON until their total number reduces the buffer
824 	 *   by 1500.
825 	 * - The pause time is fairly large at 1000 x 512ns = 512 usec.
826 	 */
827 	rx_buf_size = em_get_rx_buffer_size(hw);
828 
829 	hw->fc.high_water = rx_buf_size -
830 		PMD_ROUNDUP(RTE_ETHER_MAX_LEN * 2, 1024);
831 	hw->fc.low_water = hw->fc.high_water - 1500;
832 
833 	if (hw->mac.type == e1000_80003es2lan)
834 		hw->fc.pause_time = UINT16_MAX;
835 	else
836 		hw->fc.pause_time = EM_FC_PAUSE_TIME;
837 
838 	hw->fc.send_xon = 1;
839 
840 	/* Set Flow control, use the tunable location if sane */
841 	if (em_fc_setting <= e1000_fc_full)
842 		hw->fc.requested_mode = em_fc_setting;
843 	else
844 		hw->fc.requested_mode = e1000_fc_none;
845 
846 	/* Workaround: no TX flow ctrl for PCH */
847 	if (hw->mac.type == e1000_pchlan)
848 		hw->fc.requested_mode = e1000_fc_rx_pause;
849 
850 	/* Override - settings for PCH2LAN, ya its magic :) */
851 	if (hw->mac.type == e1000_pch2lan) {
852 		hw->fc.high_water = 0x5C20;
853 		hw->fc.low_water = 0x5048;
854 		hw->fc.pause_time = 0x0650;
855 		hw->fc.refresh_time = 0x0400;
856 	} else if (hw->mac.type == e1000_pch_lpt ||
857 		   hw->mac.type == e1000_pch_spt ||
858 		   hw->mac.type == e1000_pch_cnp) {
859 		hw->fc.requested_mode = e1000_fc_full;
860 	}
861 
862 	diag = e1000_init_hw(hw);
863 	if (diag < 0)
864 		return diag;
865 	e1000_check_for_link(hw);
866 	return 0;
867 }
868 
869 /* This function is based on em_update_stats_counters() in e1000/if_em.c */
870 static int
871 eth_em_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
872 {
873 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
874 	struct e1000_hw_stats *stats =
875 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
876 	int pause_frames;
877 
878 	if(hw->phy.media_type == e1000_media_type_copper ||
879 			(E1000_READ_REG(hw, E1000_STATUS) & E1000_STATUS_LU)) {
880 		stats->symerrs += E1000_READ_REG(hw,E1000_SYMERRS);
881 		stats->sec += E1000_READ_REG(hw, E1000_SEC);
882 	}
883 
884 	stats->crcerrs += E1000_READ_REG(hw, E1000_CRCERRS);
885 	stats->mpc += E1000_READ_REG(hw, E1000_MPC);
886 	stats->scc += E1000_READ_REG(hw, E1000_SCC);
887 	stats->ecol += E1000_READ_REG(hw, E1000_ECOL);
888 
889 	stats->mcc += E1000_READ_REG(hw, E1000_MCC);
890 	stats->latecol += E1000_READ_REG(hw, E1000_LATECOL);
891 	stats->colc += E1000_READ_REG(hw, E1000_COLC);
892 	stats->dc += E1000_READ_REG(hw, E1000_DC);
893 	stats->rlec += E1000_READ_REG(hw, E1000_RLEC);
894 	stats->xonrxc += E1000_READ_REG(hw, E1000_XONRXC);
895 	stats->xontxc += E1000_READ_REG(hw, E1000_XONTXC);
896 
897 	/*
898 	 * For watchdog management we need to know if we have been
899 	 * paused during the last interval, so capture that here.
900 	 */
901 	pause_frames = E1000_READ_REG(hw, E1000_XOFFRXC);
902 	stats->xoffrxc += pause_frames;
903 	stats->xofftxc += E1000_READ_REG(hw, E1000_XOFFTXC);
904 	stats->fcruc += E1000_READ_REG(hw, E1000_FCRUC);
905 	stats->prc64 += E1000_READ_REG(hw, E1000_PRC64);
906 	stats->prc127 += E1000_READ_REG(hw, E1000_PRC127);
907 	stats->prc255 += E1000_READ_REG(hw, E1000_PRC255);
908 	stats->prc511 += E1000_READ_REG(hw, E1000_PRC511);
909 	stats->prc1023 += E1000_READ_REG(hw, E1000_PRC1023);
910 	stats->prc1522 += E1000_READ_REG(hw, E1000_PRC1522);
911 	stats->gprc += E1000_READ_REG(hw, E1000_GPRC);
912 	stats->bprc += E1000_READ_REG(hw, E1000_BPRC);
913 	stats->mprc += E1000_READ_REG(hw, E1000_MPRC);
914 	stats->gptc += E1000_READ_REG(hw, E1000_GPTC);
915 
916 	/*
917 	 * For the 64-bit byte counters the low dword must be read first.
918 	 * Both registers clear on the read of the high dword.
919 	 */
920 
921 	stats->gorc += E1000_READ_REG(hw, E1000_GORCL);
922 	stats->gorc += ((uint64_t)E1000_READ_REG(hw, E1000_GORCH) << 32);
923 	stats->gotc += E1000_READ_REG(hw, E1000_GOTCL);
924 	stats->gotc += ((uint64_t)E1000_READ_REG(hw, E1000_GOTCH) << 32);
925 
926 	stats->rnbc += E1000_READ_REG(hw, E1000_RNBC);
927 	stats->ruc += E1000_READ_REG(hw, E1000_RUC);
928 	stats->rfc += E1000_READ_REG(hw, E1000_RFC);
929 	stats->roc += E1000_READ_REG(hw, E1000_ROC);
930 	stats->rjc += E1000_READ_REG(hw, E1000_RJC);
931 
932 	stats->tor += E1000_READ_REG(hw, E1000_TORH);
933 	stats->tot += E1000_READ_REG(hw, E1000_TOTH);
934 
935 	stats->tpr += E1000_READ_REG(hw, E1000_TPR);
936 	stats->tpt += E1000_READ_REG(hw, E1000_TPT);
937 	stats->ptc64 += E1000_READ_REG(hw, E1000_PTC64);
938 	stats->ptc127 += E1000_READ_REG(hw, E1000_PTC127);
939 	stats->ptc255 += E1000_READ_REG(hw, E1000_PTC255);
940 	stats->ptc511 += E1000_READ_REG(hw, E1000_PTC511);
941 	stats->ptc1023 += E1000_READ_REG(hw, E1000_PTC1023);
942 	stats->ptc1522 += E1000_READ_REG(hw, E1000_PTC1522);
943 	stats->mptc += E1000_READ_REG(hw, E1000_MPTC);
944 	stats->bptc += E1000_READ_REG(hw, E1000_BPTC);
945 
946 	/* Interrupt Counts */
947 
948 	if (hw->mac.type >= e1000_82571) {
949 		stats->iac += E1000_READ_REG(hw, E1000_IAC);
950 		stats->icrxptc += E1000_READ_REG(hw, E1000_ICRXPTC);
951 		stats->icrxatc += E1000_READ_REG(hw, E1000_ICRXATC);
952 		stats->ictxptc += E1000_READ_REG(hw, E1000_ICTXPTC);
953 		stats->ictxatc += E1000_READ_REG(hw, E1000_ICTXATC);
954 		stats->ictxqec += E1000_READ_REG(hw, E1000_ICTXQEC);
955 		stats->ictxqmtc += E1000_READ_REG(hw, E1000_ICTXQMTC);
956 		stats->icrxdmtc += E1000_READ_REG(hw, E1000_ICRXDMTC);
957 		stats->icrxoc += E1000_READ_REG(hw, E1000_ICRXOC);
958 	}
959 
960 	if (hw->mac.type >= e1000_82543) {
961 		stats->algnerrc += E1000_READ_REG(hw, E1000_ALGNERRC);
962 		stats->rxerrc += E1000_READ_REG(hw, E1000_RXERRC);
963 		stats->tncrs += E1000_READ_REG(hw, E1000_TNCRS);
964 		stats->cexterr += E1000_READ_REG(hw, E1000_CEXTERR);
965 		stats->tsctc += E1000_READ_REG(hw, E1000_TSCTC);
966 		stats->tsctfc += E1000_READ_REG(hw, E1000_TSCTFC);
967 	}
968 
969 	if (rte_stats == NULL)
970 		return -EINVAL;
971 
972 	/* Rx Errors */
973 	rte_stats->imissed = stats->mpc;
974 	rte_stats->ierrors = stats->crcerrs +
975 	                     stats->rlec + stats->ruc + stats->roc +
976 	                     stats->rxerrc + stats->algnerrc + stats->cexterr;
977 
978 	/* Tx Errors */
979 	rte_stats->oerrors = stats->ecol + stats->latecol;
980 
981 	rte_stats->ipackets = stats->gprc;
982 	rte_stats->opackets = stats->gptc;
983 	rte_stats->ibytes   = stats->gorc;
984 	rte_stats->obytes   = stats->gotc;
985 	return 0;
986 }
987 
988 static int
989 eth_em_stats_reset(struct rte_eth_dev *dev)
990 {
991 	struct e1000_hw_stats *hw_stats =
992 			E1000_DEV_PRIVATE_TO_STATS(dev->data->dev_private);
993 
994 	/* HW registers are cleared on read */
995 	eth_em_stats_get(dev, NULL);
996 
997 	/* Reset software totals */
998 	memset(hw_stats, 0, sizeof(*hw_stats));
999 
1000 	return 0;
1001 }
1002 
1003 static int
1004 eth_em_rx_queue_intr_enable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1005 {
1006 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1007 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1008 	struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
1009 
1010 	em_rxq_intr_enable(hw);
1011 	rte_intr_ack(intr_handle);
1012 
1013 	return 0;
1014 }
1015 
1016 static int
1017 eth_em_rx_queue_intr_disable(struct rte_eth_dev *dev, __rte_unused uint16_t queue_id)
1018 {
1019 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1020 
1021 	em_rxq_intr_disable(hw);
1022 
1023 	return 0;
1024 }
1025 
1026 uint32_t
1027 em_get_max_pktlen(struct rte_eth_dev *dev)
1028 {
1029 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1030 
1031 	switch (hw->mac.type) {
1032 	case e1000_82571:
1033 	case e1000_82572:
1034 	case e1000_ich9lan:
1035 	case e1000_ich10lan:
1036 	case e1000_pch2lan:
1037 	case e1000_pch_lpt:
1038 	case e1000_pch_spt:
1039 	case e1000_pch_cnp:
1040 	case e1000_82574:
1041 	case e1000_80003es2lan: /* 9K Jumbo Frame size */
1042 	case e1000_82583:
1043 		return 0x2412;
1044 	case e1000_pchlan:
1045 		return 0x1000;
1046 	/* Adapters that do not support jumbo frames */
1047 	case e1000_ich8lan:
1048 		return RTE_ETHER_MAX_LEN;
1049 	default:
1050 		return MAX_JUMBO_FRAME_SIZE;
1051 	}
1052 }
1053 
1054 static int
1055 eth_em_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
1056 {
1057 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1058 
1059 	dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
1060 	dev_info->max_rx_pktlen = em_get_max_pktlen(dev);
1061 	dev_info->max_mac_addrs = hw->mac.rar_entry_count;
1062 
1063 	/*
1064 	 * Starting with 631xESB hw supports 2 TX/RX queues per port.
1065 	 * Unfortunatelly, all these nics have just one TX context.
1066 	 * So we have few choises for TX:
1067 	 * - Use just one TX queue.
1068 	 * - Allow cksum offload only for one TX queue.
1069 	 * - Don't allow TX cksum offload at all.
1070 	 * For now, option #1 was chosen.
1071 	 * To use second RX queue we have to use extended RX descriptor
1072 	 * (Multiple Receive Queues are mutually exclusive with UDP
1073 	 * fragmentation and are not supported when a legacy receive
1074 	 * descriptor format is used).
1075 	 * Which means separate RX routinies - as legacy nics (82540, 82545)
1076 	 * don't support extended RXD.
1077 	 * To avoid it we support just one RX queue for now (no RSS).
1078 	 */
1079 
1080 	dev_info->max_rx_queues = 1;
1081 	dev_info->max_tx_queues = 1;
1082 
1083 	dev_info->rx_queue_offload_capa = em_get_rx_queue_offloads_capa(dev);
1084 	dev_info->rx_offload_capa = em_get_rx_port_offloads_capa(dev) |
1085 				    dev_info->rx_queue_offload_capa;
1086 	dev_info->tx_queue_offload_capa = em_get_tx_queue_offloads_capa(dev);
1087 	dev_info->tx_offload_capa = em_get_tx_port_offloads_capa(dev) |
1088 				    dev_info->tx_queue_offload_capa;
1089 
1090 	dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
1091 		.nb_max = E1000_MAX_RING_DESC,
1092 		.nb_min = E1000_MIN_RING_DESC,
1093 		.nb_align = EM_RXD_ALIGN,
1094 	};
1095 
1096 	dev_info->tx_desc_lim = (struct rte_eth_desc_lim) {
1097 		.nb_max = E1000_MAX_RING_DESC,
1098 		.nb_min = E1000_MIN_RING_DESC,
1099 		.nb_align = EM_TXD_ALIGN,
1100 		.nb_seg_max = EM_TX_MAX_SEG,
1101 		.nb_mtu_seg_max = EM_TX_MAX_MTU_SEG,
1102 	};
1103 
1104 	dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
1105 			ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
1106 			ETH_LINK_SPEED_1G;
1107 
1108 	/* Preferred queue parameters */
1109 	dev_info->default_rxportconf.nb_queues = 1;
1110 	dev_info->default_txportconf.nb_queues = 1;
1111 	dev_info->default_txportconf.ring_size = 256;
1112 	dev_info->default_rxportconf.ring_size = 256;
1113 
1114 	return 0;
1115 }
1116 
1117 /* return 0 means link status changed, -1 means not changed */
1118 static int
1119 eth_em_link_update(struct rte_eth_dev *dev, int wait_to_complete)
1120 {
1121 	struct e1000_hw *hw =
1122 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1123 	struct rte_eth_link link;
1124 	int link_up, count;
1125 
1126 	link_up = 0;
1127 	hw->mac.get_link_status = 1;
1128 
1129 	/* possible wait-to-complete in up to 9 seconds */
1130 	for (count = 0; count < EM_LINK_UPDATE_CHECK_TIMEOUT; count ++) {
1131 		/* Read the real link status */
1132 		switch (hw->phy.media_type) {
1133 		case e1000_media_type_copper:
1134 			/* Do the work to read phy */
1135 			e1000_check_for_link(hw);
1136 			link_up = !hw->mac.get_link_status;
1137 			break;
1138 
1139 		case e1000_media_type_fiber:
1140 			e1000_check_for_link(hw);
1141 			link_up = (E1000_READ_REG(hw, E1000_STATUS) &
1142 					E1000_STATUS_LU);
1143 			break;
1144 
1145 		case e1000_media_type_internal_serdes:
1146 			e1000_check_for_link(hw);
1147 			link_up = hw->mac.serdes_has_link;
1148 			break;
1149 
1150 		default:
1151 			break;
1152 		}
1153 		if (link_up || wait_to_complete == 0)
1154 			break;
1155 		rte_delay_ms(EM_LINK_UPDATE_CHECK_INTERVAL);
1156 	}
1157 	memset(&link, 0, sizeof(link));
1158 
1159 	/* Now we check if a transition has happened */
1160 	if (link_up) {
1161 		uint16_t duplex, speed;
1162 		hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
1163 		link.link_duplex = (duplex == FULL_DUPLEX) ?
1164 				ETH_LINK_FULL_DUPLEX :
1165 				ETH_LINK_HALF_DUPLEX;
1166 		link.link_speed = speed;
1167 		link.link_status = ETH_LINK_UP;
1168 		link.link_autoneg = !(dev->data->dev_conf.link_speeds &
1169 				ETH_LINK_SPEED_FIXED);
1170 	} else {
1171 		link.link_speed = ETH_SPEED_NUM_NONE;
1172 		link.link_duplex = ETH_LINK_HALF_DUPLEX;
1173 		link.link_status = ETH_LINK_DOWN;
1174 		link.link_autoneg = ETH_LINK_FIXED;
1175 	}
1176 
1177 	return rte_eth_linkstatus_set(dev, &link);
1178 }
1179 
1180 /*
1181  * em_hw_control_acquire sets {CTRL_EXT|FWSM}:DRV_LOAD bit.
1182  * For ASF and Pass Through versions of f/w this means
1183  * that the driver is loaded. For AMT version type f/w
1184  * this means that the network i/f is open.
1185  */
1186 static void
1187 em_hw_control_acquire(struct e1000_hw *hw)
1188 {
1189 	uint32_t ctrl_ext, swsm;
1190 
1191 	/* Let firmware know the driver has taken over */
1192 	if (hw->mac.type == e1000_82573) {
1193 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1194 		E1000_WRITE_REG(hw, E1000_SWSM, swsm | E1000_SWSM_DRV_LOAD);
1195 
1196 	} else {
1197 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1198 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1199 			ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
1200 	}
1201 }
1202 
1203 /*
1204  * em_hw_control_release resets {CTRL_EXTT|FWSM}:DRV_LOAD bit.
1205  * For ASF and Pass Through versions of f/w this means that the
1206  * driver is no longer loaded. For AMT versions of the
1207  * f/w this means that the network i/f is closed.
1208  */
1209 static void
1210 em_hw_control_release(struct e1000_hw *hw)
1211 {
1212 	uint32_t ctrl_ext, swsm;
1213 
1214 	/* Let firmware taken over control of h/w */
1215 	if (hw->mac.type == e1000_82573) {
1216 		swsm = E1000_READ_REG(hw, E1000_SWSM);
1217 		E1000_WRITE_REG(hw, E1000_SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
1218 	} else {
1219 		ctrl_ext = E1000_READ_REG(hw, E1000_CTRL_EXT);
1220 		E1000_WRITE_REG(hw, E1000_CTRL_EXT,
1221 			ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
1222 	}
1223 }
1224 
1225 /*
1226  * Bit of a misnomer, what this really means is
1227  * to enable OS management of the system... aka
1228  * to disable special hardware management features.
1229  */
1230 static void
1231 em_init_manageability(struct e1000_hw *hw)
1232 {
1233 	if (e1000_enable_mng_pass_thru(hw)) {
1234 		uint32_t manc2h = E1000_READ_REG(hw, E1000_MANC2H);
1235 		uint32_t manc = E1000_READ_REG(hw, E1000_MANC);
1236 
1237 		/* disable hardware interception of ARP */
1238 		manc &= ~(E1000_MANC_ARP_EN);
1239 
1240 		/* enable receiving management packets to the host */
1241 		manc |= E1000_MANC_EN_MNG2HOST;
1242 		manc2h |= 1 << 5;  /* Mng Port 623 */
1243 		manc2h |= 1 << 6;  /* Mng Port 664 */
1244 		E1000_WRITE_REG(hw, E1000_MANC2H, manc2h);
1245 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1246 	}
1247 }
1248 
1249 /*
1250  * Give control back to hardware management
1251  * controller if there is one.
1252  */
1253 static void
1254 em_release_manageability(struct e1000_hw *hw)
1255 {
1256 	uint32_t manc;
1257 
1258 	if (e1000_enable_mng_pass_thru(hw)) {
1259 		manc = E1000_READ_REG(hw, E1000_MANC);
1260 
1261 		/* re-enable hardware interception of ARP */
1262 		manc |= E1000_MANC_ARP_EN;
1263 		manc &= ~E1000_MANC_EN_MNG2HOST;
1264 
1265 		E1000_WRITE_REG(hw, E1000_MANC, manc);
1266 	}
1267 }
1268 
1269 static int
1270 eth_em_promiscuous_enable(struct rte_eth_dev *dev)
1271 {
1272 	struct e1000_hw *hw =
1273 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1274 	uint32_t rctl;
1275 
1276 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1277 	rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
1278 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1279 
1280 	return 0;
1281 }
1282 
1283 static int
1284 eth_em_promiscuous_disable(struct rte_eth_dev *dev)
1285 {
1286 	struct e1000_hw *hw =
1287 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1288 	uint32_t rctl;
1289 
1290 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1291 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_SBP);
1292 	if (dev->data->all_multicast == 1)
1293 		rctl |= E1000_RCTL_MPE;
1294 	else
1295 		rctl &= (~E1000_RCTL_MPE);
1296 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1297 
1298 	return 0;
1299 }
1300 
1301 static int
1302 eth_em_allmulticast_enable(struct rte_eth_dev *dev)
1303 {
1304 	struct e1000_hw *hw =
1305 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1306 	uint32_t rctl;
1307 
1308 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1309 	rctl |= E1000_RCTL_MPE;
1310 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1311 
1312 	return 0;
1313 }
1314 
1315 static int
1316 eth_em_allmulticast_disable(struct rte_eth_dev *dev)
1317 {
1318 	struct e1000_hw *hw =
1319 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1320 	uint32_t rctl;
1321 
1322 	if (dev->data->promiscuous == 1)
1323 		return 0; /* must remain in all_multicast mode */
1324 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1325 	rctl &= (~E1000_RCTL_MPE);
1326 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1327 
1328 	return 0;
1329 }
1330 
1331 static int
1332 eth_em_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
1333 {
1334 	struct e1000_hw *hw =
1335 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1336 	struct e1000_vfta * shadow_vfta =
1337 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1338 	uint32_t vfta;
1339 	uint32_t vid_idx;
1340 	uint32_t vid_bit;
1341 
1342 	vid_idx = (uint32_t) ((vlan_id >> E1000_VFTA_ENTRY_SHIFT) &
1343 			      E1000_VFTA_ENTRY_MASK);
1344 	vid_bit = (uint32_t) (1 << (vlan_id & E1000_VFTA_ENTRY_BIT_SHIFT_MASK));
1345 	vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, vid_idx);
1346 	if (on)
1347 		vfta |= vid_bit;
1348 	else
1349 		vfta &= ~vid_bit;
1350 	E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, vid_idx, vfta);
1351 
1352 	/* update local VFTA copy */
1353 	shadow_vfta->vfta[vid_idx] = vfta;
1354 
1355 	return 0;
1356 }
1357 
1358 static void
1359 em_vlan_hw_filter_disable(struct rte_eth_dev *dev)
1360 {
1361 	struct e1000_hw *hw =
1362 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1363 	uint32_t reg;
1364 
1365 	/* Filter Table Disable */
1366 	reg = E1000_READ_REG(hw, E1000_RCTL);
1367 	reg &= ~E1000_RCTL_CFIEN;
1368 	reg &= ~E1000_RCTL_VFE;
1369 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1370 }
1371 
1372 static void
1373 em_vlan_hw_filter_enable(struct rte_eth_dev *dev)
1374 {
1375 	struct e1000_hw *hw =
1376 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1377 	struct e1000_vfta * shadow_vfta =
1378 		E1000_DEV_PRIVATE_TO_VFTA(dev->data->dev_private);
1379 	uint32_t reg;
1380 	int i;
1381 
1382 	/* Filter Table Enable, CFI not used for packet acceptance */
1383 	reg = E1000_READ_REG(hw, E1000_RCTL);
1384 	reg &= ~E1000_RCTL_CFIEN;
1385 	reg |= E1000_RCTL_VFE;
1386 	E1000_WRITE_REG(hw, E1000_RCTL, reg);
1387 
1388 	/* restore vfta from local copy */
1389 	for (i = 0; i < IGB_VFTA_SIZE; i++)
1390 		E1000_WRITE_REG_ARRAY(hw, E1000_VFTA, i, shadow_vfta->vfta[i]);
1391 }
1392 
1393 static void
1394 em_vlan_hw_strip_disable(struct rte_eth_dev *dev)
1395 {
1396 	struct e1000_hw *hw =
1397 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1398 	uint32_t reg;
1399 
1400 	/* VLAN Mode Disable */
1401 	reg = E1000_READ_REG(hw, E1000_CTRL);
1402 	reg &= ~E1000_CTRL_VME;
1403 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1404 
1405 }
1406 
1407 static void
1408 em_vlan_hw_strip_enable(struct rte_eth_dev *dev)
1409 {
1410 	struct e1000_hw *hw =
1411 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1412 	uint32_t reg;
1413 
1414 	/* VLAN Mode Enable */
1415 	reg = E1000_READ_REG(hw, E1000_CTRL);
1416 	reg |= E1000_CTRL_VME;
1417 	E1000_WRITE_REG(hw, E1000_CTRL, reg);
1418 }
1419 
1420 static int
1421 eth_em_vlan_offload_set(struct rte_eth_dev *dev, int mask)
1422 {
1423 	struct rte_eth_rxmode *rxmode;
1424 
1425 	rxmode = &dev->data->dev_conf.rxmode;
1426 	if(mask & ETH_VLAN_STRIP_MASK){
1427 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
1428 			em_vlan_hw_strip_enable(dev);
1429 		else
1430 			em_vlan_hw_strip_disable(dev);
1431 	}
1432 
1433 	if(mask & ETH_VLAN_FILTER_MASK){
1434 		if (rxmode->offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
1435 			em_vlan_hw_filter_enable(dev);
1436 		else
1437 			em_vlan_hw_filter_disable(dev);
1438 	}
1439 
1440 	return 0;
1441 }
1442 
1443 /*
1444  * It enables the interrupt mask and then enable the interrupt.
1445  *
1446  * @param dev
1447  *  Pointer to struct rte_eth_dev.
1448  *
1449  * @return
1450  *  - On success, zero.
1451  *  - On failure, a negative value.
1452  */
1453 static int
1454 eth_em_interrupt_setup(struct rte_eth_dev *dev)
1455 {
1456 	uint32_t regval;
1457 	struct e1000_hw *hw =
1458 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1459 
1460 	/* clear interrupt */
1461 	E1000_READ_REG(hw, E1000_ICR);
1462 	regval = E1000_READ_REG(hw, E1000_IMS);
1463 	E1000_WRITE_REG(hw, E1000_IMS,
1464 			regval | E1000_ICR_LSC | E1000_ICR_OTHER);
1465 	return 0;
1466 }
1467 
1468 /*
1469  * It clears the interrupt causes and enables the interrupt.
1470  * It will be called once only during nic initialized.
1471  *
1472  * @param dev
1473  *  Pointer to struct rte_eth_dev.
1474  *
1475  * @return
1476  *  - On success, zero.
1477  *  - On failure, a negative value.
1478  */
1479 static int
1480 eth_em_rxq_interrupt_setup(struct rte_eth_dev *dev)
1481 {
1482 	struct e1000_hw *hw =
1483 	E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1484 
1485 	E1000_READ_REG(hw, E1000_ICR);
1486 	em_rxq_intr_enable(hw);
1487 	return 0;
1488 }
1489 
1490 /*
1491  * It enable receive packet interrupt.
1492  * @param hw
1493  * Pointer to struct e1000_hw
1494  *
1495  * @return
1496  */
1497 static void
1498 em_rxq_intr_enable(struct e1000_hw *hw)
1499 {
1500 	E1000_WRITE_REG(hw, E1000_IMS, E1000_IMS_RXT0);
1501 	E1000_WRITE_FLUSH(hw);
1502 }
1503 
1504 /*
1505  * It disabled lsc interrupt.
1506  * @param hw
1507  * Pointer to struct e1000_hw
1508  *
1509  * @return
1510  */
1511 static void
1512 em_lsc_intr_disable(struct e1000_hw *hw)
1513 {
1514 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_LSC | E1000_IMS_OTHER);
1515 	E1000_WRITE_FLUSH(hw);
1516 }
1517 
1518 /*
1519  * It disabled receive packet interrupt.
1520  * @param hw
1521  * Pointer to struct e1000_hw
1522  *
1523  * @return
1524  */
1525 static void
1526 em_rxq_intr_disable(struct e1000_hw *hw)
1527 {
1528 	E1000_READ_REG(hw, E1000_ICR);
1529 	E1000_WRITE_REG(hw, E1000_IMC, E1000_IMS_RXT0);
1530 	E1000_WRITE_FLUSH(hw);
1531 }
1532 
1533 /*
1534  * It reads ICR and gets interrupt causes, check it and set a bit flag
1535  * to update link status.
1536  *
1537  * @param dev
1538  *  Pointer to struct rte_eth_dev.
1539  *
1540  * @return
1541  *  - On success, zero.
1542  *  - On failure, a negative value.
1543  */
1544 static int
1545 eth_em_interrupt_get_status(struct rte_eth_dev *dev)
1546 {
1547 	uint32_t icr;
1548 	struct e1000_hw *hw =
1549 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1550 	struct e1000_interrupt *intr =
1551 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1552 
1553 	/* read-on-clear nic registers here */
1554 	icr = E1000_READ_REG(hw, E1000_ICR);
1555 	if (icr & E1000_ICR_LSC) {
1556 		intr->flags |= E1000_FLAG_NEED_LINK_UPDATE;
1557 	}
1558 
1559 	return 0;
1560 }
1561 
1562 /*
1563  * It executes link_update after knowing an interrupt is prsent.
1564  *
1565  * @param dev
1566  *  Pointer to struct rte_eth_dev.
1567  *
1568  * @return
1569  *  - On success, zero.
1570  *  - On failure, a negative value.
1571  */
1572 static int
1573 eth_em_interrupt_action(struct rte_eth_dev *dev,
1574 			struct rte_intr_handle *intr_handle)
1575 {
1576 	struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
1577 	struct e1000_hw *hw =
1578 		E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1579 	struct e1000_interrupt *intr =
1580 		E1000_DEV_PRIVATE_TO_INTR(dev->data->dev_private);
1581 	struct rte_eth_link link;
1582 	int ret;
1583 
1584 	if (!(intr->flags & E1000_FLAG_NEED_LINK_UPDATE))
1585 		return -1;
1586 
1587 	intr->flags &= ~E1000_FLAG_NEED_LINK_UPDATE;
1588 	rte_intr_ack(intr_handle);
1589 
1590 	/* set get_link_status to check register later */
1591 	hw->mac.get_link_status = 1;
1592 	ret = eth_em_link_update(dev, 0);
1593 
1594 	/* check if link has changed */
1595 	if (ret < 0)
1596 		return 0;
1597 
1598 	rte_eth_linkstatus_get(dev, &link);
1599 
1600 	if (link.link_status) {
1601 		PMD_INIT_LOG(INFO, " Port %d: Link Up - speed %u Mbps - %s",
1602 			     dev->data->port_id, link.link_speed,
1603 			     link.link_duplex == ETH_LINK_FULL_DUPLEX ?
1604 			     "full-duplex" : "half-duplex");
1605 	} else {
1606 		PMD_INIT_LOG(INFO, " Port %d: Link Down", dev->data->port_id);
1607 	}
1608 	PMD_INIT_LOG(DEBUG, "PCI Address: %04d:%02d:%02d:%d",
1609 		     pci_dev->addr.domain, pci_dev->addr.bus,
1610 		     pci_dev->addr.devid, pci_dev->addr.function);
1611 
1612 	return 0;
1613 }
1614 
1615 /**
1616  * Interrupt handler which shall be registered at first.
1617  *
1618  * @param handle
1619  *  Pointer to interrupt handle.
1620  * @param param
1621  *  The address of parameter (struct rte_eth_dev *) regsitered before.
1622  *
1623  * @return
1624  *  void
1625  */
1626 static void
1627 eth_em_interrupt_handler(void *param)
1628 {
1629 	struct rte_eth_dev *dev = (struct rte_eth_dev *)param;
1630 
1631 	eth_em_interrupt_get_status(dev);
1632 	eth_em_interrupt_action(dev, dev->intr_handle);
1633 	_rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
1634 }
1635 
1636 static int
1637 eth_em_led_on(struct rte_eth_dev *dev)
1638 {
1639 	struct e1000_hw *hw;
1640 
1641 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1642 	return e1000_led_on(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1643 }
1644 
1645 static int
1646 eth_em_led_off(struct rte_eth_dev *dev)
1647 {
1648 	struct e1000_hw *hw;
1649 
1650 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1651 	return e1000_led_off(hw) == E1000_SUCCESS ? 0 : -ENOTSUP;
1652 }
1653 
1654 static int
1655 eth_em_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1656 {
1657 	struct e1000_hw *hw;
1658 	uint32_t ctrl;
1659 	int tx_pause;
1660 	int rx_pause;
1661 
1662 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1663 	fc_conf->pause_time = hw->fc.pause_time;
1664 	fc_conf->high_water = hw->fc.high_water;
1665 	fc_conf->low_water = hw->fc.low_water;
1666 	fc_conf->send_xon = hw->fc.send_xon;
1667 	fc_conf->autoneg = hw->mac.autoneg;
1668 
1669 	/*
1670 	 * Return rx_pause and tx_pause status according to actual setting of
1671 	 * the TFCE and RFCE bits in the CTRL register.
1672 	 */
1673 	ctrl = E1000_READ_REG(hw, E1000_CTRL);
1674 	if (ctrl & E1000_CTRL_TFCE)
1675 		tx_pause = 1;
1676 	else
1677 		tx_pause = 0;
1678 
1679 	if (ctrl & E1000_CTRL_RFCE)
1680 		rx_pause = 1;
1681 	else
1682 		rx_pause = 0;
1683 
1684 	if (rx_pause && tx_pause)
1685 		fc_conf->mode = RTE_FC_FULL;
1686 	else if (rx_pause)
1687 		fc_conf->mode = RTE_FC_RX_PAUSE;
1688 	else if (tx_pause)
1689 		fc_conf->mode = RTE_FC_TX_PAUSE;
1690 	else
1691 		fc_conf->mode = RTE_FC_NONE;
1692 
1693 	return 0;
1694 }
1695 
1696 static int
1697 eth_em_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
1698 {
1699 	struct e1000_hw *hw;
1700 	int err;
1701 	enum e1000_fc_mode rte_fcmode_2_e1000_fcmode[] = {
1702 		e1000_fc_none,
1703 		e1000_fc_rx_pause,
1704 		e1000_fc_tx_pause,
1705 		e1000_fc_full
1706 	};
1707 	uint32_t rx_buf_size;
1708 	uint32_t max_high_water;
1709 	uint32_t rctl;
1710 
1711 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1712 	if (fc_conf->autoneg != hw->mac.autoneg)
1713 		return -ENOTSUP;
1714 	rx_buf_size = em_get_rx_buffer_size(hw);
1715 	PMD_INIT_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
1716 
1717 	/* At least reserve one Ethernet frame for watermark */
1718 	max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
1719 	if ((fc_conf->high_water > max_high_water) ||
1720 	    (fc_conf->high_water < fc_conf->low_water)) {
1721 		PMD_INIT_LOG(ERR, "e1000 incorrect high/low water value");
1722 		PMD_INIT_LOG(ERR, "high water must <= 0x%x", max_high_water);
1723 		return -EINVAL;
1724 	}
1725 
1726 	hw->fc.requested_mode = rte_fcmode_2_e1000_fcmode[fc_conf->mode];
1727 	hw->fc.pause_time     = fc_conf->pause_time;
1728 	hw->fc.high_water     = fc_conf->high_water;
1729 	hw->fc.low_water      = fc_conf->low_water;
1730 	hw->fc.send_xon	      = fc_conf->send_xon;
1731 
1732 	err = e1000_setup_link_generic(hw);
1733 	if (err == E1000_SUCCESS) {
1734 
1735 		/* check if we want to forward MAC frames - driver doesn't have native
1736 		 * capability to do that, so we'll write the registers ourselves */
1737 
1738 		rctl = E1000_READ_REG(hw, E1000_RCTL);
1739 
1740 		/* set or clear MFLCN.PMCF bit depending on configuration */
1741 		if (fc_conf->mac_ctrl_frame_fwd != 0)
1742 			rctl |= E1000_RCTL_PMCF;
1743 		else
1744 			rctl &= ~E1000_RCTL_PMCF;
1745 
1746 		E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1747 		E1000_WRITE_FLUSH(hw);
1748 
1749 		return 0;
1750 	}
1751 
1752 	PMD_INIT_LOG(ERR, "e1000_setup_link_generic = 0x%x", err);
1753 	return -EIO;
1754 }
1755 
1756 static int
1757 eth_em_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
1758 		uint32_t index, __rte_unused uint32_t pool)
1759 {
1760 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1761 
1762 	return e1000_rar_set(hw, mac_addr->addr_bytes, index);
1763 }
1764 
1765 static void
1766 eth_em_rar_clear(struct rte_eth_dev *dev, uint32_t index)
1767 {
1768 	uint8_t addr[RTE_ETHER_ADDR_LEN];
1769 	struct e1000_hw *hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1770 
1771 	memset(addr, 0, sizeof(addr));
1772 
1773 	e1000_rar_set(hw, addr, index);
1774 }
1775 
1776 static int
1777 eth_em_default_mac_addr_set(struct rte_eth_dev *dev,
1778 			    struct rte_ether_addr *addr)
1779 {
1780 	eth_em_rar_clear(dev, 0);
1781 
1782 	return eth_em_rar_set(dev, (void *)addr, 0, 0);
1783 }
1784 
1785 static int
1786 eth_em_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
1787 {
1788 	struct rte_eth_dev_info dev_info;
1789 	struct e1000_hw *hw;
1790 	uint32_t frame_size;
1791 	uint32_t rctl;
1792 	int ret;
1793 
1794 	ret = eth_em_infos_get(dev, &dev_info);
1795 	if (ret != 0)
1796 		return ret;
1797 
1798 	frame_size = mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN +
1799 		VLAN_TAG_SIZE;
1800 
1801 	/* check that mtu is within the allowed range */
1802 	if (mtu < RTE_ETHER_MIN_MTU || frame_size > dev_info.max_rx_pktlen)
1803 		return -EINVAL;
1804 
1805 	/* refuse mtu that requires the support of scattered packets when this
1806 	 * feature has not been enabled before. */
1807 	if (!dev->data->scattered_rx &&
1808 	    frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
1809 		return -EINVAL;
1810 
1811 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1812 	rctl = E1000_READ_REG(hw, E1000_RCTL);
1813 
1814 	/* switch to jumbo mode if needed */
1815 	if (frame_size > RTE_ETHER_MAX_LEN) {
1816 		dev->data->dev_conf.rxmode.offloads |=
1817 			DEV_RX_OFFLOAD_JUMBO_FRAME;
1818 		rctl |= E1000_RCTL_LPE;
1819 	} else {
1820 		dev->data->dev_conf.rxmode.offloads &=
1821 			~DEV_RX_OFFLOAD_JUMBO_FRAME;
1822 		rctl &= ~E1000_RCTL_LPE;
1823 	}
1824 	E1000_WRITE_REG(hw, E1000_RCTL, rctl);
1825 
1826 	/* update max frame size */
1827 	dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
1828 	return 0;
1829 }
1830 
1831 static int
1832 eth_em_set_mc_addr_list(struct rte_eth_dev *dev,
1833 			struct rte_ether_addr *mc_addr_set,
1834 			uint32_t nb_mc_addr)
1835 {
1836 	struct e1000_hw *hw;
1837 
1838 	hw = E1000_DEV_PRIVATE_TO_HW(dev->data->dev_private);
1839 	e1000_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
1840 	return 0;
1841 }
1842 
1843 RTE_PMD_REGISTER_PCI(net_e1000_em, rte_em_pmd);
1844 RTE_PMD_REGISTER_PCI_TABLE(net_e1000_em, pci_id_em_map);
1845 RTE_PMD_REGISTER_KMOD_DEP(net_e1000_em, "* igb_uio | uio_pci_generic | vfio-pci");
1846 
1847 /* see e1000_logs.c */
1848 RTE_INIT(igb_init_log)
1849 {
1850 	e1000_igb_init_log();
1851 }
1852